DNA sequence and structure requirements for cleavage of V(D)J recombination signal sequences

Increased frequency of aberrant V(D)J recombination products in core RAG-expressing mice

RAG1 and RAG2 play a central role in V(D)J recombination, a process for antigen receptor gene assembly. The truncated 'core' regions of RAGs are sufficient to catalyze the recombination reaction, although with lower joining efficiency than full-length proteins. To investigate the role of the non-core regions of RAGs in the end-joining phase of antigen receptor rearrangement, we analyzed recombination products isolated from core RAG1 and core RAG2 knock-in mice. Here, we report that the truncation of RAGs increases the frequency of aberrant recombination in vivo. Signal joints (SJs) associated with V-to-D recombination of core RAG1 knock-in mice were normal, whereas those of core RAG2 knock-in mice were highly imprecise, containing large deletions and additions, and in some cases coding sequences. In contrast, we found an elevated level of imprecise D-to-J associated SJs for both core RAG1- and RAG2-expressing mice. Likewise, sequences of coding joints (CJs) were also affected by the expression of core RAGs. Finally, sequences found at the junctions of rearranged T-cell receptor loci were highly influenced by differences in rearranging recombination signal sequence pairs. We provide the first evidence that the non-core regions of RAGs have critical functions in the proper assembly and resolution of recombination intermediates in endogenous antigen receptor loci.

DNA cleavage activity of the V(D)J recombination protein RAG1 is autoregulated

RAG1 and RAG2 catalyze the first DNA cleavage steps in V(D)J recombination. We demonstrate that the isolated central domain of RAG1 has inherent single-stranded (ss) DNA cleavage activity, which does not require, but is enhanced by, RAG2. The central domain, therefore, contains the active-site residues necessary to perform hydrolysis of the DNA phosphodiester backbone. Furthermore, the catalytic activity of this domain on ss DNA is abolished by addition of the C-terminal domain of RAG1. The inhibitory effects of this latter domain are suppressed on substrates containing double-stranded (ds) DNA. Together, the activities of the reconstituted domains on ss versus mixed ds-ss DNA approximate the activity of intact RAG1 in the presence of RAG2. We propose how the combined actions of the RAG1 domains may function in V(D)J recombination and also in aberrant cleavage reactions that may lead to genomic instability in B and T lymphocytes.

RAG proteins shepherd double-strand breaks to a specific pathway, suppressing error-prone repair, but RAG nicking initiates homologous recombination

The two major pathways for repairing double-strand breaks (DSBs), homologous recombination and nonhomologous end joining (NHEJ), have traditionally been thought to operate in different stages of the cell cycle. This division of labor is not absolute, however, and precisely what governs the choice of pathway to repair a given DSB has remained enigmatic. We pursued this question by studying the site-specific DSBs created during V(D)J recombination, which relies on classical NHEJ to repair the broken ends. We show that mutations that form unstable RAG postcleavage complexes allow DNA ends to participate in both homologous recombination and the error-prone alternative NHEJ pathway. By abrogating a key function of the complex, these mutations reveal it to be a molecular shepherd that guides DSBs to the proper pathway. We also find that RAG-mediated nicks efficiently stimulate homologous recombination and discuss the implications of these findings for oncogenic chromosomal rearrangements, evolution, and gene targeting.

Utilization of Ig heavy chain variable, diversity, and joining gene segments in children with B-lineage acute lymphoblastic leukemia: implications for the mechanisms of VDJ recombination and for pathogenesis

Sequence analysis of the immunoglobulin heavy chain genes (IgH) has demonstrated preferential usage of specific variable (V), diversity (D), and joining (J) genes at different stages of B-cell development and in B-cell malignancies, and this has provided insight into B-cell maturation and selection. Knowledge of the association between rearrangement patterns based on updated databases and clinical characteristics of pediatric acute lymphoblastic leukemia (ALL) is limited. We analyzed 381 IgH sequences identified at presentation in 317 children with B-lineage ALL and assessed the V(H)D(H)J(H) gene utilization profiles. The D(H)J(H)-proximal V(H) segments and the D(H)2 gene family were significantly overrepresented. Only 21% of V(H)-J(H) joinings were potentially productive, a finding associated with a trend toward an increased risk of relapse. These results suggest that physical location at the V(H) locus is involved in preferential usage of D(H)J(H)-proximal V(H) segments whereas D(H) and J(H) segment usage is governed by position-independent molecular mechanisms. Molecular pathophysiology appears relevant to clinical outcome in patients who have only productive rearrangements, and specific rearrangement patterns are associated with differences in the tumor biology of childhood ALL.

