ABSTRACT
The Cre DNA recombinase of bacteriophage P1 has become a useful tool for precise genomic manipulation in embryonic stem (ES) cells that have been gene modified by homologous recombination. We have re-engineered the cre gene to allow ready identification of living Cre+ cells by constructing a functional fusion between Cre and an enhanced green fluorescent protein from Aequorea victoria (GFPS65T). The GFPcre fusion gene product rapidly targeted the nucleus in the absence of any exogenous nuclear localization signal. Moreover, GFPCre catalyzed efficient DNA recombination in both a mouse 3T3 derivative cell line and in murine ES cells. Fluorescence- activated cell sorting (FACS) of transiently GFPcre-transfected ES cells not only allowed rapid and efficient isolation of Cre+ cells after DNA transfection but also demonstrated that a burst of Cre expression is sufficient to commit cells to Cre-mediated `pop-out' of loxP-tagged DNA from the genome. Thus, GFPcre allows rapid identification of living cells in which loxP- flanked DNA sequences are destined to be removed from the genome by Cre-mediated recombination without reliance on recombinational activation or inactivation of a marker gene at the target locus. In addition, the GFPcre fusion gene will prove useful in tracing tissue-specific Cre expression in transgenic animals, thereby facilitating the generation and analysis of conditional gene knockout mice.
The generation of transgenic and gene modified mice by pronuclear injection of zygotes and homologous gene targeting in embryonic stem (ES) cells has rapidly advanced understanding of gene function in mammals and increasingly will permit the development of highly useful animal models of human disease and pathology. An important tool for genome modification in vivo is the Cre site-specific DNA recombinase of bacteriophage P1 (1 -3 ). Cre catalyzes precise DNA recombination both intra- and intermolecularly between target 34 bp loxP sites and is proficient for recombination in a variety of eukaryotes. Efficient Cre-mediated excision of DNA between directly repeated loxP sites in developing transgenic animals has been harnessed both to activate expression of a dormant transgene (4 ) and to ablate a resident chromosomal gene (5 ). In both cases the timing and tissue localization of the recombination event can be precisely delimited by choosing a promoter to drive cre expression that has the spatial and temporal expression pattern desired. One other important use for Cre recombinase is the removal of unwanted DNA from the genome. For example, in some cases the selectable neo marker interferes with neighboring gene expression (6 ,7 ). Removal of the selectable marker eliminates potential complications in interpretation of an animal's phenotype after gene targeting and is easily achieved by using a selectable marker gene flanked by loxP sites so that it can be excised by transient expression of Cre recombinase either in ES cells (8 -10 ) or in fertilized zygotes (11 ) or by simply mating the gene targeted animal with a transgenic mouse having either general (4 ,12 ) or zygote-specific expression of the cre gene (13 ).
Because of the binary nature of these site-specific recombination strategies, knowledge of the expression pattern of the cre transgene is critical for evaluation of the doubly transgenic animals that result from crossing a cre mouse with one having chromosomally positioned loxP sites. Cre protein can be detected in mammalian cells by in situ immunohistochemistry using polyclonal anti-Cre antibodies (S.Gagneten and B.Sauer, unpublished work) and with specific anti-Cre monoclonal antibodies (14 ). However, direct detection of Cre in living cells cannot be achieved by immunological methods. Yet in a variety of circumstances, for example excision or `pop-out' of the selectable marker used for gene targeting in ES cells, it would be useful to identify and isolate live Cre-expressing cells for subsequent propagation or manipulation and analysis.
To provide a convenient and efficient way to identify Cre+ cells, we have fused Cre to the green fluorescent protein (GFP) of the jellyfish Aequorea victoria (15 ). GFP has rapidly become an important new reporter of gene expression in a variety of organisms (16 ,17 ). The 238 amino acid protein requires no host cofactors and emits a green fluorescence ([lambda]max 508 nm) in living cells transfected with GFP cDNA after stimulation with UV light ([lambda]max 395 nm and a much weaker excitation [lambda]max 470 nm). Alteration of GFP Ser65 to Thr (18 ) results in a protein having substantially enhanced, red-shifted fluorescence ([lambda]max 511 nm) and is maximally excited with blue light ([lambda]max 490 nm). Because many flow cytometers employ an argon laser tuned to 488 nm for excitation, GFPS65T is more suitable for use in fluorescence-activated cell sorting (FACS) compared with wild-type GFP (19 ,20 ) and we have therefore used the GFPS65T mutant for construction of the GFPcre fusion.
