An outline of the basic Screen and Insert approach is shown in Fig. 1A. A target construct is made in which a reporter gene (EGFP) is expressed from a tet-responsive promoter (TRP = TRE + CMVmin) with a loxP site positioned between the TRP and the reporter gene. Stably transfected clones, each with the target construct randomly integrated at a different site, are screened by flow cytometry to identify those with stringently tet-regulated EGFP expression. A chosen clone is then co-transfected with a Cre-expression plasmid and an insertion construct in which a promoterless GOI cassette is linked to a loxP site. Cre-mediated recombination between the insertion construct and the integrated target construct places the GOI under the control of the TRP in the chosen clone. Correct insertion events generate clones with stringently tet-regulated GOI and no GFP expression.

In the present study we have tested two of many possible ways in which the basic ScIn strategy can be implemented. We will refer to these as ScIn-1 and ScIn-2 and their key features are summarised in Fig. 1B and 1C, respectively. The target construct (pTARG4) used for both ScIn-1 and ScIn-2 carries a destabilised EGFP reporter gene (d2EGFP) driven by an optimised TRP (from pTRE-tight, Clontech). Positioned between the TRP and d2EGFP are a recognition site (FRT) for the Flp site-specific recombinase, and a mutant loxP site, lox71. The FRT site is necessary for ScIn-2 only. Lox71 was chosen because it is known to undergo unidirectional Cre-dependent recombination with lox66 26. Thus, by use of a lox66 site in the insertion construct, unimolecular Cre-mediated excision events that reverse the desired integration events, are minimised.

In ScIn-1, the GOI (e.g. luciferase) is linked to a promoterless drug resistance marker gene (e.g. hyg) by an internal ribosome entry site (IRES). With this arrangement it is possible to select in drug (hygromycin) for the desired insertion events. A limitation of ScIn-1 is that for drug selection to work the GOI must also be expressed. In ScIn-2 this limitation is avoided by use of an insertion construct in which a promoterless drug-resistance gene (e.g. gpt) is linked to the GOI (luciferase) by an FRT site. This arrangement allows insertion events to be selected in drug (Mycophenolic acid and xanthine [MPA/X]) without any accompanying GOI expression. Efficient Flp-mediated excision is then used to delete the drug-resistance marker and bring the GOI under the control of the TRP, in the presence of tet.

Target constructs (Fig. 2A) were introduced into host cells by co-electroporation with a plasmid (pBLpuroR) conferring resistance to Puromycin (see methods). Two HT1080 (human fibrosarcoma) derivatives were used as host cells: HT2 27, which expresses the original tTA protein, or Rht14 (see Methods) which expresses itTA 28, an improved tTA whose gene is modified to eliminate CpG dinucleotides and potential splice sites from the prokaryotic coding DNA and to introduce eukaryotic codon-usage. HT2 cells were used as a recipient for all target constructs except pTARG4, for which Rht14 was used. Most (>70%) of puromycin-resistant (puro^r) colonies expressed GFP as judged by fluorescence microscopy. These were expanded and tested flow cytometrically for tet-regulated GFP expression (methods). We sought to identify clones with: i) high GFP expression in the absence of tet, ii) low GFP in the presence of tet (as close as possible to the background fluorescence measured in untransfected parental cells), and iii) 'single peak' profiles, i.e. uniform expression throughout the population, whether in the presence or the absence of tet.

An initial target construct (pTARG1) used the original TRE 29. Of over 100 puro^r pTARG1 transfectants, many showed good regulation, but even the best expressed significant uninduced levels of GFP and required 144 h for full down-regulation (e.g. clone 6, Fig. 2B). To facilitate screening we used a second target construct (pTARG2) with d2EGFP as reporter. This reporter has reduced half-life compared to EGFP. Over 200 puro^r pTARG2 transfectants were screened, but none completely silenced GFP expression in the presence of tet. Flow cytometric profiles of the best-regulated clones are shown in Fig. 2C. When a target construct (pTARG3) with an improved TRE was used, 8 of 63 puro^r clones analysed had uninduced GFP expression levels of less than twice the background level, but these showed heterogeneous GFP expression in the absence of tet (Fig. 2D). Experiments with the methylation inhibitor 5-aza-2'deoxycytidine (AZC) indicated that heterogeneous expression was caused by DNA methylation of the tTA gene 30. For example, when T15, one of the clones transfected with pTARG3, was treated with AZC for 24 h, a marked increase in GFP-expressing cells was observed (Fig. 2E). Furthermore, luciferase expression in the same clone transiently transfected with a luciferase reporter gene linked to the improved TRE (pTIGHT-luc, Clontech) was also stimulated (2.6-fold) by a 24 h treatment of the cells with AZC prior to the transfection (Methods). This latter experiment suggested that methylation of the tTA gene, not the TRE, was responsible heterogeneous expression. The relatively homogeneous GFP expression seen in pTARG1 and pTARG2-transfected clones (Fig. 2C,D) probably reflects the use of lower passages of HT2 cells that were less likely to have suffered tTA gene silencing.

