Different DNA End Configurations Dictate Which NHEJ Components Are Most Important for Joining Efficiency

The nonhomologous DNA end-joining (NHEJ) pathway is a key mechanism for repairing dsDNA breaks that occur often in eukaryotic cells. In the simplest model, these breaks are first recognized by Ku, which then interacts with other NHEJ proteins to improve their affinity at DNA ends. These include DNA-PKcs and Artemis for trimming the DNA ends; DNA polymerase μ and λ to add nucleotides; and the DNA ligase IV complex to ligate the ends with the additional factors, XRCC4 (X-ray repair cross-complementing protein 4), XLF (XRCC4-like factor/Cernunos), and PAXX (paralog of XRCC4 and XLF). In vivo studies have demonstrated the degrees of importance of these NHEJ proteins in the mechanism of repair of dsDNA breaks, but interpretations can be confounded by other cellular processes. In vitro studies with NHEJ proteins have been performed to evaluate the nucleolytic resection, polymerization, and ligation steps, but a complete system has been elusive. Here we have developed a NHEJ reconstitution system that includes the nuclease, polymerase, and ligase components to evaluate relative NHEJ efficiency and analyze ligated junctional sequences for various types of DNA ends, including blunt, 5′ overhangs, and 3′ overhangs. We find that different dsDNA end structures have differential dependence on these enzymatic components. The dependence of some end joining on only Ku and XRCC4·DNA ligase IV allows us to formulate a physical model that incorporates nuclease and polymerase components as needed.

The mammalian genome is a vast target for genotoxic agents. It is estimated that the genome undergoes ϳ100,000 altera-tions/day, which can result in an estimated 10 double-stranded breaks (DSBs)/day (1)(2)(3)(4)(5). 3 The NHEJ DNA repair pathway is required to repair many of these DSBs.
The NHEJ pathway first begins with the toroid-shaped Ku heterodimer (Ku70 and Ku80) binding to the free dsDNA ends. Ku then recruits other NHEJ factors as needed (6).
Many components improve NHEJ efficiency in vivo, but the only essential component for NHEJ is the DNA ligase complex consisting of XRCC4⅐DNA ligase IV (X4⅐LIV) in a 2:1 ratio that facilitates the ligation of a 3Ј-hydroxyl to a 5Ј-phosphate at the partner DNA end (7)(8)(9). More recently, NHEJ accessory factors, XLF and PAXX, have been discovered that bind to DNA ligase IV and Ku, respectively (10 -14). Both XLF and PAXX have been shown to promote the ligation of DNA ends that do not have any terminal base pairing or microhomology (incompatible ends) (10 -12, 14). However, if the DNA ends require processing (e.g. if they are incompatible), a nuclease or polymerase may become essential. Artemis appears to be the major NHEJ nuclease (15), although other nucleases, such as APLF (also called PALF), may participate in limited cases, especially when Artemis is not present (16). Artemis has 5Ј exonuclease activity by itself and acquires endonuclease activity when complexed with autophosphorylated DNA-PK cs (15,17). A NHEJ reconstitution system in which we can observe the joined products directly following PAGE would provide information on the ligation of each strand of the duplex, allow us to determine NHEJ efficiencies, and provide detailed mechanistic insight after sequencing the junctions (Fig. 1). One earlier direct gel NHEJ reconstitution did not include a nuclease but rather focused on the ability of the ligase complex to ligate across gaps in either strand and the ability of polymerase and to add random nucleotides (nt), thereby providing new microhomology (MH) between two DNA ends (18). Another early reconstitution included all known nuclease, polymerase, and ligase components but was not sufficiently efficient to permit direct gel assessment of joining products and tested only one pair of DNA end configurations (19). None of the reconstitutions have demonstrated a role for DNA-PK cs . The study here is the first direct gel NHEJ reconstitution system that includes all major nuclease, polymerase, and ligase components and DNA-PK cs . We find that the contribution of each major component to the efficiency of DNA end joining depends on the configuration of the two DNA ends being joined in a manner that permits a coherent model for NHEJ to be proposed.

