V(D)J Recombination in Ku86-Deficient Mice: Distinct Effects on Coding, Signal, and Hybrid Joint Formation

Immunity, Vol. 7, 37–47, July, 1997, Copyright 1997 by Cell Press

V(D)J Recombination in Ku86-Deficient Mice: Distinct Effects on Coding, Signal, and Hybrid Joint Formation Molly A. Bogue, Chiyu Wang, Chengming Zhu, and David B. Roth* Department of Microbiology and Immunology Baylor College of Medicine One Baylor Plaza Houston, Texas 77030

Summary Ku, a heterodimer of 70 and 86 kDa subunits, plays a critical but poorly understood role in V(D)J recombination. Although Ku86-deficient mice are defective in coding and signal joint formation, rare recombination products have been detected by PCR. Here, we report nucleotide sequences of 99 junctions from Ku86-deficient mice. Over 90% of the coding joints, but not signal or hybrid joints, exhibit short sequence homologies, indicating that homology is required to join coding ends in the absence of Ku86. Our results suggest that Ku86 may normally have distinct functions in the formation of these different types of junctions. Furthermore, Ku862/2 joints are unexpectedly devoid of N-region diversity, suggesting a novel role for Ku in the addition of N nucleotides by terminal deoxynucleotidyl transferase. Introduction V(D)J recombination assembles antigen receptor genes from variable (V), diversity (D), and joining (J) gene segments that are separated by intervening germline DNA. These coding segments are flanked by recombination signal sequences (RSS) composed of consensus heptamer and nonamer recognition sequences separated by either a 12- or 23-nucleotide spacer. Efficient recombination requires one RSS of each spacer length, a restriction known as “the 12/23 rule,” thought to reflect a requirement for formation of a synaptic complex containing a 12/23 RSS pair (Eastman et al., 1996; Steen et al., 1996; van Gent et al., 1996b). V(D)J recombination can be divided into three basic steps—cleavage, processing of broken DNA ends, and joining (reviewed in Lewis, 1994a; Bogue and Roth, 1996). Cleavage, mediated by RAG1 and RAG2 proteins, occurs precisely between the RSS and the coding segments, giving two types of recombination intermediates—covalently sealed (hairpin) coding ends and blunt signal ends (Roth et al., 1993; Schlissel et al., 1993; McBlane et al., 1995; van Gent et al., 1995; Zhu and Roth, 1995). Signal ends are generally joined without processing, producing precise signal joints. In contrast, hairpin coding ends are opened and nucleotides are commonly lost from or added to the ends, resulting in imprecise coding joints that greatly diversify the antigen receptor repertoire. Mechanisms for nucleotide loss are *To whom correspondence should be addressed.

not well defined but may include exo- (or endo-) nuclease activity. Alternatively, nucleotide loss may occur directly as a result of the joining mechanism; such deletions would be expected to occur from recombination directed by short sequence homologies, as suggested (Feeney et al., 1994). Two major processes are responsible for adding nucleotides. Template-dependent nucleotide additions, termed P nucleotides, are thought to be generated by hairpin opening off the axis of symmetry, leaving a short single-stranded extension that can be retained in the junction (Roth et al., 1992a; Meier and Lewis, 1993; Lewis, 1994b; Zhu and Roth, 1995). Nucleotides can also be added in a templateindependent manner by terminal deoxynucleotidyl transferase (TdT; for review see Gilfillan et al., 1995). These insertions are termed N nucleotides and are usually GC rich. More than one of these processing mechanisms may be operative for a given rearrangement. The joining step of V(D)J recombination employs several ubiquitous factors also implicated in double-strand break (DSB) repair. Cell lines or mice bearing mutations in genes encoding these factors exhibit specific defects in the processing of V(D)J recombination intermediates. For example, the murine scid mutation, which affects the catalytic subunit of the DNA-dependent protein kinase (DNA-PK), specifically impairs coding joint formation while signal joint formation is relatively unaffected (Lieber et al., 1988a). Hairpin coding ends accumulate in thymocytes of scid mice, indicating that hairpin opening is defective (Roth et al., 1992a; Zhu and Roth, 1995). Since transfected hairpins can be opened normally in scid cells (Lewis, 1994b), it has been suggested that DNA-PK may regulate the accessibility of hairpin coding ends to the processing machinery (Blunt et al., 1995; Roth et al., 1995; Zhu and Roth, 1995). The catalytic subunit of DNA-PK (DNA-PKcs) is activated upon interaction with Ku, a heterodimer composed of 70 and 86 kDa subunits, which binds altered DNA structures such as hairpins, nicks, or DSB (Mimori and Hardin, 1986; Paillard and Strauss, 1991; Gottlieb and Jackson, 1993; Morozov et al., 1994). Cell lines bearing a mutation in the XRCC5 gene, which encodes Ku86, are defective in both coding and signal joint formation (Pergola et al., 1993; Taccioli et al., 1993; Getts and Stamato, 1994; Rathmell and Chu, 1994; Mizuta et al., 1996). Although coding joint sequences have not been reported, rare signal joints formed in these Ku86-deficient cell lines are aberrant, generally with loss of nucleotides (Pergola et al., 1993; Taccioli et al., 1993). These observations led us and others to propose an end protection model postulating that Ku is required for protection of broken DNA ends generated during V(D)J recombination (Getts and Stamato, 1994; Rathmell and Chu, 1994; Taccioli et al., 1994; Blunt et al., 1995; Finnie et al., 1995; Jackson and Jeggo, 1995; Roth et al., 1995). Ku86-deficient mice have recently been generated by targeted disruption of the XRCC5 gene (Zhu et al., 1996; Nussenzweig et al., 1996). As expected, formation of both coding and signal joints is severely impaired in these animals. In Ku86-deficient thymocytes, hairpin coding ends accumulate and signal ends are blunt and

Immunity 38

full length, suggesting that Ku86 is not required for the protection of V(D)J recombination intermediates (Zhu et al., 1996). Rare coding joints were detected at multiple loci in T and B cell lineages by PCR (Nussenzweig et al., 1996; Zhu et al., 1996). To learn more about the Ku86independent mechanisms responsible for formation of these junctions, we cloned and sequenced coding, signal, and hybrid joints from these mice. If the major role of Ku86 were protection of signal and coding ends, one might expect these three types of junctions to share similar structural characteristics, such as loss of nucleotides. However, we found that the sequences of coding, signal, and hybrid joints show distinct features. While coding joint formation appears to be dependent on sequence homologies, signal and hybrid joint formation are not. Furthermore, recombination products generated in Ku86-deficient mice are unexpectedly devoid of N-region diversity. These data provide the first evidence that Ku may play a role in modulating the addition of N nucleotides by TdT, perhaps by regulating the accessibility of V(D)J recombination intermediates to multiple processing and joining activities.

