Parallel Genomic Engineering of Two Drosophila Genes Using Orthogonal attB/attP Sites

Precise modification of sequences in the Drosophila melanogaster genome underlies the powerful capacity to study molecular structure-function relationships in this model species. The emergence of CRISPR/Cas9 tools in combination with recombinase systems such as the bacteriophage serine integrase ΦC31 has rendered Drosophila mutagenesis a straightforward enterprise for deleting, inserting and modifying genetic elements to study their functional relevance. However, while combined modifications of non-linked genetic elements can be easily constructed with these tools and classical genetics, the independent manipulation of linked genes through the established ΦC31-mediated transgenesis pipeline has not been feasible due to the limitation to one attB/attP site pair. Here we extend the repertoire of ΦC31 transgenesis by introducing a second pair of attB/attP targeting and transgenesis vectors that operate in parallel and independently of existing tools. We show that two syntenic orthologous genes, CG11318 and CG15556, located within a 25 kb region can be genomically engineered to harbor attPTT and attPCC sites. These landing pads can then independently receive transgenes through ΦC31-assisted integration and facilitate the manipulation and analysis of either gene in the same animal. These results expand the repertoire of site-specific genomic engineering in Drosophila while retaining the well established advantages and utility of the ΦC31 transgenesis system.

The amenability of the fruitfly's genome to targeted manipulation in combination with the vast phenotyping repertoire for this model species has enabled the precise interrogation of gene product functions. Several methodological advances have facilitated the use of directed genomic engineering in the fly. The advent of homologous recombination strategies enabled the exact targeting of genomic sequences in the fly genome, yet the stochastic nature of the occurrence of double-strand breaks (DSB) rendered this method a tedious and time-consuming venture (Gong and Golic 2003). Protocols that have made CRISPR/Cas9 (clustered regularly interspaced short palindromic repeats/CRISPR-associated 9) technology available to Drosophila genomic engineering have overcome these limitations, and single or multiple DSBs can since be exactly and efficiently induced at genomic targets (Bassett et al. 2013;Gratz et al. 2013a;Yu et al. 2013;Kondo and Ueda 2013;Ren et al. 2013;Sebo et al. 2014) . A DSB can either be inaccurately repaired by non-homologous end joining creating a palette of insertion-deletion mutations, the exact sequence of which cannot be controlled by the experimenter. Alternatively, in a gene replacement approach two DSBs release a defined genomic fragment that may harbor an entire gene or part of it (Gratz et al. 2013b). The concomittant provision of a DNA template via a homology directed repair (HDR) vector containing homology arms corresponding to the up-and downstream sequences of the released genomic fragment, a selection cassette for the identification of recombinant progeny, and sequences for their further genomic manipulation, offer an elegant means to pre-determine the precise layout of the engineered allele (Gratz et al. 2014;Port et al. 2014).
The incorporation of an attP (phage attachment) recognition site for the Streptomyces FC31 phage integrase within HDR vectors has become a standard procedure to replace an endogenous locus in Drosophila (Huang et al. 2009;Gratz et al. 2014). Subsequent integration of matching attB (bacterial attachment) site-encoding plasmids into attP-carrying flies together through germ-line expression of FC31 has greatly enhanced the speed, accuracy and reproducibility of fly transgenesis (Groth et al. 2004;Bischof et al. 2007). Once an attP founder fly line is established, FC31 transgenesis can be used to generate an unlimited number of allelic variants of the locus by restoring it with modified genomic fragments that contain mutations, in-frame fusions or other modifications of the genetic element of interest. Consequently, FC31-mediated transgene insertion enables high-throughput structure-function studies of genes and their products in Drosophila at single nucleotide and single amino acid resolution, respectively.
