Single-molecule imaging reveals the mechanism of Exo1 regulation by single-stranded DNA binding proteins

Abstract

Exonuclease 1 (Exo1) is a 5'→3' exonuclease and 5'-flap endonuclease that plays a critical role in multiple eukaryotic DNA repair pathways. Exo1 processing at DNA nicks and double-strand breaks creates long stretches of single-stranded DNA, which are rapidly bound by replication protein A (RPA) and other single-stranded DNA binding proteins (SSBs). Here, we use single-molecule fluorescence imaging and quantitative cell biology approaches to reveal the interplay between Exo1 and SSBs. Both human and yeast Exo1 are processive nucleases on their own. RPA rapidly strips Exo1 from DNA, and this activity is dependent on at least three RPA-encoded single-stranded DNA binding domains. Furthermore, we show that ablation of RPA in human cells increases Exo1 recruitment to damage sites. In contrast, the sensor of single-stranded DNA complex 1-a recently identified human SSB that promotes DNA resection during homologous recombination-supports processive resection by Exo1. Although RPA rapidly turns over Exo1, multiple cycles of nuclease rebinding at the same DNA site can still support limited DNA processing. These results reveal the role of single-stranded DNA binding proteins in controlling Exo1-catalyzed resection with implications for how Exo1 is regulated during DNA repair in eukaryotic cells.

Keywords: DNA curtains; DNA repair; nuclease; resection; single-molecule.

Fig. 1.

hExo1 is a processive DNA nuclease. (A) Schematic of the DNA curtains assay with hExo1. The flowcell surface is passivated with a lipid bilayer. DNA is affixed to the lipid bilayer, organized at nano-fabricated barriers, and extended to ∼85% of its B-form contour length. (B) Example of a DNA curtain (green) with fluorescent hExo1 molecules (magenta) in the presence (Upper) and absence (Lower) of buffer flow. Nearly all hExo1 molecules retracted with the DNA when buffer flow is turned off (Lower). (C) A histogram of the positions of individual hExo1 molecules bound to DNA (n = 435 molecules). The red line is a single-Gaussian fit to the data (the mean of the fit is 48 ± 2 kb), and the error bars indicate the SD obtained via bootstrap analysis (73). hExo1 preferentially binds the free 3′-ssDNA end but can also engage internal DNA sites. (D) Kymograph (Upper) and the corresponding particle-tracking trace (Lower) of a single hExo1 resecting from a DNA end (arrowhead indicates dissociation). (E) Kymograph (Upper) and corresponding trace (Lower) of nuclease-dead hExo1(D78A/D173A). (F) Box plots of velocities of WT hExo1 from ends (magenta, mean velocity = 8.4 ± 5.9 bp/s, n = 75) and nicks (orange, mean velocity = 9.0 ± 3.9 bp/s, n = 38), as well as for the nuclease-dead mutant (black, mean velocity = 0.1 ± 0.5 bp/s, n = 19). (G) hExo1 is a processive nuclease from both ends (magenta, mean processivity = 6.0 ± 2.9 kb, n = 75) and nicks (orange, processivity = 7.2 ± 4.2 kb, n = 36). The nuclease-dead mutant does not move (black, processivity = 0.01 ± 0.3 kb, n = 19). The velocities and processivities from nicks and ends are statistically indistinguishable (P = 0.57 for velocities, P = 0.09 for processivities) but are different from the nuclease-dead mutant (black, ***P = 2.1 × 10⁻⁸ for velocity, ***P = 2.7 × 10⁻¹⁴ for processivity). Box plots indicate the median, 10th, and 90th percentiles of the distribution. (H–J) hExo1 lifetimes at ends (n = 75), nicks (n = 39), and with nuclease-dead hExo1 (n = 19). The red line is a single exponential fit to the data. As ∼50% of the molecules still remained on the DNA after our 40-min observation window, we report the lower estimate of the hExo1 half-lives (>1,800 s for hExo1 and >1,400 s for nuclease-dead hExo1).

Fig. S1.

