FANCD2 and RAD51 recombinase directly inhibit DNA2 nuclease at stalled replication forks and FANCD2 acts as a novel RAD51 mediator in strand exchange to promote genome stability

Abstract FANCD2 protein, a key coordinator and effector of the interstrand crosslink repair pathway, is also required to prevent excessive nascent strand degradation at hydroxyurea-induced stalled forks. The RAD51 recombinase has also been implicated in regulation of resection at stalled replication forks. The mechanistic contributions of these proteins to fork protection are not well understood. Here, we used purified FANCD2 and RAD51 to study how each protein regulates DNA resection at stalled forks. We characterized three mechanisms of FANCD2-mediated fork protection: (1) The N-terminal domain of FANCD2 inhibits the essential DNA2 nuclease activity by directly binding to DNA2 accounting for over-resection in FANCD2 defective cells. (2) Independent of dimerization with FANCI, FANCD2 itself stabilizes RAD51 filaments to inhibit multiple nucleases, including DNA2, MRE11 and EXO1. (3) Unexpectedly, we uncovered a new FANCD2 function: by stabilizing RAD51 filaments, FANCD2 acts to stimulate the strand exchange activity of RAD51. Our work biochemically explains non-canonical mechanisms by which FANCD2 and RAD51 protect stalled forks. We propose a model in which the strand exchange activity of FANCD2 provides a simple molecular explanation for genetic interactions between FANCD2 and BRCA2 in the FA/BRCA fork protection pathway.

genome stability have yet to be determined.Fanconi anemia is a rare disease of bone marrow failure, de v elopmental abnormalities, and cancer predisposition.At the cellular le v el it is diagnosed by sensitivity to DNA interstrand crosslink (ICL)-inducing agents and genome instability.Fanconi anemia is a multigenic disease defined by at least 22 complementation groups, including many regulatory components, nucleol ytic activities, and homolo gy dir ected r epair (HDR) genes.The component genes suggest a coherent pathway for maintaining genome stability during DNA replication that goes beyond ICL repair and includes the response to many additional types of r eplication str ess ( 1 , 2 ).The multigenic character of the FA pathway lends itself to a comprehensi v e genetic and biochemical dissection (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14).
FANCD2 is a key regulator of the FA pathway and the focus of our current studies ( 5 , 15 ).During canonical r eplication-coupled r epair of ICLs, after a r eplication fork encounters an ICL, FANCD2 and a related protein FANCI, are phosphorylated by activated ATR kinase.A F ANCD2 / F ANCI heterodimer is also formed, and FANCD2 in this heterodimer, but not free FANCD2, is mono-ubiquitylated by the FA core complex, containing nine FA proteins, including the FANCL E3 ligase complex and se v eral associated proteins.FANCD2-ubi is involv ed in both activation of repair e v ents and also is dir ectly r equired in the later enzymatic repair steps at strand breaks ( 5 , 16-20 ).The role of ubiquitin is to enforce stable binding of F ANCD2 / F ANCI to DN A, specificall y by clamping F ANCD2-ubi / F ANCI heterodimers onto DNA for DNA repair (21)(22)(23)(24).
In addition to its role in ICL repair, FANCD2 is also involved in the recovery of stalled replication forks, irrespecti v e of the source of DNA damage causing replication stress ( 6 , 7 , 11 , 14 ).Se v eral studies implied that nonubiquitylatable F ANCD2 (F ANCD2-K561R) could not restore fork protection to patient-deri v ed FANCD2-defecti v e cells ( 7 , 25 ), Other results, howe v er, support that FANCD2 is likely to have constitutive functions, at least for low levels of r eplication str ess, such as endogenous str ess ( 1 , 2 ).With respect to ubiquitylation, the study of FANCD2 knockout and knock-in cell lines showed that cells expressing only non-ubiquitylatable FANCD2-K561R had much less se v ere phenotypes than cells with a FANCD2 knockout ( 26 ).Complementary studies showed that mutants defecti v e in the trans-acting FA core complex components responsible for ubiquitylation of FANCD2 are less sensiti v e to replication fork stalling agents than FANCD2 knockdowns or knockouts ( 27 ).Importantly, one of us reported that FANCD2 can protect the stalled forks by different mechanisms than FANCA / C / G, members of the core complex ( 28 ).FANCD2 has been shown to interact with RAD51, a key player / regulator in fork protection, and to do so in a ubiquitylation-independent but HU-stimulated manner ( 29 ).At a stalled fork, induced CMG disassembly disassociates FANCD2 and FANCI, which leads to fork instability ( 30 ).FANCD2 also has FANCI independent functions ( 11 , 27 , 31 ).FANCD2 deficient cells are HU and aphidicolin (a DN A pol ymerase inhibitor) sensiti v e, while FANCI cells are not ( 27 ).These results stimulated our interest in studies of ubiquitin-and FANCI-independent roles of FANCD2.
Fork protection, operationally, implies protection from nucleases.Se v eral nucleases hav e been implicated in nascent DNA degrada tion a t DNA structures arising a t stalled forks ( 6 , 7 , 28 , 32 , 33 ).DNA2 helicase / nuclease is of particular interest because it is essential for replication in normal yeast cells, and in metazoans, it is essential for normal embryonic de v elopment (34)(35)(36).Why DNA2 is essential has remained a matter of de bate, how e v er.While synthetic lethality with FEN1 deficiency in yeast and biochemical characterization suggests that DNA2 might function in FEN1-independent Okazaki 5 flap r emoval ( 35 , 36 ), mor e r ecent studies show that DNA2 has additional important functions, raising the question of which really makes it essential.DNA2's ability to remove long 5 (or 3 ) ssDN A fla ps could be used during non-canonical Okazaki fragment processing in the presence of Pif1 ( 37 , 38 ), also see (Hill et al., 2020, unpublished in Biorixv).It could also pr omote contr olled resection during replication fork stalling for replication fork protection, to pre v ent the accumulation of aberrant re v ersed fork intermediates or gaps and for efficient replication fork restart ( 32 , 39 , 40 ).DNA2 is thought to be especially important a t dif ficult-to-replica te sequences, such as the rDNA (41)(42)(43), telomeres (44)(45)(46), and centromeres ( 47 ).Multitasking DNA2 is also involved In DNA repair.DNA2 performs long-range resection of DSBs during homologous recombination ( 48 ), in conjunction with MRE11, to provide 3 ends for BRCA2-mediated RAD51 filament formation and strand invasion.We discovered that DNA2-deficient cells are sensiti v e to inter-or intr a-str and crosslinks induced by cisplatin or formaldehyde.Paradoxically, the depletion of DNA2 in cells deficient in FANCD2 rescued ICL sensitivity in FANCD2 mutants, in keeping with DNA2 becoming toxic in the absence of FANCD2 fork protection ( 49 , 50 ).Se v eral studies confirm tha t DNA2-media ted overresection of nascent DNA occurs at a stalled replication fork when FANCD2 is absent ( 26 , 28 , 30 , 32 , 33 , 36 , 51-54 ), suggesting that controlled resection by DNA2 at forks is essential for replication and repair, and to preserve genome stability.The question remains as to how DNA2 is precisely controlled at replication forks.Answering this question is essential to understanding how both FANCD2 and DNA2 are involved in fork protection and maintenance of genome stability and thus understanding their roles in cancer de v elopment and treatment.
RAD51 depletion can also be inhibitory to DNA2media ted degrada tion of nascent DNA in vivo ( 32 ), since RAD51 has been shown recently to promote fork reversal using its recombination activity ( 30 ).Furthermore, a dominant negati v e RAD51 mutant leads to e xcessi v e degradation of nascent DNA in a RAD51 T131P / WT heterozygote and this over-resection is pre v ented by depletion of DNA2 ( 53 ).FANCD2 and RAD51 are epistatically linked in fork protection ( 7 ).In vivo , however, it is not known if RAD51 and FANCD2 act independently or together and whether both are required to promote fork re v ersal and inhibit DNA2.Addressing the role of RAD51 in protection from degradation is difficult because of the fact that RAD51 is r equir ed f or f or k re v ersal in all pathways identified to date ( 28 , 55 ).
Recently, se v eral studies have suggested that FANCD2 and BRCA2, the RAD51 mediator, perform parallel or compensatory functions in fork protection and fork recovery after the collapse ( 25 , 56 , 57 ).Since BRCA2 is thought to stabilize RAD51 filaments, we hypothesized that FANCD2 may provide a backup source of this BRCA2 function in response to replication stress.This mechanism is supported by the fact that FANCD2 alone and F ANCD2 / F ANCI heterodimers interact physically with RAD51 ( 29 , 32 , 58-60 ).F ANCD2 / F ANCI comple xes hav e been shown to increase RAD51 le v els on DNA, but the specific contribution of FANCD2 itself and the relationship of this observation to suppression of BRCA2 −/ − defects has not been established.
In this work, we studied the mechanisms by which FANCD2 and RAD51 mediate fork protection.Our in vivo results confirm that FANCD2 is r equir ed to protect stalled replication forks from DNA2-dependent overresection after acute stress.We identified at least two potential mechanisms by which FANCD2 protects nascent DNA from nucleolytic resection in vitro : (1) FANCD2 inhibits DNA2 nuclease activity directly and (2) FANCD2 stabilizes RAD51 ssDNA filaments which pre v ent nucleolytic digestion by multiple nucleases.Surprisingly, FANCD2, promotes RAD51-mediated strand e xchange acti vity by stabilizing RAD51 on ssDNA.The ability to stimulate strand exchange suggests that FANCD2, like BRCA2, is a RAD51 mediator.Since the strand exchange activity is required for fork reversal, our work suggested that FANCD2 may also be involved in fork reversal at a stalled fork, like BRCA2 ( 28 , 30 ).This provides a novel mechanistic explanation for the dependency of BRCA2 −/ − tumors on FANCD2, and the suppression of BRCA1 / 2 −/ − phenotypes b y elev ated le v els of FANCD2 ( 25 , 56 , 57 ), Thus, our results add a major new dimension to how FANCD2 deficiency leads to loss of fork protection, leading to genome instability ( 7 ).

Reagents and materials
See Supplementary Table S1 in Supporting Material.

Cell culture
U2OS, A549 and PD20 and PD20 with FANCD2 complemented cells were cultured in DMEM medium with 10% FBS.

Nuclear fractionation
Cells (1 × 10 6 ) were harvested and washed with PBS, then lysed on ice for 20 min with 100 l H150 buffer, which contains 50 mM HEPES (pH7.4), 150 mM NaCl, 10% glycerol, 0.5% NP-40 and protease inhibitor cocktail (Roche).The lysate was spun for 10 min at 5000g, and the supernatant is the cytoplasmic fraction.The pellet was washed two times with H150 lysis buffer, and the supernatant discarded.The pellet is the nuclear fraction.The pellet was resuspended in PBS (20 l) and 20 l 2 × SDS loading buffer and boiled for western blot.

