Generation of a Useful roX1 Allele by Targeted Gene Conversion

Methods for altering the sequence of endogenous Drosophila melanogaster genes remain labor-intensive. We have tested a relatively simple strategy that enables the introduction of engineered mutations in the vicinity of existing P-elements. This method was used to generate useful alleles of the roX1 gene, which produces a noncoding RNA involved in dosage compensation. The desired change was first introduced into a genomic clone of roX1 and transgenic flies were generated that carry this sequence in a P-element. Targeted transposition was then used to move the P-element into roX1. Remobilization of the targeted insertion produced large numbers of offspring carrying chromosomes that had precisely introduced the engineered sequences into roX1. We postulate that this occurred by gap repair, using the P-element on the sister chromatid as template. This strategy was used to introduce six MS2 loops into the roX1 gene (roX1MS2-6), enabling detection of roX1 RNA by a MCP-GFP fusion protein in embryos. The roX1MS2-6 remains under the control of the authentic promoter and within the correct genomic context, features expected to contribute to normal roX1 function. The ability to replace relatively large blocks of sequence suggests that this method will be of general use.

replacement strategy "targeted gene conversion" (TGC). TGC is technically simple and capable of introducing large blocks of nonhomologous sequence. It is able to replace sequences that are more than 1 kb from a P-element insertion. The strategy that we tested relies on a P-element near the site to be mutated. However, recently developed methods for directed mutagenesis may enable a modified form of TGC in regions that lack P-elements.

MATERIALS AND METHODS
Fly culture Flies were raised on a yeast, molasses, and cornmeal diet at room temperature. Mutations are described in citations or Lindsley and Zimm (1992).

Gene conversion in females
The targeted transposition roX1 [MS2-6]T2A was mobilized in females. A total of 244 dysgenic females (roX1 [MS2-6]T2A / Binsincy; Sb p[ry + D2-3] 99/+) were mated to yw males, with approximately 10 females per vial; 25 out of 26 vials produced at least one white-eyed nonbalancer son, indicating excision. Two hundred sixty-nine excisions were mated individually to C(1)Dx y 1 f 1 females. A randomly selected subset of these was analyzed by PCR for MS2 loop incorporation and loss of P-element sequences.
DNA blotting DNA from 100 flies was extracted as described (http://www.fruitfly. org/about/methods/inverse.pcr.html). DNA was suspended in 300 ml DEPC water and treated with RNAse A and proteinase K; 15 mg DNA was digested overnight with EcoRI, concentrated, electrophoresed, and transferred to a charged nylon membrane. Blots were probed with a 32 P-labeled, 2.03-kb EcoR1-Mlu1 fragment spanning the promoter and 59 end of roX1 using previously described methods (Church and Gilbert 1984). Restriction digests of a 4.9-kb roX1 genomic clone served as a molecular weight marker.
Visualization, photography, and image processing Immunodetection of MSL1 on polytene preparations was performed as previously described (Kelley et al. 1999). MCP-GFP is removed by acetic acid fixation, preventing visualization on polytene chromosomes. To visualize MCP-GFP recruitment in embryo nuclei, homozygous roX1 MS2-6 roX2Δ; [w +mC MCP-GFP] females were mated to males carrying a p[w +mC sqh-mCherry] insertion on the X chromosome. Male embryos are distinguished by lack of mCherry signal. The 3-h to 12-h embryo collections were dechorionated, fixed in 4% paraformaldehyde with 0.1% Tween-20, DAPI-stained, and mounted with DABCO anti-fade agent in 50% glycerol. Z-stacks were recorded for individual embryos using an Olympus Fluoview FV10i scanning confocal microscope with a 60· water/oil immersion lens. Images were processed by converting to 8-bit format and importing individual Z-stacks into ImageJ. Because mCherry signal was weak and diffuse, the brightness of this channel was uniformly enhanced for reproduction ( Figure 4, C, H, and M). Consistent patterns of GFP localization were observed in images of more than 30 embryos from three collections.

