Nucleic Acids Research, 2003, Vol. 31, No. 3 899±910
DOI: 10.1093/nar/gkg171
DNA pairing is an important step in the process of
targeted nucleotide exchange
Miya D. Drury and Eric B. Kmiec1,*
Department of Chemistry and Biochemistry and 1Department of Biological Sciences, University of Delaware,
Newark, DE 19716, USA
Received October 16, 2002; Revised and Accepted November 27, 2002
ABSTRACT
Modi®ed single-stranded DNA oligonucleotides can
direct the repair of genetic mutations in yeast, plant
and mammalian cells. The mechanism by which
these molecules exert their effect is being elucidated, but the ®rst phase is likely to involve the
homologous alignment of the single strand with its
complementary sequence in the target gene. In this
study, we establish the importance of such DNA
pairing in facilitating the gene repair event.
Oligonucleotide-directed repair occurs at a low frequency in an Escherichia coli strain (DH10B) lacking
the RECA DNA pairing function. Repair activity can
be rescued by using puri®ed RecA protein to catalyze the assimilation of oligonucleotide vectors into
a plasmid containing a mutant kanamycin resistance gene in vitro. Electroporation of the preformed
complex into DH10B cells results in high levels of
gene repair activity, evidenced by the appearance
of kanamycin-resistant colonies. Gene repair is
dependent on the formation of a double-displacement
loop (double-D-loop), a recombination intermediate
containing two single-stranded oligonucleotides
hybridized to opposite strands of the plasmid at the
site of the point mutation. The heightened level of
stability of the double-D-loop enables it to serve as
an active template for the DNA repair events. The
data establish DNA pairing and the formation of the
double-D-loop as important ®rst steps in the process
of gene repair.
INTRODUCTION
DNA pairing plays a critical role in the cellular response to
DNA damage. In the most traditional repair pathway, single
strands of DNA arising from nicks, gaps or breaks in the
chromosome invade a sister homolog and hybridize to a region
of DNA that has a complementary sequence, usually
initializing its repair. This invasion forms a DNA structure
known as a heteroduplex joint containing paired strands
donated by aligned parental duplexes. Biochemical work,
focused on de®ning the activity of the bacterial recombinase
RecA protein, modeled this reaction using a single strand of
DNA and a superhelical duplex molecule, the conjoining of
which forms a structure known as a displacement loop
(D-loop) (1±6). This triple-stranded con®guration consists of
two strands paired by complementarity and a displaced strand
that can be removed by endogenous nuclease activity. If the
donor strand remains hybridized to its complement in the
recipient, it can initialize a DNA repair reaction or a
recombination event.
The D-loop structure has served as the primary experimental model for many studies aimed at elucidating the molecular
and enzymatic activities that regulate the initiation of DNA
pairing. As such, molecular requirements, including the
energy found in a superhelix in the recipient duplex, for
promoting ef®cient uptake of the single strand have been
identi®ed. In addition, the stereochemistry of the triple-helix
structure created by D-loop formation has been informative
for studies focusing on the biochemical function of RecA
protein (1,7). Energy requirements also entail ATP hydrolysis,
which produces a regulated reaction known as the D-loop
cycle (8,9).
We have been studying targeted nucleotide exchange
(TNE), a process in which a single-stranded oligonucleotide
assimilates into an episomal or chromosomal target region,
perhaps using the D-loop as a reaction intermediate. After
formation, the oligonucleotide directs the repair (or exchange)
of a targeted nucleotide in the parent duplex (10). The
speci®city of nucleotide exchange is dictated by the creation
of a mismatched base pair between the oligonucleotide vector
and a nucleotide in the coding region of the targeted gene. But
the reaction is entirely dependent on the successful pairing of
the single-stranded vector with its complement in the duplex
target. Previous data using cell-free extracts established a role
for mismatch repair genes in the repair phase of the reaction
(11,12), and genetic studies suggested that proteins with DNA
pairing activity that assimilate the single-stranded oligonucleotides into recipient plasmid targets are essential for the
TNE reaction (13±15). It seems probable, therefore, that a
D-loop or D-loop-like structure serves as the central intermediate in the overall TNE pathway. Such a hypothesis gains
support from the pioneering work of Holloman et al. (1,7). In
these studies, D-loops were formed by hybridization and the
joint molecule was transformed into Escherichia coli. The
preformation of the D-loop in phage éX174 DNA alleviated
the need for a DNA pairing function in E.coli. It is generally
*To whom correspondence should be addressed. Tel: +1 302 831 3420; Fax: +1 302 831 3427; Email: [email protected]
Nucleic Acids Research, Vol. 31 No. 3 ã Oxford University Press 2003; all rights reserved
900
Nucleic Acids Research, 2003, Vol. 31, No. 3
believed that these experiments helped de®ne `¼ one of the
most obscure aspects of genetic recombination' (7)Ðits
initiation.
The importance of elucidating the structural intermediate of
TNE emerges from the belief that the success of TNE is likely
to depend (at least in part) on the stability of the joint molecule
serving as the reaction intermediate. And factors contributing
to increased stability of the intermediate would in turn lead to
higher rates of repair. Parekh-Olmedo et al. (16) achieved
higher frequencies of repair by simultaneously targeting a
chromosomal gene with two complementary oligonucleotides.
The use of dual vectors in the TNE reaction is based upon
previously reported biochemical evidence that RecA protein
can catalyze the formation of `double-D-loops', structures in
which the two oligomers are paired, side-by-side, to parallel,
complementary sequences at the same site in a recipient
plasmid (17). These double-D-loops, also known as complement-stabilized D-loops, increase stability by providing a
second oligonucleotide `trap' to hybridize with the displaced
complementary strand of the duplex.
Since elevated frequencies of TNE were found in experiments using two oligomers by Parekh-Olmedo et al. (16), we
wondered whether the double-D-loop structure was serving as
an ef®cient, primed template for the TNE reaction. Thus, in
the present work, we set out to establish reaction conditions
for the facile formation of stable double-D-loops bearing a
single mismatched base in vitro, and asked whether a gene
repair reaction could occur within these preformed templates.