Immunofluor escence f or native Br dU staining and EdU staining
BrdU staining was carried out as described ( 61 ).Briefly, cells (1 × 10 5 labeled with BrdU and EdU as described in the legend to Supplementary Figure S5) were plated on coverslips, washed with PBS, pre-extracted with ice cold 0.5% Triton-X100 for 4 min, then fixed with 4% paraformaldehyde for 10 min, permeabilized with 0.1% Triton-X100 for 2 min and then washed with PBS 3 times.Blocking was carried out with 1% BSA in PBS for 1 h.A 1 ml click reaction containing 5 l 1 mM Azide-488 (Invitrogen), 100 l 20 mg / ml sodium ascorbate, 20 ul 100 mM CuSO 4 ) was performed to detect incorporated EdU.Then FANCD2 antibody (1:200 in blocking buffer) was added and incubated overnight a t 4 • C .For BrdU staining, slides were incuba ted with BrdU and FANCD2 primary antibody overnight at 4 • C. The slides were washed in PBS three times and then incubated with secondary antibody (1:200, Alexa Fluor 594 and 488 fr om Invitr ogen) for 1 h at room temperature.The slides were washed with PBS 3 times and mounted with Prolong Gold AntiFade Reagent with DAPI (Invitrogen P36941).

Plasmid and siRNA transfection
A549 and U2OS cells were plated the day before transfection.20 nM siRNA was used for single and 16 nM for each siRNA in co-transfection.Cells was transfected with Genmute and labeled as indicated 72 hours post-transfection.DNA2 plasmid transfection was described previously ( 36 ).

DNA fiber assay
DNA fiber spreading and staining were performed as previousl y described ( 36 ).Briefly, 1000 labeled cells (2 l, 500 cells / l) on slides were half dried, 10 l lysis buffer (0.5% SDS, 200 mM Tris-HCl pH 7.4, 50 mM EDTA) was added, followed by incubation for 6 min at room temperature.The slide was tilted to 15 degrees to allow the DNA to run slowly down the slide.Slides were air dried for at least 40 minutes and fixed for 2 min in 3:1 methanol: acetic acid in a coplin jar.Slides were dried in a hood for 20 min.Slides were treated with 2.5 M HCl for 70 min for dena tura tion and then washed with PBS 3 times and blocked with 10% goat serum in PBST (0.1% Triton-X100 in PBS) for 1 h.Slides were incubated with the rat anti-BrdU and mouse anti-BrdU antibody, 1:100, for 2 h, washed 3 times with PBS, and then incubated with secondary antibody (Goat anti-Mouse 488 and Goat anti-Rat 594, Invitrogen) at 1:200.Slides were imaged with immunofluorescence microscopy and fiber length measured by Nikon software.Statistical analyses were completed using Prism.An ANOVA test was used when comparing more than two groups followed by a Dunnett multiple comparison post-test.

Neutral COMET assay
The neutral COMET assays were performed in accordance with the manufacturer's (Trevigen) instructions.Cells were trypsinized and washed, then palleted, resuspended with low melt agarose, then dropped on the slides.After cooling down, the slides were incubated in cold lysis buffer (Trevigen) for 1 h, then incubated in running buffer for 30 min, and then subjected to electrophoresis at 21 V for 45 min.Slides were then immersed in precipita tion buf fer (Trevigen) and 70% ethanol for 30 min, respecti v ely.Slides were dried overnight and stained with SYBR green I (Thermofisher).Slides were imaged with fluorescence microscope with FITC channel.

Immunoprecipitation
For FLAG pulldown assays and immunoprecipitation assays, 293T cells were transfected with or without RAD51 vector (or FLAG-DNA2 vector) using the Polyjet (Signa-Gen SL100688) transfection reagent.24 h after transfection, the cells were incubated with or without 2 mM HU for 3 h.Cells (1 × 10 7 ) were collected and lysed by brief sonication and incubation in the immunoprecipitation (IP) buffer H150 (50 mM HEPES-KOH (pH7.4), 150 mM NaCl, 0.1% NP40 and 10% glycerol) with protein inhibitor cocktail (Thermo Fisher) for 30 min.After centrifugation (20 000g, 15 min, 4 • C), the supernatants were collected, and the protein concentration determined.Cell lysate (1 mg) was precleaned with 10 l Protein A / G beads (Thermo #88802) for 1 h.After removing beads, the lysate was incubated with 2 g (1 g / l) anti-RAD51 (ab133534 Abcam) or anti-FLAG M2 magnetic beads for FLAG pulldown (Sigma).Then 10 l Protein A / G magnetic beads were added and incubated overnight at 4 • C. The beads were washed three times with the IP buffer H150 and boiled in 1 × SDS-PAGE loading buffer directly.The DNA2 and FANCD2 were analyzed by western blot analysis.

Oligonucleotides
Oligonucleotide substrates for enzymatic assays were labeled at the 5 end with 32 P using polynucleotide kinase.The sequences are listed in the Supplementary Table S1.For DN A2 assays, single-stranded DN A was JYM945 ( 62 ).The forked substrate was designated 87 FORK.The 5 flap substrate was LU 5 FLAP.The 3 flap substrate was LU 3 .The re v ersed for k with b lunt ends consisted of 4 oligonucleotides: str and 1, str and 2, str and 3 and str and 4 in the Supplementary Table S1 ( 63 ).The reversed fork with 5 overhang consisted of strand1L, strandFANCD2, strand3, and strand4.
The MRE11 nuclease duplex substrate was formed by annealing 5 labeled JYM945 to JYM925 ( 62 ).This was also used for binding of RAD51 to dsDNA.For the EXO1 assay, a hairpin with a 3 overhang was used.
For RAD51 binding, JYM945 was used.For RAD51 strand exchange assays the single-stranded DNA was EX-TJYM925: The 60mer duplex was formed by annealing labeled JM945 to JM925 (see MRE11 substrate).

Proteins
Recombinant human RAD51 was from Abcam (ab81943) and tested for ATPase , strand exchange , and DNA binding .RuvC was Abcam (ab63828).MRE11 was the gift of Tanya Paull (UT Austin) and EXO1 (0.77 mg / ml) was a gift from Paul Modrich, Duke Uni v ersity.Sources of FANCD2 and DNA2 are described in the text or figure legend describing the experiments in which they were used.

FANCD2-his purification from E. coli
Human FANCD2 protein was purified from E. coli as previously described ( 64 ).The FANCD2 vector was transformed into BL21(DE3) CodonPlus (Agilent Technologies 230280) cells.Twenty liters of transformed cells were amplified at 30 • C, 250 rpm.FANCD2 protein was produced by adding 0.5 mM IPTG at 16 • C for 18 hours, when the cell density reached an OD 600 = 0.6.The E. coli cells were harvested and pelleted and lysed in Buffer A (50 mM Tris-HCl PH8.0, 500 mM NaCl, 5 mM 2-mercaptoethanol, 1 mM phen ylmethylsulf on yl fluoride (PMSF), 12 mM imidazole, and 10% glycerol), and disrupted by sonication.The lysate was centrifuged at 20 000g at 4 • C; the supernatant was mixed gently by the batch method with 3ml of Ni-NTA agarose beads, at 4 • C for 1 h.The beads were packed into an Econo-column, and were washed with 67 column volumes of buffer A. The His-tagged FANCD2 were eluted with a 20 column volumes linear gradient of 12-400 mM imidazole in buffer A. The peak fractions were collected.To remove His tag from the FANCD2 protein, thrombin protease (2U / mg GE healthcare) was added, and the sample was then dialyzed against 4L of buffer B (20 mM Tris-HCl, pH8.0, 200 mM NaCl, 5 mM 2-mercaptoethanol, 10% glycerol).Afterward, the sample was passed through a Q Sepharose Fast Flow (2.5 ml, GE Healthcare) column.The resin was washed with 60 column volumes of buffer B containing 250 mM NaCl.Human FANCD2 was then eluted with a 20column volume linear-gradient of 250 mM-450 mM NaCl in buffer B. The peak fractions were collected, and human FANCD2 was further purified by gel filtration chromatography on a Super de x 200 column (GE Healthcare) equilibra ted with Buf fer B containing 200 mM NaCl.The purified FANCD2 was concentrated, frozen in aliquots, and stored a t -80 • C .The concentra tion of purified FANCD2 was determined by the Bradford method, using BSA as standard.

FLAG-DNA2 purification from mammalian cells
The FLAG-DNA2 expression and purification procedure was as described previously ( 44 ).In brief, whole cell lysates were incubated with the M2 FLAG magnetic beads (Sigma) for at least 6 h in cold room.After e xtensi v ely washing with a buffer containing 50 mM Tris-Cl (pH 7.5) and 500 mM NaCl, the bound proteins were eluted with 3 × FLAG peptide (Sigma).The purity of DNA2 proteins was analyzed by 4-15% gradient SDS-polyacrylamide electrophoresis (SDS-PAGE) and Coomassie brilliant blue staining, and the concentration was determined by comparison to BSA after Coomassie blue staining of SDS gels.

Mapping the FANCD2 binding domain in DNA2
Mutant FLAG-DNA2 proteins were prepared using sitedirected mutagenesis.The N-terminal deletions were made using the HiFi DNA cloning kit from NEB to excise portions of the N-terminus of the gene, while C-terminal deletions were made by the insertion of a stop codon earlier in the gene construct.Coimmunopr ecipitations wer e performed by ov ere xpressing the DNA2 proteins in HEK-293T cells prior to making cell lysates.FANCD2 was added to the lysates to a final concentration of 2 nM protein to ensure measur able inter action with DN A2.The FANCD2:DN A2 complex was pulled down using a FANCD2 antibody attached to magnetic beads.The beads were washed prior to eluting the samples using SDS loading buffer, and the samples were analyzed by western blot using a 3 × FLAG antibody.

Strand e x change assays
Single-stranded DNA (EXTJYM925) was preincubated in the presence of RAD51 and FANCD2 in a reaction mixture containing 25 mM TrisOAc (pH 7.5), 2 mM MgCl2, 2 mM CaCl2, 2 mM ATP, 1 mM DTT and 0.1 mg / ml BSA for 5 min at 37˚C for filament formation.Following pre-incubation, dsDNA (5 labeled JYM945 annealed to JYM925) with the labeled strand complementary to the filament, was added to the reaction mixture and incubation was continued for an additional 30 min at 37˚C for strand exchange.Reactions were terminated by the addition of proteinase K and SDS to 0.5 mg / ml and 0.25% respecti v ely and incubated for 10 min at 37 • C. 1 l of Loading Buffer (2.5% Ficoll-400, 10 mM Tris-HCl, pH 7.5, and 0.0025% xylene cyanol) was added and samples were loaded on an 8% nati v e gel using 29:1 30% acrylamide solution.Gels were run at 100 V (constant voltage) for 4 h.

