ABSTRACT
The site-directed recombinase Cre can be employed to delete or express genes in
cell lines or animals. Clearly, the ability to control remotely the activity of
this enzyme would be highly desirable. To this end we have constructed
expression vectors for fusion proteins consisting of the Cre recombinase and a
mutated hormone-binding domain of the murine oestrogen receptor. The latter still binds
the anti-oestrogen drug tamoxifen but no longer 17
[beta]
-oestradiol. We show here that in embryonic stem cells expressing such
fusion proteins, tamoxifen can efficiently induce Cre-mediated recombination, thereby activating a stably integrated LacZ
reporter gene. In the presence of either 10
[mu]
M tamoxifen or 800 nM 4-hydroxy-tamoxifen, recombination of the LacZ gene is complete within 3-4 days. By placing a tamoxifen-binding domain on both ends of the Cre protein, the
enzymatic activity of Cre can be even more tightly controlled. Transgenic mice
expressing such an tamoxifen-inducible Cre enzyme may thus provide a new and useful genetic tool to
mutate or delete genes at specific times during developement or in adult
animals.
Genes can be mutated or deleted from the mouse genome by homologous
recombination in embryonic stem (ES) cells. Many genes, however, are essential
for embryonic development and a homozygous deletion of these genes will
therefore often result in non-viable embryos. To circumvent this problem, strategies have been developed
that aim at deleting or mutating genes in the animal rather than in ES cells (
1
-
3
). Such strategies employ site-specific recombinases like the Cre recombinase of bacteriophage P
1
(
4
-
6
) or the FLP recombinase from yeast (
7
). The 38 kDa Cre recombinase can delete genes which are flanked on each site by
a 34 base pair (bp) Cre recombination signal sequence called
loxP
site. Such genes can be generated by homologous recombination in ES cells (
8
). These genes will be deleted only in cells expressing the Cre recombinase.
Indeed, transgenic mice carrying a gene flanked by
loxP
sites and expressing the Cre recombinase under the control of a tissue-specific promoter, show a high percentage of gene deletion in the tissue
where Cre is expressed (
9
,
10
).
For a better understanding of the role of genes during murine development, it
would be desirable to delete genes by a Cre recombinase system which is not
only tissue-specific but also inducible (
11
). The available tissue-specific promoters, however, are not easy to combine with inducible
promoter elements. Another way to control the activity of a protein is to fuse
it with regulatory domains like the hormone-binding domain (HBD) of steroid receptors. Several transcription factors (
12
,
13
), as well as other proteins, for example the RAF kinase (
14
), or the recombinases FLP (
15
) and Cre (
16
) have been rendered functionally hormone-dependent by such a strategy (reviewed in
17
). In the absence of ligand, the HBD-fusion proteins are bound and inactivated by heat shock proteins like
Hsp90. Ligand binding releases the receptors from the inhibitory complexes
(reviewed in
19
,
19
).
The HBD of the murine oestrogen receptor is well studied and often employed in
HBD-fusion proteins. Such fusion proteins, however, can not easily be used as
control elements in transgenic mice because the ligand of HBD, 17[beta]-oestradiol, is abundant in murine plasma. Recently the murine
oestrogen receptor HBD has been mutated (
20
) so that the glycine at position 525 is replaced by arginine (G525R). This
mutated HBD has lost its binding activity for 17[beta]-oestradiol but retains binding to the anti-oestrogen drugs tamoxifen or 4-hydroxy-tamoxifen (
21
) and we therefore refer to it here as the tamoxifen-binding domain (TBD). Functional analysis of a c-Myc/TBD fusion protein has shown that it is activated by 4-hydroxy-tamoxifen yet totally unresponsive to 17[beta]-oestradiol (
22
). We have constructed expression vectors encoding fusion proteins that consists
of the Cre recombinase and TBD; we show here that the recombinase activity of
these proteins can be induced by tamoxifen.
