|Home | About | Journals | Submit | Contact Us | Français|
The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the open access version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal and Oxford University Press are attributed as the original place of publication with the correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact firstname.lastname@example.org
Comparative analysis of mutants using transfection is complicated by clones exhibiting variable levels of gene expression due to copy number differences and genomic position effects. Recombinase-mediated cassette exchange (RMCE) can overcome these problems by introducing the target gene into pre-determined chromosomal loci, but recombination between the available recombinase targeting sites can reduce the efficiency of targeted integration. We developed a new LoxP site (designated L3), which when used with the original LoxP site (designated L2), allows highly efficient and directional replacement of chromosomal DNA with incoming DNA. A total of six independent LoxP integration sites introduced either by homologous recombination or retroviral delivery were analyzed; 70–80% of the clones analyzed in hamster and human cells were correct recombinants. We combined the RMCE strategy with a new, tightly regulated tetracycline induction system to produce a robust, highly reliable system for inducible transgene expression. We observed stable inducible expression for over 1 month, with uniform expression in the cell population and between clones derived from the same integration site. This system described should find significant applications for studies requiring high level and regulated transgene expression and for determining the effects of various stresses or oncogenic conditions in vivo and in vitro.
Many studies require the ability to compare the functional consequences of introducing mutations into genes of interest. Rigorous comparative analyses require similar gene expression levels in each stable cell clone. Comparison between mutant genes is difficult using the common strategy of deriving stable cell lines as gene expression level is unpredictable due to variable copies of integrated DNA combined with position effects on gene expression. All of these confounding variables can be eliminated if each transgenic clone is derived by independently introducing the gene of interest into a pre-determined genomic locus (1).
Gene targeting by homologous recombination is widely applied in mouse embryonic stem cells (ES cells) and occasionally in other cell types to produce accurately modified loci (2). However, it is a sophisticated and time consuming procedure that can be compromised by the low recombination efficiency of some loci. Alternatively, the RMCE strategy uses bacterial and yeast DNA recombinases to target any DNA sequence into a pre-determined genomic locus that was previously modified to contain the appropriate recombinase recognition sequences (3). For example, the widely used bacteriophage P1 Cre/LoxP recombination system utilizes LoxP sites consisting of two 13 bp inverted repeats and a central 8 bp spacer sequence. The central 8 bp sequence within each LoxP site determines if a pair of LoxP sites is compatible for recombination by Cre (4). LoxP sites containing the same 8 bp core sequences recombine, but LoxP sites differing by one or more base pairs exhibit reduced or no recombination (4–6). The success of RMCE relies on the use of two heterospecific LoxP sites (LoxP sites of different sequences) that will not recombine with each other. Thus, flanking a donor gene of interest with heterospecific sites enables it to precisely replace genomic DNA flanked by identical sequences upon expression of Cre-recombinase. However, the previously published LoxP sites (L2) and LoxP511 (L1) (4) are not entirely heterospecific when tested in mammalian cells, and are consequently not ideal for RMCE (1,6). For example, deletion of the intervening sequence occurs when L2 and L1 are placed as direct repeats on the same DNA, while inversion occurs if they are placed as inverted repeats (1). Due to the wide usage of RMCE, the development of new LoxP sites with greater specificity for directional exchange is clearly needed.
RMCE can also facilitate structure-function analyses by ensuring reproducible expression of different mutant alleles from the same locus. However, the biological significance of such studies can be compromised if the expression level does not reflect that present physiologically. For that, a reliable inducible system with low basal expression and reproducible ligand-dependent induction is required. Tight regulation of gene expression is particularly important if, e.g., the gene of interest is toxic, or induces cell cycle arrest or apoptosis. The tetracycline (tet) regulatable system has proven especially valuable as it uses an inexpensive inducer (doxycycline or dox), has high dynamic range, low background in un-induced state and can be used in vivo and in vitro (7). Regulation in this system involves highly specific interaction between the Tet repressor (TetR) and Tet operator (tetO) DNA sequence (Tet response element or TRE). In the original Tet-Off system, the DNA binding domain of TetR was fused with the potent herpes simplex virus VP16 transactivation domain to form the tetracycline responsive transactivator (tTA) (8). tTA in the absence of doxycycline (dox) binds tetO to initiate transcription. The Tet-On system was later developed due to its wider application in vivo (e.g. for gene therapy and in transgenic animals) (7,9). Random mutagenesis of TetR generated a new transactivator (rtTA), which binds and transactivates gene expression in the presence of dox. Improved versions of rtTA have been developed to give tighter gene expression, increased sensitivity towards the inducer, enhanced stability and expression in mammalian cells, and more uniform transgene expression in the induced cells (10,11).