The analysis of genetic pathways, the study of gene homologs and the requirement of independent genomic modifications in the same animal requires multiple concurrent changes to its genome. However, the oneon-one compatibility of the attB/attP pair restricts the use of the FC31 platform to one locus per fly strain. The targeting of multiple attP sites at different genomic positions with non-identical transgenes in the same founder animal is not possible as each attP-landing pad is equally receptive to the insertion event. Consequently, if and where each plasmid integrates is stochastic violating the concept of site-specificity of FC31 transgenesis, one of its most compelling features. For non-linked genomic targets this problem can be solved by the manipulation of each locus of interest in an individual parental line, and their subsequent genetic combination through crossing. In contrast, linked loci on the same chromosome are not amenable to this option, particularly if their genetic distance is too small and thus the frequency of meiotic recombination to place them on the same chromosome is impracticably low.
Here we present an alternative approach to permit the manipulation and independent genomic engineering of linked loci through established FC31 integrase resources. FC31 cleaves double-stranded DNA at a central crossover dinucleotide within attB and attP sites generating a matching two-base pair 59-TT overhang in both ( Figure 1A, Table S1). Subsequently, the integrase swaps the half-sites and ligates the reciprocal partners creating hybrid attL and attR sites (Smith et al. 2004a). While the overhangs are essential for the recombination reaction, their sequences are not as long as they remain reverse-complementary to each other (Colloms et al. 2014). We have capitalized on this aspect of FC31 integration and adopted a matched attB/attP pair whose crossover dinucleotide consists of two cytosines ( Figure 1B, Table S1; here referred to as attB CC /P CC ) instead of the commonly used thymines in standard FC31 vectors for Drosophila ( Figure 1A; here referred to as attB TT /P TT ). This allowed us to use a selection of established FC31 integrase expressing fly strains with high integration efficiency without changes in the integration protocol. Existing plasmids for CRISPR/Cas9-mediated gene replacement and FC31 transgenesis were modified to encode the orthogonal attB CC /P CC pair.
This approach permitted the targeting and subsequent genomic engineering of two homologs of the adhesion GPCR (aGPCR) family (Hamann et al. 2015), CG11318 and CG15556, which are closely linked on a genomic fragment and separated through intervening genes on chromosome III. We show that our approach can be used to sequentially but also simultaneously integrate transgenes in a chromosome endowed with orthogonal attP sites using FC31 while maintaining efficiency, specificity and directionality of the targeting procedure. We demonstrate the utility of this approach by obtaining the co-transcriptional gene activity pattern of the CG11318/CG15556 gene pair in Drosophila.

Molecular reagents
All plasmids engineered herein were modified using restriction enzymes from New England Biolabs. PCRs were conducted using AccuStar DNA Polymerase (Eurogentec), primers and custom DNA fragments were synthesized by MWG Eurofins or Life Technologies. All intermediate and final constructs were DNA-sequenced to ensure no errors were introduced during the cloning procedures. The template genomic DNA used for PCR amplification throughout the study was from our stock of the w 1118 strain (Flybase ID: FBal0018186).
pHD-mW-attP CC -FRT (Addgene ID: 115158): The HDR vector contains a combination of principal elements of the pHD-DsRed-attP Figure 1 Orthogonal attP and attB site design. (A) Canonical attP/attB vectors contain a central TT dinucleotide at which the FC31 integrasemediated crossover between the two partner sequences occurs. The recombination event leads to the generation of hybrid attR and attL sites as indicated. (B) The orthogonal att site pair contains a CC instead of the TT sequence as cross-over nucleotide in both attP and attB vectors. (C,D) Sanger sequencing of genomic DNA of recombinant fly strains after insertion of attB CC+ transgenes into an attP CC+ landing site confirms that FC31 catalyzes the recombination between these non-canonical elements leading to the generation of (C) attR CC and (D) attL CC sites. Note the CC cross-over dinucleotide (boxed in blue) present in both hybrid sites. Chromatograms display the forward strand nucleotide sequence (upper strand in B), which was confirmed by corresponding reverse strand sequencing (lower strand in B; not shown).
