Single-molecule multiparameter fluorescence spectroscopy reveals directional MutS binding to mismatched bases in DNA.

Abstract

Mismatch repair (MMR) corrects replication errors such as mismatched bases and loops in DNA. The evolutionarily conserved dimeric MMR protein MutS recognizes mismatches by stacking a phenylalanine of one subunit against one base of the mismatched pair. In all crystal structures of G:T mismatch-bound MutS, phenylalanine is stacked against thymine. To explore whether these structures reflect directional mismatch recognition by MutS, we monitored the orientation of Escherichia coli MutS binding to mismatches by FRET and anisotropy with steady state, pre-steady state and single-molecule multiparameter fluorescence measurements in a solution. The results confirm that specifically bound MutS bends DNA at the mismatch. We found additional MutS-mismatch complexes with distinct conformations that may have functional relevance in MMR. The analysis of individual binding events reveal significant bias in MutS orientation on asymmetric mismatches (G:T versus T:G, A:C versus C:A), but not on symmetric mismatches (G:G). When MutS is blocked from binding a mismatch in the preferred orientation by positioning asymmetric mismatches near the ends of linear DNA substrates, its ability to authorize subsequent steps of MMR, such as MutH endonuclease activation, is almost abolished. These findings shed light on prerequisites for MutS interactions with other MMR proteins for repairing the appropriate DNA strand.

Full text

Page 1

Single-molecule multiparameter fluorescence
spectroscopy reveals directional MutS
binding to mismatched bases in DNA
Michele Cristo ´va ˜o1,2, Evangelos Sisamakis3,4, Manju M. Hingorani5, Andreas D. Marx1,
Caroline P. Jung1, Paul J. Rothwell3,*, Claus A. M. Seidel3,* and Peter Friedhoff1,*
1Institute for Biochemistry, FB 08, Justus Liebig University, Heinrich-Buff Ring 58, D-35392 Giessen, Germany,
2Department of Cell Biology and Genetics, Erasmus Medical Center, Dr. Molewaterplein 50, 3015 GE
Rotterdam, The Netherlands,3Molecular Physical Chemistry, Heinrich-Heine University, Universita ¨tsstrasse 1,
40225 Du ¨sseldorf, Germany,4Department of Applied Physics, Experimental Biomolecular Physics, Royal
Institute of Technology, SE-106 91 Stockholm, Sweden and5Molecular Biology and Biochemistry Department,
Wesleyan University, Middletown, CT 06459, USA
Received November 3, 2011; Revised January 20, 2012; Accepted January 23, 2012
ABSTRACT
Mismatch repair (MMR) corrects replication errors
such as mismatched bases and loops in DNA. The
evolutionarily conserved dimeric MMR protein MutS
recognizes mismatches by stacking a phenylalanine
of one subunit against one base of the mismatched
pair. In all crystal structures of G:T mismatch-bound
MutS, phenylalanine is stacked against thymine.
To explore whether these structures reflect direc-
tional mismatch recognition by MutS, we monitored
the orientation of Escherichia coli MutS binding
tomismatches by FRET
steady state, pre-steady state and single-molecule
multiparameter fluorescence measurements in a
solution.Theresultsconfirm
bound MutS bends DNA at the mismatch. We
found additional MutS–mismatch complexes with
distinct conformations that may have functional
relevance in MMR. The analysis of individual
binding events reveal significant bias in MutS orien-
tation on asymmetric mismatches (G:T versus T:G,
A:C versus C:A), but not on symmetric mismatches
(G:G). When MutS is blocked from binding a
mismatch in the preferred orientation by positioning
asymmetric mismatches near the ends of linear DNA
substrates, its ability to authorize subsequent steps
of MMR, such as MutH endonuclease activation, is
almost abolished. These findings shed light on pre-
requisites for MutS interactions with other MMR
proteins for repairing the appropriate DNA strand.