To minimise tTA gene methylation, the itTA-expressing HT1080 line Rht14 was made (Methods) and used for a final screening experiment. To allow subsequent testing of both ScIn-1 and ScIn-2, an FRT-tagged target construct (pTARG4) was used. Of 32 puro^r clones analysed, 7 had uninduced GFP expression levels of <2-fold above background combined with acceptable single-peak profiles (Fig. 2F), while others showed heterogeneous expression (Fig. 2G). The two best clones (Rht14-10 and Rht14-19) were chosen for further analysis.

Both Rht14-19 and Rht14-10 were found by Southern blot analysis (not shown) to carry a single copy of pTARG4. To test for stability of expression, clone Rht14-10 was cultured continuously for 42 days during which time the profile of GFP expression remained largely unchanged (Fig. 2H). A comparison of flow cytometric and immunoblot analyses of tet-induced GFP down-regulation in clone Rht14-19 illustrates one of the advantages of screening by flow cytometry, namely its sensitivity (Fig. 2I). Thus, substantial above-background GFP expression, readily detectable by flow cytometry, is undetectable on immunoblots. A second advantage of flow cytometric screening is its ability to detect and eleminate heterogeneously expressing clones, such as those in Fig. 2D and 2G, which would have been difficult or impossible to identify by bulk measurements of GFP expression (e.g. by immunoblot).

The insertion construct pIN-1 (Fig. 1B) was co-lipofected with a Cre expression construct (pMC-Cre) into 5 × 10⁵ Rht14-19 cells to generate 53 hygromycin-resistant (hygro^r) colonies. In a control experiment, identical except for the omission of pMC-Cre, no hygro^r colonies were generated. Of 48 hygro^r colonies examined, 47 had lost GFP expression and became sensitive to hygromycin when tet was added (not shown), both expected consequences of the desired integration event. DNA fragments diagnostic for the expected insertion event were detected in all (10/10) of the GFP-negative, hygro^r clones analysed by PCR and Southern blots (Fig. 3A–C). The same 10 clones also showed similar patterns of tet-regulated luciferase expression (Fig. 3D). These results show that the desired Cre-mediated insertion step occurs at a frequency of approximately10^-4 per transfected cell (Table 1) to generate tet-regulated GOI expression in a reproducible manner. Given the stringency of GFP expression in clone Rht14-19, the background of luciferase expression in the insertion clones was surprisingly high. Based on the better stringencies obtained with clone Rht14-19 using ScIn-2 (next section), this background may indicate some activation of the TRP by unidentified sequences in pIN-1.

The insertion construct pIN-2 (Fig. 1C) was co-lipofected with pMC-Cre into 5 × 10⁵ Rht14-10 or Rht14-19 cells to generate 11 and 32 MPA/X-resistant (MPA/X^r) colonies, respectively. No MPA/X^r colonies developed in the control transfections lacking pMC-Cre. Of the 11 Rht14-10-derived colonies, 9 had lost GFP expression and became sensitive to MPA/X when tet was added (not shown). Of the 32 Rht14-19-derived colonies, 22 had lost GFP expression and became sensitive to MPA/X when tet was added (not shown). Nine GFP-negative, MPA/X^r clones from each parental clone were analysed by genomic PCR, and all were positive for the 629 bp PCR product diagnostic for the expected integration (Fig. 4A,B). Four insertion clones, one derived from Rht14-10 (10IN5) and three from Rht14-19 (19IN3, 19IN4 and 19IN5), were analysed by Southern blot and all showed the expected change in structure (Fig. 4A,C). Together, these results indicate that the desired insertion event occurred at frequencies of 1.8 × 10^-5 and 4.4 × 10^-5 per transfected cell in clones Rht14-10 or Rht14-19, respectively (Table 1).

