Human Rad51 filaments on double- and
single-stranded DNA: correlating regular and
irregular forms with recombination function
Dejan Ristic1, Mauro Modesti1, Thijn van der Heijden2, John van Noort2,
Cees Dekker2, Roland Kanaar1,3and Claire Wyman1,3,*
1Department of Cell Biology and Genetics, Erasmus Medical Center, PO Box 1738, 3000 DR Rotterdam,
The Netherlands,2Kavli Institute of Nanoscience, Delft University of Technology, Lorentzweg 1,
2628 CJ Delft, The Netherlands and3Department of Radiation Oncology, Erasmus Medical Center-Daniel,
Rotterdam, The Netherlands
Received April 22, 2005; Accepted May 19, 2005
Recombinase proteins assembled into helical fila-
ments on DNA are believed to be the catalytic core
of homologous recombination. The assembly, disas-
must drive the DNA strand exchange reactions of
homologous recombination. The sensitivity of euka-
ryotic recombinase activity to reaction conditions
in vitro suggests that the status of bound nucleotide
cofactors is important for function and possibly for
filament structure. We analyzed nucleoprotein fila-
ments formed by the human recombinase Rad51 in a
variety of conditions on double-stranded and single-
filaments with extended double-stranded DNA cor-
related with active in vitro recombination, possibly
due to stabilizing the DNA products of these assays.
Though filaments formed readily on single-stranded
DNA, they were very rarely regular structures. The
irregular structure of filaments on single-stranded
DNA suggests that Rad51 monomers are dynamic
in filaments and that regular filaments are transient.
Indeed, single molecule force spectroscopy of Rad51
filament assembly and disassembly in magnetic
tweezers revealed protein association and disassoci-
ation from many points along the DNA, with kinetics
different from those of RecA. The dynamic rearrange-
ments of proteins and DNA within Rad51 nucleo-
protein filaments could be key events driving strand
exchange in homologous recombination.
Homologous recombination, the exchange of strands between
homologous DNA molecules, is a universal aspect of genome
metabolism needed to repair DNA double-strand breaks, to
ensure proper replication and chromosome segregation and
to create genetic diversity. The defining mechanistic steps of
homologous recombination are DNA strand invasion and joint
molecule formation. These reactions are catalyzed by a class
of proteins called recombinases, typified by bacterial RecA,
and including the RadA homologs in archaea and the Rad51
homologs in eukaryotes. The recombinase proteins all
assemble into helical nucleoprotein filaments on DNA. The
nucleoprotein filaments formed by bacterial RecA were first
observed more than two decades ago (1–3). Since then, this
filament structure has proven to be highly conserved.
Recombinases from all three kingdoms of life assemble on
both single-stranded and double-stranded DNA into very
similar filaments despite the limited conservation among
their amino acid sequences [structures reviewed in (4)]. Cur-
rent models of homologous recombination involve strand
exchange occurring within the recombinase nucleoprotein
filament. The mechanistic details of homology search and
the exchange of DNA strands are currently obscure, but must
require dynamic rearrangements of the nucleoprotein filament
involving both DNA and the recombinase proteins.
The in vitro activity of the eukaryotic recombinases is sens-
itive to a number of reaction conditions suggested to affect
protein or filament conformationrelated toATP binding (5–8).
The status of bound nucleotide cofactors influences DNA
binding for both RecA and Rad51, though in subtly different
ways for the recombinases from different organisms (9–14).
The formation of a recombination-competent nucleoprotein
filament in all cases requires a bound nucleotide cofactor.
*To whom correspondence should be addressed. Tel: +31 10 408 8337; Fax: +31 10 408 9468; Email: firstname.lastname@example.org
John van Noort, Department of Biophysics, Huygens Laboratory, Leiden University, 2300 RA Leiden, The Netherlands
? The Author 2005. Published by Oxford University Press. All rights reserved.
The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the open access
version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal and Oxford University Press
only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact email@example.com
Nucleic Acids Research, 2005, Vol. 33, No. 10
Conversely, ATP hydrolysis by the recombinases is dependent
on their interaction with DNA, with single-stranded DNA
stimulating the activity more strongly than double-stranded
DNA (15,16). However, the mechanistic role of ATP hydro-
lysis in recombination remains enigmatic. ATP hydrolysis is
dispensable for RecA and Rad51 catalyzed in vitro recomb-
ination reactions, which assess the pre-synaptic and synaptic
stages of recombination, including homology search, strand
invasion and joint molecule formation (7,17,18). However,
based on the phenotypes of ATPase defective recombinases,
in vivo recombination does require ATP hydrolysis (19–21).
Here, we described the effect of reaction conditions that
influence in vitro recombination on the structure of human
Rad51 nucleoprotein filaments. The filaments were observed
directly in the absence of fixatives and characterized with
scanning force microscopy (SFM). In conditions that enhance
in vitro recombination activity, we observed regular and stable
filaments with elongated double-stranded DNA. Filaments
formed on single-stranded DNA were almost all irregular in
all conditions unless they were fixed with glutaraldehyde.
In contrast, Rad51 filaments on double-stranded DNA were
irregular and apparently unstable in conditions where in vitro
recombination activity is low. Disassembly of filaments was
followed over time revealing that protein disassociationoccurs
from many places along the filament. Single molecule force
spectroscopy studies of Rad51 filament assembly and dis-
assembly also showed that stable filaments form in conditions
that favor in vitro strand exchange reactions. Furthermore,
the kinetics of filament disassembly followed in solution in
real-time indicated dissociation of Rad51 subunits from many
points at once.