Ordered DNA release and target capture in RAG transposition

Following V(D)J cleavage, the newly liberated DNA signal ends can be either fused together into a signal joint or used as donor DNA in RAG-mediated transposition. We find that both V(D)J cleavage and release of flanking coding DNA occur before the target capture step of transposition can proceed; no coding DNA is ever detected in the target capture complex. Separately from its role in V(D)J cleavage, the DDE motif of the RAG1/2 active site is specifically required for target DNA capture. The requirement for cleavage and release of coding DNA prior to either physical target binding or functional target commitment suggests that the RAG1/2 transposase contains a single binding site for non-RSS DNA that can accommodate either target DNA or coding DNA, but not both together. Perhaps the presence of coding DNA may aid in preventing transpositional resolution of V(D)J recombination intermediates.

DNA mismatches and GC-rich motifs target transposition by the RAG1/RAG2 transposase

In addition to their essential role in V(D)J recombination, the RAG proteins function as a transposase capable of inserting the V(D)J recombination intermediate, the signal end DNA fragment, into target DNA. RAG-mediated transposition has been suggested to contribute to genome instability and the development of lymphoid malignancies. Previous studies suggested that the RAG transposase exhibits a target site preference for GC rich sequences and hairpin structures. Here we demonstrate that a transposition hot spot (5'-GCCGCCGGGCC-3'), smaller portions of this hot spot and other GC rich motifs are able to target RAG-mediated transposition. Tracks of GC base pairs have been shown to have an unusually high rate of base pair breathing. Intriguingly, we find that DNA mismatches can efficiently target RAG-mediated transposition and suppress the use of other target sites. Hairpins, however, are not generally preferred targets. Our results indicate that target DNA melting may be a crucial step during RAG-mediated transposition, and that target site selection by the RAG transposase may be intimately linked to mutagenic and metabolic processes that transiently present favorable DNA structures to the transposition machinery.

V(D)J Recombination Frequencies Can Be Profoundly Affected by Changes in the Spacer Sequence

Each V, D, and J gene segment is flanked by a recombination signal sequence (RSS), composed of a conserved heptamer and nonamer separated by a 12- or 23-bp spacer. Variations from consensus in the heptamer or nonamer at specific positions can dramatically affect recombination frequency, but until recently, it had been generally held that only the length of the spacer, but not its sequence, affects the efficacy of V(D)J recombination. In this study, we show several examples in which the spacer sequence can significantly affect recombination frequencies. We show that the difference in spacer sequence alone of two V(H)S107 genes affects recombination frequency in recombination substrates to a similar extent as the bias observed in vivo. We show that individual positions in the spacer can affect recombination frequency, and those positions can often be predicted by their frequency in a database of RSS. Importantly, we further show that a spacer sequence that has an infrequently observed nucleotide at each position is essentially unable to support recombination in an extrachromosmal substrate assay, despite being flanked by a consensus heptamer and nonamer. This infrequent spacer sequence RSS shows only a 2-fold reduction of binding of RAG proteins, but the in vitro cleavage of this RSS is approximately 9-fold reduced compared with a good RSS. These data demonstrate that the spacer sequence should be considered to play an important role in the recombination efficacy of an RSS, and that the effect of the spacer occurs primarily subsequent to RAG binding.

Translocation and gross deletion breakpoints in human inherited disease and cancer I: Nucleotide composition and recombination-associated motifs

