ABSTRACT
To create a strategy for inducible gene targeting we developed a Cre-lox recombination system which responds to the synthetic steroid RU 486. Several fusions between Cre
recombinase and the hormone binding domain (HBD) of a mutated human
progesterone receptor, which binds RU 486 but not progesterone, were constructed. When
tested in transient expression assays recombination activities of all fusion proteins were responsive to RU 486, but not to the endogenous steroid progesterone. However, the observed induction of recombination activity by the synthetic steroid varied between the different
fusion proteins. The fusion with the highest activity in the presence of RU 486
combined with low background activity in the absence of the steroid was tested after stable expression in fibroblast and embryonal stem (ES) cells. We could demonstrate that its
recombination activity was highly dependent on RU 486. Since the RU 486 doses
required to activate recombination were considerably lower than doses
displaying anti-progesterone effects in mice, this system could be used as a valuable tool
for inducible gene targeting.
Gene disruption by homologous recombination has recently gained importance in
the study of gene function (
1
). However, this approach has several inherent problems, such as possible early
lethality, the inability to study the role of a gene in a specific tissue and
the inability to do so at a given point in time. The use of a recombinase
expressed in a tissue-specific manner can partially overcome these limitations (
2
). Temporally regulated recombination should permit studies in a single mouse
before and after inactivation of the appropriate gene. To develop a system for
inducible gene targeting we combined the Cre-lox system with the regulative properties of a mutant human progesterone
receptor (hPR891) that responds to a synthetic steroid (RU 486) but not to
endogenous steroid (
3
).
The
cre
gene of coliphage P1 encodes a 38 kDa site-specific recombinase of the integrase family. It efficiently promotes
intra- and intermolecular recombination at specific 34 bp sequences termed loxP
sites (
4
). Depending on the orientation of the loxP sites, intramolecular recombination
between two sites will lead to reversible excision or inversion of the
intervening DNA sequence (
5
). Intermolecular recombination can result in integration of loxP-containing circular DNA (
6
) or, if two loxP sites are on different chromosomes, in chromosomal
translocations (
7
,
8
). Cre-mediated recombination is not restricted to prokaryotes and has been
reported in yeast (
9
), plants (
10
), mammalian cells (
11
) and mice (
12
,
13
).
Steroid receptors are modular proteins organized into structurally and functionally defined domains (
14
). Fusing the C-terminal hormone binding domain (HBD) onto other transcription factors or tyrosine kinases confers hormone-dependent activities on these proteins (
15
-
18
). Recently it was shown that HBDs from estrogen, androgen or glucocorticoid
receptors confer ligand-dependent, site-specific recombination. (
19
,
20
). These observations outline the basis for direct, ligand-responsive temporal regulation of site-specific recombination. To explore the potential of this strategy we
fused the progesterone HBD to the Cre recombinase. Rather than use the wild-type progesterone HBD, a mutant human progesterone receptor (hPR891) containing a 42 amino acid C-terminal deletion was used. The hPR891 HBD was selected because it
is unable to bind progesterone or other endogenous hormones (
3
). However, it still binds the synthetic steroid RU 486, which has well-characterized anti-progesterone and anti-glucocorticoid properties (
21
). Previous studies with hPR891 showed that when fused to another transcription
factor the ligand binding domain of hPR891 confers RU 486 responsiveness (
22
). By fusing the hPR891 HBD onto Cre we sought to create a ligand-regulated recombinase that can be used in mice, since it should be insensitive to
the presence of endogenous steroids. To find an optimal Cre-hPR891HBD fusion a series of fusion proteins were constructed and tested in green monkey CV-1 and mouse embryonal stem (ES) cells.
