ABSTRACT
We used the recombination-promoting activity of the polyomavirus large T antigen (T-ag) to increase the frequency of gene targeting in rat fibroblasts. We constructed a
cell line carrying a functional polyomavirus replication origin and a
transformation-defective middle T-ag oncogene. The structure of the locus was such that homologous recombination with the targeting DNA reconstituted a functional transforming gene and converted the
cells from the normal to the transformed state. Introduction of the large T-ag with the targeting DNA promoted recombinational events that corrected
the mutation in either the target locus or the targeting DNA. The frequency of recombination was not substantially influenced by the extent of homology between the recombining
sequences. However, it was reduced when the replication origin was inactivated
in the targeting DNA, and was reduced further when the origin was inactivated
in the target locus.
Gene targeting by homologous recombination has emerged in recent years as a
powerful tool for producing specific mutations in the mammalian genome. Most
experiments have been carried out with mouse embryonic stem (ES) cells to
generate mice heterozygous or homozygous for the targeted mutation. Analysis of
the phenotypes of mutant mice has provided insight into the role of a number of
genes with unknown function (
1
-
5
). In ES cells, the frequency of gene targeting is influenced by a number of
parameters. Two types of vectors are commonly used, insertion and replacement
vectors, which differ essentially by the site of the cut with respect to the
region of homology. According to some studies, insertion vectors which are cut
at a site within the homology are more effective than replacement vectors with
the same length of homologous sequence (
6
), and targeting plasmids that carry a double-stranded gap in the region homologous to the targeted locus are more
efficient than plasmids introduced in their uncut form (
7
). With both types of vectors, the targeting efficiency increases exponentially
with the extent of homology between the vector and the target locus (
8
) and isogenic DNA targets more efficiently than does nonisogenic DNA (
9
,
10
).
Homologous recombination has also been exploited to introduce or correct mutations in somatic cell lines. The technique has been used to
study gene expression (
11
,
12
) and function (
13
,
14
) but the frequency of homologous recombination is generally lower than in ES cells. We have attempted to increase the frequency of gene targeting
in somatic cells by extension of previous work in this laboratory (
15
-
18
) which described high frequency intrachromosomal recombination promoted by the polyomavirus large T-ag. Polyomavirus is a papovavirus that can transform the growth properties
of rodent cells in culture. Transformation is the result of stable integration
of the viral DNA sequences into the host genome and subsequent expression of
the viral early region (
19
). This region encodes three distinct proteins in alternate translational
reading frames that are referred to as the large T-ag, middle T-ag and small T-ag. Large T-ag is associated with a DNA helicase activity and is
responsible for the duplex-unwinding activity that is likely to initiate viral DNA replication
(reviewed in
20
). It is known to promote amplification and excision of integrated viral
sequences (
17
,
21
-
23
). Both phenomena require a functional large T-ag, the viral replication origin and some homology within the integrated
sequences (
24
).
In the present work, we attempted to use the recombination-promoting activity of polyoma large T-ag to increase the frequency of gene targeting in rat fibroblasts.
We constructed a cell line carrying a functional polyomavirus replication
origin and a transformation-defective middle T-ag gene. The structure of the locus was such that recombination with
the targeting DNA reconstituted a functional transforming gene and converted
the cells from the normal to the transformed state.
As chromosomal target, we used a provirus stably integrated into the genome of
rat embryo fibroblasts (FR3T3). The retrovirus sequences (Fig.
1
A) were derived from Spleen necrosis virus (SNV) with splice acceptor sequences
from reticuloendotheliosis virus (strain A) (
25
). The splicing vector, pMT-dl1348, was constructed from pOri-hygro (
23
). It contained polyomavirus sequences from nucleotides (nt) 4973 (
Acc
I site converted into
Eco
RI) to 1656 (
Hin
dIII site). These sequences included the replication origin (ori) (nt 4973-173) and the middle T-ag coding sequence. The transforming potential of middle T-ag was inactivated by a 28 base-pair (bp) deletion between nt 1348 and 1377. The hygromycin-resistance gene (hygro) was separated from middle
T-ag by the splice acceptor and polylinker sequence and a
Hin
dIII-
Bam
HI fragment (nt 29-375) from pBR322. The
Bam
HI site was inactivated and replaced by a
Not
I site. pMT-dl1348 has the structure of the provirus shown in Figure
1
A except that the right-side LTR is ligated to position 2066 of pBR322 and the left-side LTR is ligated to position 4361. Details on the splice donor,
polylinker sequences and sequences required in
cis
for viral replication are provided by Dougherty and Temin (
25
).