We incorporated the Cre/LoxP and Tet-On systems into one integrated system to enable tightly regulated induction of gene expression at reproducible levels between experiments and in different clones of mammalian cells. A new LoxP site (L3) was developed to minimize unwanted intrachromosomal recombination between heterospecific LoxP sites. When tested in two different cell lines and at six independent integration sites, incoming DNA was correctly targeted at high efficiencies. Expression of the reporter gene, luciferase-green fluorescence protein fusion (LucGFP) was uniformly induced across most of the RMCE clones derived from the same integration site. Such a highly efficient gene targeting approach in combination with predictable and reproducible gene expression should find wide application in vitro and in vivo.
L3 was made from two oligonucleotides that contained 16 complementary (bold) 3-prime residues (5′-GGA CTC GAG ATA ACT TCG TAT AAA GTC TCC TAT A and 5′-CCT ATC GAT ATA ACT TCG TAT AGG AGA CTT TAT A). The oligos were made duplex by 10 cycles of PCR at an annealing temperature of 42°C. The result was cloned into pCR2.1 using TOPO cloning (Invitrogen) and confirmed by sequencing. The specificity of L3 derives from an internal non-repetitive 8 bp sequence (underlined) that deviates from wild-type at three positions (ATGTATGC).
Naming of the wild-type and LoxP511 sites are according to previously published data (12). pL1L2 and L1HyTK2L were gifts of S. Fiering (1). pL32L was made by substituting L1, bounded by XhoI and NcoI in plasmid L1HyTK2L with L3 from pCR2.1-L3, bounded by XhoI (oligonucleotide restriction site in italics above) and BspLU11 I within pCR2.1. pL3L3 was made from pL32L. L3 was removed with XhoI and PvuII and re-inserted in the position of 2L using SalI and SbfI blunted with T4 DNA polymerase.
L3HyTK2L was constructed by replacing L1 in L1HyTK2L with L3 from pL32L by AhdI and ClaI digestion. The L3HyTK2L cassette was cloned into a retrovirus backbone by inserting L3HyTK2L restricted with NotI and XbaI into pCFB-EGSH (Stratagene) digested with the same enzymes, generating RV-L3HyTK2L.
To facilitate insertion of genes into the inducible L3-2L exchange vector, we constructed L3-TRE-MCSpolyA-2L by cloning the fragment containing seven tetO sites, multiple cloning sites and a polyadenylation signal derived from XhoI and SapI/Klenow treated pTRE2 (BD Bioscience) into L3HyTK2L previously digested with XhoI and PshAI. The exchange plasmid, L3-TRE-LucGFP-2L (pLi028), was derived by cloning a BglII-NotI fragment containing LucGFP from pLuciferase-EGFP (gift from D. Buscher) into BamHI-NotI sites of L3-TRE-MCSpolyA-2L.
