Transgenesis upgrades for Drosophila melanogaster

During the past few years, Drosophila melanogaster has gained in popularity because of the availability of its genome sequence(Adams et al., 2000), its rapid life cycle, the relative ease with which it can be handled and the multitude of genetic tools that are available for its study(Greenspan, 2004). The fly's genome permits the most sophisticated manipulations of any of the known eukaryotes. Indeed, the number of existing and recently developed technological improvements, such as genome-wide transposon tagging and gene targeting (Venken and Bellen,2005) and the availability of numerous resources, including online databases such as FlyBase and stocks from fly stock centers (see Box 1 for links to some of these resources) (Matthews et al.,2005), greatly facilitate research in the field and move it forward at a relentless pace. These technologies and resources further the study of various aspects of developmental biology, genetics, cell biology,neuroscience and behavior. Indeed, the identification of novel genes and their functional characterization in vivo greatly depends on these available tools. Moreover, as most human disease genes have a counterpart in the Drosophila genome, including those involved in genetic disorders and cancer (Bier, 2005; Vidal and Cagan, 2006), the fly is also becoming increasingly popular for studying the molecular mechanisms of human disease. Much of this research relies on an efficient and reliable transgenesis system.

Transgenesis in general can be defined as a group of technologies that allow DNA to be introduced into an organism of choice. The main goal of transgenesis is to integrate a foreign piece of DNA - a transgene - into an organism's genome to result in germ line transmission (see Fig. 1), in order to study gene function. Insect transgenesis, in general, has been dominated by transposon-mediated integration (Handler and James, 2000). In Drosophila, transgenesis mainly relies on the P element transposon and this has been the foundation for most of the innovative developments within the fly field(Ryder and Russell, 2003). However, various improvements in fly transgenic techniques have been recently reported that predominantly employ the site-specific integration of transgenes at specific genomic docking sites (see glossary, Box 2) via the use of different recombinases and integrases (Groth et al.,2004; Oberstein et al.,2005; Horn and Handler,2005; Bateman et al.,2006; Venken et al.,2006; Bischof et al.,2007). Many of these advances have their origins in mouse molecular genetics (Seibler and Bode,1997; Bethke and Sauer,1997; Bouhassira et al.,1997; Groth et al.,2000; Thyagarajan et al.,2001) and have been very useful for developing new fly transgenic techniques, as discussed below.

Here, we summarize many of the current methods that are used to generate transgenic flies. We first review classical transposon-mediated transgenesis and site-specific integration methods, before describing a plethora of recent improvements that have their basis in site-specific integration systems.

Transgenesis can be performed through various techniques. In Drosophila, transgenesis mainly relies on the P element transposon, the introduction of which(Rubin and Spradling, 1982)has been one of the most important breakthroughs in germ line transgenesis in Drosophila. As such, Drosophila research has been highly dependent on P element-mediated transgenesis, even though it has two major drawbacks: the size of the DNA that can be integrated is limited and the insertion sites cannot be controlled.

P elements are transposable elements, or transposons, which were originally identified within the fly's own genome(Castro and Carareto, 2004). P elements, like other transposons, contain two terminal repeats,including inverted repeat sequences and other internally located sequence motifs absolutely required for their mobilization or transposition (see Fig. 2 and Table 1)(Beall and Rio, 1997). Mobile or autonomous P element transposons encode a functional enzymatic protein called P transposase that catalyzes transposition through both terminal repeats of the transposon. P element-mediated transgenesis requires the separation of the P transposase and the P element transposon backbone(Rubin and Spradling, 1982). A plasmid that encodes P transposase, a so-called helper plasmid, is provided in trans with another plasmid (the transgene) that contains the transposon backbone, the sequence of interest and a marker (see Fig. 2B)(Karess and Rubin, 1984). In vitro synthesized mRNA that encodes the transposase or purified transposase protein itself (Kaufman and Rio,1991) can also be co-injected with modified P elements. Co-injections limit transposase activity, which is often advantageous. Alternatively, the transposase can be expressed from a genomic source(Cooley et al., 1988),allowing the injection of a P element without a helper plasmid. Transgene expression can be rendered constitutive or inducible through the inclusion of a heat-shock promoter. A hyperactive form of Ptransposase has been isolated that results in increased transposition rates(Beall et al., 2002). In general, transposons are injected into fly strains that are devoid of the same transposon, avoiding unwanted mobilization events of transposons present in the genome, thereby ensuring the stable integration and maintenance of the injected transgene. The unbiased identification of integration events is crucial for transgenesis and predominantly relies on the incorporation of dominant markers, which are identified through screening or selection (see Table 2 and Box 3 for more information).