Human Exo1-biotin (hExo1-bio) purification and labeling. (A) Purification scheme for hExo1-bio. (B) SDS/PAGE gel showing hExo1-bio and hExo1-bio + streptavidin. Gel shift of hExo1-bio-streptavidin conjugates are indicated. The complete disappearance of the hExo1-bio band indicates that nearly 100% of the purified nucleases are biotinylated. (C) Resection assay with untagged WT hExo1 (500 pM), hExo1-bio (500 pM), or hExo1-bio (500 pM) + streptavidin (1 µg). Proteins were incubated in imaging buffer with 30 ng linearized 4.4-kb DNA (4-nt 3′ overhang) for 30 min at 37 °C. Resected DNA was separated on a 1% agarose gel and stained with SYBR green. The resection protocol has also been described previously (21). Together, these assays indicate that streptavidin-conjugated hExo1-bio retains full resection activity. (D) Snapshots at indicated times (Upper) and single-particle tracking of two representative trajectories of resection by CF488-anti-biotin–labeled hExo1 (Lower). In both trajectories, hExo1 transitions between a resecting and a paused state. These results indicate that both states are intrinsic to hExo1 and are not dependent on the nature of the fluorophore.

Fig. S2.

hExo1 requires a divalent cation to move on DNA. (A) Kymograph of hExo1-bio (magenta) in the presence of 1 mM EDTA. The white arrow indicates when the protein dissociated from DNA. (B) Example single-molecule trajectories of hExo1-bio with 1 mM EDTA. The raw data are displayed in gray, and the smoothed data are shown in black (>5-s boxcar averaging sliding window). As expected, hExo1 did not move in the absence of Mg⁺², which is required for nuclease activity.

Fig. S3.

Human Exo1-Flag (hExo1-Flag) resects DNA similarly to hExo1-bio. (A) Purification scheme and SDS/PAGE gel showing recombinant hExo1-Flag. (B) Cartoon illustration of the in situ hExo1-Flag labeling strategy. First, unlabeled hExo1-Flag is loaded on the DNA and excess protein is flushed out. Second, anti-Flag antibody-conjugated QDs are injected into the flowcell. This guarantees that hExo1 is labeled with, at most, one QD. (C) Kymograph (Upper) and single-particle trace of a resecting hExo1-Flag. Arrowheads indicate hExo1-Flag dissociation. The sequential hExo1-Flag and QD injection scheme is also shown in the kymograph. (D) Comparison of hExo1 velocity and (E) processivity. The mean hExo1-Flag velocity (9.8 ± 5.2 bp/s; n = 50) and processivity (5.6 ± 2.7 kb; n = 50) was statistically similar to hExo1-bio (P = 0.18 and P = 0.15, respectively). Within our experimental resolution, 78% (n = 39/50) of hExo1-Flag molecules paused at least once, with the remaining 22% (n = 11/50) resecting without pausing. Of those that paused, 44% (n = 17/39) paused before resection and 85% (n = 33/39) paused after resection.

Fig. S4.

hExo1 is a processive nuclease at a higher ionic strength. (A) Kymograph (Upper) and the corresponding particle-tracking trace (Lower) of a single hExo1-bio resecting a free DNA end in the presence of I = 153 mM total ionic strength (imaging buffer with 130 mM NaCl). hExo1 resects ∼8 kb and pauses before dissociating from the DNA (arrowhead indicates dissociation). (B) Box plots of hExo1 velocities in a buffer with I = 83 mM (magenta, mean velocity = 8.6 ± 5.3 bp/s, n = 113) or with I = 153 mM (orange, mean velocity = 7.2 ± 3.7 bp/s, n = 33). (C) hExo1 is a processive nuclease at both low (I = 83 mM, magenta, mean processivity = 6.4 ± 3.4 kb, n = 111) and high (I = 153 mM, orange, processivity = 5.8 ± 2.6 kb, n = 35) total ionic strengths. The velocities and processivities from both nicks and ends are statistically indistinguishable (P = 0.17 for velocities, P = 0.38 for processivities). Box plots indicate the median, 10th, and 90th percentiles of the distribution.

Fig. S5.

hExo1 resection is sequence independent. (A) GC content of our DNA substrate (derived from λ-phage) measured using a 1-kb sliding window from cosL to cosR. The GC content is increased on the cosL side relative to the cosR side. (B) Model of substrates used for testing sequence-dependent hExo1 resection activity. Oligonucleotides were ligated to the ends of the DNA to biotinylate one end and create a 3′-ssDNA overhang on the other end. The two substrates load hExo1 on opposite ends of the λ-DNA. (C) Velocities (Left) and processivities (Right) of hExo1 molecules that started at the ends of substrates 1 (magenta) and substrate 2 (black). The boxplot represents the 10th, median, and 90th percentiles of each distribution. Substrate 1 is used throughout the manuscript, and its velocity and processivity are reported in Fig. 1. For substrate 2, the velocity of hExo1 was 6.3 ± 4 bp/s (n = 31) and the processivity was 6.0 ± 3.2 kb (n = 33). The velocities and processivity were not statistically different between the two substrates (P = 0.07 for velocity, P = 0.98 for processivity). Within our resolution, there are no major differences in hExo1 behavior on substrate 1 vs. substrate 2.