Biotin pull-down assays for RAD51 and FANCD2 association with overhang DNA
The protocol was adopted from Jensen et al .Briefly, the oligonucleotide substrate Bio-RJ-PHIX-42-1 composed of the same sequence as RJ-PHIX-42-1 but containing a 3 biotin modification was obtained from IDT (Integrated DN A Technolo gies) and PAGE purified.The biotinylated 3 overhang substrate was generated by annealing Bio-RJ-PHIX-42-1 to oligonucleotide RJ-167 at a 1:1 molar ratio in STE buffer.Competitor heterolo gous dsDN A was similar ly gener ated by annealing PAGE purified oligonucleotides Oligo#90 and Oligo#60.
For pull-down, RAD51 and FANCD2 proteins were incuba ted in Buf fer S (25 mM TrisOAC pH 7.5, 1 mM MgCl 2 , 2 mM ATP, 1 mM DTT and 0.1 g / l BSA) for 15 min at 37˚C followed by the addition of 3 overhang DNA (162 nt RJ-167 annealed to 42 nt 3 Bio-RJ-PHIX-42-1) and competitor heterologous dsDNA (90mer, Oligo #90 / Oligo #60 oligonucleotides) and the reaction was incubated for an additional 5 min at 37˚C.Where DNA was omitted, TE buffer was used and similarly, respecti v e proteins stora ge b uffers were used where proteins were omitted.DNAprotein complexes were captured by adding the reaction mixtures to 2.5 l of MagnaLink Streptavidin magnetic beads (Solulink) pre-washed by excess Buffer S supplemented with 0.1% Ipegal CA-630 and rotating for 10 min at 25˚C.Bead complexes were then washed with excess Buffer S supplemented with 0.1% Ipegal CA-630.Protein was then eluted by re-suspending in 15 l of 2 × protein sample buffer and heating at 54˚C for 4 min.The elution fraction was then loaded into a Bis-Tris protein gel for western analysis.Following transfer, the membrane was cut horizontally at the 70 kDa marker to separately probe for RAD51 and FANCD2.The lower half was probed using 1:1000 diluted ␣-RAD51 (Abcam) and the upper half using 1:1000 diluted ␣-FLAG (ThermoFisher) to detect FANCD2.Anti-mouse (LI-COR) secondary antibody diluted 1:10 000 was used and membranes were imaged via Odyssey imaging system.Bands were quantified using ImageQuant (Cytiva) software.

Nuclease and DNA-dependent ATPase assays
DNA2 nuclease assay.FANCD2-His or FANCD2-His diluent was incubated in DNA2 nuclease reaction mix (50 mM HEPES-KOH, pH 7.5, 5 mM MgCl 2, 2mM DTT, 0.25 mg / ml BSA) for 30 min at 4 • C. DNA2, preincubated with substrate (87 fork, 1.5 nM molecules) for 5 min on ice, was added and the reaction was incubated for 30 min at 37 • C. See Supplementary Table S1 for substrate sequences.Following incubation, proteinase K and SDS were added to 1 mg / ml and 0.5%, respecti v ely, and incubation continued for 10 min at 37 • C. Denaturing termination dye (2X: 95% deionized formamide, 10 mM EDTA, 0.1% bromophenol blue and 0.1% xylene cyanol) was added and the mixture boiled for 5 min.Samples were run on a sequencing gel and the gel analyzed by phosphor imaging.Product formation was determined by dividing the product band by the total DNA in each lane.We calculate inhibition by determining the % product and normalizing to the control lane with respecti v e nuclease alone and no FANCD2.

Quantification and statistical analysis
Statistical analyses were completed using Prism.An ANOVA test was used when comparing more than two groups followed by a Dunnett multiple comparison posttest.A two-tailed t -test was used to compare two samples with normally distributed data.No statistical methods or criteria were used to estimate sample size or to include / exclude samples.

FANCD2 is r equir ed f or r eplication f ork protection after acute replication stress
Multiple pathways are involved in stalled replication fork repair, and forks undergo progressi v e changes in architecture during chronic stalling ( 3 , 68-70 ).Previous studies showed that FANCD2 can protect nascent DNA from degradation upon the replication stress.Howe v er, the major question is how FANCD2 protects the stalled fork and what structure it is being acted upon ( 3 , 71 ).We began by verifying over-resection in FANCD2-deficient cells using RPA2 phosphorylation as a surrogate for measuring ssDNA arising during resection in the presence of HU.Cells wer e tr eated with HU for 0-8 h and nuclear extracts wer e pr epar ed.At all-time points, we observed drastically increased RPA-p le v els (resection) in nuclear extracts of HU-treated PD20 FANCD2 −/ − deficient cells compared to FANCD2-complemented cells (PD20:FANCD2), where there was very little resection (Figure 1 A).This is consistent with other published results showing that FANCD2 protects stalled forks from nascent DNA degradation.
We next asked if knockdown of DNA2 can rescue the ov er-resection observ ed after 4 h of HU treatment, in keeping with the fact that the cisplatin and formaldehyde sensitivity of the FANCD2 −/ − PD20 patient cell line can be suppressed by DNA2 knockdown ( 50 ).Using single molecule tracking of nascent DNA before and after brief (4 h) HU tr eatments, as r eported pr e viously, we see ov er-resection in FANCD2 depleted cells (Figure 1 B), supporting use of RPA-p as the readout for resection shown in Figure 1 A. We then confirm that degradation of nascent DNA in FANCD2-deficient cells upon HU treatment can be rescued by the knockdown of DNA2 and also by the knockdown of MRE11 or EXO1 nuclease (Figure 1 B and C).WRN helicase is known to collaborate with DNA2 in resecting DNA ( 72 ), and in keeping with this, co-depleting WRN also rescues the nascent DNA degradation (Figure 1 C).These results are also consistent with previous studies ( 32 ).We verified that over-resection required fork re v ersal by SMAR-CAL1 or ZRANB3 but not cleavage by MUS81 / SLX4, as one of us had previously demonstrated (Supplementary Figure S1A and S1B) (28).We conclude that MRE11 and DNA2 may function as alternati v e nucleases or function sequentially in stalled fork processing, as they do at DSBs ( 73 ).

FANCD2 inhibits DNA2 nuclease activity in vitro providing a mechanism for FANCD2's in vivo role in fork protection
Since FANCD2 and DNA2 have been shown to inter act in vi vo in a DNA-independent fashion, suggesting a pr otein / pr otein interaction ( 49 ), we tested whether FANCD2 dir ectly r egula tes degrada tion by DNA2 nuclease.FANCD2-His was purified from SF9 insect cells (Supplementary Figure S1C) ( 74 ) and was shown to bind ds-DNA (Supplementary Figure S1D) and to be free of nuclease activities under the conditions used here (Supplementary Figure S1E).When FANCD2 was added to a DNA2 nuclease reaction containing a forked substrate, significant inhibition of DNA2 nuclease was observ ed, e v en in the presence of high le v els (5 nM) of DNA2 (Figure 1 D).Inhibition is likely due to FANCD2 protein and not reaction conditions since all reactions contained the same amount of FANCD2 diluent.In these experiments, the substrate partially mimics a stalled replication fork with singlestranded DNA arms at the dsDN A junction.DN A2 processes substrates with se v eral different configurations, such as unligated 5 flaps on Okazaki fragments or on base excision repair intermediates, single-stranded DNA, or 5 overhangs on r egr essed r eplication forks during r eplication fork stress / stalling or DSB resection during homologous recombination.As shown in Supplementary Figure 1E, S1F and S1G, FANCD2 inhibits DNA2 nuclease on each of these structures.(Sequences of all oligonucleotide substrates used are provided in Supplementary Table S1).The substrates mimicking re v ersed for k DNAs with or without a 5 overhang (Figur e 1 E) wer e generated and validated as described in Supplementary Figure S1H.
FANCD2 forms stable complexes with FANCI in vivo and in vitro , although only 20% of the FANCD2 in the cell co-IPs with FANCI ( 21 , 24 ).We next tested if FANCI also inhibits DNA2 nuclease activity.We showed that there was no inhibition of DNA2 nuclease by FANCI alone (Supplementary Figure S1I), and that addition FANCI together with FANCD2 did not further inhibit the DNA2 nuclease activity (Supplementary Figure S1J).
Since FANCD2 also pre v ents nascent strand degradation by MRE11 (7,28), we next addressed if FANCD2 inhibits MRE11 (Figure 1 F).We found that MRE11 activity on duplex DNA was inhibited by FANCD2, also consistent with a pr otein / pr otein interaction ( 74 ).We also tested the effect of FANCD2 on EXO1 (Figure 1 G).EXO1 is an exonuclease that degrades DNA from the end of a r egr essed arm, and we use double-strand DNA substrate to mimic its optimum in vitro substrate.We found that FANCD2 did not inhibit EXO1 in vitro , which does not seem consistent with the  S1).Increasing amounts of FANCD2-His were preincubated in DNA2 nuclease reaction (8 l) mix for 30 min a t 4 • C .DNA substra te (87 FORK, 15 nM, 1 l) was added, and the reaction was incubated for 30 min a t 37 observation in cells (Figure 1 C).To reconcile this we suggest that there is direct and specific inhibition of MRE11 and DNA2 by FANCD2, but that inhibition of EXO1 in vivo may be indirect.

How does FANCD2 inhibit DNA2 nuclease?
Since FANCD2 binds to DNA ( 75 ) as well as binding to DNA2, we next asked whether inhibition was mediated by a FANCD2 protein / DNA interaction and / or DNA2 / FANCD2 pr otein / pr otein interaction.To investigate w hether DN A binding by FANCD2 was involved in the inhibition of DNA2, we used a FANCD2-F1 + F3Mut, which is defecti v e, though not completely b locked, in DNA binding (Supplementary Figure S2A and B) ( 75 ).Like WT F ANCD2, F ANCD2-F1 + F3Mut showed no nuclease activity itself (Figure 2 A, controls, left) and strongly inhibited DNA2 nuclease both on the fork structure (Figure 2 A right) and on the re v ersed for k structur e (Figur e 2 B), although it only retains 10% of the ssDNA binding activity compared to WT FANCD2 (Supplementary Figure S2B), consistent with the previous characterization of the FANCD2-F1 + F3Mut protein ( 75 ).This suggests that inhibition does not occur by simply blocking the DNA substrate or competing with DNA2 for the substrate and suggests that direct pr otein / pr otein interaction may account for DNA2 inhibition.
To directly test this, we purified His-tagged human FANCD2 from E. coli and FLAG-tagged human DNA2 pr otein fr om 293T cells as described in Materials and Methods ( 64 ) (Supplementary Figure S2C).Immunoprecipitation experiments show that purified DNA2 and FANCD2 bind directly and strongly to each other (Figure 2 C,  D).Coimmunoprecipitation experiments show that the FANCD2 / DNA2 interaction is independent of DNA (Figure 2 E), further supporting that in vivo interaction may also be direct.These data are consistent with our previous finding that DNA2 and FANCD2 reciprocally co-IP in extracts of CPT-treated cells and that the interaction is independent of DNA ( 49 ).FANCD2 was pr epar ed in E. coli for the in vitro experiments and appears as a single band of non-ubiquitinylated FANCD2 on a gel, suggesting ubiquitin is not necessary for the interaction between FANCD2 and DNA2.
Using site-directed mutagenesis, we identified a region on DNA2 in the N terminus spanning amino acid (a.a.) 227 to 348 that se v er ely r educes coimmunopr ecipitation with full-length FANCD2 (Figure 2 F and Materials and Methods).This region includes the canonical DEK nuclease family acti v e site motifs ( 76 , 77 ), suggesting how the interaction might interfere with nuclease function.Thus far, point mutations introduced into the acti v e site region inactivate the catalytic activity of DNA2, so they have not been useful for further correlating the site in DNA2 required for nuclease inhibition by FANCD2.
In complementary experiments, we used a previously described complete set of contiguous fragments of FANCD2 ( 75 ) to determine the region of FANCD2 that interacts with DNA2 and that inhibits DNA2, a functional assa y f or 'interaction'.As shown in Figure 2 G, fragment F1, a.a.1-588, and fragment F4, a.a.1178-1451, both coimmunoprecipi-tated with DNA2.This may identify an interface between these subunits that interacts with DNA2.Howe v er, fragment F1 was the only sub-fragment that inhibited DNA2 (Figure 2 H).This region contains the FANCD2 ubiquitylation site (a.a.K561), and addition of a ubiquitin coding sequence to the F1 coding sequence at this site (fragment designated ubi) ( 75 ) increased the efficiency of inhibition, suggesting but not proving that ubiquitin may stabilize interaction.We note that this fragment also contains a DNA binding domain ( 75 ).We conclude that a FANCD2 F1 and F4 domain directly binds to the nuclease domain of DNA2; the interaction of F1 with DNA2 suppresses DNA2's nuclease activity.
Since the FANCD2-F1 + F3Mut protein showed residual binding to DNA, howe v er, to further strengthen the conclusion that DNA2 inhibition is through a protein / protein interaction, we investigated whether inhibition by FANCD2 is species-specific.Yeast lacks a FANCD2 ortholog, and we hypothesized that yeast DNA2 would only be inhibited by FANCD2 if inhibition was mediated by occlusion / sequestration of DNA, thus pre v enting binding by DNA2.We observed no inhibition of yeast DNA2 by FANCD2, e v en at great molar excess FANCD2, on either the forked substrate or the reversed fork substrate (Supplementary Figure S2D and S2E).Note that yeast DNA2 protein is more acti v e than human DNA2, as also reported by others ( 78 ), accounting for the concentrations used.The lack of inhibition of yeast DNA2 protein by FANCD2 supports, though it does not prove, that inhibition of hDNA2 by FANCD2 involves a species-specific and ther efor e likely a physiolo gicall y significant pr otein / pr otein interaction.

