DNA Chain Length Dependence of Formation and Dynamics of hMutS (cid:1) (cid:1) hMutL (cid:1) (cid:1) Heteroduplex Complexes*

Formation of a ternary complex between human MutS (cid:1) , MutL (cid:1) , and heteroduplex DNA has been demonstrated by surface plasmon resonance spectroscopy and electrophoretic gel shift methods. Formation of the hMutL (cid:1) (cid:1) hMutS (cid:1) (cid:1) heteroduplex complex requires a mismatch and ATP hydrolysis, and depends on DNA chain length. Ternary complex formation was supported by a 200-base pair G-T heteroduplex, a 100-base pair sub-strate was somewhat less effective, and a 41-base pair heteroduplex was inactive. As judged by surface plasmon resonance spectroscopy, ternary complexes produced with the 200-base pair G-T DNA contained (cid:1) 0.8 mol of hMutL (cid:1) /mol of heteroduplex-bound hMutS (cid:1) . Although the steady-state levels of the hMutL (cid:1) (cid:1) hMutS (cid:1) (cid:1) heteroduplex were substantial, this complex was found to turn over, as judged by surface plasmon resonance spectroscopy and electrophoretic gel shift analysis. With the former method, the majority of the complexes dissociated rapidly upon termination of protein flow, and dissociation occurred in the latter case upon challenge with competitor DNA. However, ternary complex dissociation as monitored by gel shift assay was prevented if both ends of the heteroduplex were

Repair in the Escherichia coli system is initiated by the binding of MutS to a mismatch (8 -10). Formation of a MutL⅐MutS⅐heteroduplex complex has been demonstrated by DNase I footprint analysis (11), electron microscopy (12), and surface plasmon resonance spectroscopy (SPRS) 1 (13), with assembly of this ternary complex being ATP-dependent. Several lines of evidence indicate that assembly of the ternary complex is required for subsequent steps in mismatch repair. Both MutS and MutL are required for the mismatch-dependent activation of the d(GATC) endonuclease activity of MutH, which cleaves the unmethylated strand of a hemimethylated d(GATC) site, with the ensuing strand break serving to direct repair to the unmethylated DNA strand (14). MutS and MutL are also required for the mismatch-dependent activation of DNA helicase II, which enters the helix at the strand break and initiates the excision step of repair (15).
Formation of a MutL␣⅐MutS␣⅐heteroduplex complex has been demonstrated by electrophoretic gel shift analysis with yeast mismatch repair proteins using synthetic heteroduplexes of ϳ50 base pairs (bp) in size (16,17). However, using gel shift methods and surface plasmon resonance spectroscopy, we have consistently been unable to demonstrate the corresponding ternary complex between E. coli MutS and MutL or human MutS␣ and MutL␣ and synthetic heteroduplexes of similar size. 2 This is despite the fact, as noted above, that footprint analysis, electron microscopy, and SPRS has indicated formation of a MutL⅐MutS⅐heteroduplex complex with the bacterial proteins (11)(12)(13). Since the latter experiments utilized heteroduplexes 140 bp or longer, we have examined the effect of DNA chain length on ternary complex formation using surface plasmon resonance spectroscopy and electrophoretic gel shift. We show here that efficient formation of the MutL␣⅐MutS␣⅐heteroduplex ternary complex is dependent on DNA chain length. We also show that ternary complex formation with the human proteins requires ATP, and probably its hydrolysis, and that these complexes turn over rapidly with respect to binding and release from the DNA. However, dissociation can be prevented and ternary complexes kinetically stabilized by placement of streptavidin blocks at both ends of a linear heteroduplex.

EXPERIMENTAL PROCEDURES
hMutS␣ and hMutL␣ Preparations-hMutS␣ and hMutL␣ were prepared from baculovirus constructs expressing the appropriate human cDNAs in SF9 cells. The two subunits of hMutS␣ were expressed from a single virus constructed using the Dual Bac system (Life Technologies, Inc.). Briefly, hMSH2 cDNA (provided by Bert Vogelstein, Johns Hopkins Oncology Center, Baltimore, MD) was inserted into the NcoI site just downstream of the p10 promoter, while hMSH6 cDNA (a gift from Rick Fishel, Thomas Jefferson University, Philadelphia, PA) was expressed from the polyhedrin promoter by insertion between the BamHI and SalI sites. cDNAs for human MLH1 and PMS2 (generously provided by Mike Liskay, Oregon Health Sciences University, Portland, OR) were expressed from individual viral constructs prepared using the pFastBac I system (Life Technologies, Inc.). hMLH1was expressed from the polyhedrin promoter by insertion between the BamHI and NotI sites. hPMS2 was also expressed from the polyhedrin promoter after insertion between BamHI and XbaI sites.
