Cooperative Assembly of a Protein-DNA Filament for Nonhomologous End Joining*

Background: Nonhomologous end-joining proteins can ligate DNA ends with mismatched overhangs. Results: Ku facilitates the cooperative binding of multiple XRCC4/Ligase IV (XL) and XLF molecules to DNA. Conclusion: The protein-DNA complex contains a filament of alternating XL and XLF molecules. Significance: The filament may provide a mechanism for ligating mismatched DNA ends.

complex), which contained one or two Ku molecules and three to five molecules each of XL and XLF. Surprisingly, a 1.1-MDa MEnd ligase-DNA complex had a lower electrophoretic mobility than a 2.2-MDa Ku-XL-DNA complex. These results show that the MEnd ligase-DNA complex forms a filament containing alternating XL and XLF molecules. The filament may facilitate alignment of DNA ends so that mismatched 3Ј overhangs can undergo ligation by one of several Ligase IV molecules arrayed along the DNA.
Conjugation of Tetramethylrhodamine (TAMRA) to ybbRtagged XLF and Purification-Purified ybbR-tagged XLF was labeled with the fluorophore TAMRA using TAMRA-CoA (a generous gift from Dr. Kierstin Schmidt) by Sfp phosphopantetheinyl transferase (23), purified on a 1-ml Hi-Trap heparin column (GE Healthcare) using a linear salt gradient of 150 -1000 mM NaCl in Na-PO 4 , flash-frozen, and stored at Ϫ80°C.
Preparation and Labeling of DNA Probes-To study the formation of protein-DNA complexes, blunt-ended 298-bp and 180-bp DNA probes were prepared by PCR amplification of the bacterial amp R gene and purified with a QIAquick PCR purification kit (Qiagen). For experiments in Figs. 1, 2, and 6, DNA probes were labeled with 33 P using T4 polynucleotide kinase (New England Biolabs) and purified again with nucleotide removal kit (Qiagen). For experiments in Figs. 3-5, the 180-bp DNA probe was fluorescently labeled by PCR amplification using a forward primer with 6-carboxyfluoroscein  conjugated to the 5Ј end (Elim Biopharm).
EMSA-Proteins were incubated with a labeled DNA probe for 30 min at 37°C in 10 l of end-joining buffer, which consists of 25 mM Tris (pH 7.5), 75 mM NaCl, 72.5 mM KCl, 5 mM MgCl 2 , 0.1 mM EDTA, 50 g/ml BSA, 0.05% Triton X-100, 2 mM DTT, 5% glycerol, and 5% (w/v) polyethylene glycol (molecular mass Ͼ8000 kDa). Reaction products were resolved by electrophoresis at 10 V/cm in 0.5ϫ TBE buffer containing 45 mM Tris borate (pH 8.0) and 1 mM EDTA. Initial experiments used precast commercial 5% polyacrylamide gels (Bio-Rad) (Figs. 1 and 2). Subsequent lots of the precast gels failed to stabilize the megadalton protein-DNA complexes. Therefore, we cast our own gels for other experiments and obtained highly reproducible results, using 6.5% polyacrylamide (74:1, w/w, acrylamide to bisacrylamide) (Fig. 4)  the first step, purified proteins were incubated with DNA, and the protein-DNA complexes were resolved by EMSA. In the second step, the protein-DNA complexes were cut from the EMSA gel, and the proteins were analyzed by SDS-PAGE.
For the first step, the DNA probe was labeled with 6-FAM. To limit the size of complexes on the DNA, the DNA probe was 180 bp, which was shorter than the 298-bp DNA probe used in initial EMSA experiments in Figs. 1 and 2. Proteins were incubated with 6-FAM-DNA. To measure the DNA in the protein-DNA complex, protein-DNA complexes were resolved by EMSA in parallel with dilutions of unbound 6-FAM-DNA (1/16, 1/8, 1/4, 1/2, and 1 pmol). The gel was scanned for 6-FAM fluorescence (Typhoon imager; GE Healthcare) to generate plots of background-subtracted fluorescence intensities versus known amounts of DNA. The plots were highly linear (Figs. 3C, 4C, and 5C) with Pearson's correlation coefficients r 2 Ͼ 0.99. The plot of DNA standards was used to estimate the amount of DNA in the protein-DNA complex.