Two insertion clones (10IN5 and 19IN5), one of each parental type, were chosen for Flp-mediated deletion. Cells grown with tet (without MPA/X) for at least 72 h were lipofected (without [10IN5] or with [19IN5] tet for 4 h; see methods) with the Flp expression plasmid pCAGGS-flpe (Cambion), and cultured at low density with tet until colonies appeared. Colonies were analysed by genomic PCR. For Flp-treated 10IN5 cells, 16 pools (6 clones/pool) were assayed and 9 pools produced the predicted 472 bp PCR product (Fig. 4Ac) indicative of Flp-mediated deletion (Fig. 5A). Individual clones from two positive pools (5 and 9) were analysed by the same PCR assay and 6/12 were positive for the 472 bp product (Fig. 5A). The efficiency of flp-mediated deletion was thus at least 0.14 (13/96), and probably closer to 0.28 (6/12 × 9/16), events per transfected cell. A Southern blot of two such clones (10flp9.2 and 10flp9.3) was consistent with the predicted structures (Fig. 4A,C). The 7 Flp-deleted clones showed highly stringent tet-regulated luciferase expression (Fig. 5B). Above-background levels of uninduced luciferase expression were detectable, but all clones showed induction ratios approaching 10,000-fold, consistent with the GFP inducibility of parental clone Rht14-10.

For Flp-treated 19IN5 cells, 12 pools (6 clones/pool) were assayed by PCR (Fig. 5C). Two pools (9 and 12) were positive, each due to the presence of a single clone 19flp9.2 and 19flp12.3, respectively. The relatively low efficiency of Flp-mediated deletion in 19IN5 (0.03 [2/72] events per transfected cell) probably reflects a lower efficiency of pCAGGS-flpe delivery due to the inclusion of tet in lipofection mix of that particular transfection. Southern blot (Fig. 4C) and PCR analysis (Fig. 5C), showed that 19flp9.2 was impure, the majority of cells being of parental (19IN5) type. Southern analysis of the other flp-deleted clone (19flp12.3), however, confirmed that it had undergone the expected deletion (Fig. 4C). Clone 19flp12.3 showed highly stringent tet-regulation of luciferase expression with an induction ratio of more than 10,000-fold and uninduced expression levels barely above background (Fig. 5D).

Having validated the ScIn method, we put it into practice using ScIn-2 with genes for I-SceI or Rad52 as our GOIs. For these experiments a modified insertion pIN-2 vector (pIN2-neoMCS) was used with a different selectable marker gene (neo in place of gpt) followed by a multiple cloning site (Fig. 6A). The open reading frames (ORFs) encoding I-SceI or Rad52, the former with a N-terminal Haemaglutinin (HA) tag, were cloned into pIN2-neoMCS to generate pIN2-neoSCE and pIN2-neoR52. These plasmids were co-lipofected with pMC-Cre into Rht14-10 and G418^r colonies were selected, > 96% of which had lost GFP expression. Of six such clones tested, three each of the pIN-neoSCE- and pIN-neoR52-transfected clones, all were found to become sensitive to G418 in tet and to be positive in PCR assays for targeted integration (Fig. 6B,C). In this way site-specific insertion frequencies of between 3.3 and 4.3 × 10^-4 were estimated for pIN2-neoMCS and it derivatives, at least 20-fold higher than with pIN-2.

Two clones, 10IN-SCE.1, 10IN-R52.1 (with pIN2-neoSCE and pIN2-neoRAD52 integrated, respectively) were grown in +tet medium, lipofected (- tet for 4 h; see methods) with pCAGGS-flpe and plated at low dilution in +tet medium. The majority of the resulting colonies were positive in a PCR assay for the deletion event, suggesting highly efficient Flp-mediate deletion (Fig. 6C–E). Two clones of high purity (10IN-SCE.1flp1, 10IN-R52.1flp1) as judged by PCR were analysed further. Immunoblot analyses confirmed the expected tet-regulated expression of the appropriate GOI (I-SceI or Rad52) in these clones (Fig. 6F). Expression in the presence of tet was undetectable indicating, as far as is possible by immunoblotting, highly stringent tet-regulation.