RESULTS
An autosomal roX1 MS2-6 transgene restores X chromosomal MSL1 localization RNA accumulation can be visualized in tissues or chromosome preparations by in situ hybridization. Although useful, this method is time-consuming and incompatible with living tissue. RNAs that contain stem loops from the MS2 virus can be visualized in vivo when a fusion of MCP-GFP is expressed ( Figure 1A) (Bertrand et al. 1998). A roX1 transgene was constructed with six MS2 loops (roX1 MS2-6 ) inserted in a region previously shown to be nonessential ( Figure 1B) (Stuckenholz et al. 2003;Deng et al. 2005). An autosomal copy of this transgene, p[w +mC GM roX1 MS2-6 ], rescues X-localization of a key member of the MSL complex, Male-Specific Lethal 1 (MSL1), in roX1 roX2Δ males ( Figure 1C). However, ectopic recruitment surrounding the site of transgene insertion is also observed (arrow, Figure  1C). Fully wild-type behavior of roX1 is consequently expected to require expression from the X chromosome, possibly from the roX1 locus itself.
Gene conversion by repair using a sister chromatid template During P-element mutagenesis of roX1, we obtained a series of mutations that suggested a strategy for inducing precise changes in target genes. A reporter construct containing the roX1 promoter fused to LacZ (p[w +mC roX1P-bgal]) was moved into roX1 in an effort to capture enhancers in the vicinity. This was accomplished by targeted transposition to the plArB element in roX1 mb710 (Figure S1 A, B) (Gloor et al. 1991;Heslip and Hodgetts 1994). The resulting insertion, roX1 w+tandem , retained plArB and is marked with w +mC , facilitating subsequent mutagenesis. Hybrid element insertion was used to generate roX1 Δ891 , deleted for the plArB element and 891 bp flanking the insertion site, but retained p[w +mC roX1P-bgal]   (Figure S1C). Remobilization of roX1 Δ891 produced numerous white-eyed offspring from virtually every dysgenic parent. However, only a few imprecise excisions were identified ( Figure S2A). Instead, almost 70% of excisions carried molecularly identical rearrangements exemplified by the severe roX1 SMC17A allele ( Figure S1D). These appear to be produced by a gene conversion that occurs when the double-stranded break produced by P-element mobilization undergoes gap repair using a sister chromatid template ( Figure S2B). The rearrangement generated is consistent with repair driven by homology between the roX1 promoter on the broken chromosome more than 1 kb from the break site and in p[w +mC roX1Pbgal]. Homology is also shared by terminal inverted repeats at the 59 and 39 P-element ends (P-ends). In all 38 flies recovered with this rearrangement, the 39 P-end has been precisely replaced by the 59 Pend, a structure consistent with the proposed mechanism of repair. These chromosomes lost 1.2 kb of roX1 sequence flanking the p[w +mC roX1P-bgal] insertion site and replaced it with more than 3 kb of LacZ sequence fused to the roX1 promoter. This mechanism is thus capable of efficiently replacing large regions close to P-elements.

Lack of repair utilizing a template on a different chromosome
To determine if efficient gene conversion was an intrinsic property of the roX1 locus that is independent of template location, we attempted to generate a useful allele of roX1 by introducing sequence from an engineered roX1 MS2-6 transgene situated on an autosome. Gene conversion at white (w) occurs in a small percentage of excisions when a P-element is mobilized from w and a template with homology to insertion site is present in the genome (Banga and Boyd 1992;Johnson-Schlitz and Engels 1993;Nassif et al. 1994;Lankenau et al. 1996). We attempted to introduce MS2 loops into roX1 from an autosomal p[w +mC GM roX1 MS2-6 ] template. Dysgenic males with a p[w +mC GM roX1 MS2-6 ] donor on the third chromosome and the roX1 Δ891 target site on the X chromosome were generated. To enable phenotypic detection of gene conversion, the target X chromosome was also deleted for roX2 (see Materials and Methods for full description of genotypes and matings). The roX1 Δ891 is a severe loss-of-function mutant. Conversion to roX1 MS2-6 will restore male viability and eliminate the w +mC marker in roX1 Δ891 . Approximately 100 white-eyed sons were recovered and tested by PCR for incorporation of MS2 loops, but only wild-type roX1 sequences were detected. Although  a gene conversion strategy utilizing a template situated on another chromosome may be productive in some situations, it was not useful in this instance.

