A Biochemically Defined System for Mammalian Nonhomologous DNA End Joining

A Biochemically Defined System for Mammalian Nonhomologous DNA End Joining

Molecular Cell, Vol. 16, 701–713, December 3, 2004, Copyright ©2004 by Cell Press A Biochemically Defined System for Mammalian Nonhomologous DNA End ...

724KB Sizes 0 Downloads 118 Views

Molecular Cell, Vol. 16, 701–713, December 3, 2004, Copyright ©2004 by Cell Press

A Biochemically Defined System for Mammalian Nonhomologous DNA End Joining Yunmei Ma,1,2 Haihui Lu,1,2 Brigette Tippin,2 Myron F. Goodman,2 Noriko Shimazaki,4 Osamu Koiwai,4 Chih-Lin Hsieh,3 Klaus Schwarz,5 and Michael R. Lieber1,2,* 1 University of Southern California Norris Comprehensive Cancer Center Room 5428 Department of Pathology Department of Biochemistry and Molecular Biology Department of Molecular Microbiology and Immunology 2 Department of Biological Sciences 3 Department of Urology University of Southern California Keck School of Medicine 1441 Eastlake Avenue, MC9176 Los Angeles, California 90033 4 Department of Applied Biological Science Faculty of Science and Technology Tokyo University of Science Noda, Chiba 278-8510 Japan 5 Department of Transfusion Medicine Institute for Clinical Transfusion Medicine and Immunogenetics, Ulm University of Ulm Helmholtzstrasse 10 D-89081 Ulm Germany

Summary Nonhomologous end joining (NHEJ) is a major pathway in multicellular eukaryotes for repairing doublestrand DNA breaks (DSBs). Here, the NHEJ reactions have been reconstituted in vitro by using purified Ku, DNA-PKcs, Artemis, and XRCC4:DNA ligase IV proteins to join incompatible ends to yield diverse junctions. Purified DNA polymerase (pol) X family members (pol mu, pol lambda, and TdT, but not pol beta) contribute to junctional additions in ways that are consistent with corresponding data from genetic knockout mice. The pol lambda and pol mu contributions require their BRCT domains and are both physically and functionally dependent on Ku. This indicates a specific biochemical function for Ku in NHEJ at incompatible DNA ends. The XRCC4:DNA ligase IV complex is able to ligate one strand that has only minimal base pairing with the antiparallel strand. This important aspect of the ligation leads to an iterative strand-processing model for the steps of NHEJ. Introduction DSBs in mammalian cells are repaired predominantly by either NHEJ or homologous recombination (Lieber *Correspondence: [email protected]

et al., 2004). NHEJ of DSBs is thought to begin with the binding of the heterodimer Ku to the double-stranded DNA (dsDNA) ends. Ku improves the on rate and slows the off rate of DNA-PKcs from DNA ends (West et al., 1998). A significant fraction of Artemis exists in the cell in complex with DNA-PKcs, which becomes an endonuclease after it is phosphorylated by DNA-PKcs (Ma et al., 2002). After the trimming of excess or damaged DNA, the Artemis:DNA-PKcs complex may debind and permit the binding of the ligase complex, XRCC4:DNA ligase IV, which completes the joining. Pol would be needed for any NHEJ events that require fill in of gaps or extension of the 3⬘ end at 5⬘ overhangs (Lieber et al., 2004). POL4 is responsible for a substantial fraction, but not all, of the fill-in synthesis of gaps in NHEJ events in S. cerevisiae (Wilson and Lieber, 1999). (The remaining fill-in synthesis may be due to pol delta and epsilon.) POL4 is a member of the pol X family and is most homologous to pol mu and pol lambda in mammalian cells (Tseng and Tomkinson, 2002). The only other two known mammalian proteins in the pol X family are pol beta and the lymphoid-specific enzyme, terminal deoxynucleotidyl transferase (TdT). Pol mu associates with the Ku:XRCC4:DNA ligase IV:DNA complex according to studies from one group (Mahajan et al., 2002). In crude extract studies, another group reported immunodepletion evidence, suggesting a possible role for pol lambda in NHEJ (Lee et al., 2003). Further evaluation of these partial or crude extract activities requires a more complete in vitro reconstitution and requires genetic evaluation of pol mu (Bertocci et al., 2003) and pol lambda knockouts. Biochemical purification approaches have been taken to try to isolate the NHEJ system (Hanakahi et al., 2000). But NHEJ or NHEJ-like activity has not been isolated as a single fraction. Partial fractionation attempts have relied on assays that utilize crude fractions to reconstitute activity (Udayakumar et al., 2003). In addition to these, analysis of NHEJ has relied upon genetic approaches (from yeast to mammalian systems), crude extract immunodepletion, crude extract inhibition, substrate/product analyses, and analysis of individual proteins. NHEJ reconstitution of incompatible end joining with fully purified components is one key approach that has been lacking, and here we describe such a reconstitution. Results Physical Interactions of Pol X Polymerases and the NHEJ Proteins We were interested in determining which of the polymerases might function in NHEJ. Previous work from others and from us had demonstrated that in S. cerevisiae a pol X protein, POL4, contributes significantly to NHEJ (Tseng and Tomkinson, 2002; Wilson and Lieber, 1999). Given the importance of Ku in the early phases of NHEJ, we tested all of the purified pol X family members: beta, mu, lambda, and TdT (Figure 1 and Supplemental Fig-

Molecular Cell 702

Figure 1. Polymerases Mu and Lambda, but Not the ⌬BRCT Mutants, Interact with the Ku:DNA Complex (A) Pol mu, but not the ⌬BRCT mutant of pol mu, shifted the Ku:DNA complex. Pol mu shifted DNA weakly as indicated by the diffuse and faint pol mu:DNA band shown in lane 3. The positions of the one-Ku:DNA and two-Ku:DNA complexes are indicated on the left of the gel. The supershifted species are marked on the right by arrows. Note the disappearance of the supershifted bands when pol mu ⌬BRCT was used (lane 5). In lane 6, half the amount of pol mu was used (10 nM). (B) Comparisons between pol mu, pol lambda, and a ⌬BRCT mutant of pol lambda. Pol lambda also resulted in a supershift of the Ku:DNA complex (lane 7), whereas the ⌬BRCT mutant pol lambda did not (lane 8). The supershifted species are shown on the right with brackets. In each reaction, a 5 nM 60 bp DNA probe was incubated with the protein(s) indicated above the lane for 30 min on ice. Protein concentrations: Ku, 0.5 nM; pol mu, mu ⌬BRCT (⌬B), lambda, or lambda ⌬BRCT, 20 nM. Free probe was run to the bottom of the gel.

ures S1–S3 available online at http://www.molecule.org/ cgi/content/full/16/5/701/DC1/) for binding to DNA alone and binding to Ku:DNA complexes by electrophoretic mobility shift assay (EMSA) with an end-labeled 60 bp dsDNA oligonucleotide. Pol mu alone shifts approximately 1% of the 60 bp dsDNA to yield the diffuse band in the middle of the gel (Figure 1A, lane 3). Pol mu contains a BRCA1 C-terminal (BRCT) domain, and such domains have been suggested to be important in protein:protein and protein:DNA interactions (Callebaut and Mornon, 1997). We generated a deletion mutant of pol mu that is missing the BRCT domain (pol mu ⌬BRCT) but which retains full polymerase activity (Supplemental Figure S2). Pol mu