We show here that the GFPcre fusion gene expresses Cre activity and allows the facile FACS separation of Cre+ cells after transient transfection of both a cultured fibroblast cell line and of loxP-modified ES cells.
All DNA manipulations were performed by standard procedures (21 ). Plasmid pBS377 carries the wild-type GFP gene under the control of the strong EF1[alpha] promoter. It was constructed by cloning the Asp718-BsmI GFP fragment of TU65 (16 ) into the cognate sites (underlined) of a polylinker 5'-AAT TGG ATC CAG ATA TCT
A modified cre gene [carrying an optimized translational start (23 ), a neutral S2A mutation in the coding region and convenient flanking restriction sites] was amplified by Pfu DNA polymerase (Stratagene) from bacteriophage P1 DNA using the primers 5'-TTT TCA AGC TTG GAT GGT ACC ATG GCC AAT TTA CTG ACC G-3' and 5'-TTC AGC TCT AGA GCA ATC ATT TAC GCG TTA ATG G-3' and then cloned as a HindIII-XbaI fragment into pUC19 to obtain pBS353. The NcoI-SmaI cre fragment of pBS353 replaced the neo gene of RSVneo (24 ) using HindIII-NcoI bridging oligos (5'-AGC TTG
Calf serum (CS), DMEM culture medium and leukemia inhibitory factor (ESGRO) were from Gibco BRL (Gaithersburg, MD). Fetal calf serum (FCS) was from Hyclone (Logan, UT). Cell line B-13, a derivative of NIH 3T3 cells (ATCC, Rockville, MD), carries a silent, but recombination activatable, loxP-modified lacZ cassette (B.Bethke and B.Sauer, manuscript in preparation) and was cultured in DMEM plus 10% CS. CHO-K1 cells were grown in DMEM + 10% FCS. The ES cell line H200 is a gene targeted loxP-flanked insertion at the HPRT locus (S.Gagneten, Y.Le and B.Sauer, unpublished work). Recombination between the two directly repeated loxP sites in H200 results in deactivation of the neo selectable marker. H200 cells were cultured in DMEM plus 16% FCS and 813 U/ml ESGRO and maintained on irradiated mouse embryonic fibroblasts (25 ). Cells were grown at 37oC in a humidified chamber with 5% CO2.
DNA (18 [mu]g/6 cm plate) was introduced into CHO and B-13 cells by co-precipitation with CaPO4 (26 ). The procedure was slightly modified for ES cells: the DNA (9 [mu]g)/CaPO4 co-precipitate was mixed with 1.5 * 106 trypsinized cells in 5 ml ES medium and then plated on a 6 cm gelatinized plate. The DNA co-precipitate was removed after overnight incubation and replaced with fresh medium. A CMVlacZ plasmid, p324 (27 ), was used in non-GFP control transfections. Two days post-transfection cells were trypsinized and resuspended in culture medium for flow cytometric analysis and sorting with an Elite cytometer (Coulter, Miami, FL). Argon laser (488 nm) excitation and 525 nm bandpass emission filters were used to detect GFP expression. Analysis and sorting gates were chosen to identify and isolate cells which fluoresced at levels >3 SD above the mock-transfected controls. Sort logic and coincidence decisions were chosen for highest purity. Sorted and unsorted cells were replated to allow individual colony formation. B-13-derived colonies were analyzed by an in situ [beta]-galactosidase assay (28 ) using Bluo-Gal (Gibco BRL).
Fluorescence was examined in both live and fixed (3% parafomaldehyde, 30 min) cells. Confocal microscopy was graciously performed by Dr A.Robbins (NIDDK) with a Zeiss Axiovert 100 microscope and a Zeiss LSM 410 imaging system, using a fluorescein filter set. A Nikon Optiphot-2 epifluorescence microscope with the Nikon B-2A filter block (excitation 450-490 nm, BA 520-560 nm) was used for routine monitoring of fluorescence in live cells.