CV1 cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum and 100 U/ml penicillin and streptomycin. R1 ES
cells (
23
) were grown in Dulbecco's modified Eagle's medium supplemented with 15% fetal
calf serum, 100 [mu]M non-essential amino acids, 1 [mu]M [beta]-mercaptoethanol and leukemia inhibitory factor (LIF) (ESGROtm; Gibco-BRL) at 37oC in a humidity-saturated 9% CO
2
atmosphere. The cells were cultured on confluent feeder layers mitotically
inactivated by treatment with mitomycin C, except when cells were grown in the
presence of neomycin, when they were plated onto gelatin-coated plates.
The plasmid containing NLS (nuclear localization sequence)-Cre was constructed by inserting the Cre recombinase-containing fragment from pMC1 (pIC1) (
24
) into pHD2 (G.Müller, F.Weih and G.Schütz, unpublished data). The plasmids containing NLS-lacZ and alkaline phosphatase were obtained by inserting
fragments containing the open reading frame (ORF) of NLS-lacZ (kindly provided by M.Pontoglio) and of placental alkaline
phosphatase (kindly provided by J.Sharpe) into pHD2. The target vector was
constructed by inserting the loxP-flanked
neo
cassette from ploxPneo-1 (provided by A.Nagy) into pHD2-lacZ between the promoter and the
lacZ
gene.
CrePR1-CrePR7 fusion genes were obtained by means of PCR, as described by Horton
et al
. (
25
). To construct the CrePR1 and CrePR2 fusions the Cre fragment was amplified by
PCR from pIC1 using an N-terminal oligomer A (AAG CAA CTC ATC GAT TGA TTT ACT G) and a C-terminal oligomer B (AAC TTT TTA TCG CCA TCT TCC AGC AGG C). The
hPR891 HBD fragment was amplified by PCR from pGL-VP (
22
) with oligonucleotides C (GAT GGC GAT AAA AAG TTC AAT AAA GTC A) and D (CGG CTC
GAG CTC TAG AGT CAG CAG TAC AGA TGA AGT TGT). The overlapping PCR fragments
obtained were used in a fusion PCR with oligonucleotides A and D as described
by Horton
et al
. (
25
). The resulting fragment was then subcloned into
Cla
I and
Xho
I sites of pMC1 (or pIC2 for CrePR2). The whole CrePR1 ORF was finally cloned
into the
Sma
I/
Xho
I sites of pHD2. For stable expression a
Bgl
II-
Xho
I fragment, including the CrePR1 ORF, was inserted into
Sca
I/
Xho
I-digested pZeoSV (Invitrogen) plasmid.
CrePR3-CrePR9 fusion genes were obtained by a similar method using for CrePR3
GTG AAT CTA TCG CCA TCT TCC AGC AGG C as oligomer B and CAT GGC GAT AGA TTC ACT
TTT TCA CCA G as oligomer C, for CrePR5 TTG ATC AGA TCG CCA TCT TCC AGC AGG C
as oligomer B and GAT GGC GAT CTG ATC AAC CTG TTA ATG A as oligomer C. To clone
CrePR9, in which the HBD is positioned N-terminal, the hPR891 NLS-HBD sequence was obtained by PCR with oligomers E (GGA AGA TCT TCC
ACC ATG CCC AAG AAG AAG AGG AAG GTG AGA TTC ACT TTT TCA CCA GGT CAA G) and F
(TAA ATT GGA GCA GTA CAG ATG AAG TTG T), the Cre sequence by PCR with oligomers
G (CTG TAC TGC TCC AAT TTA CTG ACC GTA) and H (AGA AGA TAA TCG CGA ACA TCT TCA
G). Fusion PCR with oligomers E and H ended with a
Bgl
II-
Nru
I fragment, which was inserted into pCre1. Subcloning this fragment into pNLS-Cre yielded CrePR7.