⌬BRCT does not shift the DNA (Figure 1A, lane 2), indicating that the BRCT domain is required for this interaction and confirming that the band is indeed due to the pol mu:DNA complex formation. Ku alone shifts the 60 bp dsDNA to form the one-Ku:DNA and two-Ku:DNA complexes, as expected (Figure 1A, lane 4). In the presence of pol mu, both the one-Ku and two-Ku:DNA complexes are supershifted (Figure 1A, lanes 6 and 7; Figure 1B, lane 6). Whereas pol mu alone shifts only 1% of the 60 bp dsDNA, pol mu shifts 35% of the one-Ku:DNA complex (Figure 1A). These results indicate that the binding of pol mu to Ku:DNA is much stronger than the binding of pol mu to DNA alone. To examine this further, we used a linear DNA frag-

Biochemical Reconstitution of Human NHEJ 703

ment that is 18 bp long (rather than the 60 bp above), a length where Ku fully occupies the DNA molecule and prevents access by other proteins (de Vries et al., 1989; Walker et al., 2001; West et al., 1998). We found no binding of pol mu to the Ku:DNA (18 bp) complexes (Supplemental Figure S4). Hence, pol mu interaction with Ku:DNA complexes requires the contact of pol mu to both Ku and free DNA simultaneously; either contact alone is insufficient to give the large amount of shift seen for the pol mu:Ku:DNA ternary complex (Figure 1A, lanes 6 and 7). When pol mu ⌬BRCT was studied, we did not detect any supershift of the Ku:DNA complexes (Figure 1A, lane 5). Because the pol mu ⌬BRCT has little or no DNA binding activity (Figure 1A, lane 2), it is not clear whether the BRCT domain of pol mu is critical for the interaction with DNA alone or both Ku and DNA when forming the pol mu:Ku:DNA ternary complex. When we examined pol lambda and a pol lambda ⌬BRCT mutant, we found a similar set of interactions as for pol mu:DNA (Figure 1B). The binding of pol lambda to DNA alone is barely detectable as a smear of retarded DNA species but is clearly not as distinct as that of pol mu (Figure 1B, compare lanes 2 and 3). The ⌬BRCT form of pol lambda (which also retains full polymerase activity [Shimazaki et al., 2002]) does not appear to bind DNA at all (Figure 1B, lane 4). Pol lambda binds the Ku:DNA complex (Figure 1B, compare lanes 5 and 7), and this interaction is fully dependent on the BRCT domain of pol lambda (Figure 1B, compare lanes 7 and 8), confirming that pol lambda is responsible for the shift. As further confirmation, polyclonal antibodies raised against pol lambda shift the pol lambda:Ku:DNA complexes into the wells without affecting the Ku:DNA complexes (Supplemental Figure S5). The two remaining pol X family members, pol beta and TdT, were also tested in similar EMSA experiments. Pol beta has no BRCT domain, and, as expected, it does not bind to the Ku:DNA complex (Supplemental Figure S6). However, TdT does have a BRCT domain, and yet it does not bind to the Ku:DNA complex (Supplemental Figure S6). This may reflect the substantial sequence variation of the BRCT domain (Callebaut and Mornon, 1997), which may also translate to a substantial variation in function. The BRCT domain of TdT is not essential for its polymerase activity in vivo (Repasky et al., 2004) or in vitro (Chang et al., 1988). Previous work had shown that pol mu binds to the Ku:XRCC4:DNA ligase IV:DNA complex (Mahajan et al., 2002), though that study did not find the simpler interaction of pol lambda and pol mu with Ku:DNA that is described here. We also tested if pol mu, pol lambda, pol beta, or TdT binds to the Ku:XRCC4:DNA ligase IV:DNA complexes. By using EMSA and immunoprecipitation of individual components in the presence and absence of DNA, we find that the Ku:DNA complex can simultaneously bind to XRCC4:DNA ligase IV and pol mu, pol lambda, or TdT (data not shown). Furthermore, we find that the BRCT domains of pol mu and lambda are necessary for the formation of the pol mu:Ku:XRCC4:DNA ligase IV:DNA complex and the pol lambda:Ku:XRCC4: DNA ligase IV:DNA complex. TdT may be recruited by the XRCC4:DNA ligase IV complex rather than by Ku:DNA (Mahajan et al., 2002). Pol beta does not form a

complex with Ku:XRCC4:DNA ligase IV:DNA, consistent with the fact that it does not bind to Ku:DNA (Supplemental Figure S6, lane 6). In summary, pol mu and pol lambda interact with the Ku:DNA complex, and these interactions are dependent on the BRCT domains of these polymerases, whereas TdT:Ku:DNA and pol beta:Ku:DNA complexes are not detectable under the same conditions. The Core NHEJ Factors in an In Vitro NHEJ System We previously showed that compatible and blunt DNA ends could be joined by using only XRCC4:DNA ligase IV (Grawunder et al., 1997; Wilson et al., 1997). Here, we were interested in the joining of incompatible DNA ends. To study this, we examined the joining of linear dsDNA substrates with a set of purified proteins (Supplemental Figure S1) that we refer to as the core components (Ku, Artemis:DNA-PKcs, and XRCC4:DNA ligase IV) because the roles of these proteins have been documented genetically. We ruled out any nuclease contamination in all components except Artemis, which is itself a nuclease (Supplemental Figure S3). For Artemis, we showed that a point mutant purified in the same manner has lost all of its endonucleolytic activity (Ma et al., 2002). The NHEJ joining reactions were done with only one species of linear dsDNA oligonucleotide present in the system, and we assayed for the joining of any two molecules. This substrate is composed of a 73 nt and a 90 nt oligonucleotide strand, which upon annealing, yields a 73 bp dsDNA with a 7 nt 5⬘ overhang and a 10 nt 3⬘ overhang (Figure 2A). There are three possible joining orientations for the two ends: joining of two different ends (head-to-tail) or the same ends (head-to-head or tail-to-tail) of two independent molecules of the substrate. The substrate was 5⬘ phosphorylated with unlabeled ATP on both strands and incubated with proteins (Figure 2A). The resulting products were then subjected to PCR analysis with one of the PCR primers radioactively labeled at the 5⬘ end. The PCR reactions were then resolved by denaturing polyacrylamide gel electrophoresis (PAGE), and subsequently the sequences of the PCR products (containing junctions of the ends of the substrate) were determined (see Experimental Procedures and Figure 2A). Including the length of the primers, the range of the joining products that can be detected is 40–81 nt with one pair of primers (YM-167 and 168 for Figures 2B and 2D) and 36–89 nts with the other pair (YM-184 and 185 for Figures 2C and 2E). We were unable to detect head-to-head or tail-to-tail joining. Rather, the only PCR products identified resulted from the joining of different ends (head-to-tail). This is presumably because any joining events that arise from the same ends (e.g., head-to-head) will form long, stable hairpin structures and will not be efficiently amplified in a PCR reaction. PCR of the control joining reaction (without any protein treatment in the joining step) yields a minimal background due to random annealing between the amplified ends (Figure 2B, lane 2). We first used the core NHEJ proteins (Ku, DNA-PKcs, Artemis, and XRCC4:DNA ligase IV) in the joining reactions (Figure 2B). In the absence of Ku, the joining of the two ends yields a similar pattern of products as the reaction with all of the core proteins (Figure 2B, compare

