Tools for Targeted Genome Engineering of Established Drosophila Cell Lines.

We describe an adaptation of φC31 integrase-mediated targeted cassette exchange for use in Drosophila cell lines. Single copies of an attP-bounded docking platform carrying a GFP-expression marker, with or without insulator elements flanking the attP sites, were inserted by P-element transformation into the Kc167 and Sg4 cell lines; each of the resulting docking-site lines carries a single mapped copy of one of the docking platforms. Vectors for targeted substitution contain a cloning cassette flanked by attB sites. Targeted substitution occurs by integrase-mediated substitution between the attP sites (integrated) and the attB sites (vector). We describe procedures for isolating cells carrying the substitutions and for eliminating the products of secondary off-target events. We demonstrate the technology by integrating a cassette containing a Cu(2+)-inducible mCherry marker, and we report the expression properties of those lines. When compared with clonal lines made by traditional transformation methods, which lead to the illegitimate insertion of tandem arrays, targeted insertion lines give more uniform expression, lower basal expression, and higher induction ratios. Targeted substitution, though intricate, affords results that should greatly improve comparative expression assays-a major emphasis of cell-based studies.

STABLE cell lines have formed an increasingly useful portion of the Drosophila melanogaster tool kit in recent years as the number of readily available lines has rapidly expanded, and many of those lines have been characterized extensively (Cherbas and Gong 2014). Over 100 diverse lines are now available through a cell line stock center maintained by the Drosophila Genomics Resource Center (DGRC); molecular characterization of many of the lines has occurred in many laboratories both as part of the modENCODE project and independently (Zurovec ; DasGupta ; Williams ; Lau ; Liu ; Schaaf ; Schwartz ; Cherbas ; Eaton ; Koppen ; Riddle , 2012; Vatolina ; Alekseyenko ; Brown ; Lee ; Wen ).

Stable transformation is a widely used tool in both flies and their cell lines; its power has increased in recent years as the random insertion of P elements has been supplemented by site-directed insertions of DNA into the chromosomes of flies. The use of integrase from the bacteriophage phiC31 to perform site-specific recombination is a particularly popular version of the latter approach (Huang ; Ejsmont and Hassan 2014). This technique is now well established in flies (Groth ; Venken ; Fish ; Huang ; Venken and Bellen 2012); it has been used for simple insertion of plasmids and much larger constructs (Venken ) via the recombination of a single attP site (either preexisting in the genome or inserted into the chromosome) with a single attB site in the targeting construct. It also has been used to mediate cassette exchange, in which a chromosomal DNA sequence bound by attP sites is exchanged for a plasmid sequence bound by attB sites (Bateman , 2012, 2013; Fujioka ; Huang ; Weng ; Sun ; Zhang ). The integrase is produced either from injected RNA (Groth ; Fish ) or from a stably integrated phiC31 integrase transcription unit that can be removed in a subsequent genetic cross (Bischof ). Targeted insertions and cassette exchanges make possible the repeated integration of constructs into an identical DNA environment, thereby eliminating variations caused by position effects.

In cell lines, phiC31 integrase–mediated targeting would confer improvements to currently used techniques beyond those seen in flies. Current techniques for stable transformation of Drosophila cell lines lead to the formation of tandem arrays of the transforming plasmid, often quite long, that are inserted by illegitimate recombination into the genome (Bourouis and Jarry 1983; Moss 1985; Cherbas ). This anomalous structure, which is also seen in transformed mammalian cells (Wurtele ; Rosser and An 2010) and to an extreme degree in a mosquito cell line (Monroe ), leads to abnormal chromatin structure, silencing of expression (Rosser and An 2010), pairing between arrays (Mirkin ), abnormal regulation caused by saturation of the supply of critical cis-acting factors, and instability in the length of the array. The resulting effects on regulation of transgene expression and the cell-to-cell variability in transformed lines, even after cloning, provide strong incentives to adapt targeted transformation techniques for cell lines. PhiC31 integrase–mediated gene targeting has been used in mammalian cell lines (Goetze ), and the integrase has been shown to function in Drosophila cell line S2 (Groth ). But targeted integration in Drosophila cell lines has proved difficult, and to our knowledge, the system has been pursued in only three laboratories: The Perrimon laboratory placed MiMIC elements, an enhancer-trap version of a phiC31 docking site, into S2R+ cells, and briefly described an integrase-mediated cassette exchange as a proof of principle (Neumuller ). The Simcox laboratory used the alternative approach of making new cell lines from flies carrying well-characterized attP docking platforms (Manivannan ). In the experiments described in this paper, we placed single copies of phiC31 docking platforms into well-characterized preexisting cell lines using P-element transformation of the cell lines and established conditions for carrying out phiC31 integrase–mediated exchange at these docking platforms. We describe here the generation of a set of tools for targeted insertion of constructs into Drosophila cell lines Kc167 and Sg4. We describe in detail cassette exchange in two of the Kc167 docking-site lines and compare the properties of the products of targeted exchange with those of stably transformed lines made with the same transgenes by more traditional means.