MATERIALS AND METHODS
The double-stranded DNA used in the SFM experiments was
made by linearization of pDERI1 (22). Digestion of this plas-
mid with ScaI produced 1821 bp blunt-ended linear double-
stranded DNA. The resulting linear DNA was purified by
phenol:chloroform:iso-amyl alcohol (25:24:1) extraction and
checked for purity by gel electrophoresis.
A sample of 810 nt single-stranded DNA was made
as follows: first a double-stranded 810 bp fragment was pro-
duced by PCR, using the URA3 gene from Saccharomyces
cerevisiae as template DNA and primers U3, which was
and Bio 5, which was 50biotinylated (50-TTTCCCGGGGGG-
CCCGGGTTCTATACTGTTGACCC). The PCR product was
purified by phenol:chloroform:iso-amyl alcohol (25:24:1)
extraction, followed by ethanol precipitation. The DNA strand
with the terminal 50-phosphate was digested by lambda
exonuclease (5 U/mg DNA). The reaction was carried out at
37?C for 1 h and stopped by the addition of a 0.1 vol of STOP
solution containing 5% (w/v) SDS, 50 mM EDTA, 30 mg/ml
Proteinase K and further incubation at 65?C for 30 min.
The resulting single-stranded DNA was resolved on a 1.5%
agarose gel and purified from the gel using a GFX? column
DNA was made as described previously (23).
The human Rad51 protein was over-expressed in Escherichia
coli. Cells were lysed in high salt. The clarified lysate was
treated with polyethylenimine. After a second clarification,
Rad51 was recovered by (NH4)2SO4salting-out and the resus-
pended pellet was purified by heparin-sepharose chromato-
graphy followed by MonoQ chromatography. The protein
was dialyzed against 300 mM KCl, 20 HEPES–NaOH
(pH 7.8), 1 mM EDTA, 2 mM DTT and 10% glycerol, and
stored at ?80?C.
Filament formation reactions
Nucleoprotein filaments were formed in 10 ml reactions con-
taining 7.5 mM DNA (concentration in nucleotides), 2.5 mM
human Rad51, 25 mM HEPES–KOH (pH 7.5), 5 mM MgCl2
or CaCl2, 2 mM nucleotide cofactor (ATP, ADP, ATPgS or
AMP-PNP) and 30 mM KCl. Reactions were carried out at
37?C for 1 h and then placed on ice. When indicated, 200 mM
(NH4)2SO4was added after 1 h of incubation at 37?C which
was continued for an additional 10 min prior to placing the
reactions on ice.
Oligonucleotide SK3 (24) was radiolabeled at the 50end and
Tris–HCl (pH 7.5), 1 mM DTT, 0.1 mg/ml acetylated BSA,
60 mM KCl, 5 mM MgCl2(or CaCl2) and 2 mM ATP (or
AMP-PNP or ATPgS). After 5 min incubation at 37?C, 10 ml
of supercoiled pUC19 plasmid DNA at 1.14 mM nucleotide
(prepared by detergent lysis and twice purified by CsCl gra-
dient equilibrium sedimentation) was added. At the desired
time point, 10 ml aliquots were withdrawn from the reaction,
mixed with 5 ml of STOP solution (0.5% SDS, 50 mM EDTA
and 30% glycerol) and deproteinized by incubation for 5 min
at 37?C with 1 mg/ml final of Proteinase K. Reaction products
were fractionated by 0.8% agarose gel electrophoresis in Tris–
borate buffer. Gels were dried on DEAE paper and analyzed
Scanning force microscopy
For imaging, reactions were diluted 15-fold in deposition buf-
fer [10 mM HEPES–KOH (pH 7.5) and 10 mM MgCl2] and
deposited on freshly cleaved mica. After ?30 s, the mica was
washed with water (glass distilled; SIGMA) and exposed to a
stream of filtered air.
For the study of human Rad51 disassembly on mica, nuc-
leoprotein filaments were formed asdescribed above including
treatment with (NH4)2SO4. Reactions were deposited on mica
and the excess buffer was removed, but the samples were not
dried. The sample on mica was then covered with buffer con-
taining 25 mM HEPES–KOH (pH 7.5), 5 mM MgCl2, 2 mM
ATP and 30 mM KCl, or 200 mM (NH4)2SO4. The nucleo-
proteinfilaments were incubated at 19?C for 1, 5or 30 min and
then washed and dried as described above.
Images were obtained on a NanoScope IIIa or a NanoScope
IV (Digital Instruments; Santa Barbara, CA) operating in tap-
ping mode in air with a type E scanner. Silicon Nanotips were
purchased from Digital Instruments (Nanoprobes). The length
of nucleoprotein filaments was measured from NanoScope
images imported into IMAGE SXM 1.62 (National Institutes
Nucleic Acids Research, 2005, Vol. 33, No. 10 3293
of Health IMAGE version modified by Steve Barrett, Surface
Science Research Centre, University of Liverpool, Liverpool,
UK). The filaments’ contours were traced manually. For the
irregular filaments, the contours were traced as a path through
the highest points along the nucleoprotein filaments.
The magnetic tweezers setup used has been described in (25).
The force applied to the bead was calculated by quantifying
thermal motion of the DNA-tethered bead and substituting it
into equipartition theorem. Using image processing, a position
thermal drift, all positions were measured relative to a poly-
styrene bead fixed to the bottom of the flow cell. Length
changes of a magnitude up to 10 nm or with a frequency
<30 Hz could be determined reliably within the spatial and
temporal noise for Z detection at the force used.