Translocations and gross deletions are important causes of both cancer and inherited disease. Such gene rearrangements are nonrandomly distributed in the human genome as a consequence of selection for growth advantage and/or the inherent potential of some DNA sequences to be frequently involved in breakage and recombination. Using the Gross Rearrangement Breakpoint Database [GRaBD; www.uwcm.ac.uk/uwcm/mg/grabd/grabd.html] (containing 397 germ-line and somatic DNA breakpoint junction sequences derived from 219 different rearrangements underlying human inherited disease and cancer), we have analyzed the sequence context of translocation and deletion breakpoints in a search for general characteristics that might have rendered these sequences prone to rearrangement. The oligonucleotide composition of breakpoint junctions and a set of reference sequences, matched for length and genomic location, were compared with respect to their nucleotide composition. Deletion breakpoints were found to be AT-rich whereas by comparison, translocation breakpoints were GC-rich. Alternating purine-pyrimidine sequences were found to be significantly over-represented in the vicinity of deletion breakpoints while polypyrimidine tracts were over-represented at translocation breakpoints. A number of recombination-associated motifs were found to be over-represented at translocation breakpoints (including DNA polymerase pause sites/frameshift hotspots, immunoglobulin heavy chain class switch sites, heptamer/nonamer V(D)J recombination signal sequences, translin binding sites, and the chi element) but, with the exception of the translin-binding site and immunoglobulin heavy chain class switch sites, none of these motifs were over-represented at deletion breakpoints. Alu sequences were found to span both breakpoints in seven cases of gross deletion that may thus be inferred to have arisen by homologous recombination. Our results are therefore consistent with a role for homologous unequal recombination in deletion mutagenesis and a role for nonhomologous recombination in the generation of translocations.

Effect of CpG methylation on RAG1/RAG2 reactivity: implications of direct and indirect mechanisms for controlling V(D)J cleavage

It has been suggested that DNA methylation/demethylation is involved in regulating V(D)J rearrangement. Although methylated DNA is thought to induce an inaccessible chromatin structure, it is unclear whether DNA methylation can directly control V(D)J recombination independently of chromatin structure. In this study, we tested whether DNA methylation directly affects the reactivity of the RAG1/RAG2 complex. Specific methylation within the heptamer of the recombination signal sequences (RSS) markedly reduced V(D)J cleavage without inhibiting RAG1/RAG2-DNA complex formation. By contrast, methylation at other positions around the RSS did not affect the reactivity of the RAG proteins. The presence of a methyl-CpG binding-domain protein inhibited the binding of the RAG1/RAG2 complex to all the methylated CpG sites that were tested. Our findings suggest that DNA methylation around the RSS may have a previously unexpected function in regulating V(D)J recombination by directly inhibiting V(D)J cleavage, in addition to its general function of inducing an inaccessible chromatin configuration.

Patterns of sequence conservation at termini of long terminal repeat (LTR) retrotransposons and DNA transposons in the human genome: lessons from phage Mu

Long terminal repeat (LTR) retrotransposons and DNA transposons are transposable elements (TEs) that perform cleavage and transfer at precise DNA positions. Here, we present statistical analyses of sequences found at the termini of precise TEs in the human genome. The results show that the terminal di- and trinucleotides of these TEs are highly conserved. 5'TG...CA3' occurs most frequently at the termini of LTR retrotransposons, while 5'CAG...CTG3' occurs most frequently in DNA transposons. Interestingly, these sequences are the most flexible base pair steps in DNA. Both the sequence preference and the degree of conservation of each position within the human LTR dinucleotide termini are remarkably similar to those experimentally demonstrated in transposable phage Mu. We discuss the significance of these observations and their implication for the function of terminal residues in the transposition of precise TEs.

The conserved CA/TG motif at Mu termini: T specifies stable transpososome assembly

The dinucleotide CA/TG found at the termini of transposable phage Mu occurs also at the termini of a large class of transposable elements, including HIV, all retroviruses and many retrotransposons. It was shown recently that mutations of this sequence block transpososome assembly, that A/T is more critical for activity than C/G, and that the hierarchy of reactivity of mutant termini follows closely the reported hierarchy of flexibility of their dinucleotide steps. In order to test the hypothesis that the terminal dinucleotide plays an essential structural role during "open termini" formation accompanying assembly, we have examined the activity of substrates carrying 100 different pairs of mismatched termini. Consistent with the flexibility hypothesis, we find that mismatched substrates are extremely efficient at assembly. A wild-type T residue on the bottom strand is essential for stable assembly, but the identity of the dinucleotide on the top strand is irrelevant for transposition chemistry. In addition, we have found a new rule for suppression of terminal defects by MuB protein, as well as a role for metal ions in DNA opening at the termini.