and anisotropywith
thatspecifically
INTRODUCTION
The DNA mismatch repair (MMR) system is responsible
for detecting and processing numerous discrepancies in
DNA, including base pair mismatches and insertion/
deletion loops arising from DNA polymerase errors or
during homeologous recombination, as well as DNA
damage lesions (1). The principal components of MMR,
MutS and MutL, are highly conserved proteins that
initiate and coordinate the appropriate DNA repair or
DNA damage signaling responses. MutS searches for
and recognizes mismatches and lesions in DNA, while
MutL serves as a molecular matchmaker (and endonucle-
ase in eukaryotes) to direct subsequent steps in the
reaction following mismatch/lesion recognition (2,3). In
Escherichia coli, MMR is directed to the erroneous
daughter strand by MutSL-dependent nicking by the
strand discrimination endonuclease MutH (4). MutS and
MutL are essential for maintaining genome integrity, as
illustrated by the fact that many mutations in these
proteins are causative of Lynch syndrome in humans
(5,6). Initiation of MMR is a complex and dynamic
process in which MutS scans vast lengths of duplex
DNA to detect relatively infrequent errors, and then gen-
erates an error-specific signal to activate MutL, which in
turn sets in motion excision of the incorrect DNA strand
and subsequent re-synthesis. MutS proteins possess two
key activities, the ability to bind DNA and distinguish a
variety of non-Watson–Crick structures and an ATPase
activity that modulates their interactions with DNA and
other proteins during initiation of MMR. In the E. coli
model system, MutS exists in a homo–dimer/tetramer
equilibrium, and the dimeric form of the protein is
*To whom correspondence should be addressed. Tel: +49 641 9935407; Fax: +49 641 9935409; Email: peter.friedhoff@chemie.bio.uni-giessen.de
Correspondence may also be addressed to Claus A. M. Seidel. Tel: +49 211 8115881; Fax: +49 211 8112803; Email: cseidel@hhu.de
Correspondence may also be addressed to Paul J. Rothwell Tel: +49211 8114715; Fax: +49 211 8112803; Email: p.rothwell@mail.cryst.bbk.ac.uk
5448–5464
doi:10.1093/nar/gks138
Nucleic Acids Research, 2012, Vol. 40, No. 12Published online 24 February 2012
? The Author(s) 2012. Published by Oxford University Press.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/
by-nc/3.0), which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Page 2

known to be MMR competent (7–10). In eukaryotes,
MutS homologs are hetero-dimeric proteins (e.g. human
MSH2/MSH6 or MutSa and MSH2/MSH3 or MutSb)
(11–13).
Crystal structures of various DNA–MutS complexes
show a hallmark 45–60?kink in the DNA at the
mismatch/lesion target site and very similar mismatch/
lesion-specific and non-specific contacts between MutS
and DNA (12,14–16). The asymmetry in prokaryotic
and eukaryotic MutS dimers and their interactions with
DNA revealed by the high-resolution structures is
striking; only one subunit of the MutS dimer makes
specific contact with only one of the mismatched bases
at the target site. A conserved phenylalanine from the
Phe-X-Glu-motif in the mismatch-binding domain of
E. coli MutS A subunit (eukaryotic MSH6) stacks on a
mismatched base from one of the DNA strands and the
conserved glutamate makes a hydrogen bond with the
same base (12,14,15). In principle, MutS could bind a mis-
matched base pair (e.g. G:T) in two orientations, with Phe
stacking either on the mismatched G in the top strand or
the mismatched T in the bottom strand. However, in
crystal structures of a G:T mismatch bound by E. coli
MutS and human MutSa, only the T (in the bottom
strand) is found stacked with the conserved Phe (12,14).
In the crystal structure of E. coli MutS bound to G:G,
only the G in the bottom strand is bound (recognized)
by MutS (15). These structural data suggest an intrinsic
bias in the orientation of MutS interaction with mis-
matched base pairs, which could have significant conse-
quences for the efficacy with which different mismatches
are recognized and repaired. Other than the information
available from crystal structures, little is known about dir-
ectional binding of MutS to mismatches. The only
in-solution quantitative data available on this issue are
from a single-molecule DNA unzipping analysis, in
which differences were observed in the peak force values
required to disrupt the Saccharomyces cerevisiae Msh2–
Msh6–G:T complex versus Msh2–Msh6–T:G complex
(17). Thus, the question as to whether MutS binding to
mismatches is directionally biased in a solution remains to
be addressed, as well as the related question of whether
MMR is influenced by which base of the mismatch is con-
tacted by the MutS Phe-X-Glu.