To test for functionality of the induced I-SceI, and as a more robust test for the stringency of tet-regulation, 10IN-SCE.1flp1 was transfected with pDRneo 31. This substrate for homologous recombination carries a functional hygromycin resistance cassette flanked by two defective neo cassettes, one being disrupted in its coding sequence by the 18 bp recognition site for I-SceI. Cleavage of the first neo cassette by I-SceI greatly stimulates its conversion to a functional neo cassette by homology directed DNA repair using the second neo cassette as a template. Before transfection with pDRneo, the purity of 10IN-SCE.1flp1 was further demonstrated by plating three million cells in G418 without tet: no colonies formed even though parental 10IN-SCE.1 cells grow in these conditions. Following transfection of 10IN-SCE.1flp1 with pDRneo, three Hygro^r clones, selected and expanded with tet in the growth medium, were tested for the presence of G418^r cells after a period (2 days) of growth with or without tet. The results (Table 2) show that I-SceI induction causes > 10⁴-fold stimulations in gene conversion, with absolute frequencies reaching 3–7% of all cells. These effects are at least as pronounced as those previously described after transient transfection of an I-SceI expression construct 31. Variations between clones in the absolute frequencies of G418^r are most likely caused by different chromosomal position effects on the integrated pDRneo. The low frequencies of G418^r colonies (1–5 × 10^-6) detected in the continuous presence of tet are likely to reflect spontaneous recombination of the pDRneo substrate rather than recombination stimulated by residual I-SceI expression: similar frequencies of spontaneous G418^r have been described for pDRneo in HT1080 32, HEK293 32 and CHO 31 cells that have never been transfected with an I-SceI gene. Furthermore, these frequencies are no higher than the spontaneous G418^r frequency (1.3 × 10^-5) observed for a hygro^r clone derived from pDRneo-transfected Rht14 cells (Table 2). This frequency was also unaffected by tet removal, excluding the possibility that tet reduced G418^r colony formation independently of its effect on I-SceI expression. Taken together, these results clearly demonstrate that I-SceI is functional and regulated with very high stringency following its targeted integration by use of the ScIn-2 method.

A plasmid (pRK5-itTA) encoding the improved transactivator itTA 28 was kindly donated by R. Sprengel, Max-Planck-Institute, Heidelberg. The itTA ORF was cloned as an EcoR1/BclI fragment into the EcoRI/BclI sites of pZeoSV (Invitrogen), to create pZeoSVitTA. pTIGHTgpt was made by inserting the gpt ORF as a BglII/BamH1 fragment from pBSgpt into the BamH1 site of pTRE-TIGHT.

Conditions used for the culture of HT1080 cells and derivatives have been described previously 63. When required the medium was supplemented with one or more of the following drugs: 5-azacytidine (1 μM), hygromycin B (100 μg/ml), mycophenolic acid (10 μg/ml), puromycin (0.4 μg/ml), tetracycline (1 μg/ml), xanthine (100 μg/ml) and zeocin (200 μg/ml). HT-2 cells have been described 27 and were also called HTET 13. A telomerase-immortalised human retinal epithelial cell line (hTERT-RPE1; Clontech) was cultured in a similar manner to HT1080 cells.

Electroporation with a Gene Pulser (BioRad) was as described 64. Rht14 cells were made by electroporation of HPRT⁺ HT1080 cells with 10 μg BglII-linearised pZeoSVitTA. Zeocin^r colonies were screened by lipofection (see below) with pTIGHTLuc (Clontech) grown in the presence or absence of tetracycline. After 48 hours, lysates were prepared and assayed for luciferase. Rht14 was chosen from among 24 colonies for its low luciferase activity in the presence of tet and its high induction ratio. Co-electroporation was used to generate clones with randomly integrated target plasmids. Target plasmid (20 μg) was linearised with PvuI (pTARG1) or BglI and SspI (pTARG2, pTARG3, pTARG4), gel-extracted, ethanol-precipitated and co-electroporated with 1 μg of similarly purified, SpeI-linearised pBL-PuroR 32, into Rht14 cells. Puro^r colonies identified as GFP-positive by fluorescence microscopy (Zeiss Axiovert S100TV microscope) were picked for further analysis.

Lipofectamine 2000 (Invitrogen) was used for delivery of insertion constructs and transient expression of Cre and Flp recombinases and luciferase. 24 hours prior to transfection, 250,000 cells were plated into 6-well plates in 2 ml of antibiotic-free (- tet, unless otherwise stated) medium. On the day of transfection a DNA/Lipofectamine/OptiMEM mix was prepared (0.5 ml/well) according to the manufacturer's instructions and incubated with the cells for 4 hours, after which the mix was replaced with fresh medium. For Cre-mediated insertion, 2 μg of insertion construct and 2 μg of pMC-Cre15 (kindly donated by H. Gu, University of Köln) were used per well. The following day, the cells were seeded into 9 cm plates (10⁵ cells per plate) and the appropriate drug selection was added after a further 24 hours. For flp-mediated excision, 4 μg of pCAGGS-flpe (Cambion) was used per well and cells were plated the following day at low density (50–500 cells per 9 cm [diam] dish) in medium with tet. To analyse the effect of AZC on luciferase expression from pTIGHT-luc, clone T15 cells were grown with or without AZC for 24 h, then plated in antibiotic-free medium and lipofected 24 h later (as above) with pTIGHT-luc (4 μg). After a further 24 h lysates were prepared and assayed for luciferase (see below).

