The RecQ helicase Sgs1 drives ATP-dependent disruption of Rad51 filaments

Abstract

DNA helicases of the RecQ family are conserved among the three domains of life and play essential roles in genome maintenance. Mutations in several human RecQ helicases lead to diseases that are marked by cancer predisposition. The Saccharomyces cerevisiae RecQ helicase Sgs1 is orthologous to human BLM, defects in which cause the cancer-prone Bloom's Syndrome. Here, we use single-molecule imaging to provide a quantitative mechanistic understanding of Sgs1 activities on single stranded DNA (ssDNA), which is a central intermediate in all aspects of DNA metabolism. We show that Sgs1 acts upon ssDNA bound by either replication protein A (RPA) or the recombinase Rad51. Surprisingly, we find that Sgs1 utilizes a novel motor mechanism for disrupting ssDNA intermediates bound by the recombinase protein Rad51. The ability of Sgs1 to disrupt Rad51-ssDNA filaments may explain some of the defects engendered by RECQ helicase deficiencies in human cells.

Figure 1.

Sgs1 binds to RPA-coated ssDNA. (A) ATP hydrolysis assays with 0, 0.25, 0.8, 2.4 μM RPA with unlabeled Sgs1. The data points represent the mean and standard deviation of three independent experiments. (B) Schematic of ssDNA curtain assay used to measure the binding and translocation activity of GFP–Sgs1 on ssDNA-RPA molecules. (C) Widefield images showing a ssDNA bound by RPA–mCherry (magenta) and GFP–Sgs1 (green). (D) Binding distribution of GFP–Sgs1 on ssDNA bound by RPA–mCherry molecules, error bars were generated by bootstrapping the data using a custom python script (N = 340).

Figure 2.

Sgs1 is a robust ssDNA motor protein. (A) Representative kymograph of GFP–Sgs1 translocation on unlabeled RPA–ssDNA. (B) Representative kymographs illustrating the translocation of GFP–Sgs1 (green) on ssDNA bound by RPA–mCherry molecules in the presence and absence free 0.1 nM RPA–mCherry, as indicated. (C) Velocity distribution of individual GFP–Sgs1 complexes translocating on RPA–ssDNA (N = 115); the data represents combined results taken from experiments with RPA–mCherry and unlabeled RPA. The data fit a Gaussian distribution and the mean was determined from the fit. (D) Survival plot used to determine the processivity of GFP–Sgs1 (N = 115); the data represents combined results taken from experiments with RPA–mCherry and unlabeled RPA. Error bars were generated by resampling the data by bootstrapping using a custom python script. All reported processivity values were determined from point in the graph at which the survival probability was equal to 0.5.

Figure 3.

Disruption of Rad51 filaments by Sgs1. (A) Representative kymograph showing GFP–Sgs1 (green) translocation on an ssDNA molecule bound by unlabeled Rad51. Rad51 displacement is revealed by rebinding of RPA–mCherry (magenta). (B) Velocities distribution for individual Sgs1 translocation events; the data represents combined results taken from experiments with GFP–Sgs1 and unlabeled Sgs1 (N = 121). The data fit a Gaussian distribution and the mean was determined from the fit. (C) Survival probability plot used to determine the processivity of Sgs1 on Rad51–ssDNA (N = 121); error bars were generated by bootstrapping. (D) Images of individual Rad51–ssDNA filaments showing embedded RPA–mCherry clusters (magenta) and bound by GFP–Sgs1 (green). (E) Graph quantifying GFP–Sgs1 binding locations on Rad51–ssDNA (N = 342).

Figure 4.

Removal of Rad51 mutants by Sgs1. (A) Kymographs illustrating GFP–Sgs1 (green) translocation on ssDNA bound by either wild-type Rad51 (left), Rad51^I345T (middle) or Rad51^K191R-ssDNA (right) in the presence of RPA–mCherry (magenta). (B) Velocities distributions for GFP–Sgs1 on Rad51^I345T-ssDNA (N = 70). (C) Survival probability plot for GFP–Sgs1 on Rad51^I345T-ssDNA (N = 70). Error bars were generated by resampling the data by bootstrapping using a custom python script. (D) Velocities distribution for GFP–Sgs1 on Rad51^K191R-ssDNA (N = 44). (E) Survival probability for GFP–Sgs1 on Rad51^I345T–ssDNA (N = 70). Error bars were generated by resampling the data by bootstrapping.

Figure 5.

Dmc1 prevents Sgs1 from binding to ssDNA. (A) Two-color widefield images of a Rad51–ssDNA (unlabeled) curtain assembled in the presence of RPA–mCherry (magenta). The contrast of these images has been adjusted to highlight the presence of the RPA–mCherry clusters above background. (B) Two-color TIRFM widefield images of a Dmc1–ssDNA (unlabeled) curtain assembled in the presence of RPA–mCherry (magenta). The contrast of these images matches the contrast shown in panel A and has been adjusted to highlight the presence of the RPA–mCherry clusters. (C) Quantification of the number of GFP–Sgs1 binding events per ssDNA molecule for Rad51–ssDNA (N = 70) and Dmc1–ssDNA (N = 86). Error bars represent the mean and standard deviation of the data set.

Figure 6.

Top3–Rmi1 slows Sgs1 translocation on Rad51–ssDNA. (A) Kymograph showing GFP–Sgs1(green)/Top3–Rmi1 translocation on Rad51–ssDNA in the presence of RPA–mCherry (magenta). (B) Velocity distribution for GFP–Sgs1/Top3–Rmi1 on Rad51–ssDNA (N = 95). (C) Survival probability plot for GFP–Sgs1/Top3–Rmi1 on Rad51–ssDNA (N = 105); error bars were generated by resampling the data by bootstrapping using a custom python script. (D) Comparison of GFP–Sgs1 translocation velocity with and without Top3–Rmi1 (P-value ≤ 0.0001). Error bars represent the 95% confidence interval for the mean of the Gaussian distribution. (E) Comparison of the processivity values for GFP–Sgs1 with and without Top3–Rmi1; the difference between the processivity values is not statistically significant (P-value = 0.0008). Error bars represent the 95% confidence interval for the half-life of exponential decay function by which the data was fit.

Figure 7.

Molecular mechanisms of Sgs1 action on ssDNA intermediates. (A) Sgs1 can act at all early stages of HR, and its antirecombinase activity would act similarly to Srs2 by channeling intermediates towards the SDSA pathway of repair. (B) Surveillance of RPA–ssDNA may allow Sgs1 to inappropriate accumulation of Rad51 at replication forks, which may otherwise give rise to replication-coupled hyperrecombination. (C) Comparison of Srs2 and Sgs1 activities on ssDNA. Srs2 and Sgs1 both translocation on RPA–ssDNA, but (i) Srs2 strips RPA from ssDNA, whereas (ii) Sgs1 does not. Srs2 and Sgs1 both load at RPA clusters present at the ends of Rad51 filaments, for Srs2 (iii) multiple loading events take place and Rad51 removal is coupled to the Rad51 ATP hydrolysis cycle. In the case of (iv) Sgs1, loading does not involve iterative binding events, and Rad51 removal is uncoupled from the Rad51 ATP hydrolysis cycle. Neither Srs2 nor Sgs1 can remove Dmc1 from ssDNA, but (v) Srs2 inhibition occurs primarily by blocking translocation, whereas (vi) Sgs1 is blocked from binding.