A genetic readout system in E.coli was chosen because of its
successful use in previous studies (11,18) and the fact that
correction results in a phenotypic change. Because we were
working in a bacterial system, we utilized the DNA pairing
function of the RecA protein to preform double-D-loops
in vitro. The complexes were then electroporated into recAde®cient cells, and the frequency of the TNE reaction was
measured by the appearance of antibiotic-resistant bacterial
colonies. Our results strongly suggest that DNA pairing is an
important step in TNE, and that the double-D-loop can serve
as an active template upon which the nucleotide exchange
event takes place.
MATERIALS AND METHODS
RecA and oligonucleotides
RecA protein was obtained from USB (Cleveland, OH).
Single-stranded oligonucleotides were obtained from
Integrated DNA Technologies, Inc. (Coralville, IA). Oligonucleotides were quanti®ed on the basis of spectrophotometric
Abs260 values and using the conversion factor of 33 mg/ml OD.
Oligonucleotides were subsequently labeled by standard
procedures using [g-32P]ATP and T4 polynucleotide kinase.
D-loop preparation
Formation of D-loops was carried out in two steps. A 15-min
pre-synaptic step occurred at 37°C using 40 nM of the
appropriate `incoming' oligonucleotide and 1.5 mM RecA
(ratio of two nucleotide bases per one RecA monomer) in a
solution containing 1 mM ATPgS, 25 mM Tris±OAc (pH 7.5),
1 mM (CH3CO2)2Mg´4H2O and 1 mM DTT. The synaptic
phase began with the addition of 10 nM supercoiled plasmid
and an additional 9 mM (CH3CO2)2Mg´4H2O for 5 min.
Double-D-loops were formed with the addition of the required
320 nM 32P-labeled-`annealing oligonucleotide' having a
length of between 30 and 50 bp. Additional incubation at 37°C
occurred for 10 min, and joint molecules were deproteinated
with the addition of 1% SDS at 4°C. Buffer exchange into
10 mM Tris, 1 mM EDTA was conducted using either
CentriSpin-20 (Princeton Separations, Adelphia, NJ) or
Chromaspin-400 (Clontech Laboratories, Inc., Palo Alto,
CA) columns. Excess unbound oligonucleotide was removed
with the Chromaspin-400 columns. The presence of joint
molecules was con®rmed by 1% agarose gel electrophoresis
prior to use in the in vivo targeting experiments. Quanti®cation
of D-loop and double-D-loop complexes was carried out using
ImageQuant 5.2 software after visualization using a Typhoon
8600 Variable Mode Imager (Molecular Dynamics Inc.,
Sunnyvale, CA) to scan for either phosphorescence or
¯uorescence.
Electroporation, plating and selection
Five microliters of DNA sample (20 ml total) was used to
transform 20 ml aliquots of electrocompetent DH10B
(Invitrogen, Carlsbad, CA) or BL21(DE3) (Clontech
Laboratories, Inc., Palo Alto, CA) bacteria using a Gibco
cell-porator apparatus (Life Technologies, Cleveland, OH)
under the following conditions: 330 mF, 4 WK, 400 V/sample.
Each mixture was transferred to a 1 ml SOC culture [2.0% (w/
v) tryptone, 0.5% (w/v) yeast extract, 10 mM NaCl, 2.5 mM
KCl, 10 mM MgCl2 and 20 mM Glc] and incubated at room
temperature for 3 h. In cases where joint complexes were not
used for transforming bacteria, 50 ng plasmid DNA, 1 mg
incoming oligonucleotide and 2 mg annealing oligonucleotide
were added to the electrocompetent cells immediately prior to
electroporation. Plasmid DNA was ampli®ed by adding
ampicillin to 100 mg/ml and an equal volume of SOC media.
These cultures underwent an additional incubation of 24 h at
room temperature with shaking at 300 r.p.m. One hundred
microliter aliquots of 10±1 and 10±2 diluted cultures were
plated onto Luria±Bertani (LB) agar plates containing 50 mg/
ml kanamycin, and 10±5 and 10±6 diluted cultures were plated
onto LB agar plates containing 100 mg/ml ampicillin. Plating
was performed in duplicate using sterile Pyrex beads. Both
sets of plates were incubated for 24 h at 37°C and colonies
were counted. Targeted conversion of the kanr gene was
determined by normalizing the number of kanamycin-resistant
colonies and dividing by the number of ampicillin-resistant
colonies, with correction ef®ciencies calculated per 105
ampicillin colonies. Resistant colonies were con®rmed by
selecting isolated clones for mini preparation of plasmid DNA
followed by sequencing using an automated ABI Prism 3100
Genetic Analyzer.
RESULTS
Assay system for targeted nucleotide exchange
A genetic readout was used as an assay system to measure
the correction ef®ciency of TNE. Plasmid pKSm4021
(Fig. 1A), bearing a point mutation in the gene that confers
kanamycin resistance onto E.coli, served as the target. A stop
codon (TAG) was created at the indicated site replacing the
Nucleic Acids Research, 2003, Vol. 31, No. 3
901
Figure 1. The assay system and sequences of modi®ed oligonucleotides. (A) Plasmid pKsm4021 has an expression cassette bearing a kanamycin and an ampicillin gene. The kanamycin gene contains a single-base transversion at nucleotide 4021, a replacement mutation of TAT (tyrosine) to TAG (stop codon).