For the second step, the DNA in the EMSA gel was stained with SYBR Gold (Invitrogen), allowing the protein-DNA complexes to be visualized on a UV transilluminator, and cut from the gel with a clean scalpel. However, for analysis of the MEnd ligase-DNA complex, SYBR Gold staining of the DNA led to loss of XLF protein. Therefore, we used the fluorescence image of the gel to cut the MEnd ligase-DNA complex from the gel (Fig. 5).
To denature the proteins in the protein-DNA complex, the gel slice was incubated in lithium dodecyl sulfate-PAGE sample buffer (Invitrogen) and heated for 30 min at 85°C. To measure the protein in the protein-DNA complex, the gel slice together with the incubation sample buffer was loaded into a well, and dilutions of unbound proteins (0.2, 0.4, 0.6, 0.8, and 1 pmol) were loaded into parallel wells. The proteins were resolved by SDS-PAGE in a denaturing 4 -12% gradient polyacrylamide gel (Invitrogen) The gel was stained with Krypton protein stain (Pierce) and scanned to generate plots of background-subtracted fluorescence intensities versus known amounts of protein. Krypton protein stain provided better sensitivity than Coomassie Blue and a wider quantitative range than silver staining. The plots were highly linear (Figs. 3D, 4D, and 5D), with r 2 Ͼ 0.97. The plots of protein standards were used to determine the amount of protein in the gel slice containing protein-DNA complex.
Replicate experiments and titrations provided estimates for the reliability of each component of the two-step gel elution procedure. DNA quantification was highly reliable with an uncertainty of 1%. Protein quantification was reliable to an uncertainty of 10%. Protein to DNA molar ratios, based on three or more replicate experiments with the full two-step gel elution procedure, revealed an overall uncertainty of 20%.
In Vitro End-joining Reactions-End-joining reactions were performed in end-joining buffer as described previously (22). The 800-bp linear DNA substrates with EcoRV-EcoRV or EcoRV-KpnI ends were prepared for end joining as described previously (22). Proteins were preincubated with DNA for 10 min at 25°C. To initiate ligation, MgCl 2 and ATP were added to final concentrations of 5 and 0.1 mM, respectively. Unless otherwise noted in the figure legends, reactions were performed in 20 l with 1 nM DNA, 5 nM Ku, 2 nM XL, and 2.5 nM XLF at 37°C for 30 min, and stopped by adding 2 l of 0.5 M EDTA (pH 8.0). DNA was purified using QIAquick PCR purification kit (Qiagen) and joining efficiencies measured by quantitative PCR as described previously (24).

Ku, XL, and XLF Bind Sequentially to DNA with Cooperative
Addition of XL and XLF-To understand how Ku, XL, and XLF form a complex on DNA, we used an EMSA to resolve protein-DNA complexes from binding reactions (see "Experimental Procedures"). Ku binding to DNA was distributive because it produced a series of protein-DNA complexes containing one, two, and three or more Ku molecules (Fig. 1A, lane 2). When XL was added, the Ku-DNA complexes were replaced by a lower mobility Ku-XL-DNA complex (Fig. 1A, lane 3).
When XLF was added, an even lower mobility MEnd ligase-DNA complex replaced the Ku-XL-DNA complex. Increasing concentrations of XLF formed progressively lower mobility  (Fig. 1D,  lanes 6 -10). Each XLF concentration formed a complex with a characteristic mobility rather than a series of complexes with different mobilities. Whereas binding of Ku to DNA was distributive, binding of Ku, XL, and XLF to DNA was cooperative and suggested binding of multiple XLF molecules.