Transposition occurs by the excision or replication of the transposon from the injected plasmid and its insertion into the host genome. Different transposons have unique insertion site characteristics. Integration events of P elements are strongly biased towards the 5′ end of genes. Hot spots - insertion sites that attract P elements at a much higher frequency than others - also exist within the Drosophila genome(Spradling et al., 1995; Bellen et al.,2004). Moreover, P elements have a narrow taxonomic activity and are non-functional outside of the Drosophilidae (Handler et al., 1993) owing to a host-specific factor that is required for transposition (Rio and Rubin,1988). To circumvent these limitations, several other transposons with a different insertional specificity and a broader host range have been identified that are suitable for germ line transformation in Drosophila (see Table 1). These include piggyBac(Handler and Harrell, 1999),identified in the cabbage looper moth Trichoplusia ni(Cary et al., 1989; Handler, 2002); the Tc1/mariner-like transposons Minos(Loukeris et al., 1995) and Mariner (Lidholm et al.,1993), isolated from Drosophila hydei(Franz and Savakis, 1991) and Drosophila mauritiana, respectively(Jacobson et al., 1986); and the hobo, Ac, Tam3 (hAT) family members Hermes(O'Brochta et al., 1996) and hobo (Blackman et al.,1989; Smith et al.,1993), isolated from the house fly Musca domestica(Warren et al., 1994) and Drosophila melanogaster, respectively(McGinnis et al., 1983). These transposons function in a variety of organisms, but their use in Drosophila transgenesis has been limited(O'Brochta and Atkinson,1996; Ryder and Russell,2003). piggyBac and Minos have been used as alternative mutagens because they have a different insertional specificity to P elements (Hacker et al.,2003; Horn et al.,2003; Thibault et al.,2004; Metaxakis et al.,2005). As Mariner elements do not remobilize efficiently(Lozovsky et al., 2002), and because hobo is present in most laboratory stocks, neither is commonly used. Finally, hobo and Hermes have been shown to cross-mobilize (Sundararajan et al.,1999). These features have limited the use of these transposable elements.

The transposon-mediated integration of transgenes has been used for numerous experiments in the fly field. These experiments can be broadly subdivided into two main groups: gene disruption methods and transgenic technologies. Gene disruption occurs when a transposon insertion interferes with the function of a gene. Transgenic technologies usually involve introducing the different components of novel techniques (see below) or performing rescue experiments.

Almost all technological progress in flies depends on our ability to transform them. Indeed, P element-mediated enhancer detection(O'Kane and Gehring, 1987; Bellen et al., 1989; Bier et al., 1989), the use of the FLP/FRT system to create mutant clones by inducing mitotic recombination (see glossary, Box 2) (Xu and Rubin,1993), the gene knockout methods in flies(Rong and Golic, 2000; Gong and Golic, 2003), the creation of molecularly defined deletions throughout the genome(Thibault et al., 2004; Ryder et al., 2004), the generation of marked mutant clones by MARCM (see glossary, Box 2)(Lee and Luo, 1999), and many other technological advances have relied on transgenesis. The recent availability of a genome-wide library of RNAi transgenic insertions that allows the knockdown of most fly genes(Dietzl et al., 2007) will also provide an invaluable tool to study gene function.

In addition, transposon-mediated phenotypic rescue of a mutation is considered to be the best and most convincing evidence that a piece of DNA contains a gene of interest. Unfortunately, traditional high-copy-number plasmids, including the P element-containing plasmids, have a limited cargo capacity of ∼20-25 kb of DNA owing to plasmid instability in bacteria. To circumvent this, P elements were engineered in a medium-copy-number cosmid backbone(Haenlin et al., 1985; Steller and Pirrotta, 1985),providing a higher cargo capacity of up to 40-50 kb. Unfortunately, the difficulties associated with obtaining integration of 30-50 kb Pelement-based cosmids did not promote the use of this methodology. As a result, a transgenic cDNA rescue based on the GAL4/UAS system (see glossary, Box 2)(Fischer et al., 1988; Brand and Perrimon, 1993) or heat-shock induction (Basler and Hafen,1989) became more popular.