DNA2 is also inhibited by RAD51 filaments
In the absence of fork protection by BRCA2 or FANCD2, DNA2-dependent degradation of nascent DNA strands can be suppressed by ov er-e xpression of RAD51 or stabilization of RAD51 filaments ( 7 , 53 , 79 ).Furthermore, FANCD2 and RAD51 show epistatic interaction in nascent DNA degradation assays, i.e. ov ere xpression of RAD51 compensates FANCD2 deficiency for the degradation of nascent DNA in cells as determined by DNA fiber tracking ( 7 ) and see also ( 4 , 6 , 7 , 32 , 53 , 80-84 ).To explore the potential molecular interplay between FANCD2 and RAD51 in regulating DNA2-mediated resection, we first looked at whether there is physical interaction between DNA2, RAD51, and FANCD2 in HU-treated cells.We show that RAD51 co-IPs with FLAG-DNA2 and with endogenous FANCD2 (Figure 3 A).Reciprocally, we immunoprecipitated RAD51 and showed that both FANCD2 and endogenous DNA2 coimmunopr ecipitated (Supplementary Figur e S3A).The RAD51 immunoprecipitate in Supplementary Figure S3A, which re v ealed two FANCD2 bands on western blotting.We propose that the slower migrating band may be the ubiquitylated form of FANCD2 while the faster band may represent unmodified FANCD2.We then repeated these experiments after treating the cell extract with Benzonase.The RAD51 / FANCD2 interaction was still observed and is ther efor e not dependent on DNA (Figur e 3 B), in keeping with previous observations ( 58 ).However, since RAD51 was not found in a DNA2 IP after treatment of extracts with  Benzonase (Figure 2 E), we conclude that the interaction of DNA2 with RAD51 is stabilized by DNA binding.
RAD51 filaments have been implicated in regulating DNA2-mediated resection.Cells heterozygous for RAD51 T131P, which fails to form stab le filaments, e xhibit DNA2dependent accumulation of ssDNA upon treatment with MMC ( 53 ).We ther efor e tested for inhibition of FLAG-DNA2 nuclease by recombinant RAD51 protein.Increasing amounts of RAD51 inhibited DNA2 nuclease activity on both fork and flap substrates (Figure 3 C and D).RAD51 also inhibits EXO1 nuclease on an overhang substrate (Figure 3 E) and has previously been shown to inhibit MRE11 ( 81 ).Nuclease inhibition r equir es ATP and Ca 2+ (Figure 3 C and E, Supplementary Figure S3B), which inhibits RAD51 ATP hydrolysis and promotes stable filament formation ( 85 ), indicating that inhibition is mediated by RAD51 filaments and not by RAD51 monomers.As verified in Figure 3 F and Supplementary Figure S3B, RAD51 filaments were formed on both ssDN A and dsDN A in the presence of ATP and are more stable in the presence of Ca 2+ than in its absence.We conclude that inhibition of DNA2 is mediated by RAD51 filaments.

FANCD2 stimulates strand e x change by high concentrations of RAD51
We were struck by the fact that BRCA2 −/ − cells and FANCD2 −/ − show non-epistatic interactions such as synthetic lethality and that ov er-e xpression of FANCD2 suppresses BRCA −/ − phenotypes ( 25 , 56 ).Furthermore, like BRCA2, FANCD2 interacts physically and robustly with RAD51 [Figure 3 and ( 29 , 58 )], and RAD51 has been shown to localize to stalled forks in cells lacking BRCA2 ( 81 ).We hypothesized that FANCD2 might, similarly to BRCA2, stimula te RAD51-media ted strand exchange ( 86 , 87 ).W hile FANCD2 does not enhance RAD51-mediated D-loop assays with resected plasmid substrates ( 31 ), complete strand exchange assays with oligonucleotides were never tested.Both r eactions ar e linked to DNA r ecombination, but mechanistically they are different.In D-loop assays strand invasion into a supercoiled DNA recipient is measured and is thought to r epr esent a sear ch for homology ( 88 , 89 ).Strand exchange assays, in contrast, measure a complete tr ansfer of DNA str ands (see schematic in Figure 4 A).As indica ted, RAD51 ca talyzes the exchange of the labeled strand in the duplex to ssDNA to form the strand exchange product ( 86 , 87 ).High concentrations of RAD51, howe v er, have been shown to be inhibitory in this assay ( 86 , 87 ).To measure strand exchange, RAD51 was incubated, in the presence or absence of FANCD2, with ssDNA (Figure 4 B, pilot experiment, see also replicates in Supplementary Figure S4), with duplex DNA with a 3 ssDNA overhang (Figure 4 C, left panel), or with duplex DNA with a 5 ssDNA overhang (Figure 4 C, right panel) to allow filament formation.Fully duplex DNA containing a 32 P labeled strand complementary to the ssDNA or respecti v e overhang DNA was then added.Stimulation of strand exchange by RAD51 is shown for ssDNA in Figure 4 B, lanes 1-4.Inhibition at high RAD51 le v els is shown in Figure 4 B, lane 5.Such inhibition is proposed to arise once ssDNA is sa tura ted with RAD51, allowing the excess RAD51 to bind to the labeled dsDN A donor, w hich inhibits the exchange ( 86 , 87 ).Supporting the hypothesis that the inhibition by high le v els of RAD51 can be due to the binding of e xcess RAD51 to duplex DNA, w e show ed that the addition of a dI-dC oligonucleotide relie v es inhibition, presumab ly by successfully competing with the labeled duplex donor for excess RAD51 binding in the assays (Figure 4 B, lane 12).We then studied whether FANCD2 stimulated RAD51 at high RAD51 concentrations, as has been shown for BRCA2 ( 86 , 87 ).As shown in Figure 4 B (lanes 6-11) and Figure 4 C, although FANCD2 has no strand exchange activity on its own, FANCD2, indeed, reproducibl y stim ulates strand exchange by high concentrations of RAD51 and does so in a concentr ation dependent manner.Str and exchange involving duplex DNA with a 3 or 5 overhang, more closely resembling a filament on resected DNA, was stimulated more efficiently than with ssDNA, suggesting that stimulation may occur on DNA with ds / ss junctions and may occur at gaps as well as at ssDNA tails (Figure 4 B, C).Se v eral controls that strand exchange was occurring were performed.Re v ersing the or der of addition of substrates, i.e. formation of RAD51 filaments on dsDNA and then addition of ssDNA, did not lead to exchange (Figure 4 D); thus, we are not observing inverse strand exchange ( 90 ).Addition of cold oligonucleotide to the stop reaction does not change the products, supporting that the strand exchange products are not formed due to dena tura tion and rena tura tion in the stop mixture (Figure 4 C, lanes labeled cold oligo) (86).We conclude that FANCD2 stimulates strand exchange at high concentrations of RAD51.