Targeted transposition of p[w +mC GM roX1 MS2-6 ]
To determine if p[w +mC GM roX1 MS2-6 ] would be utilized for gap repair if situated in roX1, targeted transposition was used to move it to the plArB insertion site in roX1 mb710 (Figure 2). Mobilization produced abundant hops to the X chromosome, 68% of which (34/50 insertions) were in roX1. The reason for the unusually high efficiency of targeting is unknown, but an interaction of roX genes in the male germ line, where transposition occurred, is suggested. Insertions on the X chromosome were characterized by in situ hybridization and PCR. The plArB was retained in tandem with 32 of the insertions. However, two precise replacements of plArB with p[w +mC GM roX1 MS2-6 ] were recovered.
Mobilization of targeted insertions to create roX1 MS2-6 Three targeted insertions in roX1 were remobilized: a replacement line (roX1 [MS2-6]R36A ) and two tandem insertions with different orientations (roX1 [MS2-6]T2A and roX1 [MS2-6]T4B ) ( Figure 2). Dysgenic males (roX1 [MS2-6]XX ; Sb p[ry + D2-3]99/ +) were mated to C(1)DX y 1 f 1 females. Mobilization is very frequent, with more than 90% of dysgenic males producing white-eyed sons, which comprise 20% of male offspring. White-eyed sons were mated individually to C(1)DX y 1 f 1 females and analyzed by PCR for repair of roX1 and inclusion of MS2 loops. Amplicons spanning the MS2 loop insertion site produce products characteristic of both wild-type roX1 (547 bp) and roX1 MS2-6 (869 bp) from targeted transpositions, but almost 99% of white-eyed offspring produced a single amplicon. A total of 352 excisions were analyzed (169 for roX1 [MS2-6]T2A , 103 for roX1 [MS2-6]T4B , 80 for roX1 [MS2-6]R36A ). Regardless of the starting line, more than 10% of white-eyed sons had incorporated MS2 loops into the repaired chromosome (Figure 3, B-E and Table 1). Amplicons from representative flies containing MS2 loops were sequenced, confirming faithful copying. The MS2 loops are 322 bp of nonhomologous sequence situated 430 bp from the point of P-element insertion ( Figure 3A). Incorporation of MS2 loops therefore requires a gene conversion tract more Hybridization to a single 4.9-kb roX1 fragment is seen in wild-type (WT) flies and in a gene conversion that did not incorporate MS2 loops or retain a P-end (roX1 + ). A single 5.2-kb fragment is detected in two precise conversions incorporating MS2 loops (lines 2A.1 and 4B.1). Hybridization to a single 5.1-kb band is observed in excision 36A.1, consistent with the reduced MS2 loop array observed by PCR. Line 2.5 is the imprecise excision depicted in (D). A 5.2-kb band from p[w +mC GM roX1 MS2-6 ] and a 2.5-kb band produced by disruption of genomic roX1 by insertion of plArB are present. than 750 bp in length. However, three flies generated by mobilization of roX1 [MS2-6]R36A produced 800 bp PCR amplicons, consistent with contraction of the MS2 loop array during gene conversion ( Figure 3E).