One of the major drawbacks of P element-mediated transgenesis is that P elements most often integrate into the 5′ regulatory regions of genes (Bellen et al.,2004), thereby causing two unwanted consequences. First, the insertion often disrupts another gene that may or may not be relevant (e.g. within the same pathway) to the gene that is being studied(Norga et al., 2003). Second,the gene within the transposon may be subject to unwanted position effects or to position effect variegation (see glossary, Box 2) dictated by the surrounding genomic environment. Insertions in the regulatory region of a gene, on which nearby cis-acting elements typically act, bring the gene into an environment that is almost certainly subject to unwanted regulation. Indeed, position effects and position effect variegation were observed early on for markers such as white(Hazelrigg et al., 1984; Levis et al., 1985) and were eventually exploited in different kinds of enhancer-trap screens to identify temporally and spatially restricted expression patterns of developmentally regulated genes (O'Kane and Gehring,1987; Bellen,1999).

Position effects can be partially neutralized through the incorporation of insulator sequences (Roseman et al.,1995). Insulators (see glossary, Box 2) tend to shield the transgene from regulatory influences imposed by the surrounding genome. Insulators, such as gypsy, have been used in some P element vectors because they are more mutagenic than other P elements that do not contain insulators (Roseman et al.,1995). They were also incorporated into P element reporter transposons developed to analyze gene regulatory sequences(Barolo et al., 2000; Barolo et al., 2004). Insulators allow for a better comparison of different transgene insertions at different loci. Yet insulators may also influence the expression of the gene that they flank within the construct and are still somewhat subject to position effects in the genome.

There are at least four alternative genetic strategies to neutralize position effects when different transgenes are being compared at the same locus. The simplest method is transgene coplacement(Siegal and Hartl, 1996),which allows any two transgenes, such as a rescue fragment and its mutant version, to be compared in the same orientation at the same locus(Fig. 3). Both transgenes are integrated into a P element that contains the site-specific recognition sites FRT and loxP, the targets of FLP and Cre recombinases, respectively (see Box 4 for more information on these recombinases). After integration of the P element, FLP can remove one transgene and Cre can remove the other. Recognition sites are oriented such that either recombination event results in an identical configuration for either transgene. This method also introduced the use of Cre recombinase into the Drosophila field(Siegal and Hartl, 1996; Siegal and Hartl, 2000). One drawback of the technique is that only two transgenes can be compared at the same locus.

A second method is based on FLP recombinase-mediated transgene remobilization (Golic et al.,1997) (Fig. 4A). First, a `donor' P element (see glossary, Box 2), containing a transgene together with the white<sup>+</sup> marker flanked by FRTsites, is integrated into the fly genome using P transposition,resulting in a donor site. Second, the transgene with the white<sup>+</sup> marker, flanked by FRT sites, is remobilized through FLP excision. This episomal fragment (see glossary, Box 2) can integrate into a second single FRT-containing `acceptor' transposon (see glossary, Box 2), which also carries another dominant marker and is located elsewhere in the genome. Successful mobilization events can be identified through screening, as relocalization usually results in changes in white⁺ marker expression owing to position effects. This strategy is facilitated if a split white⁺ marker strategy is integrated into the system(Fig. 4B): the white⁺ marker is separated into 5′ and 3′fragments, and only becomes functional after the reconstitution of these fragments through site-specific recombination within an intron located between both fragments (Golic et al.,1997). In FLP recombinase-mediated transgene remobilization, white expression is only obtained after correct mobilization and site-specific integration (Fig. 4C), facilitating the screening procedure of integration events. Interestingly, thousands of P element insertions obtained by the DrosDel project (see Box 1)were generated by the mobilization of the P{RS5} and P{RS3}transposons (Golic and Golic,1996) and were subsequently used for the generation of precise deletions (Ryder et al.,2004). Both transposons can be used as acceptor elements for in vivo FLP-mediated DNA mobilization using the split white⁺marker strategy, and they provide numerous docking sites that are dispersed all over the fly genome (Fig. 4D,E). A drawback of FLP recombinase-mediated transgene remobilization in general is that a second round of crossings for remobilization and screening has to be performed after an initial Pelement-mediated transformation to obtain the required integration events of donor elements.