A bicistronic transregulator-expressing cassette was obtained by amplifying the TetR(B/E)-KRAB (tTR or Tet-transrepressor) gene by standard PCR using the primers 5′-(B/E)-BamHI and 3′-(B/E)-BglII, followed by restriction with BamHI and BglII and ligation with BamHI-restricted and dephosphorylated pWHE120(sM2), yielding pWHE124. The polio-virus IRES element was amplified with 5′-P-IRES-SmaI and 3′-P-IRES-SmaI from pCMV-KRAB-rtTA (13), restricted with SmaI and inserted into likewise-digested and dephosphorylated pWHE124, yielding pWHE125-P. The plasmid pWHE134 containing the tricistronic transregulator-cassette with rtTA2S-M2, tTR and a neomycin selection marker separated by two IRES elements was constructed by restricting pWHE125-P with EcoRI and HpaI and ligating the fragment encoding the regulatory cassette with pIRESneo (BD Bioscience) containing the selection marker. pIRESneo had previously been restricted with BamHI, the 5′ overhangs filled-in with T4 DNA polymerase, and then restricted with EcoRI. All primer and plasmid sequences are available upon request.
The 293 HEK cells were cotransfected by electroporation with 2 µg of LoxP test plasmid and either 18 µg of GFP expressing plasmid or 18 µg of Cre-expressing plasmid, pOG231 (14). Extra-chromosomal DNA was harvested by Hirt extraction (15) 48 h post-transfection and examined by Southern hybridization after digestion with SspI or NdeI. The probe used was a 426 bp NcoI-SacII fragment corresponding to the hygromycin resistance gene common to all the LoxP plasmids.
Both HeLa and chinese hamster ovary (CHO) cells were cultured in DMEM (Cellgro) supplemented with 5% fetal bovine serum (Atlanta biologicals) and 1× non-essential amino acids (Invitrogen) in a humidified atmosphere containing 7% CO2 at 37°C.
The HeLa cell line M2K was generated by stably transfecting the cell line HRM2 (16) with pCMV-TetR(B/E)-KRAB (13) and pPUR (BD Bioscience). Details will be published elsewhere (Schätz, Knott, Hillen and Berens; manuscript in preparation). To generate the cell line M2PK, ~5 × 105 cells were transfected with 2 µg PvuII-linearized pWHE134 using Lipofectamine (Invitrogen). Twenty-four hours later, cells were seeded into 15 cm plates and selected with 800 µg/ml G418 (Invitrogen) for at least two weeks. Positive clones stably producing the transfected regulators were identified by transient transfection of 0.1 µg pUHC13-3 (8) and 0.6 µg pUHD16-1 for normalization of transfection efficiency, and 1.3 µg pWH802 as non-specific DNA. Clone with the highest induction of luciferase activity after incubation with 1 µg/ml dox for 24 h was chosen. Cells were routinely maintained in medium containing 400 µg/ml G418.
For retroviral delivery of L3HyTK2L into CHO, HeLa M2K and M2PK cells, ~5 × 105 cells were infected with culture supernatant containing RV-L3HyTK2L (as described by manufacturer). Resistant clones were selected in 400 µg/ml hygromycin B (Calbiochem) for CHO and 100 µg/ml for HeLa cells.
L3HyTK2L was also integrated into the dihydrofolate reductase gene (DHFR) locus of DR-8 (a derivative of CHOK1) by homologous recombination using the strategy previously described, generating the stable clone 146-111 (17). For dox inducibility, pWHE134 was stably integrated into 146-111 by an approach described above, generating stable line 111-134.
A total of 30 µg of exchange plasmid and 3 µg of pOG231 were electroporated into ~4 × 106 cells at 250 V 1500 µF (Hybaid). Approximately 1 × 105 cells were re-plated onto 15 cm tissue culture plate 4 days after electroporation. Ganciclovir (Moravek Biochemicals) was added the next day to a final concentration of 2 µM. The drug-containing medium was washed off and replaced with fresh medium after 3 days. Colonies were picked between 7 to 10 days after transferring into fresh medium.
Clones isolated from ganciclovir selection were screened for recombinants by PCR for loss of HyTK gene and gain of specific recombination junctions. A total of 5 ng of genomic DNA was added to a PCR mix containing 0.2 µM primers, 0.2 mM dNTP mix, 1.5 mM MgCl2 and 0.05 U/µl AmpliTaq Gold polymerase (Applied Biosystems). Amplification was performed once at 94°C for 10 min, followed by 40 cycles at 94°C for 30 s, 57°C for 30 s and 72°C for 30 s, followed by a final extension step at 72°C for 10 min. Refer to Figure 1 in results section for the position of primer pairs used. The same strategy was employed for HeLa and CHO cells. Primer sequences are available upon request.