Correction of pKsm4021 requires the replacement of the mutant G residue with a C residue. The plasmid and an (incoming) oligonucleotide are incubated in
the presence or absence of the RecA protein, followed by the addition of a second (annealing) oligonucleotide. After joint complexes are formed and isolated,
the complex is electroporated into E.coli DH10B cells, which are de®cient in RecA protein. A genetic readout of the correction ef®ciency is then analyzed by
normalizing the number of kanamycin-resistant colonies to the number of ampicillin-resistant colonies. (B) Synthetic oligonucleotides were used to direct
reversion of kanr genes to restore resistance to kanamycin. KanNT30C, KanNT70C and KanNT70G are all single-stranded oligonucleotides that target the
non-transcribed strand. KanT30C, KanT50C, KanT70C and KanTF/31G are oligonucleotides that target the transcribed strand of the plasmid target. KanTF/
31G contains a Cy5 ¯uorescent dye at its 5¢ end. pm refers to an oligonucleotide being perfectly matched to its complementary target sequence. (C) Structure
of D-loop (I) and double-D-loop (II) joint molecules as putative reaction intermediates.
wild-type TAT codon. The nucleotide exchange reaction,
directed by speci®c oligonucleotides (see below), results in
the creation of a TAC triplet and reversal of phenotype;
kanamycin sensitivity to resistance. The creation of the
TAC codon can be con®rmed by DNA sequence analyses of
kanr colonies. If the TAC codon was present, kanamycin
resistance arose from a repair event, rather than having
arisen from contamination of a wild-type plasmid, which
contains a TAT codon at the targeted position. Plasmid
pKSm4021 also contains the gene for ampicillin resistance,
which is used as an internal control and a measure of
electroporation ef®ciency.
902
Nucleic Acids Research, 2003, Vol. 31, No. 3
Table 1. TNE in DH10B and BL21 cells
Strain
DNA substrates
Kanr colonies
Ampr colonies(/105)
Correction ef®ciency [Kanr/Ampr (105)]
DH10B
DH10B
BL21
BL21
±
KanT70C/KanNT30C
±
KanT70C/KanNT30C
0
42.5
0
600
642
512
512
709
0.00
0.08
0.00
0.89
Escherichia coli strains DH10B [F± mcrA D(mrr-hsdRMS-mcrBC) f80lacZDM15 DlacX74 deoR recA1 endA1 araD139 D(ara, leu)7697 galU galK
l± rpsL nupG tonA] and BL21(DE3) [F± ompT hsdSB(rB±mB±) gal dcm (DE3)] were electroporated with either 50 ng pKSm4021 plasmid alone or 50 ng
plasmid, 1 mg T70C, and 2 mg NT30C (see Materials and Methods). Average kanamycin-resistant and ampicillin-resistant colony numbers and correction
ef®ciencies per 105 cells are presented.
The oligonucleotides used in this study are listed in
Figure 1B. KanNT30C(pm) and KanNT70C(pm) are complementary to the non-transcribed (NT) strand of plasmid
pKSm4021 and have lengths of 30 and 70 nt, respectively.
KanNT70G is also complementary to the non-transcribed
strand, except at position 4021 where a G´G mismatch is
formed with the target sequence. The second group of
oligonucleotide vectors hybridizes to the transcribed (T)
strand of pKSm4021 at the same site. These molecules differ
only in their length (30, 50 or 70 bases), but all create a C´C
mismatch with the target strand of pKSm4021. KanTF/31G
(pm) contains a ¯uorescent-Cy5 tag on the 5¢ end, is 31 bases
in length, and hybridizes with perfect complementarity (pm)
to the target site.
The experimental protocol involves the pairing of two
oligonucleotides with pKSm4021 in a reaction catalyzed by
the RecA protein. RecA protein has the capacity to transfer
oligonucleotides into the plasmid at the homologous site,
creating a structure known as a D-loop or, in our case, a
double-D-loop. Some of the oligonucleotides used in these
experiments align in perfect register with either the `transcribed or non-transcribed' strand of pKSm4021, while others
form a single mismatched base pair located in the center of the
conjoined molecules. The structure of a D-loop or a double-Dloop is represented in Figure 1C. We use the terminology
`transcribed' (T) or `non-transcribed' (NT) for convenience
simply to distinguish the strands of plasmid pKSm4021.
Once the complexes are assembled, they can be separated
from excess single-stranded molecules by using a chromaspin
column (see Materials and Methods). The isolated complexes
can be electroporated directly into DH1OB (recA±) E.coli cells
to measure their competency for nucleotide repair. Since this
strain has a disrupted RECA gene, which inactivates RecA
function, the majority of DNA pairing events must, therefore,
occur outside the cell. After electroporation of the preformed
joint molecule, the repair function is provided by the bacterial
cell. The conversion events can be monitored by growth on
agar plates containing either kanamycin or ampicillin. The
correction ef®ciency of each reaction is calculated by dividing
the number of kanr colonies by the number of ampr colonies,
the latter value normalizing for differences in electroporation
ef®ciency. Thus, we use an experimental protocol in which the
DNA pairing phase of this reaction occurs outside the cell,
while the repair of the point mutation in pKSm4021 takes place
in the cell. Such a scenario addresses the need for DNA pairing
in TNE because, in effect, electroporation of preformed, `prepaired' substrates should rescue a recA± strain for the TNE
reaction.
The importance of RecA protein in TNE was demonstrated
directly in the following experiment. Two bacterial strainsÐ
DH10B, which lacks functional RecA protein, and
BL21(DE3), which has normal RecA protein activityÐwere
electroporated with plasmid pKSm4021, which was maintained under ampicillin selection. Subsequently, oligonucleotides KanT70C and KanNT30C were co-electroporated into
each strain, respectively, and the correction of pKSm4021
monitored by the appearance of Kanr colonies. As shown in
Table 1, both E.coli strains support the repair of pKSm4021,
but at signi®cantly different levels. In DH10B cells, TNE
occurs at a low level, while in contrast, correction frequency in
BL21 is 0.89. The reduction in activity observed in DH10B
cells is attributable, in all likelihood, to the absence of active
RecA protein. This observation is consistent with the data of
Holloman and Radding (1) and forms the basis for the
implication that DNA pairing function is an important
requirement for high levels of TNE. A low level of correction
did occur in the DH10B cells, indicating that other
recombination activity(s) may promote TNE in the absence
of RecA function. Such activities could include, but would not
be limited to, strand annealing or DNA condensation, which
could juxtapose DNA sequences that could serve as repair
templates.
To test this hypothesis directly, however, we established an
experimental protocol in which double-D-loops were constructed with the same plasmid and oligonucleotides prior to
electroporation using puri®ed RecA protein to supply the
DNA pairing function ex vivo.