Polystyrene beads, as well as DNA constructs carrying a
magnetic bead at one end, were anchored to the bottom of a
flow cell as described (25). Force-extension curves were used
to identify attachment of single DNA molecules. After con-
formation of the correct contour and persistence lengths,
experiments were started by addition of human Rad51. All
measurements were carried out at 25?C.
Flow cell reactions
The reaction buffer used in the flow cell was 25 mM HEPES–
KOH(pH 7.5), 5mM MgCl2and 25 mM KCl. For the start ofa
measurement, the flow cell content was replaced by buffer
including human Rad51 and 2 mM ATP while maintaining
the DNA tether at a constant magnetic force. Note however
that due to the flow (first 60–75 s), an additional drag force
pulls on the bead. As a result, the bead is moved sideways and
downwards, increasing the tether length, but decreasing the
bead height. Human Rad51 (dis)assembly was monitored
through measurement of the height of the magnetic bead.
The structure of nucleoprotein filaments formed by the human
recombinase Rad51 was analyzed by SFM imaging. In this
way, we not only assessed if the protein binds to DNA under
the conditions tested, but also directly observed the architec-
ture of the resulting protein–DNA complexes. Significantly,
sample preparation does not require fixative agents thereby
allowing the observation of possibly dynamic structures. This
approach also resulted in capturing for observation the variety
of structures present in a reaction mixture at the time of
deposition. Comparing unfixed and fixed filaments indicated
that chemical fixation forced the accumulation of regular
Human Rad51 formed regular and irregular filaments
on double-stranded DNA depending on reaction
Nucleoprotein filaments formed in the presence of ATP and
Mg2+on 1.8 kb double-stranded DNA were irregular and
varied in length (Figure 1A). To assess DNA extension in
the filaments, their contour length was measured by tracing
a path through the highest points along the protein–DNA
complex. This was compared to the contour length of DNA
molecules in the absence of protein measured in the same way.
All length measurements are displayed as distributions in
histograms (Figure 2). The length of the filaments formed
in the presence of ATP and Mg2+indicated that many of the
DNA molecules were not significantly extended in these
complexes (Figure 2A and B). The filaments with lengths
significantly shorter than the 1.8 kb DNA (average measured
length 0.53 mm) were due to contaminating short DNA frag-
ments that were eliminated in subsequent experiments. Fila-
ments formed in the presence of ATPgS, a nucleotide cofactor
analog usually described as poorly hydrolyzable, were similar
to those formed with ATP, irregular (Figure 1B) and without
Dramatically different filaments formed in the presence of
the non-hydrolyzable ATP analog AMP-PNP (Figure 1D).
Here, there were very few filaments with intermediate lengths
between that of the bare 1.8 kb fragment (0.53 mm) and the
fully extended length (0.8 mm), expected if the DNA is com-
pletely incorporated into a filament making it 1.5 times longer
(Figure 2D). These filaments were regular and, apart from
the filaments formed on the small contaminating DNA,
had a uniform length indicating extension of the DNA to
?1.5 times that of a B-form.
In addition to varying the nucleotide cofactor, we also
formed Rad51 filaments in buffer conditions known to influ-
ence the in vitro strand exchange reactions. Rad51-catalyzed
strand exchange is stimulated by the addition of ammon-
ium sulfate to a reaction including ATP and Mg2+(8) and
the (partial) substitution of Ca2+for Mg2+in an ATP-
containing reaction (7). Filaments were formed as described
above by addition of protein and nucleotide cofactor to DNA,
where ammonium sulfate was added subsequently. These con-
ditions, both addition of ammonium sulfate (Figure 1C) and
substitution of Ca2+for Mg2+(Figure 1E), resulted in filaments
with a regular appearance, similar to those formed in the
presence of AMP-PNP. Addition of ammonium sulfate resul-
ted in elongated filaments, though with a distribution centered
at shorter than fully extended length (Figure 2C). Careful
inspection of images of these filaments suggests their less
than complete extension may be due to less than complete
coverage of DNA by Rad51, but where DNA is covered the
structures are regular (see apparent gaps in filaments shown in
Figure 1C). Substitution of Ca2+for Mg2+resulted in filaments
with a length distribution centered near the fully extended
form (Figure 2E).
Human Rad51 rarely formed regular filaments on
We also analyzed the structure of Rad51 filaments formed on
single-stranded DNA,the complexthat initiates joint molecule
formation and is presumably active in identifying regions
of sequence homology. Rad51 was incubated with single-
stranded DNA, a natural plasmid sequence ?800 nt long, in
the same set of conditions and the same protein to DNA ratio
(equivalent here to one Rad51 monomer per 3 nt of DNA) as
was done for double-stranded DNA. In the absence of protein
cross-linking agents, regular filaments were only rarely
observed. Filaments were classified as regular or irregular
3294Nucleic Acids Research, 2005, Vol. 33, No. 10
based on their appearance and the percentage of regular and
irregular filaments in the various conditions was tabulated
(Figure 3). Regular or partially regular filaments were
observed when ATP in the presence of Ca2+or AMP-PNP
was used as a nucleotide cofactor. For AMP-PNP, only a
small fraction of the filaments were regular (85%, n = 60,
of the filaments were irregular, see example in Figure 3D).