Cells grown on 35 mm dishes were washed once with phosphate-buffered saline (PBS) and lysed in 200 µl lysis buffer [50 mM Tris–HCl (pH 8.0), 5 mM EDTA, 150 mM NaCl, 0.5% NP-40, 1× complete protease inhibitor mix (Roche Diagnostics)] and incubated for 30 min at 4°C. A total of 20 µl of cell lysate was then mixed with 100 µl of luciferase reagent [20 mM Tricine, 1.07 mM MgCO3, 2.67 mM MgSO4, 0.1 mM EDTA, 33.3 mM DTT, 270 µM coenzyme A (USBiological), 530 µM ATP and 470 µM luciferin (Molecular Probes)] before reading on the luminometer (Lumat LB9507). Luciferase activity was normalized to total protein content (BioRad Laboratories). GFP fluorescence was analyzed by passing single cell suspensions in PBS through the FACScan (Becton Dickenson).
The population doubling is calculated based on the formula 3.34 (logNi-logNf), where Ni is the initial cell count and Nf is the final cell count.
The success of directional RMCE relies on the use of two heterospecific LoxP sites ideally exhibiting no recombination with each other. The natural LoxP (L2) site and the mutant LoxP511 (L1) are commonly used (1,12), but they differ by only 1 bp in the central 8 bp spacer region. A plasmid bearing L1 and L2 sites (pL1L2) can recombine illegitimately when cotransfected with the Cre-recombinase plasmid (compare Lane 1 and Lane 2 in Figure 2c). As such recombination reduces the effectiveness of RMCE, we designed a new LoxP site (designated LoxP257 or L3; see Figure 2a) that differs from the original LoxP site by 3 nt. A plasmid carrying a pair of L3 sites (pL3L3) recombines efficiently with itself in the presence of Cre-recombinase (compare lanes 3 and 5 in Figure 2c). More importantly, no recombination was detected in a plasmid containing L3 and the inverse L2 (2L) site (pL32L) in the presence of Cre (compare lanes 4 and 6). Thus, the newly designed L3 site paired with L2 is more ideal for RMCE than is the pair L1 and L2.
We next perform a vigorous test for the efficiency of cassette exchange mediated by L3 and L2. L3 and L2 are integrated into the genome of CHO and HeLa cells by homologous recombination or retroviral delivery. CHO was chosen because it is a commonly used cell line for the expression of recombinant proteins of pharmaceutical value (18–20) and because of our ongoing studies of DNA replication control in the DHFR locus. HeLa was chosen because it is commonly used for analyses of a variety of biological processes and is readily transfected. We made use of a chimeric hygromycin phosphotransferase-thymidine kinase (HyTK) gene for both positive and negative selections (1). The HyTK fusion gene, which confers resistance to hygromycin and sensitivity to ganciclovir, was flanked by L3 and L2 (L3HyTK2L). This cassette was first inserted into the DHFR locus in CHO cells by homologous recombination. The DHFR locus was chosen due to our interest in its replication domain and because it can be amplified to high levels in the presence of methotrexate (21), which may be used to enhance production of proteins of biological interest (21,22). Also, robust procedures are available for homologous recombination between a cosmid carrying the 3′ half of the DHFR coding sequence with DR-8 (a DHFR-deficient CHO cell line hemizygous for a 3′ truncated DHFR gene) (23,24) (Figure 3a). This reconstructs the full-length DHFR gene, which is readily identified by growth of cells in medium lacking thymidine, hypoxanthine and glycine (17). We modified the targeting vector to include the L3HyTK2L cassette downstream of the DHFR gene. Southern blot analysis using probes spanning different regions of the DHFR locus confirmed the reconstruction of the DHFR gene and a probe for HyTK confirmed the integration of the L3HyTK2L cassette (Figure 3b), this clone is now named 146-111.