n and pGX-attP vectors previously published by (Gratz et al. 2014) and (Huang et al. 2009). It harbors two multiple cloning sites on both sides of the replacement/mini-white marker element that are flanked by type IIS restriction sites, AarI (59 MCS) and SapI (39 MCS), respectively, to seemlessly insert homology arms for homology directed repair after CRISPR/Cas9-mediated cleavage of genomic sequences (Gratz et al. 2014). The mini-White marker element is flanked by two FRT sites for its subsequent removal by FLP recombinase expression. In addition, the replacement cassette contains a modified attP FC31 docking site with the central cross-over nucleotides changed from TT to CC (attP CC ). This way, pHD-mW-attP CC -FRT with its selection marker, marker removal sites and attP integration elements can be used in parallel and thus in combination with pHD-DsRed-attP.
pU6-gRNAs: CRISPR/Cas9 cutting sites 59 and 39 of the CG11318 and CG15556 loci suitable to remove all exons, UTRs and the promoter regions were identified by 'CRISPR Optimal Target Finder' (Gratz et al. 2014) ( Table 1). The genomic sequence of all CRISPR/Cas9 cleavage sites were confirmed by DNA sequencing of PCR fragments encompassing the suggested sites prior to cloning. Target-specific sequences for CG11318 and CG15556 gRNAs were synthesized as 59-phosphorylated oligonucleotides, annealed, and ligated into the BbsI sites of the pU6-BbsI-chiRNA vector (Gratz et al. 2013a).
CG11318-GAL4 reporter vector: A 4.6 kb fragment corresponding exactly the the genmic CG11318 sequence removed through the CRISPR/Cas9 cuts was amplified off genomic DNA with primers tl_768F/tl_769R, which contained NotI and AscI restriction sites, respectively. The DNA fragment was double digested with NotI and AscI and inserted into pGE-attB TT -DsRed to generate a wild-type CG11318 rescue vector (pTL784). In order to insert a GAL4.2 transcription factor cassette at the transcriptional start site of CG11318, a 1.6 kb AgeI/NsiI fragment of pTL784 was subloned into pTL550 (pMCS5 derivative with KanR; MoBiTec; pTL785). This subclone was outward PCR-amplified using primers tl_824F/tl_825R to generate a 4.6 kb amplicon. An 1.6 kb fragment encoding the optimized GAL4 cassette was amplified off pBPGal4.2::p65d (Pfeiffer et al. 2010) using primers tl_822F/tl_823R. Both PCR fragments were appended with primer-encoded BglII and NheI sites on either end, respectively, digested with BglII/NheI and ligated generating clone pTL787. A 3.2 kb AgeI/NsiI fragment of this clone was re-transferred into the CG11318 rescue vector pTL784 to construct the final CG11318-GAL4 reporter allele plasmid pTL789 (attB TT+ , loxP + , DsRed + ).
CG15556 LexA reporter vector: A 4.0 kb fragment corresponding exactly to the genomic CG15556 sequence removed through the CRISPR/Cas9 cuts was amplified off genomic DNA with primers tl_827F/tl_828R, which contained NotI and AscI restriction sites, respectively. The DNA fragment was double digested with NotI and AscI and inserted into pGE-attB CC (pTL788) to generate a wild-type CG15556 rescue vector (pTL790). In order to insert a LexA transcription factor cassette at the transcriptional start site of CG15556, a 1.4 kb EcoRI fragment of pTL790 was subloned into pTL550 (pTL791). This subclone was outward PCR-amplified using primers tl_834F/tl_835R to generate a 4.5 kb amplicon, which was appended with a BstEII site and was re-circularized at an AatII site introduced through both primers (pTL792). The so modified 1.4 kb EcoRI fragment of pTL792 was re-introduced into pTL790 generating pTL793. An 1.7 kb fragment encoding the LexA cassette with primer-inserted AatII and BstEII sites was amplified off pBSK-LexA-VP16-SV40 (Diegelmann et al. 2008) using primers tl_836F/tl_837R, cut with AatII/BstEII and inserted into pTL793 to generate the final CG15556-LexA reporter allele plasmid pTL794 (attB CC+ , FRT + , mW + ) ( Table 2,  Table 3).