Filaments formed in the presence of ATP and Ca2+were
occasionally like those formed on double-stranded DNA and
classified as regular (Figure 3E). However, most of them had
an intermediate appearance, not completely regular but not as
irregular asfilaments formed,forinstance, with ATPandMg2+
(compare Figure 3A and C with Figure 3E). All other condi-
tions resulted in irregular filaments. Notably, the addition of
ammonium sulfate to filaments formed with ATP and Mg2+
caused apparent disassociation of protein from DNA in the
irregular filaments present before addition of ammonium sul-
fate. However, treatment of the reactions producing irregular
filaments with glutaraldehyde resulted in all of the filaments
becoming regular (compare Figure 3A and B). This suggests
that the irregular structures observed on mica were a sampling
of dynamic filaments in solution with subunits rearranging
or associating and disassociating. Fixation by glutaraldehyde
may have trapped a structure that was only transiently present
or that was the sum of dynamic interactions that occur in the
time the cross-linking agent was present.
Human Rad51 recombination activity varied with
The structure of Rad51 filaments varied based on the DNA
substrate and the reaction conditions. The conditions used
were previously described to effect in vitro recombination
activity, but have not all been directly compared in the same
assay. In order to correlate filament structure with activity
directly, we tested all of these conditions with our Rad51
protein in a D-loop assay (Figure 4). As judged by D-loop
product accumulation, Rad51 is most active in the presence of
AMP-PNP or ATP and Ca2+. Indeed, in these conditions
product signal at the time labeled 0 (Figure 4) indicated
that product was formed in even the very short time it took
to remove an aliquot after mixing all components. Only weak
Figure 1. SFM images of filaments formed by human Rad51 on double-stranded DNA in the presence of various nucleotide cofactors. (A and B) Examples of
of regular filaments formed on double-stranded DNA in conditions including ATP and MgCl2treated with (NH4)2SO4(C), AMP-PNP (D) and ATP and CaCl2(E).
All images are 1 · 1 mm and height is represented by color in the range of 0–3 nm, red to yellow as shown in the scale bar.
Nucleic Acids Research, 2005, Vol. 33, No. 10 3295
activity was observed in the presence of ATP or ATPgS and
Mg2+. Little D-loop formation was detected when ammonium
sulfate was added to reactions with ATP and Mg2+. The Rad51
activity measured by D-loop formation does not require ATP
hydrolysis as this was most active with the non-hydrolyzable
analog AMP-PNP or with ATP and Ca2+, where ATP hydro-
lysis is considerably inhibited (7). Addition of ammonium
sulfate inhibited D-loop formation here, though it has been
shown to stimulate Rad51 activity in a different in vitro assay
[(8) and see Discussion]. The lack of D-loop product correl-
ated with the observation that Rad51 filaments formed on
single-stranded DNA in the presence of ATP and Mg2+
were disrupted by ammonium sulfate in our experiments.
Dynamic disassembly of human Rad51 filaments
correlates with ATP hydrolysis
ATP hydrolysis is required in vivo for Rad51 function in
homologous recombination (20). We observed that filaments
formed in the presence of ATP were very irregular.The irregu-
lar structure of these filaments suggested that they might be
dynamic, with active exchange of free and DNA-bound Rad51
along the filament. To test if this was the case, we character-
ized the transition from regular to irregular filaments. Ammo-
nium sulfate does not significantly inhibit the ATPase activity
of Rad51 [(14); and data not shown], but does result in the
formation of regular filaments. We therefore reasoned that
removing ammonium sulfate would provide an experimental
tool to control the transition from regular to irregular fila-
ments. For this purpose, filaments were formed with Rad51
and double-stranded DNA in conditions including ATP and
Mg2+, followed by addition of ammonium sulfate as before.
After deposition onto mica, excess buffer was removed and
replaced with a buffer lacking ammonium sulfate and Rad51.
The filaments deposited on mica were incubated in this buffer
for varying times before observation by SFM. As can be seen
in Figure 5A, incubation without ammonium sulfate for 1, 5
and 30 min resulted in increasing disorder along the filaments,
apparent loss of protein from DNA and eventually completely
protein-free DNA. Filament length also decreased over this
time indicating that DNA extension is not maintained as the
filaments become irregular (Figure 5B). In contrast to RecA
(26), Rad51 dissociation did not occur from one end, but
apparently from many sites along the DNA. The effect of
ammonium sulfate on stability of the Rad51 filaments was
specific and not a general effect of salt in the buffer. Substi-
tution of KCl for ammonium sulfate did not stabilize the fila-
ments, irregular filaments appeared just as in the absence of
ammonium sulfate or when it was removed without additional
salt (Figure 5D). Incubation of Rad51 filaments on mica per se
does not result in their dissociation, as regular filaments
remained stable in buffer including ammonium sulfate but
without Rad51 even after 30 min of incubation (Figure 5C).
ThoughRad51binds toDNA inthe presence ofADP,irregular
filaments were also formed when ADP was used as a cofactor
The experiments incubating filaments on mica suggested
that filaments disassembled over the course of ?30 min by
dissociation of Rad51 from many points. It is notable that the
Rad51 protein in filaments bound to mica appeared to behave
the same as it did in solution. The appearance of filaments on
Figure 2. Distribution of filament and DNA lengths measured from SFM
images. The length of DNA molecules or filaments in the indicated conditions
was measured and their distribution is plotted in histograms. (A) Bare DNA
without any bound protein. (B) Filaments formed by Rad51 in the presence of
ATP and MgCl2. (C) Filaments formed by Rad51 in the presence of ATP and
MgCl2and subsequently treated with (NH4)2SO4. (D) Filaments formed by
Rad51 in the presence of AMP-PNP. (E) Filaments formed by Rad51 in the
presence of ATP and CaCl2. The average measured length of the bare DNA
fragment (blue line) and the length of this DNA fragment if it is extended
1.5 times in a filament (green line) are indicated in the histograms.