Infected SF9 cells for hMutS␣ isolation were grown by Kemp Biotechnologies, Inc. (Frederick, MD). Frozen cell pellets were thawed and suspended (10 ml/g of cells) in 25 mM HEPES-KOH, pH 7.5, 0.1 mM EDTA, 1 mM dithiothreitol containing 1 g/ml each of aprotinin, leupeptin, Pefabloc (Roche Molecular Biochemicals), and E64 (Peptides International). Cells were lysed with five strokes of a Dounce B pestle, and the extract supplemented with KCl to 200 mM. After centrifugation (30,000 ϫ g, 15 min), hMutS␣ was isolated from the supernatant by minor modifications of the method used previously for isolation of the HeLa cell activity (18). Recombinant hMutS␣ preparations obtained in this manner had a purity of 98% or better, were characterized by a 1:1 subunit stoichiometry, and were fully active in mismatch repair as judged by in vitro complementation of nuclear extracts derived from MSH2-deficient human cells.
For hMutL␣ purification, SF9 cells were continuously cultured in 800 ml of serum-free HyQ-SFX medium (HyClone, Inc.) at 27°C in 2.8-liter Fernbach flasks. When culture density reached 1 ϫ 10 6 /ml, cells were co-infected with hMLH1 and hPMS2 baculovirus constructs at a multiplicity of infection of 5. Infected cells were harvested 60 h later by centrifugation at 4,000 rpm for 10 min in a Sorvall RC-3B centrifuge. Cell pellets were suspended in 120 ml of 20 mM KPO 4 , pH 7.6, 5 mM KCl, and 1 mM MgCl 2 containing 0.1% phenylmethylsulfonyl fluoride (relative to a saturated solution in isopropanol) and 1 g/ml each of aprotinin, leupeptin, and E64. After incubation on ice for 10 min, cells were lysed with 20 strokes using a Dounce B pestle, the suspension adjusted to 100 mM KCl, and then clarified by centrifugation at 20,000 ϫ g for 10 min. The supernatant was frozen in liquid N 2 and stored at Ϫ80°C (fraction I). Forty ml of fraction I was thawed and loaded at 4°C onto a 5-ml heparin HiTrap column (Amersham Pharmacia Biotech) equilibrated with 25 mM HEPES-KOH, pH 7.5, 100 mM KCl, 1 mM EDTA, and 10% (v/v) glycerol at flow rate of 1.5 ml/min. After wash with starting buffer, the column was eluted with a 60-ml gradient of KCl (100 -450 mM) in 25 mM HEPES-KOH, pH 7.5, 1 mM EDTA, and 10% (v/v) glycerol. hMutL␣ fractions, which eluted at ϳ230 mM KCl, were diluted to 80 mM KCl with 20 mM KPO 4 , pH 7.4, 0.1 mM EDTA, 10% (v/v) glycerol and loaded onto a 1-ml Mono Q column (Amersham Pharmacia Biotech) equilibrated with 20 mM KPO 4 , pH 7.4, 0.1 mM EDTA, 10% (v/v) glycerol (buffer A) containing 80 mM KCl at a flow rate of 0.5 ml/min. After wash with starting buffer, the column was eluted with a 20 ml gradient of KCl (80 -380 mM) in buffer A. hMutL␣ fractions, which eluted at ϳ220 mM KCl, were diluted to 80 mM KCl with buffer A and loaded onto a Mono S column (Amersham Pharmacia Biotech) equilibrated with buffer A containing 80 mM KCl at a flow rate of 0.5 ml/min. After wash with starting buffer, the column was eluted with a 20-ml KCl gradient (80 -380 mM) in buffer A. hMutL␣ fractions, which eluted at ϳ150 mM KCl, were pooled, and aliquots frozen in liquid N 2 and stored at Ϫ80°C. Purification from 40 ml of extract yielded 1.6 mg of hMutL␣ with an MLH1:PMS2 subunit ratio of 1:1 and an estimated purity of 98%. Activity of such preparations are comparable to that of hMutL␣ isolated from HeLa cells (19), as judged by in vitro complementation of nuclear extracts derived from cells deficient in hMLH1.