consistently observed at similar levels in bone marrow from Ku86 1/2, Ku862/2 and scid mice. We wondered whether the smaller PCR product might represent a particular type of rearrangement not significantly impaired in Ku86-deficient or scid mice. Given its size and failure to hybridize to D-region-specific probes (data not shown), we considered the possibility that this smaller product might be derived from hybrid joints. As illustrated in Figure 1, joining of the signal end created by cleavage 59 of the D coding element to the J H4 coding end would result in formation of a hybrid joint, which does not contain the D coding region, and would be amplified by the same primers used to detect coding joints. To search for such junctions, we rehybridized the blot shown in Figure 2A with an oligonucleotide probe specific for a hybrid joint. This probe specifically detects the smaller PCR product, which is present at similar abundance in all samples (Figure 2B). Nucleotide sequence analysis of cloned IgH D-JH4 PCR products revealed the presence of coding and hybrid joints (shown separately in Figure 3) in bone marrow

Results Formation of Perfect Hybrid Joints in Ku86-Deficient Bone Marrow Precursor Cells The strategy for amplification of IgH D-JH4 rearrangements is shown in Figure 1. The PCR primers employed can simultaneously detect standard coding joints and nonstandard junctions, termed hybrid joints (Lewis et al., 1988). Examination of PCR products derived from bone marrow IgH D-JH4 rearrangements of Ku861/2 mice revealed the presence of two distinct classes of products that hybridized to a JH4-specific probe (Figure 2A). A larger product, predominant in the Ku861/ 2 sample (lane 1), was quite rare in Ku862/2 or scid bone marrow (lanes 3–6) (see longer exposure in the bottom panel of Figure 2A). A lower molecular weight product was

Figure 1. Standard Coding Joints and Nonstandard Hybrid Joints Are Generated at the IgH Locus IgH D segments (shaded box) are flanked by 12-RSS on both sides (open and hatched triangles), while the J region (open box) is flanked 59 by a 23-RSS (closed triangle). Coding and hybrid joint formation can occur at this locus without violating the 12/23 rule and can be detected by PCR using the same primer set (primers shown as horizontal arrows) and J H4-specific probe (asterisk). Standard coding joints would be dependent on cleavage at the 39 12-RSS of the D region and the J 23-RSS (open vertical arrows), while hybrid joint formation would be dependent on cleavage at the 59 12-RSS of the D region and the J 23-RSS (closed vertical arrows).

Figure 2. PCR Analysis of IgH D-JH4 Recombination Products Semiquantative PCR for analysis of IgH recombination products was performed as described (Zhu et al., 1996) using 100 ng of bone marrow DNA isolated from individual mice from Ku861/2 day 16 (lane 1); Ku862/2 day 10 (lane 3), day 16 (lane 4), and 2 months (lane 5); and from scid day 19 (lane 6). The PCR primers employed can simultaneously detect standard coding joints and nonstandard hybrid joints as illustrated in Figure 1. The predicted size of a perfect hybrid junction, which lacks the entire D coding region, is 83 bp where a coding joint would be somewhat larger (taking into account 17–23 nucleotides of coding region as well as any nucleotide loss or addition). (A) PCR products were blotted and hybridized to a probe specific for the JH4 coding region. The exposure in the top panel is optimized to show Ku861/2 fragments representing coding joints (cj) and hybrid joints (hj), while the longer exposure in bottom panel shows hybrid joints as well as rare products consistent with the size of coding joints in day 16 Ku862/2 (lane 4) and scid (lane 6) samples. (B) The same blot was stripped and hybridized with a hybrid-specific probe. Relevant size markers (1 kb ladder, GIBCO-BRL) are indicated in base pairs. It should be noted that the levels of hybridization seen in lanes 3 and 4 with the hybrid-specific probe are very similar (B), while the levels of hybridization are different in these lanes when hybridized with the JH4-specific probe (A). This apparent discrepancy can be explained by the relatively high proportion of deleted coding joint sequences (see Figure 3) that are recovered from the sample in lane 4 and are not detected by the hybrid-specific probe. These deleted coding joint sequences are exactly the same size as the undeleted hybrid joints, leading to a more intense signal with the J H4-specific probe.

V(D)J Recombination Products in Ku86-Deficient Mice 39

Figure 3. Sequences of IgH D-JH4 Rearrangements Derived from Bone Marrow DNA PCR products amplified from bone marrow DNA preparations from individual Ku862/2 mice (days 10, 16, and two months), Ku861/2, and scid mice were cloned. Plasmid DNA from colonies that hybridized to a J H4-specific probe were sequenced. Sequences of hybrid joints (A) and coding joints (B) are displayed separately for clarity. The 59 12-RSS of the D coding region is shown in lowercase letters, while nucleotides from D and J coding regions as well as P and N nucleotides are shown in capital letters. Junctional region nucleotides have been assigned according to the criteria in Experimental Procedures. Nucleotides that could be assigned to more than one category are sequence homologies and are thus underlined. Deleted nucleotides are indicated by dots. Note that the DFL16.2 coding region is 6 bp longer than the other D regions shown. Abbreviations are as follows: 59 12-RSS of D coding region (12-RSS); D coding region (D); P nucleotides (P); N-nucleotide additions (N); J H4 coding region (JH4), frequency (f); D coding segment usage (Ds).

preparations from Ku86-deficient mice. Remarkably, all 29 hybrid joints recovered from three Ku862/2 mice are perfect, with no loss or addition of nucleotides from either the signal or the coding end (Figure 3A). An imperfect hybrid joint with nine nucleotides deleted from the JH4 coding end was detected in our scid bone marrow sample. Hybrid joints with deletions from coding or signal ends have been previously detected in scid cell lines

(Lieber et al., 1988a) and in wild-type mice and cell lines (Lewis et al 1988; Sollbach and Wu, 1995; VanDyk et al., 1996). Because hybrid joints are much less abundant than coding joints in Ku861/2 bone marrow DNA (Figure 2A), these junctions were not recovered in this sample. Overall, these results are consistent with recent experiments using artificial recombination substrates that demonstrated that hybrid joints are formed with similar

Immunity 40

efficiency in Ku86-deficient, scid, and wild-type fibroblasts (Han et al., 1997). Collectively, these data indicate that while coding joint formation is severely impaired in both scid and Ku86-deficient mice, formation of hybrid joints is relatively unaffected. Thus, joining of a coding end to a signal end can bypass the requirement for Ku86 or DNA-PKcs. Formation of Rare IgH and TCRd Coding Joints in Ku86-Deficient Mice Analysis of IgH D-JH4 rearrangements from bone marrow DNA of Ku86-deficient mice (Figure 2A, lower panel) revealed rare PCR products of sizes expected for coding joints (lane 4). Nucleotide sequence analysis of cloned PCR products from this sample revealed several distinct IgH D-JH4 coding joints (7 unique sequences of 19 total) shown in Figure 3B. We also cloned and sequenced TCRd coding joints from thymocyte DNA since PCR products of sizes appropriate for coding joints were detected in Ku86-deficient mice (Nussenzweig et al., 1996; Zhu et al., 1996). This rearrangement was chosen because a large number of junction sequences from both scid and wild-type thymocytes are available for comparison (Kienker et al., 1991; Carroll et al., 1993a, 1993b; Fish and Bosma, 1994). As shown in Figure 4, 25 Dd2-Jd1 coding joint sequences were obtained, 15 of which represent unique junctions. The presence of numerous repeated junctions in Ku862/2 samples compared with very few repeats in scid and Ku861/2 samples (Figures 3B and 4) suggests that coding joint diversity may be more limited in Ku86-deficient mice. This interpretation is consistent with the observation that TCRd rearrangements are less frequent in Ku862/2 mice than in scid mice (Nussenzweig et al., 1996; Zhu et al., 1996). These data demonstrate that authentic coding joints can form, although rarely, in the absence of Ku86. That these coding joints are derived from hairpin coding end intermediates (Roth et al., 1992a; Meier and Lewis, 1993; Zhu and Roth, 1995) is supported by the frequent presence of P nucleotides, which are normally less than three nucleotides in length (Meier and Lewis, 1993). However, in Ku862/2 junctions, a number of abnormally long P nucleotide insertions (up to eight nucleotides in length) were observed. Such long P insertions are also observed in scid coding joints, as seen in Figures 3 and 4 (see also Kienker et al., 1991; Schuler et al., 1991) and are thought to result from aberrant processing of hairpin intermediates. These data suggest that hairpin coding ends can be processed—albeit infrequently—in the absence of Ku86, generating junctions that (except for occasional long P insertions) are like coding joints formed in normal mice. However, a more detailed analysis of these sequences revealed several unique features, described below.