3296 Nucleic Acids Research, 2005, Vol. 33, No. 10
mica after removal of ammonium sulfate was the same as
the appearance of filaments deposited from solution reactions
including ATP. However, it was not possible to follow these
apparent dynamics directly and we could not exclude possible
effects of the surface on this reaction and its apparent kinetics.
In order to study the dynamics without possible interference
from surface immobilization, we followed Rad51 filament
assembly and disassembly on single DNA molecules held
in magnetictweezers. An 8kb double-stranded DNA molecule
was tethered between a surface and a magnetic bead such that
the position of the bead and the force exerted upon it could be
manipulated and recorded. In the experiments reported, DNA
with at least one nick near the magnetic bead was used,
resulting in a torsionally unconstrained DNA molecule. At a
constant stretching force, enough to extend the DNA molecule
but not distort its structure, the assembly of a Rad51 filament
on this DNA will increase the length of the tethered molecule.
Assembly (or disassembly) of the filament was followed in
Figure 3. SFM images of filaments formed by human Rad51 on single-stranded DNA in the presence of various nucleotide cofactors. Examples of protein–DNA
complexes formed on single-stranded DNA in the presence of ATP and MgCl2deposited directly (A) and after fixation with glutaraldehyde (B). Examples of
filaments formed on single-stranded DNA in the presence of ATPgS (C), AMP-PNP (D), and ATP and CaCl2(E). The percentage of filaments appearing irregular
(suchasinAandC)orregularislistedbelowtheimages.Allimagesare1 · 1mmandheightisrepresentedbycolorintherangeof0–3nm,redtoyellowasshownin
the scale bar.
Nucleic Acids Research, 2005, Vol. 33, No. 103297
real-time by recording changes in the height of the magnetic
bead. Upon addition of buffer containing Rad51, ATP and
Mg2+, filament assembly was very fast. The dynamics of
this process could not be properly followed when 830 nM
Rad51 was added to the flow cell as the DNA tether was
already extended to 1.35 times the length of bare DNA when
the inflow of Rad51 was stopped (at ?75 s, Figure 6A). This
is equivalent to ?70% of the DNA covered by a filament
extending the helix 1.5timesits B-form lengththat hadformed
while flushing the flow cell. Note that the assembly rate could
have been affected by the extra extension of the double-
stranded DNA due to the drag of the flow. At lower levels
of Rad51, 166 nM added to the flow cell, the dynamics of
filament assembly could be separated from flushing the flow
cell. However, full extension to 1.5 times the original DNA
length was not observed in these conditions (Figure 6B). Very
similar time courses, fast assembly for 830 nM Rad51 and
slower assembly for 166 nM Rad51 were observed in several
independent experiments (3 and 11 respectively).
The non-linear elongation of DNA versus time in the pres-
ence of 166 nM Rad51 indicated that the rate-limiting step in
filament assembly was not a slow DNA binding or nucleation
event followed by more rapid filament growth (Figure 6B).
Rather it appeared that nucleation events of Rad51 binding
to DNA occurred at multiple sites, the frequency of which
depended on the concentration of free Rad51 (a complete
quantitative analysis of these and other tweezers experiments
will be published elsewhere). Filaments formed in these con-
ditions, continuous presence of Rad51 and ATP, remained
stable and elongated, within the detection limits of the tweez-
ers, for several hours. Addition of ammonium sulfate after the
formation of filaments did not affect their elongation. How-
ever, addition of ammonium sulfate together with Rad51 pre-
vented any obvious filament formation, consistent with SFM
imaging of reactions assembled in this way (data not shown).
As observed by SFM, elongated filaments were formed when
Rad51 was added to the flow cell together with AMP-PNP or
with ATP and Ca2+instead of Mg2+(data not shown).
Disassembly of a Rad51–DNA filament formed in the
presence of ATP was followed after replacing the buffer in
the flow cell with one that lacked ATP and Rad51. The bead
height, representing end-to-end distance of the tethered
molecule decreased over the course of 50 min to the original
length of bare DNA (Figure 6C). The exponential decay of this
curve indicated that filament dissociation did not occur from
one end or a single point, but that Rad51 dissociated from
many points along the molecule. As expected from the SFM
imaging experiments, Rad51 filaments treated with ammo-
nium sulfate or formed with ATP and Ca2+did not dissociate
whenthebuffer wasreplaced withonelacking Rad51andATP
(observed for several hours to over-night). These single
molecule dynamic measurements of filament disassembly in
solution confirmed the snap shot SFM images of filament
disassembly on mica.