Materials and Methods

Cell culture

Kc167 and Sg4 cells were obtained from the collection of the DGRC; the former is a clone of Kc (Echalier and Ohanessian 1969; Bourouis and Jarry 1983), and the latter is a clone of S2 made by D. Arndt-Jovin (Schneider 1972). Kc167 cells were grown in serum-free CCM-3 Medium (GE Healthcare HyClone) unless otherwise indicated; Sg4 cells were grown in Shields and Sang M3 Insect Medium with added bactopeptone and yeast extract (M3 + BPYE) (Cherbas ) supplemented with 10% heat-inactivated fetal calf serum. General procedures for cell culture were as described previously (Cherbas ).

Cells were cloned by a modification of a procedure described previously (Cherbas ; Cherbas and Cherbas 2007). A feeder layer was prepared from cells of the parental line (Kc167 or Sg4) by pelleting cells, resuspending them in 5 ml of Robb’s saline (Robb 1969) in a 25-cm2 T-flask, and exposing them to 60 kR of gamma rays (cesium source). The irradiated cells were transferred to M3 + BPYE supplemented with penicillin (100 units/ml), streptomycin (100 μg/ml), and heat-treated fetal calf serum (5% for Kc167, 10% for Sg4) at a final concentration of 1.5 × 106 cells/ml. This feeder-cell suspension was plated in 96-well plates, 100 μl/well. Cells to be cloned were dispensed individually into the wells using a fluorescence-activated cell sorter (see later). After approximately 2 weeks, clones were picked, scaled up in their normal medium (CCM-3 for Kc167, M3 + BPYE + 10% serum for Sg4) as described previously (Cherbas et al. 1994a), and used for analysis and for the preparation of frozen stocks. Cloning efficiency was typically 10–20%. Although Kc167 cells and their derivatives are normally maintained in CCM-3, their cloning efficiency was near zero if they were dispensed by the cell sorter into a feeder-cell suspension in CCM-3; for this reason, we used M3 + BPYE with 5% serum for the feeder-cell suspension and reverted to CCM-3 for expansion of the growing clones.

Plasmid construction

Sequences for all the plasmids constructed for the experiments described in this paper are deposited in GenBank. We constructed two types of docking sites in the P-element vector Carnegie4 (Rubin and Spradling 1983), with and without gypsy insulator elements flanking a pair of parallel phiC31 attP sites. Both docking sites contain a nuclear eGFP expression cassette (driven by an promoter) between the attP sites. Maps for the two docking-site transposons are shown in Figure 1A. Vectors for targeting to the docking sites are shown in Figure 1B; these plasmids each contain a pair of parallel phiC31 attB sites flanking a methotrexate-resistance marker, with a herpes simplex virus (HSV) TK expression cassette conferring ganciclivir sensitivity located outside the attB sites. One of the vectors also contains a Gateway insertion cassette for use in inserting fragments to be transported to the docking site. Figure 1C shows an attB-bound region containing Mt-mCherry that was targeted to these docking sites. In act-phiC31 integrase, the coding sequence for phiC31 integrase was placed under the control of a strong constitutive promoter from . Sequences for all these plasmids are deposited in GenBank; critical portions came from the following plasmids: insulators and eGFP from pStinger (Barolo ), P-element ends from Carnegie4 (Rubin and Spradling 1983a), actin promoter and dihydrofolate reductase (DHFR) coding sequence from pUC-act-DHFR (Segal ), metallothionein promoter from pRmHa-1 (Bunch ), phiC31 integrase coding sequence from pET11phiC31polyA (Groth ), HSV TK coding sequence from pAL119-TK (Dewey ) (purchased from AddGene), attP sites from pXLBacII-attP-yellow forward (gift from Koen Venken), attB sites from attB-P[acman]-ApR (Venken ), Gateway entry cassette (purchased from Invitrogen), and mCherry coding sequence from pmCherry Vector (purchased from Clontech). Except where otherwise indicated, all the source plasmids were obtained from the vector collection of the DGRC.

Figure 1

Constructs used in this paper. Sequences for these plasmids and for the integrase-expression plasmid are deposited in GenBank. eGFP, DHFR, and HS-TK are each expressed from an Act5C promoter; mCherry is expressed from a MtnA promoter. (A) The P transposons used as docking platforms. Each contains a GFP-expression cassette between parallel attP sites; the encoded fluor is eGFP with a nuclear localization signal. IPPI also contains insulator elements flanking the attP sites. Only the P-element transposons are shown. (B) Vectors for integration by replacement in the docking sites. Each contains a MTX-resistance marker (DHFR) between parallel attB sites for positive selection and a HSV TK-expression cassette outside the attB-bounded region for counter selection against cells that have acquired the entire plasmid by illegitimate integration. The DHFR encoded in these plasmids is resistant to MTX and is of prokaryotic origin (Bourouis and Jarry 1983). HSV TK is a thymidine kinase derived from herpes simplex virus, and its expression renders cells sensitive to GCV. The two vectors differ by the presence in B-DHFR-GW-B-TK of a cassette for inserting constructs using the Gateway system (Life Technologies); B-DHFR-B-TK has a limited number of unique sites for inserting constructs, including an EcRI site upstream of the DHFR transcription unit and a ClaI site downstream of DHFR. (C) A Cu2+-inducible mCherry targeting sequence used for integration into the docking sites in experiments described in this paper. The targeting plasmid was made by cloning an Mt-mCherry transcription unit into the Gateway entry vector pENTRB (Life Technologies) and then using the Gateway reaction to place the transcription unit into B-DHFR-GW-B-TK. Only the fragment bounded by attB sites is illustrated here.