Results
Resection-dependent Compatible DNA Ends Require Artemis and Are Strongly Stimulated by Ku and DNA-PK cs -Ligation of DNA ends with MH of even 1 base pair can be more efficiently ligated by the X4⅐LIV complex than blunt DNA ends. We wondered whether DNA ends with 3Ј overhangs without terminal MH but with internal MH would undergo ligation. We incubated a 5Ј-radiolabeled DNA substrate with a (CCC CTT TTT T-3Ј) overhang and a substrate with a (GGG G-3Ј) overhang with Ku, DNA-PK cs , Artemis, and X4⅐LIV, as indicated (Fig. 2). These two DNA ends have the potential of generating 1-4-bp MH. The 3Ј end of the bottom strand of the labeled duplex and the 5Ј end of the bottom strand of the unlabeled duplex were biotinylated so that we could block one end of each duplex by adding streptavidin. The string of Ts in the 3Ј overhang is intended to prevent base pairing until the Ts are resected. In this paper, we use the phrase "resection-dependent" to refer to joining that depends on any loss of nucleotides, regardless of DNA strand polarity. Supporting this point, we observed that NHEJ ligation products are not generated unless the reactions contain the Artemis nuclease (Fig. 2, lanes 2-4 and lane 6).
However, pairwise or more complete reactions have measurable ligation. Approximately 2% of the substrate is converted to ligated product when Artemis and X4⅐LIV are present (Fig. 2, lane 5). The addition of Ku does not change NHEJ efficiency (Fig. 2, lane 7). When Artemis and DNA-PK cs are used along with X4⅐LIV, NHEJ efficiency increases to 11% (Fig. 2, lane 8). The addition of Ku to DNA-PK cs , Artemis, and X4⅐LIV further increases NHEJ efficiency to 18% (Fig. 2, lane 9). We describe this event as resection-dependent (RD)-compatible end NHEJ because ligation can occur efficiently if the Artemis⅐DNA-PK cs complex is present to efficiently remove the nonhomologous portion of the overhang so that the region of MH can be used to stabilize the DNA ends for ligation.
X4⅐LIV Stimulates DNA-PK cs -independent Artemis Activity-We were curious why there was resection and NHEJ products in Artemis and X4⅐LIV conditions without DNA-PK cs (Fig. 2, lane 5). Artemis has intrinsic 5Ј exonuclease activity and endonuclease activity when complexed with DNA-PK cs . We incubated the (CCC CTT TTT T-3Ј) overhang substrate with either Artemis alone; DNA-PK cs and Artemis; Artemis and X4⅐LIV; or DNA-PK cs , Artemis, and X4⅐LIV (Fig. 3A). It has been reported that the C-terminal region of Artemis (amino acids 485-495) FIGURE 1. NHEJ reconstitution workflow. NHEJ proteins are added to labeled DNA substrate, and NHEJ products are resolved by denaturing PAGE. NHEJ products are cut out and PCR-amplified. PCR products are TA cloned into a pGEM-T vector and transformed into E. coli. Vector is then isolated from colonies, and the junctions are sequenced. may interact with the DNA binding domain of DNA ligase IV (20,21). We find that Artemis resection of the 3Ј overhang increases when X4⅐LIV is added (Fig. 3A, lane 2 versus lane 4). We next tested whether the Artemis resection requires the presence of a partner substrate that may be able to generate a 3Ј flap if the two overhangs transiently anneal by forming C-G base pairs prior to resection of the dTs (Fig. 3B). We find that Artemis resection of the 3Ј overhang is stimulated by the addition of X4⅐LIV, independent of a second partner substrate (Fig.  3B, lane 4).
