Nanofabricated racks of DNA for visualizing 1D diffusion of Mlh1–Pms1. (a) Illustration of some predicted structures for Mlh1–Pms1 based on structural and biochemical data17–20. The NTDs, CTDs, central pore and linker arms are indicated. (b) Diagram of the nanofabricated rack device used for making the double-tethered DNA curtains (Supplementary Methods)15. The rack consists of linear barriers to lipid diffusion, which align the lipid-tethered DNA molecules, followed by an array of antibody-coated pentagons that provide immobile anchor points for the second end of the DNA. Pattern elements are ~20 nm tall, and the bilayer is ~5 nm thick. (c) Top and bottom, YOYO1-stained λ-DNA curtains (green; 48,502 base pairs) with and without QD-tagged Mlh1–Pms1 (magenta), respectively (Supplementary Videos 1 and 2). DNA-bound proteins were not detected in reactions using incorrect antibody-epitope pairs (not shown), and QDs alone did not bind DNA (not shown). (d) Kymograms illustrating the motion of Mlh1–Pms1. The lower panel shows photocleavage (arrowhead) of a DNA during data collection. The proteins disappear from view when the DNA breaks, showing that the proteins and DNA are not adsorbed to the bilayer.

Quantitative analysis of Mlh1–Pms1 diffusion. (a) Examples of MSD plots generated from tracking the motion of Mlh1–Pms1. (b) Diffusion coefficients derived from the MSD plots and categorized according into the different nucleotide conditions used for the measurements. Diamonds (♦), the mean values; the mean ± s.d. are color coded (graph and inset; N ≥ 25 for all reported diffusion coefficients). The cumulative diffusion coefficients for Mlh1–Pms1 and the previously measured values for Msh2–Msh6 are shown together for comparison14, and the upper boundaries for the theoretical diffusion coefficients based on models either with or without a rotational component are also shown14,44. (c) Diffusion coefficients for Mlh1–Pms1 determined at different concentrations of NaCl.

Diffusion of Mlh1–Pms1 and Msh2–Msh6 along nucleosome-bound DNA. (a) Mlh1–Pms1 (magenta) is shown diffusing along a DNA bound by recombinant nucleosomes (green); white, regions of overlapping signal. The nucleosomes were labeled with QDs either after conducting the Mlh1–Pms1 diffusion experiment (upper and middle panels) or before the addition of Mlh1–Pms1 (lower panel). Examples of bypass and bounded diffusion are highlighted; identical results were obtained ± ATP. (b) Msh2–Msh6 (red) is shown diffusing on a DNA molecule with unlabeled nucleosomes (green); yellow, regions of overlap. Upper and middle panels, results with unlabeled nucleosomes. The number designations (either 1 or 2) in the middle panel indicate number of QD-tagged Msh2–Msh6 molecules trapped in each region of the kymogram (6 total). Lower panel, Msh2–Msh6 colliding with QD-labeled nucleosomes. See Supplementary Methods for additional experimental details. (c) Illustrates how the structures of Mlh1–Pms1 (left) and Msh2–Msh6 (right) may influence nucleosomal encounters18–20,24,45. The molecules are drawn to scale. The trajectory of the DNA leaving the nucleosome surface has been modified for illustrative purposes. Mlh1–Pms1 steps over the nucleosome (solid arrow). Nucleosome bypass by Msh2–Msh6 might occur by occasional hopping or two-dimensional (2D) sliding (dashed arrows)8, but neither mechanism allows free mobility on a nucleosomal array.

DNA-binding activity of fluorescently tagged Mlh1–Pms1. (a) Gel mobility shifts were performed as described in Online Methods. Reactions contained 120 nM Mlh1–Pms1 and 60 nM 5′-³²P-labeled 40-mer dsDNA. Antibodies were preincubated with Mlh1–Pms1, as indicated. (b) Nitrocellulose filter binding assays were performed with 100 pM 5′-³²P-labeled 3-kb linear plasmid at the indicated concentrations of Mlh1–Pms1, and the percentage of bound DNA was determined by dividing the background subtracted counts measured for each filter by the total amount of DNA in the reactions. (c) The effects of antibodies and antibody-labeled QDs on the binding activity of Mlh1–Pms1 (20 nM) as determined by the nitrocellulose filter-binding assay. All data points are reported as mean ± s.d.

End-dependent dissociation of Mlh1–Pms1 from DNA. (a) Kymogram of Mlh1–Pms1 (magenta) diffusing on a DNA molecule (unlabeled) anchored by both ends to the flow-cell surface. Flow is cycled on and off as indicated. When flow is on, Mlh1–Pms1 is pushed to the downstream anchored end of the DNA but does not dissociate. (b) Kymograms of Mlh1–Pms1 dissociating from the free blunt ends (generated by SfoI digest) of single-tethered DNA molecules. (c) Mlh1–Pms1 (magenta) was bound to DNA (green) stained with YOYO1 and was pushed repeatedly to the end of the molecule (as indicated) to verify that it did not dissociate. A DSB was introduced by laser illumination in the absence of flow. Upon breaking, the DNA retracts from the surface, as indicated by the sudden disappearance of the green signal. Flow was resumed to extend the broken DNA and push Mlh1–Pms1 toward the free end of the molecule. Mlh1–Pms1 immediately dissociates upon encountering the free DNA end (see inset). (d) Mlh1–Pms1 was bound to a looped DNA molecule (i) and slid along the arc formed by the DNA until stopping at the loop apex (ii). The protein remained at the DNA apex (iii and iv) but continued sliding down the DNA upon induction of a DSB (v), and the protein immediately dissociated upon reaching the newly generated free end (vi), leaving behind the naked DNA (vii). (e) Kymogram of Mlh1 bound to a double-tethered DNA molecule. In the absence of flow the proteins diffuse rapidly along the DNA, but when flow is applied, they are pushed to the end of the DNA and rapidly dissociate. Experiments in a–e were collected from isolated DNA molecules, as described14. The DNA in a and b was located with YOYO1, but the dye was removed before data acquisition to avoid unintentional photocleavage, and the experiments were conducted at 150 mM NaCl. The experiments in c, d and e were conducted at 50 mM NaCl. In all cases, identical results were obtained ±1 mM ATP.