Distinct RPA domains promote recruitment and the helicase-nuclease activities of Dna2

The Dna2 helicase-nuclease functions in concert with the replication protein A (RPA) in DNA double-strand break repair. Using ensemble and single-molecule biochemistry, coupled with structure modeling, we demonstrate that the stimulation of S. cerevisiae Dna2 by RPA is not a simple consequence of Dna2 recruitment to single-stranded DNA. The large RPA subunit Rfa1 alone can promote the Dna2 nuclease activity, and we identified mutations in a helix embedded in the N-terminal domain of Rfa1 that specifically disrupt this capacity. The same RPA mutant is instead fully functional to recruit Dna2 and promote its helicase activity. Furthermore, we found residues located on the outside of the central DNA-binding OB-fold domain Rfa1-A, which are required to promote the Dna2 motor activity. Our experiments thus unexpectedly demonstrate that different domains of Rfa1 regulate Dna2 recruitment, and its nuclease and helicase activities. Consequently, the identified separation-of-function RPA variants are compromised to stimulate Dna2 in the processing of DNA breaks. The results explain phenotypes of replication-proficient but radiation-sensitive RPA mutants and illustrate the unprecedented functional interplay of RPA and Dna2.


INTRODUCTION
To prepare the expression construct for his-tagged full-length trimeric yeast S. cerevisiae RPA, a sequence coding for 6x-his tag was introduced to the 5'-end of RFA1 in p11d-sctRPA vector (a kind gift from M. Wold, University of Iowa) 34 . To this point, primers AA07 and AA08 were annealed and cloned into NcoI site in p11d-sctRPA, creating p11d-his-sctRPA. See Supplementary Table S1 and S2 for the list and sequences of all primers used in this study. All point mutants of full-length trimeric yeast RPA were introduced into this construct using mutagenic primers and QuikChange II-XL site-directed mutagenesis kit (Agilent Technologies) according to manufacturer's recommendations. To prepare Rfa1, Rfa2, Rfa3, Rfa1-F, Rfa1-FAB, Rfa1-ABC and Rfa1-AB expression constructs, the respective DNA sequences were amplified by PCR using the corresponding primers (Supplementary Table S1 and S2) and ligated into pET21 vector. These constructs contained a T7-tag at the N-and 6x-his tag at the Cterminus. Point mutants of Rfa1 and Rfa1-AB were also introduced into pET21-T7-Rfa1-his and pET21-T7-Rfa1-AB-his, respectively, using mutagenic primers as described above for the trimeric mutants.
All yeast recombinant trimeric full-length his-tagged wild type RPA and mutants (I14S, K45E, This was followed by elution with elution buffer (wash buffer II containing 300 mM imidazole). The eluate was then diluted 8-fold with dilution buffer (50 mM HEPES-KOH pH 7.5, 10% glycerol, 2 mM β-ME, 94 mM KCl) and loaded on a pre-equilibrated HiTrap Q column. The loaded column was washed first with dilution buffer containing 100 mM KCl, and then 170 mM KCl, before finally eluting the protein in a salt gradient from 170 mM to 800 mM KCl under identical buffer conditions. The eluate fractions were analyzed on a 12% polyacrylamide gel to identify the best ones, which were pooled, aliquoted, snap frozen in liquid nitrogen and stored at -80 °C. Recombinant Rfa2 and Rfa3 were also purified as above, except that the wash buffer I contained 1 M KCl. Recombinant Rfa1, Rfa1-ABC and Rfa1-C were purified using the same procedure as Rfa2 and Rfa3, but diluted after the first Ni-NTA step and loaded onto the HiTrap Q column at pH 8.5. These fragments were further washed with wash buffers with pH 8.5 and eluted as described above in buffers with pH 7.5. Recombinant Rfa1-F, Rfa1-AB and Rfa1-FAB constructs were purified using Ni-NTA and HiTrap-Heparin column, since they did not bind to the HiTrap Q column. The Ni-NTA purification step was done as described for the fragments (Rfa1, Rfa1-ABC and Rfa1-C) above, but washed with wash buffer II containing 40 mM imidazole.
The heparin step was performed as described for HiTrap Q purification of Rfa1, Rfa1-ABC and Rfa1-C. Point mutants of Rfa1 (I14S and K494E) and Rfa1-AB (S191K, Y193A and Y193D) were purified using Ni-NTA only using the above described procedures for their corresponding wild type proteins.
Recombinant S. cerevisiae wild type and nuclease-dead Dna2 variants were expressed from a modified pGAL:DNA2 vector that contained N-terminal FLAG and HA-tags and a C-terminal 6x-his tag, in the S. cerevisiae strain WDH668 36,37 . Yeast cells (4 l) were grown to approximately OD600 ∼0.6 in a standard synthetic medium lacking uracil, and supplemented with glycerol (3%, vol/vol) and lactic acid (2%, vol/vol) as carbon sources. Expression of Dna2 was induced with galactose (2%) for 6 h and the cells were lysed in a freezer mill (Spex Sample Prep). All Dna2 variants were purified by Ni-NTA and FLAG (Sigma, A2220) affinity chromatography as described previously 29 . Human mitochondrial SSB was expressed from pMALT-P-mtSSB in 1 l of BL21-DE3 pLysS E. coli cells upon induction with 0.4 mM IPTG for 3 h at 37 °C, and purified using amylose affinity resin (New England Biolabs) and HiTrap Blue HP as described previously 28 . Recombinant S. cerevisiae Sgs1 protein was expressed from the pFB-MBP-Sgs1-his vector in Spodoptera frugiperda 9 (Sf9) insect cells by using Bac-to-Bac baculovirus expression system (Invitrogen) and purified using Amylose and Ni-NTA affinity chromatography 38 .