Incompatible 3Ј Ends Cannot be Joined in a Biochemical System Containing Ku, DNA-PK cs , Artemis, and X4⅐LIV Complex Alone-We next attempted to increase the complexity of in vitro NHEJ by testing substrates with no regions of MH in the overhangs (i.e. incompatible ends). Joining of a DNA end with a (CCC CTT TTT T-3Ј) overhang to an end with a (TTT T-3Ј) overhang showed marginal levels of NHEJ (Fig. 5A, lanes 7-9). Similarly, NHEJ of a DNA end with a (CCC CTT TTT T-3Ј) overhang with a blunt-ended DNA end pair showed marginal levels of NHEJ (Fig. 5B, lanes 7-9). Because of the small percentage of total signal, the visible but faint NHEJ product bands in Fig. 5 (A and B, lanes 7-9) were not quantitatively greater than the gel background. These data suggest that Artemis⅐DNA-PK cs resection activity alone is not sufficient to promote NHEJ of fully incompatible DNA ends. In contrast, blunt end ligation required only Ku and X4⅐LIV (Fig. 5C, lane 3). The addition of DNA-PK cs actually inhibited ligation ϳ2-fold irrespective of whether Artemis is also included (Fig. 5C, lane 3 versus lane 6 versus lane 7 versus lane 9). These data suggest that NHEJ of blunt-ended substrates in our biochemical system proceeds by direct ligation without any requirement for MH or terminal base pairing after resection.
Pol but Not Pol Stimulates NHEJ of 3Ј Incompatible DNA Ends-Because Ku, Artemis, DNA-PK cs , and X4⅐LIV were not sufficient for NHEJ of incompatible 3Ј overhangs, we tested whether Pol X family polymerases can stimulate NHEJ. The Pol X family members consist of Pol , Pol , Pol ␤, and terminal deoxynucleotidyl transferase (TdT). Pol ␤ and TdT do not participate in NHEJ, except for the role of TdT in NHEJ in pre-B or pre-T cells (4). Pol has the ability to add nucleotides in either a template-dependent or template-independent manner (22). Conversely, Pol is primarily a template-dependent polymerase, with much less template-independent activity (22). The addition of Pol or Pol did not significantly increase NHEJ of RD-compatible ends (Fig. 6, A, lanes 2-4, and B, lane 2 versus lane 4). However, Pol was able to promote two sequential ligation events, suggesting that streptavidin does not completely block one end from ligation and that Pol may be promoting NHEJ (Fig. 6, A, lane 3, and B, lane 4). These data suggest that resection, and not nucleotide addition, are sufficient for NHEJ of RD-compatible DNA ends, but Pol may be able to add nt that would generate occult MH (4).
NHEJ of 3Ј incompatible DNA ends increased from ϳ0.1 to 4.6% with the addition of Pol but not Pol (Fig. 6A, lanes 6 -8). A similar increase from undetectable to 6.0% was observed with the addition of Pol in the NHEJ of a 3Ј overhang with blunt-ended DNA partner (Fig. 6A, lanes 10 -12). Furthermore, Pol does not provide a DNA end stabilization function, as indicated by the lack of NHEJ improvement upon Pol inclusion in dNTP-free conditions (Fig. 6B, lanes 2-4, 6 -8, and 10 -12). (Weaker intensity bands that migrated between the substrate and NHEJ product are likely DNA hairpinned products from NHEJ of two molecules of the identical duplex followed by melting and intramolecular annealing of each strand (18).) These data suggest that Pol may be adding nucleotides to help form regions of MH for NHEJ of substrates with incom- patible 3Ј overhangs. Because Pol does not stimulate NHEJ, the contribution of Pol most likely proceeds via its templateindependent addition.
A smaller effect was observed in the NHEJ of blunt-ended DNA. Pol only marginally increased NHEJ from 14.2% to 20.4%, whereas Pol did not change NHEJ (Fig. 6A, lanes 14 -16). These data support our other findings suggesting that blunt-ended NHEJ may proceed through direct ligation even though the Artemis⅐DNA-PK cs complex, which is capable of resecting nucleotides from the blunt ends, is present.