We next used retroviruses to insert single copies of the L3HyTK2L cassette into HeLa (Figure 3c) or CHO cells genome (data not shown). Southern blot analysis using a probe against HyTK demonstrated single copy integration at random sites in the genome of different clones (Figure 3d). Four HeLa clones and one CHO (CHO #4) clone exhibiting high and stable expression of HyTK [assayed by western blot for TK expression, (data not shown); polyclonal antibody against TK kindly provided by Ian R Wickersham] were used for subsequent recombination experiments.
To test the efficiency of recombination at the genomic site, plasmid DNA containing L3 and L2 sites was electroporated with a Cre-expressing plasmid (14) into the previously constructed CHO and HeLa parental lines containing L3HyTK2L cassette (L3-2L parent). LoxP sites, L3 and L2, were placed as inverted repeats to eliminate the possibility of recombination and deletion of the selection gene. Recombinants were selected based on resistance to ganciclovir after the replacement of the TK gene (Figure 1). A clone that is resistant to ganciclovir can either lose the TK gene by RMCE or has retained TK but silenced its expression or mutationally inactivated it. We identified recombinants that underwent RMCE using PCR to identify clones that lost the TK gene and gained the predicted recombinant junctions (Figure 1 and Table 1). Single copy integration of the incoming DNA was also confirmed by Southern blot analysis (Supplementary Data).
The overall frequency of correct recombination was 81% for CHO and 69% for HeLa cells (Table 1). The efficiency of recombination does not appear to be affected by the size of the incoming DNA, as fragments between the lengths of 100 to 4000 bp recombined with similar frequencies (Table 1). All clones that were negative for recombination junctions were positive for HyTK by PCR, suggesting that they arise due to silencing of the TK gene. No clone was tested positive for inversion between LoxP sites or illegitimate deletion of the TK sequence (Table 1).
It has been previously reported that gene expression is uniform between RMCE clones derived from the same parental line (1,25). Since it is often essential to regulate the level of gene expression, we further tested the uniformity and stability in gene expression by comparing gene induction in RMCE clones containing an integrated copy of L3-TRE-LucGFP-2L. The Tet-On system we used has been modified as follows to enable stringent on-off regulation. The transactivator (rtTA) is a modified bacterial TetR protein fused to VP16 transcriptional activation domain (rtTA2S-M2). It activates gene transcription by binding to TRE in the presence of dox (10,11); in the absence of dox, a Tet-transrepressor (tTR), TetR(B/E)-KRAB, binds TRE and actively represses gene expression (13).
We tested two ways of generating rtTA and tTR expressing stable cell lines. In the first system, rtTA and tTR were linked by an IRES element in a single plasmid (pWHE134) and were transcriptionally controlled by one promoter (single plasmid Tet-On system exemplified by the M2PK HeLa and CHO 111-134 in Figure 4a). In the second system, rtTA and tTR genes were on separate plasmids and their expression were driven from two independent promoters (two plasmid Tet-On system in HeLa cell line M2K in Figure 5a).
Gene expression was induced at a saturating dose of dox (2 µg/ml) for 72 h followed by a luciferase assay to determine the maximum induction attained by each RMCE clone. To compare uniformity in gene expression between clones, cells were induced at a non-saturating dose of dox followed by single cell analysis for GFP fluorescence using flow cytometry. We first tested the dose response to dox by titrating the concentration of dox and measuring the mean fluorescence units at 3 days (Figure 4b and c) after induction. Flow cytometric analysis showed uniform GFP expression in the induced populations and the fluorescence increased in a dose-dependent manner in CHO 111-134 cells from 12.5 to 200 ng/ml. A similar titration experiment was performed for HeLa cells (data not shown). Dox used between 50 and 100 ng/ml for CHO 111-134 and HeLa cells, respectively, fell within the linear range of induction and was chosen to test the uniformity in gene expression in the RMCE clones.