DNA substrate preparation
For DNA binding experiments, ssDNA oligonucleotide (93 nt-long, X12-3HJ3) 28 was labeled at the 3' terminus with [α-32 P] dCTP (Perkin Elmer) and terminal transferase (New England Biolabs), according to standard protocols. Unincorporated nucleotides were removed using Micro Bio-Spin P-30 Gel Columns (Bio-Rad). For oligonucleotide-based nuclease assays and DNA recruitment assays, either of two 5'-overhanged DNA substrates (45 nt ssDNA and 48 bp dsDNA, oligonucleotides X12-3 HJ1S and   X12-3 TOPL, or 30 nt ssDNA and 31 bp dsDNA, oligonucleotides PC92 and X12-4SC, as indicated) were used 28,29 . Recruitment measurements using mass photometry used a DNA substrate with a 25 nt 5'-overhang and a 48 bp dsDNA region (assembled from oligonucleotides X12-3 HJ1S 28  DNA substrates were used 28,29 . The dsDNA construct for the magnetic tweezers assays was prepared as described 28,39 . A 6.6 kbp dsDNA fragment was excised from pNLrep plasmid 40 , using the restriction enzymes BamHI and BsrGI. Simultaneously, a 63 nt gap was formed by utilizing the Nt.BbvCI nicking enzyme. The gap enabled annealing of a 65 nt long oligonucleotide with a 25 nt complementary sequence and a 40 nt polythymidine 5' tail. Subsequently, 0.6 kbp long dsDNA handles containing either multiple biotin or digoxigenin modifications were ligated to each end of the 6.6 kbp fragment to support specific anchoring to the streptavidin-coated magnetic bead and the antidigoxigenin-covered glass coverslip.
Proteins were added and incubated at 30 °C for 30 min. Loading dye (5 µl; 50% glycerol, bromophenol blue) was added to the reactions and the products were separated on 6% polyacrylamide gels (ratio acrylamide:bisacrylamide 19:1, Bio-Rad) in TAE buffer (40 mM Tris, 20 mM acetic acid and 1 mM EDTA) at 4 °C. The gels were dried on 17 CHR (Whatman), exposed to a storage phosphor screen (GE Healthcare) and scanned by a Typhoon Phosphor Imager (FLA 9500, GE Healthcare). Binding of Dna2-E675A to the Y structure (oligonucleotides X12-3HJ3 and X12-3TOPL, 0.1 nM, in molecules) was performed under identical conditions with 2 mM magnesium acetate and 1 mM ATP but without added salt.

Nuclease assays
Unless and otherwise indicated, all oligonucleotide-based nuclease assays (15 µl

Helicase assays
Unless otherwise indicated, all oligonucleotide-based helicase assays (15 µl volume) with Dna2-E675A were performed in the reaction buffer used for the nuclease assays supplemented with 50 mM KCl, using Y-shaped DNA as substrate (oligonucleotides X12-3HJ3 and X12-3TOPL or PC92 and X12-4SC, as indicated, 0.1 nM, in molecules). Recombinant proteins were added as indicated. The reactions were incubated at 30 °C for 30 min and stopped using 5 µl 0.2% stop buffer containing 0.2% SDS, 150 mM EDTA and 30% glycerol, and 1 µl proteinase K (14-22 mg/ml, Roche), and incubated at 37 °C for 10 min. To avoid re-annealing of the substrate, the stop solution was supplemented with a 20-fold excess of the unlabeled oligonucleotide with the same sequence as the 32 P-labeled one. The products were separated by 10% polyacrylamide gel electrophoresis in TBE buffer, dried on 17 CHR chromatography paper (Whatman) and analyzed as described above. Plasmid based helicase assays (15 µl volume) were performed in the same way, except 2.2 kbp-long DNA substrate (0.1 nM, in molecules) was used. The products were separated by 1% agarose (Sigma, A9539) gel electrophoresis in TAE buffer, squeezed, dried on DE81 paper (Whatman) and analyzed.

DNA recruitment assays
The assays were performed as the DNA binding assays with short or long 5'-overhanged DNA substrate (1 nM, in molecules), except that the reactions contained 150 mM NaCl and the DNA was always precoated with the respective RPA variant on ice for 5 min. The reactions performed with wild type Dna2 contained 3 mM EDTA instead of 5 mM magnesium acetate to inactivate the nuclease. The products were separated on 4% polyacrylamide gels and analyzed as described above.