Molecular Cell 704

Figure 2. In Vitro NHEJ between Double-Stranded DNA Molecules with Incompatible Ends (A) Schematic diagram of the in vitro NHEJ assay. A dsDNA oligonucleotide was mixed with NHEJ proteins to allow end joining between the incompatible ends of two molecules of the duplex substrate. After the end joining step, the junctions of products were amplified with 35 cycles of PCR with one of the primers 5⬘ end labeled. The resulting PCR products were resolved by denaturing PAGE. For some NHEJ assays (e.g., Figure 3B, top two groups of sequences), the primary PCR products (radioactive) were cloned directly into the vector (bulk cloning). In other cases, PCR products of desired sizes were excised from the denaturing gel, eluted out of the polyacrylamide gel, reamplified by a second round of PCR, and then cloned. The sequence of individual molecular junctions was determined as described in the Experimental Procedures. (B) DNA end joining with the core components (Ku, Artemis:DNA-PKcs, and XRCC4:DNA ligase IV) and DNA pol X family polymerases pol lambda and TdT. The substrate was incubated with different protein complexes as indicated. The products were subjected to a PCR analysis with primers YM-167 and YM-168, and the resulting PCR products were resolved by denaturing PAGE. A PCR control reaction (with ddH2O as template) is in lane 1. Proteins were added in the following amounts: Ku, 2.5 pmol; Artemis, 2.5 pmol; DNA-PKcs, 1.25 pmol; XRCC4:DNA ligase IV, 1 pmol; DNA pol lambda, 1 pmol; and TdT, 1 pmol. In lane 10, a 10 nt single-stranded DNA oligonucleotide, HJ30, was added to

Biochemical Reconstitution of Human NHEJ 705

lanes 3 and 7). The absence of the nuclease components, DNA-PKcs or Artemis, or the use of a catalytic mutant of Artemis results in a higher proportion of longer products (Figure 2B, compare lanes 4 and 5 with 7; Supplemental Figure S7), and this is consistent with the nucleolytic role of the Artemis:DNA-PKcs complex in overhang processing (Ma et al., 2002). Without the XRCC4:DNA ligase IV complex in the joining reaction, joining is reduced to background levels (Figure 2B, compare lanes 6 and 7), underscoring the crucial role of XRCC4:DNA ligase IV in the ligation step of NHEJ. DNA ligases I and III are unable to substitute for the XRCC4:DNA ligase IV complex (Figure 2E and Supplemental Figure S8). When the components are all present, the joining products distribute in the predicted range of 40–81 nts (Figure 2B, bands in lane 7). In summary, the core components, Ku, Artemis:DNAPKcs, and XRCC4:DNA ligase IV, are able to join incompatible DNA ends. The XRCC4:DNA ligase IV complex is critical for this joining. The lack of the Artemis:DNAPKcs complex results in longer products, as expected. The Roles of Pol X Family DNA Polymerases Because of the physical interaction between Ku and the pol X family members (Figure 1 and Supplemental Figure S5), we tested pol X family DNA polymerases for effects in the in vitro DNA end joining system (see Supplemental Figure S2 for polymerase activity). When the NHEJ core component proteins are supplemented with DNA pol lambda, products of higher molecular weight (MW), defined as ⬎81 nt (or the size of the full-length product), are also detected (Figure 2B, lane 8). This is presumably because the addition of pol lambda preserves at least part of the 5⬘ overhangs as well as contibutes to junctional addition (see below and Figure 3). TdT, a lymphoid-specific and template-independent pol X family member, adds nucleotides nearly randomly to the coding ends in V(D)J recombination (Chang and Bollum, 1986; Komori et al., 1993). Therefore, we also tested TdT in our joining reactions. The addition of TdT further increases the abundance of the high MW products (Figure 2B, lane 11), indicating that TdT and pol lambda work synergistically. The addition of pol beta to the core protein components does not increase the amount of the high MW

products (Figure 2C, compare lanes 1 and 2). However, the addition of pol mu (Figure 2C, lane 3) and pol lambda (Figure 2C, lane 4) alone or the two polymerases together (Figure 2C, lane 5) do result in the generation of the high MW products. In the absence of TdT, the pattern of the high MW bands in the presence of both pol mu and pol lambda is very similar to the one that arises when only pol mu is added (Figure 2C, compare lanes 3 and 5 and compare lanes 4 and 5). When TdT is added to the joining reaction along with the other pol X members, the production of the high MW products is slightly enhanced. When TdT is present in the reaction, the pattern of the high MW products from pol mu and pol lambda is similar to when only pol lambda (along with TdT) is used (Figure 2C, compare lanes 7 and 9 and lanes 8 and 9). In summary, in the absence of TdT, the pol mu plus pol lambda reactions yield a pattern very similar to pol mu alone, whereas in the presence of TdT, pol lambda determines the pattern of DNA end modification as if pol mu were not present. The Effect of Ku in the Presence of DNA Polymerases One role of Ku has been hypothesized to be in the synapsis of the two broken DNA ends (Ramsden and Gellert, 1998). However, biochemical and microscopy evidence has not confirmed this (DeFazio et al., 2002; Yaneva et al., 1997). Consistent with a lack of any synaptic role, we do not detect a role of Ku when only the core components are used (Figure 2B, compare lanes 3 and 7). However, when DNA polymerases mu, lambda, and TdT are present in the joining reactions, the presence of Ku significantly stimulates the formation of the high MW products (Figure 2C, compare lanes 9 and 10). We also examined the role of Ku in the presence of a single DNA polymerase, pol lambda, in the joining assay (Figure 2D). Again, the presence of Ku enhances the abundance of the high MW products (Figure 2D, compare lanes 2 and 6). The addition of TdT increases the amount of the high MW bands even further (Figure 2D, lane 7) as described above. Interestingly, the ⌬BRCT form of pol lambda is much reduced in its ability to contribute to the high MW products in the presence of Ku (Figure 2D, lane 8) to an extent that is similar to when Ku is absent (Figure 2D, lane 2). This is consistent with

test for any effect. The designation “full-length product (81 nt)” refers to the theoretical product formed by direct ligation of the two ends without any loss from either overhang and without any addition. The designation “shortest product (40 nt)” refers to the sum of the length of the two PCR primers, which is the theoretically shortest product. Products with both overhangs resected (but not any further resection) would be the length of the full-length product (81 nt) minus the length of the overhangs (7 nt and 10 nt) or 64 nt here (but 72 nt in Figure 2C). The designation “high MW products” refers to ends longer than the full-length product. (C) The effect of different DNA pol X family members on in vitro NHEJ. The substrate was incubated with the core NHEJ components and different combinations of pol X polymerases as indicated. The gel image shows the PCR products (from primers YM-184 and YM-185) of the joined DNA oligonucleotide substrate. Protein amounts are as follows: Ku, 2 pmol; Artemis, 4 pmol; DNA-PKcs, 1 pmol; XRCC4:DNA ligase IV, 1 pmol; DNA pol mu, 1 pmol; DNA pol beta, 0.5 pmol; pol lambda, 0.5 pmol; and TdT, 0.5 pmol. (D) The effect of Ku on in vitro NHEJ assay in the presence of pol lambda and the role of the BRCT domain of pol lambda. The ⌬BRCT mutant protein is indicated in lane 8. The amount of each protein added in the reactions is the same as in Figure 2B except TdT, for which 0.1 pmol was used. The assay was carried out as described in (B), and the gel image of the PCR products is shown. The position of the high molecular weight products is indicated on the right with a bracket. The labeled free PCR primer and the primer extension products are not shown. A 50 bp ladder was labeled with T4 PNK and [␥-32P]ATP and is indicated by M. Sizes of the marker bands (in nts) are indicated on the left. (E) DNA ligase I and III do not substitute for XRCC4:DNA ligase IV. The gel image shows the PCR results of end-joining products resulting from the enzymatic activities of the indicated proteins. The PCR reaction with H2O as template (negative control) is shown in lane 1. ⌬BRCT indicates the pol mu mutant lacking the BRCT domain. The sizes of the DNA standards (lane M) are labeled on the left. The position of the high molecular weight products is marked on the right.