In the ®rst iteration, double-D-loops were formed using an
unlabeled incoming oligonucleotide and a 32P-labeled annealing oligonucleotide. As shown in Figure 2A, double-D-loops
are visualized by the change in position of the 32P-labeled
annealing oligonucleotide. The position of the 32P-labeled
annealing oligonucleotide is altered from a fast moving species
to one that co-migrates with the superhelical DNA once
incorporated into the plasmid. In our reaction protocol, the
annealing oligonucleotide will hybridize to the complement
strand of the D-loop only if RecA protein remains associated
with the initial joint molecule consisting of the plasmid DNA
and the incoming oligonucleotide (this fact is con®rmed by
lane 7 below). In the experiment presented in Figure 2A,
KanNT70G was used as the incoming oligonucleotide and
either 32P-labeled KanT30C or 32P-labeled KanT50C was used
as the annealing oligonucleotide. Under these reaction conditions, double-D-loop formation is dependent on the presence
of RecA protein and both DNA substrates (Fig. 2A, lanes 1 and
2). In this case, double-D-loops bear two mismatched base
Nucleic Acids Research, 2003, Vol. 31, No. 3
pairs, one created by the non-transcribed strand with
KanNT70G and one created by the transcribed strand with
KanT30C or KanT50C. Figure 2A, lanes 3, 4 (30mer), 5 and 6
(50mer), respectively, demonstrates the ef®ciency of separating the double-D-loop complexes from excess, unbound,
903
32P-labeled oligonucleotides by using the chromaspin column
technique (see Materials and Methods). Of particular interest is
the observation that a signi®cant percentage of stable doubleD-loops were recovered due, in all likelihood, to the enhanced
level of complex stability engendered by the 50mer. We
904
Nucleic Acids Research, 2003, Vol. 31, No. 3
observed (Fig. 2A) that a higher level of double-D-loops are
isolated when the annealing oligonucleotide has a length of 50
bases, rather than 30 bases. Lane 7 illustrates the results of a
reaction in which the 30mer (KanNT30C) was used as the
initial or incoming oligonucleotide. No double-D-loops are
formed, suggesting that the 30mer cannot enter the complex or
does not remain in the complex as the incoming oligomer, but
it can assimilate if the 70mer and the plasmid DNA have
combined to form the initial D-loop structure.
To show that the annealing oligonucleotide co-migrates
with supercoiled plasmid DNA after agarose gel electrophoresis, we carried out a double-D-loop reaction using a
¯uorescently tagged, single-stranded oligonucleotide.
KanTF/31G, which contains a Cy5 ¯uorescent moiety
attached to the 5¢ end, was used as the annealing oligonucleotide. This molecule was synthesized with this ¯uorescent tag
so that its location in an agarose gel could be detected at a
wavelength of 670 nm. Lane 1 contains a 1 kb ladder used as a
double-stranded molecule. By forming double-D-loops and
removing excess oligonucleotide, the presence of oligonucleotide can be identi®ed because of Cy5 emission at
670 nm (Fig. 2B, lane 2). When the same gel is stained with
SyberGreen, which demarcates the position of doublestranded DNA (pKSm4021) with emission at 526 nm, a
green ¯uorescent band is visualized not only on the 1 kb DNA
ladder (lane 1*), but also at the position of superhelical
plasmid DNA (lane 2*). Using ImageQuant software, the two
scans of the gel can be overlaid, identifying the co-migration
of oligonucleotide vector and plasmid pKSm4021 (lane 2+).
To more accurately de®ne the oligonucleotide components
of the double-D-loop, we used two new oligonucleotides
modi®ed at the 5¢ end. KanNTF70C(pm) is analogous to
KanNT70C (pm), except it contains a Cy5 ¯uorescent tag on
its 5¢ end (see Fig. 1B). KanTF/31C is the same as KanTF/
31G, except that it has a Cy3 ¯uorescent tag on its 5¢ end
instead of a Cy5 tag. It also creates a C´C mismatch instead of
being perfectly matched to the transcribed strand (Fig. 2C).
We used KanNTF70C and KanTF/31C in double-D-loop
reactions, following each oligonucleotide by its unique
¯uorescent label. As shown in Figure 2C, double-D-loop
formation is dependent on the presence of RecA protein (lane
1), plasmid DNA (lane 2) and incoming oligonucleotide (lane
3). In these three lanes, no detectable ¯uorescence is observed.
Lane 4 represents a reaction in which KanNTF70C and
KanTF/31C are both present and, as seen in the far left panel,
green ¯uorescence indicates the presence of Cy3-labeled
KanTF/31C oligonucleotides. The center panel (lane 4)
con®rms the presence of the Cy5-labeled KanNTF70C (red).
Data in the far right panel con®rm that both oligonucleotides
co-localize (lane 4, yellow band). Lanes 5 and 6 represent
reaction mixtures in which either KanTF/31C or KanNTF70C
is used in conjunction with a partner oligonucleotide that does
not contain a ¯uorescent tag (see Fig. 1B). Taken together, the
data establish the presence of both oligonucleotides in the
double-D-loop complex.
Complexes from Figure 2A were electroporated into E.coli
and the correction ef®ciency was determined by selection on
plates containing either ampicillin or kanamycin. As shown in
Nucleic Acids Research, 2003, Vol. 31, No. 3
Figure 2D, kanamycin-resistant colonies were observed when
the initial mixtures contained all the appropriate reaction
components. The double-D-loop reactions containing
KanT70C and KanNT30C were tested for correction activity
in the isolated (chromaspin) and non-isolated forms. Both
exhibited gene repair activity at approximately the same level
indicating that, in all likelihood, the corrected plasmids
conferring antibiotic resistance are likely to be those that
contained preformed double-D-loops. The combination of
KanNT70G/KanT50C exhibited a higher level of nucleotide
905
exchange than the combination of KanT70C/KanNT30C,
perhaps re¯ecting the increased percentage (40% higher) of
isolated double-D-loops from the combination of KanNT70G
and KanT50C with passage through a chromaspin column.