Fluorescence microscopy

To screen clones for expression of GFP, we examined clones growing in the original 96-well plates into which they had been sorted using a BD Pathway 435 High-Content Bioimager. For photomicrography, we placed 1 ml of growing cells into a 35-mm petri dish with a poly-d-lysine-coated glass bottom (MatTek Corp.); after the cells had settled onto the surface, they were visualized using an Applied Precision PersonalDV live cell imaging system. Both of these microscopes are housed in the Light Microscopy Imaging Center of Indiana University, Bloomington.

Fluorescence-activated cell sorting

All cell sorting and cloning were carried out in the Flow Cytometry Core Facility of Indiana University, Bloomington. Populations for sorting were selected for single cells by light scatter and for living cells either by light scatter (FSC-A vs. SSC-A) or by exclusion of propidium iodide dye. eGFP was excited with a 488-nm 100- or 30-mW laser, and emission was detected with a 530/30 bandpass filter. mCherry was excited by a 561-nm 150-mW laser, and emission was detected at 610/10. Propidium iodide was excited with a 561-nm 150-mW laser, and emission was detected at 582/15. Cloning, with or without fluorescence selection, was done on a FACSAria IIu cell sorter (BD Biosciences). Analysis without cloning was carried out on either the FACSAria IIu or an LSR II flow cytometer (BD Biosciences).

Molecular analysis by PCR

For digital-drop PCR (ddPCR), DNA was prepared from approximately 1.5 × 106 cells using a QIAamp DNA Micro Kit (Qiagen), yielding a final volume of 30 µl. For all other forms of PCR, we used a cell lysate (Gloor ) modified as follows: ∼1.5 × 106 cells were centrifuged and the culture medium removed. Pelleted cells were resuspended in 50 µl of squishing buffer [10 mM Tris-Cl, pH 8.2, 1 mM EDTA, 25 mM NaCl, and 200 µg/ml Proteinase K (Qiagen)] and incubated at 37° for 30 min. The lysate was then heated to 95° for 2 min to inactivate the Proteinase K.

Copy number was determined by ddPCR. Here 8.5 µl of DNA was digested with 10 units of EcoRI-HF (New England BioLabs) for 1 hr at 37°. Following a 20-min incubation at 65° to inactivate the enzyme, 1 μl of the digest was used for each 20-μl ddPCR reaction. Primer sequences are provided in Supporting Information, File S1. All copy number variation assays were duplexed with an reference assay; the region is known to be present in four copies in Kc167 cells (Cherbas and Cherbas 1997; Lee ). Reactions were set up using 2× ddPCR Super Mix for Probes (Bio-Rad), each 20× primer and probes [copy number variation (CNV) assay and reference assay] and digested DNA in a final volume of 20 μl. ddPCR was set up and performed as described by Hindson . Thermal cycling conditions for reaction emulsifications (Eppendorf Mastercycler) were 95° for 1 min and 94° for 30 sec and 62° for 30 sec (40 times, 50% ramp speed) and 98° for 10 min, followed by a 4° hold.

For clones that had a single copy of a docking site, the insertion site of the docking-site transposon was mapped by splinkerette PCR (Potter and Luo 2010) using 25 μl of cell lysate in place of the purified genomic DNA in the original protocol.

We used conventional PCR for additional characterization of docking sites and targeted insertion. Sequences of primers used for all PCR techniques are provided in File S1.

Transfections for integrase-mediated insertion into the docking sites

Docking-site lines were transfected using Lipofectamine LTX with PLUS Reagent (Life Technologies) and a mixture of the integrase-expression plasmid and an attB-targeting plasmid according to the manufacturer’s protocol. Four days after transfection, we began selection with methotrexate (MTX) as described previously (Cherbas ). After the MTX-sensitive cells had died and been replaced with a MTX-resistant population (∼2 weeks), we added ganciclivir (GCV, 20 mM final concentration) while continuing the MTX selection. Approximately 2 weeks later, we removed the selective agents and cloned GFP− cells.

Data availability

All cell lines and plasmids described in this paper are available from the DGRC (https://dgrc.bio.indiana.edu). Plasmid sequences are available from GeneBank (accession numbers KT894021-KT894026). File S1 contains sequences of PCR primers.