FANCD2 promotes strand e x change activity by enhancing ss-DNA binding of RAD51
We next interrogated the mechanism of FANCD2 stimulation of RAD51.BRCA2 DNA binding is r equir ed for stimulation of strand exchange, and BRCA2 is thought to stimulate strand exchange in several ways: by stabilizing RAD51 filaments through inhibiting RAD51 DNA-dependent AT-Pase, by promoting the handoff of ssDNA from RPA to RAD51, and by nucleating filament formation on ssDNA while inhibiting filament formation on duplex DNA.To determine if FANCD2 DNA binding was r equir ed to stimulate strand exchange, we tested if the FANCD2 DNA binding mutant described above stimulated strand exchange ( 75 ).Although FANCD2-F1 + F3Mut showed an approximately ten-fold reduction in ssDNA binding at 10 nM (Supplementary Figure S2B), FANCD2-F1 + F3Mut protein can still stimulate strand exchange (Figure 5 A), which suggests that the DNA binding activity of FANCD2 is not r equir ed in promoting strand exchange, or that weak binding is sufficient.We next determined if FANCD2 inhibits RAD51 DNA-dependent ATPase.Surprisingly, unlike BRCA2, FANCD2 does not inhibit RAD51 DNAdependent ATPase (Figure 5 B), and thus may not be acting to stabilize RAD51 / ssDNA filaments by blocking the ATPase.
We finally tested if FANCD2 plays a role in targeting RAD51 pr efer entiall y to ssDN A by inhibiting nucleation on dsDNA.We carried out DNA binding experiments using biotin-streptavidin pull-downs (Figure 5 C).We first  l reactions were performed as in panel D using 60 nM RAD51 preincubated with (lanes 3-6) or without (lane 2) increasing concentrations of FANCD2 (FD2).After capture, both proteins were separately probed by western analysis.The histogram shows quantification of the RAD51 western analysis.Assays wer e r epeated two times.( F ) FANCD2 stimulates RAD51 filament formation on 3 overhang DNA in the absence of dsDNA competitor.Reactions were performed as in panel E using 60 nM RAD51 preincubated with (lanes 3-7) or without (lane 2) increasing concentrations of FANCD2.TE buffer in lieu of dsDNA was added with 3 overhang DNA for all samples and incubated for an additional 5 min a t 37˚C .Both proteins were separately probed for western blot anal ysis.Histo gram shows quantification of the RAD51 western blot analysis.Assays wer e r epeated two times.( G ) FANCD2 stimulates recruitment of RAD51 to DNA but BRC repeat double mutant of FANCD2 inhibits recruitment of RAD51 to DNA.Top: map positions of FXXA motifs in BRCA2 and FANCD2.Bottom: The conditions were the same as for panel F. demonstra ted tha t dsDNA inhibits RAD51 binding to a biotin-labeled 3 overhang substrate (Figure 5 D).We then added FANCD2 and found that FANCD2 did not stimula te the associa tion of RAD51 with the 3 overhang substrate in the presence of excess dsDNA (Figure 5 E).This is unlike what has been demonstrated for BRCA2, which has been shown to specifically overcome dsDNA inhibition of RAD51 binding to overhang DNA in a similar assay ( 86 , 87 ), Thus, FANCD2 is not stimulating RAD51 by reducing binding to dsDNA and is more likely stabilizing RAD51 / ssDNA filaments.Supporting this interpretation, FANCD2 alone does stimulate the accumulation of RAD51 / ssDNA complexes in the absence of ds-DNA (Figure 5 F), similar to the previous characterization of RAD51 / F ANCD2 / F ANCI interaction with DNA ( 58 ).BRC repeats have been shown to be required to stimulate strand ex change b y BRCA2.The FANCD2 protein carries two FXXA BRC consensus site motifs at F1127 and F1320, respecti v ely ( 86 , 88 , 89 ).To test if they were required to stimulate RAD51 DNA binding, a FANCD2 F1127A / F1320A m utant DN A was constructed, and the m utant protein expressed and purified.As shown in Figure 5 G, the wild-type FANCD2 protein stimulated RAD51 DNA binding in the biotin pull-down assay but the BRC-mutant FANCD2 protein reproducibly inhibited binding of RAD51 (Figure 5 G, compare lanes 5, 8, and 9).Based on the results in Figures 4 and 5 , we suggest that FANCD2 stimulates strand ex change b y directly pr omoting, either thr ough nucleation, assembly, or filament stabilization, RAD51 / ssDNA filament forma tion, ra ther than by competing with dsDNA.This FANCD2-media ted stabiliza tion does not involve inhibition of RAD51 ATPase and does not r equir e optimum FANCD2 DNA binding activity.Stimulation of strand exchange with these characteristics suggests that the molecular role of FANCD2 in fork protection ma y in volve stimula tion of forma tion of or stabiliza tion of RAD51 filaments, and thus indirect inhibition of DNA2 nuclease, in addition to the direct inhibition of DNA2 shown in Figure 2 .The results further suggest that interaction of FANCD2 with RAD51 protein [Figure 3 and ( 58 )] contributes to strand exchange, and ther efor e that this may contribute to the ability of ele vated le v els of FANCD2 to suppress some BRCA2deficiencies in fork protection ( 56 , 91 ).
FANCD2 has been demonstrated by iPOND to increase four to fiv e fold on nascent DNA in the presence of HU ( 92 ).To reconcile the roles of FANCD2 in vivo , we further verify the association of FANCD2 with replication forks stalled by HU.We labeled nascent DNA strands with the thymidine analog BrdU in the presence of HU under conditions that specifically mark nascent ssDNA (single-stranded DNA) and used immunofluorescence to monitor localization of FANCD2 and ␥ H2AX ( 61 ).We find that ␥ H2AX foci and FANCD2 foci co-localize with Br dU-mar ked nascent ss-DNA foci (Supplementary Figure S5A).We then compared the le v el of associa tion of FANCD2 with replica tion forks in the absence and presence of HU by immunofluorescence (see Materials and Methods).We detected FANCD2 / EdUassociated foci only after HU treatment (Supplementary Figure S5B).To determine how FANCD2 affects RAD51 filament stability, we performed immunofluorescence to look at RAD51 foci number in HU treated U2OS cells.We found that the number of RAD51 foci is dramatically decreased in FANCD2 depleted cells, as well as after RAD51 inhibitor B02 treatment (Supplementary Figure S5C).Altogether, the results show that FANCD2 responds to replication stress and stabilizes RAD51 filaments, as we suggested in the in vitro assays.

FANCD2 may play different roles at different types of DNA damage
As shown in Figure 1 A, we observed substantial overresection in the absence of FANCD2 after HU treatment of cells.This suggests that the damage, likely consisting of helicase / polymerase uncoupling and fork reversal damage subsequent to stalling ( 55 , 93 , 94 ), is pr otected fr om resection by FANCD2.Such a role for FANCD2 in negati v e regulation of resection, howe v er, needs to be reconciled with previous elegant studies showing that FANCD2 is actually r equir ed for r esection for r epair after e xtensi v e ICL-induced stalling, which may induce a different type of damage, including DSBs and or gaps ( 13 , 95 , 96 ).
To support the proposal that FANCD2 can have two opposing effects on resection in response to different types of damage, we carried out time courses of CPT (Camptothecin) treatment in PD20 or FANCD2 depleted cells and respecti v e FANCD2-complemented cells.While low le v els of CPT can simply induce fork slowing or stalling (through topological stress) ( 97 ), high dose CPT rapidly induces DSBs when the replication fork encounters sites of the CPT-induced Top1-DNA cleavage complexes ( 98 , 99 ).We observed that in the absence of FANCD2, there is overresection at the earliest time point (Figure 6 A, lane 2 compared to lane 8).At later times, FANCD2 becomes a positi v e regulator of resection (Figure 6 A, lanes 5 and 6 compared to lanes 11 and 12), howe v er, presumab ly at rapidly accum ulating 'colla psed forks', since FANCD2 has been shown to recruit CtIP to process the stalled fork ( 13 , 95 , 96 ).Cells lacking FANCD2 respond to e xtensi v e cispla tin trea tment similarly as to CPT (Figure 6 B).Neutral COMET assays support a greater abundance of DSBs in CPT treatment than in HU and a more rapid increase in DSBs during CPT treatment than in HU (Figure 6 C, D).We cannot distinguish whether the initial damage is being remodeled during the time course or if different structures arise independently as stalled forks ar e r emodeled in response to damage.Taken together, our results suggest that FANCD2 is required to protect from over-resection after damage on forks transiently stalled by HU, probably largely on re v ersed for ks or gaps.FANCD2, howe v er, ma y pla y a different role during chronic stalling that leads to fork collapse to DSBs or to other types of damage, such as gaps due to Prim-Pol activity, when FANCD2 may actually be r equir ed for resection and r epair (Figur e 6 A, B).This interpr etation is consistent with recent proposals tha t dif fer ent r epair mechanisms may function on HU and CPT damage ( 28 , 33 , 61 ).

DISCUSSION
FANCD2 has been studied for decades and much has been learned about its cellular functions and structure.W ha t is unkno wn, ho we v er, despite e xtensi v e cellular and structural characterization, are the biochemical activities it uses to accomplish and coordinate its di v erse in vivo roles.We and others have shown that FANCD2 is a fork protection factor that pre v ents MRE11 and DNA2 mediated resection a t replica tion f orks upon HU f ork stalling and remodeling into re v ersed for ks.The replication for k protection function of FANCD2 may be distinct from its canonical roles in the Fanconi anemia pathway since different substrates may be inv olved.Ho w FANCD2 protects the stalled fork and what kind of DNA structures it acts upon remain unclear.Here, our study utilizes biochemical assays to uncover three ways that FANCD2 protects the stalled forks or gaps: (i) FANCD2 directly binds to DNA2's nuclease domain to inhibit its nuclease activity; (ii) FANCD2 stabilizes RAD51 filaments to pre v ent non-specific DN A degradation by m ul-tiple nucleases; (iii) apart from regulating nascent strand degradation, FANCD2 stabilizes RAD51 filaments on ss-DN A to stim ulate strand e xchange acti vity, which may parallel BRCA2, explaining the previous finding that FANCD2 compensates for BRCA2's loss.

FANCD2 directly inhibits DNA2 in vitro identifying a noncanonical role for FANCD2 in protecting stalled forks from degradation
Our in vitro nuclease inhibition assay may indicate that FANCD2 directly binds the nuclease domain of DNA2 to pre v ent DNA2 mediated over-resection at stalled forks, independent of FANCD2's roles in ICL repair.Se v eral observations are consistent with the conclusion: (i) FANCI did not efficiently stimulate the FANCD2-mediated inhibition of DNA2, nor did FANCI inhibit DNA2 significantly on its own.Failure to stimulate FANCD2 inhibition might be explained if the FANCD2 that inhibits DNA2 is in a dimeric form, which has also been reported to fail to interact with FANCI ( 24 ).Our finding does not exclude the possibility, though, that FANCD2 / I heterodimers may participate when present.(ii) FANCD2 stably and specifically binds DNA2.Ubiquitylation is not essential for interaction but seems to augment inhibition (Figure 2 ).(iii) FANCD2 does not suppress the nuclease activity of the heterologous yeast DNA2, further suggesting that FANCD2 / DNA2 pr otein / pr otein interaction is important in downregulating the nuclease.We identified two DNA2 interaction domains in FANCD2, the F1 and F4 fragments of FANCD2.Furthermore, we found that FANCD2-F1, but not FANCD2-F4, inhibits DNA2.In a complementary experiment, w e show ed that the FANCD2-interaction domain within DNA2 lies in the N-terminal region comprising its nuclease catalytic site.Thus, we propose that the direct binding of FANCD2-F1 to the nuclease domain of DNA2 may hinder the nuclease acti vity.(i v) FANCD2 inhibits DNA2 e v en when FANCD2 DNA binding is compromised.Recent reports found that FANCD2 is purified as a dimer ( 21 ) and suggested that the dimer is not capable of DNA binding ( 24 , 100 ).Howe v er, our wild-type FANCD2 preparation does bind dsDNA (Supplementary Figure S1D).It was proposed that the DNA binding defect in the FANCD2 dimers might be due to sequestration of the DNA binding domain ( 24 , 100 ).We do not know if our preparation contains monomeric or dimeric FANCD2, as gel filtration experiments were inconclusi v e to date.

