For the targeting cross, the number of P{donor}*/hs-FLP, hs-I-SceI virgin females directly determines the scale of the whole targeting experiment. Genes that are resistant to homologous recombination may require collecting and sorting >15,000 virgins from the targeting cross (Larsson et al. 2004), which is extremely labor intensive due to the time-sensitive nature of virgin collection and the genotyping process. To eliminate this major bottleneck in scaling up the targeting cross, we modified the original hs-FLP, hs-I-SceI stocks by replacing their Y chromosomes and balancer chromosomes with ones that contain hs-hid transgenes (Grether et al. 1995). We named these modified stocks “6934-hid” and “6935-hid” (Figure 1b) after the original stock numbers. Ubiquitous expression of the cell-death gene hid induced by heat shock causes strong lethality. As illustrated in Figure 1c, in a targeting cross using 6934-hid, all male progeny and those female progeny carrying hs-hid balancer chromosome are eliminated. Since P{donor}*/hs-FLP, hs-I-SceI females are the only genotype that survives, 6934-hid and 6935-hid completely eliminate the time-sensitive virgin collection and genotyping process.

For screening and mapping crosses, we found their throughput was often severely limited by the high background of false positives, which may represent >95–99.9% of preliminary candidates (J. Huang, W. Zhou, and Y. Hong, unpublished results, and see below). Therefore, we introduced a negative selection marker into the current ends-out targeting scheme, so the majority of nontargeted integrations may be directly eliminated before they are subject to any further screening and mapping efforts. Ectopic expression of another cell-death gene reaper (rpr), similar to hid, also causes strong lethality (White et al. 1996). As illustrated in Figure 2b, a UAS-Rpr module can be tagged to the 3′ end of a transgenic donor DNA fragment (e.g., P{crbmEosFP^KI}). Once the donor DNA fragment is recombined into the target gene locus, UAS-Rpr will be lost due to homologous recombination. In contrast, nontargeted integrations will likely retain the donor DNA fragment with an intact UAS-Rpr module (“Rpr⁺”). By using proper Gal4 driver stocks to set up the screening cross, Rpr⁺/Gal4 false-positive candidates will be directly eliminated due to the ectopic expression of Rpr.

To implement the UAS-Rpr selection, we made a new set of ends-out targeting vectors, pRK1 and pRK2 (Figure 2a) that were based on the integration and modification of pEndsOut2 and pBS70W (available from http://dgrc.cgb.indiana.edu/). Both pRK1 and pRK2 contain a UAS-Rpr module, while pRK2 also has a GMR (Hay et al. 1994)-enhanced w⁺ marker to further facilitate the recovery of targeting candidates. We made two targeting constructs, a crbmEosFP^KI knockin construct (Figure 2b) and a dArf6^KO knockout construct (Figure 2c), on the basis of pRK1 and pRK2, respectively. Crb is a transmembrane protein essential for developing cell polarity (Tepass et al. 1990). We plan to study the trafficking and dynamics of Crb by tagging it with a photo-convertible fluorescent protein mEosFP (Wiedenmann et al. 2004). dArf6 (Arf51F) is a small GTPase that may play key roles in Drosophila muscle and nervous system development, although no dArf6 mutants are currently available. dArf6^KO targeting aims to delete 2.158 kb of the dArf6 locus that includes all the coding exons plus the 3′-UTR (Figure 2c). We obtained multiple transgenic lines from both targeting constructs at normal frequency (supplemental Table 2), indicating that UAS-Rpr in pRK1 and pRK2 was sufficiently silent in the absence of the Gal4 driver and did not adversely affect the routine P-element-based transgenic process. All the transgenic donor lines were larval or pupal lethal when crossed with neuronal-specific drivers Gal4⁴⁷⁷ and Gal4²²¹ (supplemental Table 2) (Grueber et al. 2003). Gal4²²¹/UAS-Rpr also consistently produced very few adult escapers of a fully penetrated wing inflation phenotype (supplemental Figure 1d).