A third genetic trick to neutralize position effects is P element replacement or targeted transposition(Gloor et al., 1991; Lankenau and Gloor, 1998). An`acceptor' P element inserted at one location is replaced by a second`donor' P element, integrated at another location, through in vivo gap repair (Fig. 5). The technique requires homologous recombination between the 10-20 bp footprints of the 31 bp inverted terminal repeat that remain after excision of the acceptor element and the homologous counterpart of the donor element. The homologous recombination event is promoted owing to a double-stranded gap that is generated after the excision of the acceptor element. Various donor Pelements can be targeted to the same locus, allowing different transgenes to be directly compared with each other. A drawback of the technique is that replacement can occur in both directions, requiring additional molecular verification. The technique has not been used to perform structure-function analysis of differently mutagenized transgenes but has been proven useful for converting existing lacZ enhancer-detectorP elements into GAL4 drivers (Sepp and Auld,1999; de Navas et al.,2006) that allow more versatile reporter analysis using the GAL4/UAS system (Fischer et al.,1988; Brand and Perrimon,1993).

The best but most labor-intensive way to eliminate position effects is in vivo gene targeting through homologous recombination. Gene targeting in Drosophila can be performed using two strategies: `ends-in' or insertional gene targeting (Rong and Golic, 2000) and `ends-out' or replacement gene targeting(Gong and Golic, 2003)(Fig. 6). Insertional gene targeting results in the insertion of the entire targeting sequence into the region of homology. This results in a duplication that can be resolved during a second round of homologous recombination(Fig. 6A)(Rong et al., 2002). Replacement gene targeting results in the substitution of an endogenous DNA sequence with exogenous DNA through a double-reciprocal recombination event between two stretches of homologous sequence(Fig. 6B). Both strategies require the introduction of a `donor' element, which contains the gene-targeting cassette, through transgenesis prior to in vivo homologous recombination, and require extensive screening. Although the techniques have not been used to compare the phenotypic outcome of different transgenes at the same locus, they are gaining in popularity for creating targeted mutations(O'Keefe et al., 2007).

Recent efforts have focused on making gene targeting more efficient in Drosophila through the use of site-specific zinc-finger-nuclease-stimulated gene targeting(Bibikova et al., 2003; Beumer et al., 2006). Zinc-finger nucleases are protein fusions between the Fok1 nuclease and(generally) three zinc-finger DNA-binding domains that introduce sequence specificity. Because each zinc finger recognizes 3 bp, zinc-finger nucleases can be designed to bind to a unique segment of 9 bp. As these nucleases need to dimerize at the target site before they can cut the target DNA, a recognition site of 18 bp is effectively required, a sequence that is likely to be unique in the fly genome. Thus, cutting by zinc-finger nucleases can be directed to specific target sites to create a double-stranded break, resulting in increased gene-targeting efficiency when a linearized donor targeting element is introduced.

Although all the strategies of site-specific integration described above are elegant and useful, they have not been used extensively. The main drawbacks are that they allow only a limited number of transgenes to be compared and are too labor-intensive, as they require transgenesis of a donor construct prior to extensive genetic screening to obtain the required site-specific transgenic insertion event. Hence, the primary goal of true-targeted transgenesis is to achieve efficient site-specific integration upon injection of the DNA without the need for further manipulations.

This strategy was pioneered in the fly field using the bacteriophageΦC31 integrase, which can integrate transgenic constructs at defined docking sites (Groth et al.,2004). Moreover, ΦC31 integrase-mediated transgenesis allows large DNA fragments to be integrated into the fly genome, well beyond the fragment sizes that can be introduced by P element-mediated integration (Venken et al.,2006). As discussed in more detail below, this approach has also introduced a user-friendly DNA modification platform, called recombineering,into Drosophila research.

ΦC31 integrase catalyzes the recombination between the phage attachment(attP) site present in its own bacteriophage genome and a bacterial attachment (attB) site present within the bacterial host genome(Thorpe and Smith, 1998) (see Box 4). Previous work has shown that the ΦC31 integrase can catalyze the site-specific integration of attB-containing plasmids into so-called attP-containing`docking' or `landing' sites that have been introduced into mammalian cell lines (Groth et al., 2000; Thyagarajan et al., 2001). Interestingly, attB-containing plasmids integrate more readily into attP-containing genomic docking sites than do attP sites in the reciprocal reaction, indicating that the integration reaction is asymmetric in nature (Thyagarajan et al.,2001; Belteki et al.,2003). This phenomenon was recently confirmed in Drosophila (Nimmo et al.,2006).