DNAs-Oligodeoxyribonucleotides were purchased from Oligos Etc. (Wilsonville, OR), and when indicated were radiolabeled at the 5Јterminus with T4 polynucleotide kinase and [␥-32 P]ATP (3000 Ci/mmol, PerkinElmer Life Sciences) to a specific activity of 1 ϫ 10 6 cpm/pmol. Poly d(I)⅐d(C) was purchased from Amersham Pharmacia Biotech.
A 41-bp G-T heteroduplex used for gel shift analysis was prepared by mixing, in a 100-l volume, 80 nM top strand 5-32 P-d(AGCCGAATTT-TTAGACTCGATAGCTTGCTAGCAATTCGGCG) with 120 nM unlabeled bottom strand 5Ј-biotin-d(CGCCGAATTGCTAGCAAGCTGTCG-AGTCTAAAAATTCGGCT). Strands were annealed in 10 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 150 mM NaCl by heating at 99°C in a PerkinElmer Life Sciences Gene Amp 9600 thermocycler for 2 min and cooling to 25°C over a period of 90 min. The corresponding A⅐T homoduplex was prepared in a similar manner by annealing 5Ј-biotind(CGCCGAATTGCTAGCAAGCTATCGAGTCTAAAAATTCGGCT) with the top strand above. Identical 41-bp substrates were prepared for SPRS by annealing 1 M each of top and bottom strands described above in a 100-l volume. These 41-bp DNAs correspond to coordinates 5612-5652 of the f1MR phage DNAs used for preparation of in vitro mismatch repair substrates (9).
PCR-derived 200-bp G-T heteroduplex and homoduplex DNAs were prepared after strand separation by denaturing HPLC as described previously (20). Briefly, a biotinylated viral strand sequence obtained by strand separation of a 200-bp PCR product derived from phage f1MR1 (0.3 M) was annealed in 100 l with 0.2 M 200-nucleotide, 32 P-labeled complementary strand sequence derived by PCR from f1MR1 or f1MR3 (9). Duplexes were annealed as described above. Nonradioactive 200-bp homoduplex and heteroduplex DNAs for SPRS were prepared in a similar manner by annealing 0.5 M 5Ј-biotinylated f1MR1 viral strand sequence with 0.5 M complementary strand sequence derived from f1MR1 or f1MR3.
100-bp homoduplex and G-T heteroduplexes used for gel shift analysis were prepared by denaturing HPLC strand separation of a 150-bp PCR product derived from coordinates 5582-5732 of bacteriophages f1MR1 and f1MR3 (9). Forward and reverse primers for PCR were d(CGCTTTCTTCCCTTCCTTTCTCG) and d(AAGTTTTTTGGGGTCG-AGGT). The 32 P-labeled, 150-residue viral strand sequence from f1MR1 was combined with the complementary sequence (0.45 M each in 100 l) prepared from f1MR1 or f1MR3 to yield homoduplex or G-T heteroduplex, respectively. DNAs were annealed as described above in 20 mM Tris acetate, pH 7.9, 10 mM magnesium acetate, 50 mM potassium acetate, and 1 mM dithiothreitol. Resulting duplexes were cleaved with 20 units of BanII for 1 h at 37°C. After dilution to 400 l with buffer B (10 mM Tris-HCl, pH 8.0, 1 mM EDTA) containing 300 mM NaCl, the DNA was loaded onto a Gen-Pak Fax column (4.6 ϫ 100 mm) equilibrated with this buffer at a flow rate of 0.55 ml/min. After washing 5 min with buffer B containing 0.3 M NaCl, the DNA was eluted with a gradient of NaCl (0.3-1 M) in buffer B over a 35-min period. 5Ј-Biotinylated 110-bp DNAs for surface plasmon resonance spectroscopy were prepared in a similar manner from PCR products derived from region 5570 -5732 of f1MR1 and f1MR3 using the same reverse primer described above and 5Ј-biotin-d(GCCCGCTCCTTTCGCTTTCT) as forward primer. The biotinylated viral strand sequence from f1MR1 was annealed with the complementary strand sequence (1 M each in 100 l) prepared from f1MR1 or f1MR3, subjected to BanII cleavage, and the 110-bp heteroduplex and homoduplex purified as described above.
3Ј-32 P-Labeled 200 bp homoduplex and heteroduplex DNAs that were tagged with 5Ј-biotin at both ends were prepared as described previously (20).