Our aim was to engineer roX1 without leaving vector or P-element sequence behind. However, homology at P-ends can support gap repair, leading to predictable rearrangements. When the tandem insertion roX1 [MS2-6]T2A is mobilized, homology-dependent gap repair can restore roX1 with no P-element sequences or with a 39 P-end retained (Figure 3, B and C). Flies that retain the 39 P-end also duplicate the 59 end of roX1 and, depending on the length of repair tract, have fulllength wild-type roX1 (roX1 P39 ) or roX1 with MS2 loops (roX1 P39MS2-6 ) ( Figure 3C). Retention of the 39 P-end is also possible following mobilization of the replacement line roX1 [MS2-6]R36A ( Figure S5). When the tandem insertion roX1 [MS2-6]T4B is mobilized, the 39 end of p[w +mC GM roX1 MS2-6 ] as well as the entire plArB element may be retained ( Figure S4). All of these alternative outcomes were readily identified by PCR (Table 2). Eight out of 18 MS2 loop-containing excisions of roX1 [MS2-6]T2A retained a 39 P-end. However, one of these is an imprecise excision that is mutated for w +mC but retains both P-elements in tandem ( Figure 3D). In agreement with the structure determined by PCR, this line also produced both 547-bp and 869-bp PCR amplicons when tested for presence of MS2 loops in roX1. Two out of 12 excisions of roX1 [MS2-6]T4B retained the 39 P-end and plArB ( Figure S4). No residual P-element sequences were detected in the eight excisions of roX1 [MS2-6]R36A examined (Table 2). We conclude that the overwhelming majority of excisions are repaired by a mechanism consistent with template-directed gap repair. Sixty-one percent of these had eliminated all vector sequences.
To confirm the structure of rearranged chromosomes, representative lines were analyzed by DNA blotting using the roX1 promoter region as probe ( Figure 3F). Excisions 2A.1 and 4B.1 are conversions to roX1 MS2-6 that retain no P-ends. Each produces a single 5.2-kb hybridizing EcoR1 fragment, consistent with introduction of 322 bp MS2 loops into the 4.9-kb genomic EcoR1 fragment. Line 36A.1, which displayed contraction of the MS2 loop array, shows a single hybridizing band at 5.1 kb (Figure 3 E and F). Line 2.1 retains no P-element sequences and has repaired roX1 without incorporating MS2 loops. As expected, a single 4.9-kb band is detected in this line. In contrast, the imprecise excision line 2.5 described has two hybridizing bands. The EcoR1 fragment present in p[w +mC GM roX1 MS2-6 ] is 5.2 kb, and a 2.5-kb band, consistent with insertional disruption of the chromosomal roX1 gene, is also present.
The roX1 MS2-6 is functional in dosage compensation The roX1 is functionally redundant with roX2 for dosage compensation. We tested the engineered roX1 MS2-6 allele for roX activity by determining adult male survival after recombination with roX2D, a deletion of roX2 (Menon and Meller 2012). Male flies inheriting roX1 MS2-6 roX2D chromosomes derived from three independent gene conversions were fully viable ( Table 3).

Mobilization of targeted insertions in females
Although roX1 MS2-6 was produced with high efficiency, excision was performed in males. Because roX1 is X-linked, no alternative template for repair is present. It is possible that mobilization in females would be less efficient because of selection of the homolog, rather than the sister chromatid, as the repair template. To test this idea, we mobilized the tandem insertion roX1 [MS2-6]T2A in females. Only 3 out of 131 white-eyed sons incorporated MS2 loops into the roX1 locus. This efficiency, 2.3%, contrasts with more than 10% MS2 loop incorporation in the offspring of dysgenic males. Two of the three lines contained a 39 P-end, and thus represent an alternative rearrangement.