The central domain of core RAG1 preferentially recognizes single-stranded recombination signal sequence heptamer

RAG1 and RAG2 initiate V(D)J recombination by introducing DNA double strand breaks between each selected gene segment and its bordering recombination signal sequence (RSS) in a two-step mechanism in which the DNA is first nicked, followed by hairpin formation. The RSS consists of a conserved nonamer and heptamer sequence, in which the latter borders the site of DNA cleavage. A region within RAG1, referred to as the central domain (residues 528-760 of 1040 in the full-length protein), has been shown previously to bind specifically to the double-stranded (ds) RSS heptamer, but with both weak specificity and affinity. However, additional investigations into the RAG1-RSS heptamer interaction are required because the DNA substrate forms intermediate conformations during the V(D)J recombination reaction. These include the nicked and hairpin products, as well as likely base unpairing to produce single-stranded (ss) DNA near the cleavage site. Here, it was determined that although the central domain showed substantially higher binding affinity for ss and nicked versus ds substrate, the interaction with ss RSS was particularly robust. In addition, the central domain bound with greater sequence specificity to the ss RSS heptamer than to the ds form. This study provides important insight into the V(D)J recombination reaction, specifically that significant interaction of the RSS heptamer with RAG1 occurs only after the induction of conformational changes at the RSS heptamer.

The C-terminal portion of RAG2 protects against transposition in vitro

The assembly of antigen receptor genes by V(D)J recombination is initiated by the RAG1/RAG2 protein complex, which introduces double-strand breaks between recombination signal sequences and their coding DNA. Truncated forms of RAG1 and RAG2 are functional in vivo and have been used to study V(D)J cleavage, hybrid joint formation and transposition in vitro. Here we have characterized the activities of the full-length proteins. Unlike core RAG2, which supports robust transposition in vitro, full-length RAG2 blocks transposition of signal ends following V(D)J cleavage. Thus, one role of this non-catalytic domain may be to prevent transposition in developing lymphoid cells. Although full-length RAG1 and RAG2 proteins rarely form hybrid joints in vivo in the absence of non-homologous end-joining factors, we show that the full-length proteins alone can catalyze this reaction in vitro.

V(D)J recombination: RAG proteins, repair factors, and regulation

V(D)J recombination is the specialized DNA rearrangement used by cells of the immune system to assemble immunoglobulin and T-cell receptor genes from the preexisting gene segments. Because there is a large choice of segments to join, this process accounts for much of the diversity of the immune response. Recombination is initiated by the lymphoid-specific RAG1 and RAG2 proteins, which cooperate to make double-strand breaks at specific recognition sequences (recombination signal sequences, RSSs). The neighboring coding DNA is converted to a hairpin during breakage. Broken ends are then processed and joined with the help of several factors also involved in repair of radiation-damaged DNA, including the DNA-dependent protein kinase (DNA-PK) and the Ku, Artemis, DNA ligase IV, and Xrcc4 proteins, and possibly histone H2AX and the Mre11/Rad50/Nbs1 complex. There may be other factors not yet known. V(D)J recombination is strongly regulated by limiting access to RSS sites within chromatin, so that particular sites are available only in certain cell types and developmental stages. The roles of enhancers, histone acetylation, and chromatin remodeling factors in controlling accessibility are discussed. The RAG proteins are also capable of transposing RSS-ended fragments into new DNA sites. This transposition helps to explain the mechanism of RAG action and supports earlier proposals that V(D)J recombination evolved from an ancient mobile DNA element.

Common physical properties of DNA affecting target site selection of sleeping beauty and other Tc1/mariner transposable elements

Sleeping Beauty (SB) is the most active Tc1/mariner-type transposable element in vertebrates, and is therefore a valuable vector for transposon mutagenesis in vertebrate models and for human gene therapy. We have analyzed factors affecting target site selection of SB in mammalian cells, by generating transposition events from extrachromosomal plasmids to chromosomes. In contrast to the local hopping observed when transposition is induced from a chromosomal context, mapping of 138 unique SB insertions on human chromosomes showed a fairly random genomic distribution, and a 35% occurrence of transposition into genes. Inspection of the DNA flanking the sites of element integration revealed significant differences from random DNA in both primary sequence and physical properties. The consensus sequence of SB target sites was found to be a palindromic AT-repeat, ATATATAT, in which the central TA is the canonical target site. We found however, that target site selection is determined primarily on the level of DNA structure, and not by specific base-pair interactions. Computational analyses revealed that insertion sites tend to have a bendable structure and a palindromic pattern of potential hydrogen-bonding sites in the major groove of the DNA. These features appear conserved in the Tc1/mariner family of transposons and in other, distantly related elements that share a common catalytic domain of the transposase, and integrate fairly randomly. No similar target site preference was found for non-randomly integrating elements. Our results suggest common factors influencing target site selection of a wide range of transposable elements.