Individual ES cell colonies were expanded into 24-well dishes and DNA was prepared after cells had become confluent (29 ). Retention of loxP-flanked DNA after H200 transfection was detected with the PCR primers a (5'-ATA GCC GAA TAG CCT CTC CAG C-3') and c (5'-TAA CAG CGT CAA CAG CGT GCC-3'). Primers b (5'-GTA GCC AAC GCT ATG TCC-3') and d (5'-ACA GTA GCT CTT CAG TCT G-3') were used to detect excision events. DNA was amplified with Taq DNA polymerase: 30 s, 61oC; 60 s, 72oC; 30 s, 94oC, for either 30 cycles (primers a + c) or 35 cycles (primers b + d).
GFP carrying the S65T mutation was fused to the N-terminus of Cre. The GFPS65T mutant was chosen as a fusion partner for Cre both to give enhanced fluoresence and to permit excitation with blue light rather than UV in order to eliminate UV-induced damage to mammalian cells or mouse zygotes after visualizing expression of the GFPcre fusion gene (30 ). Initial work in Escherichia coli and in Saccharomyces cerevisiae showed that expression of GFPcre resulted in cellular fluorescence and Cre-mediated recombination (data not shown). The mammalian expression vectors pBS448 and pBS500 (Fig. 1 ) place the GFPcre fusion gene under the control of the strong RSV and EF1[alpha] promoters respectively. We first examined expression of the fusion gene in cultured mammalian cells by transient expression. Although of prokaryotic origin, Cre protein efficiently targets the nucleus of mammalian cells due to specific determinants in the wild-type Cre protein that direct nuclear localization/retention (S.Gagneten and B.Sauer, manuscript in preparation). We thus suspected that in cells transfected with the GFPcre fusion gene, only the nucleus would exhibit strong fluorescence. This is exactly the result observed. Transfection of CHO cells with pBS448 resulted in strong green fluorescence almost exclusively in the nucleus of productively transfected cells, whereas no subcellular localized fluorescence occurred after transfection with the wild-type GFP construct pBS377 (Fig. 2 ). These results indicated that fluorescence might be a convenient tag to identify Cre expression in living cells after transient expression.
Transient expression of Cre recombinase efficiently evicts loxP-flanked DNA from the mammalian genome (31 ,32 ). The frequency of Cre-catalyzed excision is roughly equivalent to the DNA transfection efficiency, indicating that a brief burst of Cre expression commits a transfected cell to Cre-mediated excision of DNA. Because transient expression of the GFPcre fusion gene results in green fluorescence of cells that take up and express the GFPcre construct, we asked whether GFPcre-transfected cells could be sorted, based on their fluorescence, in order to enrich for cells that are destined to undergo Cre-mediated recombination.
To facilitate this analysis we exploited cell line B-13, an engineered derivative of NIH 3T3 cells containing a single copy of a recombination activatable lacZ gene (Fig. 3 ). B-13 cells are LacZ- (white) due to a loxP-flanked STOP cassette (4 ) inserted into the open reading frame at the N-terminus of the lacZ gene. Upon Cre-mediated excision the STOP cassette is removed and the lacZ reading frame is restored to allow production of [beta]-galactosidase, thus producing blue colonies.
Although previous work had established that transient expression of Cre recombinase was sufficient for productive recombination in ES cells, the absolute frequency of obtaining such marker pop-outs was only a few percent, thus necessitating the use of a negative selectable marker, tk, interposed between the loxP sites in order to obtain an efficient yield of pop-outs (8 ). It is likely that this low absolute efficiency in detection of recombination derives not from an inherent inability of the recombinase to catalyze recombination, but rather from the low transfection efficiency of ES cells that results in only a small percentage of the transfected cell population actually expressing the enzyme.