Molecular Cell 706

Figure 3. Sequences of the Junctions Formed from Incompatible DNA Ends in Intermolecular NHEJ Reactions (A) DNA sequences from joining reactions with only the core NHEJ components. The sequences resulted from DNA eluted from lane 7 of the gel for Figure 2B (see Experimental Procedures). Dotted lines mark the edges of the double stranded portion of the substrate. (B) DNA sequences from joining reactions with the core components and indicated pol X family DNA polymerases. Only the first two sets of reactions (the core components plus pol lambda or pol mu only) were derived from reactions equivalent to lanes 3 and 4 of Figure 2C by cloning in bulk (see Figure 2A). The other sets of sequences were derived from excised bands from lanes of the gel for Figure 2C. Numbers to the left of each set of sequences correspond to lane numbers in Figure 2C. The name of the pol X polymerases (in addition to the core NHEJ components) used for each set

Biochemical Reconstitution of Human NHEJ 707

the fact that the ⌬BRCT pol lambda has lost its ability to form the pol lambda:Ku:DNA complex (Figure 1B). Interestingly, the ⌬BRCT mutant of pol mu contributed to high MW products with smaller sizes compared to those from wild-type (wt) pol mu (Figure 2E, compare lanes 9 and 10). When TdT was absent, pol mu and pol lambda together still failed to form high MW products efficiently without Ku (Figure 2E, compare lanes 4 and 8). These junctional effects are an important functional confirmation of the binding studies described earlier (Figure 1). Sequence Analysis of Joined Products from Reactions Containing the Core NHEJ Components Sequences of junctions are derived from the cloning of bulk reaction products (equivalent to one entire lane in Figures 2B–2D) and from cutting out specific regions from a given lane (see Experimental Procedures). The results from the sum of the regions within a lane correspond well to the bulk cloning of the entire lane. After the excision of a specific region of a lane, a secondary PCR step was introduced to further amplify molecules eluted out of gel segments prior to the cloning step (Figure 2A). Hence, the number of specific sequences does not necessarily reflect the actual abundance of a junction after the joining reaction. Such information is best discerned from the gel profiles themselves (Figures 2B–2D). For comparisons between different lanes, equivalent regions were taken (e.g., high MW region) with the same boundaries. Sequences of junctions attributable to the four NHEJ core components show several interesting features (Figure 3A contains sequences from Figure 2B, lane 7). We find that the 5⬘ overhang is resected more often than the 3⬘ overhang (Figure 3A). This is consistent with the known difference in cleavage patterns at 5⬘ and 3⬘ overhangs by the Artemis:DNA-PKcs nuclease complex: at 5⬘ overhangs, the Artemis:DNA-PKcs complex preferentially cleaves at the junction of the single- and doublestranded DNA to yield a blunt end, and at 3⬘ overhangs, Artemis:DNA-PKcs tends to leave 4–5 nt of the 3⬘ overhang attached to the dsDNA portion (Ma et al., 2002). Consistent with NHEJ events observed in vivo, some of the events also show the use of terminal microhomology (Lieber et al., 2004). Interestingly, some of the junctions have part of the 3⬘ overhangs retained. These can not be preserved by direct fill-in reactions by DNA polymerases because this would violate the catalytic polarity of all DNA polymerases (5⬘ to 3⬘). Furthermore, no polymerase was added to this reaction (Figure 2B, lane 7) and the possibility of contaminant polymerase activities in Ku, DNA-PKcs, and XRCC4:DNA ligase IV was ruled out in a polymerase

assay (Supplemental Figure S2, lane 5). Some of these 3⬘ overhangs (the ones without asterisks) may be involved in the partial annealing by microhomology with the other DNA end. After the trimming of the potential flaps, at least one of the strands may be ligated and, thus, the 3⬘ overhangs appear in the final junctions. Though this is a plausible mechanism for these joining events that preserve 3⬘ overhangs, a small subset of these junctions that lack junctional microhomology (the ones with asterisks) cannot be explained by this mechanism, and we propose a model below that provides possible insight. Sequence Analysis of Joined Products from Reactions Containing the Core NHEJ Components and Pol X Polymerases Sequences of junctions joined by the core components plus the pol X polymerases are quite distinct from those generated by only the core components (Figure 3B shows sequences from the reactions in Figure 2C). For reactions containing pol mu or pol lambda alone, junctional sequences show some additions, consistent with the known TdT-like activity of both pol mu (Dominguez et al., 2000) and pol lambda (Ramadan et al., 2003). These results are distinct from reactions with no polymerase, where no additions are seen (Figure 3A). We have focused our attention here on the more physiologic cases where both pol mu and pol lambda are present because both are distributed ubiquitously in mammalian cells. The pattern of joining in the presence of pol mu plus pol lambda is similar to pol mu alone (in the absence of TdT) (Figure 2C, lanes 3–5). When we examine the sequences from reactions containing both pol mu and pol lambda, the junctional additions for a subset of sequences are longer than those for pol mu or pol lambda alone (Figure 3B). This illustrates that although the bulk level of addition is similar for pol mu plus pol lambda relative to pol mu alone, the TdT-like activities of these two polymerases do not compete with one another in this NHEJ reconstitution system but, in fact, can result in junctions with longer additions than those generated by each polymerase individually. Therefore, pol mu and pol lambda appear to be capable of acting on the same junction during the course of a single NHEJ reaction. Whether this occurs concurrently (such as with one polymerase acting at each of the two DNA ends) or sequentially cannot be determined here. As noted earlier, the joining pattern when pol mu, pol lambda, and TdT are all present is more similar to that of pol lambda plus TdT than to that of pol mu plus TdT (Figure 2C, compare lanes 7–9). The distribution of the length of the sequences agrees with the observations from the gel. The junctional additions by pol mu plus TdT have a wider distribution in length than that from

of sequences (or lanes of Figure 2C) are indicated on the left. The last set of sequences was derived from a joining reaction where Ku was omitted (⫺Ku). The sequence of the end with the 5⬘ overhang and the end with the 3⬘ overhang of the substrate is shown at the top. Only the top strand of each NHEJ junctional sequence is shown. The continuous sequence of each PCR product was divided into the left half, the right half, and the junctional additions in between. Capital letters in the sequences indicate sequences originating from the dsDNA portion of the substrate, and lower letters indicate sequences from the overhangs. Bold italic letters represent nts that can be assigned either to the left end or the right end and were arbitrarily assigned to the left (microhomology). The length of each PCR product and the number of clones sequenced are indicated on the right side of the figure. An internal deletion is indicated by underlined dashed lines in one of the sequences.

Molecular Cell 708

Figure 4. Estimation of the In Vitro NHEJ Joining Efficiency with Various Amounts of Double-Stranded Oligonucleotides that Resemble NHEJ Products (A) As one method for estimating the efficiency of the in vitro NHEJ reactions, we did PCR reactions amplifying known amounts of “joined” product (synthesized oligonucleotides that mimic the NHEJ product with overhangs fully preserved for lanes 9–13) with the same amount of NHEJ substrate as was used in the standard NHEJ assays (Figure 2). (B) The titrations used are 1 molecule of joined product per 100 (lane 9), 1000 (lane 10), 104 (lane 11), 105 (lane 12), and 106 (lane 13) molecules of substrate. (Corresponding dilutions of a shorter joined product that resembles NHEJ product with overhangs fully resected are shown in lanes 4–8.) We compared the amount of PCR products from these reactions (lanes 9–13) with that of the full NHEJ system (Ku ⫹ Artemis:DNA-PKcs ⫹ XRCC4:DNA ligase IV ⫹ pol mu ⫹ pol lambda ⫹ TdT) products (lane 3). The intensity of the joining products (after PCR) in lane 3 falls between the intensity of the positive PCR control reactions originating from 1 joined product per 1000 (which is the same as 10 per 10,000) and 1 per 100 (or 100 per 10,000) pairs of substrate ends (compare lane 3 with lanes 9 and 10). When one-tenth of the NHEJ assay products was used in the PCR reaction (lane 2), then the estimates yield the corresponding results. Therefore, we estimate the efficiency of our in vitro complete NHEJ reactions to be between 10 and 100 joining products per 10,000 pairs of DNA ends (or between 0.1–1%). A PCR reaction with water as the template (negative control for PCR) is shown in lane 1. Sizes of DNA standards (in nts) are indicated on the left.