The vectors were introduced by DNA transfection into D17.2G, a dog helper cell
line which supplies
trans
-functions for the packaging of defective retrovirus (
26
). Selection for hygromycin resistance was done in the presence of 70 [mu]g/ml hygromycin (Eli Lily & Co.). Virus harvested from cultures of hygromycin-resistant D17.2G cells was used to infect FR3T3 cells.
Cells were grown at 37oC in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf
serum. To express the polyomavirus large T-ag, the cells were transfected with pneo-LT1 (
15
) and G418 selection was carried out as described previously (
27
).
The arrangement of viral sequences within the cellular DNA was analyzed by Southern blotting. The DNA was digested by restriction endonucleases
and fractionated by agarose gel electrophoresis. The fragments were transferred
onto nylon membranes and hybridized to a
32
P-labeled probe. The DNA was labeled by random priming to a specific
activity of ~10
8
c.p.m./[mu]g DNA.
We used a spleen necrosis virus-based retroviral vector, containing the hygromycin-resistant gene, to introduce a transformation-defective middle T-ag oncogene into FR3T3 cells. Colonies of hygromycin-resistant cells were isolated, established into
cell lines and the cellular DNA was analyzed by Southern blot hybridization.
The number of proviral integrations was determined in each cell line by
cleaving the DNA with
Bgl
II, an enzyme cutting outside the insert. One such cell line, designated A3-2, had a single insert (see below) and was selected for further studies.
Like all of the cell lines established in this manner, A3-2 had a normal phenotype indistinguishable from that of FR3T3.
We wished to determine whether polyomavirus large T-ag could promote homologous recombination between the provirus and various DNA targeting constructs. To this end, A3-2 cells were transfected with pneo-LT1, the vector carrying the large T-ag gene, and the targeting vectors depicted in Figure
1
. pdl13 and pdl13hy carried a 15-bp deletion (nt 209-223) that inactivated transformation. However, as shown below, homologous recombination
between the transfected DNA and the provirus reconstituted an intact coding
region for middle T-ag and converted the cells from the normal to the transformed state.
First, we attempted to target the provirus using a protocol for transient expression of the transfected DNA. Forty-eight hours post transfection, the cells were plated at a density of 150
000 cells per dish, grown to confluence and observed for morphological
transformation. Foci of transformed cells appeared after 2 weeks in culture
(Table
1
). Interestingly, the efficiency of transformation was not reduced by decreasing
the degree of homology between the provirus and the targeting vector. A few
foci appeared in the absence of large T-ag. However, they did not contain an intact middle T-ag-coding sequence (not shown) and were considered to be
spontaneous transformants (see below). The combination of pdl13 and pneo-LT1 did not produce any significant transformation when transfected into
FR3T3 cells (not shown). Consequently, we expected no more than 5-10% of the foci in the assay to be transformants arising independently of
homologous recombination in the provirus.
In parallel experiments, the transfected cells were selected for G418-resistance. After selection, the colonies exhibiting a transformed
morphology were scored as transformants (Table
1
). Colonies were also picked at random, independently of their phenotype,
transferred into 15-mm Linbro microplates and observed for morphological transformation. Some
of the colonies, while having a normal phenotype at the time of selection,
yielded small foci of transformed cells as they reached confluence in
microplates. Transformants were scored in 17 out of 115 clones when large T-ag was present in the transfection (Table
1
). No transformant appeared when
neo
was transfected alone. From previous transfection experiments (
23
,
27
), we expected about half of the G418-resistant clones to express large T-ag.
Table 1
To analyze changes in the provirus structure, the cellular DNA was isolated from
representative transformants and examined by Southern blot hybridization. An
intact middle T-ag-coding sequence contains an
Sst
I site (nt 1373) that can be detected by a 742-bp fragment (Fig.
1
F). This fragment was seen in all of the dl13hy transformants except in
dl13hy.4, suggesting that the latter was a spontaneous transformant (Fig.
2
A). Correction of the 28-bp deletion was expected to produce a 1998-bp
Sst
I fragment. Such a fragment was not detected in the transient expression
experiment (Fig.
2
A), but the 1580-bp fragment was consistent with a gene conversion or double crossover
event that eliminated the 419-bp pBR segment from the provirus (Fig.
