MutS homolog sliding clamps shield the DNA from binding proteins

Sliding clamps on DNA consist of evolutionarily conserved enzymes that coordinate DNA replication, repair, and the cellular DNA damage response. MutS homolog (MSH) proteins initiate mismatch repair (MMR) by recognizing mispaired nucleotides and in the presence of ATP form stable sliding clamps that randomly diffuse along the DNA. The MSH sliding clamps subsequently load MutL homolog (MLH/PMS) proteins that form a second extremely stable sliding clamp, which together coordinate downstream MMR components with the excision-initiation site that may be hundreds to thousands of nucleotides distant from the mismatch. Specific or nonspecific binding of other proteins to the DNA between the mismatch and the distant excision-initiation site could conceivably obstruct the free diffusion of these MMR sliding clamps, inhibiting their ability to initiate repair. Here, we employed bulk biochemical analysis, single-molecule fluorescence imaging, and mathematical modeling to determine how sliding clamps might overcome such hindrances along the DNA. Using both bacterial and human MSH proteins, we found that increasing the number of MSH sliding clamps on a DNA decreased the association of the Escherichia coli transcriptional repressor LacI to its cognate promoter LacO. Our results suggest a simple mechanism whereby thermal diffusion of MSH sliding clamps along the DNA alters the association kinetics of other DNA-binding proteins over extended distances. These observations appear generally applicable to any stable sliding clamp that forms on DNA.

Sliding clamps on DNA consist of evolutionarily conserved enzymes that coordinate DNA replication, repair, and the cellular DNA damage response. MutS homolog (MSH) proteins initiate mismatch repair (MMR) by recognizing mispaired nucleotides and in the presence of ATP form stable sliding clamps that randomly diffuse along the DNA. The MSH sliding clamps subsequently load MutL homolog (MLH/PMS) proteins that form a second extremely stable sliding clamp, which together coordinate downstream MMR components with the excision-initiation site that may be hundreds to thousands of nucleotides distant from the mismatch. Specific or nonspecific binding of other proteins to the DNA between the mismatch and the distant excision-initiation site could conceivably obstruct the free diffusion of these MMR sliding clamps, inhibiting their ability to initiate repair. Here, we employed bulk biochemical analysis, singlemolecule fluorescence imaging, and mathematical modeling to determine how sliding clamps might overcome such hindrances along the DNA. Using both bacterial and human MSH proteins, we found that increasing the number of MSH sliding clamps on a DNA decreased the association of the Escherichia coli transcriptional repressor LacI to its cognate promoter LacO. Our results suggest a simple mechanism whereby thermal diffusion of MSH sliding clamps along the DNA alters the association kinetics of other DNA-binding proteins over extended distances. These observations appear generally applicable to any stable sliding clamp that forms on DNA.
Sliding clamps on DNA have been conserved throughout evolution and play essential roles in coordinating DNA replication, repair, and the cellular damage response (1)(2)(3). Among notable examples are the MutS homolog (MSH) 4 proteins, which operate as a homo-or hetero-dimer and specifically recognize mismatched nucleotides, lesions, or structures within the DNA (4). MSH proteins bind and hydrolyze ATP (5)(6)(7). Importantly, DNA mismatch or structure recognition provokes ATP binding by MSH proteins that ultimately results in the formation of a stable sliding clamp (8 -13). These MSH sliding clamps move by one-dimensional (1D) thermal (Brownian) diffusion with a lifetime of up to 10 min while in intermittent contact with the DNA backbone (14 -17). The MSH sliding clamps that function in mismatch repair (MMR) act as a platform to load MutL homolog (MLH/PMS) proteins, which then form a second extremely stable freely diffusing ATP-bound sliding clamp that communicates mismatch recognition along the DNA to an excision-initiation site, which may be hundreds to thousands of nucleotides distant from the mismatch (4,18).
Unrelated DNA-binding proteins may become roadblocks if they bind between the mismatch and the downstream excisioninitiation site. Single-molecule imaging of molecular motors that use the energy of ATP hydrolysis to move along the DNA have demonstrated that these proteins can actively remove roadblocks (19). Alternatively, some proteins that rely on thermal fluctuation-driven motion may hop over roadblocks on DNA (20). Remarkably, MSH sliding clamps that rely entirely on 1D thermal diffusion to move along the DNA have been shown to disassemble nucleosomes (21). The mechanism of MSH-dependent nucleosome disassembly was not immediately obvious since the numerous histone octamer-DNA interactions that comprise a nucleosome would appear to require significant energy to release (22).
One hypothesis is that multiple randomly diffusing MSH sliding clamps might increasingly occupy nucleosomal DNA that is transiently unwrapped as a result of thermal fluctuation (23). Such 1D thermal motion of multiple MSH sliding clamps was projected to provide an explicit mechanism for influencing the binding of other proteins over distances that are larger than the total DNA footprint of the MSH particles (21,23). Here we have used bulk analysis and single-molecule fluorescence imaging to examine the properties of MSH sliding clamps on DNA. We found that increasing the numbers of MSH sliding clamps on DNA decreased LacI repressor association to its cognate LacO promoter. We also show that the number of MSH sliding clamps already bound to the DNA affected the loading of additional MSH sliding clamps at a mismatch. Together these observations appear consistent with the ability of MSH sliding clamps to affect the association kinetics of additional DNAbinding proteins.