PAXX, but Not XLF, Stimulates NHEJ of Blunt DNA Ends-We next wondered whether XLF and PAXX would improve joining in our biochemical NHEJ system. We incubated XLF and/or PAXX with Ku, DNA-PK cs , Artemis, X4⅐LIV, and Pol (Fig. 7). We observed that XLF and PAXX have no stimulatory effect on the NHEJ of RD-compatible ends (Fig. 7A, lanes [2][3][4][5]. It is possible that at RD-compatible ends, other DNA ligase factors may be unnecessary. NHEJ of incompatible 3Ј ends was not improved with XLF or PAXX (Fig. 7B, lanes 2-5). The same was observed for the NHEJ of DNA with a 3Ј overhang with a blunt-ended DNA partner (Fig. 7B, lanes 7-10). However, NHEJ of blunt-ended DNA was stimulated ϳ2-fold by PAXX but not XLF (Fig. 7B, lanes [12][13][14][15]. We had already reported the finding that XLF does not improve blunt-end ligation (23). These new data suggest that the role of PAXX may be to help align the 5Ј-phosphate and 3Ј-hydroxyl by interacting with Ku so that the ligase can ligate efficiently.
NHEJ of a 5Ј Overhang to a Blunt End Is Stimulated by XLF and PAXX-Next we tested the effect of XLF and PAXX on the NHEJ of a 5Ј overhang substrate with the radiolabel on the blunt-ended 3Ј end (Fig. 8A). This substrate can undergo head to tail, head to head, or tail to tail ligation events; however, both the head to head and tail to tail events will quickly form hairpins. The hairpinned DNA will migrate much faster than the linear products formed from head to tail events. Because we did not observe any hairpinned bands between the NHEJ product and substrate bands, we assumed that the preferred NHEJ event was the (5Ј-TTT TTT CCC) head ligating to the blunt-ended tail end of the substrate (Fig. 8). XLF and PAXX marginally improved the low levels of NHEJ observed in Ku and X4⅐LIV reactions (Fig. 8B, lanes 2 -5). In reactions with Ku, Artemis, and X4⅐LIV, XLF and PAXX did not improve NHEJ alone, but marginally improved NHEJ with XLF and PAXX together (Fig.  8B, lanes 6 -9). However, NHEJ was stimulated individually by XLF and PAXX in reactions containing Ku, DNA-PK cs , Artemis, and X4⅐LIV (Fig. 8B, lanes 10 -12). The addition of both XLF and PAXX together increases NHEJ ϳ3-fold from 1% (XLF or PAXX alone) to 3% (XLF and PAXX) (Fig. 8B, lanes 11-13). These data suggest that XLF and PAXX are able to promote NHEJ of a 5Ј overhang with a blunt-ended DNA partner. Because the ligation product band migrates approximately at the 152-nt position, it is likely that the 5Ј overhang is mostly processed prior to ligation (Fig. 8B, lane 13). In addition, the NHEJ products are larger when Artemis is not present (Fig.  8B, lanes [3][4][5] or when Artemis 5Ј exonuclease activity is dominant in DNA-PK cs -free conditions (Fig. 8B, lanes 6 -9).
We next tested the role of Pol and Pol in the NHEJ of the 5Ј overhang to a blunt end. Both Pol and Pol only marginally stimulated NHEJ (Fig. 8C, lanes 2-4). These data suggest that the addition of nucleotides is not used to join DNA with blunt ends and 5Ј overhangs.
NHEJ of RD-compatible Ends Reveals a Preference for Resection to Generate MH for Ligation-We next sequenced the junctions from the joining products by cutting out bands from the PAGE gels, PCR amplifying, and TA cloning them into a pGEM-T vector for sequencing (Fig. 1). We found that NHEJ of RD-compatible ends is maximal in Ku, DNA-PK cs , and Artemis conditions (Figs. 2, lane 9; 5A, lane 2; and 6B, lane 2). The sequences from these junctions showed that the preferred modification is to resect the 3Ј dTs from the (CCC CCT TTT TT-3Ј) overhang to create ϳ3-4-nt MH to stabilize ligation (Tables 1 and 2). Surprisingly, the addition of Pol did not result in any sequences with additional nucleotides (Table 1). In all junctions analyzed (n ϭ 99), irrespective of NHEJ proteins involved, the end processing never proceeded into the duplex portion but rather stopped at or before the 3Ј overhang (supplemental Table S1).