1
F). In addition,
Hin
dIII digestions revealed that 10 out of 12 samples analyzed contained no other
fragment than the 4.7-kb fragment already present in the parental A3-2 cell line (not shown). This indicated that, at least in some
cases, neither large T-ag nor the targeting DNA had to become integrated into the cellular DNA to
induce transformation and that reconstitution of a functional middle T-ag-coding sequence occurred by homologous recombination between the
provirus and the transfected DNA.
The intensity of some of the
Sst
I fragments on Southern blots (e.g. dl13h.9, dl13h.25; Fig.
4
A) suggested that some sequences were amplified. Amplification could occur by
intrachromosomal recombination between the two retroviral LTRs followed by
replication of the circular DNA containing the polyomavirus origin of DNA
replication. To investigate this possibility, PCR analyses were performed with
oligonucleotide primers (primers 1 and 5; Fig.
5
A) designed to amplify a product from circular DNA molecules. Four out of 10
samples analyzed yielded PCR products. Two of them (dl13h.9 and dl13h.25)
yielded a 4.0-kb fragment that had the structure shown in Figure
5
B. Such a molecule could be produced if the dl13h DNA underwent homologous
recombination in the LTR repeats following transfection. Since primer 1 did not
recognize the dl13 mutation, the latter must have been corrected prior to
recombination. However, primer 1' too amplified a product from the same samples (Fig.
4
C), indicating that the targeting DNA persisted in the transfected cells for
many generations. Two samples (dl13h.22 and dl13h.33) yielded a 4.4-kb product with the structure shown in Figure
5
C. This product, which did not contain a functional middle T-ag, arose by homologous recombination in the proviral LTR. Transformation
of these cell lines was presumably triggered by correction of the dl13 mutation
in the targeting DNA. A 1580-bp
Sst
I fragment can be detected in sample dl13h.33 (Fig.
4
A). It is not known yet whether the DNA is free or stably integrated in the
cellular genome.
We wished to determine whether recombination in the provirus required a
functional polyomavirus replication origin. To this end, we constructed a
vector similar to dl13.hy but with a deletion in the origin region (nt 4973-93) that inactivated replication. Transfection of this DNA into A3-2 with pneo-LT1 yielded 514 G418-resistant colonies, five of which exhibited a
transformed phenotype, i.e. ~1/3 the number of transformants obtained with dl13.hy, the vector
containing a functional origin (Table
1
). Analysis of these transformants by Southern blotting and PCR revealed that
they all carried a provirus lacking the
Sst
I site at nt 1373 and that the dl13 mutation had been corrected (not shown).
In another approach, we constructed pMT-dl1348ori
-
, a retroviral vector identical to pMT-dl1348 except for a deletion in the replication origin (nt 5285-17), and we established two cell lines, ori
-
.4 and ori
-
.33, that carried the mutant middle T oncogene with a defective replication
origin. Transfection of the targeting DNA (dl13hy) with large T-ag resulted in a few transformants (five out of a total of 222 colonies
analyzed), i.e. ~6.5 times less than in the A3-2 cell line. When these transformants were analyzed by Southern
blotting and PCR, they too appeared to have corrected the dl13 mutation in the
targeting DNA (not shown).
In this paper, we describe a system in which homologous recombination occurs
between transfected DNA and a specific chromosomal locus. We constructed a rat cell line containing a functional polyomavirus origin of DNA replication and a transformation-defective middle T oncogene. The structure of the locus was such that
recombination with the targeting DNA reconstitutes a functional transforming
gene and converts the cells from the normal to the transformed state.
Homologous recombination occurs at a high frequency in the presence of the
polyoma large T-ag and yields products that are compatible with a gene conversion process
in which the targeting DNA acts as the donor of information. Alternatively,
correction of the middle T-ag mutation could occur by a double crossover event resulting in a
reciprocal exchange between the middle T-ag locus and the targeting DNA. Although one cannot formally distinguish
between these alternatives, it is noteworthy that we have been unable to
recover any reciprocal product from such exchange by PCR. There may be several
reasons for this. One is the low frequency of reciprocal exchange-if any-between the locus and the targeting DNA. Another is the unlikely
integration of the product in the absence of selection, and finally, the
possibility that the product is not detected by the oligonucleotide primer
(primer 1') because the recombination process always leads to the repair of the
dl13 mutation in the targeting DNA.