F ork pr otection involv es well-contr olled resection by DNA2
We propose here that DNA2 plays essential roles not only in a well-defined, non-canonical Okazaki fragment processing pathway but also in the replication fork protection pa thway.In both pa thways , as a nuclease , it must be precisel y and tightl y r egulated to pr e v ent aberrant processing, as w e show ed here f or f ork protection by FANCD2.Unrestrained resection gi v es rise to DNA breaks, chromosomal rearrangements, and aneuploidy, which are hallmarks and dri v ers of cancer.DNA2 is especially interesting because it appears to be involved in multiple, distinct pathways of protection ( 28 ).
We have previously shown that eliminating the replication checkpoint by deletion of the DNA replication checkpoint mediator RAD9 53BP1 or both RAD9 53BP1 and MRC1 Claspin rescues the inviability of dna2 -defecti v e yeast cells ( 101 ).This strongly supports that DNA2 is essential for resolution of DNA replication stress.We proposed that the checkpoint is activated by replication stress to prolong G2 and allo w repair.Ho we v er, in the absence of DNA2, repair cannot occur, so the checkpoint leads to irre v ersib le cell cycle arrest and cell death.In the absence of the checkpoint, cells, though stressed, can continue to divide ( 101 ).Supporting an essential role for human DNA2 in fork protection, Thangavel et al. reported that DNA2 drives the processing of re v ersed for ks upon stalling and mediates for k restart.Controlled DNA2 resection at re v ersed for ks may mediate repair and thus contribute to the survival of cancer cells that would otherwise be eliminated by apoptosis or senescence ( 37 ).On the other hand, resection is not always beneficial, over-resection at the stalled fork leads to deleterious le v els of ssDNA, activating the checkpoint and leading to genome instability and cell death.In this scenario, RAD51-, BRCA2-and FANCD2-mediated fork protection is requir ed to pr e v ent ov er-resection after for k uncoupling and for k re v ersal.Interestingly, the paradox is that while preserving genome stability to pre v ent tumorigenesis in normal cells, the fork protection mechanism in tumor cells confers stability of stalled fork, leading to increased tumor growth and possibly conferring chemoresistance.In fact, there is e xtensi v e e vidence that ov er-e xpression of DNA2 in cancer cells correlates with poor prognosis ( 102 ).DNA2 can thus serve as a biomarker for patient stratification for studies of new drug targets with respect to efficacy and possible drug resistance.
The re v ersed for k is not the sole substrate of DNA2, howe v er.Uncontrolled resection by DNA2 may also lead to long ssDN A ga ps at forks.Specificall y, on the lagging strand, DNA2 may continue to process the 5 end of nascent Okazaki fragment DNA to create long ssDNA ga ps.ssDN A ga ps can also arise on the leading strand, which initiate from repriming by PRIMPOL ( 103), then extended by DNA2 mediated r esection.Thus, over-r esection by DNA2 in the absence of FANCD2 may be deleterious not only during fork reversal and DSB repair but also at lagging or leading strand gaps.In keeping with this proposal, DN A2 shRN A knockdown leads to lengthened replication tracts in DN A fiber studies, w hich could be due to gapped daughter chromosomes (104)(105)(106)(107)(108).More intriguingl y, though poorl y understood, PARP has been implicated in protecting from ssDNA damage ( 105 , 109 , 110 , 111 ).W hether FANCD2 regula tion of DNA2 also functions in these newl y a pprecia ted pa thways of gap repair at stalled forks remains for further investigation.
Our results suggest a role for FANCD2 in modulating DNA2 resection of any or all of these stalled replication fork intermediates.Our results indicate new directions for experiments in resolving the role of DNA2 in these various processes and their relati v e significance to cell viability and cancer.
Further significance of our findings is based on the recent demonstra tion tha t FANCD2 inhibition of DNA2 resection, surprisingly, plays a significant role in restraining accumulation of cytoplasmic DNA and induction of the cGAS-STING pathway in response to replication fork stalling in vivo ( 112 ).This identifies new roles for FANCD2 and DNA2 in immune secretion and inflammation.This new link between the immune response and DNA repair pathways ( 113 ) lends further significance to understanding the mechanism by which FANCD2 controls DNA2.

FANCD2 also regulates resection through RAD51
Previously, the FANCD2 / I heterodimer was shown to stabilize RAD51 filaments, but we have demonstra ted tha t FANCD2 alone can stabilize RAD51 on ssDNA and that ablation of FXXA motifs found in putati v e BRC motifs within FANCD2 pre v ents RAD51 filament stabilization (Figure 5 ).Recent studies of replication initiation have rev ealed that e v en low le v els of stabilization can have critical regulatory outcomes in the cell ( 114 ).This uncovers a mechanism by which RAD51 ov ere xpression, in addition to FANCD2, may suppress the DNA2-dependent component of nascent DNA degradation in BRCA2 and FANCD2 deficient cells ( 6 , 7 ).Like FANCD2, w e show ed that RAD51 filaments also inhibit DNA2 nuclease, and this r equir es stable RAD51 / ssDNA filaments (Figure 3 ).We suggest that RAD51 filaments may more generally retard degradation by multiple nucleases, since we found that RAD51 filaments also pr otect fr om EXO1-mediated nascent DNA degradation, and others have demonstrated in vitro inhibition of MRE11 nuclease by RAD51 ( 4 , 81 , 115-116 ).
We reasoned that the nuclease inhibition by RAD51 filaments was not the sole role of FANCD2 at stalled for ks, gi v en comple x roles that RAD51 plays, including in promoting fork reversal.The fact that FANCD2 and BRCA2 are synthetically lethal and that FANCD2 ov ere xpr ession suppr esses the r eplication fork protection defect of BRCA −/ − cells, suggested that FANCD2 might have a parallel function to BRCA2 ( 25 , 56 ).Ther efor e, we tested whether FANCD2 had activities similar to BRCA2 protein, a known RAD51 mediator.We found that FANCD2 stimulates RAD51 strand e xchange, ov ercoming inhibition of strand ex change b y high le v els of RAD51, and does so with similar stoichiometry to that reported for BRCA2 ( 86 , 87 ).BRCA2 stimulates strand exchange on one level by acting as a mediator in the exchange of RPA for RAD51 on resected overhangs.Second, BRCA2 promotes RAD51 ssDNA filament formation through inhibition of non-producti v e or inhibitory binding of RAD51 to dsDNA, presumably by competition between BRCA2 and RAD51 for DNA binding ( 86 , 87 ).BRCA2 uses BRC repeats 1-4 to inhibit RAD51 ATPase and thus to stabilize filaments, but BRCA2 also uses BRC repeat 6-8 to promote nucleation of RAD51 on ssDNA and thus stimulate strand exchange ( 88 , 89 ).We did not find that FANCD2 inhibited RAD51 ATPase, nor did it overcome the inhibition of RAD51 binding to ssDNA by dsDNA (biotin pull-down assays).Thus, FANCD2 is acting differently from BRCA2 or MMS22L / TONSL ( 117 ), which also stimulates strand exchange.Furthermore, the DNAbinding-defecti v e FANCD2-F1 + F3Mut protein, was e v en more efficient than WT FANCD2 in enhancing strand exchange, also differing from BRCA2, which needs to bind to DN A to stim ula te RAD51.The (a t least partial) independence of FANCD2 DNA binding in strand exchange suggests that stimulation of strand exchange by FANCD2 involves a significant FANCD2 / RAD51 pr otein / pr otein interaction and that this in turn helps stabilize RAD51 filaments.Interestingly, the F ANCD2 / F ANCI complex stabilizes RAD51 filaments, and FANCI DNA binding motifs are necessary, but the FANCD2 DNA binding motifs are not necessary ( 58 ), in accordance with our observa tion tha t FANCD2 DNA binding mutants stimulate strand e xchange e v en in the absence of FANCI.Stimulation of strand exchange by FANCD2 most likely involves stabilization in some way of the RAD51 filament, perhaps by pre v enting end r elease, as suggested pr eviously for the F ANCD2 / F ANCI complex ( 58 ) or by altering the filament structure in multiple ways, as demonstrated for RAD51 paralogs ( 98 , 118-124 ).This proposal is supported by the fact that FANCD2 carrying mutations affecting two FXXA consensus BRC motifs, such as are found in the BRC1 and BRC2 motifs of BRCA2, fail to stimulate RAD51 binding.
One likely mechanism of FANCD2 stimulation of RAD51 filament formation is to provide a chaperone for RAD51 filament assembl y.FANCD2, namel y, has been shown to act as a histone chaperone in nucleosome assembly.The histone chaperone function of FANCD2 is stimulated by histone H3K4 methylation mediated by BOD1L and SETD1A.Strikingly, in BOD1L or SETD1A depleted cells or in cells with inactivated FANCD2 chaperone function, RAD51 filaments are destabilized, and stalled forks are e xcessi v el y degraded by DN A2 ( 52 ).Our r esults ar e consistent with the suggestion that FANCD2 might assist stab le comple x formation between RAD51 and DNA, i.e. filament nuclea tion, elonga tion or stabiliza tion, in addition to promoting histone association and appropriate chromatin structure to protect stalled forks.A similar chaperone-like function has also been proposed for the RAD51 paralogs (122)(123)(124)(125), and many histone chaperones have been shown to chaperone additional proteins into assemblies,

Possible steps in r eplication f ork protection mediated by FANCD2-stimulated strand RAD51 e x change activity
How does strand exchange support fork protection and r estart of forks?Ther e ar e se v eral possibilities ( 126 ).(i) Stabilizing RAD51 filaments on the ssDNA arising at uncoupled replication forks could stimulate strand exchange promoting fork reversal processes ( 55 , 98 , 117 , 127 ), The re v ersed for ks might be the substrates for FANCD2controlled DNA2-mediated processing, leading to replication fork restart ( 32 ).Two studies did show that cells deficient in FANCD2 failed to restrain synthesis in the presence of HU or a phidicolin, possibl y by failing in fork reversal ( 8 , 13 ).(ii) RAD51 could promote reannealing of the ssDN A immediatel y behind the stalled helicase, essentially zipping up the unwound DNA, helping to promote for k re v ersal.For k re v ersal could slow replication forks for repair; (iii) FANCD2 stimulation of RAD51 strand exchange may be important for post-r eplication r epair ( 3 , 71 ).While it was originally thought that the structure of the forks at these extensively damaged chromosomes might be DSBs, recent evidence suggests that they may also be unrepaired gaps after repriming, by Prim-Pol, at leading strand blocks or by pol ␣−primase at blocks on the lagging strand, especially ( 103 , 108 ).FANCD2 might stimulate RAD51mediated strand exchange events in post-r eplication r epair and template switch at these sites, especially if BRCA2 is defecti v e ( 128 ).(i v) In BRCA2 −/ − cells e xcessi v e resection crea tes a substra te tha t r equir es MUS81 for restart ( 68 ).The MUS81-cleaved intermediate, a one-ended DSB, may then be r epair ed by template switch post-replication repair, which r equir es RAD51, and / or br eak-induced r eplication (BIR), which r equir es pol ␦, or by translesion synthesis.FANCD2 could participate in such RAD51-mediated failsafe mechanisms of completing replication (Figure 7 ).(v) Yet another repair mechanism at stressed replication forks is dependent on post-r eplication r epair of gaps introduced by r epriming downstr eam of lesions on the leading strand, instead of DSBs and could also compensate for or substitute for over-resected re v ersed for ks.In the presence of FANCD2 we see increasing phospho-RPA during stalling which implies activation of ATR, which is presumably necessary for repair.ATR has been shown by fiber tracking to activate PRIMPOL and promote downstream repriming leading to gaps in nascent DNA (Quinet and Vindigni, 2019).In FANCD2-depleted cells, this pathway cannot be activa ted ef fectively because RPA-p does not accumulate (Figure 1 A, 8 h time point), blunting the checkpoint and recovery and leading to genome instability.FANCD2 has also been reported to counteract NHEJ at IR-induced DSBs, and FANCD2 deficient cells show increased toxic NHEJ, decr eased r esection, and decr eased r ecombinational r epair ( 129 ).FANCD2 has also been implicated in counteracting Ku70 inhibition of repair ( 130 ).(vi) The strand exchange stimulation function of FANCD2 may be its important contribution to the late stages of ICL repair by the FA pathway, which involves repair of DSBs ( 131 ), rather than or in addition to its role at re v ersed for ks or gaps.Thus, RAD51 strand exchange stimulation might also explain the minor defect in DSB repair pathways in the absence of FANCD2 r eported pr e viously ( 132 ).The role of FANCD2 in for k protection may be compensated for by either BRCA2, for HR-like fork protection, or in a by-pass mechanism by pol theta / CtIP recruitment for alt-EJ ( 25 ).Future studies will be aimed at understanding whether the RAD51 filament stabilization is more important for fork reversal or protection from over-resection or for fork restoration.