CV1-5B cells were plated into 6 cm dishes, grown for 24 h and then transfected
with 1 [mu]g NLS-Cre or an equimolar amount of the different test DNAs. In each
transfection 1 [mu]g pHD2-AP (encoding the placental alkaline phosphatase gene) was included as
an internal control. Transfection was performed by calcium phosphate precipitation as
previously described (
26
). Cells were washed in phosphate-buffered saline (PBS) 12 h after transfection and 100 nM RU 486 or 1 [mu]M progesterone was added to the medium. Cells were harvested 24 h later and fixed in PBS containing 1% formaldehyde
(Merck) and 0.2% glutaraldehyde (Serva). They were stained overnight with a
solution containing 4 mM potassium hexacyanoferrate (III), 4 mM potassium hexacyanoferrate (II), 2 mM MgCl
2
and 0.4 mg/ml X-gal (Biomol). After heat inactivation of endogenous alkaline phosphatase
(AP) activity (30 min, 65oC) cells were stained for AP activity with fast red tablets (Boehringer
Mannheim). Red and blue cells of an area of 120 * 2 mm were then counted. [beta]-Galactosidase activity from cytoplasmic extracts was measured
in the presence of orthonitrophenyl-[beta]-D-galactoside as described (
27
).
Stably transfected subclones of the CV1 cell line (
28
) were obtained by the electroporation of 10
7
cells in 0.8 ml PBS containing linearized plasmid DNA (10 [mu]g CrePR plasmids or 4 [mu]g target vector). A BioRad Gene Pulser was set at 960 [mu]F and 300 V. Cells were grown for 24 h prior to starting selection
with 250 [mu]g/ml Zeocin (Invitrogen) or 400 [mu]g/ml G418 in the medium.
Electroporation of R1 ES cells was after trypsinization and resuspension in PBS
at a concentration of 1 * 10
7
/ml. For each individual transfection 0.8 ml cells (5.6 * 10
7
) were mixed with 20 [mu]g linearized DNA in an electroporation cuvette with a 0.4 cm electrode gap.
Cells were electroporated with a BioRad Gene Pulser set at 240 V, 500 [mu]F. R1 cells were first electroporated with
Sca
I-linearized pPGKpaX1 plasmid (target construct; Fig.
1
B). The cells were plated onto three 10 cm gelatinized plates. After 24 h cells
were fed with medium additionally supplemented with puromycin at 1 [mu]g/ml final concentration. After selection with puromycin for 10 days
colonies were picked for expansion and a second round of transfection with the
CrePR expression plasmid.
To test the activity of the different fusion proteins we constructed a target
vector and stably integrated it into CV-1 cells (Fig.
1
A). This vector contains an ORF of the
Escherichia coli
[beta]-galactosidase gene which is not expressed, since it is separated
from the promoter sequence by a neomycin resistance gene flanked by two loxP
sites. In the presence of active Cre recombination between the loxP sites
excises the neomycin resistance gene and its polyadenylation signal, leading to expression of the
E.coli
[beta]-galactosidase gene. Cells which undergo recombination will turn blue
after staining with X-gal. We analyzed 24 G418-resistant clones obtained after transfection of the target vector
into CV-1 cells. Of the 16 clones that showed Cre-dependent [beta]-galactosidase expression, one clone (5B) was chosen for
further work since it contained only one integrated vector copy (data not
shown).
We first compared Cre and NLS-Cre on 5B cells by transient transfection. NLS-Cre has the SV 40 large T NLS at its N-terminus, a modification previously employed to increase the
activity of Cre in mammalian systems (
24
). Figure
2
shows that NLS-Cre is indeed more active than unmodified Cre. As a further control, a
construct that transiently expresses [beta]-galactosidase without the need for recombination (pHD2-lacZ) was included. Figure
2
indicates that [beta]-galactosidase expression achieved by NLS-Cre-mediated genomic recombination is at least as efficient
as that achieved by direct transient expression from pHD2-lacZ. No [beta]-galactosidase activity was observed in 5B cells in the absence
of transfected plasmids (data not shown).
In order to obtain a CrePR fusion protein which is well regulated by RU 486
several fusions of the Cre recombinase and the hPR891 HBD were constructed.