Results

Insertion of docking sites into the genome of Drosophila cell lines

Commonly used methods of transforming Drosophila cells generate multiple copies in tandem arrays. To insert single copies of a docking site into cultured cells, we used P-element transposition exactly as described previously (Segal ). We began with Kc167 cells, which were used in the earlier work on P-element transposition in cells, and subsequently repeated the procedure with the S2 derivative Sg4. In both cases, the transfection efficiency was low (as expected with electroporation), and to clone stably transformed GFP-expressing cells, we found it necessary to include an intermediate sorting step. We collected GFP-expressing cells 4 days after transfection, a time when much of the expression is still coming from plasmids not stably incorporated into the genome, to generate a population enriched for transformed cells. GFP+ cells were cloned from the enriched population ∼7–10 days later. Once the clones were large enough to visualize, we screened them for GFP expression in a fluorescence microscope. This last step was used to eliminate roughly 30% of the clones, which we presume derived from cells that either were transiently expressing GFP at the time of cloning and/or whose autofluorescence caused them to be scored as GFP+ by FACS. Autofluorescence, a significant source of error in FACS because the range of GFP fluorescence from cells carrying a single copy of the transposon overlaps the range of autofluorescence, is easily distinguished from GFP fluorescence in microscopy because autofluorescence is punctate and cytoplasmic, while GFP expression in these cells is nuclear.

The number of copies of the docking site was determined for each GFP-expressing clone by ddPCR, and clones carrying a single docking site per cell (6–40% of GFP-expressing clones in three experiments with Kc167, 20–30% in two experiments with Sg4) were expanded, saved as frozen stocks, and used for further analysis. For each single-copy clone, the insertion site of the docking site was determined by splinkerette PCR, a PCR protocol devised for the sequencing of sequences flanking P-element insertions (Potter and Luo 2010), and duplicate clones were discarded. Mapping of the insertion sites was confirmed by PCR using primers from the genomic regions flanking the insertion site. Table 1 lists the docking-site clones that we recovered: 1 IPPI insertion and 10 PP insertions in Kc167 and 6 PP insertions in Sg4.

Table 1

Positions of docking site insertions in Drosophila cell lines

Name of line	Molecular coordinate of insertion	Direction of insertion	Nearest annotated gene
Kc167-PP-16F	X:18,094,755	5′ toward centromere	RhoGAP16F
Kc167-PP-21B	2L:161,526	5′ toward centromere	spen
Kc167-PP-21D	2L:479,848	5′ toward centromere	cbt
Kc167-PP-50Aa	2R:13,343,299	5′ toward telomere	CR44206
Kc167-PP-50Ab	2R:13,337,584	5′ toward centromere	CR44206
Kc167-PP-52E	2R:16,125,298	5′ toward telomere	spin
Kc167-PP-61C	3L:635,370	5′ toward telomere	CR43334
Kc167-PP-89B	3R:≥16,189,680	5′ toward telomere	sra
Kc167-PP-93E	3R:<21,591,314	5′ toward telomere	InR
Kc167-PP-99A	3R:29,287,394	5′ toward telomere	CG14506 (10 kb away)
Sg4-PP-3A	X:2,545,583	5′ toward centromere	trol
Sg4-PP-27F	2L:7,423,926	5′ toward centromere	CR43857
Sg4-PP-49B	2R:12,589,942	5′ toward telomere	Sin3A
Sg4-PP-57B	2R:20,961,063	5′ toward centromere	hbn
Sg4-PP-70F	3L:14,757,988	5′ toward centromere	Trl
Sg4-PP-84E	3R:8,106,183	5′ toward telomere	puc
Kc167-IPPI-66D	3L:8,686,703	5′ toward telomere	h (7 kb away)

Names of lines are in the format [parental line]–[type of docking site (PP or IPPI)] – [site of insertion (given as the polytene region containing the insertion site)]. Molecular coordinates refer to the D. melanogaster genome, release 6. In cases where the coordinate is not given precisely, our sequencing reached within a few bases of the recombination site but did not cross the junction between the docking site and the chromosomal sequence. The direction of the insert is shown with 5′ taken as the left end of the map shown in Figure 1A.

Each docking-site clone has significant variation in the intensity of the GFP signal of individual cells and includes a small fraction of cells in which no GFP expression is detected either by FACS or by fluorescence microscopy (Figure 2). In Kc167-PP-93E, a typical docking-site line, approximately 1% of the population has no detectable GFP fluorescence. To characterize the GFP-null cells in Kc167-PP-93E, we stained cells with Hoechst 33342 and analyzed by FACS to estimate their position in the cell cycle. There was no significant difference in the distribution of Hoechst 33342 staining between the population as a whole and the GFP-null portion of the population (data not shown); hence, loss of GFP fluorescence does not appear to be associated with a stage of the cell cycle. We separately cloned GFP-expressing and GFP-null subpopulations of Kc167-PP-93E. GFP-expressing cells gave rise to healthy clones, each of which had a subpopulation of GFP-null cells indistinguishable from that of parental population (1.16 ± 0.21% for four clones). By contrast, GFP-null cells gave rise to small, unhealthy-looking clones composed overwhelmingly of GFP− cells. These observations suggest that occasionally Kc167-PP-93E cells permanently lose their ability to express GFP, but those cells grow poorly, leading to a steady-state level of a few percent. We speculate that loss of GFP expression occurs when a chromosomal rearrangement removes all or part of the docking site. The only docking-site lines in which significantly more than 1–2% of cells fail to express GFP are two lines in which the transposon is inserted near the tip of chromosome arm 2L (Kc167-PP-21B and -21D) (Figure 2); perhaps loss of one copy of this region is either more frequent or less deleterious than loss of other regions in which docking sites have inserted.