In Drosophila, recombination is mediated via ΦC31 integrase,provided through an mRNA source, between an attP docking site,previously integrated with a transposon into the fly genome, and an attB site present in an injected plasmid(Groth et al., 2004)(Fig. 7A). Three so-called pseudo-attP docking sites have been identified within the Drosophila genome. As one of these pseudo-sites is located in the endogenous transposable element copia, the true number of available pseudo-sites is likely to be high(Kaminker et al., 2002). Fortunately, these pseudo-sites were shown not to be receptive to attB plasmids, as all integration events were at the desired attP sites (Groth et al.,2004). However, rare non-specific integrations have been documented in Drosophila (Venken et al., 2006; Nimmo et al.,2006; Bischof et al.,2007). The ΦC31 integrase-mediated transformation technique has also recently been introduced successfully in the yellow fever mosquito Aedes aegypti (Nimmo et al.,2006).

After the original report describing two attP P element docking sites (Groth et al., 2004),numerous additional docking sites have been created. One set is embedded in a piggyBac backbone (Venken et al., 2006), whereas a second set is embedded in a Marinerbackbone (Bischof et al.,2007). Venken et al. (Venken et al., 2006) observed that one out of seven docking sites tested was not receptive, suggesting that the genomic position of the docking site can affect integration efficiency. This was not observed for the 19 sites tested by Bischof et al. (Bischof et al.,2007). A detailed characterization and comparison of all the available docking sites will allow us to determine which ones are the most useful for specific purposes, such as cDNA overexpression, RNAi, genomic rescue or promoter/enhancer analysis.

Although the first reports used mRNA-encoded ΦC31 integrase to integrate the DNA (Groth et al.,2004; Bateman et al.,2006; Venken et al.,2006), Bischof et al. (Bischof et al., 2007) recently reported an efficient germ line ΦC31 integrase source that is driven by nanos or vasa regulatory elements. Interestingly, through ΦC31 integrase-mediated transgenesis,different ΦC31 integrase sources have been incorporated at the same docking sites. Additionally, the same ΦC31 integrase source was integrated into different docking sites, allowing the most efficient genomic ΦC31 integrase source to be selected. In the same study, a Drosophilacodon-optimized ΦC31 integrase was described that performs better than the non-optimized version (Bischof et al.,2007).

Site-specific integration using a single recombination site results in the integration of the vector backbone, which may interfere with transgene expression (Chen et al.,2003). This can be minimized through marker genes strategically positioned between transgene and vector backbone(Venken et al., 2006). Alternatively, appropriately engineered recombinase sites in both the docking site and integration plasmid can be used to remove unwanted vector backbone sequence after correct integration events are isolated(Bischof et al., 2007). Finally, the integration of the backbone can be directly avoided through recombinase-mediated cassette exchange (RMCE)(Baer and Bode, 2001).

In RMCE, both docking site and transgene are flanked by a recombination site (see Box 4 and Fig. 7). Double reciprocal cross-over results in the integration of a transgene without its vector backbone. However, two sets of directly oriented loxP or FRTsites will favor deletion over RMCE. This problem can be overcome with sites,called spacer variants, that support recombination between themselves but not with others (see Box 4 for more information). The use of RMCE with spacer variants was initially utilized in the mouse in Cre- (Bethke and Sauer,1997; Bouhassira et al.,1997) and FLP- (Seibler et al., 1998) based genetic engineering. This approach has been recently exploited in Drosophila for both recombinases(Oberstein et al., 2005; Horn and Handler, 2005)(Fig. 7B). For example, RMCE has been used elegantly to perform structure-function analysis of the eve2 (eve - FlyBase) enhancer with a lacZ reporter(Oberstein et al., 2005). An alternative way to ensure that RMCE avoids the deletion or integration of plasmid backbone when employing FLP or Cre, is to use inverted recombination sites (as shown for ΦC31 integrase in Fig. 7C), which was pioneered in the mouse using Cre (Feng et al.,1999). This strategy eliminates the deletion problem but causes inversions.

Interestingly, RMCE events can be locked using the ΦC31 integrase system, as these recombination reactions are unidirectional. Pioneered in the yeast Schizosaccharomyces pombe(Thomason et al., 2001) and in mouse (Belteki et al.,2003), a modification of this system using inverted attrecombination sites was recently described for Drosophila(Bateman et al., 2006)(Fig. 7C). This study demonstrated that unmarked constructs can be integrated through RMCE, as site-specific integration events are identified by loss of the marker(Bateman et al., 2006).