1993) and mice (see below). Although there are limitations to the degree of nucleotide loss one can detect using a PCR-based approach, comparison of the average number of nucleotides deleted from any given coding end in Ku862/2 and Ku861/ 2 mice did not reveal statistically significant differences. On the contrary, our analysis of coding joints revealed numerous junctions containing full-length coding ends. For example, a considerable proportion of Dd2-Jd1 coding joints from Ku86-deficient mice (15 of 25 total junctions; 5 of 15 unique junctions) have lost no nucleotides from either coding end, and overall, 22 of 25 (88%) of the junctions have at least one full-length coding segment. In fact, there are more full-length ends in coding joints of Ku86deficient mice than in littermate controls or C.B-17 wildtype mice (19 of 31; 42%). Similarly, our analysis of the IgH locus revealed a high proportion of junctions containing full-length D or J segments. The lack of excessive deletion and the preservation of many full-length ends that participate in joining indicate that Ku is not required for their protection. This is also consistent with the complete preservation of coding and signal ends in hybrid joints from Ku86 2/2 bone marrow (Figure 3A) and cultured cells (Han et al., 1997) and with our previous observation that hairpin coding ends and undeleted signal ends accumulate in thymocytes of Ku86-deficient mice (Zhu et al., 1996). Together, these data indicate that Ku86 is not required for the protection of coding ends. One notable feature of these junctional sequences is that almost all (41 of 44 junctions; 93%) D-J coding joints of IgH and TCRd from Ku86-deficient mice have short sequence homologies (Figures 3B and 4). In fact, all of the IgH D-JH4 coding joint sequences contain junctional homologies. This is in sharp contrast to Ku86 1/2 IgH junctions, none of which have homologies of even a single nucleotide. Considering all coding joint sequences from Ku861/ 2 and C.B-17 (wild-type) mice, only 4 of 42 (9.5%) IgH and TCRd coding joints have short sequence homologies. Because the lack of N regions observed in the Ku862/2 coding joints (see below) might bias the data set, enriching for sequences with short homologies (Gu et al., 1990; Feeney, 1992; Komori et al., 1993; Gerstein and Lieber, 1993; Gilfillan et al., 1993; Zhang et al., 1995), we compared the Ku862/2 sequences to scid coding joints that have fewer sequences containing N regions than Ku861/ 2 or C.B-17 coding joints. Even when sequences with N-region diversity were omitted from the comparison, only 20 of 40 total scid sequences (50%) show junctional homologies. Taken together, these results indicate that in the absence of Ku86, the mechanism responsible for coding joint formation may preferentially employ short sequence homologies.

Full-Length Coding Segments and Short Sequence Homologies Predominate in Coding Joints from Ku86-Deficient Mice One simple prediction of the end protection model is that junctions involving coding ends isolated from Kudeficient mice might exhibit excessive loss of nucleotides, as observed in signal joints generated in Ku86deficient fibroblasts (Pergola et al., 1993; Taccioli et al.,

Signal Joints in Ku86-Deficient Mice Lack Short Sequence Homologies To examine the characteristics of signal joint formation in Ku86-deficient mice, we cloned and sequenced these reciprocal rearrangement products. Full-length signal joints are quite infrequent in Ku862/2 mice and were not detected by standard PCR amplification (Nussenzweig

V(D)J Recombination Products in Ku86-Deficient Mice 41

Figure 4. Sequences of TCR Dd2-Jd1 Coding Joints Derived from Thymocyte DNA PCR products amplified from thymocyte DNA preparations from pooled samples from day 16 Ku862/2 and Ku861/2 mice were cloned and sequenced. The germline sequence for the Dd2 coding region with its flanking 12-RSS (in small letters) and the Jd1 coding region are shown in the top line. Heptamer and nonamer nucleotides of the RSS are underlined only in the germline sequence. Previously published (Fish and Bosma, 1994) sequences from newborn scid mice and adult C.B-17 wild-type mice are shown for direct comparison. Sequences of Group I and II rearrangements are separated by a blank line, with Group I rearrangements listed first in each case. The space between the 12-RSS and the D coding region is to facilitate alignment of those sequences with added nucleotides (open-and-shut events [Lewis and Hesse, 1991; Fish and Bosma, 1994] with modification); N and P nucleotides resulting from these events are shown in parentheses. See Figure 3 legend and Experimental Procedures for abbreviations, nucleotide assignment criteria, and Group I and II designations. (*) Sequence detail of the Ku862/2 rearrangement containing a large insert. The insert is shown in small letters, while P (in parenthesis) and adjacent coding region nucleotides are shown in capital letters. Over- and underlines indicate pairs of direct repeats; pairs are numbered. Such duplication junctions have been observed in fibroblasts (in the absence of TdT), and a homology-dependent model for their formation has been proposed (Roth et al., 1985, 1989).

Immunity 42

Figure 5. Sequences of TCR Dd2-Jd1 Signal Joints from Thymocyte DNA PCR products amplified from the same Ku861/2 and Ku862/2 thymocyte DNA preparations (as used for TCRd coding joints in Figure 4) as well as from DNA preparations from different ages of scid mice were cloned and sequenced. To preserve the D-to-J orientation, sequences are presented 39 to 59. Germline sequences of the 39 Dd2 23-RSS and the 59 Jd1 12-RSS are shown in the top line in lowercase letters; nucleotides of the heptamer and nonamer are underlined only in the germline sequence. Non-RSS nucleotides are in capital letters. The nucleotides in parenthesis are derived from the Dd2 coding region. Nucleotides of homology are underlined. See Experimental Procedures for nucleotide assignment criteria. (*) The sequence of the unusually large insert (39 bp) is TGACGATTCACACCGATCACTTGACCGGTGCTCGTATCC.

et al., 1996; Zhu et al., 1996). Rare TCR Dd2-Jd1 signal joints were amplified by two-stage PCR from the same Ku862/2 thymocyte DNA samples used above. Signal joint sequences are shown in Figure 5. The majority of scid and Ku861/2 signal joints are perfect without nucleotide loss or addition, consistent with previous results (Carroll et al., 1993a). This is in contrast to the Ku862/2 sequences, of which only 3 of 26 junctions were perfect signal joints. However, many of the junctions contained one full-length signal end with the other end exhibiting abnormal nucleotide loss. These characteristics are consistent with signal joints formed from artificial recombination substrates in Ku86-deficient cell lines (Pergola et al., 1993; Taccioli et al., 1993). A single signal joint containing nucleotides from the Dd2 coding region (nucleotides in parenthesis) is present in our Ku86 2/2 sequence set. This type of aberrant joint, which has been observed previously, could be generated by imprecise cleavage by RAG proteins (Carroll et al., 1993a, 1993b). Although there are several opportunities for homologydirected joining in these signal ends, only 1 out of 26

signal joints showed any junctional homology (two nucleotides). This is in sharp contrast to coding joints from Ku862/2 mice, in which over 90% of the sequences contain short homologies. Thus, while it appears that short sequence homologies may play an important role in joining coding ends in Ku862/2 mice, the mechanism of signal joint formation appears to be distinct.