XLF Mutations That Disrupt the MEnd Ligase-DNA Complex Also Disrupt End Ligation-To determine whether XLF must bind both DNA and XL to form a MEnd ligase-DNA complex, we studied two XLF mutant proteins, XLF 1-224 and XLF L115A . XLF  , which deletes C-terminal amino acids 225-299 and disrupts binding to DNA (13), failed to form a MEnd ligase-DNA complex and failed to stimulate blunt-toblunt and blunt-to-3Ј overhang end ligation (Fig. 2, A, lanes 7-9, and C). XLF L115A , which disrupts binding to XRCC4 (13),

Cooperative Assembly of a Nonhomologous End-joining Complex
formed a barely detectable MEnd ligase-DNA complex, consistent with its weak stimulation of blunt-to-blunt and bluntto-3Ј overhang end ligation (Fig. 2, B, lanes 7-9, and C). Thus, defects in formation of the MEnd ligase-DNA complex correlated with defects in blunt end and mismatched end ligation.
A Two-step Gel Elution Procedure Measures the Stoichiometry of Protein-DNA Complexes-To determine the stoichiometry of the protein-DNA complexes, we developed a two-step gel elution procedure ("Experimental Procedures"). The first step resolved the protein-DNA complex by EMSA. The second step excised the complex from the EMSA gel and resolved the DNAbound proteins by SDS-PAGE.
To validate the procedure, we performed a pilot experiment by incubating Ku with DNA and using an EMSA to resolve the protein-DNA complex containing a single Ku molecule (Fig.  3A, lanes 6 and 7). In the second step, we excised the Ku-DNA complex from the EMSA gel (marked by the dotted lines) and used SDS-PAGE to resolve the proteins contained in the gel slice (Fig. 3B).
Ku is an obligate heterodimer of Ku70 and Ku80. Because the measured amounts of Ku70 and Ku80 in the gel slice were 0.35 pmol and 0.22 pmol, respectively, we estimated the amount of Ku to be the average of the two values, 0.29 pmol. This was nearly identical to the amount of DNA, 0.27 pmol, and consistent with the predicted 1:1 molar ratio of Ku to DNA (Fig. 3E).
Ku and XL Form a Ku-XL-DNA Complex Containing Multiple Ku and XL Molecules-Addition of increasing concentrations of XL to Ku and DNA formed a Ku-XL-DNA complex of progressively lower mobility without evidence of multiple bands (Fig. 4A, lanes 7-9). Thus, binding of Ku and XL to DNA was cooperative. We scanned the gel for fluorescence and quantified the 6-FAM-DNA in the Ku-XL-DNA complex (Fig.  4, A and C). Proteins in Ku-XL-DNA gel slices were resolved by SDS-PAGE and quantified by Krypton protein stain using dilutions of unbound Ku and XL (Fig. 4, B-E).
As in the pilot experiment, the amount of Ku70 was somewhat higher than the amount of Ku80. XL was purified as a stable protein complex of XRCC4 and Ligase IV, and the expected molar ratio of XRCC4 dimer to Ligase IV was 1:1. The measured XRCC4 to Ligase IV molar ratios were somewhat less than 1:1. These deviations indicated an uncertainty of 10% in protein quantification.
In the Ku-XL-DNA complex, each DNA molecule bound four to seven molecules of Ku and XL (Fig. 4E). The presence of multiple Ku molecules in the Ku-XL-DNA complex was surprising because in the absence of XL, the Ku concentration used in the binding reaction generated mostly a one Ku-DNA complex (Fig. 4A, lane 6). For successive 2-fold increases in XL concentration, the XL to Ku molar ratio in the Ku-XL-DNA complex increased slowly from 1.0 to 1.1 to 1.2. Thus, Ku binding to DNA recruits XL (Fig. 1) boxes (a, b, and c) were excised from the EMSA gel, denatured, and resolved by SDS-PAGE. C, standard curve for DNA quantification. D, standard curves for protein quantification. E, DNA and protein quantification of the protein-DNA complexes.
EMSA gel matrix to generate sharper bands. This decreased the uncertainty in protein to DNA molar ratio measurements, which require matching the gel region scanned for DNA and the gel slice analyzed for protein. Because the new gel matrix might affect recovery of proteins from the EMSA gel slice, we checked the protein-to-protein molar ratio in the Ku-XL-DNA gel slices from the new gel matrix (Fig. 5A,  lanes 7 and 8) and confirmed that it remained near 1:1 (data not shown).