Fig. 2.

RPA, but not SOSS1, rapidly dissociates hExo1 from DNA. (A) Kymograph of hExo1 displacement (magenta, Top) by hRPA-GFP (green, Middle). (Bottom) Merged images. The orange line indicates when hRPA-GFP enters the flowcell, and white arrowheads indicate hExo1 dissociation. (B) Kymograph of hExo1 (magenta, Top) with injection of fluorescent SOSS1 (green, Middle). (Bottom) Merged images, and the orange line indicates when SOSS1 entered the flowcell. (C) Lifetime of DNA end-bound hExo1 in the presence of 1 nM RPA (half-life = 18 ± 1 s, n = 90) and (D) 1 nM SOSS1 (half-life >2,000 s, n = 52). Red lines, single exponential fits to the data. (E) Distribution of hExo1 velocities (Left; 9.8 ± 4.5 bp/s, n = 47) and processivities (Right; 5.5 ± 3.0 kb, n = 41) with 1 nM SOSS1 in the imaging buffer. Box plots indicate the median, 10th, and 90th percentiles of the distributions. (F) Diagram of hRPA, SOSS1, and various truncations used in this study. DNA binding domains are shown in light green (hRPA) or blue (SOSS1). (G) Half-lives of hExo1 in the presence of several hRPA truncations, SOSS1, or a mock injection containing RPA storage buffer. Right column indicates the half-lives normalized to those of hExo1 with hRPA.

Fig. S6.

hExo1 is rapidly removed by both human and yeast RPA. (A) Kymographs of hExo1-bio (magenta) when 1 nM hRPA (Left) or 1 nM yRPA (Right) is injected into the flowcell. In both kymographs, the WT RPAs are not labeled. The dashed line represents the time when RPA is injected and white arrowheads mark hExo1 dissociation events. Representative kymographs of (B) hExo1-bio at a high ionic strength (I = 153 mM) or (C) hExo1(D78A/D173A)-bio (magenta, Top) plus hRPA-GFP (green, Middle). (Bottom) Merged images. hRPA displaces WT and nuclease-dead hExo1. (D) Lifetime of hExo1 on DNA with various RPAs. Red line, fit to an exponential model. Table S1 summarizes hExo1 half-lives in the presence of various RPAs and prokaryotic SSBs.

Fig. S7.

Unlabeled hExo1 is inhibited by hRPA. (A) Cartoon illustration of the experiment (Upper). To monitor resection catalyzed by unlabeled hExo1, the DNA substrate was prepared with a 3′-72 nt polyT and was terminated with a digoxigenin (dig, white square in cartoon illustration). The 3′-ssDNA end was labeled with an anti-dig conjugated QD. hExo1-catalyzed resection converts dsDNA to ssDNA, which appears as an overall shortening of the DNA at these flow rates. Kymograph (Lower) hExo1-catalyzed resection on naked DNA. (B) Cartoon illustration of the experiment as above (Upper) after injection of 1 nM hRPA. After injection of hRPA, hExo1 is rapidly displaced by hRPA, and the QD does not move. Kymograph (Lower) shows unlabeled hExo1 resection after injection of 1nM RPA (orange line). (C) Velocity (Left) and processivity (Right) of the QD-labeled ssDNA in the absence (red) or with 1 nM RPA (green). The velocity of the QD was 1.9 ± 1.6 nm/s (n = 60) in the absence of RPA and 0.2 ± 1.6 nm/s (n = 39) with RPA. Likewise, the processivity was 0.9 ± 0.7 μm (n = 54) in the absence of RPA and 0.1 ± 0.3 μm (n = 39) in the presence of RPA. These were significantly different (**P < 0.01), indicating that RPA displaces unlabeled hExo1.

Fig. S8.

Effects of SSB and gp32 on hExo1 resection. (A) Kymograph of hExo1 displacement by WT SSB. The dashed line indicates when SSB was injected and the white arrow indicates hExo1 dissociation. (B) Kymograph (Upper) and corresponding particle-tracking trace (Lower) of hExo1 on injection of gp32. Green and red lines indicate the start and stopping point of the molecule, respectively. (C) Lifetime of hExo1 in the presence of 1nM SSB (half-life = 74 ± 1.3 s, n = 31) and (D) 1 nM gp32 (half-life = 790 ± 90 s, n = 34). The red lines are single exponential fits of the data. Table S1 further summarizes the fit parameters. (E) Velocity (Left) and processivity (Right) of hExo1 in the absence (purple) or presence (gray) of gp32. The velocity of hExo1 in the presence of gp32 was 7.4 ± 3.8 bp/s (n = 27, P = 0.06), whereas the processivity was 5.5 ± 1.8 kb (n = 29, P = 0.09). These values were not statistically different from those of hExo1 in the absence of SSBs.