Western and Southern Blotting of Gel-shifted Complexes-A nitrocellulose membrane was placed on polyacrylamide gels and a NA45 DEAE-cellulose membrane (Schleicher & Schuell) placed on top of the nitrocellulose. Protein and DNA were then electrophoretically transferred in 50 mM Tris, 376 mM glycine, and 20% methanol for 1 h at 100V. Under these conditions proteins are retained by the nitrocellulose, but DNA passes through to be retained on the DEAE membrane. Radiolabeled DNA was visualized by autoradiography of the DEAE membrane. Mismatch repair polypeptides were identified by Western blot using monoclonal anti-hMLH1 (PharMingen), monoclonal anti-hPMS2 (PharMingen), monoclonal anti-hMSH2 (Calbiochem), or goat anti-MSH6 (N-20, Santa Cruz Biotechnology Inc.). Nitrocellulose membranes were incubated in blocking buffer (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 5% (w/v) nonfat dried milk) for 30 min, followed by a 1-h incubation with the appropriate antibody diluted 1/500 into blocking buffer. After three 10-min washes with blocking buffer, membranes were incubated for 30 min with a 1/500 dilution of anti-mouse Ig horseradish peroxidase conjugate (Amersham Pharmacia Biotech) or anti-goat IgG peroxidase conjugate (Sigma) as appropriate. Immune complexes were detected by ECL. When indicated nitrocellulose membranes were stripped by incubating a 65°C for 4 h in 62.5 mM Tris-HCl, pH 6.8, 100 mM 2-mercaptoethanol, 2% (w/v) sodium dodecyl sulfate. After washing three times for 10 min with blocking buffer, the membrane was probed with a second antibody.
Surface Plasmon Resonance Spectroscopy-Surface plasmon resonance measurements were performed on a BIAcore 2000. Streptavidin sensor chips were derivatized with the 41-, 110-, and 200-bp homoduplex or G-T heteroduplex DNAs described above, in which one strand was tagged with a 5Ј-terminal biotin. Solutions contained 20 mM Tris-HCl, pH 7.6, 1 mM dithiothreitol, 0.005% surfactant P20, 5 mM MgCl 2 , 100 mM KCl and ATP, and hMutS␣ and hMutL␣ as indicated. Measurements were performed at 25°C at a flow rate of 20 l/min, and samples were maintained at 4°C prior to injection. Dissociation kinetics were monitored by coinjecting 120 l of ATP supplemented reaction buffer immediately following protein association. SA chips were regenerated by a 20-l injection of 0.5% sodium dodecyl sulfate.
DNA-bound protein monitored by SPRS is expressed as equivalents of MutS␣ (M r ϭ 258,000; Refs. 18 and 21-24) bound/mol of chip-bound DNA. This value was calculated from the mass ratio, which is given by (RU exp )/(0.79 ϫ RU DNA ), where RU exp is the output signal due to protein binding and RU DNA corresponds to the amount of DNA bound to the chip in resonance units. The factor 0.79 corrects for the fact that the refractive index increment for a typical protein is 79% of that obtained with an equivalent mass of DNA (25). Mass ratio values were converted to molar ratios using the molecular weights of the DNA and protein in question. Since this method is based on several assumptions (for example, that the relative refractive index increment cited above is generally valid and that all chip-bound DNA is equally accessible to protein flow), it can only be regarded as approximate. However, we have also used this method to estimate relative binding stoichiometries of hMutS␣ and hMutL␣ in ternary complexes with DNA, and these values should be quite accurate.

Mismatch-, ATP-, and Chain Length-dependent Formation of a hMutL␣⅐hMutS␣⅐Heteroduplex
Complex-Ternary complexes of yeast MutS␣ and MutL␣ with synthetic heteroduplexes of ϳ50 bp in size have been demonstrated by electrophoretic gel shift (16,17). However, we have been unable to reproducibly detect ternary complexes involving bacterial MutS and MutL, or human MutS␣ and MutL␣, utilizing synthetic heteroduplexes 41 bp in length. Since there is abundant evidence for ternary complex formation between the bacterial mismatch repair proteins and heteroduplexes of 143 bp or longer (11)(12)(13), we reasoned that this problem might be due to the small size of synthetic heteroduplexes.