NHEJ Junctions from the Joining of 3Ј Incompatible Ends Reveal a Preference for Resection and Nucleotide Addition to
Provide MH for Ligation-On the other hand, NHEJ of DNA ends with no potential microhomology on the 3Ј overhangs (incompatible 3Ј ends) was very weak without Pol (Fig. 6, A,  lane 7, and B, lane 8). Hence, the predominant products show resection of nucleotides from the (CCC CCT TTT TT-3Ј) overhang and Pol synthesis of primarily dAs which creates 2-to 4-nt MH with the 3Ј (TTT T-3Ј) on the other DNA end (Tables  1 and 2). When Pol was not included in the reactions, the preferred junctions involved complete resection of the (CCC CCT TTT TT-3Ј) overhang until the stretch of two dAs were exposed 3 bps into the duplex (supplemental Table S2). These dAs were then used to provide MH with the (TTT T-3Ј) overhang on the other DNA end. One unique sequence was the result of 3-bp MH and a ligation over a G-T mismatch (Table 1). Not surprisingly, X4⅐LIV was able to ligate over this mismatch (18).
NHEJ of a 3Ј Overhang with a Blunt-ended Partner Reveals a Preference for Resection and Ligation without MH-We next examined the junction sequences from the NHEJ of a 3Ј overhang ligated with a blunt-ended partner. NHEJ efficiency was best in reactions that contained Pol (Fig. 6, A, lane 11, and B,  lane 12). Sequences of these junctions revealed that the DNA ends undergo resection on both ends of the DNA without significant nucleotide addition. Surprisingly, only 2 of the 13 sequenced junctions showed the addition of nucleotides (Table  1). Furthermore, only one of those sequences resulted from a 3-bp MH with ligation over a G-T mismatch (Table 1). Overall, the sequencing results indicate that the preferred repair process is to resect the overhang and resect into the duplex without the use of MH ( Table 2).
NHEJ of Blunt-ended DNA Reveals Preference for Direct Ligation-NHEJ of blunt-ended DNA is a highly efficient process that only requires Ku and X4⅐LIV for maximal activity (Fig.  5C, lane 3). The addition of other NHEJ factors does not improve NHEJ and does not increase the diversity of junctions generated (supplemental Table S4). Thus, the preferred joining product from two blunt DNA ends is direct ligation (Tables 1  and 2).
NHEJ of a 5Ј Overhang to a Blunt End Reveals Preference for Resection and Direct Ligation-NHEJ of a 5Ј overhang to a blunt end is maximal in Ku, DNA-PK cs , Artemis, X4⅐LIV, XLF, and PAXX conditions (Fig. 8B, lane 13). Ensuing sequence analysis shows that the 5Ј overhang is completely resected in most joints without processing of the blunt-ended partner (Table 1). Ligation then occurs without utilizing MH (Tables 1 and 2). In fact, sequencing of the Ku, Artemis, and X4⅐LIV condition also shows that ligation without MH is preferred (supplemental Table S5).