The single plasmid Tet-On system in HeLa cells (M2PK in Figure 4) generates low basal luciferase activity in the un-induced state and high level of luciferase activity at high dox concentration in all the four RMCE clones tested (Figure 4d). The fold increase in luciferase activity ranged from 3 × 104- to 7 × 104-fold over the un-induced state in these clones. Three of four RMCE clones exhibited greater than 90% GFP expression while one exhibited greater than 70% expression after induction with sub-maximal dose of dox (Figure 4e). Three clones had mean GFP fluorescence ranging between 90–110 and one clone had a mean GFP fluorescence of about 70. These data show that within single clones, individual cells exhibited similar expression levels, though we did detect some inter-clonal variability.
In CHO 111-134, all five clones derived by RMCE at the DHFR locus showed high luciferase activities upon induction at 2 µg/ml dox whereas the expression of luciferase was close to background in the absence of dox (Figure 4d). The fold increase in luciferase activity range from 6 × 103 to 4 × 104. As shown in Figure 4e, five out of five RMCE clones were greater than 92% GFP positive after induction at sub-maximal dose. Clone 1 had the highest mean GFP fluorescence (236) while clone 3 showed the lowest mean GFP fluorescence (94). Other RMCE clones had mean GFP fluorescence 105, 107 and 138, respectively. These results indicated that the reporter gene expression in these RMCE clones derived from the L3HyTK2L CHO parental cell line had substantial inducibility and uniformity.
In the two plasmid Tet-On system (M2K HeLa cells in Figure 5), basal luciferase activity was relatively low and uniform across the four clones analyzed. The maximum fold induction in luciferase activity ranged from 1 × 104 to 3 × 104 in the clones analyzed (Figure 5b). Three of the four RMCE clones tested show greater than 80% green cells after induction and exhibited mean GFP fluorescence between 100 and 110 (Figure 5c). One of the clones showed greater than 50% green cells and mean GFP fluorescence of 86 upon induction.
The RMCE approach enables genes to be independently introduced into a single genomic locus. Comparison between genes obtained from independent RMCE experiments is possible if the level of gene expression is preserved in clones obtained from independent RMCE experiments. Using the same L3-2L parent (HeLa M2K), we compared GFP expression in clones obtained from two independent RMCE experiments performed at the same genomic locus (Figure 5d). Seven out of ten clones obtained in the second RMCE experiment had similar mean GFP fluorescence level as 3 out of 4 of the clones obtained in the first RMCE experiment.
We evaluated whether the inducible phenotype was stable over time using one representative RMCE clone from M2PK and one from M2K HeLa. These clones were continuously passaged for 1 month before analysis. The basal GFP level remained low, and dox addition produced robust GFP induction when assayed 1 month later. Uniformity of expression in M2PK #1 was very high over this time period (i.e. 90% of cells exhibited GFP fluorescence after 1 month, compared to 92% originally). (Figure 6). However, the mean GFP fluorescence and the percentage of GFP expressing cells after induction declined slightly in HeLa M2K #3 (Figure 6).
The highly inducible nature of this system suggested potential applications for transiently marking individual cells to ‘pulse label’ them. As one example, we previously generated a very stable histone H2BGFP fusion protein that is readily incorporated into chromatin to generate brilliantly fluorescent chromosomes and nuclei (26). We reasoned that if the H2BGFP gene was induced for several cell doublings to uniformly label the chromatin in each cell, and it was then turned off, the fluorescence of the cell would be diminished by a factor of two each time the cell divided since nucleosomes segregated randomly when DNA is replicated (27). Thus, fluorescence microscopy or flow cytometry could be used to determine how many times a cell doubled by merely measuring its fluorescence intensity relative to that at the beginning of the experiment.
Figure 7 shows a flow cytometric analysis of the fluorescence intensity of HeLa cells followed for seven population doublings. The decrease in fluorescence per population doubling is precisely that expected for a 2-fold decrease per cell division. Importantly, there was no decrease in fluorescence observed in cells that were held in quiescence by thymidine block (Figure 7a). This shows that the decreased fluorescence required entry into and progression through S-phase, and was not the result of H2BGFP degradation.