Mutational analysis of all conserved basic amino acids in RAG-1 reveals catalytic, step arrest, and joining-deficient mutants in the V(D)J recombinase

Although both RAG-1 and RAG-2 are required for all steps of V(D)J recombination, little is known about the specific contribution of either protein to these steps. RAG-1 contains three acidic active-site amino acids that are thought to coordinate catalytic metal ions. To search for additional catalytic amino acids and to better define the functional anatomy of RAG-1, we mutated all 86 conserved basic amino acids to alanine and evaluated the mutant proteins for DNA binding, nicking, hairpin formation, and joining. We found several amino acids outside of the canonical nonamer-binding domain that are critical for DNA binding, several step arrest mutants with defects in nicking or hairpin formation, and four RAG-1 mutants defective specifically for joining. Analysis of coding joints formed by some of these mutants revealed excessive deletions, frequent use of short sequence homologies, and unusually long palindromic junctional inserts, known as P nucleotides, that result from aberrant hairpin opening. These features characterize junctions found in scid mice, which are deficient for the catalytic subunit of DNA-dependent protein kinase (DNA-PKcs), suggesting that the RAG proteins and DNA-PKcs perform overlapping functions in coding joint formation. Interestingly, the amino acids that are altered in 12 of our mutants are also mutated in human inherited immunodeficiency syndromes. Our analysis of these mutants provides insights into the molecular mechanisms underlying these disorders.

Targeted transposition by the V(D)J recombinase

Cleavage by the V(D)J recombinase at a pair of recombination signal sequences creates two coding ends and two signal ends. The RAG proteins can integrate these signal ends, without sequence specificity, into an unrelated target DNA molecule. Here we demonstrate that such transposition events are greatly stimulated by--and specifically targeted to--hairpins and other distorted DNA structures. The mechanism of target selection by the RAG proteins thus appears to involve recognition of distorted DNA. These data also suggest a novel mechanism for the formation of alternative recombination products termed hybrid joints, in which a signal end is joined to a hairpin coding end. We suggest that hybrid joints may arise by transposition in vivo and propose a new model to account for some recurrent chromosome translocations found in human lymphomas. According to this model, transposition can join antigen receptor loci to partner sites that lack recombination signal sequence elements but bear particular structural features. The RAG proteins are capable of mediating all necessary breakage and joining events on both partner chromosomes; thus, the V(D)J recombinase may be far more culpable for oncogenic translocations than has been suspected.

Assembly of the RAG1/RAG2 synaptic complex

Assembly of antigen receptor genes by V(D)J recombination requires the site-specific recognition of two distinct DNA elements differing in the length of the spacer DNA that separates two conserved recognition motifs. Under appropriate conditions, V(D)J cleavage by the purified RAG1/RAG2 recombinase is similarly restricted. Double-strand breakage occurs only when these proteins are bound to a pair of complementary signals in a synaptic complex. We examine here the binding of the RAG proteins to signal sequences and find that the full complement of proteins required for synapsis of two signals and coupled cleavage can assemble on a single signal. This complex, composed of a dimer of RAG2 and at least a trimer of RAG1, remains inactive for double-strand break formation until a second complementary signal is provided. Thus, binding of the second signal activates the complex, possibly by inducing a conformational change. If synaptic complexes are formed similarly in vivo, one signal of a recombining pair may be the preferred site for RAG1/RAG2 assembly.

Valpha gene replacement in a TCRalpha knock-in mouse

Using a TCRalpha chain knock-in mouse, we demonstrate that V-gene replacement can operate in the T cell receptor alpha locus. Functional TCRalpha chain transcripts generated by Valpha-gene replacement at the site of the Valpha-embedded heptamer were identified in splenic T cells. This finding shows that Valpha-gene replacement can likely be used to shape the peripheral T cell repertoire. The conservation of the embedded heptamer in most Valpha segments adds support to the notion that V-gene replacement is a mechanism maintained to diversify the immune system and that argues that it is common to B and T cells.