Recombination Products from Ku86-Deficient Mice Lack N-Region Diversity A conspicuous feature of the Ku862/2 coding, signal, and hybrid joints is the virtual absence of N-nucleotide additions. Only 1 of 25 Ku86 2/2 TCRd coding joint sequences contained a single nucleotide that could not be assigned to a coding region or as a P nucleotide, in contrast to Ku861/2 and C.B-17 TCRd coding joint sequences in which the majority of sequences (23 of 31) have N-region diversity (Figure 4). The extra nucleotide in the Ku862/2 TCRd coding joint sequence set could have been added in a TdT-independent manner, since

V(D)J Recombination Products in Ku86-Deficient Mice 43

such rare nucleotide additions are also observed in TdTdeficient mice (Gilfillan et al., 1993; Komori et al., 1993). The two unusually large inserts in the Ku862/ 2 TCRd coding and signal joint sequences (Figures 4 and 5) do not appear to have been generated by TdT because of their size and low GC content. The presence of these inserts is more readily explained by a homology-dependent joining mechanism (see legend to Figure 4) and by oligonucleotide capture (Roth et al., 1989; Lieber, 1992; Carroll et al., 1993b). Thus, severe deficiency of N-nucleotide addition is evident in coding joints derived from Ku86-deficient mice. Furthermore, although signal and hybrid joints generated in wild-type mice may contain N nucleotides (Lieber et al., 1988b; Carroll et al., 1993a, 1993b; Lewis, 1994a; Cande´ias et al., 1996), none of the Ku862/2 hybrid or signal joints contained N-region diversity (Figures 3A and 5). Taken together, these data indicate that N-nucleotide addition is severely impaired in the absence of Ku86. TdT Transcript Levels Are Similar in Thymocytes of scid and Ku86-Deficient Mice Since Ku has been implicated in transcriptional regulation (Li et al., 1995; Giffin et al., 1996), one obvious explanation for the lack of N-region diversity is that Ku86 is required for transcriptional activation of the TdT gene. Since thymocytes from Ku86-deficient and scid mice are arrested at the same developmental stage (Zhu et al., 1996) and scid-derived coding and signal joint sequences often contain N nucleotides, scid thymocytes are the most appropriate comparison for TdT transcript levels. TdT transcript levels were assessed by semiquantitative RT–PCR using serial dilutions of cDNA derived from thymocyte RNA. We found that TdT transcript levels were comparable in thymocyte RNA samples from one-month-old scid and Ku862/2 mice, as seen in Figure 6. Similar results were obtained with independent thymocyte RNA preparations from scid and Ku862/ 2 (day 12) mice; furthermore, RT–PCR reactions based on equal amounts of RNA (Figure 6) and on equal cell numbers (not shown) gave similar results. These results indicate that Ku86 is not required for the transcriptional activation of the TdT gene. Discussion Here, we report the first nucleotide sequences of coding joints generated in the absence of Ku86. The structural

Figure 6. TdT RT–PCR of Thymocyte RNA from scid and Ku862/2 Mice Serial dilutions of cDNA derived from reverse transcription of thymocyte RNA (0.5 mg) from one month old scid or Ku862/2 mice were amplified by PCR, blotted, and hybridized with an internal oligonucleotide probe. Four 10-fold dilutions are shown for each RNA preparation, followed by a blank lane. The negative control (neg) contained all components except RNA. The expected 320 bp product is reverse transcriptase dependent (not shown).

characteristics of coding, signal, and hybrid joints suggest that different joining mechanisms are responsible for the formation of these three types of junctions in Ku86-deficient mice (summarized in Table 1). A large proportion of coding, signal, and hybrid joints contain at least one full-length end, indicating that Ku86 is not required for the protection of these ends. Unexpectedly, junctions from Ku86-deficient mice are devoid of N-region diversity, suggesting a novel role for Ku in the addition of N nucleotides by terminal deoxynucleotidyl transferase. These data support the contention that Ku is involved in processing and joining V(D)J recombination intermediates. A RAG-Mediated Joining Model for Hybrid Joint Formation The majority of rearrangements at the IgH D-JH4 locus recovered from Ku86-deficient mice are perfect hybrid joints (Figure 3A), which are present at similar levels in Ku861/2 and Ku862/ 2 bone marrow DNA (Figure 2). Together with experiments using artificial substrates, these data indicate that joining of a coding end to a signal end to form a hybrid joint can efficiently bypass the requirement for Ku86. We have previously suggested that in the absence of Ku86, coding and signal ends are trapped in a stable synaptic complex that contains RAG1 and RAG2 (Zhu et al., 1996). One possibility is that RAG proteins may mediate formation of hybrid joints within this complex (Han et al., 1997). In the normal cleavage reaction, the RAG nuclease first introduces a nick immediately 59 to the RSS and then uses the resulting 39 hydroxyl group as a nucleophile to attack the phosphodiester bond on the bottom strand, forming a hairpin coding end and a blunt signal end (McBlane et al., 1995) as shown in Figure 7A. Simply reversing the second step would allow joining of a signal end to a hairpin coding end, a process analogous to “disintegration” reactions performed by retroviral integrases (Chow et al., 1992) which operate by the same chemical mechanism as the RAG nuclease (van Gent et al., 1996a). As shown in Figure 7A, rejoining of a corresponding coding/ signal end pair simply regenerates the substrate. However, joining of a pair of noncorresponding ends would yield a hybrid junction as illustrated in Figure 7B. Because free ends are not required, this model explains both the efficient formation of hybrid joints in the absence of Ku86 and the complete lack of end processing observed in the hybrid joints reported here (Figure 3A). Distinct Pathways for Coding and Signal Joint Formation in Ku86-Deficient Mice Remarkably, over 90% of the coding joints recovered from both B and T lineages in Ku86-deficient mice contain short sequence homologies. The simplest interpretation of these results is that coding joint formation might be dependent upon (or facilitated by) base-pairing interactions in the absence of Ku86. This hypothesis is supported by recent studies of plasmid end joining in Saccharomyces cerevisiae, where joint formation in cells deficient for either the Ku70 or Ku86 homolog is impaired, and all junctions recovered contain short sequence homologies (Boulton and Jackson, 1996a,

