The toposiomerase IIIalpha-RMI1-RMI2 complex orients human Bloom’s syndrome helicase for efficient disruption of D-loops

Homologous recombination (HR) is a ubiquitous and efficient process that serves the repair of severe forms of DNA damage and the generation of genetic diversity during meiosis. HR can proceed via multiple pathways with different outcomes that may aid or impair genome stability and faithful inheritance, underscoring the importance of HR quality control. Human Bloom’s syndrome (BLM, RecQ family) helicase plays central roles in HR pathway selection and quality control via unexplored molecular mechanisms. Here we show that BLM’s multi-domain structural architecture supports a balance between stabilization and disruption of displacement loops (D-loops), early HR intermediates that are key targets for HR regulation. We find that this balance is markedly shifted toward efficient D-loop disruption by the presence of BLM’s interaction partners Topoisomerase IIIα-RMI1-RMI2, which have been shown to be involved in multiple steps of HR-based DNA repair. Our results point to a mechanism whereby BLM can differentially process D-loops and support HR control depending on cellular regulatory mechanisms.

linear fit to the data (slope: 455 ± 21 nM -1 ; intercept: 1040 ± 254)). These control experiments confirm that the detection method and the D-loop construct concentration applied in the kinetic experiments (30 nM) are suitable for precise quantification required for kinetic analysis. (b) Kinetic profiles of two control experiments monitoring DL4 unwinding by BLM CR in the absence of ssDNA trap strand. After stopping the reaction, 3 µM I-trap strand was added to the mixture to prevent reannealing of unwound DNA molecules. As expected, the omission of the ssDNA trap strand during the reaction markedly accelerated the decomposition of all multi-stranded DNA species (cf. Fig. 2b). (c) Kinetic profiles of two control experiments monitoring DL4 unwinding by BLM in the presence of 15 µM trap strand (5 times higher than in Fig. 2c). Elevation of trap strand concentration slightly reduced the amplitude of the first phase of DL4 decomposition, similar to that previously observed for RecQ 1 , suggesting an increase in the fraction of the non-productive unwinding runs (DLN, Fig. 3). DNA species are color coded as in Fig. 3. Experiment sets 1 and 2 are shown as solid and open circles, respectively. DNA species are described in Fig. 2b and in Methods. Source data are provided as a Source Data file.   Fig. 6a), originating from unwinding the leftward dsDNA arm of DL4 (Fig. 5a, Supplementary Fig. 7).
Additional bands are identified as the four stranded L4F (shows fluorescein signal only) and L4FC (fluorescein and Cy3 signal) and as the three-stranded LF3 (shows fluorescein signal only) and LFC3 (fluorescein and Cy3 signal) DNA structures ( Supplementary Fig. 6a). Considering possible D-loop binding orientations, L4FC is a product of DL3 (initially present in the DL4 preparation and bound by BLM) processing occurring via the DLL pathway ( Supplementary Fig. 7, DL3 -DLL pathway). In addition, L4FC can also arise from the DL5L structure ( Supplementary Fig. 7, PE pathway) due to enzyme rebinding. Similarly, L4F, LF3 and LFC3 will form due to enzyme rebinding and subsequent processing of intermediate DNA structures, ultimately yielding the final reaction products L4F, LF3 (fluorescein signal only) and ssDNA (Cy3 signal only). For all constructs, three independent experiments were performed and showed reproducible kinetic profiles. Densitometry of DNA species DL4, DL3, 3T and DL5L obtained from the Cy3 signal and DL5L obtained from the fluorescein signal are shown in Figs. 6a and 7a. Densitometry of DNA species LF3 + LFC3, L4F + L4FC, detectable via the fluorescein signal, is shown in Supplementary Fig. 9. Source data are provided as a Source Data file.  Fig.  6b. Importantly, a band migrating slower than DL4 was not observed for any of the investigated helicase constructs. Based on Fig. 5a and Supplementary Fig. 6b, unwinding of the rightward dsDNA arm of the DL4 structure should lead to the formation of the DL5R structure, migrating slower than DL4 ( Supplementary Fig. 6b). Lack of this structure indicates that the investigated helicases do not efficiently catalyze the initial unwinding of the rightward dsDNA arm and/or the amount of the formed products is below of our detection limit.