We have combined the powers of RMCE and regulated gene expression into a single highly efficient system that should facilitate structure-function studies, enable production of biopharmaceuticals, or provide a means of tracking cell proliferation in real time. The Cre/LoxP recombination system is preferred for site-specific recombination because it exhibits higher efficiency of recombination over the Flp/FRT system in specific cell types such as mouse embryonic stem cells and primary cells (28–31). However, the system we describe could as easily be implemented using engineered Flp recombination target sequences designed not to recombine with each other. We first engineered a new LoxP site to minimize intrachromosomal exchanges that generate deletions, and to maximize exchanges between the chromosomal and plasmid donor LoxP targets. To minimize the need for drug selection markers, we introduced into the chromosomal target a HyTK fusion gene. This enables positive selection to identify clones containing the LoxP target, and then counterselection using gancyclovir to select for loss of the chromosomal target and concomitant replacement with the donor plasmid. This is especially valuable for situations in which introduction of extraneous transcriptional control sequences needs to be avoided. We also used an optimized Tet-On system to enable robust controlled expression of the donated sequences. This should allow for analyses of phenotypic effects of the introduced gene when it is expressed at normal, sub- or supraphysiologic levels.
We tested the system by first analyzing the stringency of recombination with the newly characterized L3 LoxP site. A pair of heterospecific recombination sites that do not recombine with each other is necessary for efficient RMCE (6). When used with the original LoxP site (L2), the first LoxP variant (LoxP511 or L1) that was generated greatly enhanced the frequencies of exogenous gene integration when coupled with positive selection (12). However, using negative selection as a means to enrich for recombinants, the efficiencies were markedly different (1,6). Deletion of intervening sequences without replacement occurred when L2 and L1 were placed as direct repeats, suggesting interaction between L1 and L2 (6). When placed as inverted repeats, inversion of the intervening sequence resulted in an equal proportion of clones in either direction (1). This is not desirable, especially when directionality of the integrant is important. We solved the problem of recombination between L2 and L1 by generating a new LoxP site, LoxP257 (L3), which was designed based on findings from in vitro recombination assays (5). L3 differs from L2 by three bases within the 8 bp spacer sequence. When used in combination with L2 and L3 generated recombinants at high frequency, suggesting the feasibility of using L3 and L2 for RMCE. The finding was consistent with a previous study showing that a modified LoxP sequence (LoxP2272) with two base changes from L2 also enabled high efficiency exchange (6). Other than having a high inter-molecular recombination efficiency, the probability of selecting a recombinant clone among the ganciclovir resistant clones is increased when TK expression from the L3HyTK2L parent line is high and stable. Using this criterion for the selection of the L3HyTK2L cell line, we observed high efficiency, accurate RMCE in two different mammalian cell lines at six independent genomic loci. We also observed little size dependence on RMCE frequency or accuracy. The high frequency of correctly recombined clones expedites screening for desirable clones.
We used homologous recombination to introduce L3HyTK2L into several single copy loci, including the intergenic region of the CHO DHFR locus. Our data demonstrate that the DHFR locus was favorable for RMCE and for tightly regulated gene induction. Recombination occurred at very high efficiency (50–100%) and we readily obtained clones exhibiting robust inducible expression (up to 104-fold by luciferase, and >100-fold by GFP analyses). This locus offers a number of advantages for other studies. For example, it can be readily amplified using selection for methotrexate-resistance (21). As CHO cells are often used to produce biopharmaceuticals, the combination of amplification and highly regulated expression could provide substantial advantages for reducing production costs, enabling stable maintenance of transgenes, and obtaining very high level expression of potentially toxic proteins. In studies to be presented elsewhere, we found the L2 and L3 sites to be useful for generating genetically engineered ES cells, mouse embryo fibroblasts, and mice with different p53 point mutations or fusion genes with high efficiency (F. Toledo, C. Lee and G.M. Wahl, manuscript in preparation).