Immunity 44

bound RAG proteins. In this regard, it is noteworthy that many of these signal joints (approximately 70%) contain at least one full-length end. One way in which signal ends might escape from a synaptic complex could involve rare endonucleolytic cleavage events that would disrupt or remove an RSS, promoting dissociation of bound RAG proteins. Occurrence of such an event at one RSS should be sufficient to allow release of both ends from the complex, as RAG proteins do not bind to an isolated signal end (Hiom and Gellert, 1997). Differential processing of coding and signal end intermediates is a hallmark of the normal V(D)J recombination process (reviewed in Lewis, 1994a). The substantial differences in the characteristics of coding and signal joints isolated from Ku86-deficient mice indicates that differential processing and joining of coding and signal ends is maintained in the absence of Ku86. This situation contrasts with observations in xrcc4 mutant cell lines, which are also defective in coding and signal joint formation. Nucleotide sequences of coding, signal, and hybrid joints formed in xrcc4 mutant cell lines suggests that all three types of joints are formed in a common pathway that utilizes short sequence homologies (Li et al., 1995). These data indicate that the XRCC4 protein and Ku86 have distinct functions in V(D)J recombination.

Figure 7. A RAG-Mediated Joining Model for Formation of Hybrid Joints (A) Forward and reverse RAG-mediated reactions. In the forward reaction, the RAG nuclease (shaded oval) nicks precisely between the RSS (triangle) and the coding flank, generating a free 39OH, which then attacks the phosphodiester bond of the opposite coding strand. A hairpin coding end and a blunt signal end are produced (McBlane et al., 1995). In the reverse reaction, the 39OH of the signal end could be used as a nucleophile by the RSS-bound RAG proteins to attack the phosphodiester bond of the corresponding hairpin coding end. Repair of the nick would regenerate the original substrate. (B) Hybrid joint formation mediated by RAG proteins. Following RAG-mediated cleavage, the four ends remain in a DNA–protein complex (hairpin coding ends boxed and signal ends terminate in triangles; corresponding coding/signal pairs are indicated by opened or shaded boxes or triangles). The 39OH at a signal end could be used to attack a noncorresponding coding end. Repair of the resulting nicked structure would generate a perfect hybrid joint.

1996b). However, overrepresentation of junctions containing short homologies is not observed in products of extrachromosomal end joining in Ku86-deficient fibroblasts (Liang and Jasin, 1996; Kabotyanski et al., unpublished data), indicating that use of short sequence homologies is not a general feature of end joining in mammalian cells lacking Ku86. The signal joint sequences obtained in Ku86-deficient thymocytes (this work) and mutant fibroblasts (Pergola et al., 1993; Taccioli et al., 1993) rarely contain junctional homologies of even a single nucleotide. Thus, these signal joints appear to form by a joining mechanism distinct from that used to form coding joints, suggesting that Ku has different roles in the formation of these two types of junctions. The characteristic loss of nucleotides from signal joints formed in the absence of Ku86 may reflect escape of the ends from protection provided by

N-Nucleotide Addition Is Dependent on Ku86 Perhaps the most striking feature of junctions from Ku86-deficient mice is the virtual absence of N nucleotides, as summarized in Table 1. In our collection of 99 sequences of coding, signal, and hybrid junctions (all of which are substrates for N-nucleotide addition in wildtype mice), only one junction contained a single extra nucleotide that could not be assigned to a coding or signal end or as a P nucleotide (also excluding the two large inserts we suspect are TdT-independent). This contrasts with the situation in wild-type and Ku861/2 rearrangements, in which 81% of coding joints (IgH and TCRd) contain N regions. We have considered several potential explanations for the lack of N nucleotides in Ku86-deficient mice. The most straightforward possibility is that Ku86 is required, either directly or indirectly, for transcriptional activation of the TdT gene. However, the presence of similar levels of TdT transcripts in Ku862/2 and scid mice (which have N regions) indicates that expression of TdT mRNA is not grossly affected by the absence of Ku86 (Figure 6). Since several studies have established a good correlation between TdT mRNA levels and TdT activity in wild-type mice (Tillinghast et al., 1989; Coleman et al., 1992), active TdT is likely to be present in Ku862/2 thymocytes. An alternative possibility is that N nucleotides might be added normally but fail to appear in the junctions. Since TdT catalyzes the addition of nucleotides to the 39 hydroxyl group, perhaps addition of N nucleotides blocks the joining reaction in Ku86-deficient precursor cells by generating 39 single-stranded extensions. However, the frequent presence of P nucleotides in Ku862/2 coding joints indicates that at least some ends with single-stranded extensions can be joined. A related possibility is that short 39 extensions created by TdT could be used as homologies to direct joining and thus would

V(D)J Recombination Products in Ku86-Deficient Mice 45

Table 1. V(D)J Recombination Products Formed in Ku86-Deficient Mice Coding Joints

Signal Joints

Hybrid Joints

Decreased frequency Many with one or both full-length ends Lack of N nucleotides Homology-dependent joining

Decreased frequency Many with one full-length end Lack of N-nucleotides Homology-independent joining

Normal Both ends full-length Lack of N nucleotides Homology-independent joining (RAG-mediated?)

not appear as N regions. However, in the case of long 39 extensions, one would expect some residual N nucleotides to be retained in the junction. Taken together with the lack of N additions in signal joints, which rarely show junctional homologies, it seems unlikely that N nucleotides are simply “masked” by the use of homology. Based on the arguments given above, we favor another possibility, namely that Ku may be important for facilitating the interaction of TdT with V(D)J recombination intermediates. For example, Ku, as a DNA end binding heterodimer, could act in a direct fashion to actively facilitate loading of TdT onto DNA ends or to stimulate the activity of TdT once bound. Alternatively, perhaps Ku-dependent remodeling of the postcleavage complex is necessary to promote access of TdT to the coding and signal ends. According to this model, in the absence of Ku86, TdT would be excluded from the DNA–protein complex containing V(D)J recombination intermediates. This is consistent with our proposal that other V(D)J end processing activities such as hairpin opening and signal end joining require Ku86 (Zhu et al., 1996). Since Ku86 is a component of the DNA-PK holoenzyme, the requirement for Ku suggests that DNA-PKcs may also be involved in promoting access of coding ends to TdT. This hypothesis is supported by the relatively low frequency of coding joints with N regions derived from scid mice compared to age-matched controls observed previously (Pennycook et al., 1993; Araki et al., 1996) and in this study (25% of scid sequences with N nucleotides versus 81% for Ku861/2 and wild-type mice). Since the murine scid allele is not null (Blunt et al., 1996; Danska et al., 1996), the observation that the defect in N-nucleotide addition in scid mice is less severe than in Ku86-deficient mice may reflect some partial activity of the mutant form of DNA-PKcs. Together, the observations that N-nucleotide addition is completely blocked in Ku86-deficient mice and is impaired in scid mice suggest a previously unexpected role for DNA-PK in processing of V(D)J recombination intermediates. Experimental Procedures Mice and Preparation of DNA and RNA The generation of mice with a targeted disruption of the XRCC5 gene, encoding the Ku86 protein, has been described (Zhu et al., 1996). Mice heterozygous for this mutation were crossed, producing homozygous mutant mice and littermate controls. For simplicity, these mice will be referred to as Ku861/2 and Ku862/2. C.B-17 Icrscid/scid mice were bred and used for comparison. DNA was prepared as described (Roth et al., 1992b) from singlecell suspensions of homogenized thymus or bone marrow. Samples were pooled where noted in the event cell numbers were limiting. Thymocyte RNA was prepared from individual mice with the Ultraspec-II RNA isolation system (BIOTECX). First-strand cDNA synthesis was prepared from RNA samples with the SuperScript Preamplification System (Gibco BRL) using an oligo(dT) primer.