Additional bands are identified as the four-stranded RFC4 (fluorescein and Cy3 signal) and the three-stranded RFC3 (fluorescein and Cy3 signal) and RF3 (fluorescein signal only) structures. These structures will form due to enzyme rebinding and subsequent processing of intermediate DNA structures. For all constructs, three independent experiments were performed and showed reproducible kinetic profiles. Densitometry of DNA species DL4, DL3, 3T obtained from the Cy3 signal are shown in Figs. 6b, 7b. Densitometry of DNA species RFC4, R3F and RFC3, detectable via the fluorescein signal, is shown in Supplementary Fig. 9. Source data are provided as a Source Data file. . k U represents a lower bound for non-productive events and also for the unwinding rates as reaction kinetics are not resolved before the first 10 s. In the DLR pathway, unwinding of DL4 and DL3 (with rate constant k U ) will lead to the formation of the DL5R (purple) and DL4R (olive) structures, respectively. These structures, if present, would be observed in R-trap experiments ( Supplementary  Figs. 5, 6b). Importantly, our results ( Supplementary Fig. 5) indicate that none of the investigated helicase constructs uses the DLR pathway. Unwinding of the invading strand in the DL5L or L4FC structures (PI pathways) will lead to formation of 3T and ssDNA, respectively. A helicase binding orientation that leads to the unwinding of the DL5L structure outward of the invading strand, using it as the tracking strand, will lead to the formation of the L4FC structure. The PN pathway contains all unwinding events that are non-productive (due to premature termination of unwinding starting from any binding orientation or due to reannealing of the unwound dsDNA arm upstream of the strand invasion in L-trap experiments). Slow rebinding of enzyme to DNA products (inhibited by excess ssDNA trap strands, occurring with rate constant k R ) and subsequent unwinding leads to further processing of the given structures (these processes are exemplified by the disappearance of 3T or DL5L after their accumulation). Pathways involving DL5R and DL4R are not shown in the model as the DLR pathway was not observed. The model was used in global fits to results shown in Figs. 6-8, Supplementary Fig.  13.
The black star indicates the Cy3 label in the invading strand, whereas the orange and purple stars/strands represent L-trap and R-trap (with fluorescein label), respectively.  Fig 2c). TRR decreased (but did not completely abolish) the fraction of DL4 in a concentration-dependent manner, leading to the formation of 3T. The fractions of DL3 and ssDNA did not change during the experiment, indicating that binding of TRR to DL4 partially disrupts the strand invasion, but only when the 5' end of the strand invasion is in the double-stranded form. (c) Changes in fractions of DNA species present in DL4 preparations were followed in time after addition of ATP (1 mM final) to a preincubated (5 min 37°C) mixture of TRR and the DL4 preparation. The final concentration of TRR was 1.2 µM after addition of ATP. Reactions were set up and analysis was performed as for the experiments monitoring the activity of helicases (cf. Fig. 2c). DL4 fractions were lower, while 3T fractions were higher than that observed without TRR (Fig. 2c, Supplementary  Fig. 8a), as expected based on panel b. The fractions of different DNA species remained constant during the investigated time regime, highlighting that the passive DL4 disruption activity of TRR reached an equilibrium during the preincubation period. Importantly, these results indicate that the D-loop processing measurements (Fig. 8, Supplementary Fig. 11) Supplementary Fig. 8a. In line with panel b, TRR alone significantly decreased the fraction of DL4 (p = 0.03) and significantly increased the fraction of 3T (p = 0.003) compared to the no protein control, whereas the moderate increase in the DL3 fraction is statistically not significant. Importantly, TRR + BLM behaved similar to TRR alone; the small differences between fractionation of DNA species with TRR or TRR + BLM were found to be statistically not significant.
These control experiments, together with panel b, indicate that in experiments where TRR and BLM are incubated with the D-loop structure in the absence of ATP for 3 min, TRR-facilitated D-loop disruption reaches an equilibrium and this is not affected by the presence of BLM. Source data are provided as a Source Data file. Supplementary Fig. 11. Representative electrophoretograms of L-trap D-loop unwinding experiments with and without TRR, monitoring Cy3 and fluorescein signals (a) D-loop processing measurements of BLM FL were done as in Fig. 5 and Supplementary Fig. 5 with or without 1.2 µM TRR. DNA species were annotated according to Supplementary Fig. 6. The presence of TRR inhibits the formation of DL4L molecule, indicating the inhibition of the DLL pathway. (b) Representative Cy3 and fluorescein (Flu) electrophoretograms obtained with the indicated helicase constructs with or without 1.2 µM TRR.