Fig. 3.

hRPA depletion increases hExo1 localization to DNA break sites but decreases resection in human cells. (A) Representative images of laser-induced GFP-hExo1 foci in the presence of siControl or siRPA2. The white circle indicates site of laser damage. (B) Quantification of A with at least 20 cells per biological replicate. (C) Western blot showing levels of RPA2 after siRPA2 treatment. Loading control: β-tubulin. (D) Quantification of EdU+ cells (S phase) in the presence and absence of RPA2. (E) Quantification of hExo1 ChIP efficiency at DNA sites that are 80 or 800 nt downstream of two different AsiSI-induced breaks (DSB1 and DSB2). (F) qPCR-based resection assays were carried out with siControl-, siRPA2-, siDNA2-, and siDNA2+siRPA2-treated cells. All error bars indicate SEM from three biological replicates.

Fig. S9.

Exo1-GFP colocalizes with DNA damage. (A) RPA2 knockdown does not affect the cell cycle. Representative images of EdU-stained DNA in siControl- and siRPA2-treated cells. Quantification of EdU foci for two biological replicates (>200 cells per replicate) is included in the main text (Fig. 3D). (Scale bar, 10 µm.) (B) Exo1-GFP colocalizes with the DNA damage marker γH2AX at sites of laser-induced DNA damage. (Scale bar, 10 µm.)

Fig. 4.

yExo1 is a processive nuclease. (A) Kymograph (Upper) and resection track (Lower) showing yExo1 processing of a DNA end. (B) Comparison of yExo1 (orange) and hExo1 (purple) velocities (n = 57 for yExo1 and n = 113 for hExo1) and (C) processivities (n = 57 for yExo1 and n = 111 for hExo1). The P values were 0.004 and 0.15 for the velocities and processivities, respectively. For yExo1, the mean velocity was 10.9 ± 5.4 bp/s and the average processivity was 5.6 ± 2.6 kb (n = 57). (D) Kymograph of a yExo1 molecule in the presence of 1 nM yRPA. The orange line indicates when yRPA enters the flowcell. (E) yExo1 lifetimes in the absence of yRPA (orange, half-life >1,800 s; n = 58) and (F) with 1 nM yRPA injected into the flowcell (green, half-life = 52 ± 2 s, n = 76). The red lines are single exponential fits to the data.

Fig. 5.

RPA promotes distributive Exo1 activity. (A) hExo1 resection in the presence of WT hRPA. Assays were performed with 10 ng of 4.5-kb linear DNA with 4-nt 3′ ssDNA overhangs, 62.5, 125, 250, or 500 pM hExo1, and 100 nM hRPA. Samples were deproteinized, separated on a 1% agarose gel in nondenaturing conditions, and then stained with SYBR green. (B) Resection reactions performed as above with 75, 150, 300, or 600 pM yExo1 and 100 nM yRPA. (C) Lifetime of hExo1 binding events when both hExo1 and hRPA are continuously injected into the flowcell. (D) Lifetime of yExo1 binding events in the presence of yRPA, when both proteins are continuously injected into the flowcell. For both C and D, the data were best described by two characteristic timescales (red, biexponential fit, see accompanying text). (E) Schematic of multiple turnover experiments. A 1:1 mixture of magenta- and green-labeled Exo1, as well as 1 nM RPA is continuously flowed over DNA curtains. (F) Kymographs (Upper) and the corresponding single-molecule trajectories (Lower) of several hExo1 and (G) yExo1 reloading at the same resection tract. Both magenta and green Exo1 molecules bind at the same site on the DNA molecule. After the resection experiments, hRPA-GFP was injected into the flowcell to stain the resulting ssDNA tracts (blue, Right).

Fig. 6.

A model for hExo1 regulation by human SSBs. SOSS1 cooperatively loads hExo1 on DNA ends. As hExo1 resects the DNA, SOSS1 is replaced by hRPA. hRPA displaces hExo1, but another hExo1 molecule can rebind at the site. This cycle of binding, displacement, and rebinding results in distributive resection. Additional factors (e.g., BLM, MRN, and PCNA; not included for clarity) may also promote resection in the presence of hRPA.