Addition of increasing concentrations of XLF to Ku, XL, and DNA formed MEnd ligase-DNA complexes of progressively lower mobility (Fig. 5A, lanes 9 -11). For successive 2-fold increases in XLF concentration, the number of XLF molecules increased from 2.6 to 3.6 to 5.1, and the number of XL molecules increased from 3.8 to 4.1 to 4.6 (Fig. 5, B-E). Thus, Ku and XL binding to DNA recruits XLF (Fig. 1), and XLF binding recruits XL. This suggests that the MEnd ligase-DNA complex contains a filament of alternating XL and XLF molecules.
The MEnd ligase-DNA complex contained fewer Ku molecules with increased XL and XLF binding, declining from 2.0 to 1.3 to 1.1. This result was unexpected because Ku has a higher affinity for DNA than either XL or XLF (Fig. 1, B and C) and because the Ku-XL-DNA complex contained more Ku molecules with increased XL binding. Thus, assembly of progressively larger MEnd ligase-DNA complexes resulted in an increasing XL to Ku molar ratio, whereas by contrast, the assembly of progressively larger Ku-XL-DNA complexes maintained an XL to Ku molar ratio close to 1:1.
The MEnd Ligase-DNA Complex Forms More Efficiently Than the Ku-XL-DNA Complex-Although the two-step gel elution procedure was designed to ensure its accuracy, we wanted a direct and more sensitive method for estimating the number of XLF molecules in the MEnd ligase-DNA complex. Therefore, we labeled XLF with TAMRA, which increased XLF detection sensitivity by 10-fold. This eliminated the gel excision step by allowing in situ quantification of XLF in the EMSA gel. The DNA probe was labeled with 33 P rather than 6-FAM to avoid interference with the TAMRA signal.
To confirm that TAMRA labeling preserved the MEnd ligase-DNA complex, we added unlabeled XLF or TAMRA- labeled XLF in increasing concentrations to Ku, XL, and DNA and resolved the complexes by EMSA. XLF and TAMRA-XLF generated similar patterns of bands with two ranges of mobility (Fig. 6A, lanes 5-16). Only the lower mobility bands contained TAMRA (Fig. 6A, right panel), confirming that XLF and TAMRA-XLF formed similar MEnd ligase-DNA complexes.
Close inspection of the gel showed that the lower mobility bands did migrate into the gel, particularly for lower concentrations of XLF (Fig. 6A, lanes 5-7 and 11-13). To confirm that the band represented a well defined protein-DNA complex and not protein-DNA aggregates, we allowed the free DNA to migrate off the bottom of the gel (Fig. 6B).
An XLF concentration of only 3 nM was sufficient to form the same amount of MEnd ligase-DNA and Ku-XL-DNA complexes despite much higher concentrations of XL (100 nM) and of Ku (20 nM) (Fig. 6A, lane 5). Furthermore, an XLF concentration of only 13 nM resulted in disappearance of the Ku-XL-DNA complex (Fig.  6A, lane 7). TAMRA-XLF also formed a MEnd ligase-DNA complex and suppressed the Ku-XL-DNA complex, albeit to a slightly lesser degree. This is probably due to a small decrease in DNA binding because the smear of radioactivity from direct binding of XLF to DNA was less prominent for TAMRA-XLF (Fig. 6A, compare lanes 10 and 16). Nevertheless, for both XLF and TAMRA-XLF, the MEnd ligase-DNA complex formed much more readily than the Ku-XL-DNA complex.
To study the MEnd ligase-DNA complex further, we added increasing concentrations of XL to Ku and unlabeled XLF, or to Ku and TAMRA-labeled XLF. The results for unlabeled and TAMRA-labeled XLF were similar, showing that TAMRA labeling preserves the MEnd ligase-DNA complex (Fig. 6C,  compare lanes 8 -13 with lanes 14 -19). XL at 25 nM formed a MEnd ligase-DNA complex with minimal formation of Ku-XL-DNA or Ku-DNA complexes (Fig. 6C, lanes 10 and 16). This confirmed the result in Fig. 6A, showing that the MEnd ligase-DNA complex forms more readily than the Ku-XL-DNA complex. Moreover, increasing XL concentrations recruited additional XLF molecules.