PCR, Cloning, and Sequencing DNA samples were amplified by semiquantitative PCR for the analysis of IgH and TCRd D-J recombination products, as described (Zhu et al., 1996). Briefly, we amplified IgH coding joints with a degenerate primer (DHL; Schlissel et al., 1991) recognizing most members of the murine D region family and a primer specific for JH4 (DR217; Zhu et al., 1996). These PCR products were detected using endlabeled internal oligonucleotide probes specific either for J H4 (DR218; Zhu et al., 1996) or for a perfect hybrid junction (HYB) specific for 8 nucleotides of the 59 D 12-RSS and 9 nucleotides of the JH4 coding region. TCRd rearrangements were amplified using primers specific for Dd2-Jd1 coding joints (DR6 [Zhu and Roth, 1995] and DR53 [Zhu et al., 1996]) and signal joints (first stage: DR124 and DR162 [Zhu et al., 1996]; second stage: 1 ml of first stage PCR into 50 ml fresh reaction mix with DR21 [Zhu et al., 1996] and DR162). TdT mRNA quantitation was performed as described (Bogue et al., 1992). All PCR amplifications were performed with a PE 9600 thermocycler with the following conditions: 13 reaction buffer supplied by the vendor with 2 mM MgCl2 and 1 unit of Taq Polymerase (Perkin Elmer) in 50 ml total volume; 948C for 30 sec, 658C for 15 sec, 728C for 15 sec. Amplification was for 30 cycles except for TdT and the first stage PCR for signal joints that were 25 cycles. PCR products were separated on 6% polyacrylamide gels and transferred to GeneScreen Plus (DuPont). The following end-labeled oligonucleotide probes were employed: DR218 for IgH, DR2 for TCRd coding joints (Zhu and Roth, 1995), and DR50 for TCRd signal joints. Hybridization was detected using a Molecular Dynamics PhosphorImager. PCR products were cloned by the TA cloning method (Invitrogen); colonies from vector-transformed bacteria were screened by colony hybridization using the above probes. Plasmids from positive colonies were subjected to cycle sequencing using ThermoSequenase with 33P-labeled dideoxynucleotides (Amersham). Oligonucleotides for PCR primers and probes not previously published: DR50, TCACCCTGCAGTTTTTGTAC; DR124, AGGCCAGCAA GTGGAGGTCATATC; and HYB, CTACTGTGATTACTATG.

Assignment of Junctional Nucleotides Nonambiguous nucleotides were assigned according to published germline sequences for IgH (Sakano et al., 1980; Chang et al., 1992, and references therein) and TCRd (Chien et al., 1987; Kienker et al., 1991). Nucleotides that could be assigned to more than one category are underlined, representing short sequence homologies, as described (Gu et al., 1990; Feeney, 1992; Gilfillan et al., 1993; Komori et al., 1993). Such ambiguous nucleotides from coding joints were arbitrarily assigned to the following regions in decreasing order of preference: D coding, J coding, P nucleotides of D region, or P nucleotides of J region. In some cases, P nucleotides could have been derived from either coding end; in this event, P nucleotides were equally distributed between each coding end. These ambiguous P nucleotides also represent sequence homologies and are thus underlined. Ambiguous nucleotides in signal joints were arbitrarily assigned to the RSS of the D region. By default, those nucleotides that did not fit under any of these categories were categorized as N nucleotides. It should be noted that there are two types of coding joints generated from Dd2-Jd1 rearrangements in wild-type mice, designated Groups I and II by Fish and Bosma (1994). Group I rearrangements retain at least some of the D coding region, while Group II rearrangements lack the entire coding Dd2 coding region and may exhibit nucleotide loss into the 59 12-RSS. Both groups are abundant in thymocyte DNA from wild-type and scid mice (Fish and Bosma, 1994) as shown.

Immunity 46

Acknowledgments

V(D)J recombination in vitro obeying the 12/23 rule. Nature 380, 85–88.

We thank Paul Hasty and Dae-Sik Lim for maintaining the Ku86deficient mice and Larissa Gomelsky and Mary Lowe for technical and secretarial assistance, respectively. We are especially grateful to the following for their helpful discussions and critical review of the manuscript: Martin Gellert, Susan Gilfillan, Mark Landree, and Susanna Lewis. This work was supported in part by a National Institutes of Health Post Doctoral Fellowship (T32-AI07495) to C. Z. and by a grant from the American Cancer Society (DB-118). D. B. R. is a Charles E. Culpeper Medical Scholar.

Feeney, A.J. (1992). Predominance of VH-D-JH junctions occurring at sites of short sequence homology results in limited junctional diversity in neonatal antibodies. J. Immunol. 149, 222–229.

Received March 3, 1997; revised May 2, 1997. References Araki, R., Itoh, M., Hamatani, K., and Abe, M. (1996). Normal D-JH rearranged products of the IgH gene in SCID mouse bone marrow. Int. Immunol. 8, 1045–1053. Blunt, T., Finnie, N.J., Taccioli, G.E., Smith, G.C.M., Demengeot, J., Gottlieb, T.M., Mizuta, R., Varghese, A.J., Alt, F.W., Jeggo, P.A., et al. (1995). Defective DNA-dependent protein kinase activity is linked to V(D)J recombination and DNA repair defects associated with the murine scid mutation. Cell 80, 813–823. Blunt, T., Gell, D., Fox, M., Taccioli, G.E., Lehmann, A.R., Jackson, S.P., and Jeggo, P.A. (1996). Identification of a nonsense mutation in the carboxyl-terminal region of DNA-dependent protein kinase catalytic subunit in the scid mouse. Proc. Natl. Acad. Sci. USA 93, 10285–10290. Bogue, M., and Roth, D.B. (1996). Mechanism of V(D)J recombination. Curr. Opin. Immunol. 8, 175–180. Bogue, M., Gilfillan, S., Benoist, C., and Mathis, D. (1992). Regulation of N-region diversity in antigen receptors through thymocyte differentiation and thymus ontogeny. Proc. Natl. Acad. Sci. USA 89, 11011–11015. Boulton, S.J., and Jackson, S.P. (1996a). Saccharomyces cerevisiae Ku70 potentiates illegitimate DNA double-strand break repair and serves as a barrier to error-prone DNA repair pathways. EMBO J. 15, 5093–5103. Boulton, S.J., and Jackson, S.P. (1996b). Identification of a Saccharomyces cerevisiae Ku80 homologue: roles in DNA double strand break rejoining and in telomeric maintenance. Nucleic Acids Res. 24, 4639–4648. Cande´ias, S., Muegge, K., and Durum, S.K. (1996). Junctional diversity in signal joints from T cell receptor beta and delta loci via terminal deoxynucleotidyl transferase and exonucleolytic activity. J. Exp. Med. 184, 1919–1926. Carroll, A.M., Slack, J.K., and Chang, W.-T. (1993a). Biased T-cell receptor d element recombination in scid thymocytes. Mol. Cell. Biol. 13, 3632–3640. Carroll, A.M., Slack, J.K., and Mu, X. (1993b). V(D)J recombination generates a high frequency of nonstandard TCR Dd-associated rearrangements in thymocytes. J. Immunol. 150, 2222–2230. Chang, Y., Paige, C.J., and Wu, G.E. (1992). Enumeration and characterization of DJH structures in mouse fetal liver. EMBO J. 11, 1891–1899. Chien, Y.-H., Iwashima, M., Wettstein, D.A., Kaplan, K.B., Elliott, J.F., Born, W., and Davis, M.M. (1987). T cell receptor d gene rearrangements in early thymocytes. Nature 330, 722–727. Chow, S.A., Vincent, K.A., Ellison, V., and Brown, P.O. (1992). Reversal of integration and DNA splicing mediated by integrase of human immunodeficiency virus. Science 255, 723–726. Coleman, M.S., Yang, B., and Sorscher, D. (1992). Regulation of terminal deoxynucleotidyl transferase gene expression in mice and men. Crit. Rev. Euk. Gene Express. 2, 237–250. Danska, J.S., Holland, D.P., Mariathasan, S., Williams, K.M., and Guidos, C.J. (1996). Biochemical and genetic defects in the DNAdependent protein kinase in murine scid lymphocytes. Mol. Cell. Biol. 16, 5507–5517. Eastman, Q.M., Leu, T.M.J., and Schatz, D.G. (1996). Initiation of