According to results obtained from previous HBD fusion protein studies, the
preferred fusion construct has the HBD as close as possible to the sequence encoding the required activity (
18
). Mutational analysis of the bacteriophage Cre recombinase showed that
mutations distributed over the entire protein can lead to completely or severely reduced activity (
29
). Therefore, hPR891 HBDs fused to both N- and C-temini of Cre were generated. Four different HBD fusions to the C-terminus of Cre were tested, (CrePR1-CrePR3 and CrePR5; Fig.
2
). They differ in the length of the PR D domain or the presence of a synthetic
SV40 large T NLS at the N-terminus of Cre. The D domain of steroid receptors is proposed to function
as a hinge region and plays a role in nuclear localization. A single N-terminal HBD-Cre fusion was constructed (CrePR9; Fig.
2
). Finally, a double fusion to both termini of Cre was also tested (CrePR7; Fig.
2
).
The constructs were transfected into 5B cells and their activity was tested in
the presence of 100 nM RU 486 and in its absence (Fig.
2
). CrePR1, which includes the complete D domain, showed ~1% of the activity of the wild-type NLS-Cre recombinase in the absence of RU 486. Treatment with RU 486 increased the
activity to 44% (Fig.
2
). Shortening (CrePR3) or removing (CrePR5) the D domain did not enhance the efficiency of the system. The background of
recombination was comparable with the background exhibited with CrePR1, whereas the induction by RU 486 was significantly
lower.
Fusing the hPR891 HBD to the N-terminus of Cre recombinase reduced Cre activity in the presence of the
hormone. Interestingly, the observed inducibility was higher, due to a strongly
reduced background in the absence of RU 486. Compared with NLS-Cre, cells cultured with or without RU 486 displayed 8.5 or 0.07%
activity respectively. Fusing the HBD to both termini of the recombinase
completely abolished the background. Adding the anti-hormone restored the activity, but only to 1%.
In transient transfection experiments the presence of a NLS at the N-terminus of wild-type Cre enhances the recombination activity by ~2-fold (Fig.
2
;
24
). The requirement for a NLS in CrePR constructs was tested by comparing CrePR1
with CrePR2. No differences in activities were detected in the induced or in the uninduced states. This may reflect the fact that HBDs appear to carry nuclear
localization functions.
It is expected that all CrePR (hPR891) fusions cannot be activated by
progesterone (
3
). Several of the fusion proteins were tested for this by incubating transfected
cells with 1 [mu]M progesterone. None of them responded to progesterone (Fig.
2
).
To determine the RU 486 inducibility of genomically expressed CrePR fusion
proteins stable cell lines were established. Three fusions (CrePR1, CrePR3 and
CrePR9) were cloned into pZeoSV (Fig.
1
A), electroporated into 5B cells and selected for Zeocin resistance. From each
electroporation over 50 clones were expanded and tested for recombination
activity by culturing them for 3 days with or without 100 nM RU 486. The cells
were then stained with X-gal in order to assay the frequency of recombination.
As observed in the transient transfection experiments, stably expressed CrePR9
was much less active than CrePR3, while CrePR3 was less active than CrePR1 (data not shown). Three clones expressing
CrePR1 (5B-3, 5B-7 and 5B-32) were selected for further analysis, since they showed a
high recombination frequency combined with good RU 486 inducibility.
To determine the recombination rate more precisely we measured the accumulation of recombination events over time. Cells were cultured for 9 days in the presence or absence of RU 486. At different time
points the percentage of blue cells was determined after X-gal staining. After 9 days clone 5B-32 displayed the highest background (0.15% blue cells) and the
highest recombination activity when induced with RU 486 (40% blue cells). For clones 5B-3 and clone 5B-7 the percentages were 0.13 and 30% and 0.03 and 32% respectively. Thus a 200- to 1000-fold induction of CrePR-mediated recombination by RU 486 was observed
(Fig.