ATP-bound MutS homologs reduce LacI binding to its cognate LacO site
The random diffusion characteristics of MSH sliding clamps on DNA suggested that they might influence the association of heterologous DNA-binding proteins over distances greater than their footprint. To address this possibility we examined the effect of three different MSH proteins, Thermus aquaticus TaMutS, E. coli EcMutS, and human (Homo sapiens) HsMSH2-HsMSH6, on the binding of the E. coli transcription factor LacI to its cognate LacO site using surface plasmon resonance (SPR) (Fig. 1, a-c, left panels; supporting Fig. S1). In this system the proximal end of a 98-bp mismatched DNA was attached to the SPR surface utilizing a biotin-streptavidin linkage. The mismatch was located 15 bp from the surface and the 21-bp LacI LacO site (24 -26) was located 12 bp from the distal end that was blocked by digoxigenin-antidigoxigenin (dig-antidig). This DNA substrate could retain up to three MSH sliding clamps with a footprint of ϳ26 bp in the presence of ATP (9,14,(27)(28)(29)(30). However, there is only enough space for LacI to bind if there are two or less MSH sliding clamps present on the DNA. Previous studies have demonstrated that single or multiple sliding clamps are not formed on duplex DNA in the absence of a mismatch (9,11). Following subtraction of the intrinsic MSH dissociation kinetics, we find that the association rate of LacI (k on⅐LacI ) decreased with increasing MSH concentration in the presence of ATP (Fig. 1, a-c, middle and right panels; supporting Fig. S1). As a control we show that increasing MSH concentration in the absence of ATP does not affect LacI binding (supporting Fig. S2). These observations are consistent with the conclusion that k on⅐LacI is inhibited with increasing ATP-bound MSH sliding clamps on the DNA.