Using a strategy similar to that employed for the delivery of Flp/FRT tagged sites into the genome of mammalian cells (25,32), we integrated the LoxP sites at single copy into random positions in the genome of HeLa and CHO cells by retroviral delivery (Figure 3c). An advantage of the retroviral system is that it ensures delivery of the entire transgene at high efficiency and at single copy (Figure 3d) into the genome. Active transcription sites are the preferred retroviral integration sites (33,34). Consistent with this, we were able to select retroviral integrations at high efficiency, an advantage over gene targeting by homologous recombination.
An important advantage of RMCE approach that is valuable for structure-function studies and production of recombinant proteins of pharmaceutical value is the ability to repeatedly obtain clones exhibiting predictable and uniform gene expression at high frequencies (1,25,35). We combined that with the improved version of the Tet-On inducible system to evaluate the reproducibility of gene induction at each locus (11). By single cell analysis, GFP was uniformly expressed within the population and it was induced in a dose-responsive manner. The improved rtTA (rtTA2s-M2) demonstrated many superior properties over the original rtTA. Beside the ability to induce graded transgene expression, induction exhibited a greater dynamic range, higher sensitivity to dox and faster induction rate (within 24 h, data not shown) compared to the original rtTA system (Figure 4b and c) (11,36). Recombinants derived from the same L3-2L parental line have the reporter gene inserted at the same genomic locus. As expected, the majority of the RMCE clones obtained in one experiment showed similar levels of expression upon induction (Figures 4 and and5).The5).The uniformity in gene expression was also found between clones obtained from independent RMCE experiments (Figure 5d). Hence, clones that reproducibly induce different transgenes can be readily obtained through multiple rounds of RMCE.
Maintaining stable transgene expression is as important as high transgene expression. Since transgene expression may be subjected to silencing, especially when viral promoter is used to drive gene expression (37,38) and when expression is not maintained under selective pressure (39). We measured the stability of expression in HeLa cells after at least 30 population doublings in culture. We observed that the level of inducibility, and the fraction of cells expressing the transgene, was stable over the time period studied when transcription of rtTA and tTR were coupled to that of the neomycin resistance gene by the IRES elements and by maintaining the culture under G418 selection. Hence, the use of IRES to mediate coexpression of multiple genes will gain popularity in both in vitro and in vivo applications (40,41).
The tight and reversible gene expression regulated by the Tet-On system was further demonstrated by switching H2BGFP expression on and then off to monitor cell doubling (Figure 7) as a function of fluorescence units. H2BGFP is incorporated into the chromatin after synthesis, and tracks with nucleosomes that are segregated randomly to daughter cells during each division (27). Being a stable protein, H2BGFP fluorescence should approximately halve with each division if its expression is prevented. Consistent with this prediction, after H2BGFP expression was turned off for different numbers of population doublings, flow cytometric analysis revealed the expected 2-fold decay per population doubling (Figure 7). Hence, the inducible expression of H2BGFP can be used to measure population doublings within a defined time frame. For example, as stem cells are expected to divide fewer times than transit amplifying cells, the retention of H2BGFP fluorescence should be able to be used as a marker to localize putative stem cells in situ and as a means to isolate them by fluorescence based cell sorting (42–44).
In conclusion, the LoxP/Cre recombination system offers reproducible and directional DNA integration into specific loci in the genome at high frequency. The single plasmid Tet-On system, which ensures stable expression of the transactivator and transrepressor, allows reproducible and graded induction of transgenes to a similar level in the RMCE clones derived from single and independent recombination experiments.
Supplementary Data are available at NAR Online.
We would like to thank Rose Rodewald and David J. Chambers for technical assistance, Joyce Hamlin, Frank Toledo, Kurt Krummel, Mengjia Tang and Mark Wade for helpful discussions, and Dirk Buscher, Steven Fiering and Ian R Wickersham for generously providing materials. This work was supported by grants from the NIH (CA061449 and 048405 to G.M.W.) and the Wilhelm Sander Stiftung (2002.124-1 to W.H.). W.H. and C.B. thank the Fonds der chemischen Industrie. This material is based upon work supported under an A*STAR oversea graduate scholarship awarded to E.T.W. Funding to pay the Open Access publication charges for this article was provided by NIH grant CA061449.
Conflict of interest statement. None declared.