Discussion
Double-stranded DNA breaks arise from ionizing radiation, reactive oxygen species, replication errors, and inadvertent cleavage by nuclear enzymes and by exogenous chemicals. These breaks can create diverse DNA end structures that must be repaired. We have developed a NHEJ reconstitution assay that can be used to compare NHEJ efficiencies and analyze the junctions by analyzing sequences to determine the effects of the proteins involved. It had been previously determined that the ligation of compatible DNA ends is a highly efficient process that requires only the X4⅐LIV complex and is stimulated by Ku when there is a 1-nt gap to ligate over (18). Thus, we were now interested in determining how complex 5Ј and 3Ј overhangs were processed and resolved. We find that NHEJ of incompatible DNA ends is less efficient compared with NHEJ of compatible DNA ends. This inefficient process is improved with the addition of Pol for 3Ј incompatible ends. PAXX alone promotes the NHEJ of blunt ends, whereas PAXX and XLF synergistically enhance NHEJ of a 5Ј overhang to a blunt end. We have summarized these observations in the diagram in Fig. 9 (the red squares represent known protein interactions, and the red stars represent interactions that increase Artemis activity). The conclusions drawn from these experiments support the flexibility and differing NHEJ protein requirements that we previously hypothesized might exist (24). Ku Contribution to NHEJ-We find that NHEJ of all DNA ends without MH requires Ku. We previously showed that NHEJ of ends with MH does not require Ku (18), and as the ends become more difficult to stabilize, perhaps by transient end annealing, Ku is required. Ku has a high affinity for DNA ends (K D ϭ 5.9 ϫ 10 Ϫ10 M) (25) and can promote the binding of X4⅐LIV to the DNA end when it is present (26). This explains why Ku is important for all of the substrates we test here in our study.
Artemis⅐DNA-PK cs Contribution to NHEJ-The Artemis⅐DNA-PK cs complex is required for hairpin opening of the coding joints in V(D)J recombination (27). Consequently, both Artemis and DNA-PK cs -deficient cells fail to cleave this hairpin, consistent with DNA-PK cs activation of Artemis (4). In the context of NHEJ, Artemis-deficient cells are radiosensitive, but the majority of the DSBs in these cells are repaired efficiently (28,29). Approximately 20% of ionizing radiation-induced DNA DSBs required Artemis for repair (30). Here, we find that Artemis activity is required for NHEJ of DNA substrates with incompatible 5Ј and 3Ј overhangs. In contrast, blunt-ended DNA undergoes direct ligation. The Artemis⅐DNA-PK cs complex is able to resect the overhang to reveal regions of MH in RD-compatible ends (Table 1). For DNA ends with incompatible 3Ј overhangs, resection occurs, but that is not sufficient for efficient NHEJ because there are no regions of MH (Fig. 5, A  and B). Often, the resection goes deeper into the duplex to allow exposure of other regions of MH (supplemental Table S3). Artemis⅐DNA-PK cs activity on 5Ј overhangs prefers resection of the overhang to result in a blunt DNA end (27). Thus, the 5Ј overhang is processed to a blunt-ended substrate, in which direct ligation is favored.  (20,21,31), although to a lesser extent than by DNA-PK cs . DNA ligase IV may be able to stimulate Artemis at DNA ends because X4⅐LIV has an affinity for ligatable DNA ends. (Hairpin DNA ends likely do not bind X4⅐LIV.) The dependence of some DSBs on DNA-PK cs and not other DSBs may be relevant to immunoglobulin class switch recombination. During Ig class switch recombination, DNA lesions generated by activation-induced deaminase lead to DSBs. We speculate that after the 5Ј nuclease, ExoI, acts, many 3Ј overhangs are common, although this may depend on which downstream Ig switch region is being joined to the upstream S DNA end. This would explain why DNA-PK cs is variably required for some Ig switch recombinations but not others (32,33).