Docking simulations to model the interaction between RPA1_AB domains and DNA2
To generate a model of the assembly between the N-terminal domain of Rfa1 and the small region of Dna2 upstream of the OB fold, a structure between of the complex between the orthologous subunits in human (PDB:5EAY) was used as template to produce a homology model using rosettaCM standard protocol 41 (Fig. 4a). To model the interaction between the domains A and B of Rfa1 with Dna2, no structure of homologous complex was available. Consequently, a rigid-body free docking simulation was performed using the InterEvDock2 server 42 , which takes into account the physicochemical nature of protein surfaces and co-evolutionary information. InterEvDock2 computes a consensus between 3 complementary scores, Frodock 43 , SOAP-PP 44 and InterEvScore 45 to identify the most likely interfaces (http://bioserv.rpbs.univ-paris-diderot.fr/services/InterEvDock2/). The free docking strategy used human rather than yeast proteins because it would have required to use homology models as inputs of the free docking protocol likely decreasing docking performance. Docking simulations were performed using the X-ray structures of human RPA1 (PDB:1JMC) and DNA2 (PDB:5EAN). The input structures were the crystal structure of the tandem OB domains A and B of human RPA70 and the crystal structure of human DNA2 (Fig. 6a). The docking protocol was performed as described previously following the server standard protocol 46,47 . The co-alignments between both DNA2 and RPA1 partners automatically generated by the server contained 101 sequences belonging to species ranging from metazoans to fungi.
From the results archive, the 50 best decoys of every 3 scores (Frodock, SOAP-PP, InterEvScore) used in the consensus selection of the docking models were considered. There were two major regions at the surface of DNA2 involved in the most likely docked models ( Supplementary Fig. 6a). The first surface patch involved a surface close to the 3' end of the ssDNA molecule co-crystallized with DNA2 and to the N-terminal OB-fold of DNA2. Since it had previously been shown that the OB domain in DNA2 contributes to the binding of RPA 22 , a second docking simulation was run to further sample around this first surface patch. A second docking run was constrained so as to consider only models for which residues D111, D114 and F115 of Dna2 (in the OB domain) are at less than 15 Å from any atom of RPA1. The Top5 models generated in the second run were selected and relaxed using Rosetta 48 to remove steric clashes through using a standard relax protocol under native coordinate constraints and using the beta_nov15 scoring function. The model scored in rank 3 by InterEvScore was selected based on the calculated interface energy and visual inspection of the interface. This model was converted into a model of the complex between the S. cerevisiae Dna2 and Rfa1-AB partners using comparative modelling by RosettaCM. The resulting model was used to design a disruptive mutant on Y193 and S191 likely to prevent the Rfa1-AB module to interact with DNA2 while keeping its affinity for ssDNA.

Mass photometry characterization of protein complexes
Mass Photometry measurements were performed on a OneMP mass photometer (Refeyn Ltd). To prepare the measurements, borsilicate microscope coverslips (No. Protein Standard, Invitrogen) was measured the same day. All samples were measured in triplicates.

Magnetic tweezers
Magnetic tweezers measurements were carried out in a custom-built magnetic tweezers setup at room temperature 49,50 . The DNA construct were boundby their biotinylated end to 2.8 µm streptavidin-coupled magnetic beads (Dynabeads M280, Thermofischer Scientific) and flushed into the flow cell. Its bottom glass slide was covered with antidigoxigenin to enable the specific attachment of the digoxigenin-modified end of the construct. After a short incubation time, the excess of the magnetic beads was flushed away. Lowering the magnets allowed to stretch DNA molecules tethered to the magnetic beads.
The measurements were performed at 30 Hz using video microscopy and real-time GPU accelerated image analysis 50 . Magnetic forces were calibrated using fluctuation analysis 51 . DNA unwinding experiments were performed in reaction buffer (25 mM Tris-acetate pH 7.5, 2 mM magnesium acetate, 50 mM KCl, 1 mM ATP, 1 mM DTT, 0.1 mg/ml BSA) containing 5 nM nuclease-dead Dna2 E675A and 20 nM wild type RPA or fragments. Analysis was performed using a custom-written MATLAB program 52 .

RPA specifically promotes both helicase and nuclease activities of Dna2
Both nuclease and helicase activities of S. cerevisiae Dna2 are known to be stimulated by RPA 14,29 . In accord, we show that RPA promoted the nuclease activity of wild type Dna2 (Fig. 1a-c), and the helicase activity of the nuclease-dead Dna2-E675A (Fig. 1e, f). Human mitochondrial ssDNA binding protein (mtSSB), which bears no sequence similarity to RPA, and rather resembles prokaryotic SSB, in contrast inhibited both activities of Dna2 (Fig. 1b, d-f). DNA unwinding by the Dna2 partner Sgs1 (Supplementary Fig. 1a) under the same conditions was instead promoted similarly by either mtSSB or RPA (Supplementary Fig. 1b-d). RPA activates DNA unwinding by Sgs1 and stimulates its processivity in singlemolecules assays 39,53 . A proportion of this apparent stimulatory effect likely involves simple prevention of DNA reannealing, which does not require specific interactions between Sgs1 and mtSSB. In contrast, specific functional and physical interactions clearly underpin the interplay of Dna2 and RPA.