Figure 2

Expression of GFP in docking site lines. FACS-generated histograms are shown on the left; fluorescence photomicrographs on the right. Kc167 is the untransformed parental line, exhibiting only autofluorescence. Representative clonal docking site lines are shown below. The vertical blue line at 100 units GFP is provided for visual alignment. GFP-null cells are estimated at about 10% of the population in Kc167-PP-21D and 1% in Kc167-PP-93E. Any GFP-null cells in Kc167-IPPI-66D are masked by the overlapping range of GFP expression. Bar, 25 μm.

The cell-to-cell variation of GFP expression is easily visible in fluorescence microscopy (Figure S1) (note that the FACS and imaging data are not directly comparable because the linearity properties of the two techniques are quite different). The basis for this variability is unknown, but we observed a similar level of variation in another cell line that has a single copy of an unrelated GFP transgene: Jupiter (Karpova ) (data not shown). Thus, it appears to be generally true that in a clonal population of cells in which a single copy of a transgene expresses a stable fluorescent protein, the level of transgene expression is not uniform; we do not know the mechanism of this variation.

Targeted insertion by cassette exchange

We describe here the insertion of targeting constructs into two docking-site lines: Kc167-IPPI-66D and Kc167-PP-93E. We expect that similar procedures will give successful insertions in the remaining Kc167 and Sg4 docking-site lines, but they have not yet been tested.

When cells containing a docking platform marked by GFP expression are challenged with a plasmid that contains a MTX-resistance marker between two attB sites, in combination with an integrase-expressing plasmid, correctly integrated products should be resistant to MTX and should fail to express GFP; we initially tried to select targeted integration using either or both of these properties. Two problems, described individually, necessitated modifications to this scheme:

Loss of GFP was usually associated with loss of the entire docking platform rather than with targeted substitution. This became obvious when we cloned GFP− cells and examined the DNA of individual clones using PCR primers targeted to genomic sequences flanking the docking-platform insertion site. Both Kc167 and Sg4 are tetraploid at most loci (Lee ); hence, PCR reaction of each docking-site line produces a small amplicon from the three wild-type copies of the region and a much longer amplicon from the copy carrying the docking platform. When GFP− cells were cloned from a population that had been challenged with an attB targeting plasmid, however, the same PCR reaction almost always failed to produce any amplicon other than the small wild-type product. We conclude that loss of the genomic region containing the docking site (see earlier) occurs at a much higher frequency than targeted substitution; selection of GFP− cells was of little utility in isolating the correctly targeted products.

MTX selection of the transfected population efficiently selected for cells that carried the targeting construct. When we used PCR directed at attB, attL, and attR sites, however, we found that the vast majority of clones made from this population retained attB sites, and only a very small fraction had the recombinant attL and attR sites generated by targeted integration (data not shown). We conclude that illegitimate recombination (in which the entire targeting plasmid is integrated) occurs at much higher frequency than targeted substitution.

To permit selection against cells in which the entire attB plasmid has been incorporated by illegitimate recombination, we modified our original attB vectors by adding a HSV TK transcription unit, whose product renders cells sensitive to GCV; these modified vectors are shown in Figure 1 and were used in all the remaining experiments described in this paper. Because targeted substitution incorporates only the region flanked by attB sites, GCV selection should kill cells in which the entire plasmid is incorporated and spare those in which only targeted replacement has occurred. Thus, treatment with MTX to kill cells in which the targeting plasmid has not been incorporated plus GCV to kill cells in which the targeting plasmid has been incorporated illegitimately should enrich the population for cells with the intended targeted substitution only. The expected reactions are diagrammed in Figure 3, and the expected properties of their products are shown in Table 2.

Figure 3

Design of a targeted substitution experiment. (A) The desired reaction, catalyzed by phiC31 integrase, in which the act5C-eGFP marker of the docking site is replaced by Mt-mCherry plus a MTX-resistance marker. (B) The result of an illegitimate recombination (not catalyzed by phiC31-integrase) in which the entire targeting plasmid is incorporated at a random chromosomal site. See Table 2 for a summary of the predicted properties of these two transformation products.

Table 2

Properties of targeted and illegitimate products

	Parental line (PP or IPPI)	Targeted insertion	Illegitimate insertion	Both targeted and illegitimate insertions
Fluorescence:
GFP (nuclear)	+	−	+	−
mCherry (cytoplasmic)	−	+	+	+
Selection:
MTX   resistant	−	+	+	+
GCV   resistant	+	+	−	−
PCR:
attP	+	−	+	−
attB	−	−	+	+
attL	−	+	−	+
attR	−	+	−	+

We began by targeting the two docking-site lines with the empty attB vector B-DHFR-B-TK (Figure 1). We selected with MTX, followed by MTX + GCV. MTX selection was complete within 2 weeks, as reported previously (Cherbas ). GCV selection was inefficient, as shown by the fact that control populations transfected with the attB vector in the absence of integrase displayed only a transient slowing of growth when treated with GCV. The MTXR GCVR population then was cloned to isolate homogeneous populations of cells containing only the correctly targeted insert.