Using surface plasmon resonance spectroscopy, specific binding of hMutS␣ to a 200-bp G-T heteroduplex (see "Experimental Procedures") was evident in the presence of ATP⅐Mg 2ϩ , with the heteroduplex signal approximately 3 times that observed with an otherwise identical A⅐T homoduplex (Fig. 1). Although hMutL␣ did not bind detectably to the 200-bp heteroduplex under these conditions, passage of a mixture of hMutS␣ and hMutL␣ over the chip resulted in a substantial enhancement of the mass of heteroduplex-bound protein, as compared with that observed with hMutS␣ alone. This mass enhancement is mismatch-dependent since it was not observed with the A⅐T homoduplex control, and experiments described below demonstrate that it is not a simple consquence of increased hMutS␣ binding due to presence of hMutL␣. This effect requires the simultaneous presence of hMutS␣ and hMutL␣ since no enhancement of heteroduplex-bound protein was observed in experiments in which the two proteins were passed over the chip in a sequential manner (data not shown).
The hMutL␣-dependent mass enhancement was only observed in the presence of ATP, where it displayed a strong dependence on DNA chain length (Fig. 2). In the absence of ATP (lower panel), ϳ1 hMutS␣ heterodimer was bound/41-bp G-T heteroduplex, and this value increased to approximately two equivalents with the 200-bp heteroduplex. Under these conditions in the absence of ATP, the protein mass bound by 41-, 100-, and 200-bp heteroduplexes was unaffected by inclusion of hMutL␣ along with hMutS␣, as compared with that observed with hMutS␣ alone.
As observed previously (18,20,26), the presence of ATP resulted in a dramatic reduction in hMutS␣ binding to the 41-bp heteroduplex, and the presence of hMutL␣ was without effect (Fig. 2, compare upper and lower panels). However, the extent of hMutS␣ binding to 100-and 200-bp heteroduplexes in the presence of ATP was similar to that observed in the absence of nucleotide, and the presence of hMutL␣ resulted in a substantial enhancement of protein mass bound by these two heteroduplex DNAs. Based on experiments described below, and previous observations with bacterial and yeast mismatch repair proteins (11-13, 16, 17), we have concluded that this enhancement of heteroduplex-bound protein mass reflects mismatch-, ATP-, and hMutS␣-dependent presence of hMutL␣ in a nucleoprotein complex with heteroduplex DNA. The stoichiometries of formation of this hMutL␣⅐hMutS␣⅐heteroduplex ternary complex will be considered below.
Ternary complex formation was demonstrable in the presence of ATP⅐Mg 2ϩ , but we have been unable to detect a hMutL␣dependent increase in the mass of heteroduplex-bound protein in the presence of AMPPNP⅐Mg 2ϩ (data not shown). This finding, which is similar to previous observations with bacterial MutS and MutL (13), strongly suggests that ternary complex formation is dependent upon ATP hydrolysis by one or both proteins.

Apparent hMutS␣ and hMutL␣ Affinities and Stoichiometry of Ternary Complex Formation-Ternary complex formation
requires ATP⅐Mg 2ϩ , conditions that reduce the affinity of hMutS␣ for a mispair (18,20,26). The apparent specific affinity of hMutS␣ for the 200-bp G-T heteroduplex and A⅐T homoduplex in the presence of ATP was estimated by SPRS (Fig. 3,  upper panel). Under these conditions, hMutS␣⅐heteroduplex and hMutS␣⅐homoduplex formation was hyperbolic, with apparent K d values of 205 and 420 nM, respectively. These data were also evaluated after subtraction of the hMutS␣⅐homoduplex values from those observed with the heteroduplex in order to correct interactions with heteroduplex for mismatch-independent binding. Correction in this manner also yielded an excellent hyperbolic fit with an apparent K d of 140 nM and an asymptotic value of 3 equivalents of the MSH2⅐MSH6 heterodimer bound per DNA at saturation (Fig. 3, upper panel). Although the stoichiometry of hMutS␣⅐heteroduplex formation calculated from SPRS can be regarded as only approximate (see "Experimental Procedures"), the finding that the limiting stoichiometry of heteroduplex binding exceeds unity, even after correction for homoduplex effects, may seem surprising given that the DNA contains a single mismatch. There are several possible explanations for this effect. For example, multiple copies of hMutS␣ may oligomerize on a heteroduplex in a mismatch-dependent reaction. A second possibility is based on the observation that hMutS␣ can leave a mismatch in the presence of ATP by movement along the helix contour to dissociate at DNA termini (20,27,28). Such a mechanism would account for the loading of multiple copies of hMutS␣ onto a heteroduplex, provided that the migrating species fail to reach a DNA terminus before another hMutS␣ binds to the mispair (27).