Pol and Contribution to NHEJ-Pol X family members consist of Pol , , ␤, and TdT. Pol ␤ is involved in base excision repair and lacks the BRCT domain that is responsible for its interaction with Ku (19). TdT is only expressed in pre-B and pre-T lymphocytes and contributes to increasing immune diversity by adding random nucleotides to coding ends in V(D)J recombination. Thus, TdT has no role in NHEJ outside the context of V(D)J recombination. Pol is known to be involved in NHEJ and has the ability to add nucleotides in a templateindependent and -dependent manner (18). Pol on the other hand, is primarily a template-dependent polymerase. The structural differences in loop 1 of these polymerases have been implicated as the flexible region that promotes template-independent activity (22). Our study shows that Pol , but not Pol ,  1-4), pHC115/119 (lanes 5-8), or pHC115/120 (lanes 9 -12). Blunt end ligations were performed with 20 nM *HC121/102 and 20 nM pHC115/120 (lanes 13-16). B, 100 M dNTPs were incubated with DNA substrates 20 nM *HC101/102 and either 20 nM pHC115/116 ( lanes  1-4), pHC115/119 (lanes 5-8), or pHC115/120 (lanes 9 -12). These are representative gels of multiple similar experiments that have confirmed these results. plays a major role in promoting the NHEJ of 3Ј incompatible ends (Fig. 6). Pol strongly stimulates NHEJ of 3Ј overhangs because nucleotides can be added that can form MH with the 3Ј overhang on the other DNA partner ( Table 1). The lack of obvious regions of MH in the NHEJ of 3Ј incompatible overhangs reduces the ability of Pol to find a stable template to initiate synthesis. In contrast, 5Ј incompatible ends do not require Pol (Fig. 8C). The 5Ј overhang is resected into a blunt end efficiently by the Artemis⅐DNA-PK cs complex. This blunt end then undergoes direct ligation as is preferred for blunt ends (Table 1).
In vivo studies of POL4 (homolog to mammalian Pol X family members) mutants in yeast have also demonstrated that POL4 is indispensable for 3Ј overhangs but not for 5Ј overhangs (34). Furthermore, Pol X family members complement POL4 mutants, suggesting that the yeast model is translatable to some aspects of mammalian cells (35). Pol has likely evolved to be able to add nucleotides in a template-independent manner to help generate the MH required for ligation. Another study on C57BL/6 MEFs demonstrated that Pol was largely responsible for the nucleotide addition in short (1-2 nt) 3Ј incompatible ends (36). The observation that overhangs are resected by Artemis⅐DNA-PK cs followed by the addition of nucleotides by Pol supports their importance in increasing junctional diversity during V(D)J recombination.
XLF and PAXX Contribution to NHEJ-XLF and PAXX are the most recent NHEJ components discovered, and these interact with XRCC4 and Ku, respectively (10,11,37). Single molecule studies have shown that XRCC4/LIV/XLF may assemble at the DNA end with Ku (38). Although XLF-deficient cells are more radiosensitive and biochemical evidence has shown that XLF stimulates NHEJ of incompatible 3Ј ends (10,23), XLF may only be required for the repair of a subset of DNA ends (12). To date, PAXX has been shown to improve blunt end ligation (Fig. 7B) (11). We find that XLF and PAXX stimulate NHEJ of a 5Ј overhang to a blunt end (Fig. 8B). XLF and PAXX do not seem to play a significant role in the NHEJ of substrates with 3Ј overhangs. The role of XLF and PAXX may thus be to stabilize ends that do not utilize MH for ligation.
Concluding Comments-As summarized in Table 2, some DNA end configurations may be ligatable with the involvement of a small number of NHEJ proteins and others require more NHEJ proteins. Our biochemical reconstitution supports this view, which is consistent with in vivo studies.

Experimental Procedures
Oligonucleotides and DNA Substrates-Oligonucleotides used in this study were synthesized by Integrated DNA Technologies, Inc. (San Diego, CA). They were purified and 5Ј endradiolabeled with [␥-32 P]ATP (3,000 Ci/mmol) (PerkinElmer Life Sciences) or 3Ј end-radiolabeled using [␣-32 P]dNTP (3,000 Ci/mmol) (PerkinElmer Life Sciences). The oligonucleotides were labeled and purified as previously described (15). Oligonucleotides were quantified by using a UV spectrophotometer to determine the absorbance at 280 nm. The concentration of dsDNA was determined by annealing known concentrations of both oligonucleotides. For unlabeled DNA, equal amounts of each oligonucleotide were annealed. For radiolabeled dsDNA, a 20% excess of unlabeled oligonucleotide was added to the radiolabeled oligonucleotide to ensure that all labeled substrates were in duplex conformation.