FANCD2 may act differently at different types of DNA damage
We observed that while FANCD2 inhibits resection at HUstalled forks, FANCD2 becomes r equir ed for r epair at mor e se v ere types of damage that block forks, such as CPT and cisplatin (Figures 1 and 6 ).This observation is reminiscent of e v ents at forks stalled by 24 h at ICLs (inter-strand crosslink), where FANCD2 recruits CtIP, which augments resection by DNA2 / BLM, channeling repair of obligate DSB intermediates in the ICL repair pathway into HR instead of toxic NHEJ, and / or recruits pol theta ( 13 , 95 , 96 ).
To recapitulate, our model (Figure 7 ) taking cumulati v e data into account, we suggest that DNA2 is r equir ed for resection of transiently stalled re v ersed for ks to promote restoration of acti v e for ks without colla pse to DSBs or ga ps.FANCD2 is r equir ed to keep DNA2 / MRE11 mediated resection in a range consistent with preserving genome stability and restoring forks, as demonstrated previously by increased chromosomal aberrations in its absence ( 7 ).Howe v er, e v en if FANCD2 is present, e xtensi v e or prolonged stalling, or strong fork blocking lesions, such as CPT or cisplatin, lead to the emergence of genome destabilizing structur es that r equir e an additional r epair mechanism(s) involving resection.FANCD2 then becomes essential for resection.

Synthetic viability and fanconi anemia
This study began as a discovery of synthetic viability between DNA2 and FANCD2 defects.We and others have shown that additional FANC alleles show over-resection that is reduced by depletion of DNA2, suggesting that additional proteins are required to fully reconstitute FANCD2r egulated r esection at stalled forks ( 33 ).Inter estingly, different mechanisms of fork protection are seen with different types of damage ( 33 ) and with different fork protection factors ( 28 ).Many additional genes show synthetic viability with FA complementation groups FA-A, FA-C, FA-I, FA-D2, FA, such as BLM helicase ( 133 ).One of the major functions of BLM is to complex with DNA2 in double-strand end resection ( 134 , 135 ).BLM is also deleterious for fork protection and replication fork restart in FANCD2 deficient cells, and depletion of BLM rescues restart in that case ( 7 ).It is possible that BLM is also r equir ed at re v ersed for ks to promote degradation by DNA2 (and / or EXO1, which is stimulated BLM).These results have significant impact on our goal of using inhibition of DNA2 to increase the therapeutic index for treating Fanconi anemia patients with v arious cancers b y pr otecting normal, non-cancer ous cells from chemotherapeutics that stall replication forks and by limiting inflammation.

Figure 1 .
Figure 1.FANCD2 pre v ents DNA2 and MRE11 mediated nascent DNA degradation.( A ) FANCD2 pre v ents resection in response to HU. PD20 cells and PD20:FANCD2 complemented cells wer e tr eated with 4 mM hydroxyurea (HU) for 0-8 hours.HU was added and samples were taken at 0, 2, 4, 6 and 8 h.Nuclear extract was pr epar ed, as illustrated by the H3 loading control, and analyzed for resection by western blot.'RPA2-p long' refers to long exposure time monitoring RPA2 T21 phosphorylation; 'RPA2-p short' is a short exposure time monitoring RPA2 T21 phosphorylation.The RPA2-p and histone H3 blot intensity were measured with ImageJ.The RPA2-p value was normalized to histone H3.The graph value is the average of two experiments; error bar is standard deviation.( B , C ) Over-resection of nascent DNA in HU-treated FANCD2 deficient cells is reduced by depletion of nucleases.U2OS cells were co-transfected with 12 nM siRNA for each indicated gene.72 h post-transfection, cells were pulsed by CldU and IdU, followed by 4 mM HU for 4 h, as indicated on the top of the panel.The cells were harvested and analyzed by a fiber spreading assay.The IdU and CldU track lengths were measured, and the ratio was graphed ( ≥150 fibers were analyzed).One-way ANOVA test was performed, n = 2.Western blots show the level of knockdown.( D ) DNA2 nuclease activity is inhibited by FANCD2-His on a fork structure with a 30 NT 5 ssDNA overhang and a 13 NT 3 overhang (87 FORK, Supplementary TableS1).Increasing amounts of FANCD2-His were preincubated in DNA2 nuclease reaction (8 l) mix for 30 min a t 4 • C .DNA substra te (87 FORK, 15 nM, 1 l) was added, and the reaction was incubated for 30 min a t 37 • C .All reactions contained equal amounts of FANCD2 diluent.Lane 1, DNA alone; lane 2, DNA2 alone; lanes 3-8, DNA2 (1 nM) plus 13, 25 and 50 nM FANCD2 in duplicate, respecti v ely.( E ) Inhibition of DNA2 by FANCD2 on re v ersed for ks.Reactions were as in panel C. Lane 1 and 7, DNA alone; lane 2 and 8, DNA2 alone; lanes 3-5 and lanes 9-11, DNA2 (1 nM) plus 5, 9 and 18 nM FANCD2, respecti v ely.Verification of the re v ersed for k structur es is pr esented in Supplementary Figur e S1F. ( F ) FANCD2-His inhibits MRE11 on a dsDNA substrate.FANCD2-His was preincubated in MRE11 nuclease reaction mix on ice for 30 min.Then substrate was added to activate the reaction at 37 • C for 30 min.dsDNA substrate is shown at the top of the panel.Quantification shows the degradation le v els.( G ) FANCD2 does not inhibit EXO1 on an overhang substrate.FANCD2 was added where indicated (3, 6, 11, 23, 46 and 91 nM).FANCD2-His was preincubated in EXO1 nuclease reaction mix on ice for 30 min.Then substrate was added to activate the reaction at 37 • C for 30 min.EXO1 was 0.77 nM.
Figure 1.FANCD2 pre v ents DNA2 and MRE11 mediated nascent DNA degradation.( A ) FANCD2 pre v ents resection in response to HU. PD20 cells and PD20:FANCD2 complemented cells wer e tr eated with 4 mM hydroxyurea (HU) for 0-8 hours.HU was added and samples were taken at 0, 2, 4, 6 and 8 h.Nuclear extract was pr epar ed, as illustrated by the H3 loading control, and analyzed for resection by western blot.'RPA2-p long' refers to long exposure time monitoring RPA2 T21 phosphorylation; 'RPA2-p short' is a short exposure time monitoring RPA2 T21 phosphorylation.The RPA2-p and histone H3 blot intensity were measured with ImageJ.The RPA2-p value was normalized to histone H3.The graph value is the average of two experiments; error bar is standard deviation.( B , C ) Over-resection of nascent DNA in HU-treated FANCD2 deficient cells is reduced by depletion of nucleases.U2OS cells were co-transfected with 12 nM siRNA for each indicated gene.72 h post-transfection, cells were pulsed by CldU and IdU, followed by 4 mM HU for 4 h, as indicated on the top of the panel.The cells were harvested and analyzed by a fiber spreading assay.The IdU and CldU track lengths were measured, and the ratio was graphed ( ≥150 fibers were analyzed).One-way ANOVA test was performed, n = 2.Western blots show the level of knockdown.( D ) DNA2 nuclease activity is inhibited by FANCD2-His on a fork structure with a 30 NT 5 ssDNA overhang and a 13 NT 3 overhang (87 FORK, Supplementary TableS1).Increasing amounts of FANCD2-His were preincubated in DNA2 nuclease reaction (8 l) mix for 30 min a t 4 • C .DNA substra te (87 FORK, 15 nM, 1 l) was added, and the reaction was incubated for 30 min a t 37 • C .All reactions contained equal amounts of FANCD2 diluent.Lane 1, DNA alone; lane 2, DNA2 alone; lanes 3-8, DNA2 (1 nM) plus 13, 25 and 50 nM FANCD2 in duplicate, respecti v ely.( E ) Inhibition of DNA2 by FANCD2 on re v ersed for ks.Reactions were as in panel C. Lane 1 and 7, DNA alone; lane 2 and 8, DNA2 alone; lanes 3-5 and lanes 9-11, DNA2 (1 nM) plus 5, 9 and 18 nM FANCD2, respecti v ely.Verification of the re v ersed for k structur es is pr esented in Supplementary Figur e S1F. ( F ) FANCD2-His inhibits MRE11 on a dsDNA substrate.FANCD2-His was preincubated in MRE11 nuclease reaction mix on ice for 30 min.Then substrate was added to activate the reaction at 37 • C for 30 min.dsDNA substrate is shown at the top of the panel.Quantification shows the degradation le v els.( G ) FANCD2 does not inhibit EXO1 on an overhang substrate.FANCD2 was added where indicated (3, 6, 11, 23, 46 and 91 nM).FANCD2-His was preincubated in EXO1 nuclease reaction mix on ice for 30 min.Then substrate was added to activate the reaction at 37 • C for 30 min.EXO1 was 0.77 nM.

Figur e 2 .
Figur e 2. FANCD2 / DN A interaction and FANCD2 / DN A2 contribute to DN A2 nuclease inhibition.( A ) Inhibition of hDN A2 by the hFANCD2-F1 + F3Mut DNA binding mutant protein -87 FORK substrate.Conditions were as in Figure 1 except that mutant FANCD2 was used.Lanes 1-5, FANCD2 wild type (WT) and F1 + F3Mut alone, lane 6, DNA2 (1 nM) alone; lanes 7-10, DNA2 (1 nM) plus 9, 18 35, and 70 nM FANCD2-F1 + F3Mut protein, respecti v ely.The dashed line shows that lanes 6-10 were run on a different gel.( B ) Inhibition of hDNA2 by the hFANCD2-F1 + F3Mut DNA binding mutant protein -re v ersed for k substra te.Quantifica tion a t the bottom shows the degrada tion le v els.( C ) His-ta gged FANCD2 and FLAG-ta gged DNA2 interact in vitro .FLAG-DNA2 was ov er-e xpressed in 293T cells and purified by binding to M2 FLAG beads.The beads were washed with lysis buffer and then incubated with purified FANCD2 protein (1 g / ml) at 4 • C for 1 h.Beads were washed with PBS 5 times and then in 2XSDS loading buffer follow ed by w estern blotting.Empty M2 beads incubated with FANCD2 served as negati v e control.( D ) FLAG-DN A2 is imm uno-precipitated by Histagged FANCD2 in vitro . 1 g / ml DNA2 protein was incubated with FANCD2-His bound to Ni-NTA beads at 4 • C for 1 hour.Beads were washed 5 times with PBS, and then boiled in 2XSDS loading buffer for western blotting.Empty Ni-NTA beads were incubated with FLAG-DNA2 as negati v e control.( E ) DNA2 and FANCD2 interact in the absence of DNA.FLAG-DNA2 was over expressed (o.e.) in wild type U2OS cells, and nuclear extract was pr epar ed; Benzonase was added to remove DNA.FLAG-DNA2 was pulled down with M2-beads.The beads were washed with nuclear prepara tion buf fer and eluted with FLAG peptide.The elution was pr epar ed for immunoblot.See Materials and Methods.( F ) FANCD2 interacts with the N terminal domain of DN A2. (Top) Ma p of DN A2 domains and truncations of DN A2. (Bottom) Coimm unoprecipitation of the full length (FL) DN A2 and truncated DN A2proteins using FANCD2 antibody.Cell lysates with ov ere xpressed FLAG-DNA2 protein were supplemented with 2nM FLAG-FANCD2.Pull-down was performed using FANCD2 antibody.Products were separated using SDS-PAGE and imaged by western blot analysis using 3XFLAG antibody, re v ealing both FANCD2, as indicated, and DNA2 full length and deletion proteins.( G ) Mapping of the FANCD2 inhibitory domain-An N terminal fragment F1 and C terminal fragment F4 of FANCD2 bind to DNA2. ( H ) The N terminal, DNA2-interacting domain of FANCD2 inhibits DNA2 nuclease.DNA2 nuclease assays were performed as in Figure1using the FANCD2 fragments indicated( 75 ).Assays were performed in duplicate using 0.2 nM DNA2 and 30 nM FANCD2 and were repeated four times.Ubi (lanes 6 and 7) indicates addition of ubiquitin to fragment F1( 75 ).