3
A). The differences in recombination activity may result from different levels of
CrePR1 expression in the 5B-derived clones. The results obtained by X-gal staining were confirmed by
in vitro
enzymatic assay measurements (Fig.
3
B) and by Southern blot analysis (Fig.
3
C). The latter demonstrated the predicted excision of the loxP-flanked neomycin resistance cassette. Quantification of this Southern
analysis using a phosphorimager showed 45% recombination for clone 5B-32 after 9 days exposure to RU 486. This value is very similar to the
percentage of blue staining cells (40%), thus confirming that
in situ
staining for [beta]-galactosidase activity is a faithful reflection of recombination.
Since RU 486 has anti-progesterone and anti-glucocorticoid activities (
21
,
30
), the use of CrePR fusions for gene targeting in animals should be performed
using the lowest possible concentration of RU 486. To estimate the required RU 486 concentration which can
activate the CrePR fusions we determined recombination activity by measuring
enzymatic [beta]-galactosidase activity after culturing 5B-3, 5B-7 or 5B-32 cells in the presence of variable amounts of
the steroid. As shown in Figure
4
, substantial activation is obtained with 1 nM RU 486. Induction with 100 pM
RU486 is observable, although it is weak (Fig.
4
).
Based on the above studies, we selected CrePR1 as the best fusion protein for
use in mice. To this end we stably introduced a recombination reporter into ES cells (Fig.
1
B) and then selected a clone (X3P) that contained a single integrated copy (data
not shown). CrePR1 was then stably introduced into X3P cells using a PGK expression
vector carrying a neomycin resistance gene (Fig.
1
B). Of 96 neomycin-resistant clones selected 83 showed RU486-inducible recombination, as assessed by staining for [beta]-galactosidase expression (data not shown). However, all
clones also showed variable levels of staining in the absence of induction, indicative of a background level of ligand-independent recombination. This can be seen in Figure
5
A, which shows [beta]-galactosidase staining of a representative clone. Strong induction
by RU 486 to nearly complete recombination is also apparent, as is the fact that exposure to RU 486 does not induce morphological
changes in ES cells. Figure
5
B shows Southern analysis of the time courses of RU486 administration.
Quantification (Fig.
5
C) revealed that uninduced recombination in clone X3P1 was 4.7% at the start of
the incubation and 5.5% after 120 h, whereas in X3P12 it was 2.0% at the start
and 2.0% at the end. RU486 treatment of X3P1 resulted in over 80% recombination
within the time course of the study, whereas X3P12 was ~50% recombined after 120 h.
To develop a system in which gene targeting could be induced when desired we
combined the use of the Cre-lox system with the regulative properties of the mutated human
progesterone receptor hPR891. We therefore fused the hPR891 HBD domain to the
phage P1 Cre recombinase. General observations on fusions between steroid
hormone binding domains and heterologous proteins (
18
) showed that the closer the HBD is fused to an active domain the stronger is
its inhibition. Since active regions were described in both the N- and the C-termini of Cre, we designed several fusion proteins which differed
in the position of the HBD or the length of the hinge region (D domain). Their
activities were compared in transient transfection experiments with the
activity of a NLS-containing Cre (NLS-Cre). The highest activity in combination with good inducibility
was observed when the HBD, including most of the D domain, was fused to the C-terminal of Cre (CrePR1). An N-terminal fusion was less active, although it displayed higher
induction due to a lower background. A double fusion (CrePR7) resulted in the
tightest regulation, but also showed a very low activity when induced with the
anti-hormone.
The presence of a NLS enhances the activity of Cre ~2-fold. This is consistent with previous observations (
24
). However, the presence of a NLS had no influence on CrePR1 activity. Although
the previously described hPR NLS is almost completely removed in CrePR1, a
further mechanism for nuclear localization could be mediated by the hPR HBD,
which is not dependent on activation of the DNA binding domain (
31
).