RPA promotes Dna2 beyond recruitment
Previously, RPA was shown to recruit Dna2 to DNA 21 , and the recruitment function was thought to explain the apparent stimulation of Dna2 enzymatic activities 21,22 . It was proposed that Dna2 immediately replaces RPA on the DNA substrate 21 , although this conclusion was also disputed 16,22 . We observed that low concentrations of Dna2 (5-10 nM) were inefficient to bind overhanged DNA on their own (Fig. 2a, lanes 3 and 4, Fig. 2b). The same Dna2 concentrations led to a clear electrophoretic mobility shift when RPA was pre-bound to the substrate (Fig. 2a, lanes 9 and 10), showing that RPA facilitates the binding of Dna2 to DNA. In contrast, no DNA binding by Dna2 was observed when DNA was pre-bound by mtSSB (Fig. 2a, b), suggesting that mtSSB prevents the access of Dna2 to DNA. To distinguish whether Dna2 replaced RPA or whether both proteins bound DNA simultaneously, we characterized the formed complexes by mass photometry. This technique allows to estimate the molecular weight of biomolecules and their complexes at nanomolar concentrations by evaluating the backscattered light from individual complexes in a microscope setup (Fig. 2c). RPA alone exhibited a single dominant peak in the mass spectrum at a molecular weight 115 ± 27 kDa (theoretical value 114 kDa).
Low concentrations of Dna2 alone showed molecular weight of 166 ± 28 kDa (theoretical value 172 kDa), which did not change when DNA was added, indicating inefficient DNA binding (Fig. 2d, Supplementary Fig. 2a). When RPA was added to Dna2 without DNA, we observed only mass peaks corresponding to single RPA and Dna2 molecules (Fig. 2c). When DNA was added to the Dna2 and RPA sample, we observed a new species at 302 ± 67 kDa, corresponding to the combined molecular weights of Dna2, RPA, and DNA (theoretical total weight of 311 kDa). The species corresponding to Dna2 and DNA (202 kDa) was not observed. RPA-DNA species were positioned between the peaks of single RPA and Dna2s and were not easily distinguishable. Therefore, we conclude that RPA recruits Dna2 to DNA, and remains a component of the nucleoprotein complex at least as an intermediate. RPA is therefore positioned to promote Dna2 activities in principle also downstream of recruitment.
To establish whether RPA promotes Dna2 beyond recruitment, we prebound Dna2-E675A to DNA, using high Dna2 concentrations to achieve efficient binding (Fig. 2e), employing conditions to limit DNA unwinding by Dna2-E675A alone. We then activated unwinding reactions by adding ATP, or ATP in combinations with SSB or RPA (Fig. 2f). In this setup, when Dna2-E675A did not need to be recruited, RPA still promoted the helicase activity of Dna2-E675A in a specific manner, much more than SSB that acts non-specifically by preventing DNA reannealing. Therefore, RPA specifically promotes the Dna2 activities beyond facilitating recruitment to DNA, which is further supported by mutational analysis later in this study.

The Rfa1-FAB domains stimulate the Dna2 nuclease
To define how RPA stimulates Dna2, we set out to prepare RPA point mutants and truncations. As some of these mutants exhibited reduced ssDNA binding that could have impaired the established protein purification procedure, we modified the RPA preparation protocol to include an affinity step upon adding a his-tag on the N-terminus of Rfa1 ( Supplementary Fig. 3a We first tested point mutants in the N-terminal domain of Rfa1 (Rfa1-F), including I14S and K45E ( Fig. 3a), which were initially identified in a screen for RPA mutations resulting in radiation sensitivity but normal growth otherwise 54 . This phenotype suggested proficiency in DNA replication, but potential defect in DNA repair processes such as recombination. Additionally, the rfa1-I14S mutation was later found to be synthetically lethal in a dna2 helicase-deficient background, and was found to impair physical interaction with Dna2 33 . We also selected mutations mapping to the Rfa1-C domain, which impair ssDNA binding, creating RPA-K494E and RPA-CCAA (C505A, C508A) ( Fig. 3a, b, and Supplementary Fig. 3g) 54,55 . Accordingly, the RPA-CCAA mutations are lethal in yeast 55 . As anticipated, the I14S and K45E mutations did not impair ssDNA binding of the RPA heterotrimer, while the K494E and CCAA mutations reduced ssDNA binding ~3-fold, based on apparent KD (Fig. 3a, b, Supplementary   Fig. 3h). We then analyzed the ability of the RPA variants to stimulate the Dna2 nuclease. We observed that RPA-I14S, and to a lesser degree RPA-K45E, reduced but did not eliminate the stimulation of the Dna2 nuclease, while the mutations in the Rfa1-C domain did not have a notable effect (Fig. 3c). The RPA-K45E mutant was previously found to be impaired also in its interaction with the MRX complex in yeast 56 . These experiments demonstrated that the capacity of RPA to bind ssDNA does not directly correspond to the ability to stimulate the Dna2 nuclease. As the N-terminal domain of Rfa1 does not significantly bind DNA, its contribution to the stimulation of the Dna2 nuclease likely reflects specific physical and functional interactions with the Dna2 nuclease in agreement with previous data 22,33 .
Next, we expressed and purified the RPA subunits Rfa1, Rfa2 and Rfa3 individually, as well as various truncations of the main subunit Rfa1 (Fig. 3d, Supplementary Fig. 3i, j). As anticipated, the variants differed in affinity to ssDNA (Fig. 3d, e, Supplementary Fig. 3k, l). We observed that Rfa1 alone was fully sufficient to catalyze the stimulation of the Dna2 nuclease, while the Rfa2 and Rfa3 subunits possessed no detectable stimulatory activity (Fig. 3f, g). Within Rfa1, we found that the Nterminal Rfa1-F, along with Rfa1-A and Rfa1-B domains, were essential for the stimulation of the Dna2 nuclease, while the Rfa1-C domain, which provides the strongest contribution towards ssDNA binding, was entirely dispensable (Fig. 3h-j) 33 . We also noted that the Rfa1-F domain needs to be part of the same polypeptide together with Rfa1-AB, as no stimulation was observed when individual truncations were combined (Fig. 3k, Supplementary Fig. 3 m, n).