The plasmid pBSKS+MerG525R (
22
) contains the mutated (G525R) coding sequence of the ligand-binding domain of the murine oestrogen receptor (amino acid 281-599) cloned into the
Bam
HI-
Eco
RI sites of Bluescript (Stratagene). The plasmid pTZ-creN carrying the 1050 bp coding sequence of the Cre recombinase was a
kind gift of Lars Nitschke. At the 3' end of the Cre coding sequence, a glycine-serine encoding linker and a
Bgl
II site were introduced by PCR using a 5' primer (5'-ACTCGCGCCCTGGAAGGGATT-3') lying just 13 bp upstream of the
Cla
I site of pTZ-creN and a 3' primer (5'-ACGG
The expression vector pAN-H carries the human [beta]-actin promotor followed by a
Hin
dIII cloning site and the polyadenylation site of SV40 as well as ampicillin and
neomycin resistance genes. To obtain the Cre-TBD expression vector pANCreMer (Fig.
1
a), the CreMer coding sequence of pTZ-CreMer was cloned as a
Hin
dIII fragment into the
Hin
dIII-linearized pAN-H plasmid. The plasmid pUCHMR carrying the hygromycin resistance
gene has been described previously (
23
). To construct the pANMerCreMer vector, we cloned a linker containing an ATG
start codon and
Bgl
II +
Sal
I sites into pTZ-Cre just in front of the Cre reading frame. The modified pTZ-Cre plasmid was linearized with
Bgl
II +
Sal
I digestion and a PCR fragment of the murine oestrogen receptor (Mer) cDNA
derived from pBSKS+MerG525R was ligated into this plasmid to yield pTZMerCre.
From the latter plasmid, we isolated a
Hin
dIII-
Bam
HI fragment which was coligated with a
Bam
HI-
Hin
dIII fragment from pANCreMer into the
Hin
dIII linearized pAN-H plasmid to obtain the vector pANMerCreMer (Fig.
1
b).
The cell line, MS4pAM (A.-M. Ayral
et al
., in preparation), is a derivative of the 129/Ola embryonic stem cell line
E14.1 and carries a single copy of the Cre activity indicator construct pAMA.
The pAMA vector carries a neomycin resistance gene and a modified [beta]-galactosidase (ATG-
loxP
-stop-pA-
loxP
-LacZ) gene (Fig.
1
c), the reading frame of which is disrupted after the second codon by two
loxP
sites flanking a 304 bp sequence with a translational stop and a
polyadenylation site (Fig.
1
d). Transcription of the ATG-
loxP
-stop-pA-
loxP
-LacZ gene is initiated at a polyoma enhancer-Tk promoter cassette and terminates at the polyA site present
between the two
loxP
sites. Cre-
loxP
-mediated recombination will lead to the deletion of the translational stop
and poly-A site and generate the ATG-
loxP
-LacZ gene expressing [beta]-galactosidase.
ES cells were cultured in high-glucose (4.5 mg/ml) DMEM medium containing 10 mM sodium pyruvate, 20 mM l-glutamine, 120 [mu]M [beta]-mercaptoethanol, 50 U/ml penicillin, 50 U/ml
streptomycin and 1* MEM non-essential amino acids supplemented by 15% fetal calf serum and
supernatant of LIF-producing CHO cells at a 1:300 dilution. The ES cells were grown either on
a dense monolayer of mitomycin C-treated, G418-resistant primary mouse embryo fibroblasts or on gelatin-coated Petri dishes.
The MS4pAM cells were harvested by trypsinisation, washed once with phosphate-buffered saline (PBS) and resuspended in PBS at 2 * 10
7
cells/ml. For stable transfection, 30 [mu]g of
Sfi
1-linearized vectors pANCreMer or pANMerCreMer and 10 [mu]g of
Hin
dIII-linearized pUCHMR were added to 1 ml of the cell suspension and the cells
were transfected by electroporation as described previously (
24
). After 1 day in culture, the transfectants were selected in medium containing
200 U/ml hygromycin. Growing transfectants were split and incubated either in
the absence or presence of 10 [mu]M tamoxifen. The cells were then stained with X-Gal by the method of Buhler
et al
. (
25
). The amount of [beta]-galactosidase in the lysate of the transfectants was determined by
the [beta]-Galactosidase Enzyme Assay System (Promega).