Cells were grown in 6-well plates with or without tet for 48–72 hours (except where longer times are indicated) prior to FACS analysis. Cells, generally no more than 80% confluent on the day of analysis, were trypsinised and resuspended in cold phosphate buffered saline (PBS) at a density of 1000 cells per μl. 40,000 cells were analysed in a FACScan machine (Becton Dickinson) with an argon laser tuned to 488 nm (FL-1; with fluorescence channel FL-3 as a control). Acquisition, storage, and analysis of data were carried out using CellQuest software (Becton Dickinson).

Immunoblots were carried out as described previously for Rad52 65. For GFP detection, a primary monoclonal antibody was used (Clontech, 632380; 1/1000 dilution) and a secondary, horseradish peroxidase-conjugated goat anti-mouse immunoglobulin antibody (Sigma, P-0447; 1/1000 dilution) was used. For HA-tagged I-SceI detection, the primary antibody was a rat monoclonal (Roche, 3F10; 1/500 dilution) and the secondary a horseradish peroxidase-conjugated goat anti-rat immunoglobulin antibody (Sigma, A-9037; 1/1000 dilution)

Cell pellets (100–10,000 cells) were resuspended in a 25 μl Taq polymerase buffer (Qiagen) containing pronase (0.6 μg/μl), incubated at 50°C for one hour, 95°C for 10 minutes and then placed on ice. A mix (25 μl) containing nucleotides (Pharmacia, 0.5 μM), Taq polymerase (Qiagen, 1.25 U) and oligonucleotides (100 ng each) dissolved in in Taq polymerase buffer was added. Annealing temperatures (T_a) for the various primer pairs were 60°C (O2/O4 and O1/O3), 63°C (O2/O5), and 57°C (O2/O6 and O2/O7). Temperature sequences were: 95°C (10 min.) then 30 cycles of 95°C, T_a, and 72°C (1 min each) then 72°C (10 min.). Primers were: O1 (5'-ACGAGGCCCTTTCGTCTTCA-3'), O2 (5'-TTTAGTGAACCGTCAGATCGCC-3'), O3 (5'-CCACGGCTTACGGCAATAATGC-3'), O4 (5'-CCCCTTTTTGGAAACGAACAC-3'), O5, (5'-AGAAGGCGATAGAAGGCGATGC-3'), O6 (5'-ACTCGAACTGCATACAGTAG-3') and O7 (5'-CCAAAGATAAAGCCTTGGAC-3').

As positive controls for the DNA template, ~10 pg of the indicated plasmid DNA, or genomic DNA from 1000 pelleted cells of the indicated cell line, were used. TetNeo cells have a tet-regulated neo gene and are described in detail elsewhere 30. Briefly, plox66Neo was inserted by Cre-mediated recombination into tTERT-RPE1 cells with an integrated pTARG4 supporting tightly regulated GFP. PCR products were analysed by standard agarose gel electrophoresis with 1 kb ladder marker DNA (Invitrogen).

Standard methods were used, as previously described 61. The GFP probe was a 727 bp NcoI fragment from pd2EGFP [Clontech]). Marker DNA was end-labelled 1 kb ladder (Invitrogen).

A commercial kit (Promega) was used according to manufacturers instructions. Briefly, cells were seeded at 10⁵ cells per 3.5 cm (diam.) well, incubated in medium with or without tet for 48 hours, then washed with PBS and agitated gently for at least 20 minutes with 500 μl passive lysis buffer. The cell lysate was cleared by centrifugation and 20 μl added to 100 μl of luciferase assay reagent in a luminometer tube. The relative light units (RLU) were measured immediately in a Biorbit 1253 luminometer. The average of ten RLU measurements taken at 2-second intervals was recorded.

To estimate the recombination frequency when I-SceI was not expressed, 3 million cells were distributed in three 15 cm (diam.) plates and selected in G418 and tetracycline and colonies counted after 14 d. To determine the frequency after I-SceI induction, cells were grown in the absence of tet for 48 h and then plated (at 50, 100, 250 cells per 9 cm [diam.] plate) in G418 and tetracycline and colonies counted after 14 d. To account for the plating efficiency, the proportion of G418^r clones was calculated as a fraction of the total number of colonies formed when compared to the amount generated on equivalent unselected (no G418) plates. Rht14/DRneo cells were analysed in the same way except that tet was either present or absent from the growth medium throughout, and for 48 h prior to, G418 selection.