pol lambda plus TdT or pol mu plus pol lambda plus TdT. The latter two combinations generate similar lengths of sequences. When Ku is omitted (Figure 2C, lane 10), we observe some distinct changes. First, we detect long runs of homopurines and homopyrimidines (Figure 3B, last set of sequences). This is similar to what is observed for TdT-dependent junctional additions formed within human cells (Gauss and Lieber, 1996). In addition, we see a clear preference for G incorporation as evidenced by the fact that the junctions are quite GC rich. The apparently dominant participation of TdT here relative to the complete reaction (Ku, Artemis:DNA-PKcs, XRCC4:DNA ligase IV, polymerases mu and lambda, and TdT) suggests that Ku is not required for TdT loading at the DNA ends. Instead, because Ku associates with pol mu or pol lambda when bound to DNA (Figure 1), the absence of Ku seemed to allow TdT to dominate junctional addition (Figure 3B, last set of sequences). Estimates for the efficiency of end joining in this system (including the pol X polymerases) can be done by PCR reconstruction experiments (Figure 4), by calculations based on the yield of unique clones (described in the Supplemental Results), or by determining upper limits based on direct gel assays of joining of endlabeled NHEJ substrates (Supplemental Figure S9). All

three methods are consistent with a joining efficiency of between 10 and 100 joining events per 10,000 pairs of DNA ends, or 0.1–1%. This is reasonably efficient given that the reaction represents intermolecular joining of incompatible DNA ends. We also examined the joining of four other pairs of incompatible ends by using this in vitro reconstitution system (Figure 5). These included long (5–10 nt) and short (2–4 nt) 5⬘ and 3⬘ overhangs in various combinations. As with the pair of ends described in detail above, the core components (Ku, Artemis:DNA-PKcs, and XRCC4:DNA ligase IV) yield some joining of these incompatible ends. Depending on the type of ends (overhang and length), the amount of joining and length of the products can be much greater with pol mu, pol lambda, and TdT present. This is consistent with what was observed in the more detailed analysis above. The effect of the polymerases on junctional length is more prominent for long overhangs than short overhangs, as expected (Figure 5). Independent Ligation of the Two Strands at a dsDNA Junction by the XRCC4:DNA Ligase IV Complex Some of the junctions that form in vivo and in our purified system in vitro have no microhomology. Such joinings occur in vivo quite often, but the nature of the end align-

Biochemical Reconstitution of Human NHEJ 709

Figure 5. End Joining Reactions for Other DNA End Combinations The ends of the standard NHEJ substrate YM46/YM-159 were modified to generate different configurations. The joining of these ends (on a single duplex substrate) by the specified proteins/protein complexes was analyzed as described in the text. The substrate used for lanes 4–6 (5⬘–3⬘ L) is YM-46/YM-159. L indicates long (7 and 10 nt) incompatible overhangs; S indicates short (2 and 4 nt) incompatible overhangs. The 5⬘ or 3⬘ nature of the overhangs is indicated.

ment (if not blunt) has been uncertain (Pfeiffer et al., 2000). A few of the in vitro events are ones where part of the 3⬘ overhang is preserved (e.g., Figure 3A, the two sequences with asterisks). A 3⬘ overhang cannot be “filled in” by any DNA polymerase. Moreover, no DNA polymerases were added to this reaction set (Figure 3A). How could a 3⬘ overhang be ligated to another strand given that there is no opportunity for the antiparallel strand to provide alignment? Does this mean that this strand is ligated in some form of single-strand ligation? In order to test whether the XRCC4:DNA ligase IV complex could ligate one strand without ligating the other, we designed the following series of experiments. As a positive control, we created an oligonucleotide substrate with compatible dsDNA ends (AAAA/TTTT) (Figure 6A). For comparison with XRCC4:DNA ligase IV, we used T4 DNA ligase, which shows intramolecular and intermolecular ligation activity to generate circular and multimeric products, respectively (Figure 6A). The identity of the circular products was confirmed by Exo V digestion (Figure 6C). Compared to T4 DNA ligase, XRCC4:DNA ligase IV forms circles and tandem multimers quite efficiently when an equivalent amount of protein is tested (Figure 6A, compare lanes 2 and 5). In order to make one strand unligatable while monitoring ligation of the other strand, we created a series of molecules that had gaps of 1, 2, or 3 nt on the top strand (and 3, 2, or 1 bp of annealing between the two strands) (Figure 6A, top). Despite gaps of 1 and 2 nt on the top strand, the bottom (long circular) strand is ligated efficiently (Figure 6A, lanes 9–12 and lanes 15–18), and the efficiency of ligation decreases with increasing gap size. Even T4 DNA ligase is able to carry out this ligation, though its activity on nicks with an opposing 2 nt gap is relatively weak (Figure 6A, lane 14). Neither T4 nor the XRCC4:DNA ligase IV complex can ligate nicks opposite of a strand with a 3 nt gap and only 1 bp of potential annealing between the two strands. Ku has very little or no stimulatory effect under these conditions, consistent with two previous studies (Chen et al., 2000; Kysela et al., 2003). In these experiments, only the long (circular) bottom strand has a 5⬘ phosphate (acquired upon radioactive labeling of that strand); in contrast, the top strand has a 5⬘OH and, hence, is unligatable even though it can participate in terminal alignment. Interest-

ingly, the efficiency of intermolecular ligation (which forms multimers) decreases less rapidly with increasing gap size than does the efficiency of intramolecular ligation (which forms circular products) (Figure 6A, compare lanes 3, 9, and 15, or 5, 11, and 17). This may reflect the steric strain in the generation of such short circles. These results show that the XRCC4:DNA ligase IV complex is able to ligate one strand, whereas the other strand is not only unligatable (because of the 5⬘OH) but also is not even configured in the form of a nick (on the top strand). The lack of ligation when the gap is 3 nt long reflects the lack of sufficient hydrogen bonding needed to align the two DNA ends. It appears that at least two AT base pairs (four hydrogen bonds) are required under these conditions to align the two ends for efficient ligation. We next tested the effect of varying the gap size when the hydrogen bonding between the two ends was maintained constant at 2 AT base pairs (Figure 6B, lanes 1–9). When the gap size is 2 or 3 nt, tandem molecules and circles are readily detectable. However, when the gap size is 4 nt, the amount of circular and tandem products are both reduced. Therefore, the contacts between the DNA and the XRCC4:DNA ligase IV complex are sensitive to the double strandedness in this portion of the junction. Putative dimers of XRCC4 in the ligase complex may only recognize dsDNA (Modesti et al., 2003), and the distance between the dimers may have a role in the efficiency of ligation. The ligation of a nick opposite of a 2 or 3 nt gap by T4 DNA ligase is much less efficient (Figure 6B, lanes 2 and 5) compared to the ligation by XRCC4:DNA ligase IV, suggesting that the XRCC4:DNA ligase IV complex has evolved to tolerate substantial gaps at an unstable pair of ends. Given that four hydrogen bonds (two AT base pairs) are adequate for alignment, we were interested in whether three hydrogen bonds would be sufficient by replacing the two AT base pairs with one GC base pair (Figure 6B, lanes 10–12). None of these reactions generated any ligated products, indicating that three hydrogen bonds are inadequate. We also tested five hydrogen bonds by using GA/CT terminal microhomology (Figure 6B, lanes 13–15), and this resulted in more efficient ligation than did four hydrogen bonds (Figure 6B, compare lanes 3 and 15). Hence, four hydrogen bonds is the