To evaluate the effectiveness of UAS-Rpr without any bias, we first carried out crbmEosFP^KI knockin experiments without UAS-Rpr selection (similar to Figure 1a). From ~5 × 10⁴ screening cross progenies, we recovered 270 male candidates, of which 14 were mapped to the third chromosome where crb is located. We then screened 125 non-third chromosome candidates and all 14 third chromosome candidates for the presence of UAS-Rpr by crossing them into Gal4²²¹. As summarized in Table 1, only 3/125 of the non-third chromosome candidates (i.e., false positives) are Rpr^[−],while 11/14 of the third chromosome candidates are Rpr^[−] of which 7 were confirmed by PCR to have the correct targeting events (supplemental Figure 1a). Extrapolating from these data, selecting against UAS-Rpr in the screening cross of crbmEosFP^KI would eliminate >96% (253/263) of false positives (Table 1). In addition, UAS-Rpr selection also eliminates tandem-insertion mutants (Gong and Golic 2003), which can be difficult to distinguish from true targeting candidates by simple PCR assays (supplemental Figure 1, b and c). The homologous recombination frequency of crbmEosFP^KI is ~1/7000 if only considering the male candidates.

We then decided to carry out a large-scale dArf6^KO targeting experiment by taking full advantage of the new reagents and methods described here, as we failed at dArf6^KO targeting on the basis of the original pEndsOut2 vector by screening ~1.6 × 10⁵ screening cross progenies (W. Zhou and Y. Hong, unpublished results). We recloned the same 5′ and 3′ homologous arms into the pRK2-based vector. By using 6935-hid to set up the targeting cross, we easily collected >2 × 10⁴ virgin females. Twelve thousand of them were mated with w/Y; Pin/CyO; Gal4^221[w−] males to set up the screening cross (Figure 1d). From >7 × 10⁵ screening cross progenies (Table 2) we recovered 315 w⁺ males, of which 5 were verified as specific targeting candidates by PCR (Table 2, supplemental Figure 2, a and b). As a control, 200 virgin females from the same targeting cross were crossed with regular w/Y, Pin/CyO males. Of 124 w⁺ male candidates recovered, 1.6% (2/124) were Rpr^[−], but none harbored the true targeting event (Tables 1 and 2). Thus, UAS-Rpr selection achieved an impressive ~60-fold reduction of false positives. Effectively, dArf6^KO targeting was accomplished at a scale equivalent to screening/mapping >18,000 (315 × 60) preliminary candidates in the absence of UAS-Rpr selection. In addition, the dramatically reduced number of preliminary candidates, combined with their dark-red eye color due to GMR-enhanced w⁺ expression in pRK2, made the screening process much easier and faster. The homologous recombination frequency of dArf6^KO can be estimated as ~1/140,000 if only considering the male candidates. Homozygotes of dArf6^KO are viable but are male and female sterile, so it is possible that dArf6 only plays a specific and indispensable role in germline development. When we were preparing this article, a P-element-induced deletion allele of dArf6 was published and Dyer et al. (2007) observed the same male and female sterile phenotypes in their homozygous dArf6 mutant flies.

Compared with the “rapid scheme” in which preliminary candidates were screened for the loss of FRT sites (Rong et al. 2002; Gong and Golic 2003), UAS-Rpr selection is more efficient since it directly eliminates false positives. In addition, we found that the majority of the false positives (57–87%) had damaged FRT sites (Table 1); therefore, they could only be eliminated by UAS-Rpr selection but not by the FRT test. Since the I-SceI sites are positioned rather close to the FRT sites in ends-out targeting constructs, we speculate that the frequent FRT damage seen here was most likely due to the double-strand DNA repair process triggered by the premature cut of I-SceI sites (Bellaiche et al. 1999; Gong and Golic 2003). Separating FRT and I-SceI sites further away in future pRK-based targeting vectors should further reduce the frequency of false positives. Consistently, the Golic lab reported that the frequency of false positives was low using the pW25 targeting vector in which FRT and I-SceI were separated by 100–150 bp (Gong and Golic 2004). Since pRK-based vectors may not be suitable for making Gal4 knockin alleles, pW25 series vectors should be excellent alternatives.