Reduced efficiency of MS2 loop incorporation could result from use of roX1 + on the balancer chromosome as the repair template. Alternatively, it could reflect differences in the repair process in the male and female germ lines. For example, if repair tracts tend to be shorter in females, then inclusion of MS2 loops would be less frequent. To address these possibilities, we searched for P-element sequences on the repaired chromosomes. Retention of P-ends is expected when a sister chromatid template is utilized. We examined 125 randomly selected white-eyed offspring (including three with MS2 loops) for the presence of a 39 end; 103 out of 125 (82.4%) retained the 39 end. We then selected 29 flies at random (out of 125) and tested for the junction between the 39 end of p[w +mC GM roX1 MS2-6 ] and plArB. Twenty-one (72.4%) retained the junction. These findings are consistent with the idea that template-directed gap repair in females strongly favors copying of the sister chromatid.
Visualization of roX1 localization in roX1 MS2-6 embryos To visualize roX1 distribution in embryos, roX1 MS2-6 roX2D stocks carrying p[w +mC MCP-GFP] were generated. Females (roX1 MS2-6 roX2D; [w +mC MCP-GFP]) were mated to males carrying an X-linked n p[w +mC Sqh-mCherry] insertion. All embryos from this mating carry the roX1 MS2-6 roX2D X chromosome and a single copy of p[w +mC MCP-GFP], but females display weak mCherry expression throughout (Figure 4, compare 4C and 4H). MCP-GFP is recruited to a single, large, subnuclear domain in male (Figure 4, A-E) but not female (Figure 4, F-J) embryos. MCP-GFP in males overlaps the nuclear DAPI signal, and the domain occupied is a size consistent with Xchromosome painting ( Figure 4E). Examination of confocal Z-stacks from individual embryos reveals single GFP foci in virtually every nucleus (File S1, avi file).

DISCUSSION
The roX RNAs occupy a central position in fly dosage compensation. Full upregulation of X-linked genes does not occur in male roX1 roX2 mutants, and the MSL proteins mislocalize to ectopic autosomal sites (Meller and Rattner 2002;Deng and Meller 2006). Although autosomal roX transgenes rescue roX1 roX2 males, these transgenes also recruit MSL proteins to flanking autosomal chromatin, which is then modified in a manner similar to that at compensated X-linked genes (Kelley et al. 1999;Henry et al. 2001;Kelley and Kuroda 2003;Oh et al. 2004;Larschan et al. 2007). These observations suggest that position of roX genes on the X chromosome contributes to their normal function. More generally, the presence of complex or distant regulatory elements, or a requirement for a specific chromatin context, may contribute to deficiencies in the function of transgenics. Our objective was to generate an allele of roX1 that would function normally yet be readily visualized by GFP. The engineered allele roX1 MS2-6 supports full male viability in a roX2D background. Visualization of roX1 MS2-6 RNA with MCP-GFP reveals punctate labeling of a subnuclear domain in male embryos and does not require lengthy histological protocols, making roX1 MS2-6 a new resource for detection of roX1 localization.
The absence of readily accomplished homologous recombination in Drosophila is a notable drawback in a powerful model organism. Groundbreaking studies more than a decade ago established a technique for homologous recombination in flies, but this process remains labor-intensive (Rong and Golic 2000;Gao et al. 2008;Huang et al. 2009;Wesolowska and Rong 2010). More recently, a strategy for reinsertion of large clones that has been modified by recombineering has been shown to be quite efficient (Bateman et al. 2013). This and similar strategies that use site-specific recombination leave vector remnants or recombination sites within the genome (Crown and Sekelsky 2013). In contrast, we have introduced an engineered change with no residual vector sequences. Alternative rearrangements that retain a P-end can be predicted and easily detected by PCR.
We have named this new strategy Targeted Gene Conversion (TGC) to reflect the two-step process required: targeted transposition followed by gene conversion. TGC is a variation of older techniques that utilized repair-mediated gene conversion to engineer Drosophila genes. These relied on transposon mobilization to generate doublestranded breaks that were then repaired using a template supplied by the homolog (Gloor et al. 1991;Johnson-Schlitz and Engels 1993), by a transposon at another position in the genome (Nassif et al. 1994;Lankenau et al. 1996;Merli et al. 1996), or by DNA injected into dysgenic embryos (Banga and Boyd 1992). The efficiency of this process, typically not exceeding a few percent of excised chromosomes, has limited its use. In contrast, almost all excisions of targeted insertions containing the template are repaired using the template, and 10% of these incorporated MS2-6 loops into roX1.