Molecular Cell 710

Figure 6. XRCC4:DNA Ligase IV Can Ligate Only One Strand of Two Partially Annealed DNA Ends (A) XRCC4:DNA ligase IV can ligate the single nick generated by annealing two dsDNA ends by 2 bp microhomology. DNA substrates with different lengths of base pairing between the two ends were incubated with T4 DNA ligase or XRCC4:DNA ligase IV, and products were resolved by 5% denaturing PAGE. Proteins added in the reactions were as follows: Ku, 0.25 pmol; XRCC4:DNA ligase IV, 0.25 and 1 pmol for “⫹” and “⫹⫹” marked lanes, respectively; and T4 DNA ligase, 1 pmol. (B) XRCC4:DNA ligase IV can ligate the single nick created by annealing two dsDNA ends by 2 bp microhomology that generates a 4 nt gap and the nick created by annealing two ds DNA ends by 4 bp microhomology that generates a 2 nt flap. 1 pmol of T4 DNA ligase or XRCC4:DNA ligase IV were used in each reaction. The assay was done as described in (A). Dotted lines on the continuous gel images separate reactions with different substrates. Relevant nucleotide sequences around the double-strand breaks are shown on the diagrams of the substrates. The asterisk on the long oligonucleotide indicates labeling at the 5⬘ end. 50 bp ladder was labeled by T4 PNK and [␥-32P]ATP and is designated with M at the top of each gel. Sizes of the marker bands (in nts) are indicated on the left. DNA substrates were made by annealing two short (18–20 nts) oligonucleotides to a 5⬘ 32P-labeled long strand (95 nt). The oligonucleotides used in the figure are specified in the Supplemental Experimental Procedures. (C) Confirmation of the identity of the circular product by Exo V digestion. Atter the joining of cohesive ends by T4 DNA ligase (lane 3), the products were incubated with 11 units of Exo V at 37⬚C for 1 hr. Only the circular products (monomer and dimer) are resistant to the Exo V digestion.

minimal requirement in this in vitro test for single-strand ligation by the XRCC4:DNA ligase IV complex, and more hydrogen bonds improve the joining. One independent test of the single-strand ligation by XRCC4:DNA ligase IV utilizes flap substrates (Figure 6B, lanes 16–24). 1 and 2 nt flaps on the top strand do not diminish the ligation of the bottom strand (Figure 6B, lanes 21 and 24), providing yet another indication that one strand can be ligated without a ligatable nick on the other strand at a double-strand DNA break. Discussion The Order of Nucleolytic, Polymerization, and Ligation Steps of Mammalian NHEJ is Flexible The joining events by the core components (Ku, Artemis:DNA-PKcs, and XRCC4:DNA ligase IV) of this in vitro NHEJ system are diverse. The core system is entirely dependent on the ligase complex and is affected by the Artemis:DNA-PKcs complex. The formation of some junctions does not require microhomology at all, consis-

tent with in vivo NHEJ joinings (Gerstein and Lieber, 1993). A few of these can not have occurred by blunt end ligation. Without a polymerase, TdT-like activity also can not be invoked. The most reasonable explanation for some of the events is ligation of one strand followed by nucleolytic processing and ligation of the other strand (Figure 7 and see below). This means that the order of the key enzymatic steps can vary not only from one junction to another but also from one strand to the other in the same junction. A second major finding from the in vitro NHEJ system is that Ku has a functional role in affecting the junctions when pol mu or pol lambda is present, and this Ku effect is dependent on the BRCT domain of the polymerases (Figure 2D). In vivo, the absence of Ku results in a less severe phenotype than that of the ligase IV null (Ferguson and Alt, 2001). This in vivo difference is reflected extremely well in the in vitro system described here, where the reliance on the ligase complex is absolute. A third major finding of this study is the cooperative and competitive interactions between pol mu, lambda,

Biochemical Reconstitution of Human NHEJ 711

Figure 7. Iterative Processing Model for NHEJ Independent ligation of the two strands increases the number of ways in which the joining can occur in NHEJ. Ligation on one strand can precede nucleolytic or polymerization action on the other strand, resulting in the iterative processing model for the joining of a DSB by the NHEJ pathway. This model is in contrast to previous models in which ligation of the two nicks by XRCC4:DNA ligase IV concludes the NHEJ pathway.

and TdT (Figures 2B–2E). Pol mu and pol lambda are present in all mammalian cells; in early lymphoid cells, TdT is also present. Our findings agree well with the known features of the corresponding null mice (see below). Given the flexibility of the order of the three key enzymatic steps (nucleolytic action, polymerization, and ligation) in NHEJ, one can schematize the many variations of the steps for the two strands at any junction (Figure 7 and Supplemental Figure S10). Single-Strand Ligation by XRCC4:DNA Ligase IV, Ligation on One Strand Can Precede Nucleolytic Processing on the Other Strand Given that ligation can occur on one strand despite an unligatable configuration on the other, nucleolytic processing on the unligatable strand would be required in some cases before it could be ligated. We know that the Artemis:DNA-PKcs complex is able to cleave flaps (Y.M. and M.R.L. unpublished data). The ability to ligate one strand before nucleolytic processing on the other may explain joining events where there is no apparent terminal microhomology. Some of these events may simply be blunt end ligations. Others could be cases where a polymerase functions in a TdT-like manner and creates terminal microhomology. (If the added nucleotides are, by chance, complementary to nucleotides at the other dsDNA end, then the utilization of microhomology would be impossible to document by simple inspection of the DNA sequence; it would simply appear that the two ends were joined without microhomology.) However, some joinings (two asterisked ones from Figure 3A and one not shown) occur within a 3⬘ overhang under circumstances where there was no microhomology. Additionally, no polymerase was added to these reactions. Ligation of one strand prior to the processing of the other strand could conceivably explain these events in the following way. After ligation of the top strand (Supplemental Figure S11), nucleolytic processing of the other strand might then result in a large gap. Secondary structure formation (stem and loop) on the top strand might then permit the configuration of a ligatable nick on the bottom strand (Supplemental Figure S11). The two resulting strands would have a region of mismatch, which if not repaired by mismatch repair would, upon DNA replication in the cell or during PCR in the assay, yield two different daughter molecules. Ligation of the 5⬘ and 3⬘ overhangs to one another might generate the