We observe that the human recombinase Rad51 forms regular
stable filaments with extended double-stranded DNA in con-
ditions that promote an efficient in vitro strand exchange reac-
tion (7,8). In vitro assays for recombinase functions measure
the accumulation of DNA products after de-proteination,
thus the conditions that stabilize these double-stranded DNA
structures would enhance product accumulation. The regular
filament form appears to be trapped when ATP, or a structural
mimic, is bound but not hydrolyzed. AMP-PNP is a chemical
analog of ATP that can bind to but is not hydrolyzed by
ATPases such as Rad51. Substituting Ca2+for Mg2+in reac-
tions involving human Rad51 results in a reduced ATPase rate
that maintains human Rad51 in an ATP-bound form (7). The
effects of ammonium sulfate are more complex. It islikelythat
one effect of ammonium sulfate is to mimic an ATP-bound
form of Rad51. Although ammonium sulfate does not inhibit
the ATPase activity of human Rad51 [(14), and data not
shown], it probably does influence conformation at the
Figure 4. D-loop assay for recombination function. Human Rad51 was incubated with radiolabeled single-stranded oligonucleotide, supercoiled plasmid and the
followed in time as indicated.
3298 Nucleic Acids Research, 2005, Vol. 33, No. 10
ATPase active site. The atomic structures of both the archaeal
recombinase PfRad51 (27) and the yeast ScRad51 (28) have a
sulfate ion (presumably scavenged from the crystallization
solution), not a nucleotide, bound in their ATPase active
sites. Either the bound sulfate ion alone or in the presence
of ADP could induce the same protein conformation as bound
ATP (14). Additionally, it is expected that ammonium sulfate
has a more general role, as increased ionic strength prevents
aggregation of Rad51–DNA filaments in vitro. Accordingly,
incubation with KCl also prevents non-specific aggregation
but does not result in regular filaments [(5,8) and Figure 5C].
The importance of nucleotide cofactors for filament struc-
ture and stability is clear in the atomic structures of ScRad51
and Methanococcus volte RadA in filaments on DNA (28,29).
In these two structures, the ATPase active site isat the junction
of adjacent monomers in a filament. In the MvRadA filament
structure, the AMP-PNP nucleotide is clearly resolved as a
bridge between adjacent RadA monomers. It is easy to ima-
gine that hydrolysis of ATP would, in simple terms, break this
bridge or change the conformation of the adjacent recom-
binase monomers so that the subunit interface becomes less
The structural requirements for filaments on single-stranded
DNA that are needed to form joint molecules are often
assumed but have not yet been defined. We observed mostly
irregular filaments in all conditions on single-stranded DNA,
even in conditions that resulted in most active recombination.
Regular filaments have been reported by others in these
conditions (5,30), but only after fixation which could have
trapped otherwise dynamic protein in a regular structure.
Indeed after treatment of half of a reaction that produced
irregular filaments on single-stranded DNA with glutaralde-
hyde, exclusively regular filaments were observed (Figure 3A
and B). The irregular appearance of the filaments suggests
presented as histograms of filament lengths measured from the reactionsshown in (A). The length of the bare DNA fragment (blue line) and the length of this DNA
fragment if it is extended to 1.5 times its length in a filament (green line) are indicated in the histograms. (C–E) SFM images of control reactions showing the
are 1.5 · 1.5 mM in (A) and 1 · 1 mM in (C, D and E) and height is represented by color in the range of 0–3 nm, red to yellow.
Nucleic Acids Research, 2005, Vol. 33, No. 10 3299
dynamic association and disassociation of subunits. Our data
do not rule out regular filaments as the active species in joint
molecule formation, a dynamic transition to short-lived regu-
lar forms would explain their rarity. However, it is difficult to
imagine the strand exchange steps of homologous recombina-
tion occurring without dynamic rearrangement of protein and
DNA in the filaments.
The description we provide here of Rad51-DNA filament
structure and stability should form the basis for subsequent
investigation into the effect of the recombination mediators on
dynamic filament assembly and disassembly. The recomb-
ination reaction mediated by eukaryotic proteins that can be
completed in vitro by the bacterial recombinase RecA alone
requires, in addition to the recombinase, a host of additional
eukaryotic proteins, often called recombination mediators
(31,32). The suggested roles of the recombination mediators
involve promoting the formation of recombinase filaments and
stabilizing appropriately formed filaments on single-stranded
DNA. For example, the mediator BRCA2 promotes Rad51
binding to recombination-competent DNA structures (33).
Othermediators, the Rad51paralogs, aresuggested tostabilize
Rad51-DNA filaments since one of them, human XRCC2,
is an ADP/ATP exchange factor for Rad51 that would help
maintain Rad51 in an ATP-bound state (34). Rad54 can also
stabilize Rad51 filaments on single-stranded DNA (35–37).
The transition to unstable and disassembling filaments after
joint molecule formation is also a step where the recombina-
tion mediators could provide regulation and control.
We followed disassembly of filaments formed on double-
stranded DNA, and demonstrated that Rad51 filaments dis-
assembled in conditions that correlate with transition from an
ATP to an ADP-bound form of Rad51. Filament disassembly
followed on mica suggested that Rad51 monomers disassoci-
ate from DNA at multiple places. Quantitative analysis of
Rad51 filament disassembly from the single molecule force
spectroscopy experiments also indicated that Rad51 disasso-
ciation occurred from many places along the molecule. Rad51
apparently nucleates filament formation at many sites along
DNA with, we assume, independent initial orientation of
monomers probably resulting in filament segments with
opposite polarity running into each other. Our observation
of multiple disassembly points could be due to dissociation
of Rad51 monomers from internal sites of filaments. However,
this could also reflect monomer dissociation exclusively from
the ends of many short independently nucleated filaments.