In this study, we investigated binding of E. coli MutS
to DNA containing different mismatches using ensem-
ble fluorescencespectroscopy
multiparameter fluorescence
order to gain information on the associated structural
dynamics (not provided by the static crystal structures)
and to specifically address the possibility that MutS
binds mismatches with a preferred orientation. We per-
formed the single-molecule fluorescence spectroscopy on
free diffusing molecules to avoid possible complications
due to unwanted interactions of MutS with the surface.
This is the first time that smMFD has been used to study
the MutS protein, and this approach allowed us to detect
multiple DNA–MutS species that otherwise are not distin-
guishable in ensemble or other single-molecule studies
(18–22). We show that in a solution, MutS does interact
with a mismatched base pair in two different orientations
and single-molecule
(smMFD),detection in
and that, depending on the type of mismatch, one orien-
tation is indeed preferred strongly over the other. We
discuss the implications of these findings in light of
proposed models of the MMR mechanism.
MATERIALS AND METHODS
Proteins
The cysteine-free (CF) E. coli MutS dimer variant
MutS-CF/D835R was expressed as the His-tagged fusion
protein and purified using Ni-NTA affinity chromatog-
raphy followed by gel filtration (9). Purification of recom-
binant proteinsMutL and
essentially as described elsewhere (23–25).
MutH wasperformed
Fluorophore-labeled DNA
The sequences of the 42 bp oligonucleotides used in fluor-
escence experiments were top strand, 50-TAT TAA TTT
CGC GGG CTC GAX AGC TTC ATC CTC TAC GCC
GGA-30and bottom strand, 50-TCC GGC GTA GAG
GAT GAA GCT YTC GAG CCC GCG AAA TTA
ATA-30. X:Y represents the type of base pair at the
central position (e.g. in G:T, X=G and Y=T).
Labeled DNA duplexes were prepared by annealing com-
plementary oligonucleotides (Purimex, Germany) contain-
ing Alexa Fluor 594 (A) at position 9 on the X strand and
Alexa Fluor 488 (D) on position 8 on the Y strand
attached toathymine
(Figure 1A). See also Figure 6 for additional oligonucleo-
tides used in the study.
viaaflexibleC6-linker
Modeling the 42bp G:T heteroduplex DNA
A 3D model of the 42-bp DNA (X:Y) structure in B-form
was generated using the 3D-DART webserver (26). Two
structural models of the MutS-bound kinked DNA were
generated by using DNA distortion parameters obtained
from the NDB server (ndbserver.rutgers.edu) using the
crystal structure of MutS and a 30-bp G:T DNA (PDB
code 1e3m). Parameters of positions 2–16 from the top
strand (chain E) and positions 15–29 from bottom
strand (chain F) were used for positions 14–28 of the
top strand and positions 57–71 of the bottom strand
forming the 42-bp G:T-DNA (positions 57–71 and
14–28 for the T:G DNA, respectively). Parameters for
B-DNA (obtained from the 3D-DART webserver, see
above) were used for all other positions. Sequence-
dependent structural bending of the chosen DNA
sequence is not significant as judged by the program
DIAMOD (27) MutS was docked to this modeled DNA
by superposition of the phosphate backbone atoms of the
DNA in the co-crystal structure onto the respective atoms
of the modeled DNA using PyMOL (DeLano Scientific
LLC,version 1.2r2).The
fluorophores were modeled as described in (28). They
aredepictedascloudsrepresenting
freedom of the fluorophore dependent on the linker
length and steric clashes with the DNA (for details see
Supplementary Data Sections 1.1 and 1.2). The distance
between mean dye positions Rmpwas determined as 85A˚
accessiblevolumesof
therotational
Nucleic Acids Research, 2012,Vol.40, No. 125449

Page 3

for dsDNA (Figure 1A). The distance dependence for
Fo ¨ rster Resonance Energy Transfer (FRET) from an
excited D molecule to an A molecule is given by the
Fo ¨ rster equationE=1/(1+(RDA/R0)6),
FRET efficiency, E, is related to distance between D and
A, RDAand the Fo ¨ rster radius R0for a specific DA pair
(for Alexa Fluor 488 (D) and Alexa Fluor 594 (A)
R0=53.2A˚).