Imaging
Third instar larval whole-mount specimen were fixed in 4% paraformaldehyde in PBS for 30 min at room temperature, rinsed phosphatebuffered saline with 0.1% (v/v) Triton X100, mounted on a slide using Vectashield H1000 (Vector Laboratories, Burlingame, USA) and a coverslip, whose edges were sealed with nail varnish. Confocal images were obtained with a Leica SP8 system. Z-stacks were collected at 4 mm intervals while capturing the entire larvae in XY dimensions. Individual Z-planes and maximum projections of the Z-stacks were inspected for expression pattern analysis.

Data availability
The fly strains described in this article are available upon request; the plasmids will be deposited at Addgene. Supplemental material available at Figshare: https://doi.org/10.25387/g3.6860723.

Construction of an orthogonal attB CC /attP CC pair
We have previously investigated structure-function relationships of the neuronal aGPCR homolog latrophilin/dCIRL using genomic engineering of the dCirl locus in Drosophila (Scholz et al. 2015;. In quest of additional potential fly aGPCR homologs we identified two genes on chromosome III that contain open reading frames encoding seven transmembrane-spanning (7TM) and GPCR autoproteolysis-inducing (GAIN) domains, the combination thereof being the molecular tell-tale signature of the aGPCR family ( Figure 2B) (Langenhan et al. 2013). CG15556 and CG11318 display high sequence conservation (data not shown) necessitating the construction of single and double knockout animals to account for possible functional redundancy. However, as both genes are closely linked on a 25 kb genomic fragment and separated through two additional genes ( Figure 2A) we sought to remove each gene separately through CRISPR/Cas9-assisted homologous recombination.
To allow for later gene-specific rescue and modification of each locus independently through FC31-mediated integration, we generated a set of vectors encoding an attB CC /attP CC site pair that can function orthogonally to attB TT /attP TT sites contained in standard genomic engineering vectors (Huang et al. 2009;Gratz et al. 2013a;Gratz et al. 2014): i. A homology-directed repair vector harboring an attP CC site, in which the central crossover dinucleotide was changed from TT to CC (pHD-attP CC -FRT-mW-FRT). To facilitate the selection of recombinant flies that were also targeted with pHD-DsRed-attP TT (Gratz et al. 2014), the plasmid additionally contains a FRT-flanked mini-White selection cassette for removal through FLP recombinase expression rendering all main characteristics of the vector orthogonal to pHD-DsRed-attP TT . All other elements including the multiple cloning sites for homology arm insertion are identical to pHD-DsRed-attP TT ( Figure S1A).
CRISPR/Cas9 targeting of CG11318 and CG15556 with canonical attP TT and novel attP CC sites CG11318 and CG15556 were individually targeted through CRISPR/ Cas9-mediated homology directed repair (Gratz et al. 2014) ( Figure  2C-E). Chimeric guide RNAs (gRNAs) for the gene targeting were selected to completely remove each gene, 59 and 39 UTRs and potential promoter regions ( Figure 2C). Homology arms were about 1 kb in length for each HDR vector and placed to immediately edge the Cas9 cleavage sites (Figure 2D,E). In a first round, CG11318 and CG15556 were individually targeted with a standard HDR plasmid (Gratz et al. 2014) replacing the genes with an attP TT docking site, whereas in a separate targeting round of CG15556 an HDR vector encoding the attP CC variant was used. Cremediated DsRed cassette removal in CG11318 recombinants and FLPmediated mini-White excision (Golic and Lindquist 1989) from CG15556-targeted animals was performed yielding single CG11318 KO attP TT+ and CG15556 KO attP CC+ knockout/knockin founder animals for both aGPCR loci. Subsequent PCR genotyping and sequencing confirmed the correct insertion of the replacement cassettes into the CG15556 ( Figure 2F; Table 4) and CG11318 ( Figure 2G; Table 4) loci.
To generate a chromosome lacking both genes, we selected a CG11318 KO attP TT+ founder strain, crossed it to a vasa-Cas9 background and targeted CG15556 by CRISPR/Cas9 as described above to yield CG15556 KO attP CC+ , CG11318 KO attP TT+ double mutant founders, which were verified by PCR genotyping and sequencing ( Figure  2H; Table 4). mW and DsRed markers were subsequently removed from these recombinants through consecutive rounds of Cre and FLP recombination to yield CG15556 KO attP CC+ mW -, CG11318 KO attP TT+ DsRedfounders (Table 4).