Filament disassembly is likely to be as important as assem-
bly in authentic recombination reactions in vivo. In vitro
assays for recombinase functions measure the accumula-
tion of joint molecule intermediates or strand exchange pro-
ducts after de-proteination, which eliminates any differences
among the reactions with respect to disassociation of the
recombinase. However, recombination in vivo does not stop
at joint molecule formation or strand exchange. To complete
the process, other proteins need access to the joint molecule to
extend the invading strand by polymerization or extend the
heteroduplex by Holliday junction migration and eventually to
resolve the recombined molecules. A recombinase filament
covering the joint molecule is likely to inhibit these processes.
Indeed, a dynamic RecA filament, capable of hydrolyzing
ATP, is required for in vitro recombination coupled replica-
tion of E.coli (38). Similar replication coupled reactions are
essential recombination process in both prokaryotes and
Though our data do not yet address the function of Rad51
filament dynamics for identifying homology or strand
invasion, a comparison with another DNA–protein polymer
may be informative here. Though the ultimate result is a site-
specific, and not a general recombination event, some analo-
gies can be made between the phage Mu transposition reaction
and Rad51-mediated recombination, specifically with respect
to MuB function and activity. Similar to Rad51, MuB bound to
ATP forms polymers on DNA and ATP hydrolysis is coupled
to polymer disassembly (40,41). In addition, MuB monomers
readily exchange between polymers on DNA in both an
end-dependent and end-independent manner (40,42). The
MuB-DNA polymer functions to direct the site of phage
Mu insertion. It is suggested that the dynamic assembly and
disassembly on DNA effectively allow MuB to sample mul-
tiple DNA locations until a suitable one is found, a MuB-DNA
polymer is formed and remains stable enough for transposition
Figure 6. Assembly and disassembly of humanRad51 filamentson double-strandedDNA followed in real-time in a magnetictweezers setup. (A) Time course ofa
exchange occured for the first 60–75 s.
3300Nucleic Acids Research, 2005, Vol. 33, No. 10
to proceed (40). Similarly, dynamic Rad51-DNA interactions
could result in transiently regular filaments, which would
allow Rad51-bound single-stranded DNA to sample for
homologous regions in the double-stranded target DNA,
thus implying that a regular protein filament would stabilize
the formation of a proper duplex joint molecule, promoting
heteroduplex extension and favoring further steps of homo-
logous recombination. Formation of a regular protein filament
with extended DNA is a required step in current models of
homologous recombination. We do not dispute the importance
of helical nucleoprotein filaments in eukaryotic homologous
recombination, but suggest that important questions remain
regarding the functional lifetime of this structure as well as
the identity of the reaction step, pre-synaptic or post-synaptic,
This work was supported by grants from the Netherlands
Organization for Research FOM-ALW (to C.W., J.v.N., R.K.
and C.D.), the European Commission, the Association for
International Cancer Research, and the Dutch Cancer Society
(to R.K. and C.W.). Funding to pay the Open Access publica-
tion charges for this article was provided by the above grants.
Conflict of interest statement. None declared.
of complexes between recA protein and duplex DNA by electron
microscopy. J. Mol. Biol., 157, 87–103.
2. Dunn,K., Chrysogelos,S. and Griffith,J. (1982) Electron microscopic
visualization of RecA-DNA filaments: evidence for cyclic extension of
duplex DNA. Cell, 28, 757–765.
3. Flory,J. and Radding,C.M. (1982) Visualization of RecA protein and its
association with DNA: a priming effect of single-strand binding protein.
Cell, 28, 747–756.
4. Yu,X., VanLoock,M.S., Yang,S., Reese,J.T. and Egelman,E.H. (2004)
5. Liu,Y., Stasiak,A.Z., Masson,J.Y., McIlwraith,M.J., Stasiak,A. and
RAD51 protein. J. Mol. Biol., 337, 817–827.
6. Yu,X., Jacobs,S.A., West,S.C., Ogawa,T. and Egelman,E.H. (2001)
Domain structure and dynamics in the helical filaments formed by RecA
and Rad51 on DNA. Proc. Natl Acad. Sci. USA, 98, 8419–8424.
recombination protein Rad51 by modulating its ATPase activity.
Proc. Natl Acad. Sci. USA, 101, 9988–9993.
8. Sigurdsson,S., Trujillo,K., Song,B., Stratton,S. and Sung,P. (2001)
Basis for avid homologous DNA strand exchange by human Rad51 and
RPA. J. Biol. Chem., 276, 8798–8806.
9. Namsaraev,E.A. and Berg,P. (1998) Binding of Rad51p to DNA.
Interaction of Rad51p with single- and double-stranded DNA.
J. Biol. Chem., 273, 6177–6182.
10. Gupta,R.C., Bazemore,L.R., Golub,E.I. and Radding,C.M. (1997)
Activities of human recombination protein Rad51. Proc. Natl Acad. Sci.
USA, 94, 463–468.
11. Kim,H.K., Morimatsu,K., Norden,B., Ardhammar,M. and Takahashi,M.
(2002) ADP stabilizes the human Rad51-single stranded DNA complex
and promotes its DNA annealing activity. Genes Cells, 7, 1125–1134.
12. De Zutter,J.K. and Knight,K.L. (1999) The hRad51 and RecA proteins
show significant differences in cooperative binding to single-stranded
DNA. J. Mol. Biol., 293, 769–780.
binding properties of Saccharomyces cerevisiae Rad51 protein.
J. Biol. Chem., 274, 2907–2915.
14. Tombline,G. and Fishel,R. (2002) Biochemical characterization of the
human RAD51 protein. I. ATP hydrolysis. J. Biol. Chem., 277,
15. Cox,M.M. (2003) The bacterial RecA protein as a motor protein.
Annu. Rev. Microbiol., 57, 551–577.