long products seen when no polymerase is present (Figure 2B, lanes 4 and 5). Ku as a Recruitment Protein for the Nucleolytic, Polymerization, and Ligation Steps of NHEJ: Physical and Functional Interaction between Pol Mu or Pol Lambda and Ku:DNA Our studies indicate a significant role of Ku in both the physical and functional recruitment of pol mu and pol lambda. Previous work demonstrated the physical recruitment by Ku of the DNA-PKcs component of the nuclease complex (Gottlieb and Jackson, 1993; Hammarsten and Chu, 1998; West et al., 1998; Yaneva et al., 1997) and of the ligase complex (Chen et al., 2000; Nick McElhinny et al., 2000). These two Ku-ligase recruitment studies described conflicting data about how this interaction affects the ligation efficiency. Our reconstitution studies are relevant to this. Ku appears to have only a small effect in the reaction that contains only the core components (Figure 2B), a fact that is inconsistent with it being a synapsis factor (Ramsden and Gellert, 1998). In contrast, Ku has a more substantial effect on the formation of high MW products when pol mu and pol lambda are present (Figures 2C and 2D). These functional data are consistent with Ku recruiting pol mu and pol lambda, as well as the ligase complex. This is a functional role (rather than merely a physical DNA binding role) for Ku in eukaryotic NHEJ for the joining of incompatible DNA ends. The recruitment of the nuclease (Artemis:DNA-PKcs), the polymerases (pol mu and pol lambda), and the ligase complex (XRCC4:DNA ligase IV) by Ku:DNA complexes suggests that Ku functions independently to recruit each of the key enzyme complexes for the three key steps of NHEJ. It is not yet clear whether Ku:DNA complexes can bind all three of these components at the same time, though it can bind at least two simultaneously: Ku:DNA complexes bind to the ligase complex and to pol mu or pol lambda concurrently (Mahajan et al., 2002; data not shown). Based on the studies here, it is not necessary that the three activities (nuclease, polymerase, and ligase) be recruited to a DNA end in a certain order to achieve joining. The Role of Pol X DNA Polymerases in NHEJ Pol X family polymerases have been suspected to be involved in NHEJ for the following reasons. First, the pol X family member, Pol4, is responsible for a substantial

Molecular Cell 712

amount of junctional addition in yeast NHEJ (Tseng and Tomkinson, 2002; Wilson and Lieber, 1999). Second, genetic studies of pol mu-deficient animals suggest its role in V(D)J recombination (Bertocci et al., 2003). Third, the association of pol X family polymerases with NHEJ proteins Ku and XRCC4:DNA ligase IV suggest a relationship to NHEJ (Mahajan et al., 2002). In our studies here, we observe that pol mu and pol lambda, but not pol beta, are able to participate in the in vitro NHEJ reactions. TdT, when present, is also responsible for significant junctional addition. The interplay between pol mu, pol lambda, and TdT is interesting. In pol mu knockout mice, there is shortening of the light chain coding ends, but not of the heavy chain coding ends (Bertocci et al., 2003). TdT is known to be absent during murine light chain gene rearrangement. In pol lambda null mice, there is shortening of only the heavy chain coding joints (B. Bertocci, J.C. Weill, and C.A. Reynaud, personal communication), and TdT is present in murine pre-B cells during heavy chain gene rearrangement. These observations agree with our DNA end-joining results very well. When TdT is absent, pol mu plus pol lambda result in the same outcome as pol mu alone. In the presence of TdT, pol lambda dominates pol mu, resulting in the same outcome as pol lambda plus TdT without any pol mu (Figure 2C). The fact that pol mu knockout mice have shorter light chain coding ends also suggests that pol mu has a higher affinity to DNA ends (or to the Ku:DNA complex) than pol lambda does. The presence of pol mu in pol lambda knockout mice does not seem to correct the shortening of the heavy chain coding ends. Perhaps this is because TdT competes with pol mu for the binding to DNA ends, whereas TdT and pol lambda can function synergistically. (Such a cooperative role between TdT and pol lambda is consistent with recent yeast twohybrid system data and in vitro biochemical association studies that suggest that these two proteins can form a physical complex [N.S. and O.K., unpublished data].) The above assumptions still require further biochemical and genetic studies, but the agreement of this in vitro system with the corresponding genetic studies is quite compelling. The double pol mu/TdT and pol lambda/ TdT knockouts and the triple knockout will be interesting, based on the biochemical findings here. Pol mu and pol lambda permit a wide range of products to be observed, including retention of both overhangs. However, complete retention of both overhangs requires template-independent extension of a 3⬘OH. This might then permit some microhomology to be created between the two DNA ends. Pol mu and pol lambda are both documented to have some low but detectable amount of terminal transferase-like (template-independent) activity (Dominguez et al., 2000; Ramadan et al., 2003). It is likely that the N nucleotides observed in TdT knockout mice can be accounted for by pol mu and/or pol lambda (Komori et al., 1993).

Polymerase mu (aa 1–494) and the ⌬BRCT-truncated form (aa 133–494) were purified as recombinant proteins from E. coli as described previously (Dominguez et al., 2000). Polymerase lambda (aa 1–575) and the ⌬BRCT-truncated form (aa 133–575) were purified as recombinant proteins from E. coli as described (Shimazaki et al., 2002). The proteins were purified with a Ni-NTA column (Qiagen, Valencia, CA), a HiTrap Heparin column (Amersham Biosciences), and a Superose 6 column (Amersham Biosciences). Terminal transferase (aa 1–510) and the ⌬BRCT-truncated form (aa 101–510) were purified as recombinant proteins as described (Chang et al., 1988). Purified polymerase beta was a gift from Dr. R. Sobol and Dr. S. Wilson (Sobol et al., 1996).

Experimental Procedures

DeFazio, L.G., Stansel, R.M., Griffith, J.D., and Chu, G. (2002). Synapsis of DNA ends by DNA-dependent protein kinase. EMBO J. 21, 3192–3200.

Electrophoretic Mobility Shift Assay and Oligonucleotides See the Supplemental data. Protein Purification The purification of Ku, DNA-PKcs, and Artemis has been described (Ma et al., 2002). XRCC4:DNA ligase IV complex was purified as described (Nick McElhinny et al., 2000).

NHEJ Assay 0.25 pmol of the NHEJ substrate (YM-46/YM-159) was incubated with different combinations of proteins as indicated in NHEJ buffer (25 mM Tris (pH 8.0), 50 mM KCl, 2.5 mM MgCl2, 1 mM DTT, 0.05 mg/ml BSA, 0.25 mM ATP, and 10 ␮M dNTPs) in a total volume of 10 ␮L. After the protein DNA mixture was incubated at 37⬚C for 2 hr, the mixture was treated with or without organic extraction and 10% of the joined products were amplified by 35 cycles of PCR (94⬚C; 45 sec; 55⬚C; 1 min 15 sec; 72⬚C; 30 sec) with one of the primers labeled (YM-168 or YM-184, Figure 2A). The PCR products were resolved by 8% or 10% denaturing PAGE. The gel was dried, exposed in a PhosphorImager cassette, and the screen was scanned in PhosphorImager SI445 (Amersham Biosciences, Piscataway, NJ). Sequencing of the Junctions of the NHEJ Products and Ligaton Assay See the Supplemental data. Acknowledgments We thank Drs. Frederick J. Bollum and Lucy M. Chang for providing the TdT; Drs. S. Wilson and R. Sobel for pol beta; Dr. U. Grawunder for earlier work related to ligases I and III; and Drs. D. Shibata, S. Raghavan, and anonymous reviewers for comments. We thank C.A. Reynaud, B. Bertocci, and J.C. Weill for sharing data prior to publication. We apologize for uncited work due to space limitations. The work was supported by National Institutes of Health grants to M.R.L. (NIH GM 43236 and CA100504) and to M.F.G. (NIH ES012259). Received: July 13, 2004 Revised: October 6, 2004 Accepted: November 12, 2004 Published: December 2, 2004 References Bertocci, B., Smet, A.D., Berek, C., Weill, J.C., and Reynaud, C.A. (2003). Immunoglobulin kappa light chain gene rearrangement is impaired in mice deficient for DNA polymerase mu. Immunity 19, 203–211. Callebaut, I., and Mornon, J.P. (1997). From BRCA1 to RAP1: a widespread BRCT module closely associated with DNA repair. FEBS Lett. 400, 25–30. Chang, L.M., and Bollum, F.J. (1986). Molecular biology of terminal transferase. CRC Crit. Rev. Biochem. 21, 27–52. Chang, L.M., Rafter, E., Rusquet-Valerius, R., Roy, N.K., Cheung, L.C., and Bollum, F.J. (1988). Expression and processing of recombinant human terminal transferase in the baculovirus system. J. Biol. Chem. 263, 12509–12513. Chen, L., Trujillo, K., Sung, P., and Tomkinson, A.E. (2000). Interactions of the DNA ligase IV-XRCC4 complex with DNA ends and the DNA-dependent protein kinase. J. Biol. Chem. 275, 26196–26205.