Figure 3 .
Figure 3. RAD51 filaments pr otect DNA fr om DNA2 mediated degradation.( A ) Co-immunoprecipitation of FANCD2 and RAD51 with DNA2 (FLAG-DNA2 pull-down).FLAG-tagged DNA2 or empty vector was transfected into A549 cells, 24 hours later cells wer e tr eated with 2 mM HU for 3 h and then harvested.Cells were lysed and pull down carried out with FLAG M2 beads; the beads were washed with lysis buffer then boiled with SDS loading buffer ; imm unopr ecipitants wer e analyzed by western blotting of a 12% acrylamide gel with the indicated antibodies.( B ) RAD51 and FANCD2 interact in the absence of DNA.U2OS cell nuclear extract was pr epar ed as described in Materials and Methods.Extracts were treated with Benzonase.10 l RAD51 antibody or IgG control added to 500 g nuclear extract and incubated overnight at 4 • C. 10ul Protein A agarose beads were added and then incubated for 1 hour at room temperature.The beads were washed with nuclease buffer and boiled with SDS sample buffer for western blot.( C ) RAD51 filaments inhibit DNA2 on the forked substrate (87 FORK).Increasing amounts of RAD51 were preincubated with 4 nM 87 FORK substrate prior to the addition of 5nM DNA2.Controls show DNA2 activity in the presence of 100 nM RAD51 and in the absence of ATP (lane 8) or Ca 2+ (lane 9).( D ) RAD51 filaments inhibit DNA2 on a 5 or 3 flap.The indicated amounts of RAD51 were incubated with 4 nM of the respecti v e flap substrate prior to the addition of 5nM DNA2.( E ) RAD51 inhibits EXO1 nuclease on a 3 overhang substrate.( F ) Ca 2+ enhances DNA binding activity of RAD51. 1 nM ssDN A or dsDN A was incubated in a 10 l reaction mixture containing 200 nM RAD51, 25 mM TrisOAc (pH 7.5), 1 mM MgCl 2 , 2 mM CaCl 2 and 2 mM ATP (except where CaCl 2 was omitted), 1 mM DTT, 0.1 BSA mg / ml, as indicated.The reaction was incubated for 10 min at 37 • C(53) and samples wer e mix ed with 1 l loading buffer (2.5%Ficoll-400, 10 mM Tris-HCl, pH 7.5 and 0.0025% xylene cyanol).Products were analyzed on a 5% nati v e gel (29:1 acrylamide / bisacrylamide in TAE), constant voltage, 60V, in the cold room for 1h (lanes 9-14) or 2h (lanes 1-8) followed by phosphor imaging.

Figure 4 .
Figure 4. FANCD2 stimulates RAD51-mediated strand exchange.( A ) Schematic of strand exchange assay: Single-stranded or 3 or 5 overhang DNA is incubated in the presence of RAD51 to form filaments.The filaments are then incubated with a duplex DNA with a labeled strand complementary to the filament.Product formation, r epr esenting complete strand exchange, is monitor ed using a nati v e acrylamide gel.( B ) FANCD2 stimulates strand exchange on ssDNA by high le v els of RAD51.Quantification is shown below each gel.Lane labeled B contains DNA markers for each relevant DNA species as indicated in the schematic on the left and was pr epar ed by annealing oligonucleotide EXTJYM925, JYM925, and 5 labeled JYM945; the lane labeled P is the marker for the position of the exchanged strand product (EXTJYM925 and 5 labeled JYM945).Lanes 1-5: 4 nM ssDNA (100 nt, oligonucleotide EXTJYM925) was incubated with indicated amounts of RAD51 for 5 min at 37˚C and the 5 labeled dsDNA (60mer, JYM925 / JYM945 oligonucleotides) (final concentration 4 nM) was added and incubation continued for an additional 30 min at 37˚C for strand exchange.Lanes 6 and 7, as in lanes 1-5 with indica ted concentra tions of FANCD2 in the absence of RAD51.Lanes 8-11: RAD51 plus FANCD2 at the indica ted concentra tions present during both the 5 preincubation with ssDNA and after addition of dsDNA.Lane 12: 10 nM of dI-dC competitor present during preincubation of RAD51 and ssDNA.Histogram shows quantitation.( C ) FANCD2 stimulates strand exchange on 3 and 5 overhang DNA by high le v els of RAD51.Reactions performed as in panel B; howe v er, 4 nM 3 ov erhang DNA (162 nt RJ-167 annealed to 42 nt RJ-PHIX-42-1) (Left) or 4 nM 5 overhang DNA (162 nt RJ-167 annealed to 42 nt RJ-PHIX-42-2) (right) as indicated, were incubated in the presence of indicated amounts of RAD51 for 5 min at 37˚C to form filaments. 5 labeled dsDNA (40mer, RJ-Oligo1 / RJ-Oligo2 in the case of 3 overhang DNA or RJ-Oligo4 / RJ-Oligo3 in the case of 5 overhang DNA) (final concentration 4 nM) was added and incubation continued for an additional 30 min at 37˚C for strand exchange.Lane 1, no protein; lanes 2-3: RAD51 alone; lanes 4-7: RAD51 plus FANCD2 at the indicated concentrations present during both the 5 preincubation with 3 or 5 overhang DNA and after addition of dsDNA.Lane 8: 10-fold excess of unlabeled heterologous ssDNA (40 nt, oligonucleotide RJ-Oligo2) complementary to labeled strand of dsDNA was added to the stop solution to rule out that the product observed was due to denaturation and annealing during the deproteinization / termination step.(% Product r epr esents the value with unstimulated exchange subtracted.)The graph shows quantification for both assays.The assays wer e r epeated twice.( D ) Inverse strand exchange assay.In lanes 1-4, the exchange assay was conducted as in the legend to A-C.In lanes 5-7 the double-stranded DNA was preincubated with RAD51 and then ssDNA was added.

Figure 5 .
Figure 5. FANCD2 stimulates strand exchange activity by enhancing ssDNA binding of RAD51.( A ) FANCD2-F1 + F3Mut stimulate strand exchange.Left panel: Lanes 1-6, titration of WT FANCD2 at 100 nM RAD51 and 2 nM of both ssDNA and dsDNA.Assays were performed with oligonucleotides as in panel B, Figure 4 .Right panel: Lanes 1-6: Titration of F1 + F3 FANCD2 mutant stimulation of RAD51 as for FANCD2 WT.Right panel: Lanes 7-10: Indicated amounts of FANCD2-F1 + F3 mutant incubated in the absence of RAD51 during both the preincubation with ssDNA and after the addition of dsDNA.Graph shows quantification for FAND2 WT and FANCD2-F1 + F3 stimulation assays.( B ) Wild-type FANCD2 and FANCD2-F1 + F3 mutant do not inhibit the DNA-dependent ATPase activity of RAD51.FANCD2 WT and F1 + F3 mutant concentration is 6, 11, 23, 46, 91 nM.300 nM RAD51 and 900 nM ssDNA added to the reaction.( C ) Schematic of biotinylated DNA pull-down assay, B: B-biotin; S: S-streptavidin.( D ) Assembly of RAD51 onto biotinylated 3 overhang DNA is suppressed by heterolo gous dsDN A competitor.RAD51 and FANCD2 proteins at indicated concentrations were incubated for 15 min at 37˚C followed by the addition of 3 overhang DNA (162 nt RJ-167 annealed to 42 nt 3 Bio-RJ-PHIX-42-1) and competitor heterologous dsDNA (90mer, Oligo #90 / Oligo #60 oligonucleotides) and incubated for an additional 5 min at 37˚C.Where DNA was omitted, TE buffer was used and similarly, respecti v e protein stora ge b uffers wer e used wher e proteins wer e omitted.20 l r eactions wer e performed as in panel F using 60 nM RAD51 either in the absence (lane 2) or presence (lanes 3-5) of excess competitor heterolo gous dsDN A (90mer, Oligo#90 / Oligo#60 oligonucleotides).After capture of protein / DNA complexes, western blotting was performed.Histogram shows quantification.Assays were repeated two times.( E ) FANCD2 does not rescue RAD51 filament formation on 3 overhang DNA in the presence of heterologous dsDNA competitor (40 nM).20l reactions were performed as in panel D using 60 nM RAD51 preincubated with (lanes 3-6) or without (lane 2) increasing concentrations of FANCD2 (FD2).After capture, both proteins were separately probed by western analysis.The histogram shows quantification of the RAD51 western analysis.Assays wer e r epeated two times.( F ) FANCD2 stimulates RAD51 filament formation on 3 overhang DNA in the absence of dsDNA competitor.Reactions were performed as in panel E using 60 nM RAD51 preincubated with (lanes 3-7) or without (lane 2) increasing concentrations of FANCD2.TE buffer in lieu of dsDNA was added with 3 overhang DNA for all samples and incubated for an additional 5 min a t 37˚C .Both proteins were separately probed for western blot anal ysis.Histo gram shows quantification of the RAD51 western blot analysis.Assays wer e r epeated two times.( G ) FANCD2 stimulates recruitment of RAD51 to DNA but BRC repeat double mutant of FANCD2 inhibits recruitment of RAD51 to DNA.Top: map positions of FXXA motifs in BRCA2 and FANCD2.Bottom: The conditions were the same as for panel F.

Figure 6 .
Figure 6.FANCD2 acts differently on different types of DNA damage.( A ) FANCD2 is r equir ed for r esection in CPT tr eated cells as r e v ealed by time course of treatment of PD20 FANCD2 −/ − cells with CPT.Assays for resection are as in Figure 1 A. PD 20 FANCD2 −/ − cell and PD20 complemented with FANCD2 (PD20:FD2) were treated with 2 M CPT for indicated times.Nuclear extract was pr epar ed and used for the western blots for proteins and protein modifications as indicated in the figure.( B ) FANCD2 is required for resection in cisplatin treated cells after prolonged exposure.A549 cells were transfected with 40 nM FANCD2 siRNA or control scrambled siRNA (siNC).After 48 h, 2 M CPT or 10 M cisplatin was added as indicated and cells incubated for 16h.Samples were taken and whole cell lysates (GAPDH control) were prepared for western blot.RPA2-p and ␥ H2AX were monitored as shown.(C and D).Neutral comet assay of DNA damage in HU treated cells compared to CPT-treated cells over time suggests greater number of DSBs in the CPT treated cells.

Figure 7 .
Figure 7. Model for multiple roles of FANCD2 in fork protection studied in this work.At a stalled replication fork upon moderate stress, FANCD2 can protect the r egr essed arm by directly inhibiting DNA2 / MRE11 or stabilizing RAD51 on ssDNA to pre v ent digestion of nascent DNA by various nucleases.Not shown is that FANCD2s strand exchange activity might also aid RAD51 mediated for k re v ersal.In prolonged stress or at CPT or cisplatin induced damage, FANCD2 can recruit CtIP to the broken fork to facilitate resection and HDR.FANCD2 may also promote RAD51 mediated strand exchange reactions to restart forks either together with BRCA2 or by itself.