Integration of attB CC transgenes into attP CC landing pads
We next tested whether the novel attB CC /attP CC pair can be used for FC31-mediated recombination. We selected a CG15556 KO attP CC+ founder line and independently injected two constructs carrying cognate attB CC (CG15556 Rescue ; CG15556p LexA ) for FC31 mediated integration and recovered 2 and 3 recombinant founder animals, respectively ( Table 5). Sequencing of the genomic site of the attB/attP recombination for both transgenic integrants confirmed the formation of hybrid attR CC and attL CC sites on either side of the inserted DNA fragment ( Figure 1C,D; Table S1). This demonstrates that the attB CC / attP CC pair with exchanged overlap dinucleotides allows for directional transgene incorporation in Drosophila ( Figure 1B).
We continued to remove the mini-White marker cassette by standard FLP expression (Golic and Lindquist 1989) showing that the pHDR-mW-attP CC -FRT and its matching pGE-attB CC -FRT-mW partner vector can be used to complete FC31 assisted allele construction including transgenesis marker removal ( Table 5).

Orthogonality of attB TT+ and attB CC+ transgene integration
To evaluate the precision at which attP TT and attP CC sites, concomittantly present in the genome, are targeted we injected CG15556 KO w -attP CC+ , CG11318 KO DsRed -attP TT+ embryos expressing FC31 in the germline with either CG15556p LexA attB CC or CG11318p GAL4 attB TT reporter vectors, and recovered independent integrants from each injection (Table 5).
After expanding a balanced stock from each of the resultant CG15556p LexA+ mW + or CG11318p GAL4+ DsRed + founder animals, genomic DNA was harvested and subjected to PCR genotyping and DNA sequencing to assess into which attP integration site the vector was inserted. We found that the accuracy of attB TT /attP TT and attB CC /attP CC integration was complete (5 correct integrants/5 recovered integrants for CG11318p GAL4 ; 1/1 for CG15556p LexA ; Table  5). Notably, we observed no attB TT /attP CC or attB CC /attP TT recombinations in addition to correctly targeted integration events (0 wrong integrants/6 recovered integrants; Table 5).
This was corroborated by expression analyses of CG15556p LexA or CG11318p GAL4 reporter animals. Each CG11318p GAL4+ founder stock was crossed to 20xUAS-6xmCherry-HA or 20xUAS-6xGFP partners and the expression pattern or their progeny was analyzed by fluorescence microscopical inspection. All offspring exhibited identical gene expression throughout the gastrointestinal canal at all developmental stages (5/5), and none displayed gene activity in the CG15556 expression domain in the Malpighian tubule system (0/5; Figure 3A). Similar results were obtained for CG15556p LexA founder crosses with 13xLex-Aop2-6xmCherry-HA reporters, which all showed expression in Malpighian tubules (1/1) but none in the gut (0/1; Figure 3B). Both receptor gene expression patterns confirm whole transcriptome microarray and RNAseq datasets made available through the Flyatlas (Chintapalli et al. 2007) and modENCODE (Graveley et al. 2011) projects.
In sum, this indicates that the overlap dinucleotide difference in both attP sites allows for a sufficiently high specificity of cognate attB vector integration using standard FC31 expression and injection protocols, and that both attB/attP pairs function orthogonally to each other. Nonetheless, future projects that will increase the sample size of parallel targetings with this approach are warranted to gain a definitive estimate on the specificity of both attB/attP site pairs.
Integration of two transgenes in attP TT+ and attP CC+ animals Based on these results we finally tested whether two transgenes, one endowed with an attB TT+ and the other with an attB CC+ site, can be genomically inserted in the same animal. To this end we repeated the FC31 recombinations in two regimes: We successfully recovered double recombinants with both regimes (Table 5) demonstrating that the dual targeting of two loci is feasible in succession but also simultaneously, although for the latter expectedly at the expense of efficiency (Table 5). The precision of attB TT+ and attB CC+ transgene integration into their respective genomic landing pads within the same genome was confirmed by PCR genotyping ( Figure S2). In addition, we crossed a founder line carrying the linked CG15556p LexA CG11318p GAL4 transcriptional reporter alleles to a strain with matching lexAop-mCherry and UAS-6xGFP transgenes and observed expression domains for both genes in third instar larvae ( Figure 3C) and adults (not shown) that were identical to the ones obtained from the single reporter expression assays ( Figure 3A,B).