Transposons are generally characterized by a low cargo capacity, limiting the amount of DNA that can be integrated and mobilized. Transgene size limitation can be overcome by the incorporation of a site-specific integration system, such as ΦC31 integrase, as shown in a chicken cell culture system(Dafhnis-Calas et al., 2005),or by gene targeting at the Hprt1 locus in mouse embryonic stem (ES)cells (Heaney et al., 2004). Unfortunately, an intermediate cell culture system supporting both in vitro gene manipulation and subsequent germ line transmission, similar to mouse ES cells, is not available for Drosophila. Moreover, there is a strong negative correlation between the upper size limit of the insert and plasmid copy number: large DNA fragments are unstable when present in high-copy-number vectors in bacteria. Therefore, low-copy-number plasmids, such as P1 bacteriophage (Sternberg,1990), bacterial artificial chromosomes (BACs)(Shizuya et al., 1992) and P1 artificial chromosomes (PACs) (Ioannou et al., 1994), were developed to maintain large inserts. Unfortunately, these plasmids interfere with both cloning and microinjection procedures, which require high DNA concentrations. This can be circumvented by the use of a specialized plasmid backbone - the conditionally amplifiable BAC- that has two origins of replication(Wild et al., 2002): an oriS for low-copy propagation and an oriV for high-copy induction. Importantly, the manipulation of large DNA fragments in these vectors has been facilitated through recent developments in in vivo recombination-mediated genetic engineering, also known as recombineering(Copeland et al., 2001; Heintz, 2001; Muyrers et al., 2001; Sawitzke et al., 2007).

Three recently developed technologies - recombineering, the ability to amplify plasmid copy number at will, and ΦC31 integrase-mediated transgenesis - have recently been integrated into a single transformation system (Venken et al., 2006). This system provides an efficient recombineering platform for Drosophila, permitting the integration of large DNA fragments into the fly genome. Selected fragments that encode the gene of interest are obtained in the amplifiable BAC backbone through recombineering-mediated gap repair, which can be performed at low copy number(Fig. 8A). Gap repair in high-copy and medium-copy plasmids has an upper size limit of ∼30 and 80 kb, respectively (Lee et al.,2001). However, by maintaining the plasmid at low copy number,large fragments of up to 102 kb can be efficiently gap-repaired(Venken et al., 2006), as observed by others (Kotzamanis and Huxley,2004). Interestingly, one 133 kb fragment that encodes one of the largest genes in the fly genome, Tenascin major, was reconstituted from two different BACs, each containing part of the gene(Venken et al., 2006). Gap-repaired DNA was induced to high copy number, isolated, and integrated into the fly genome using both P transposase and ΦC31 integrase: P transposase was used to integrate gap-repaired fragments of up to 39 kb, whereas ΦC31 integrase was used to integrate gap-repaired fragments of up to 133 kb.

A similar gap-repair approach was recently used to generate transgenes for Drosophila in vivo (Takeuchi et al., 2007). The gap-repaired constructs were obtained in flies through homologous recombination into the endogenous locus after the in vivo linearization of the transgene between both homology arms using the meganuclease I-SceI (Fig. 8B). The technique relies on endogenous fly enzymes, rather than on bacterial enzymes, to mediate the gap repair. The authors observed an upper size limit of 28 kb for correct gap repair.

An important reason for the development of recombineering is the ease with which DNA can be modified. Indeed, restriction enzymes and DNA ligase are not user-friendly when handling large DNA fragments, as the occurrence of unique cutting sites decreases with increasing plasmid size. Recombineering does not suffer from those limitations and allows BACs to be modified more rapidly using PCR products or oligonucleotides that contain the desired mutation as recombination templates (Copeland et al.,2001; Court et al.,2002) (Fig. 9A). This strategy can easily be combined with positive/negative selectable markers, such as galK and thyA(Warming et al., 2005; Wong et al., 2005). Positive/negative selectable markers are targeted to the desired site for mutagenesis during a first round of recombineering using selection, and are then replaced by the desired mutation or tag during a second round of recombineering using counterselection(Fig. 9B). An alternative way to modify DNA constructs uses the recombinogenic protein RecA, also known as RecA-assisted modification (Yang et al.,1997; Gong et al.,2002), a methodology that is somewhat different from traditional recombineering. In a first recombination step, a modifying plasmid is integrated into the target plasmid, resulting in a co-integrate that becomes resolved during a second round of recombination(Fig. 9C). The technique allows the integration and deletion of fragments within a genomic fragment(Misulovin et al., 2001) and has been applied on a high-throughput level in the mouse field to create an atlas of gene expression in the mouse central nervous system(Gong et al., 2003). BAC modification was pioneered in the mouse field because most mouse genes tend to have multiple distant regulatory regions and are therefore too large to be handled using high-copy plasmid backbones(Heintz, 2001). The efficient recombineering-mediated tagging of genes in a genomic context has also been recently demonstrated in Caenorhabditis elegans and C. briggsae (Dolphin and Hope,2006; Sarov et al.,2006).