The most promising fusions (CrePR1, CrePR3 and CrePR9) were tested by stable expression in a CV1-derived cell line (5B). Both clones CrePR3 and CrePR9 showed poor to
moderate inducibilities. Clone CrePR1 showed better RU 486 induction
properties, however, the recombination rate was still low (3-4.5%/day; Fig.
3
). We believe that the low recombination rate is probably due to low expression
of the CrePR1 construct in the CV1 clones, since its expression, as determined
by RNase protection experiments, was very weak (data not shown).
In contrast, activation of CrePR1 in ES cells was more efficient (Fig.
5
), suggesting that inducible gene targeting in cells or animals is feasible. A
recombination rate of 10-30% cells/day (Fig.
5
) might be too slow for gene disruption in embryogenesis, where rapid
recombination in all cells is required, however, enhanced Cre expression levels
should enhance the rate of recombination. The use of strong and specifically
activated promoters are presently being tested to investigate control at the
level of expression and protein activity.
The Cre-hPR891 system presents several advantages for use in mice. First, it does
not respond to any endogenous hormone, even in the presence of 1 [mu]M progesterone, where we could not detect any activation. Second, RU 486 has
been shown to be rapidly and widely distributed after oral or i.v.
administration (
32
). Finally, the abortifacient properties and toxicity of RU 486 require higher
doses than that required for induction of CrePR1 (Fig.
4
;
33
,
30
).
In vivo
studies in rats demonstrate that RU 486 administered at doses of 400 [mu]g/kg or lower cannot induce abortion or prevent nidation (
30
). Mice are less sensitive to RU 486 and need 3-fold higher doses to induce abortion or prevent nidation (
30
). We show here that CrePR(891) is inducible at a concentration of 1 nM. As a
mouse serum concentration of 1 nM corresponds to an orally administrated dose
of ~5 [mu]g/kg (
22
), the system should work in the mouse with 100- to 1000-fold lower concentrations of RU 486 than needed for abortion or anti-nidation. Toxicological studies on RU 486 showed no
teratogenic or mutagenic effects in rats and mice or during embryogenesis (
33
).
Previous work has shown that HBDs from estrogen, glucocorticoid and androgen receptors can be fused to FLP recombinase to generate ligand-inducible recombinase-HBD fusion proteins (
19
). Also, work suggesting that a fusion between Cre and the estrogen receptor is estrogen inducible has been published (
20
). We demonstrate here that Cre is strongly regulated by RU 486 when it is fused
to the hPR891 HBD. Others have reported the use of inducible Cre recombinase
under the control of the interferon-responsive Mx promoter (
34
). The strategy presented here has an important advantage. Induction of the
CrePR fusion relies on the use of a synthetic steroid that binds to a mutated
HBD. Thus interference by endogenous mechanisms is limited.
The high inducibility of the reported system, combined with the described advantages of the inducing agent, suggests it as a useful system for
inducible gene targeting in cells or animals. Introduction of a CrePR fusion into animals containing a gene locus which can be inactivated
via recombination between two loxP sites is under way.
We thank W.Fleischer for synthesis of oligonucleotides and photographic work. We
thank Drs M.Pontoglio, J.Sharpe, A.Nagy, S.Tsai and H.Gu for the gift of NLS-lacZ, placental alkaline phosphatase and the ploxPneo-1, pGLVP and pIC-Cre plasmids respectively. We are indebted to Drs
T.Mantamadiotis and B.Lutz for helpful discussions and critical reading of the
manuscript. FT is in receipt of a Long-Term Fellowship from EMBO. P-OA is a recipient of an EU Human Capital and Mobility Fellowship.
This work was supported by the Deutsche Forschungsgemeinschaft through SFB 229, the Fonds der Chemischen Industrie, BMFT-Projekt 0310 681, and by European Community grant no. BI02-CT93-0319.
REFERENCES
Return