Like hMutS␣ binding to the 200 bp G-T heteroduplex in the presence of ATP, ternary complex formation was a hyperbolic function of hMutL␣ concentration (Fig. 3, lower panel), characterized by an apparent K d of 70 nM in the presence of a hMutS␣ concentration of 200 nM. At this MutS␣ concentration, which is approximately equal to the K d for binding to heteroduplex DNA (above), the SPRS results shown in Figs. 1-3 indicate a relative ternary complex stoichiometry of ϳ0.6 and 0.8 mol of hMutL␣/mol of hMutS␣ for the 100-and 200-bp heteroduplexes, respectively. This calculation is based on the extent of hMutS␣ binding to heteroduplex DNA after correction for nonspecific complexes formed with homoduplex controls (Figs. 1 and 3). It is noteworthy that a similar hMutL␣-dependent enhancement of heteroduplex-bound mass was observed at 800 nM hMutS␣ (4 times the K d ), providing additional evidence that the observed mass increase is due to hMutL␣ binding hMutS␣⅐hMutL␣⅐Heteroduplex Complexes rather than increased hMutS␣ binding in the presence of the MutL homolog. It is important to note that we have consistently found the 110-bp heteroduplex to be somewhat less effective than the 200-bp DNA in supporting ternary complex formation, as judged by the additional mass enhancement observed in the presence of hMutL␣.
hMutS␣⅐hMutL␣⅐Heteroduplex Complexes Are Dynamic-The kinetic lifetimes of hMutL␣⅐hMutS␣⅐heteroduplex complexes were monitored with SPRS by terminating protein flow and continuing wash with reaction buffer containing ATP and Mg 2ϩ (Fig. 4, upper curve). Dissociation of the ternary complexes from the 200-bp G-T heteroduplex was multiphasic, with decay curves fitting well to a sum of two exponentials. The major amplitude (Ϸ60% of the complexes) dissociated rapidly (t1 ⁄2 ϳ 1 s), a second component (Ϸ20%) dissociated more slowly (t1 ⁄2 ϳ 10 s), and the residual (Ϸ20%) dissociated so slowly that a rate could not be estimated. Complexes prepared with homo-duplex DNA in the presence of hMutS␣ and hMutL␣ displayed similar multiphasic dissociation kinetics (Fig. 4 lower curve). The major species (Ϸ60%) dissociated rapidly with a t1 ⁄2 of ϳ1 s, the second component (Ϸ16%) dissociated more slowly (t1 ⁄2 ϳ 19 s), with the residual (Ϸ24%) dissociating too slowly to permit an estimate of lifetime.
The relative extents of binding to heteroduplex and homoduplex DNAs (Figs. 1, 3, and 4) indicate that 60 -70% of the protein mass bound to the heteroduplex under these conditions is dependent on the presence of a single mismatch. We therefore think it likely that the more rapidly dissociating heteroduplex species (t1 ⁄2 values of 1 and 10 s, 80% by mass) correspond to several classes of mismatch-dependent ternary complexes. This implies that the levels of these species observed by SPRS prior to termination of protein flow correspond to a dynamic steady state, i.e. the complexes are turning over rapidly via dissociation and reassociation. It is important to emphasize that the multiphasic dissociation kinetics observed with both heteroduplex and homoduplex DNAs imply the existence of several distinct types of specific and nonspecific complexes.
hMutL␣⅐hMutS␣⅐Heteroduplex Ternary Complexes by Gel Shift Analysis-The chain length dependence of formation and the nature of hMutL␣⅐hMutS␣⅐heteroduplex ternary complexes was also examined by electrophoretic gel shift assay. As observed by SPRS, electrophoretic assay indicated that hMutL␣ did not bind detectably to 41-, 100-, or 200-bp G-T heteroduplexes (Fig. 5). In the presence of hMutS␣, specific complexes were evident with each of these heteroduplexes, and the presence of both proteins led to production of one or more supershifted species. A hMutL␣-dependent supershifted complex was produced with the 41-bp heteroduplex, as well as its homoduplex control. Although production of this species required presence of hMutS␣, the lack of a mismatch requirement indicates that it is nonspecific in nature. Three supershifted species were observed with 100-and 200-bp DNAs in the presence of hMutS␣ and hMutL␣ (these were in addition to the hMutS␣⅐heteroduplex binary complex, which was evident at a low level with the 200-bp substrate). Two of these (Fig. 5, asterisks) were produced with both homoduplex and heteroduplex, and as observed with 41-bp DNAs, production of these nonspecific complexes was dependent on the presence of both hMutS␣ and hMutL␣. However, with both 100-and 200-bp DNAs, a heteroduplex-specific, supershifted complex was also produced (Fig. 5, arrows). Combined Southern and Western blot analysis confirmed the presence of both hMutS␣ and hMutL␣ in specific and nonspecific ternary complexes produced with the 100-bp G-T heteroduplex (Fig. 6).