Table 1
To determine whether use of the GFPcre gene would facilitate retrieval of Cre-mediated DNA pop-outs, we examined GFPcre-catalyzed recombination in ES cell line H200 that carries a gene targeted loxP-flanked marker cassette at the HPRT locus (Fig. 5 A). Transfection of H200 with the EF1[alpha]-GFPcre construct pBS500 resulted in ~1% of the cells exhibiting green fluorescence after 2 days, as determined by fluorescence microscopy. Similarly, FACS analysis indicated fluorescence in ~1.4% of the transfected cells (Fig. 4 B) compared with the control (Fig. 4 A). After sorting, FACS analysis revealed a dramatic enrichment to ~95% fluorescent cells (Fig. 4 C).
We have constructed a functional GFPcre fusion gene and shown that transient expression after DNA transfection of mammalian cells results in cells that concomitantly exhibit bright nuclear fluorescence and are phenotypically Cre+. Moreover, enrichment of GFPcre fluorescent cells by FACS resulted in a population of cells in which the vast majority are committed to Cre-mediated excisive recombination. We have applied this procedure to ES cells that carry a chromosomal loxP-flanked DNA cassette and shown that Cre+ cells destined to pop-out the DNA between the two directly repeated loxP sites could be easily and efficiently obtained by transient expression of the GFPcre construct and subsequent FACS. This is particularly valuable for situations in which only a low percentage of the cells productively take up DNA, as is often the case for ES cells.
The task of ridding the ES cell genome of a no longer needed selectable marker after homologous targeting is greatly facilitated by the use of GFPcre. Rapid identification and enrichment of phenotypically Cre+ cells by FACS obviates the need for a negative selectable marker like tk to be present on the DNA segment targeted for removal by Cre. This is advantageous not only because it avoids a negative drug selection step, but also because it frees the tk marker to be used for positive/negative selection, if desired, in the initial gene targeting step in order to enhance recovery of homologous recombinants (33 ). As an alternative to the use of FACS, purification of GFPCre+ ES cells in good yield may also be achieved using direct visualization by fluorescence microscopy and micromanipulation (T.Larson, Y.Le and B.Sauer, unpublished work).
Use of the GFP fusion approach is advantageous for the detection and isolation of rare transfection events by flow cytometry. High specificity can be attained by elimination of background fluorescence from non-specific binding with antibodies or the fluorescent substrates used to detect reporter gene expression. Moreover, it should be possible to precisely select cell populations based on the level of GFPcre transgene expression by simple differential gating of the sort histogram. This has certain practical applications. For example, in certain cases ES cells may be modified to carry a selectable marker flanked by directly repeated loxP sites and also a third loxP site in cis (5 ) designed to delete the targeted gene by Cre-mediated recombination (a conditional knockout). To remove the selectable marker in ES cells without deleting the targeted gene itself thus requires partial Cre-mediated excision. Because the recombinational potential of a transfected cell correlates with the level of Cre expression (34 ), FACS-mediated retrieval of cells expressing only a moderate level of Cre should enhance the recovery of cells that have excised only the selectable marker, leaving the targeted gene intact but flanked by two directly repeated loxP sites.
In addition to simple marker removal, Cre can also be used to engineer more extensive, precisely determined chromosomal deletions by targeting (by homologous recombination) correctly oriented loxP sites to span the exact chromosomal interval to be deleted. Although the efficiency of large deletion generation (>200 kb) in the mammalian genome by Cre may vary between different loci, the use of GFPcre to identify functionally Cre+ cells will clearly allow rapid enrichment of the desired deletions by eliminating non-expressing or weakly expressing cells. Finally, because expression of GFP does not appear to be toxic in transgenic animals (35 ), the GFPcre gene should also prove useful in tracing tissue-specific expression of Cre recombinase in transgenic animals.
We are grateful to Dr A.R.Robbins for confocal microscopy, to M.Klessman for assistance with cell culture and to J.Njoroge for FACS expertise. We thank Dr E.Ginns for thoughtful comments on the manuscript.
*To whom correspondence should be addressed. Tel: +1 301 402 4567; Fax: +1 301 496 0839; Email: sauerb@helix.nih.gov
+Present address: Food and Drug Administration, Bethesda, MD 20892, USA
Cells
Number of colonies
Excisive recombination (%)
White
Blue
Total
Not sorted
390
61
451
13.5
FACS sorted
51
460
511
90
REFERENCES