To estimate the molar ratio of XLF to DNA, we performed in situ measurements of the EMSA gels for TAMRA-XLF and 33 Plabeled DNA (Fig. 6, A, C, and D). These measurements showed that the MEnd ligase-DNA complex could assemble with as few as one XLF molecule, and as many as five XLF molecules (Fig.  6E).
The MEnd Ligase-DNA Complex Has an Anomalously Low Electrophoretic Mobility-MEnd ligase-DNA complexes always had a lower mobility than Ku-XL-DNA complexes (Figs.  1, 2, and 5). The largest fully analyzed Ku-XL-DNA complex contained 5.7 Ku and 7.1 XL molecules (Fig. 4E). Based on molecular masses for Ku (150 kDa), XL (185 kDa, from XRCC4 homodimer, 76 kDa, and Ligase IV, 109 kDa), the protein molecular mass of the Ku-XL-DNA complex was 2.2 MDa. The smallest fully analyzed MEnd ligase-DNA complex contained 2.0 Ku, 3.8 XL, and 2.6 XLF (homodimer, 66 kDa) molecules, corresponding to a total molecular mass of 1.1 MDa (Fig. 5E). Although the molecular mass of proteins in the MEnd ligase-DNA complex was half the molecular mass of proteins in the Ku-XL-DNA complex, it migrated with a lower mobility in the EMSA. Indeed, the smallest MEnd ligase-DNA complex containing only a single XLF molecule migrated more slowly than the largest Ku-XL-DNA complexes (Fig. 6A). Thus, the Ku-XL-DNA complexes and MEnd ligase-DNA complexes exhibited an anomalous difference in mobility.
Larger Ku-XL-DNA and MEnd Ligase-DNA Complexes Are Associated with Increased DNA End-joining Efficiency-We measured end joining under conditions favoring intramolecular ligation of a linear DNA substrate. Increasing XLF or XL concentrations led to increased joining of two blunt ends or mismatched blunt and 3Ј overhanging ends (Fig. 7). The increased joining efficiency could be due to increased size of the complexes or to increased numbers of complexes. Indeed, increased size and increased number both occurred with increasing XLF or XL concentrations (Figs. 1-6).

DISCUSSION
Here we show that Ku, XL, and XLF bind DNA cooperatively to form a megadalton protein-DNA complex. The biochemistry recapitulated observations in living cells. We used XRCC4 in a complex with Ligase IV because XRCC4 requires Ligase IV for stable localization to double-strand breaks in living cells (26). In our EMSAs, as in living cells, stable recruitment of XLF to double-strand breaks required Ku, XL, and an intact C-terminal DNA binding domain in XLF (25,27).
Formation of the MEnd ligase-DNA complex correlated with end joining. The absence of Ku or XL prevented complex formation and as shown previously, abolished end joining (22,28). Mutant proteins XLF 1-224 and XLF L115A , which disrupted XLF interactions with DNA and with XRCC4, respectively, disrupted complex formation and decreased joining efficiency. Increasing XL and XLF concentrations increased the size of the MEnd ligase-DNA complex and increased endjoining efficiency.
Structural studies show that XRCC4 and XLF can form a helical filament of alternating XRCC4 and XLF molecules (16 -19). Depending on crystallization conditions, the helical struc-  3-7, lanes 8 -12, and lanes 14 -18). For lanes 13 and 19, the XL concentration was 100 nM. D, DNA amount was plotted against TAMRA-XLF fluorescence intensity. Data are results from scanning gels in A and C for [ 33 P]DNA and TAMRA-XLF. E, the relative ratios of XLF to DNA are expressed as multiples of unknown variables x and y for the gels in A and C, respectively. The amounts of Ku, XL, XLF, and DNA in the binding reactions for A were present in the same molar ratios, at lower absolute concentrations as the binding reactions for Fig. 5. Therefore, we set x ϭ 1, its minimum value. Because the binding reactions in A, lanes 15 and 16, and C, lanes 18 and 19,  ture undergoes a full turn of six XRCC4/XLF pairs every 720 -860 Å. Because the crystal structures lacked DNA as well as Ku and Ligase IV, they were consistent with several models for how XRCC4 and XLF might bind to DNA.