We performed PCR directed at attB, attL, and attR to distinguish between targeted insertions (which contain attL and attR but no attB), illegitimate insertions (which contain attB but no attL or attR), and a mixture of the two. Most of the clones from populations selected with MTX and GCV and failing to express GFP contained both correct and illegitimate insertions (data not shown). The integrase-expression plasmid was incorporated into the genome in only a very small minority of clones.

In a separate experiment, we targeted B-DHFR-mCherry-B-TK to the same two docking-site lines (Table 3). The results were similar to those seen with the empty vector, but in this case, we also assayed Cu2+-inducible expression of mCherry in clones that contained only a correct insert by the attL/attR/attB test. Surprisingly, a significant number of clones failed to express any detectable mCherry; this was particularly true when a relatively high level of act-integrase was used in the transfection and when the docking site Kc167-IPPI-66D was targeted. We speculate that following targeted insertion, a secondary integrase reaction involving cryptic attB and/or attP sites sometimes leads to a rearrangement of the DNA within the docking site; this notion is supported by the facts that these clones showed correct formation of attL and attR sites and loss of GFP expression and that loss of mCherry expression seems to be correlated with the amount of integrase-expression plasmid used in the transfection. Further experiments will be required to determine the nature of the rearrangement, but we are encouraged that such secondary rearrangements seem to occur relatively rarely at low concentrations of act-integrase.

Table 3

Products of RMCE transfection in Kc167 docking site lines

	IPPI-66D (1 µg integrase)	IPPI-66D (2 µg integrase)	PP-93E (1 µg integrase)	PP-93E (2 µg integrase)
Total	14	10	12	37
Passed B/R/L test:	11/14	9/10	12/12	33/37
Integrase lost	10/11	10/10	12/12	33/35
mCherry expressed	7/10	0/5	6/6	6/7

Data refer to clonal lines derived from the procedure described earlier targeting B-DHFR-Mt-mCherry-B-TK to the docking platforms in Kc167-PP-93E and Kc167-IPPI-66D. Transfections included 1 or 2 µg of act-integrase. All clones were MTX resistant and had no detectable GFP. The table indicates the number of clones that were expanded and analyzed and the fraction that passed the tests for correct integration only (attB absent, attL and attR present by PCR assay), lack of stable integration of the integrase plasmid (integrase absent by PCR assay), and expression of mCherry (FACS analysis of Cu2+-treated cells). For each fraction, the denominator is the number of clones from the previous step that were tested.

On the basis of these experiments, we recommend the following procedure for targeted replacement in our IPPI and PP docking site lines:

Cotransfect 1 ml of a docking-site cell line, at about 3–5 × 106 cells/ml, using Lipofectamine LTX with PLUS Reagent. For each transfection, use 0.5 μg act-phiC31 integrase plus 2 μg of B-Mt-mCherry-DHFR-B-TK (8.0 kbp) or an equivalent molar concentration of a similar attB-targeting plasmid.

After 2 days, transfer the cells to 5 ml of medium containing MTX (2 × 10−7 M final concentration). Change the medium every 4 days, retaining the MTX and diluting the cells as necessary to keep their concentration <1 × 107/ml, until healthy MTX-resistant cells dominate the population; this process generally takes ∼2 weeks.

Add GCV (20 μM final concentration) to the medium, and continue maintaining the cells in the presence of MTX and GCV, changing the medium every 4 days, until the growth rate is clearly depressed compared to a control culture containing only MTX; this process generally takes ∼2–3 weeks.

Using FACS, clone GFP− cells. When growing clones are clearly visible, use a fluorescence microscope to confirm the absence of GFP, and expand the GFP− clones. In the case of Mt-mCherry targeting, we were able to get similar results with a much smaller background of GFP+ clones by treating the population with CuSO4 for 20 hr prior to cloning and then cloning GFP− mCherry+ cells.

Prepare DNA from candidate clones, and use PCR to confirm the presence of attL and attR sites, the absence of attB sites, and the absence of the integrase plasmid. (We omit the targeting of attP sites for PCR because the background products of this reaction in untransformed Kc167 cells made the interpretation of attP results unreliable.)

Test candidate clones for expression of the targeting transgene (mCherry in our example) to eliminate clones with secondary rearrangements.