This increase may indicate that the complex bearing
KanNT70G and KanT50C is more amenable to the repair
reaction because the half-life of the whole complex within the
cell is increased. Another possibility for the higher level of
correction observed with the KanNT70G/KanT50C combination is that two mismatched base pairs are created when these
906
Nucleic Acids Research, 2003, Vol. 31, No. 3
oligonucleotides hybridize to the target. Finally, the puri®ed
complexes from lane 6 (Fig. 2A) are more active than the
unpuri®ed complexes, represented in lane 5 (Fig. 2A), and
thus again, the data suggest that the corrected kanamycin
genes are likely to have arisen from the preformed double-Dloop complexes.
Colonies exhibiting kanamycin resistance as a result of the
action of KanNT70G/KanNT50C were processed for DNA
sequence analyses (Fig. 2E). Isolated plasmid molecules from
these colonies were found to contain the converted TAC
codon rather than the TAG codon. In addition, no TAT codons
were found, validating the assay and suggesting that
kanamycin resistance arose from a DNA repair process and
not from wild-type plasmid contamination. Resistant colonies
generated by the electroporation of the 70/30mer oligonucleotide combination showed the same DNA sequence results
(data not shown). These results suggest that puri®ed double-Dloops can serve as templates for the gene repair reaction and
that the speci®c nucleotide base targeted for repair is, in fact,
altered.
In the next series of experiments, we compared the
ef®ciency of repair generated from double-D-loop complexes
in which the 70mer or the 30mer served as the oligonucleotide
that created the single mismatch with the transcribed (T)
strand. Preformed double-D-loops created through the combinations of KanT70C/KanNT30C or KanNT70C/KanT30C
were isolated and electroporated into DH10B cells for genetic
readout. When KanT70C and KanNT30C are used, a single
mismatch (C´C) is made between the incoming 70mer and the
`transcribed' strand of pKSm4021 (Fig. 3A). While the overall
percentage of 32P-labeled oligonucleotide transferred into the
complex by the action of RecA protein is estimated to be
4.5%, relative to the amount of labeled excess oligo (Fig. 3B,
lane 3), half of the preformed complexes were successfully
isolated by the chromaspin technique (Fig. 3B, lane 4). When
electroporated into E.coli, this combination of oligonucleotides produces a correction frequency of 15.23. Under these
reaction conditions, plasmid DNA is limiting and it is likely
that most of it will be assembled into joint complexes (see
Fig. 4). When KanNT70C and KanT30C are used to preform
the double-D-loop, the incoming oligonucleotide, KanNT70C,
hybridizes completely to the non-transcribed strand of
pKSm4021 and KanT30C forms a C´C mismatch to the
transcribed strand (Fig. 3A, lanes 5 and 6). After electroporation into DH10B cells, the isolated molecules give rise to a
correction ef®ciency of 16.10. Since the two values (15.23 and
16.10) are similar, we conclude that the length of the
oligonucleotide creating the mismatched base pair with the
transcribed strand of pKSm4021 has little impact on the
overall correction frequency.
A series of controls in which several reaction conditions
were altered either before or after the formation of the doubleD-loop were carried out. For most experiments, KanT70C and
KanNT30C were used as incoming and annealing oligonucleotides, respectively. As shown in Figure 4A, the
formation of double-D-loops does not occur when either
plasmid DNA or RecA protein is excluded from the reaction
(lanes 1 and 2), whereas a complete reaction mixture results in
double-D-loop formation (lane 4). In one modi®ed reaction,
isolated double-D-loop complexes were heated to 89°C for
5 min and, as shown in Figure 4A (lane 3), the population of
joint molecules was no longer observed. Figure 4B illustrates
the correction ef®ciencies of samples 1±4 and, consistent with
earlier data, only sample 4 produces a signi®cant correction
frequency. The heat-treated sample does not give rise to kanr
colonies after electroporation into DH10B cells. Furthermore,
in a reaction in which pKSm4021, KanT70C and KanNT30C
were electroporated simultaneously without the in vitro pairing step (sample 5), no detectable correction events took place,
as evidenced by the correction ef®ciency value of 0.0. The
same results were observed when only the plasmid was
electroporated into E.coli (sample 6), and align with the results
presented in Figure 2. When a non-homologous 74mer
(Hyg3S/74NT) (13) replaced KanT70C as the incoming
oligonucleotide, no gene repair of pKSm4021 was found
(lane 7). The formation of D-loops, as opposed to double-Dloops, only produced a low correction ef®ciency of 0.53 (lane
9). No correction ef®ciencies were observed when using a
Figure 2. (Previous three pages) Length effect of annealing oligonucleotide in TNE. (A) Double-D-loops were formed by adding either a 30mer (KanT30C)
(lanes 3 and 4) or 50mer (KanT50C) (lanes 5 and 6) as the annealing oligonucleotide after the initial synaptic step in which the incoming oligonucleotide
(KanNT70G) is paired to plasmid pKanSm4021. The presence of double-D-loops are visualized by the migration of the 32P-labeled oligonucleotide on an
agarose gel. An amount of 5 ml joint complex sample is loaded on a 1% agarose gel with the addition of 13 loading dye (0.25% bromophenol blue, 0.25%
xylene cyanol, 25% ®coll), and electrophoresed at 97 V for 2 h at room temperature. No double-D-loops are formed when a 30mer is used as an incoming
oligonucleotide and a longer 70mer as the annealing oligonucleotide (lane 7). (B) Double-D-loops were formed using KanTF/31G, a Cy5-labeled annealing
oligonucleotide perfectly matched to the target strand (lanes 2, 2* and 2+). To ensure formation of a double-D-loop complex, the gel was stained with
SyberGreen to monitor the presence of the double-stranded plasmid target. Lanes 1, 1* and 1+ represent a DNA ladder with no base pair periodicity.