de Vries, E., vanDriel, W., Bergsma, W.G., Arnberg, A.C., and van der Vliet, P.C. (1989). HeLa nuclear protein recognizing DNA termini and translocating on DNA forming a regular DNA-multimeric protein complex. J. Mol. Biol. 208, 65–78. Dominguez, O., Ruiz, J.F., Lain de Lera, T., Garcia-Diaz, M., Gonza-

Biochemical Reconstitution of Human NHEJ 713

lez, M.A., Kirchhoff, T., Martinez, C., Bernad, A., and Blanco, L. (2000). DNA polymerase mu (Pol mu), homologous to TdT, could act as a DNA mutator in eukaryotic cells. EMBO J. 19, 1731–1742. Ferguson, D.O., and Alt, F.W. (2001). DNA double-strand break repair and chromosomal translocations: lessons from animal models. Oncogene 20, 5572–5579. Gauss, G.H., and Lieber, M.R. (1996). Mechanistic constraints on diversity in human V(D)J recombination. Mol. Cell. Biol. 16, 258–269. Gerstein, R.M., and Lieber, M.R. (1993). Coding end sequence can markedly affect the initiation of V(D)J recombination. Genes Dev. 7, 1459–1469. Gottlieb, T., and Jackson, S.P. (1993). The DNA-dependent protein kinase: requirement for DNA ends and association with Ku antigen. Cell 72, 131–142. Grawunder, U., Wilm, M., Wu, X., Kulesza, P., Wilson, T.E., Mann, M., and Lieber, M.R. (1997). Activity of DNA ligase IV stimulated by complex formation with XRCC4 protein in mammalian cells. Nature 388, 492–495. Hammarsten, O., and Chu, G. (1998). DNA-dependent protein kinase: DNA binding and activation in the absence of Ku. Proc. Natl. Acad. Sci. USA 95, 525–530. Hanakahi, L., Bartlet-Jones, M., Chappell, C., Pappin, D., and West, S.C. (2000). Binding of inositol phosphate to DNA-PK and stimulation of double-strand break repair. Cell 102, 721–729. Komori, T., Okada, A., Stewart, V., and Alt, F. (1993). Lack of N regions in antigen receptor variable region genes of TdT-deficient lymphocytes. Science 261, 1171–1175. Kysela, B., Doherty, A.J., Chovanec, M., Stiff, T., Ameer-Beg, S.M., Vojnovic, B., Girard, P.M., and Jeggo, P.A. (2003). Ku stimulation of DNA ligase IV-dependent ligation requires inward movement along the DNA molecule. J. Biol. Chem. 278, 22466–22474. Lee, J.W., Blanco, L., Zhou, T., Garcia-Diaz, M., Bebenek, K., Kunkel, T.A., Wang, Z., and Povirk, L.F. (2003). Implication of DNA polymerase lambda in alignment-based gap filling for nonhomologous DNA end joining in human nuclear extracts. J. Biol. Chem. 279, 805–811. Lieber, M.R., Ma, Y., Pannicke, U., and Schwarz, K. (2004). The mechanism of vertebrate nonhomologous DNA end joining and its role in V(D)J recombination. DNA Repair (Amst.) 3, 817–826. Ma, Y., Pannicke, U., Schwarz, K., and Lieber, M.R. (2002). Hairpin opening and overhang processing by an Artemis:DNA-PKcs complex in V(D)J recombination and in nonhomologous end joining. Cell 108, 781–794. Mahajan, K.N., Nick McElhinny, S.A., Mitchell, B.S., and Ramsden, D.A. (2002). Association of DNA polymerase mu with Ku and DNA ligase IV: role of in end-joining double-strand break repair. Mol. Cell. Biol. 22, 5194–5202. Modesti, M., Junop, M.S., Ghirlando, R., van de Rakt, M., Gellert, M., Yang, W., and Kanaar, R. (2003). Tetramerization and DNA ligase IV interaction of the DNA double-strand break protein XRCC4 are mutually exclusive. J. Mol. Biol. 334, 215–228. Nick McElhinny, S.A., Snowden, C.M., McCarville, J., and Ramsden, D.A. (2000). Ku Recruits the XRCC4-Ligase IV Complex to DNA Ends. Mol. Cell. Biol. 20, 2996–3003. Pfeiffer, P., Goedecke, W., and Obe, G. (2000). Mechanisms of DNA double-strand break repair and their potential to induce chromosomal aberrations. Mutagenesis 15, 289–302. Ramadan, K., Maga, G., Shevelev, I.V., Villani, G., Blanco, L., and Hubscher, U. (2003). Human DNA polymerase lambda possesses terminal deoxyribonucleotidyl transferase activity and can elongate RNA primers: implications for novel functions. J. Mol. Biol. 328, 63–72. Ramsden, D.A., and Gellert, M. (1998). Ku protein stimulates DNA end joining by mammalian DNA ligases: a direct role for Ku in repair of DNA double-strand breaks. EMBO J. 17, 609–614. Repasky, J.A., Corbett, E., Boboila, C., and Schatz, D.G. (2004). Mutational analysis of terminal deoxynucleotidyl transferase-mediated N-nucleotide addition in V(D)J recombination. J. Immunol. 172, 5478–5488. Shimazaki, N., Yoshida, K., Kobayashi, T., Toji, S., Tamai, K., and Koiwai, O. (2002). Over-expression of human DNA polymerase

lambda in E. coli and characterization of the recombinant enzyme. Genes Cells 7, 639–651. Sobol, R.W., Horton, J., Kuhn, R., Gu, H., Singhal, R., Prasad, R., Rajewsky, K., and Wilson, S. (1996). Requirement of mammalian DNA polymerase-beta in base-excision repair. Nature 379, 183–186. Tseng, H.M., and Tomkinson, A.E. (2002). A physical and functional interaction between yeast Pol4 and Dnl4-Lif1 links DNA synthesis and ligation in nonhomologous end joining. J. Biol. Chem. 277, 45630–45637. Udayakumar, D., Bladen, C.L., Hudson, F.Z., and Dynan, W.S. (2003). Distinct pathways of nonhomologous end joining that are differentially regulated by DNA-dependent protein kinase-mediated phosphorylation. J. Biol. Chem. 278, 1631–1635. Walker, J.R., Corpina, R.A., and Goldberg, J. (2001). Structure of the Ku heterodimer bound to DNA and its implications for doublestrand break repair. Nature 412, 607–614. West, R.B., Yaneva, M., and Lieber, M.R. (1998). Productive and Nonproductive Complexes of Ku and DNA-PK at DNA Termini. Mol. Cell. Biol. 18, 5908–5920. Wilson, T., and Lieber, M.R. (1999). Efficient processing of DNA ends during yeast nonhomologous end joining: evidence for a DNA polymerase beta (POL4)-dependent pathway. J. Biol. Chem. 274, 23599–23609. Wilson, T.E., Grawunder, U., and Lieber, M.R. (1997). Yeast DNA ligase IV mediates non-homologous DNA end joining. Nature 388, 495–498. Yaneva, M., Kowalewski, T., and Lieber, M.R. (1997). Interaction of DNA-dependent protein kinase with DNA and with Ku: biochemical and atomic-force microscopy. EMBO J. 16, 5098–5112.