The binding of LacI is altered by the number of MutS sliding clamps on DNA
SPR studies are unable to determine how the MSH sliding clamps affect the ability of LacI to bind the mismatched DNA. We consider at least two possibilities: 1) MSH sliding clamps saturate the DNA effectively obscuring the LacO site, or 2) subsaturating MSH sliding clamps affect LacI binding at a distance. We used a single-molecule total internal reflection fluorescent (smTIRF) microscope system to determine the number of MSH sliding clamps on the mismatched DNA and their effect on LacI binding. Cy3-labeled TaMutS was utilized in these studies because we have previously demonstrated specific mismatch binding and the formation of stable ATP-bound sliding clamps on defined mismatched DNA substrates (14,17). The 98-bp DNA was attached to a passivated surface via a biotin-Neutr-Avidin linkage and modified to contain a ϩdT mismatch 15 bp from the proximal end with a Cy5 fluorophore 9 bp distal of the mismatch. The LacO site was located similarly to the SPR mismatched DNA substrate ( Fig. 2a; supporting Table S1). The Cy5 location was previously shown to have no influence on mismatch binding activity or the formation of TaMutS sliding clamps (14,17). The position of the LacO ensured that when bound by LacI the TaMutS sliding clamps may access only 65 bp of the 98-bp mismatched DNA.
LacI binding was determined by observing the time-averaged FRET produced by Cy3-TaMutS on the Cy5-labeled LacOmismatched DNA (14,17). FRET is inversely correlated with the length of accessible DNA and thus depends on the presence or absence of bound LacI. We calculate time-averaged FRET efficiencies of Cy3-TaMutS on the full-length 98 bp LacO-mismatched DNA of E 1 ϳ0.32, E 2 ϳ0.34, and E 3 ϳ0.30 for one, two, and three single fluorophore-labeled sliding clamps, respectively ( Fig. 2; supporting Fig. S3). The binding of LacI to the LacO-mismatched DNA reduces the effective length that Cy3-TaMutS sliding clamps may occupy, resulting in a calculated increase in the time-averaged FRET efficiency of E 1⅐LacI ϳ0.60 and E 2⅐LacI ϳ0.46 for one and two single-labeled Cy3-TaMutS sliding clamps, respectively. The site size of three TaMutS sliding clamps on the LacO-mismatched DNA is expected to fully exclude LacI binding.
The number of TaMutS sliding clamps on single LacO-mismatched DNA molecules was determined by counting the number of Cy3 photobleaching steps ( Fig. 2b; supporting Fig.  S3). These were separated into single (E 1 ), double (E 2 ), and triple (E 3 ) fluorophores with associated FRET efficiency (E). Based on the labeling efficiency we expected 30% of the TaMuS dimers might contain two Cy3 fluorophores, potentially influencing the sliding clamp count. However, in the absence of LacI the binned FRET efficiencies fit to normal curves with means of E 1 ϭ 0.31 Ϯ 0.08, E 2 ϭ 0.30 Ϯ 0.12, and E 3 ϭ 0.21 Ϯ 0.08, which closely correlated with the calculated time-averaged FRET efficiencies ( Fig. 2b; supporting Fig. S3). Prebinding LacI to the LacO-mismatched DNA followed by loading Cy3-TaMutS sliding clamps resulted in DNA molecules that contained one or two fluorophores with FRET efficiencies that correlated extremely well with the calculated time-averaged FRET efficiencies (see Fig. 2a compared with Fig. 2b for one fluorophore, E 1⅐LacI ϭ 0.60 Ϯ 0.10, or compared with supporting Fig. S3 for two fluorophores, E 2⅐LacI ϭ 0.47 Ϯ 0.07). We conclude that simple Cy3 fluorophore counting provides a reasonable approximation of the number of TaMutS sliding clamps, and that time-averaged FRET efficiency is a practical indicator of the presence or absence of LacI on the LacO-mismatched DNA.
We noted that the distributions of binned FRET efficiency with and without LacI overlapped for both the one-or twofluorophore cases. A stringent criterion for distinguishing LacIbound DNA is to identify molecules with a FRET efficiency that is greater than 2 S.D. from the mean (E 1⅐LacI Ͼ 0.4, E 2⅐LacI Ͼ  Table S1) containing a ϩdT (TaMutS) or G/T (EcMutS and HsMSH2-HsMSH6) mismatch was anchored to the surface of the streptavidin-coated SPR (Biacore) chip and the remaining end blocked by dig-antidig as described previously (38). Each curve is color-coded and comes in pairs Ϯ LacI injection. Association curves were processed as described in supporting MutS sliding clamps shield the DNA 0.33). We then examined 40 m ϫ 80 m fields containing ϳ60 well-resolved molecules and found that when one Cy3 fluorophore was associated with the Cy5-labeled LacO-mismatched DNA, 55% of the molecules displayed a time-averaged FRET efficiency consistent with LacI bound to the DNA (E 1⅐LacI Ͼ 0.4) (Fig. 2c). When two Cy3 fluorophores were associated with the Cy5 LacO-mismatched DNA, 37% of the molecules displayed a time-averaged FRET efficiency consistent with LacI bound to the DNA (E 2⅐LacI Ͼ 0.33) (Fig. 2c). As expected, we did not observe any molecules that displayed altered time-averaged FRET when three or four Cy3 fluorophores were associated with the Cy5 mismatched DNA, although these events were  Table S1). Binding of Cy3-TaMutS to the mismatch results in a high FRET signal (E ϳ 0.8) that in the presence of ATP resolves into a sliding clamp with time-averaged FRET (E ϳ 0.3) as described previously (17). The binding of LacI and/or multiple sliding clamps alter the time-averaged FRET efficiency as shown below the middle and right molecules and as described in the text.

MutS sliding clamps shield the DNA
relatively rare (Fig. 2c). We note that with the four Cy3 fluorophore case at least one of the TaMutS sliding clamps must contain two fluorophores. These results support the conclusion that increased numbers of MSH sliding clamps exclude LacI transcription factor binding to DNA.