DISCUSSION
Here, we demonstrate a simple, easily adaptable and efficient system for the separate and repeated manipulation of two linked genetic loci in Drosophila (Figure 4). This protocol capitalizes on the currently most widely used genomic engineering toolkit for this model species, the FC31 integrase assisted transgenesis method (Bischof et al. 2007). Altogether, the results indicate that the specificity, directionality and recombination efficiency of the orthogonal attB CC /attP CC site pair introduced here can be used in conjunction with the canonical attB TT /attP TT system and established FC31 resources to handle two transgenesis targets independently and simultaneously.
This confirms results obtained in E. coli using an array of similar orthogonal attB/attP site pairs including the attB CC /attP CC version used here in Drosophila (Colloms et al. 2014). The successful recombination of attB CC /attP CC elements suggests that also in metazoan cells the parallel alignment of the recombination sites during synapsis of the DNA strands is not influenced by the nature of the crossover nucleotides as long as they are reverse-complementary to each other (Smith et al. 2004b). Likewise, our work further implies thatin addition to attB/ attP sites carrying TT/AA and CC/GG sequencesother attB/attP pairs with asymmetric central overlap dinucleotides (GT/CA; CT/GA; TC/ AG; CA/GT) will likely function as precise and independent, i.e., orthogonal, targeting addresses for FC31 integrase as established in E. coli (Smith et al. 2004b;Colloms et al. 2014). Therefore, this rationale can expand the parallel modifiability of genes in Drosophila to up to six genomic locations by a simple modification to the attP landing pads and associated attB-containing integration vectors.
Notably, site-directed transgenesis through integrases orthogonal to FC31 and with specific recognition sequences offers an alternative approach for the separate handling of multiple loci during genomic engineering. Several other such recombinases were shown to operate in Drosophila (reviewed in (Venken et al. 2016)). For example, recently, the mycobacteriophage integrase system BxbI was introduced for Drosophila transgenesis and shown to operate in parallel to FC31 (Huang et al. 2011;Voutev and Mann 2017). However, the combination of two integrases requires successive rather than parallel integration of transgenes doubling the processing time required for the establishment of doubly recombinant fly strains. In addition, the system is not widely used yet by academic laboratories and not offered by commercial suppliers for fly transgenesis services limiting the potential of this elegant tool and the principle of establishing orthogonality to FC31 transgenesis at the level of the employed integrase systems. Nonetheless, the insights and feasibility of alternative attB/attP site usage by FC31 provided in the current work is molecularly separate from those tools. It may thus even be combined with non-FC31 integrase mediated transgenesis tools, and in addition also be adopted by non-FC31 systems to expand their own recombinatorial transgene/target logic. Figure 4 Work flow diagram highlighting vector use and eye color screening paradigm for identification of successful transgene integrations using two sets of orthogonal attB/attP sites. Steps 1-5 and 7 require the sequential targeting of the loci of interest to deposit both an attP TT and an attP CC site in the same genome. Once an attP TT + attP CC founder stock has been constructed, step 6 can be performed sequentially but, importantly, also in parallel for both targets allowing for high-throughput contemporaneous engineering of two genes. Steps that entail plasmid construction are boxed and suitable vector backbones are indicated. Note that in animals that have co-received transgenes marked with mW + and 3xP3-DsRed + the latter marker is masked by the w + color in the eye (see arrowheads in fluorescence micrograph). However, the strong ocellar DsRed expression reliably indicates the presence of 3xP3-DsRed + transgenes (arrow in micrograph). Genotype of the photographed fly was w 1118 ; +; attP CC {CG15556-rescue mW + }CG15556 KO , attP TT {CG11318-rescue 3xP3-DsRed + } CG11318 KO .