Role of DNA end distortion in catalysis by avian sarcoma virus integrase

Retroviral integrase (IN) recognizes linear viral DNA ends and introduces nicks adjacent to a highly conserved CA dinucleotide usually located two base pairs from the 3'-ends of viral DNA (the "processing" reaction). In a second step, the same IN active site catalyzes the insertion of these ends into host DNA (the "joining" reaction). Both DNA sequence and DNA structure contribute to specific recognition of viral DNA ends by IN. Here we used potassium permanganate modification to show that the avian sarcoma virus IN catalytic domain is able to distort viral DNA ends in vitro. This distortion activity is consistent with both unpairing and unstacking of the three terminal base pairs, including the processing site adjacent to the conserved CA. Furthermore, the introduction of mismatch mutations that destabilize the viral DNA ends were found to stimulate the IN processing reaction as well as IN-mediated distortion. End-distortion activity was also observed with mutant or heterologous DNA substrates. However, further analyses showed that using Mn(2+) as a cofactor, processing site specificity of these substrates was also maintained. Our results support a model whereby unpairing and unstacking of the terminal base pairs is a required step in the processing reaction. Furthermore, these results are consistent with our previous observations indicating that unpairing of target DNA promotes the joining reaction.

Determinants for hairpin formation in Tn10 transposition

Tn10 transposition involves the formation of a hairpin intermediate at the transposon termini. Here we show that hairpin formation exhibits more stringent DNA sequence requirements at the terminal two base pairs than either transpososome assembly or first strand nicking. We also observe a significant DNA distortion at the terminal base pairs upon transpososome assembly by chemical nuclease footprinting. Interestingly, mutations at these positions do not necessarily inhibit the formation of the distortion. However, it remains a possibility that the inhibitory effect of these mutations is due to a defect in protein-DNA interactions subsequent to this deformation. Terminal base pair mutations also inhibited strand transfer, providing evidence that transposase interactions with the terminal residues on both 'transferred' and 'non-transferred' strands are important for hairpin formation. We also demonstrate that mutation of a highly conserved tyrosine residue that is a component of the YREK motif, Y285, results in a phenotype comparable to that of the terminal base pair mutations. In contrast, a mutation at another conserved position, W265, is shown to relax the specificity of the hairpin formation reaction.

The RAG proteins in V(D)J recombination: more than just a nuclease

V(D)J recombination is the process that generates the diversity among T cell receptors and is one of three mechanisms that contribute to the diversity of antibodies in the vertebrate immune system. The mechanism requires precise cutting of the DNA at segment boundaries followed by rejoining of particular pairs of the resulting termini. The imprecision of aspects of the joining reaction contributes significantly to increasing the variability of the resulting functional genes. Signal sequences target DNA recombination and must participate in a highly ordered protein-DNA complex in order to limit recombination to appropriate partners. Two proteins, RAG1 and RAG2, together form the nuclease that cleaves the DNA at the border of the signal sequences. Additional roles of these proteins in organizing the reaction complex for subsequent steps are explored.

A C-terminal region of RAG1 contacts the coding DNA during V(D)J recombination

The site-specific DNA rearrangement process, called V(D)J recombination, creates much of the diversity of immune receptor molecules in the adaptive immune system. Central to this reaction is the organization of the protein-DNA complex containing the proteins RAG1 and RAG2 and their DNA targets. A long-term goal is to appreciate the three-dimensional relationships between the proteins and DNA that allow the assembly of the appropriate reaction intermediates, resulting in concerted cleavage and directed rejoining of the DNA ends. Previous cross-linking approaches have mapped RAG1 contacts on the DNA. RAG1 protein contacts the DNA at the conserved heptamer and nonamer sequences as well as at the coding DNA adjacent to the heptamer. Here we subject RAG1, covalently cross-linked to DNA substrates, to partial cyanogen bromide degradation or trypsin proteolysis in order to map contacts on the protein. We find that coding-sequence contacts occur near the C terminus of RAG1, while contacts made within the recombination signal sequence occur nearer the N terminus of the core region of RAG1. A deletion protein lacking the C-terminal DNA-contacting region is still capable of making the N-terminal contacts. This suggests that the two binding interactions may exist on two separate domains of the protein. A trypsin cleavage pattern of the native protein supports this conclusion. A two-domain model for RAG1 is evaluated with respect to the larger recombination complex.