Feeney, A.J., Victor, K.D., Vu, K., Nadel, B., and Chukwuocha, R.U. (1994). Influence of the V(D)J recombination mechanism on the formation of the primary T and B cell repertoires. Semin. Immunol. 6, 155–163. Finnie, N.J., Gottlieb, T.M., Blunt, T., Jeggo, P.A., and Jackson, S.P. (1995). DNA-dependent protein kinase activity is absent in xrs-6 cells: implications for site-specific recombination and DNA doublestrand break repair. Proc. Natl. Acad. Sci. USA 92, 320–324. Fish, S.M., and Bosma, M.J. (1994). Abnormal deletions in the T cell receptor d locus of mouse thymocytes. Mol. Cell. Biol. 14, 4455– 4464. Gerstein, R.M., and Lieber, M.R. (1993). Extent to which homology can constrain coding exon junctional diversity in V(D)J recombination. Nature 363, 625–627. Getts, R.C., and Stamato, T.D. (1994). Absence of a Ku-like DNA end binding activity in the xrs double-strand DNA repair-deficient mutant. J. Biol. Chem. 269, 15981–15984. Giffin, W., Torrance, H., Rodda, D.J., Pre´fontaine, G.G., Pope, L., and Hache´, R.J.G. (1996). Sequence-specific DNA binding by Ku autoantigen and its effects on transcription. Nature 380, 265–268. Gilfillan, S., Dierich, A., Lemeur, M., Benoist, C., and Mathis, D. (1993). Mice lacking TdT: mature animals with an immature lymphocyte repertoire. Science 261, 1175–1178. Gilfillan, S., Benoist, C., and Mathis, D. (1995). Mice lacking terminal deoxynucleotidyl transferase: adult mice with a fetal antigen receptor repertoire. Immunol. Rev. 148, 201–219. Gottlieb, T.M., and Jackson, S.P. (1993). The DNA-dependent protein kinase: requirement for DNA ends and association with Ku antigen. Cell 72, 131–142. Gu, H., Fo¨rster, I., and Rajewsky, K. (1990). Sequence homologies, N sequence insertion and JH gene utilization in VHDJH joining: implications for the joining mechanism and the ontogenetic timing of Ly1 B cell and B-CLL progenitor generation. EMBO J. 9, 2133–2140. Han, J.-O., Steen, S.B., and Roth, D.B. (1997). Ku86 is not required for protection of signal ends or for formation of nonstandard V(D)J recombination products. Mol. Cell. Biol. 17, 2226–2234. Hiom, K., and Gellert, M. (1997). A stable RAG1–RAG2–DNA complex that is active in V(D)J cleavage. Cell 88, 65–72. Jackson, S.P., and Jeggo, P.A. (1995). DNA double-strand break repair and V(D)J recombination: involvement of DNA-PK. Trends Biochem. Sci. 20, 412–415. Kienker, L.J., Kuziel, W.A., and Tucker, P.W. (1991). T cell receptor g and d gene junctional sequences in scid mice: excessive P nucleotide insertion. J. Exp. Med. 174, 769–773. Komori, T., Okada, A., Stewart, V., and Alt, F.W. (1993). Lack of N regions in antigen receptor variable region genes of TdT-deficient lymphocytes. Science 261, 1171–1175. Lewis, S.M. (1994a). The mechanism of V(D)J joining: lessons from molecular, immunological and comparative analyses. Adv. Immunol. 56, 27–150. Lewis, S.M. (1994b). P nucleotide insertions and the resolution of hairpin DNA structures in mammalian cells. Proc. Natl. Acad. Sci. USA 91, 1332–1336. Lewis, S.M., and Hesse, J.E. (1991). Cutting and closing without recombination in V(D)J joining. EMBO J. 10, 3631–3639. Lewis, S.M., Hesse, J.E., Mizuuchi, K., and Gellert, M. (1988). Novel strand exchanges in V(D)J recombination. Cell 55, 1099–1107. Li, G.C., Yang, S.-H., Kim, D., Nussenzweig, A., Ouyang, H., Wei, J., Burgman, P., and Li, L. (1995). Suppression of heat-induced hsp70 expression by the 70-kDa subunit of the human Ku autoantigen. Proc. Natl. Acad. Sci. USA 92, 4512–4516. Li, Z., Otevrel, T., Gao, Y., Cheng, H.-L., Seed, B., Stamato, T.D., Taccioli, G.E., and Alt, F.W. (1995). The XRCC4 gene encodes a novel protein involved in DNA double-strand break repair and V(D)J recombination. Cell 83, 1079–1089.

V(D)J Recombination Products in Ku86-Deficient Mice 47

Liang, F., and Jasin, M. (1996). Ku80-deficient cells exhibit excess degradation of extrachromosomal DNA. J. Biol. Chem. 271, 14405– 14411.

A. (1993). Double-strand signal sequence breaks in V(D)J recombination are blunt, 59-phosphorylated, RAG-dependent, and cell cycle regulated. Genes Dev. 7, 2520–2532.

Lieber, M.R., Hesse, J.E., Lewis, S., Bosma, G.C., Rosenberg, N., Mizuuchi, K., Bosma, M.J., and Gellert, M. (1988a). The defect in murine severe combined immune deficiency: joining of signal sequences but not coding segments in V(D)J recombination. Cell 55, 7–16.

Schlissel, M.S., Corcoran, L.M., and Baltimore, D. (1991). Virustransformed pre-B cells show ordered activation but not inactivation of immunoglobulin gene rearrangement and transcription. J. Exp. Med. 173, 711–720.