However, in FRET experiments not Emp (Emp=1/
(1+(Rmp/R0)6)) but rather a mean FRET efficiency hEi
is measured, which is an average over all pairs of
spatial dyepositions,RDA(i),
E
h i ¼ hR6
the Fo ¨ rster equation and solving for the inter-dye
distance, the FRET-averaged distance between the dyes
hRDAiE can be calculated as hRDAiE=R0 (1/hEi–1)1/6.
The corresponding theoretical values for hEi and hRDAiE
are denoted as hEimodeland
Considering the volume sterically accessible to the dye in
the above model of dsDNA, the computed value is
hEimodel=0.063, which corresponds to
83A˚ (28). In our case,
h
wherethe
and isgivenby
0=R6
0+R6
DAðiÞi (21,28) (22). By inserting hEi in
RDA
himodel
E
, respectively.
RDA
himodel
E
=
RDA
imodel
E
is close to the
Rmp=85A˚. Considering the two MutS:DNA complexes
in Figure 1B, we calculate for GA:TD and for TA:GD
bound to MutS nearly identical values of E
and the RDA
h
h imodel=0.20
imodel
E
=67A˚, which is close to Rmp=69A˚.
Single-molecule fluorescence spectroscopy
In all experiments, double-labeled DNA (7.5–15 pM) was
incubated with MutS (150nM dimer) in 25mM Tris
(pH7.5), 125mM KCl and 5mM MgCl2and 0.5mg/ml
BSA in the absence or presence of ADP, ATP or
AMP-PNP (1mM). The experiments were carried out
with a confocal epi-illuminated set-up (29) with spectral
detection windows for donor (520/66nm) and acceptor
(630/60nm). The fluorescently labeled molecules were
excitedbya linearlypolarized,
Argon-ionlaser(476.5nm,
Supplementary Data Section 1.3). Fluorescence bursts
are distinguished from the background of 3–3.5kHz by
applying threshold intensity criteria defined by 0.05ms
interphoton time and 100 photons minimum per burst.
An additional selection criterion was applied to reduce
active-mode-locked
73MHz,150ps) (see
A
B
C
D
Figure 1. MutS-binding orientation and DNA bending with mismatched DNA monitored by fluorescence spectroscopy. (A) Sequence and molecular
model of the 42-bp DNA containing a central mismatch (XA:YD, black) and an Acceptor (Alexa Fluor 594, red) and Donor dye (Alexa Fluor 488,
green) in the top or bottom strand. In the molecular model the dyes are depicted as clouds, representing possible positions of the freely rotating dye
(21). The distance between the mean dye positions is Rmp=85A˚. (B, C) Model of the complex between E. coli MutS and XA:YDDNA when Phe-36
(indicated with a blue hexagon) is stacked on the mismatched X base (B) or the mismatched Y base (C). Subunit A and B of MutS are shown in blue
and orange and the DNA bases interacting which each monomer are shown in blue and orange, respectively. In both binding orientations (B or C),
the distance between mean dye positions is Rmp=69A˚; however, only in (B, Model I) the donor dye is in contact with the protein. (D) Steady-state
fluorescence emission spectra of 10nM GA:TD, TA:GDand GA:GDin the absence (blue curve) or presence (magenta curve) of 125nM MutS dimer.
5450Nucleic Acids Research, 2012,Vol.40, No. 12

Page 4

the number of bursts containing donor only mol-
ecules taken into consideration in the analysis (see
Supplementary Data Section 2.1). A detailed description
for the calculation for the experimentally recovered FRET
parameters is given in Supplementary Data Section 2.2.
A description of the fitting procedure for Photon
DistributionAnalysis(PDA)
Supplementary Data Section 2.3, whereas a detailed de-
scription of the analysis techniques can be found in
ref. (30,31).
isgiven inthe
Stopped-flow fluorescence measurements
All experiments were performed in 25mM Tris (pH 7.5),
125mM KCl and 5mM MgCl2 by using an Applied
Photophysics SX20 with dual detection stopped flow ap-
paratus (deadtime 1.1ms), with Alexa Fluor 488 excitation
at 493nm and Alexa Fluor 594 excitation at 593nm.