The sequences of the oligonucleotides used in this study are as follows:    Protein Expression and Purification-Purification of Ku, DNA-PK cs , and Artemis was done as previously described (15). Ku70-His/Ku80 was purified from Hi-Five insect cells (Invitrogen, B855-02) by Ni-NTA affinity chromatography, dsDNA (oligo) affinity chromatography, and anion exchange chromatography. Endogenous DNA-PK cs was purified from HeLa cells using a series of anion exchange, cation exchange, dsDNA (oligo) affinity, and size exclusion chromatography. Artemishis was purified from Sf9 insect cells (Life Technologies, 11496-015) by Ni-NTA affinity chromatography and anion exchange chromatography. Native Pol and Pol were purified as described previously (18). The X4⅐LIV complex was purified as described (39). Briefly, X4⅐LIV was purified from Hi-Five insect cells by Ni-NTA affinity chromatography, anion exchange, size exclusion, and cation exchange chromatography. XLF-myc-His protein was purified from 293T cells by Ni-NTA affinity and anion exchange chromatography as described (40). PAXX was . Diagram of end complex. This diagram shows how the various NHEJ proteins might associate at the ends. For simplification, we have only depicted the ligation of the top strand, but the bottom strand will also undergo processing. Red stars represent interactions that stimulate Artemis activity. Red squares represent known protein-protein binding interactions. Ku interacts with LIV at the region containing the two BRCT domains (41). The region between the BRCT domains of LIV interacts with the helical domain of X4 (7)(8)(9). The N-terminal head domain of XLF interacts with the N-terminal head domain of X4 (10). The C terminus of PAXX interacts with Ku (11). The N-terminal BRCT domain of Pol interacts with the Ku⅐DNA complex (19). The FAT domain in DNA-PK cs interacts with Ku (42). Artemis is activated by its interaction with DNA-PK cs through its C terminus (amino acids 402-403). The C terminus of Artemis (amino acids 485-495) interacts with the N-terminal head domain of LIV (20,21,31), and the current study indicates that this stimulates Artemis activity (Fig. 2).
purified from a pHAT4 vector encoding for His-PAXX in BL21(DE3) Escherichia coli using Ni-NTA affinity chromatography, followed by TEV protease-mediated cleavage of the N-terminal his-tag, and size exclusion chromatography as described (11). NHEJ Assay-In vitro NHEJ assays were performed in a volume of 10 l with a buffer composition of 25 mM Tris-HCl (pH 8.0), 75 mM KCl, 10 mM MgCl 2 , and 1 mM DTT, 10% PEG 8000. The reactions consisted of 20 nM of 32 P-labeled DNA and 20 nM of unlabeled DNA substrate incubated with 0.5 mM ATP and 200 nM streptavidin (in reactions with biotinylated substrates) at 37°C for 60 min, unless specified otherwise. In addition, 100 M dNTPs were added to reactions containing a polymerase. The streptavidin is used to block the biotinylated end of the DNA from proteins. Phenol-chloroform extraction was immediately performed on the reactions to deproteinize the samples. The reactions were then analyzed on an 8% denaturing PAGE. The gels were then dried and exposed to a PhosphorImager screen overnight. The screen was scanned, and quantification was performed in Quantity One one-dimensional analysis (Bio-Rad) by dividing the NHEJ product band intensity by the total signal in each lane.
Junction Sequence Analysis-The dried denaturing PAGE gels were exposed to X-ray film for at least 24 h. The developed film was then overlaid on the gel to cut out individual NHEJ ligation bands. The bands were incubated in 100 l of 10 mM Tris (pH8) -1 mM EDTA at 37°C overnight to allow ample time for DNA diffusion. PCR was performed on the samples using primers (HC105 and HC114 for 5Ј-labeled substrates and JG187 and JG188 for 3Ј-labeled substrates) that flank the junction. The PCR product was then TA-cloned into a pGEM-T Easy vector by following the product manual (catalog no. TM042, Promega). TA-cloned product was diluted 2-fold in ddH 2 O and transformed into DH10B competent cells, and transformants were plated on LB/Amp/X-gal agar plates. White colonies were selected for plasmid mini-prep and Sanger sequencing.