Directed mutagenesis has been improved by the use of zinc finger nucleases (ZFN) and, more recently, TALENs and CRISPR/Cas9 nucleases, to introduce double-stranded breaks at specific sites (Bibikova et al. 2002;Christian et al. 2010;Bassett et al. 2013;Gratz et al. 2013). When repair templates with the desired changes are present, these sequences may be introduced by gene conversion (Gaj et al. 2013). The ability to rapidly generate mutations at a specific site is a clear strength of these methods. However, the efficiency of gene conversion decreases sharply with an increased distance from the break site (Moehle et al. 2007).
The potential for efficient replacement of longer sequences is anticipated to be a primary strength of TGC. Generation of roX1 SMC17A required resection of more than 1.2 kb from the break site, followed by copying more than 3 kb of sequence, including the Figure 4 The roX1 MS2-6 supports focal recruitment of MCP-GFP in male embryonic nuclei. Embryos were generated by mating roX1 MS2-6 roX2D; [w +mC MCP-GFP] females to males carrying an X-linked [w +mC Sqh-mCherry] transgene. Sons (roX1 MS2-6 roX2D/Y; [w +mC MCP-GFP]/+) lack w +mC Sqh-mCherry (A-E). Females express mCherry (F-J). A wildtype embryo reveals autofluorescence limited to the vitelline membrane (K-N). Details in (E) reveal MCP-GFP recruitment to a single domain within the male nucleus, consistent with X-chromosome painting. MCP-GFP recruitment is absent in the female nucleus (I, J). Each set of panels is derived from a single Z-plane image. The brightness of mCherry signals was uniformly enhanced for reproduction (C, H, and M). See Materials and Methods for details of photography and image processing.
n  (654) Male survival is based on the number of females emerging from each mating. Total numbers of flies recovered are provided in parentheses.
entire LacZ gene, into the break. Because this precise rearrangement accounted for 68% of excisions, TGC readily replaced large blocks of sequence. Gene conversion was also easily achieved on mobilization of roX1 [MS2-6]T or roX1 [MS2-6]R but, in this instance, no resection of broken ends was necessary to uncover homology with p[w +mC GM roX1  ]. Instead, incorporation of MS2 loops requires a repair tract to extend at least 750 bp from the break and to accommodate 322 bp of nonhomologous sequence. Ten percent of excisions incorporate MS2 loops, consistent with a previous study that documented conversion tracts extending almost 2 kb (Nassif and Engels 1993). The roX1 Δ891 and three targeted roX1  insertions are readily mobilized by transposase, with more than 90% of dysgenic males producing white-eyed offspring. This is not a general feature of P-element insertions in roX1, because only 20% of dysgenic roX1 mb710 males produce ry offspring (V. H. Meller, unpublished results). Despite high mobility, recovery of imprecise excisions was remarkably low. Four out of 56 excisions of roX1 Δ891 and 4 out of 352 excisions of targeted roX1  insertions were imprecise. The apparent high mobility and bias against imprecise excision are likely attributable to the presence of an alternative template for gap repair that excludes w +mC .
A clear limitation of our strategy is the need to move the template sequence close to the target site. We have accomplished this by targeted transposition, but targeted transpositions typically comprise a few percent of new insertions and require a P-element at the target site. The exceptionally rich coverage of P-element insertions in Drosophila makes this feasible in many instances. Alternatively, recently developed techniques that use engineered nucleases, such as TALENs, CRISPR/Cas9, or ZFNs, could be used to introduce a landing site, such as attP, at the desired location (Groth et al. 2004;Gaj et al. 2013). Integration of a selectable marker and template flanked by P-ends would generate a mutagenic precursor for TGC without the need for a preexisting P-element ( Figure S6).