MutS sliding clamps on the DNA affect the k on of LacI
To determine whether MutS sliding clamps affect the on or off rate of LacI-binding kinetics, we developed a single-molecule protein-induced fluorescence enhancement (smPIFE) system (31). A Cy3 fluorophore was placed 2 bp distal to the LacO site on the 98-bp mismatched DNA ( Fig. 3a; supporting Table  S1). LacI binding induced a PIFE signal (supporting Fig. S4a) in which the off⅐LacI depended upon protein concentration and could be used to determine the on rate (k on⅐LacI ; supporting information and supporting Fig. S4b) and off rate ( on⅐LacI ϭ 1/k off⅐LacI ) to calculate the equilibrium dissociation constant (K D ⅐LacI ϭ k off⅐LacI /k on⅐LacI ϭ 119 Ϯ 22 pM), which was similar to historical reports (24). To count the number of ATP-bound sliding clamps we labeled TaMutS or EcMutS with Alexa Fluor 647 (Thermo Fisher Scientific) as described previously (14,17).
For these studies the off⅐LacI was determined with 1 nM LacI, which was then followed by counting the number of Alexa Fluor 647 photobleaching events to determine the number of fluorophores on single LacO-mismatched DNA molecules ( Fig.  3b; supporting Fig. S5 and supporting information). The resulting off⅐LacI histograms were individually binned for one-and two-fluorophore photobleaching events (17). To account for the possibility of unlabeled and double-labeled MutS dimers the off⅐LacI histograms were fit to a double exponential (Fig. 3c). We then normalized the total probability density to 1 and based on the experimentally determined labeling efficiency, fixed the ratio of double fluorophore bleaching events to single fluoro-phore bleaching events to reduce the number of variables to two (supporting Fig. S5 and supporting information). These fits could either be used directly or subjected to maximum likelihood estimation to garner the off⅐LacI when one ( off⅐LacI-1 ) or two ( off⅐LacI-2 ) MutS sliding clamps were present on the LacOmismatched DNA (supporting Table S3). These off values were then used to calculate the k on⅐LacI in the presence of one MutS sliding clamp or two MutS sliding clamps and displayed individually with standard deviation error for experimental comparison (Fig. 3c). These results demonstrate that increased numbers of MutS sliding clamps decrease the k on of LacI association to its cognate LacO-binding site.

Reduction in LacI binding is consistent with a predictive model for freely diffusing MSH sliding clamps on DNA
We developed a model that quantitatively describes the reduction of k on⅐LacI as a result of any number of sliding clamps and DNA lengths. This model treats MSH sliding clamps as particles that engage in free 1D diffusion along the DNA, with the only requirement that their footprints cannot overlap (23). A similar model was originally described by Lewi Tonks (32) and is generally recognized as a 1D Tonks gas model. Importantly, the properties of a 1D Tonks Gas/Sliding Clamp can be systematically calculated with this model (supporting information, model for freely diffusing MutS sliding clamps, and supporting Table S2).
To test the application of the Tonks Gas/Sliding Clamp model in the absence of additional proteins and to establish its main parameter, the footprint size () of the MutS sliding clamps, single-molecule fluorescence imaging was used to count the number of MutS sliding clamps on a DNA containing a single mismatch (Fig. 4a; supporting Table S1). In these studies a 95-bp DNA was linked at the proximal end to a passivated The k on⅐LacI was determined from the off⅐LacI data shown in supporting Table S3 and color-coded similarly as follows: red hash, least squares calculation using one-fluorophore data; blue hash, least squares calculation using two-fluorophore data; green hash, maximum likelihood calculation using one-fluorophore data; purple hash, maximum likelihood calculation using two-fluorophore data (supporting information).