Average fractions of DNA species determined from multiple experiments are shown in Fig. 8 and Supplementary Fig. 13. Experiments were performed 3 times independently and showed identical patterns for DNA species. Source data are provided as a Source Data file. Fig. 5 and Supplementary Fig. S11 but, instead of a 3 µM total trap stand concentration, 4.5 µM was used (2.25 µM of each of I-trap and L-trap). These experiments confirmed that the absence of the DLL pathway is not caused by depletion of trap strand by TRR. (b) R-trap experiments performed as in Supplementary Fig. S6 in the presence of 1.2 µM TRR. The absence of the DL4R DNA species (Fig. 5) with or without TRR ( Supplementary Fig. 6) indicates that TRR does not increase the preference of BLM for the DLR pathway (for pathways see Supplementary Fig.  7). Results were in line with previous independent results and each experiment was therefore performed in a single run. Supplementary Fig. 13. Determined fractions of DNA species from DL4 ( Supplementary Fig. 11), DL3 and 3T unwinding experiments for BLM CR and RecQ with or without 1.2 µM TRR.

Supplementary Fig. 12. DL4 L-trap experiment of BLM FL + TRR in the presence of increased trap stand concentration and DL4 R-trap experiment with BLM FL + TRR (a) Experiments performed as in
Solid lines show global fits using the extended model shown in Supplementary Fig. 7. DNA species are color coded as in Fig. 6. For BLM FL data see Fig. 8a. Determined parameters are shown in Supplementary Table 3. Means ± SEM (n = 3 for DL4 experiments, n = 2 for 3T and DL3 experiments) are shown on all panels for the detected DNA species at each time point, determined from independent experiments with individual protein constructs. Source data are provided as a Source Data file. Supplementary Fig. 14. Control DL4 processing L-trap measurements performed with BLM FL with or without 1.2 µM TRR at increased time resolution (a) Representative Cy3 electrophoretograms from L-trap experiments performed as in Supplementary  Fig. 11, but here the reaction kinetics were followed in 2-s intervals for 18 s. Experiments were independently performed twice and showed identical results. Reaction profiles were in line with results of Supplementary Fig. 11 and reveal that the absence of DL4L in experiments performed with 1.2 µM TRR is not caused by the increased unwinding rate of BLM and support the hypothesis that TRR increases the preference of BLM FL for disruption of strand invasions (DLI and DLI' pathways on Supplementary Fig. 7). (b-c) Kinetic profiles of the fractional changes of (b) DL5L (Flu signal) and (c) DL4 (Cy3 signal) determined from panel a. Means ± SEM for two independent experiments are shown. (b) In the absence of TRR, after initial accumulation the fraction of DL5L decreased in time as seen in Supplementary Fig. S11. In the presence of TRR only a faint signal at the position of DL5L was detectable (panel a) and its fraction increased only marginally during the experiment. Fitting of single exponential decay function to the declining phase of data obtained without TRR revealed observed rate constant of 0.07 ± 0.04 s -1 , close to the k r rebinding rate ( Supplementary Fig. 7) obtained via global model fitting (Supplementary Table 3). (c) The fraction of DL4 decreased in line with kinetics observed in previous experiments ( Supplementary Fig. S11). Fitting of a single exponential decay function revealed observed rate constants of 0.33 ± 0.02 s -1 and 0.55 ± 0.09 s -1 for BLM FL and BLM FL + TRR, respectively, in line with the k u unwinding rates determined via global fitting of the model to data in Supplementary Fig. S11 (see also Supplementary Table 3). These results confirm that TRR enhances the D-loop unwinding activity of BLM FL . Importantly, the ratio between the fraction change of DL5L and DL4 after 2 s is 0.38, indicating in a model-independent manner that BLM FL maintains a balance between pathways leading to D-loop disruption and stabilization (see also Fig. 8b, Supplementary  Table 3). In contrast, in the presence of TRR, DL4 is processed via other pathways, i.e., DLI and DLI', as predicted by modeling (Fig. 8b, Supplementary  a Values reported were determined from global fitting of the model shown in Fig. 3 to DL4, DL3 and 3T unwinding data for each helicase construct (Fig. 4a). Best-fit values  fitting standard errors (determined by the software used for fitting) are shown. b Values of k U also represent lower bounds for the rate constant of unwinding and non-productive unwinding runs. c Determined from direct fluorescence anisotropy titrations using ss54-FLU ( Supplementary Fig. 3a). d Determined from competitive titration experiments ( Supplementary Fig. 3b-e).  Supplementary Fig. 7 to DL4 data from L-trap and R-trap experiments and DL3 and 3T unwinding data from I-trap experiments (Fig. 3a, RecQ unwinding data obtained from ref. 1 ) for each helicase construct (Figs. 6-7). b Values of k U also represent lower bounds for the rate constant of unwinding and non-productive unwinding runs. Best-fit values and estimation of fit robustness as error are shown. The modeling software was unable to determine fitting standard errors for most parameters. Thus, fitting uncertainty was characterized as the SD of best-fit values from 5 fitting runs starting from differentially initialized fraction ratios.