The advent of site-specific integration combined with recombineering and other methodologies will impact the fly field in numerous ways. These techniques make it possible to carry out structure-function studies at a higher resolution with fewer transgenes, as position effects can be mitigated using some of these approaches. Moreover, we are no longer confined to the study of single genes but can now tackle entire gene complexes that play key roles in development. Through repeated rounds of mutagenesis via recombineering, one can dissect the in vivo role of each gene and each regulatory region within a gene complex. Similar manipulations are now also possible for larger genes and for loci that have previously had no available mutations to study. These loci can now be identified and studied through the introduction of small deletions by FLP/FRT recombination(Parks et al., 2004; Ryder et al., 2004).

Combinations of the different methodologies described here should also greatly enhance our ability to manipulate the fly genome. For example, P replacement with an attP-containing P element could be used to convert many of the existing P elements into a useful docking site for ΦC31 integrase-mediated transgenesis or RMCE. Alternatively, gene targeting of recombination sites at desired locations might allow the site-specific integration of any DNA fragment. Finally,ΦC31 integrase-mediated transgenesis can be used to insert the donor constructs that are required for gene targeting. These are just a few examples of possible future fly manipulations.

In another vein, these technologies will also help to improve genome-wide studies of Drosophila. For example, one could try to identify optimal genomic sites for the integration of all RNAi constructs. These sites should permit the optimal expression of RNA hairpin loops in all tissues at all developmental stages to allow the efficient RNAi-mediated knockdown of any gene. This approach may alleviate some of the potential drawbacks that are associated with P transposase-mediated transgenesis of RNAi constructs, such as poor transgene expression or misregulation(Dietzl et al., 2007). Furthermore, many different genomic DNA fragments containing cis-regulatory elements that drive GAL4 expression could be integrated at the same docking site to allow the labeling and manipulation of specific cell populations. Finally, the integration of overlapping duplications of defined areas of the X chromosome into the same docking site would be a useful way to map essential genes on the X chromosome.

Each transgenesis technique might result in unwanted side effects. Transposases might cause multiple insertion and excision events before the final transgenic insertion is stabilized. Recombinases and integrases may recognize pseudo-sites localized within the genome, as previously identified for Cre in the mammalian genome(Thyagarajan et al., 2000). Moreover, at high doses, Cre has been demonstrated to result in undesired effects in both vertebrates (Schmidt et al., 2000; Loonstra et al.,2001) and Drosophila(Heidmann and Lehner, 2001). Detrimental side effects have been observed for ΦC31 integrase in mammalian cell culture (Ehrhardt et al.,2006; Liu et al.,2006), although ectopic expression in vivo in mice and Drosophila indicates that it has minimal side effects in these organisms (Belteki et al.,2003; Raymond and Soriano,2007; Bischof et al.,2007). Even gene targeting in Drosophila has not been spared from artifacts, and second-site mutations have been shown to cause interference with phenotypic characterization(O'Keefe et al., 2007). Interestingly, so far no such observations have been documented for FLP, which is widely used within the fly community(Blair, 2003).

Unfortunately, the solution to an important problem such as efficient site-specific integration, immediately results in the creation of new challenges: the handling of thousands of fly strains associated with typical high-throughput projects, as well as the maintenance of thousands of new stocks. Recent methods for the automated microinjection of fly embryos for high-throughput in vivo RNAi experiments have been developed(Zappe et al., 2006), and it should now be possible to adapt this technology for DNA microinjections. However, no solution has yet been developed to maintain numerous additional fly stocks, except through the expansion of existing or new stock centers.

We apologize to those whose work could not be cited owing to space limitations. We thank Karen Schulze for comments on the manuscript. K.J.T.V. and H.J.B. are supported by the NIH and the Howard Hughes Medical Institute.