One MutS Two MutS Three MutS
MutS sliding clamps shield the DNA surface via 5Ј-biotin-NeutrAvidin and the distal end was blocked by 3Ј-dig-antidig antibody (Fig. 4a). Blocking both ends (surface and dig-antidig) effectively retains freely diffusing MutS sliding clamps on the mismatched DNA as described previously (14,17). An extrahelical ϩdT mismatch was positioned at 15 bp from the proximal end and the DNA was labeled 7 bp from the distal end with Cy5 (Fig. 4a). Cy3-labeled TaMutS was used in this analysis because we have previously demonstrated specific mismatch binding and the formation of stable ATP-bound sliding clamps on similar mismatched DNA substrates (14,17).
We first determined that the location of the Cy5 does not influence TaMutS mismatch binding or the formation of an ATP-bound sliding clamp by SPR (supporting Fig. S6). These results suggest the ATP-bound TaMutS sliding clamps loaded onto the DNA are the consequence of genuine mismatch recognition events.
Our previous studies showed that ATP-bound Cy3-TaMutS sliding clamps rapidly diffuse along the length of short oligonucleotides producing a time-averaged FRET with the Cy5-labeled DNA (Fig. 4b) (14,17). These observations indicated that the number of ATP-bound TaMutS sliding clamps could be unequivocally determined by monitoring the FRET signal that results from Cy3-TaMutS located on the Cy5-mismatched DNA and then counting the number of Cy3 photobleaching events over a 100-s observation ( Fig. 4b; supporting Fig. S7).
We found that the fraction of TaMutS sliding clampassociated fluorophores depended on protein concentration (Fig. 4c, data points). The Tonks Gas/Sliding Clamp model predicts that the effective on-rate for binding a new TaMutS clamp will be affected by number of MutS sliding clamps already diffusing on the DNA. This prediction allows for calculation of the concentration-dependent number of MutS sliding clamp on the DNA (supporting information). The general agreement with the measured and calculated distribution of MutS sliding clamps provides support for the Tonks Gas/Sliding Clamp model (Fig. 4c, lines). In addition, the interpolated footprint size ( ϭ 26 bp) corresponds well with previous biochemical determinations (33).
The Tonks Gas/Sliding Clamp model may also be used to calculate the k on⅐LacI in the presence of MutS sliding clamps that may transiently occlude the LacO-binding site and reduce accessibility to LacI (Fig. 3; supporting information). Moreover, the model permits calculation of on-rates for LacI (k on⅐LacI ) at different DNA lengths and numbers of MSH sliding clamps bound to the DNA (Fig. 4d). The ratio of k on⅐LacI with zero MSH sliding clamps (k on⅐LacI-0 ) to that with N number of MSH sliding clamps on the DNA (k on⅐LacI-N ) provides a convenient measure of any reduction in LacI binding (Fig. 4d). For example, the Tonks Gas/Sliding Clamp model predicts at least a 2-fold reduction in k on⅐LacI-N (k on⅐LacI-0 /k on⅐LacI-N ϭ 2) when two MutS sliding clamps are confined to 160 bp or less, and when three MutS sliding clamps are confined to 235 bp or less. A 10-fold reduction in k on⅐LacI-N (k on⅐LacI-0 /k on⅐LacI-N ϭ 10) is predicted when two MutS sliding clamps are confined to 100 bp or less, and three MutS sliding clamps are confined to 140 bp or less.
For the specific DNA used in Fig. 3 we calculated a distribution of predictions using several hypothetical MSH footprints ( ϭ 24, 25, or 26 bp) and LacI footprints ( ϭ 21 or 25 bp) and assuming that either the 5Ј-biotin-NeutrAvidin or dig-antidig linkages increased the DNA length by an additional 2 bp, which resulted in a mean with upper and lower quartile for k on⅐LacI-0 / k on⅐LacI-N ( Fig. 4e; supporting information). We find excellent agreement with the theoretical mean predicted by the Tonks Gas/Sliding Clamp model for two MutS sliding clamps when compared with the experimentally determined average of the k on⅐LacI-2 shown in Fig. 3c (k on⅐LacI-0 /k on⅐LacI-2 ϭ 9.6 Ϯ 2.0 S.D.; Fig. 4e, red asterisk in two MutS). For one MutS sliding clamp, we noted a larger effect on the experimental ratio than predicted by the Tonks Gas/Sliding Clamp model (k on⅐LacI-0 / k on⅐LacI-1 ϭ 2.6 Ϯ 0.9 S.D.; Fig. 4e, red asterisk in one MutS). Our calculations using this model assume that the DNA is a stiff rod and the MutS proteins are inelastic spheres with defined site size. However, small changes on the effective DNA length that may be occupied by a MutS sliding clamp, the site size of either the MutS or the LacI proteins (elasticity or hydration) and the mechanical properties of the DNA compared with lattice models (34) may account for the differences in experimental and predicted effects of one MutS sliding clamp on k on⅐LacI .