Properties of clones carrying targeted insertions

Stably transformed lines are routinely generated in Drosophila cell lines by illegitimate recombination between tandem arrays of transfected plasmids and random genomic sites (Moss 1985; Cherbas ). We expected that insertion of a single copy of a transgene into a docking site might give both increased stability of the transgene structure and more nearly normal chromatin structure and expression regulation than illegitimate insertion of tandem arrays of the same construct. For this reason, we compared the properties of cells containing either targeted or illegitimate insertions of an identical Mt-mCherry transcription unit. To maximize the homogeneity of the cell populations, we restricted ourselves to clonal lines. For illegitimate insertions, we used transfections that included the attB-Mt-mCherry donor plasmid but no source of phiC31-integrase; transformed cells were selected for resistance to MTX and then cloned. For targeted insertions, we used the procedure described earlier. Examples of mCherry expression patterns are shown in Figure 4. For each line, a FACS histogram of mCherry expression is shown in untreated cells and in cells treated for 20 hr with 1 mM CuSO4 to induce the expression of the Mt promoter. Autofluorescence was estimated from Kc167 cells (Figure 4A), which do not contain a coding sequence for mCherry. The two targeted lines (Figure 4, B and C) have low background expression in the absence of Cu2+ treatment and a strong induction by Cu2+; the two lines, in which Mt-mCherry was targeted to two different docking sites, differ both in the intensity and uniformity of mCherry expression. Replicate clones in which Mt-mCherry was targeted to a single docking site gave indistinguishable patterns of mCherry expression (data not shown). Photomicrography (Figure S1) confirms both the variation in intensity of mCherry expression and the complete absence of nuclear GFP that is expected in targeted substitution. By contrast, clones with illegitimate arrays of Mt-mCherry (Figure 4, D–F, and Figure S1) show a very broad range of background expression that is shifted to higher mCherry expression following Cu2+ treatment and retention of nuclear GFP expression; the three illegitimate clones shown in Figure 4 differ in the intensity of mCherry expression (both with and without Cu2+ treatment) and in the variation among cells in the population. In general, the targeted transformants show lower background expression, a higher induction ratio, and lower variation among cells than the illegitimate transformants.

Figure 4

Expression of a transgene in transformed clones. mCherry expression (arbitrary units) was measured by FACS analysis and is shown as histograms of cell populations with or without Cu2+ treatment (1 mM CuSO4, 20 hr) and as mean expression and Cu2+ induction ratio. (A) Kc167 (parental line); these cells do not contain an mCherry coding sequence; background autofluorescence measured in these cells is subtracted to give the mean mCherry expression estimates shown in the −Cu and +Cu columns. (B and C) Targeted substitution of Mt-mCherry into the docking site lines Kc167-IPPI-66D and Kc167-PP-93E, respectively. (D–F) Illegitimate insertions of the Mt-mCherry plasmid, derived from transformation of Kc167-IPPI-66D (D) or Kc167-PP-93E (E and F) in the absence of φC31-integrase, followed by MTX selection and cloning.

Discussion

We report here a series of transformants made from Kc167 and Sg4, each carrying a single copy of a docking site bearing two phiC31 attP sites designed for integrase-mediated cassette exchange. In this paper, we provide protocols for targeting transgenes to the docking sites and describe vectors for preparation of attB-bounded constructs for this purpose.

Distribution of P-element insertion sites

Data in this paper provide the first mappings of P-element insertion sites in somatic cells. FlyBase release FB2015_03 includes mappings of 67,543 separate P-element insertions into the germ line. Those sites are not randomly distributed; they show a preference for localized structural features (Liao ) and an association with DNA replication origins (Spradling ) and transcriptional activity (Fontanillas ). P-element “hot spots” are usually located in promoter regions (Spradling ), but only 2% of promoters accounted for over 40% of P-element insertions in a recent survey of over 18,000 independent transpositions (Spradling ).

Although our data for insertions into somatic cell lines are much too sparse to permit statistical comparison with those for germ-line insertion sites, even these restricted data are sufficient to show that the somatic insertions map preferentially in the vicinity of germ-line hot spots despite large differences in the patterns of transcription and of replication origins between germ cells and somatic cell lines (Cherbas ; Eaton ; Graveley ; Brown ). To illustrate this point, we note that most 2-kb segments of the genome have no known germ-line P-element insertions; using data from FlyBase, we surveyed 100 such segments and found that 82 of them contained no mapped germ-line sites (data not shown). Yet 13 of the 17 insertions into Kc167 or Sg4 lay within 1 kb of at least one known germ-line site, and 8 of the 17 lay within 1 kb of at least 10 germ-line sites. Figure S2 shows a typical example of a docking-site insertion that is clearly located in a hot spot for germ-line insertions.

Targeting: problems specific to cell lines

Spontaneous integration of exogenous plasmids into random positions in the genome is widely observed—and used—in Drosophila cell lines but not in flies. Our observations indicate that these illegitimate insertions are far more common in Kc167 cells than the integrase-mediated substitutions. When targeted replacement is done in flies, illegitimate insertions of the entire plasmid do not occur, and scoring the loss of a marker from a docking site and the appearance of a marker from the donor cassette was sufficient to give efficient identification of flies with the correct structure (Bateman ). In the cell lines, however, loss of the docking-site marker and appearance of a marker from the targeting plasmid is not sufficient to give useful enrichment for correct insertions, and it is necessary to include a counterselection, such as GCV, against cells with illegitimate insertions.

The efficiency of GCV selection as reported in mammalian cells is quite variable (Seibler ; Converse ; Chakraborty ), with the strength of the promoter driving HSV TK probably playing an important role. A similar cassette exchange experiment in a Drosophila cell line in which HSV TK was driven by an -GAL4 driver combined with a UAS-promoter gave efficient GCV selection (Manivannan ), and we suspect that the difference in selection efficiency between the two results from the amplification of TK expression produced by the GAL4/UAS system. We chose to use the promoter, a standard strong constitutive promoter in these cell lines, without the amplification conferred by the GAL4/UAS system in order not to preclude other possible uses of GAL4/UAS in subsequent experiments with the targeted cells. The consequence of this decision is that GCV selection provides enrichment for cells with the intended structure, but to achieve a pure population, it is necessary to clone the cells and use PCR to identify correct insertions. The use to which this technique is put will determine the relative value of strong GCV selection, which may make cloning unnecessary, and the weaker GCV selection conferred by the promoter, which permits one to use the GAL4/UAS system for other purposes in the resulting line.