RPA-SSL mutant of the Rfa1-F domain fails to stimulate the Dna2 nuclease
To better define the contribution of the residues around I14 of Rfa1 to the stimulation of the Dna2 nuclease, we performed molecular modeling to predict the impact of the mutation on the structure of Rfa1-F. The residue I14 is located in a conserved helix embedded in the N-terminal domain, and is thus unlikely to mediate protein-protein interaction directly (Fig. 4a-c). Rather, mutation of this residue could result in alteration of the surrounding structure including residues that are exposed to the surface, impacting thus the interplay with Dna2. We next replaced two additional conserved residues in the vicinity of I14 in the same helix, namely F11 and F15, to prepare the RPA-SSL (F11S, I14S, F15L) mutant (Fig. 4d, Supplementary Fig. 4a). The triple mutations within the RPA heterotrimer were anticipated to exacerbate the effect of I14S. While the RPA-I14S mutant bound ssDNA indistinguishably from wild type RPA, the SSL mutant exhibited ~1.6-fold lower ssDNA binding affinity (Fig. 4e, Supplementary Fig. 4b). This reduction of ssDNA binding was however more subtle than that of the RPA-K494E and RPA-CCAA variants (Fig. 3b, ~3-fold reduction), which did not affect the stimulation of the nuclease. Therefore, the decrease in ssDNA binding of RPA-SSL is unlikely to be responsible for any potential effects on the stimulation of the Dna2 nuclease.
Next, we compared the capacity of wild type RPA and the RPA-I14S and RPA-SSL mutants to stimulate Dna2 to cleave 5'-overhanged DNA. The SSL variant was dramatically impaired to promote the Dna2 nuclease, much more so than RPA-I14S (Fig. 4f, g). As described above, Rfa1 alone can stimulate the Dna2 nuclease, similarly as heterotrimeric RPA (Fig. 3g). Accordingly, the I14S mutation within the Rfa1 subunit alone had a similar negative effect, comparable to the I14S mutation within the heterotrimer. Rfa1-K494E mutant was able to stimulate the Dna2 nuclease better than the Rfa1-I14S, despite its lower DNA binding capacity ( Supplementary Fig. 4c-f), in agreement with results from experiments where these mutations were analyzed within the RPA trimer (Fig. 3a-c). We conclude that the integrity of the region surrounding residues F11, I14 and F15 in the N-terminal Rfa1-F domain is essential for RPA to stimulate the nuclease activity of Dna2.

Dna2 recruitment and nuclease stimulation by RPA are genetically separable
The stimulation of the Dna2 nuclease was previously attributed to the recruitment function of RPA 21,22 .
However, we observed that the RPA-I14S and RPA-SSL mutants were not impaired in their function to recruit Dna2 to overhanged DNA (Figure 4h, i, Supplementary Fig. 4g), which challenges the previous model. To understand the relationship between the stimulation of the Dna2 nuclease and recruitment to DNA in more detail, we next analyzed the capacity of the various Rfa1 domains to mediate recruitment. RPA could recruit Dna2, while Rfa1-FAB, and Rfa1-ABC were moderately impaired in this function. In contrast, Rfa1-AB could not recruit Dna2 at all, despite its strong DNA binding capacity, demonstrating the importance of both Rfa1-F and Rfa1-C domains of Rfa1 for Dna2 recruitment (Fig.   4j, Supplementary Fig. 4h). Therefore, while the DBD-C domain of Rfa1 promotes Dna2 recruitment, it is dispensable for nuclease stimulation. Under limiting concentrations, recruitment of Dna2 will likely be prerequisite for Dna2 to degrade DNA, however our experiments suggest that it may not be sufficient. The functions of RPA to promote Dna2 nuclease and recruit Dna2 are thus at least in part genetically separable.
In DNA end resection, RPA promotes the degradation of 5'-terminated ssDNA, while it inhibits 3'-ssDNA degradation, enforcing thus the correct polarity of resection. This specificity is likely dependent on the directional orientation of RPA when bound to ssDNA 13,57 , and the strong physical interaction between Dna2 and Rfa1-F at the N-terminus of Rfa1 33 . When degrading the 5'-terminated strand, Rfa1-F would be the first RPA domain that Dna2 encounters on ssDNA, while the Rfa2-D would be the first encountered domain from the other side when degrading the 3'-terminated strand. We observed that RPA-SSL was indistinguishable from wild type RPA in its function to protect 3'-terminated ssDNA from DNA degradation by Dna2. In contrast, the low affinity ssDNA binders RPA-K494E and RPA-CCAA were slightly impaired in 3'-ssDNA protection (Figure 4k, Supplementary Fig. 4i). The function of RPA to protect 3'-DNA from degradation therefore appears be a direct consequence of DNA binding of RPA and its polarity. When encountering RPA from the 3' DNA side, a lack of specific physical and functional interactions makes RPA a block for Dna2 degradation, but only when bound tightly to ssDNA.
We conclude that the Rfa1 domains that mediate Dna2 recruitment do not match those required to promote the Dna2 nuclease, and the stimulation of the Dna2 nuclease is thus not a simple consequence of Dna2 recruitment to substrate DNA. We identify mutations within a helix of the N-terminal Rfa1 domain, leading to the RPA-SSL mutant, which is proficient to facilitate Dna2 recruitment and the protection of the 3'-terminated ssDNA, but it is strongly impaired in the stimulation of the Dna2 nuclease activity acting on 5'-terminated DNA.