Several pieces of data indicate that the MEnd ligase-DNA complex contains a filament of alternating XL and XLF molecules bound to the DNA. First, XL and XLF bound to DNA in a 1:1 molar ratio (Fig. 5). Second, XL recruited XLF, and XLF recruited XL (Fig. 6). Third, XRCC4/XLF co-crystals contain filaments of alternating XRCC4 and XLF molecules. Finally, the dimensions of the XRCC4/XLF co-crystal are consistent with the stoichiometry of the largest MEnd ligase-DNA complex, which contained one Ku molecule and five molecules each of XL and XLF bound to 180 bp of DNA (Fig. 5). Because Ku binds 20 bp of DNA (29), one XL/XLF pair would bind about 30 bp of DNA, approximating the distance spanned by the head domains in the XRCC4/XLF co-crystal.
We detected MEnd ligase-DNA containing two Ku molecules, indicating that Ku can interrupt regions of alternating XL and XLF molecules. This complex may form in intact cells, which contain thousands of Ku molecules and only hundreds of XL and XLF molecules (12,30,31).
In the absence of XLF, we detected cooperative formation of a Ku-XL-DNA complex. Ku binding to DNA recruited XL, and XL binding recruited Ku. As many as six molecules each of Ku and XL bound to DNA in a 1:1 molar ratio. It is unlikely that the complex contained adjacent Ku molecules because Ku does not interact with itself while bound to DNA (6). Thus, the data  suggest that Ku-XL-DNA is a filament of alternating Ku and XL molecules (Fig. 8).
The electrophoretic mobility of a protein-DNA complex generally decreases with increasing protein molecular mass (32,33). However, interference with normal DNA "reptation" (i.e. snake-like movement) through the pores of the gel (34) can decrease mobility further. For example, this occurs with proteins that force DNA into a bent configuration (35,36).
Surprisingly, the electrophoretic mobility of a 2.2-MDa Ku-XL-DNA complex was greater than a 1.1-MDa MEnd ligase-DNA complex. This could not be explained by synapsis of two MEnd ligase-DNA complexes. None of the many reaction conditions we tested formed a higher mobility MEnd ligase-DNA complex or supported detectable levels of end joining (Figs. 1-6). 3 Thus, Ku-XL-DNA and MEnd ligase-DNA must form structures with different effects on reptation through the gel. For example, the Ku-XL-DNA may be more flexible than the MEnd ligase-DNA. Alternatively, Ku-XL-DNA may have a smaller radius of gyration than MEnd ligase-DNA.
Cooperative assembly of the protein-DNA filaments suggests that two Ku-XL-DNA, or two MEnd ligase-DNA filaments can undergo end-to-end fusion. Such a structure would allow sliding of the two DNA ends, so that one of the Ligase IV molecules arrayed along the DNA could ligate the ends when they slide into juxtaposition. Ku-XL-DNA, which supports ligation of two DNA ends with cohesive overhangs, may not form a structure that properly aligns DNA ends with mismatched overhangs. By contrast, MEnd ligase-DNA, with its differences from Ku-XL-DNA, both in the molar ratio of XL to Ku and in electrophoretic mobility, must form a distinct structure that successfully aligns mismatched ends.
Cooperative formation of a protein-DNA filament occurs in both nonhomologous end joining and homologous recombination. Our EMSAs showed cooperative formation of Ku-XL-DNA and MEnd ligase-DNA filaments with increasing concentrations of XL or XLF forming progressively larger complexes. Similarly, previously reported EMSAs show cooperative formation of a protein-DNA filament with increasing concentrations of the homologous recombination RecA homolog Dmc1 (37).
Interactions with other NHEJ proteins can occur prior to formation of the full MEnd ligase-DNA complex. In living cells, DNA-PKcs stabilizes the recruitment XRCC4 and/or XL to double-strand breaks (38). For nonligatable ends, DNA-PKcs can recruit the Artemis nuclease to remove damaged terminal nucleotides, and XRCC4 can recruit polynucleotide kinase/ phosphatase to prepare terminal 5Ј-OH or 3Ј-phosphate groups for ligation (30).