Donor and acceptor fluorescence was separated by two
bandpass filters with the following ranges: HQ 520/
35nm (green) and HQ 645/75nm (red). The signal was
normalized to the maximum of each measurement.
Kinetics of MutS binding to GA:TDwere measured by
mixingequal volumes of
pre-incubated with nucleotide (2mM of ADP, ATP or
AMP-PNP), with DNA (30nM) (mixing ratio 1:1). The
effect of ATP or ADP on a GA:TD–MutS complex was
determined by pre-incubating DNA (30nM) and MutS
(500nM dimer) and challenging the complex with 2mM
of ADP or ATP (mixing ratio 1:1). Dissociation kinetics of
MutS from GA:TDwere measured by mixing preformed
GA:TD–MutS complex with 75-fold molar excess of a trap
(G:T-22, see also Figure 7) and ADP or ATP (mixing ratio
1:1). In all experiments, the final concentrations are
250nM MutS dimer, 1mM ATP and 15nM DNA.
The curves were fitted to a single, double or triple
exponential using OriginLab 8.5: single: y=y0+Ae?x/t,
double:y ¼ y0+A1e?x=t1+A2e?x=t2
y ¼ y0+A1e?x=t+A2e?x=t2+A3e?x=t3;y0corresponds to the
offset, A1, A2, A3correspond to the amplitudes of decay
constants t1, t2and t3and t=1/kobs.
MutS (500nM dimer),
or triple:
Steady-state fluorescence measurements
Steady state fluorescence measurements (FRET and an-
isotropy) were performed by incubating MutS (125 or
250nM dimer)with double-labeled
(10nM), in 25mM Tris/HCl, pH 7.5, 125mM KCl and
5mM MgCl2in the absence or presence of ADP (1mM)
for 5min at 20?C. Fluorescence anisotropy (r) measure-
ments were performed with double- or single-labeled
42-bp DNA as described for the steady-state fluorescence
measurements but using the polarization module of
Fluoromax 4 with ?ex=470nm and ?em=517nm for
the donor (rD), and ?ex=570nm and ?em=615nm for
the acceptor (rA) (1.8nm slit width; signals averaged
for 2s).
DNA,XA:YD
Competition titration experiments
MutS-CF/D835R (50nM dimer) was pre-incubated with
GA:TD DNA (50nM), and the complex was then
competed with increasing concentrations of unlabeled
42-bp duplexes containing a G:T or T:G mismatch, 22
or 6bp from the 30end (G:T-22, T:G-22, G:T-6, T:G-6).
FRET was monitored at three excitation/emission wave-
lengths (in nm), FD(?ex=475 / ?em=525), FA(575/625)
and FDA(475/625), using a Tecan Infinite M200 fluores-
cence plate reader. FDA/FAwas plotted against the con-
centration of the competitor and the data were fit using
the parameter estimation tool of COPASI (32).
Substrates for MutH activity assays
Long heteroduplex DNA (390 or 406bp) was generated by
annealing two single-stranded DNA fragments generated
from PCR products after ?-exonuclease treatment (33).