We anticipate that parahomologous targeting (Cherbas and Cherbas 1997) also may present problems for targeted substitution in fly cell lines. When cells are challenged with an exogenous plasmid that contains a long stretch of homology with the fly genome (e.g., 10 kb or more), illegitimate recombinations are concentrated in the chromosomal region containing the homologous sequence probably because of somatic pairing between the plasmid and the chromosome. It therefore would not be surprising if competition between pairing and integrase-mediated recombination would significantly reduce the efficiency of targeted recombination when the targeting plasmid contains a long segment of chromosomal DNA. We have not tested this prediction.

Targeting: properties of the resulting lines

Using two of the docking site lines, one with and one without insulator elements, we have characterized the expression of a Cu2+-inducible Mt-mCherry marker inserted into the docking site. The properties of the targeted transgene are remarkably different from the properties of clones carrying the same transgene transformed by current methods in their stability and in their expression properties.

Currently, stable transformation of Drosophila cell lines is usually done by introducing the exogenous DNA using any of a variety of techniques and then selecting cells in which the DNA is incorporated into the genome. Incorporation into the genome occurs by illegitimate insertion, generally preceded by the formation of long arrays of the exogenous plasmid via homologous recombination. The insertion of these long arrays into the chromosome has been described extensively only for S2 cells transfected using calcium phosphosphate–DNA coprecipitates (Moss 1985; Cherbas ); more fragmentary data indicate that a similar process occurs in Kc167, though the length of the array may vary with the transfection procedure (Bourouis and Jarry 1983; Lee 1990). The products of these transformations are reasonably stable, as observed at the level of Southern blots (Moss 1985), but there is substantial variation from cell to cell. Such variation is to be expected in an uncloned population, but it occurs even in clonal populations (Fehon ; Lee 1990) (Figure 4 and Figure S1), consistent with the tendency of tandem arrays to undergo frequent deletions and expansions via homologous recombination. A single copy of a transgene inserted by targeted substitution would be expected to be much more stable and homogeneous; our measurements confirm this expectation (Figure 4). Insertion of a single copy of a transgene by P-element transposition (Segal ) also produces clones with reasonably homogeneous expression (Figure 2); this procedure has been available for 20 years but has been rarely used because it is much more cumbersome than illegitimate insertion. The targeted substitution procedure described in this paper is predicated on the insertion of single copies of a docking platform by P-element transposition; once the docking site lines are isolated and characterized, constructs may be targeted to these docking sites. The targeting step requires more time and effort than illegitimate integration and can be recommended only for applications where increased homogeneity and improved transcriptional regulation of the targeted inserts can compensate for the extra work. As more of these targeted transformants become available, we expect that they may be particularly valuable as substrates for rapidly emerging CRISPR/Cas techniques for mutation and replacement (Bassett ).

Our approach differs from that developed in the Simcox laboratory (Manivannan ) in several respects, each conferring both advantages and disadvantages. First, and most important, we have chosen to start with existing, well-characterized cell lines rather than establishing new cell lines from existing fly stocks carrying well-characterized docking sites. This makes it possible to employ the extensive data already available for the parental lines. Second, we have chosen to use attP sites (and therefore attB sites) in parallel rather than opposing orientation. This fixes the orientation of the resulting insertion. Third, we have used the promoter to drive expression of constitutive markers such as GFP, MTX, and GCV. The relative advantages of the two promoter systems were discussed earlier.

We do not yet have sufficient data to evaluate the effect of the insulator elements in the IPPI docking site and to compare position effects between IPPI docking site insertions and the uninsulated PP docking sites. The single IPPI insertion that we recovered produces relatively homogeneous expression from a targeted substitution of Mt-mCherry. However, it has proved to be difficult to recover more IPPI insertions because the level of GFP expression from IPPI docking sites is so low as to be difficult to distinguish from autofluorescence in cell sorting; cells that have GFP expression that is clearly higher than the autofluorescence signal almost always turn out to have several copies of the IPPI transposon. PP inserts have proved much easier to isolate because many of these inserts have much higher levels of GFP expression; it is quite likely that the PP insertions that we have isolated do not represent the full range of levels of GFP expression from this structure but instead only the more strongly expressed insertion sites, which are easier to isolate. Our observations are consistent with the predicted pattern, in which IPPI gives expression of GFP (and presumably of other markers substituted for the GFP cassette) that is relatively insensitive to position effects and PP gives a broader range of expression that is heavily influenced by the position into which it is inserted. Until more insertions of IPPI are isolated, however, and expression data are available from a number of IPPI and PP insertions, we can only speculate about the effectiveness of the insulators and the possible value of PP insertions for assessing cis-acting elements in flanking chromosomal regions.

Note added in proof: See Manivannan (pp. 1319–1328) in this issue for a related work.