Lieber, M.R., Hesse, J.E., Mizuuchi, K., and Gellert, M. (1988b). Lymphoid V(D)J recombination: nucleotide insertion at signal joints as well as coding joints. Proc. Natl. Acad. Sci. USA 85, 8588–8592. Lieber, M.R. (1992). The mechanism of V(D)J recombination: a balance of diversity, specificity, and stability. Cell 70, 873–876. McBlane, J.F., van Gent, D.C., Ramsden, D.A., Romeo, C., Cuomo, C.A., Gellert, M., and Oettinger, M.A. (1995). Cleavage at a V(D)J recombination signal requires only RAG1 and RAG2 proteins and occurs in two steps. Cell 83, 387–395. Meier, J.T., and Lewis, S.M. (1993). P nucleotides in V(D)J recombination: a fine-structure analysis. Mol. Cell. Biol. 13, 1078–1092. Mimori, T., and Hardin, J.A. (1986). Mechanism of interaction between Ku protein and DNA. J. Biol. Chem. 261, 10375–10379.

Schuler, W., Ruetsch, N.R., Amsler, M., and Bosma, M.J. (1991). Coding joint formation of endogenous T cell receptor genes in lymphoid cells from scid mice: unusual P-nucleotide additions in VJ-coding joints. Eur. J. Immunol. 21, 589–596. Sollbach, A.E., and Wu, G.E. (1995). Inversions produced during V(D)J rearrangement at IgH, the immunoglobulin heavy-chain locus. Mol. Cell. Biol. 15, 671–681. Steen, S.B., Gomelsky, L., and Roth, D.B. (1996). The 12/23 rule is enforced at the cleavage step of V(D)J recombination in vivo. Genes to Cells 1, 543–553. Taccioli, G.E., Rathbun, G., Oltz, E., Stamato, T., Jeggo, P.A., and Alt, F.W. (1993). Impairment of V(D)J recombination in double-strand break repair mutants. Science 260, 207–210.

Mizuta, R., Taccioli, G.E., and Alt, F.W. (1996). The V(D)J recombination defect in the xrs-6 cell line results from a point mutation in the Ku80 gene. Int. Immunol. 8, 1467–1471.

Taccioli, G.E., Gottlieb, T.M., Blunt, T., Priestley, A., Demengeot, J., Mizuta, R., Lehmann, A.R., Alt, F.W., Jackson, S.P., and Jeggo, P.A. (1994). Ku80: product of the XRCC5 gene and its role in DNA repair and V(D)J recombination. Science 265, 1442–1445.

Morozov, V.E., Falzon, M., Anderson, C.W., and Kuff, E.L. (1994). DNA-dependent protein kinase is activated by nicks and larger single-stranded gaps. J. Biol. Chem. 269, 16684–16688.

Tillinghast, J.P., Russell, J.H., Fields, L.E., and Loh, D.Y. (1989). Protein kinase C regulation of terminal deoxynucleotidyl transferase. J. Immunol. 143, 2378–2383.

Nussenzweig, A., Chen, C., da Costa Soares, V., Sanchez, M., Sokol, K., Nussenzweig, M.C., and Li, G.C. (1996). Requirement for Ku80 in growth and immunoglobulin V(D)J recombination. Nature 382, 551–555.

van Gent, D.C., Mizuuchi, K., and Gellert, M. (1996a). Similarities between initiation of V(D)J recombination and retroviral integration. Science 271, 1592–1594.

Paillard, S., and Strauss, F. (1991). Analysis of the mechanism of interaction of simian Ku protein with DNA. Nucleic Acids. Res. 19, 5619–5624. Pennycook, J.L., Chang, Y., Celler, J., Phillips, R.A., and Wu, G.E. (1993). High frequency of normal DJH joints in B cell progenitors in severe combined immunodeficiency mice. J. Exp. Med. 178, 1007– 1016. Pergola, F., Zdzienicka, M.Z., and Lieber, M.R. (1993). V(D)J recombination in mammalian cell mutants defective in DNA double-strand break repair. Mol. Cell. Biol. 13, 3464–3471. Rathmell, W.K., and Chu, G. (1994). A DNA end-binding factor involved in double-strand break repair and V(D)J recombination. Mol. Cell. Biol. 14, 4741–4748. Roth, D.B., Porter, T.N., and Wilson, J.H. (1985). Mechanisms of nonhomologous recombination in mammalian cells. Mol. Cell. Biol. 5, 2599–2607. Roth, D.B., Chang, X.-B., and Wilson, J.H. (1989). Comparison of filler DNA at immune, nonimmune, and oncogenic rearrangements suggests multiple mechanisms of formation. Mol. Cell. Biol. 9, 3049– 3057. Roth, D.B., Menetski, J.P., Nakajima, P.B., Bosma, M.J., and Gellert, M. (1992a). V(D)J recombination: broken DNA molecules with covalently sealed (hairpin) coding ends in scid mouse thymocytes. Cell 70, 983–991. Roth, D.B., Nakajima, P.B., Menetski, J.P., Bosma, M.J., and Gellert, M. (1992b). V(D)J recombination in mouse thymocytes: Doublestrand breaks near T cell receptor d rearrangement signals. Cell 69, 41–53. Roth, D.B., Zhu, C., and Gellert, M. (1993). Characterization of broken DNA molecules associated with V(D)J recombination. Proc. Natl. Acad. Sci. USA 90, 10788–10792. Roth, D.B., Lindahl, T., and Gellert, M. (1995). How to make ends meet. Curr. Biol. 5, 496–499. Sakano, H., Maki, R., Kurosawa, Y., Roeder, W., and Tonegawa, S. (1980). Two types of somatic recombination are necessary for the generation of complete immunoglobulin heavy-chain genes. Nature 286, 676–683. Schlissel, M., Constantinescu, A., Morrow, T., Baxter, M., and Peng,

van Gent, D.C., Ramsden, D.A., and Gellert, M. (1996b). The RAG1 and RAG2 proteins establish the 12/23 rule in V(D)J recombination. Cell 85, 107–113. van Gent, D.C., McBlane, J.F., Ramsden, D.A., Sadofsky, M.J., Hesse, J.E., and Gellert, M. (1995). Initiation of V(D)J recombination in a cell-free system. Cell 81, 925–934. VanDyk, L.F., Wise, T.W., Moore, B.B., and Meek, K. (1996). Immunoglobulin DH recombination signal sequence targeting. J. Immunol. 157, 4005–4015. Zhang, Y., Cado, D., Asarnow, D.M., Komori, T., Alt, F.W., Raulet, D.H., and Allison, J.P. (1995). The role of short homology repeats and TdT in generation of the invariant gd antigen receptor repertoire in the fetal thymus. Immunity 3, 439–447. Zhu, C., Bogue, M.A., Lim, D.-S., Hasty, P., and Roth, D.B. (1996). Ku86-deficient mice exhibit severe combined immunodeficiency and defective processing of V(D)J recombination intermediates. Cell 86, 379–389. Zhu, C., and Roth, D.B. (1995). Characterization of coding ends in thymocytes of scid mice: implications for the mechanism of V(D)J recombination. Immunity 2, 101–112.