Discussion
Our results provide consistent evidence that stable freely diffusing ATP-bound MSH sliding clamps can influence the association of other binding proteins with the DNA. The major effect of the multiple MSH sliding clamps is to reduce the k on of a DNA-binding protein. This would effectively increase the K D of a DNA-binding protein ultimately reducing its equilibrium binding as well as its localization to the region occupied by stable sliding clamps. There appeared to be little effect on the k off of LacI when one MSH sliding clamp was present, although the number of observations was low as a result of the pM K D of LacI. A unique effect on k on implies that the exclusion effect introduced by multiple MSH sliding clamps would only affect unbound DNA-binding proteins or previously bound DNA-binding proteins immediately following equilibrium dissociation.

MutS sliding clamps shield the DNA
These observations support a previous hypothesis that the disassembly of nucleosomes by HsMSH2-HsMSH6 was the result of MSH sliding clamps inhibiting nucleosome DNA rewrapping kinetics around the histone octamer (21,23). This concept also explains the synergistic effect of HsMSH2-HsMSH6 -catalyzed nucleosome disassembly when histone post-translational modifications that reduced nucleosome DNA wrapping stability were present (21,23). Recent studies from our group demonstrated that MSH sliding clamps load MLH/PMS proteins onto the DNA in a cascade of extremely stable ATP-bound sliding clamps (18). It seems possible that these two sliding clamps together may synergistically reduce the localized association of proteins that could inhibit communication and/or excision processes during mismatch repair.
There are a number of other stable sliding clamps that associate with DNA in all organisms. In eukaryotes these include the replicative processivity factor proliferating cell nuclear antigen (PCNA) and the DNA damage response complex RAD9-HUS1-RAD1 (9-1-1 complex) that are ubiquitous in dividing cells and are essential for genome maintenance and stability (1,2). It is likely that biological evolution has selected for these stable sliding clamp complexes based at least in part on the useful physical property that they may freely diffuse over long distances.

Proteins and DNA
T. aquaticus TaMutS(C42A,T469C) was expressed, purified, and labeled with Cy3 or Alexa Fluor 647 as described previously with a labeling efficiency per monomer of 54.5 or 45%, respectively (14,17). E. coli EcMutS(D835R,R840E) containing a C-terminal formylglycine-generating enzyme (FGE) hexaamino acid recognition sequence and a hexa-histidine tag was expressed, purified, and labeled with Alexa Fluor 647 by Hydrazinyl-iso-Pictet-Spengler (HIPS) ligation as described previously (35,36) with a labeling efficiency per monomer of 26%. The human HsMSH2-HsMSH6 heterodimer was purified as described previously (37). Mismatch binding and ATP-dependent sliding clamp formation was determined by surface plasmon resonance (SPR), as described previously (8,38). PAGE gel purified DNA oligonucleotides (supporting Table S1) were purchased from Midland Reagents (Midland, TX) or Integrated DNA Technologies (Coralville, IA). Fluorophore labeling of DNA and purification by reverse phase HPLC on a C18 column (Agilent) was performed as described previously (14,17). Complementary oligonucleotides were annealed and purified by HPLC on a Gen-Pack ion exchange column (Waters) as described previously (39).

SPR-binding kinetic analysis
A 98-bp mismatched DNA containing a 5Ј-biotin at one end and a 5Ј-dig at the other end was attached to a streptavidincoated SPR chip (Biacore) as described previously (supporting Table S1) (38). MSH protein binding at indicated concentrations was performed in SM Buffer (20 mM Tris-HCl, pH 7.8, 100 mM NaCl, 5 mM MgCl 2 , 0.1 mM DTT. 0.1 mM EDTA, 0.5 mM ATP) and minus glucose, glucose oxydase/catylase (GOD/ CAT), and trolox, but with 0.005% Surfactant P20 (GE Health-care, BR100054), 0.2 mg/ml acetylated BSA (Promega, R3961), and 25 nM antidig (Roche) at 10 l/min and 23°C (EcMutS and HsMSH2-HsMSH6) or 35°C (TaMutS). For studies that examined the effect of MSH sliding clamps on LacI binding, following MSH binding in the presence of ATP (see MutS/MSH Injection, Fig. 1; supporting Fig. S1), LacI (0.5 nM for TaMutS; 1.5 nM for EcMutS and human HsMSH2-HsMSH6) was injected (see LacI Injection, Fig. 1; supporting Fig. S1) and binding compared with the absence of LacI. The MSH dissociation curve in the absence of LacI was subtracted from the rate curve in the presence of LacI to obtain the LacI association curve (supporting Fig. S1). The LacI association curve was fit to a single component binding exponential to obtain the k on⅐LacI at each concentration of initial MSH binding/loading (supporting Fig. S5).