RPA-I14S and RPA-SSL mutants promote the Dna2-E675A helicase
The Dna2 motor may not function as a helicase to unwind dsDNA in vivo, but rather as a ssDNA translocase. The motor facilitates movement of Dna2 along ssDNA, accelerating its degradation by the Dna2 nuclease domain 19,30,32 . We used the nuclease-deficient Dna2-E675A mutant as a tool to study the effect of the RPA variants on the Dna2 motor activity. We first employed oligonucleotide-based DNA, with 45 nt-long ssDNA arms and 48 base pairs of dsDNA. We observed that the RPA-K45E, RPA-I14S and RPA-SSL mutants were indistinguishable from wild type RPA to promote the helicase of Dna2-E675A, while the ssDNA-binding mutants RPA-K494E and RPA-CCAA were moderately impaired (Fig. 5a,   b). Next, when monitoring the unwinding of 2.2 kbp-long dsDNA, RPA-SSL still promoted Dna2-E675A helicase as wild type RPA, while the RPA-K494E and RPA-CCAA did not support unwinding at all (Fig. 5c).
To define DNA unwinding by Dna2-E675A more quantitatively, we used single molecule magnetic tweezers experiments 52 . To this point, 6.1 kbp-long DNA molecules were attached on one side to a surface of a fluidic cell, and on the other side, to a magnetic bead. An internal 5'-tailed ssDNA of 40-nt in length was left free to allow for RPA binding and the loading of Dna2-E675A (Fig. 5d). The DNA was first incubated with RPA to allow binding to the 5'-overhang. Then, the reaction was initiated by adding Dna2-E675A. The DNA unwinding was quantified from the apparent DNA length measured in this set up due to different specific extensions of double-and single-stranded DNA (Fig. 5d). Despite RPA being pre-bound to the 5'-overhang, no DNA unwinding by Dna2-E675A was detected without RPA in solution (Supplementary Fig. 5a). Upon supplementing free RPA into the solution, in addition to RPA pre-bound to the overhang, we observed long-range DNA unwinding at a mean rate of 44±3 bp/s and a mean processivity 2.3±0.3 kbp similar to those obtained previously (Fig. 5e-g) 29 . Importantly, the RPA-I14S mutant supported DNA unwinding similarly as wild type RPA (Fig. 5h-j), in agreement with our bulk unwinding assays. We conclude that the RPA-I14S and RPA-SSL mutants are not affected in their function to promote the motor of Dna2.