The following primers (biomers.net, Germany) were used
for each substrate: G:T-22-L, TCA TCC TCG GCA CCG
TCA C and p-TCC GGC GTA GAG GAT GAA GCT
CTC GAG CCC GCG AAA TTA ATA (top strand),
p-TCA TCC TCG GCA CCG TCA C and TCC GGC
GTA GAG GAT GAA GCT TTC GAG CCC GCG
AAA TTA ATA (bottom strand); G:T-6-L, TCA TCC
TCG GCA CCG TCA C and p-AAG CTC TCG AGC
CCG CGA AAT TAA TAC GAC TCA CTA TAG GGG
(top strand) and p-TCA TCC TCG GCA CCG TCA C
and AAG CTT TCG AGC CCG CGA AAT TAA TAC
GAC TCA CTA TAG GGG (bottom strand); T:G-22-L,
TCA TCC TCG GCA CCG TCA C and p-TCC GGC
GTA GAG GAT GAA GCT ATC GAG CCC GCG
AAA TTA ATA (top strand) and p-TCA TCC TCG
GCA CCG TCA C and TCC GGC GTA GAG GAT
GAA GCT GTC GAG CCC GCG AAA TTA ATA
(bottom strand); T:G-6-L, TCA TCC TCG GCA CCG
TCA C and p-AAG CTA TCG AGC CCG CGA AAT
TAA TAC GAC TCA CTA TAG GGG (top strand) and
p-TCA TCC TCG GCA CCG TCA C and AAG CTG
TCG AGC CCG CGA AAT TAA TAC GAC TCA
CTA TAG GGG (bottom strand); G:C-22-L, TCA TCC
TCG GCA CCG TCA C and p-TCC GGC GTA GAG
GAT GAA GCT CTC GAG CCC GCG AAA TTA ATA
(top strand) and p-TCA TCC TCG GCA CCG TCA C
and TCC GGC GTA GAG GAT GAA GCT CTC GAG
CCC GCG AAA TTA ATA (bottom strand). p stands for
a 50phosphorylated primer.
Mismatched-provoked activation of MutH by MutS
and MutL
MutH endonuclease activation by MutS and MutL was
assayed on long DNA substrates (390 or 406bp) contain-
ing a G:T, T:G or G:C base pair at position 385 and a
single unmethylated GATC site at position 210. DNA
(10nM) was incubated with MutH (500nM), MutL
(500nM monomer) and
dimer) in 10mM Tris-HCl (pH 7.9), 5mM MgCl2, 1mM
ATP and 125mM KCl at 37?C. DNA substrate and
cleavage products were analyzed by 6% polyacrylamide
gelelectrophoresisfollowed
staining. MutH endonuclease activity was scored by the
appearance of products from cleavage of both DNA
strands at the unmethylated GATC site.
MutS-CF/D835R(250nM
byethidium bromide
Nucleic Acids Research, 2012,Vol.40, No. 125451

Page 5

RESULTS AND DISCUSSION
Design and modeling of dye-labeled DNA to study MutS
binding and kinking DNA
Binding and bending of mismatched DNA by MutS was
monitored using double-labeled DNA and performing
bulk or single molecule FRET measurements. We used
Alexa Fluor 488 (D) and Alexa Fluor 594 (A) as a
FRET pair (DA) attached via a C6linker, respectively,
to the bottom and top strand of a 42bp double-stranded
oligonucleotide termed XA:YD(Figure 1A). The Fo ¨ rster
radius for the specific dye pair at the specific labeling pos-
itions was determined to be R0=53.2A˚. A molecular
model of the dye labeled oligonucleotide in the B-DNA
form was generated and the mean dye positions were
determined in the volume sterically accessible to the
coupled dyes (green and red clouds in Figure 1A, respect-
ively) as described in the Materials and Methods section
(21,28,34). According to our model, the FRET-averaged
distance between the two fluorophores in free DNA,
RDA
h
0.063.
Next, we modeled the structure of the complex between
MutS and the dye-labeled DNA (XA:YD) based on the
crystal structure of E. coli MutS with a G:T DNA (PDB
code 1e3m, (14). We generated two models for XA:YD, in
which the conserved Phe-36 in subunit A of MutS stacked
either on the mismatched base in the top strand (X)
(Model I, Figure 1B) or the bottom strand (Y) (Model
II, Figure 1C). In both cases, the FRET-averaged
distance,RDA
h
for free DNA to 67A˚ for MutS-bound DNA. However,
differences in dye–protein interactions between the two
DNA–MutS complexes are readily apparent. In Model I
(Phe-36 stacked on base X; Figure 1B), ?8% of the donor
dye cloud is within 5A˚distance to the mismatch-binding
domain of MutS subunit B, whereas no contact is detected
in Model II (Phe-36 stacked on base Y; Figure 1C). In
contrast, 3% and 23% of the acceptor dye cloud is
within 5A˚ distance to the DNA-binding clamp domains
of MutS in Models I and II, respectively. These structural
models indicate that (i) the decrease in sterically available
volume for the dye molecules and possible quenching
effects due to dye–protein interactions could result in
changes in fluorescence anisotropy and (ii) differential
changes in donor and acceptor fluorescence anisotropy
could distinguish the two DNA–MutS complexes from
each other. Thus, in addition to FRET serving as a
monitor for DNA binding/bending, anisotropy of the
donor and acceptor dye emission may serve as a
monitor of MutS orientation on DNA.