Single-molecule FRET and photobleaching analysis
Cy3-TaMutS (100 nM) was incubated with indicated Cy5-DNA (supporting Table S1) for 5 min in SM Buffer (plus 0.0025% P20) to load TaMutS sliding clamps. After washing away free MutS with SM Buffer plus 0.0025% P20 with oxygen scavenging system (OSS) (0.8% w/v D-glucose, 146 units/ml glucose oxydase, 2170 units/ml catylase, 2 mM trolox) and 1 nM LacI, time-averaged FRET was detected after 1 min as described previously using a prism-type laser excitation smTIRF microscopy system (14,17) containing an Olympus IX-71 with watertype 60ϫ objective (N.A. 1.2), a Photometrics DV2 two-channel dual color separation system and a Princeton Instruments ProEM 512 Exelon charge-coupled device recorder. The number of Cy3 fluorophore photobleaching steps were counted as described previously (14,17) and compared with the number and distribution of fluorophores predicted based on the TaMutS monomer labeling efficiency. Briefly, laser intensity may be increased and the resulting step loss of fluorophore signal(s) until zero fluorescence is an indicator of the number of fluorophores on a single molecule. Fluorescent images and FRET following 532 nm DPSS LASER excitation were analyzed using IDL software (ITT VIS) and MATLAB (The MathWorks) scripts. Data analysis methods may be found in the supporting information.

Single-molecule PIFE
We assembled a prism-type laser excitation total internal reflection fluorescence (TIRF) microscopy system (14, 17) containing an Olympus IX-71 with water-type 60ϫ objective (N.A. 1.2), a Photometrics DV2 two-channel dual color separation system and a Princeton Instruments ProEM 512 Exelon chargecoupled device recorder. The DNA substrate was identical to the SPR substrate except it was labeled by Cy3 near the LacObinding site that when bound by LacI induced protein-induced fluorescence enhancement (PIFE) (supporting Table S1) (31). Fluorescent images were analyzed using IDL (ITT VIS) and MATLAB (The MathWorks) scripts following 532 nm (PIFE) and 635 nm (MSH fluorophore photobleaching and counting) DPSS laser excitation (Crystal Laser, 2 milliwatt) at 250 -400 ms time resolution. The off (period of protein-free DNA) and on (period of protein-bound DNA) were determined by measuring fluorescence intensity changes upon LacI binding (supporting Fig. S4a), where the cumulative distributions at several MutS sliding clamps shield the DNA concentrations of LacI were fit to a single exponential decay (supporting Fig. S5b). The k on⅐LacI (( off⅐LacI ) Ϫ1 ) and k off⅐LacI (( on⅐LacI ) Ϫ1 ) were found to be linear and used to obtain the K D ⅐LacI (supporting Fig. S4, b and c). We note that two measures for off , the time between the initial injection to the first LacI-binding event or the time between LacI-binding dissociation and reassociation, gave nearly identical k on between 0 and 1 nM LacI (compare supporting Fig. S4b and supporting Fig.  S4c). Above 1 nM LacI the off was increasingly short relative to our time resolution and fluorophore PIFE intensity fluctuations, thus escalating the measurement error.
The loading of MSH sliding clamps followed by LacI protein binding was carried out in SM Buffer plus 0.0025% P20 with the Alexa Fluor 647-TaMutS (250 nM, or 315 nM) or Alexa Fluor 647-EcMutS (30 nM, or 40 nM) added to the flow cell for 10 min, followed by a 10ϫ volume wash with SM Buffer to eliminate free MSH protein, followed by LacI (1 nM) injection. The off in the presence of MSH sliding clamps was obtained as the time between the initial LacI injection and the first LacI-binding event PIFE induced by LacI binding. The number of Alexa Fluor 647 fluorophores was obtained from the number of photobleaching steps as described previously (14,17). The methods utilized for data analysis may be found in the supporting information.