(C) Double-D-loops were formed using both Cy5-labeled and Cy3-labeled oligonucleotides. Additional oligonucleotides were used in conjunction with those
listed in Figure 1B. Kan NTF70C contains a Cy5 tag at its 5¢ end and is perfectly matched to the non-transcribed strand of the plasmid. KanTF/31C contains
a Cy3 tag on its 5¢ end and a C´C mismatch with the plasmid. The following combinations were tested in the following reactions: no RecA protein (lane 1),
no pKanSm4021 (lane 2), no incoming oligonucleotide, NTF70C (lane 3), NTF70C (Cy5) + TF31C (Cy3) (lane 4), NTF70C (Cy5) + T30C (lane 5) and
NT70C + TF31C (Cy3) (lane 6). The presence of Cy3-labeled oligonucleotides in double-D-loops are shown in green when visualized at 580 nm, whereas
Cy5-labeled oligonucleotides are red when visualized at 670 nm. When both oligonucleotides are incorporated into double-D-loops, yellow bands appear in
the gel. (D) The bar graph indicates the correction ef®ciency using double-D-loops containing either 30 or 50mer as the annealing oligonucleotide. After
monitoring double-D-loop complex formation on an agarose gel, an additional 5 ml sample was transfected into DH10B competent cells. The designation
`chromaspin' indicates that these double-D-loops were isolated by passage through a chromaspin column, removing excess oligonucleotide. The correction
ef®ciencies were determined by normalizing the number of kanamycin-resistant colonies to the number of ampicillin-resistant colonies. (E) Kanamycinresistant colonies were picked for colony PCR, and the PCR products were sequenced as described in Materials and Methods. Wild-type pKSm4021 has a
sequence of TAT, while the mutant pKSm4021 plasmid contains a sequence of TAG at the target position. All kanamycin-resistant colonies tested were found
to have a TAC sequence, exhibiting a perfect replacement of a G base at position 4021. The standard deviation for these experiments and others that follow
was calculated using data from four independent experiments.
Nucleic Acids Research, 2003, Vol. 31, No. 3
907
Figure 3. Alternate mismatches in double-D-loop intermediates. (A) Joint molecules were formed under standard conditions so that either the incoming or
annealing oligonucleotide created the mismatch with the target base; in either case, the mismatch targeted the transcribed strand of plasmid pKSm4021.
(B) Correction ef®ciency was measured in samples lacking plasmid or RecA protein (lanes 1 and 2). The formation of joint double-D-loop complexes were
quanti®ed and compared with the correction ef®ciencies for samples 4 and 6. The amount of correction is similar, despite the mismatch location change from
the long incoming oligonucleotide to the short annealing oligonucleotide. KanT70C* and KaNT30C complexes observed a correction ef®ciency of 15.23 6
3.45, whereas KanNT70C and KanT30C* observed a correction ef®ciency of 16.10 6 0.76 after excess oligonucleotide removal. The amount of complex in
these samples transfected into the competent cells was also similarÐ1.64 and 1.2%, respectively.
D-loop formed with a 30mer oligonucleotide. The results of all
of these control experiments suggest that a DNA pairing event
creating a double-D-loop reaction intermediate is a critical
step in the pathway of TNE.
DISCUSSION
We have used synthetic oligonucleotides to correct a mutation
in the kanamycin gene contained in plasmid pKSm4021. Using
the DNA repair pathways in E.coli, the gene is corrected and
its expression enables the bacteria to grow on plates laden with
kanamycin. A simple assay system (11) in which the plasmid
and the oligonucleotides are introduced into E.coli at various
times and in various combinations facilitates an evaluation of
reaction parameters. In this study, we examine how the
preformation of joint molecules comprised of a plasmid and
oligonucleotides in¯uences the repair of the kanamycin
mutation. More speci®cally, we focus on the creation of
double-D-loopsÐa conjoined molecule in which oligonucleotides hybridize to opposite strands of the target helix forming a
complement-stabilized structureÐand its capacity to serve as
a template for gene repair. Our results indicate the following:
®rst, the creation of double-D-loops through a DNA pairing
reaction greatly enhances the number of kanamycin-resistant
bacterial cells arising from the repair activity in E.coli;
secondly, the frequency of repair is elevated when the length
of the oligonucleotides bound within the double-D-loop
complex is increased; and thirdly, either oligonucleotide can
serve as the molecule forming the mismatch with the target,
but the highest level of correction is attained when both
908
Nucleic Acids Research, 2003, Vol. 31, No. 3
Figure 4. Parameters of the reaction. (A) The two oligonucleotides, KanT70C and KanNT30C, were used to study joint complex formation in the absence of
either plasmid (lane 1) or RecA protein (lane 2). A stable double-D-loop formed only under standard reaction conditions (lane 4). In one case, the sample was
heated to 89°C for 5 min before loading on the gel (lane 3). (B) The same samples from the agarose gel in (A) were transfected into DH10B cells. Additional
samples were also transfected into DH10B cells (samples 5±9). Plasmid and oligonucleotides were added without RecA, ATPgS or Mg2+ (sample 5). Sample
6 contained only plasmid, and sample 7 contained a non-speci®c oligonucleotide with no complementarity to the kanamycin sequence. This oligonucleotide
was a 74mer (17,18) and was used in place of a 70mer oligonucleotide. D-loops were formed in samples 8 and 9 using KanT30C and KanT70C, respectively,
prior to electroporation in DH10B cells.
oligonucleotides form mismatches with both strands at the
speci®c site.
The success of the gene repair reaction depends largely on
the DNA pairing step, which, in our experimental protocol, is
catalyzed by the action of a RecA protein prior to electroporation into E.coli. The reaction is also dependent on each
oligonucleotide bearing sequence homology to the target
sequence, and the dissociation of the preformed complex
eliminates gene repair activity as measured by the appearance
of kanr colonies. Preformation of this complex requires a strict
order of addition in which the longer incoming oligonucleotide is ®rst assembled with RecA protein in the presence of
ATP-gS, followed by the addition of the shorter (annealing)
oligonucleotide. These reaction conditions insure that RecA
will not catalyze the reverse reaction and, as a side reaction,
anneal the two juxtaposed oligonucleotides, leading to the
disablement of the double-D-loop structure. We are aware,
however, that once the complex enters the cell, there is still a
possibility that the structure of the complex is modi®ed in
some fashion. But all of the controls presented argue against
this notion and in favor of the hypothesis that the pre-pairing,
preformation of double-D-loop complexes greatly enhances
the frequency of gene repair. This conclusion is also supported
by the fact that isolated/puri®ed double-D-loops are as
effective in producing corrected plasmids as double-D-loops
that have not been puri®ed.