16. Tombline,G., Heinen,C.D.,Shim,K.S. and Fishel,R. (2002) Biochemical
characterization of the human RAD51 protein. III. Modulation of
DNA binding by adenosine nucleotides. J. Biol. Chem., 277,
17. Rehrauer,W.M. and Kowalczykowski,S.C. (1993) Alteration of the
nucleoside triphosphate (NTP) catalytic domain within Escherichia coli
J. Biol. Chem., 268, 1292–1297.
18. Kowalczykowski,S.C. and Krupp,R.A. (1995) DNA-strand exchange
promoted by RecA protein in the absence of ATP: implications for the
mechanism of energy transduction in protein-promoted nucleic acid
transactions. Proc. Natl Acad. Sci. USA, 92, 3478–3482.
19. Symington,L.S. (2002) Role of RAD52 epistasis group genes in
homologous recombination and double-strand break repair.
Microbiol. Mol. Biol. Rev., 66, 630–670.
20. Stark,J.M., Hu,P., Pierce,A.J., Moynahan,M.E., Ellis,N. and
Jasin,M. (2002) ATP hydrolysis by mammalian RAD51 has a key
role during homology-directed DNA repair. J. Biol. Chem., 277,
21. Konola,J.T., Logan,K.M. and Knight,K.L. (1994) Functional
characterization of residues in the P-loop motif of the RecA protein ATP
binding site. J. Mol. Biol., 237, 20–34.
22. Ristic,D., Wyman,C., Paulusma,C. and Kanaar,R. (2001) The
architecture of the human Rad54-DNA complex provides evidence for
protein translocation along DNA. Proc. Natl Acad. Sci. USA,
23. van der Heijden,T., van Noort,J., van Leest,H., Kanaar,R., Wyman,C.,
dsDNA. Nucleic Acids Res., 33, 2099–2105.
24. Mazin,A.V., Zaitseva,E., Sung,P. and Kowalczykowski,S.C. (2000)
homologous pairing. EMBO J., 19, 1148–1156.
Dual architectural roles of HU: formation of flexible hinges and rigid
filaments. Proc. Natl Acad. Sci. USA, 101, 6969–6974.
26. Bork,J.M., Cox,M.M. and Inman,R.B. (2001) RecA protein filaments
disassemble in the 50to 30direction on single-stranded DNA.
J. Biol. Chem., 276, 45740–45743.
27. Shin,D.S., Pellegrini,L., Daniels,D.S., Yelent,B., Craig,L., Bates,D.,
Yu,D.S., Shivji,M.K., Hitomi,C., Arvai,A.S. et al. (2003) Full-length
and control by BRCA2. EMBO J., 22, 4566–4576.
28. Conway,A.B., Lynch,T.W., Zhang,Y., Fortin,G.S., Fung,C.W.,
Symington,L.S. and Rice,P.A. (2004) Crystal structure of a Rad51
filament. Nature Struct. Mol. Biol., 11, 791–796.
archaeal recombinase RADA: a snapshot of its extended conformation.
Mol. Cell, 15, 423–435.
30. Benson,F.E., Stasiak,A. and West,S.C. (1994) Purification and
EMBO J., 13, 5764–5771.
31. Sung,P., Krejci,L., Van Komen,S. and Sehorn,M.G. (2003) Rad51
recombinase and recombination mediators. J. Biol. Chem.,
32. Wyman,C.,Ristic,D.and Kanaar,R. (2004) Homologous recombination-
mediated double-strand break repair. DNA Repair (Amst.), 3, 827–833.
33. Yang,H., Li,Q., Fan,J., Holloman,W.K. and Pavletich,N.P. (2005)
The BRCA2 homologue Brh2 nucleates RAD51 filament formation at a
dsDNA–ssDNA junction. Nature, 433, 653–657.
hXRCC2 enhances ADP/ATP processing and strand exchange by
hRAD51. J. Biol. Chem., 279, 30385–30394.
35. Tan,T.L., Kanaar,R. and Wyman,C. (2003) Rad54, a Jack of all trades in
homologous recombination. DNA Repair (Amst.), 2, 787–794.
36. Mazin,A.V., Alexeev,A.A. and Kowalczykowski,S.C. (2003) A novel
function of Rad54 protein. Stabilization of the Rad51 nucleoprotein
filament. J. Biol. Chem., 278, 14029–14036.
37. Wolner,B. and Peterson,C.L. (2005) ATP-dependent and ATP-
independent roles for the RAD54 chromatin remodeling enzyme during
Nucleic Acids Research, 2005, Vol. 33, No. 103301
recombinational repair of a DNA double strand break. J. Biol. Chem., Download full-text
38. Xu,L. and Marians,K.J. (2002) A dynamic RecA filament permits
DNA polymerase-catalyzed extension of the invading strand in
recombination intermediates. J. Biol. Chem., 277,
Nature, 404, 37–41.
assembly and disassembly pathways of the MuB transposition target
complex. EMBO J., 21, 1477–1486.
41. Greene,E.C. and Mizuuchi,K. (2002) Direct observation of
single MuB polymers: evidence for a DNA-dependent conformational
42. Greene,E.C. and Mizuuchi,K. (2004) Visualizing the assembly and
disassembly mechanisms of the MuB transposition targeting complex. J.
Biol. Chem., 279, 16736–16743.
3302 Nucleic Acids Research, 2005, Vol. 33, No. 10