There are other arguments in favor of a gene conversion process. Single reciprocal exchanges are less frequent than gene conversions, which suggests that double reciprocal exchanges are even less likely (
28
,
29
). Single reciprocal recombination of an insertion vector occurs more frequently than double reciprocal recombination of a replacement vector with crossover junctions on both the long and short
arms (
6
). Jasin
et al
. (
30
) have described a system of high frequency recombination in which the
chromosomal DNA acts as the donor of information in a gene conversion process.
Whatever the recombination pathway might be between the middle T locus and the
targeting DNA, reciprocal exchange is not uncommon in rat cells. Previous work
from this laboratory has shown that crossovers between homologous sequences
lying as inverted repeats on the same chromosome occur more frequently than
gene conversion events (
18
,
31
). Whether this is due to the structure or configuration of the DNA is uncertain but, in the presence of large T-ag, intrachromosomal recombination such as excision of sequences by crossover in the LTRs can occur at high rates (
23
and this study). It is possible that one pathway (nonreciprocal versus
reciprocal recombination) is favored under certain conditions such as in
extrachromosomal recombination.
Contrary to our expectation, we have found that the frequency of homologous
recombination in A3-2 is not substantially influenced by linearization of the targeting vector
nor by the extent of homology between the latter and the target locus. Several
studies have reported that introduction of a double-stranded break near one of the recombining sequences enhances the rate of
recombination (reviewed in
10
). The recombination frequency varies with the length of homology (
33
,
34
) and is influenced by the position of the double-strand break with respect to homologous sequences (
7
,
30
). It is possible that in large T-ag-mediated recombination, linearization of the DNA has no effect
because the sequences close to the polyoma replication origin are already in a
configuration favorable for homologous recombination. Previously, it has been
proposed that the role of large T-ag in recombination is to melt and unwind the DNA so as to create single-stranded regions in the vicinity of the viral origin (
15
). There is evidence that a single-stranded DNA molecule can recombine extrachromosomally with double-stranded DNA and that it can directly integrate into the genome
without first being converted to a double-stranded form (
35
).
We have shown that the recombination frequency is reduced when the replication
origin is inactivated in the targeting DNA, and that it is reduced further when
the origin is inactivated in the proviral DNA. Thus, whereas the origin is
important in both DNAs, the presence of a functional origin in the chromosomal
locus is particularly critical. The reason for this is not understood, nor is
the reason why correction of the targeting DNA with the locus as the donor of
information occurs more frequently than the opposite. One possibility is that
minichromosomes introduced into cells by transfection do not have the same
structure or configuration as resident chromosomes and are more prone to
homologous recombination. Interestingly, the placement of a double-strand break in both of the homologous recombining sequences has a
synergistic effect and can increase the recombination rate as much as 100-fold (
36
-
39
). When the break is placed in only one homologous sequence, that sequence acts
predominantly as the recipient of information in nonreciprocal exchanges (
37
). In the SV40/COS cell system described by Jasin
et al
. (
30
), the chromosomal DNA, which lacks a functional SV40 replication origin, acts
as the donor of information in a high-frequency gene conversion process. In this system, up to 25% of the
successfully transfected cells recombine the transfected DNA containing the
SV40 origin. Other studies have shown that in intrachromosomal recombination,
the sequences undergoing gene conversion can act as the donor as well as the
recipient of genetic information (
18
). Whether this particular behavior is typical of intrachromosomal recombination
is still unresolved.
The authors thank C. Bergeron for excellent technical assistance, J. Dougherty
for retroviral vectors and C. Gélinas for fruitful discussions. This work was supported by the Medical
Research Council of Canada. V.F. was supported by a postdoctoral fellowship
from the `Région Rhône-Alpes'.
Experiment
Transfection
Recipient
Targeting
Large T-ag
a
No. of G418
No. of transformants
cell line
DNA
clones
Circular DNA
Linear DNA
e
1
Transient exp.
b
A3-2
dl13hy
+
42
41
dl13hy
-
3
dl13
+
48
43
dl657
+
59
80
dl1016
+
63
54
2
G418 selection
c
A3-2
dl13hy
+
608
17
dl13hy
-
521
1
dl13hy.ori
-
+
514
5
3
G418 selection
d
A3-2
dl13hy
+
115
17
A3-2
dl13hy
-
113
0
ori
-
.4
dl13hy
+
116
2
ori
-
.4
dl13hy
-
120
0
ori
-
.33
dl13hy
+
106
3
ori
-
.33
dl13hy
-
117
0
REFERENCES
Return