The critical characteristic of the double-D-loop structure in
promoting high levels of gene repair is most likely its inherent
stability. This stability is enabled because both oligonucleotides are hybridized to their complementary strand at the target
Nucleic Acids Research, 2003, Vol. 31, No. 3
site. Previous data have shown that D-loops are kinetically
stable in superhelical plasmid DNA, even after the removal of
RecA proteins (17,19±21), and that this stability is due to the
slow dissociation step once the joint is formed. In these
studies, however, the incoming oligonucleotide is preassembled into a RecA ®lament, but our work establishes a
more simple strategy for double-D-loop formation in which
only the incoming molecule is pre-bound by RecA. Recent
data from Parekh-Olmedo et al. (22) indicate that these stable
double-D-loops, serving as intermediates in the gene repair
reaction, produce a higher level of correction compared with
correction levels attained using a single oligonucleotide.
Initialization of the gene repair reaction depends on DNA
pairing. In this scenario, the oligonucleotides are positioned
into homologous register with the target sequence with
subsequent cellular activities leading to the exchange of a
single nucleotide. Here we have emphasized the importance of
the pairing step in achieving substantial levels of gene repair.
And DNA pairing events that lead to the creation of a more
stable joint molecule (here, the double-D-loop) can increase
correction ef®ciency, as compared with TNE ef®ciencies
obtained without a pre-pairing phase. Similar observations
were made for single D-loop structures in the pioneering
studies of Holloman and Radding (1) and Holloman et al. (7).
This work was important not only for showing that genetic
information can be transferred from single-stranded DNA to
progeny, but also for establishing a transformation system in
E.coli that simulated the early steps of recombination (7). We
have used this system and extended their observations to
implicate the double-D-loop as a putative intermediate in
the gene repair reaction. The correction of the mismatched
base pairs in E.coli was originally reported by Razin
et al. (23), a reaction used extensively to biochemically
de®ne the components of the mismatch repair system
(reviewed in 24).
The DNA mismatch repair pathway corrects aberrant base
pairs created through the hybridization of the oligo(s) with the
plasmid sequence (11,12,25). While our objective in the
current study is not to elucidate more fully this pathway in the
gene repair reaction, we are aware that the majority of
correction events herein arise from a C´C base mismatch
formed in the double-D-loop. And it is widely accepted that
the C´C mismatch is corrected in the least ef®cient fashion,
due perhaps to the structural con®guration adopted by this
speci®c base mispair (26±30). Currently, we are investigating
the repair of other mismatched base pairs in order to establish
a `base repair hierarchy'. But it is also important to note that
the template upon which such repairs take place is dissimilar
to the one used to establish the current hierarchy. In our case,
the mismatch repair enzymes would encounter a double-Dloop con®guration rather than a double helix containing a
simple mismatch. The ef®ciency with which mismatches are
corrected is believed to be determined by structural properties
of individual mispairs, with those corrected most ef®ciently
forming a structure that induces a rigid deformation of the
helix. In contrast, those corrected more poorly deform the
helix into a more dynamic state based on cooperative
hydrogen binding and reduced interhelical stacking. The latter
conformation enables a higher intrinsic ¯exibility that allows
C/C mispairs, for example, to escape recognition by the
mismatch repair enzymes (26,31). Evidence has been put
909
forward indicating that double-D-loops containing heterologous inserts, perhaps a single mismatched base pair, adopt
an `anti-rotational lock' conformation (20). As such, there is a
strong potential that a quadruplex evolves primarily through
guanine base pairing, a structure will clearly produce a more
rigid conformation, which in turn would obviate the accepted
hierarchy. Future studies focused on the DNA conformation of
the double-D-loop will hopefully shed more light on this
possibility, but, regardless of its structural characteristics, we
now establish the double-D-loop as an active template upon
which gene repair can take place. Furthermore, the pathway
taken to form this intermediate depends on DNA pairing
activities.
ACKNOWLEDGEMENTS
We are grateful to members of the Kmiec laboratory for
comments during the course of the work. We acknowledge the
continuing efforts of Ms Elizabeth Feather in manuscript
preparation and Mr Eric Roberts in graphics. This work was
supported by NIH grant RO1 DK56134 and NIH training grant
T32 GM-08550.
REFERENCES
1. Holloman,W.K. and Radding,C.M. (1976) Recombination promoted by
superhelical DNA and the recA gene of Escherichia coli. Proc. Natl
Acad. Sci. USA, 73, 3910±3914.
2. Hotchkiss,R.D. (1971) Toward a general theory of genetic recombination
in DNA. Adv. Genet., 16, 325±348.
3. Hotchkiss,R.D. (1974) Models of genetic recombination. Annu. Rev.
Microbiol., 28, 445±468.
4. Miller,R.C.,Jr (1975) Replication and molecular recombination of
T-phage. Annu. Rev. Microbiol., 29, 355±376.
5. Radding,C.M. (1973) Molecular mechanisms in genetic recombination.
Annu. Rev. Genet., 7, 87±111.
6. Lacks,S. (1966) Integration ef®ciency and genetic recombination in
pneumococcal transformation. Genetics, 53, 207±235.
7. Holloman,W.K., Wiegand,R., Hoessli,C. and Radding,C.M. (1975)
Uptake of homologous single-stranded fragments by superhelical DNA: a
possible mechanism for initiation of genetic recombination. Proc. Natl
Acad. Sci. USA, 72, 2394±2398.
8. Shibata,T., Ohtani,T., Iwabuchi,M. and Ando,T. (1982) D-loop cycle. A
circular reaction sequence which comprises formation and dissociation of
D-loops and inactivation and reactivation of superhelical closed circular
DNA promoted by recA protein of Escherichia coli. J. Biol. Chem., 257,
13981±13986.