To obtain a tamoxifen-inducible Cre recombinase we constructed the expression vector pANCreMer
in which the reading frame of the Cre gene was fused with that of TBD, the
mutated (G525R) HBD of the murine oestrogen receptor (Fig.
1
a). Compared with the wild-type, the mutated TBD has a 1000-fold lower affinity for 17[beta]-oestradiol, whereas its affinity for tamoxifen or 4-hydroxy-tamoxifen remains unchanged (
20
).
The ES cell line MS4pAM had been previously transfected with the recombination-dependent [beta]-galactosidase expression vector pAMA (Fig.
1
c). In this vector, the reading frame of the [beta]-galactosidase (LacZ) gene is disrupted after the second codon by two
loxP
sites flanking a 304 bp sequence with a translational stop and a
polyadenylation site (Fig.
1
d). As a consequence, this vector cannot express [beta]-galactosidase. Cre-
loxP
-mediated recombination, however, would lead to deletion of the stop codon
and the polyadenylation site, thereby restoring the open reading frame of the
ATG-
loxP
-LacZ fusion gene and allowing [beta]-galactosidase expression. The MS4pAM cell line can thus be
used as an indicator line for the activity of the Cre recombinase.
The MS4pAM cells were co-transfected with pANCreMER and the hygromycin vector pUCHMR. Two weeks
after selection in hygromycin-containing medium, we obtained 52 hygromycin-resistant transfectants of MS4pAM. These cells were cultured for 2
days in the presence or absence of 10 [mu]M tamoxifen and analyzed for Cre recombinase activity by X-Gal staining. Five transfectants scored Cre-positive in this assay and three (ES41, ES33 and ES5) were
chosen for further analysis. Cultures of these stable transfectants with normal
medium contained only a few blue cells (Fig.
2
a, c and e) whereas after a 3 day culture in the presence of tamoxifen, most
cells of ES41 and ES33 (Fig.
2
b and d) and >50% of ES5 (Fig.
2
f) stained with X-Gal. This result showed that the recombinase activity of Cre-TBD in three ES cell transfectants could be regulated by tamoxifen
in a majority of the cells.
We next cultured 2 * 10
3
cells of either ES5 or ES41 for 3 days with various doses of tamoxifen (Fig.
3
a). After culture, the cells were lysed and the amount of [beta]-galactosidase activity in the lysate was determined by an enzymatic
assay. An optimum of [beta]-galactosidase activity was reached in both Cre-TBD transfectants at 8 [mu]M tamoxifen. Higher concentrations of tamoxifen were toxic
to the cells and resulted therefore in declining [beta]-galactosidase activity.
To control the Cre enzyme more tightly, we constructed the vector pANMerCreMer
(Fig.
1
b) expressing a fusion protein in which a TBD is appended to both ends of the
Cre enzyme. This vector was stably transfected into MS4pAM cells and 12 of 92
obtained hygromycin-resistant transfectants scored Cre-positive. Two of these transfectants (ES22 and ES26) were analyzed
for the induction of Cre-activity. Surprisingly, the TBD-Cre-TBD protein was still enzymatically active. When exposed for
various times to 800 nM 4-hydroxy-tamoxifen, both TBD-Cre-TBD transfectants showed a kinetic of Cre induction
similar to that of the Cre-TBD transfectants (Fig.
5
). The Cre activity, however, did not increase when these cells were cultured in
the presence of 80 nM 17[beta]-oestradiol. Furthermore, when ES22 and ES26 cells were kept in
culture for >8 weeks, the background level of Cre activity did not increase
(data not shown). This demonstrates that the TBD-Cre-TBD protein is under a more stringent control and should therefore
be more suitable when expressed in transgenic mice than the Cre-TBD protein.