The DBD-A and DBD-B domains specifically promote the Dna2 helicase
RPA promotes DNA unwinding by Dna2-E675A in a species-specific manner, beyond its capacity to recruit Dna2 to DNA (Fig. 1e, f, 2f). To define which regions of RPA are required to stimulate this unwinding activity, we tested the RPA fragments and subunits in unwinding assays. While heterotrimeric wild type RPA promoted the unwinding of 2.2 kbp-long dsDNA, the large subunit Rfa1 was incapable to do so (Fig. 5k, Supplementary Fig. 5b). This stands in contrast with the nuclease assays, where Rfa1 could replace the RPA heterotrimer. Unexpectedly, we found that truncations of the Rfa1 subunit, resulting in Rfa1-FAB, Rfa1-AB and Rfa1-ABC subunit combinations could readily promote DNA unwinding, in contrast to full-length Rfa1 (FABC) (Fig. 5k). Therefore, the minimal RPA region required to promote the Dna2-E675A helicase corresponds to the central Rfa1-AB domains, which typically do not associate with other proteins. Our results also indicated that the N-terminal Rfa1-F and Cterminal Rfa1-C can be inhibitory for DNA unwinding by Dna2-E675A. We conclude that the Rfa1 domains required for long-range DNA unwinding are different from those required for DNA recruitment and the stimulation of the Dna2 nuclease.
We noted that the full-length Rfa1 subunit could still facilitate apparent DNA unwinding of short oligonucleotide-based DNA (Fig. 5l, Supplementary Fig. 5c), albeit less than heterotrimeric RPA, in accord with the conclusion that Rfa1 is insufficient to stimulate the Dna2 helicase function, unlike the nuclease function. We reasoned factors other than DNA unwinding per se likely affect the product formation, such as DNA reannealing or Dna2 recruitment. To define the function of RPA in promoting DNA unwinding more accurately, we turned again to the single-molecule experiments. In agreement with the bulk assays with long DNA, we failed to observe DNA unwinding by Dna2-E675A in the presence of Rfa1 (Fig. 5m, n, Supplementary Fig. 5b). In contrast, we observed DNA unwinding with Rfa1-FAB and Rfa1-ABC, highlighting once more the function of Rfa1-AB domains of Rfa1 to promote DNA unwinding by Dna2-E675A. The processivity and rates of DNA unwinding in conjunction with the tested RPA/Rfa1 variants were very similar to wild type, with one exception (Fig. 5m, n, Supplementary Fig. 5d-i). Unexpectedly, the rates of DNA unwinding by Dna2-E675A in the presence of Rfa1-FAB were even higher than with heterotrimeric RPA, suggesting again that the Rfa1-C subunit may be inhibitory for DNA unwinding.
To understand the role of Rfa1-AB in promoting DNA unwinding by Dna2-E675A, we applied structure modeling. Based on sequence analysis and published mammalian DNA2 and RPA structures 4,22 , we performed 3D docking to model the potential Dna2 and Rfa1-AB interaction (Supplementary Fig. 6a).
In the obtained model, a conserved and exposed tyrosine residue (Y193) of Rfa1-A could act as an anchor to interact with an apolar site on Dna2, near the region where the 3'-end of the threaded DNA lies in DNA2 (Fig. 6a, Supplementary 6a). To block this potential interaction, we designed Rfa1-AB-Y193A and Rfa1-AB-Y193D mutants to disrupt the tyrosine, and Rfa1-AB-S191K to create a steric hindrance at the interaction site. We then expressed and purified the Rfa1-AB variants (Fig. 6b, Sup-plementary Fig. 6a). As indicated in the structure model, the mutations are located on the surface of Rfa1-AB on the opposite side of the ssDNA binding site. The mutations consequently did not notably affect the binding of Rfa1-AB to ssDNA (Fig. 6b, c, Supplementary Fig. 6c).
We next analyzed the Rfa1-AB mutants for their capacity to promote long-range DNA unwinding by Dna2-E675A. We observed that all mutants exhibited reduced DNA unwinding in conjunction with Dna2-E675A compared to wild type Rfa1-AB. In particular, the Rfa1-AB-Y193D mutant was ~2.5fold less efficient in promoting Dna2-E675A unwinding activity in assays with various protein concentrations, as well as in a kinetic experiment (Fig. 6d, e), validating the docking model. Next, we combined the S191K and Y193D mutations in the heterotrimeric RPA, creating the SKYD mutant ( Supplementary   Fig. 6d). Although RPA-SKYD bound ssDNA similarly as wild type (Fig. 6f, g and Supplementary Fig.   6e), it notably reduced DNA unwinding by Dna2-E675A (Fig. 6h, i). Therefore, residues located on the outer surface of the major DNA binding domain A of RPA, around Y193, are not involved in ssDNA binding, but are likely required to transiently interact with Dna2 to promote its motor function 22 . The RPA-SKYD mutant only weakly reduced nucleolytic degradation of 2.2 knt-long ssDNA ( Fig. 6j and Supplementary Fig. 6f), in contrast to RPA-SSL, which brought about a dramatic DNA degradation defect, in accord with Fig. 4f, g that utilized oligonucleotide-based DNA. These experiments confirm that Rfa1 residues within DBD-F primarily promote the Dna2 nuclease, while Rfa1 residues within DBD-A promote the motor activity of Dna2.

RPA-SSL and SKYD mutants are impaired in stimulating DNA end resection.
It has been established that Dna2, Sgs1 and RPA comprise a minimal protein complex capable of DNA end resection in vitro 26,27 . Both nuclease and motor activities of Dna2 are required for maximal DNA end resection capacity 30,32 . To verify the importance of the newly-identified RPA variants for a physiologically relevant process such as DNA end resection, we explored the capacity of the RPA variants to function with Dna2 and Sgs1 in reconstituted resection assays. We observed that RPA-SSL and RPA-SKYD mutants both showed notably reduced capacity to resect DNA, despite their proficient ssDNA binding affinity (Fig. 7a, b). These results confirm that RPA, beyond its single-stranded DNA binding functions, has specific functions to recruit and activate Dna2 helicase-nuclease activities to promote DNA end resection in DSB repair.