Escherichia coli MutS exists in dimer/tetramer equilib-
rium with a reported KDranging from 0.2 to 2mM (7,8).
In this study, we used a tetramerization-deficient but fully
repair proficient variant of E. coli MutS (MutS-CF/
D835R) (9,10) to avoid potential complications in data
analysis due to binding of MutS tetramers to DNA (35).
Steady-state fluorescence spectroscopy experiments with
double-labeled DNA containing a G:T, T:G and G:G
mismatch (XA:YD format; Figure 1A) displayed low
FRET (LF) in the absence of protein (blue curves in
imodel
E
, is 83A˚ and the FRET efficiency, hEimodel, is
imodel
E
, is expected to shorten from 83A˚
Figure 1D), as expected from the large distance between
the two dyes shown in the model. Upon addition of
125nM MutS to the solution, there was a significant
increase in energy transfer as indicated by the concomitant
decrease of donor and increase of acceptor fluorescence.
Only the GA:TD–MutS complex showed the high FRET
(HF) expected from the structural models described
above. In contrast, the TA:GD–MutS complex showed
only a donor quench and the GA:GD–MutS complex
showed a weak FRET increase. It should be noted that
in the available crystal structures, all three complexes have
DNA kinked by 60?; thus, there are subtle conformational
differences among these complexes that are not discernible
in crystal structures. Since bulk FRET measurements
could not allow us to distinguish different populations
of free and bound DNA, we performed a more detailed
analysis using smMFD as described below.
smMFD analysis of free DNA
In order to directly detect the heterogeneity of distinct
DNA–MutS complexes, we employed a single-molecule
approach coupled with MFD. smMFD experiments
avoid ensemble averaging by counting and analyzing
single-molecule events one at a time. As outlined above,
our experimental setup was designed to monitor DNA–
MutS complexes by differences in FRET between two dye
molecules, and sense orientation-specific binding of MutS
to DNA by changes in fluorescence anisotropy (Figure 1).
In our single-molecule fluorescence microscope, individu-
ally labeled DNA molecules (15pM) freely diffuse
through a confocal detection volume and generate brief
bursts of fluorescence over the short dwell time (?1ms).
During these transits, fluorescence intensities (F), lifetimes
(?), anisotropies (r) and diffusion times of both the donor
(D) and acceptor (A) dyes are probed. The selected bursts
are characterized offline to determine all fluorescence par-
ameters. In this work, we performed MFD analysis in two
steps. In Step 1, we generated 2D fluorescence parameter
histograms of single-molecule bursts (MFD plot) to gain a
semi-quantitative overview of distinct species with distinct
apparent mean FRET efficiencies
anisotropies rD(Figure 2A–E). In Step 2, PDA was used
for a quantitative and rigorous analysis of the apparent
mean FRET efficiency probability distributions to calcu-
late also the corresponding inter-dye distances
(Figure 2F–J). In addition, PDA allows us to determine
the actual number of species in a probability distribution
by consideringthe stochastic
emission, which leads to the broadening of the observed
hEai distributions (photon shot noise) (see Supplementary
Data Section 2.3) (20). Due to the presence of many
similar species, we preferred an analysis procedure
(for all details see Supplementary Data Section 2.3: fit
model 2), where the relevant FRET states were identified
from the measurements in which they were predominant
and thus able to be characterized with high confidence (see
Supplementary Tables S1.4–S1.6). In a joint analysis of all
histograms, the PDA fits were most stable, when only
relative species fractions were left to vary, i.e. we used
fixed distances and variable fractions. Using both 2D
hEai and donor
RDA
hiE
character of photon
5452 Nucleic Acids Research, 2012,Vol.40, No. 12