The functional analysis of Cre-TBD-expressing ES cells demonstrates that the Cre recombinase can be
fused at its C-terminus to another protein without losing its enzymatic activity. By
fusion to the TBD derived from the oestrogen receptor, a Cre-mediated recombination can be rendered ligand-dependent. In two independent CreTBD transfectants, ES33 and ES41,
Cre-mediated recombination of the LacZ-reporter gene could be induced by the anti-oestrogen drugs tamoxifen and 4-hydroxy-tamoxifen in the micromolar and nanomolar range,
respectively. The different sensitivities of the two drugs is in agreement with
the finding that the 4-hydroxy-tamoxifen binds to HBD of the oestrogen receptor with a 10-100-fold higher affinity than tamoxifen (
21
,
27
). In the presence of these HBD-ligands, the Cre-mediated LacZ gene recombination and activation occurred in most
cells of the ES33 and ES41 culture showing that the ligand-bound Cre-TBD fusion protein is an efficient recombinase.
The [beta]-galactosidase activity in the cell culture increased from the 15 h
time point onward and followed a time course which was very similar to that
observed in cells expressing a fusion protein (FLP-LBD) in which the FLP recombinase of yeast was fused to the ligand-binding domain of the human oestrogen receptor (
15
). Furthermore, rat fibroblasts expressing a myc-TBD fusion protein undergo apoptosis with similar kinetics when cultured
in the presence of 4-hydroxy-tamoxifen (
22
). Therefore, after ligand binding and release from the Hsp90 complex, the
different HBD-containing fusion proteins apparently can exert their diverse biological
functions with similar efficiency.
The control exerted by the TBD on the Cre enzyme is tight, but not absolute.
When the Cre-TBD-transfected ES cells were kept for >10 weeks in normal medium,
between 5% and 10% of the cells stained with X-Gal (data not shown). The background Cre activity may be due to a cleavage
of the linker between the Cre and TBD by intracellular proteases. To overcome
this problem, we designed a Cre enzyme carrying a TBD at both of its termini.
ES cells expressing this TBD-Cre-TBD fusion protein showed a much lower basal Cre activity. The TBD-Cre-TBD protein, however, was still a functional enzyme
which upon ligand binding could mediate site-directed recombination with a kinetic similar to that exerted by the Cre-TBD protein.
We have designed the Cre/TBD system for the purpose of inducing the deletion or
expression of genes in animals. The Cre-TBD or TBD-Cre-TBD recombinases can be activated by tamoxifen or 4-hydroxy-tamoxifen but not by the endogenously expressed
oestrogen. Therefore, these proteins should be better suited for
in vivo
experiments than Cre fusion proteins carrying the wild-type oestrogen binding domain (
16
). The two drugs can be easily administered in drinking water to transgenic
mice. At the dose required for Cre-TBD or TBD-Cre-TBD activation, these compounds are not toxic. Indeed breast
cancer patients (
28
), as well as mice (
29
,
30
), have been treated with these drugs over prolonged time.
As these fusion proteins are efficient recombinases, the deletion or activation
of a murine gene flanked by
loxP
sites should require only a few days of exposure to the drug. With such a short
period of treatment, the side effect of these anti-oestrogen drugs should remain relatively mild and should rapidly vanish
upon withdrawal of the drug. However, whether Cre fusion proteins also work
efficiently in different cell types of an animal has to await the analysis of
Cre/TBD-trangenic mouse lines. By using a tissue-specific promotor in front of the CreMer or MerCreMer cDNA the
fusion proteins can be expressed not only in an inducible but also in an tissue-specific manner. Similar to the inducible tetracyclin repressor systems (
31
,
32
), the Cre/TBD system thus may be developed into a useful new genetic tool to
mutate or delete genes from the genome of a living organism.
We would like to thank Dr Stewart for communication of unpublished data and Drs
Leclercq, Nielsen, Nitschke and Shaw for critical reading of this manuscript.
This work was supported by a grant from the Deutsche Forschungsgemeinschaft
(SFB388).
REFERENCES
Return