9. Shibata,T., Ohtani,T., Chang,P.K. and Ando,T. (1982) Role of
superhelicity in homologous pairing of DNA molecules promoted by
Escherichia coli recA protein. J. Biol. Chem., 257, 370±376.
10. Brachman,E.E. and Kmiec,E.B. (2002) The `biased' evolution of targeted
gene repair. Curr. Opin. Mol. Ther., 4, 171±176.
11. Cole-Strauss,A., Gamper,H., Holloman,W.K., Munoz,M., Cheng,N. and
Kmiec,E.B. (1999) Targeted gene repair directed by the chimeric RNA/
DNA oligonucleotide in a mammalian cell-free extract. Nucleic Acids
Res., 27, 1323±1330.
12. Rice,M.C., Bruner,M., Czymmek,K. and Kmiec,E.B. (2001) In vitro and
in vivo nucleotide exchange directed by chimeric RNA/DNA
oligonucleotides in Saccharomyces cerevisae. Mol. Microbiol., 40,
857±868.
13. Liu,L., Rice,M.C. and Kmiec,E.B. (2001) In vivo gene repair of point and
frameshift mutations directed by chimeric RNA/DNA oligonucleotides
and modi®ed single-stranded oligonucleotides. Nucleic Acids Res., 29,
4238±4250.
14. Liu,L., Cheng,S., van Brabant,A.J. and Kmiec,E.B. (2002) Rad51p and
Rad54p, but not Rad52p, elevate gene repair in Saccharomyces
cerevisiae directed by modi®ed single-stranded oligonucleotide vectors.
Nucleic Acids Res., 31, 2742±2750.
910
Nucleic Acids Research, 2003, Vol. 31, No. 3
15. Liu,L., Rice,M.C., Drury,M., Cheng,S., Gamper,H. and Kmiec,E.B.
(2002) Strand bias in targeted gene repair is in¯uenced by transcriptional
activity. Mol. Cell. Biol., 22, 3852±3863.
16. Parekh-Olmedo,H., Krainc,D. and Kmiec,E.B. (2002) Targeted gene
repair and its application to neurodegenerative disorders. Neuron, 33,
495±498.
17. Sena,E.P. and Zarling,D.A. (1993) Targeting in linear DNA duplexes
with two complementary probe strands for hybrid stability. Nature
Genet., 3, 365±372.
18. Gamper,H.B., Parekh,H., Rice,M.C., Bruner,M., Youkey,H. and
Kmiec,E.B. (2000) The DNA strand of chimeric RNA/DNA
oligonucleotides can direct gene repair/conversion activity in mammalian
and plant cell-free extracts. Nucleic Acids Res., 28, 4332±4339.
19. Jayasena,V.K. and Johnston,B.H. (1993) Complement-stabilized D-loop.
RecA-catalyzed stable pairing of linear DNA molecules at internal sites.
J. Mol. Biol., 230, 1015±1024.
20. Belotserkovskii,B.P., Reddy,G. and Zarling,D.A. (1999) DNA hybrids
stabilized by heterologies. Biochemistry, 38, 10785±10792.
21. Belotserkovskii,B.P. and Zarling,D.A. (2002) Peptide nucleic acid (PNA)
facilitates multistranded hybrid formation between linear doublestranded DNA targets and RecA protein-coated complementary singlestranded DNA probes. Biochemistry, 41, 3686±3692.
22. Parekh-Olmedo,H., Drury,M. and Kmiec,E.B. (2002) Targeted
nucleotide exchange in Saccharomyces cerevisiae directed by short
oligonucleotides containing locked nucleic acids. Chem. Biol., 9,
1073±1084.
23. Razin,A., Hirose,T., Itakura,K. and Riggs,A.D. (1978) Ef®cient
correction of a mutation by use of chemically synthesized DNA. Proc.
Natl Acad. Sci. USA, 75, 4268±4270.
24. Marra,G. and Schar,P. (1999) Recognition of DNA alterations by the
mismatch repair system. Biochem. J., 338 (Pt 1), 1±13.
25. Rice,M.C., May,G.D., Kipp,P.B., Parekh,H. and Kmiec,E.B. (2000)
Genetic repair of mutations in plant cell-free extracts directed by speci®c
chimeric oligonucleotides. Plant Physiol., 123, 427±438.
26. Hunter,W.N., Brown,T., Anand,N.N. and Kennard,O. (1986) Structure of
an adenine-cytosine base pair in DNA and its implications for mismatch
repair. Nature, 320, 552±555.
27. Cornelis,A.G., Haasnoot,J.H., den Hartog,J.F., de Rooij,M.,
van Boom,J.H. and Cornelis,A. (1979) Local destabilisation of a DNA
double helix by a T±T wobble pair. Nature, 281, 235±236.
28. Ho,P.S., Frederick,C.A., Quigley,G.J., van der Marel,G.A.,
van Boom,J.H., Wang,A.H. and Rich,A. (1985) G.T wobble base-pairing
Ê atomic resolution: the crystal structure of
in Z-DNA at 1.0 A
d(CGCGTG). EMBO J., 4, 3617±3623.
29. Brown,T., Hunter,W.N., Kneale,G. and Kennard,O. (1986) Molecular
structure of the G.A base pair in DNA and its implications for the
mechanism of transversion mutations. Proc. Natl Acad. Sci. USA, 83,
2402±2406.
30. Hare,D., Shapiro,L. and Patel,D.J. (1986) Extrahelical adenosine
stacks into right-handed DNA: solution conformation of the d(C-GC-A-G-A-G-C-T-C-G-C-G) duplex deduced from distance geometry
analysis of nuclear Overhauser effect spectra. Biochemistry, 25,
7456±7464.
31. Werntges,H., Steger,G., Riesner,D. and Fritz,H.J. (1986) Mismatches in
DNA double strands: thermodynamic parameters and their correlation to
repair ef®ciencies. Nucleic Acids Res., 14, 3773±3790.