These observations confirm the chain length dependence of specific ternary complex formation observed in SPRS experiments. By contrast, nonspecific complexes of the sort observed by gel shift assay were not detected as a mass enhancement in SPRS experiments with homoduplex DNA (Fig. 1). The reason for this is not clear, but as noted above, the multiphasic dissociation kinetics observed by SPRS with homoduplex DNA could be indicative of several classes of nonspecific complex.
The stability of ternary complexes was evaluated by challenge of preformed complexes with polyd(I)⅐d(C). As shown in Fig. 7, polyd(I)⅐d(C) challenge resulted in a dramatic reduction in ternary complex formation with the 200-bp G-T (lanes 4 and  5). The yield of binary hMutS␣⅐heteroduplex complexes was also extremely low under these conditions when ATP was present in the reaction. However, distinct results were obtained when the two ends of the duplex were blocked with biotinstreptavidin complexes. The presence of end blocks at both heteroduplex termini stabilized binary hMutS␣⅐heteroduplex hMutS␣⅐hMutL␣⅐Heteroduplex Complexes complexes in the presence of ATP (compare lanes 2 and 3 with lanes 8 and 9), confirming previous observations in this respect (20,27).ThebiterminalendblockalsostabilizedhMutL␣⅐hMutS␣⅐ heteroduplex ternary complexes, but a single end block did not; that fraction of the heteroduplex that contained only one end block was recovered as free DNA after polyd(I)⅐d(C) challenge, whereas heteroduplexes with streptavidin-biotin blocks at both duplex termini were not (compare lanes 10 and 11). These observations are consistent with the conclusion discussed above that ternary complexes are dynamic in nature.
Due to the size of the heteroduplex and the presence of streptavidin end blocks, the supershifted complex observed in the presence of hMutS␣ and hMutL␣ was not resolved into specific and nonspecific components (lanes 10 and 11, compare with lane 9). However, a surprising effect of streptavidin-biotin end blocks became evident upon examination of nonspecific interactions with the 200-bp homoduplex. Two classes of nonspecific complex are produced with homoduplex DNA in the presence of hMutS␣, hMutL␣, and ATP (Fig. 5, lane 8; Fig. 7, lane 6). Unexpectedly, the corresponding nonspecific complexes were not detectable under conditions where the homoduplex molecules were blocked at both ends with streptavidin-biotin complexes (lane 12). These observations suggest that presence of free duplex DNA termini have an important role in the production of nonspecific complexes that are observed by gel shift assay. DISCUSSION Although MutS homologs bind readily to small synthetic heteroduplexes (10, 18, 23, 24, 29 -31), the experiments described here indicate that formation of the hMutL␣⅐hMutS␣⅐ heteroduplex requires ATP (and probably its hydrolysis) and depends on DNA chain length. The 200-bp heteroduplex used in this report supports efficient ternary complex formation, 100-and 110-bp heteroduplexes appear to be somewhat less effective in this regard, and we have been unable to detect mismatch-dependent ternary complex formation with a 41-bp heteroduplex substrate. We have also obtained similar results with E. coli MutS and MutL. 3 The SPRS and gel shift analyses described here show that the hMutL␣⅐hMutS␣⅐heteroduplex ternary complex is dynamic, undergoing rapid dissociation and reassociation in the presence of ATP⅐Mg 2ϩ at near physiological ionic strength. In fact, the kinetics of dissociation of ternary complexes are similar to those observed with the binary hMutS␣⅐heteroduplex in the presence of ATP. As observed for ternary complexes (Fig. 4), dissociation of binary complexes as monitored by SPRS was multiphasic (data not shown); the major amplitude dissociated rapidly (Ϸ60% of complexes, t1 ⁄2 ϭ 3 s), a second species dissociated more slowly (Ϸ26%, t1 ⁄2 ϭ 50 s), and a third component (Ϸ14%) dissociated too slowly to determine an accurate rate. These observations are similar to those obtained previously by Galio et al. (13) in SPRS studies of bacterial MutS and MutL. As described here for hMutS␣ and hMutL␣, these earlier studies led to the conclusion that ternary complexes of MutS, and MutL, with a 149-bp heteroduplex turn over rapidly as compared with the lifetime of MutS⅐heteroduplex complexes. In the case of the human system, we have also shown that hMutL␣⅐ hMutS␣⅐heteroduplex complexes can be kinetically stabilized by placement of a physical block at each end of a linear DNA. The simplest interpretation of this finding is that turnover of the ternary complex depends on movement of one or both mismatch repair proteins along the helix with dissociation occurring at free ends. By contrast, the recent work of Hsieh and colleagues (32) has led to the conclusion that bacterial MutL stabilizes mismatch-bound MutS in the presence of ATP, resulting in a much longer lifetime for the MutL⅐MutS⅐heteroduplex ternary complex as compared with that of the MutS⅐ heteroduplex (32). The basis of these differing conclusions is uncertain, although different methods were used to monitor dissociation kinetics. Our conclusions and those of Galio et al. (13) are based on real time analysis using SPRS, whereas Schofield et al. (32) monitored dissocation kinetics by gel shift assay after addition of a heteroduplex trap.