DISCUSSION
Here we define the functional interactions between the eukaryotic ssDNA binding protein RPA and the S. cerevisiae Dna2 nuclease-helicase, which participates in diverse DNA metabolic processes including replication and recombination. RPA is known to specifically promote Dna2 recruitment to overhanged or flapped DNA, and promote both nuclease and helicase activities of Dna2 16,21,22,26,27,29 . Human mitochondrial SSB, which resembles prokaryotic ssDNA binders, instead inhibits all Dna2 activities, which underlines the specific nature of the interaction between cognate yeast RPA and yeast Dna2. Dna2 was shown to bind the large Rfa1 subunit of RPA 33 , and the stimulatory activity of RPA on the enzymatic activities of Dna2 was thought to be a consequence of enhanced recruitment. Our data show that the stimulation of the Dna2 nuclease and helicase activities is more direct than anticipated. We demonstrate that RPA has the capacity to specifically promote both enzymatic activities of Dna2 in addition to its recruitment function.
Previously, Dna2 was shown to physically interact with the N-terminal (Rfa1-F) and less efficiently with the additional Rfa1-A, -B and -C subunits 22,33 . We show that both Rfa1-F and Rfa1-C domains are required for efficient recruitment of Dna2 to overhanged DNA when coupled with the central Rfa1-AB domains (Fig. 7c). In contrast, the Rfa1-AB construct alone, which binds DNA with a high affinity, fails to recruit Dna2. The contribution of Rfa1-F and Rfa1-C domains towards recruitment of Dna2 is likely dependent on the described physical interactions 22,33 . Whether RPA remains a part of the nucleoprotein complex upon Dna2 recruitment has been controversial, likely due to difficulties interpreting the mobilities of protein-bound complexes in electrophoretic shift assays 16,21,22 . Using mass photometry, we observed that upon the recruitment of Dna2, RPA remains a part of the nucleoprotein complex as an intermediate. RPA is thus positioned to be capable to stimulate also the enzymatic activities of Dna2 that occur downstream of recruitment (Fig. 7c).
We show that Rfa1 alone can promote the Dna2 nuclease equally well as heterotrimeric RPA. The Rfa1-C domain is inhibitory in this regard. On the other hand, the N-terminal Rfa1-F domain, coupled with Rfa1-AB, has a critical function to promote cleavage of 5'-terminated ssDNA, although it does not contribute to ssDNA binding (Fig. 7d, Supplementary Fig. 7). We identify the RPA-SSL variant with mutations in a conserved helix embedded in the N-terminal Rfa1-F domain. The mutations likely affect the local structure that extends to residues exposed to the surface. The resulting mutant is proficient in DNA binding and Dna2 recruitment, but strongly impaired in the stimulation of DNA degradation by Dna2, as observed with both oligonucleotide-based and plasmid-length DNA substrates. Therefore, recruitment of Dna2 and stimulation of its nuclease activity are genetically separable.
As with the Dna2 nuclease, RPA promotes the Dna2 helicase function beyond recruitment. The stimulation of DNA unwinding by RPA also cannot be explained by non-specific prevention of DNA reannealing. When nuclease-deficient Dna2-E675A, was prebound to DNA, making recruitment irrelevant, RPA promoted DNA unwinding much more than mtSSB. Under these conditions, mtSSB was expected to prevent reannealing equally well as RPA. Likewise, in single molecule assays, RPA prebound to the ssDNA overhang did not support long-range DNA unwinding by Dna2-E675A in absence of additional free RPA. These experiments collectively suggest that RPA is required for the stimulation of the motor activity of Dna2, and becomes essential for the unwinding of long DNA substrates. Surprisingly, the RPA-SSL mutant that largely failed to promote the Dna2 nuclease activity stimulated DNA unwinding by Dna2-E675A equally well as wild type RPA, showing that that the stimulation of the nuclease and helicase activities involves different regions of RPA. In contrast to recruitment and nuclease stimulation, Rfa1 alone could not promote long-range DNA unwinding by Dna2. Unexpectedly, the Rfa1-AB, as well as the Rfa1-ABC and Rfa1-FAB fragments activated the helicase activity of Dna2-E675A almost as well as the heterotrimeric RPA. Therefore, when using only Rfa1, the N-terminal Rfa1-F and C-terminal Rfa1-C domains were surprisingly inhibitory for long-range DNA unwinding by Dna2-E675A. Structure modeling and three-dimensional molecular docking identified a potential interaction site between the surface of Rfa1-A and Dna2. We could validate this model by mutating a conserved tyrosine residue in Rfa1-AB (Y193), which was important for the ability of the Rfa1 fragment to promote DNA unwinding. Consequently RPA-SKYD, which carried the S191K mutation in addition to Y193D within the RPA heterotrimer, was proficient to bind ssDNA, but impaired to stimulate DNA unwinding. These experiments showed that Dna2 also interacts with the Rfa1-A domain, which is quite unusual for RPA-interacting proteins.
Together, our data show that a certain level of ssDNA binding activity by RPA is prerequisite for Dna2 stimulation. However, the full complement of ssDNA binding capacity of RPA may not be necessary, as mutants with reduced ssDNA binding retained Dna2 stimulation. Rather, the stimulatory activity of RPA depends additionally on specific interactions between distinct RPA domains and Dna2. Most surprisingly, we found that different RPA domains are responsible for stimulating the diverse Dna2 functions including DNA binding, DNA nuclease and DNA helicase activities, showing that the functions of Dna2 are intertwined with RPA much more than previously assumed ( Fig. 7d and Supplementary   Fig. 7). We show that both RPA-SSL (strongly impaired to promote the Dna2 nuclease, but not the helicase activity) and RPA-SKYD (strongly impaired to promote the Dna2 helicase, but with only a weak effect on the nuclease activity) showed reduced capacity to function in DNA end resection together with Dna2 and Sgs1, underpinning the importance of both enzymatic activities of Dna2 in DSB repair.    k Quantification of nuclease assays such as shown in Supplementary Fig. 3m, n. Error bars, SEM; n=3. e Quantitation of ssDNA (93 nt, 0.1 nM, in molecules) binding by wild type RPA and RPA-SSL mutant in assays such as shown in Supplementary Fig. 3c and 4b. RPA is replotted as in Fig. 3b for reference. Error bars, SEM; n=3.