The molecular basis of the chain length dependence for ternary complex formation is not clear, but there are a number of potential explanations for this effect. One possibility is that the presence of flanking homoduplex is necessary to accommodate both hMutS␣ and hMutL␣. The formation of nucleoprotein complexes containing hMutL␣ that we have detected require hMutS␣; however, the ATPase of bacterial MutL is known to be activated in the absence of MutS by single strands and to a After electrophoresis, protein and DNA were electrotransferred to nitrocellulose and DEAE membranes (see "Experimental Procedures"). DNA bound to the DEAE membrane was visualized by autoradiography and nitrocellulose-bound mismatch repair polypeptides identified by Western blot. In the experiment shown at the top, the nitrocellulose membrane was probed as indicated with anti-MLH1, and then with anti-MSH6 after stripping. The experiments shown in the center and at the bottom were performed in a similar manner except that stripping was not used; parallel gels were examined individually for DNA and MSH2, or for DNA and PMS2. Specific complexes are indicated by arrows and nonspecific complexes by asterisks. In the experiment shown at the top, the nonspecific complex that runs more slowly than the specific component is barely visible in the DNA and MSH6 panels, but is evident in the MLH1 panel. The faster migrating species in lane 1 of each DNA panel corresponds to the hMutS␣⅐DNA complex. That portion of the gel where free DNA runs is not shown.

FIG. 7. The hMutL␣⅐hMutS␣⅐heteroduplex ternary complex is dynamic.
Gel shift analysis was performed as described under "Experimental Procedures" in the presence or absence of 0.5 mM ATP using 3Ј-32 P-labeled 200-bp G-T heteroduplex (lanes 1-5 and 7-11) or a control 200-bp A⅐T homoduplex (lanes 6 and 12). Both DNAs contained a 5Ј-terminal biotin on each DNA strand. Reactions in the right panel contained 0.5 mg/ml streptavidin and were preincubated 10 min prior to addition of mismatch repair proteins to allow conjugation of biotin. Reactions were initiated by addition of hMutS␣ and hMutL␣ as indicated, to a final concentration of 200 nM. After 10 min at room temperature, a poly(dI)⅐d(C) competitor was added to a final concentration of 25 g/ml. Reactions were terminated after an additional 5-min incubation subjected to polyacrylamide gel electrophoresis (see "Experimental Procedures"). The location of the specific ternary complex formed in the absence of streptavidin is shown by an arrow. Mobilities of free DNA with 0, 1, or 2 bound streptavidin molecules are also indicated. lesser extent by duplex DNA (33,34), implying presence of a DNA binding center. MutL and hMutL␣ are large asymmetric proteins (Stokes radii of 61 and 74 Å, respectively (Refs. 11 and 19)) and are potentially capable of occluding a substantial segment of helix. It is also possible that ternary complex formation involves oligomerization (or polymerization along the helix) of hMutS␣ or hMutL␣. Indeed, we have concluded that ternary complexes with 200-bp heteroduplex DNA contain several copies of each heterodimer, but as discussed above, the presence of multiple protein copies can also be explained by a mechanism that invokes movement of repair protein complexes along the helix contour. A third interesting possibility is that the chain length requirement is indicative of a major DNA conformational transition associated with ternary complex formation, e.g. the opening of a significant length of helix or the introduction of a substantial bend, perhaps due to a partial wrapping of DNA about one of the repair activities.