Molecular characterization of the genomic breakpoint junction in the t(11;18) (q21;q21) translocation of a gastric MALT lymphoma

The t(11;18) translocation, fusing the API2 and MALT1 genes, is one of the most frequent chromosomal translocations associated with mucosa-associated lymphoid tissue (MALT) lymphomas. The translocation breakpoints have been cloned and characterized at the mRNA sequence level. Although the genomic organization of the API2 gene has been described, hitherto the genomic sequence of MALT1 remains unknown. To gain some insight into the mechanism that generates this translocation, we cloned and sequenced an API2-MALT1 fused transcript as well as genomic DNA of the t(11;18) translocation from a MALT lymphoma. We localized the API2 breakpoint within intron 7. Nucleotide sequence analysis revealed that the genomic breakpoint junction possesses the consensus heptamers of immunoglobulin V(D)J recombination signal sequences, all the matches being completely present on the API2 allele and five of seven matches on the MALT1 allele. These data suggest that the translocation in the MALT lymphoma might have been mediated in part by an aberrant V(D)J recombination event.

Separation-of-function mutants reveal critical roles for RAG2 in both the cleavage and joining steps of V(D)J recombination

The only established physiological function of the V(D)J recombinase, comprising RAG1 and RAG2, is to perform DNA cleavage. The molecular roles of RAG2 in cleavage, the mechanisms used to join the broken DNA ends, and the identity of nuclease(s) that open the hairpin coding ends have been unknown. Site-directed mutagenesis targeting each conserved basic amino acid in RAG2 revealed several separation-of-function mutants that address these questions. Analysis of these mutants reveals that RAG2 helps recognize or cleave distorted DNA intermediates and plays an essential role in the joining step of V(D)J recombination. Moreover, the discovery that some mutants block RAG-mediated hairpin opening in vitro provides a critical link between this biochemical activity and coding joint formation in vivo.

Conditional RAG-1 mutants block the hairpin formation step of V(D)J recombination

Hairpin formation serves an important regulatory role in V(D)J recombination because it requires synapsis of an appropriate pair of recombination sites. How hairpin formation is regulated and which regions of the RAG proteins perform this step remain unknown. We analyzed two conditional RAG-1 mutants that affect residues quite close in the primary sequence to an active site amino acid (D600), and we found that they exhibit severely impaired recombination in the presence of certain cleavage site sequences. These mutants are specifically defective for the formation of hairpins, providing the first identification of a region of the V(D)J recombinase necessary for this reaction. Substrates containing mismatched bases at the cleavage site rescued hairpin formation by both mutants, which suggests that the mutations affect the generation of a distorted or unwound DNA intermediate that has been implicated in hairpin formation. Our results also indicate that this region of RAG-1 may be important for coupling hairpin formation to synapsis.

The RAG proteins and V(D)J recombination: complexes, ends, and transposition

V(D)J recombination proceeds through a series of protein:DNA complexes mediated in part by the RAG1 and RAG2 proteins. These proteins are responsible for sequence-specific DNA recognition and DNA cleavage, and they appear to perform multiple postcleavage roles in the reaction as well. Here we review the interaction of the RAG proteins with DNA, the chemistry of the cleavage reaction, and the higher order complexes in which these events take place. We also discuss postcleavage functions of the RAG proteins, including recent evidence indicating that they initiate the process of coding end processing by nicking hairpin DNA termini. Finally, we discuss the evolutionary and functional implications of the finding that RAG1 and RAG2 constitute a transposase, and we consider RAG protein biochemistry in the context of several bacterial transposition systems. This suggests a model of the RAG protein active site in which two divalent metal ions serve alternating and opposite roles as activators of attacking hydroxyl groups and stabilizers of oxyanion leaving groups.

Mutations of acidic residues in RAG1 define the active site of the V(D)J recombinase

The RAG1 and RAG2 proteins collaborate to initiate V(D)J recombination by binding recombination signal sequences (RSSs) and making a double-strand break between the RSS and adjacent coding DNA. Like the reactions of their biochemical cousins, the bacterial transposases and retroviral integrases, cleavage by the RAG proteins requires a divalent metal ion but does not involve a covalent protein/DNA intermediate. In the transposase/integrase family, a triplet of acidic residues, commonly called a DDE motif, is often found to coordinate the metal ion used for catalysis. We show here that mutations in each of three acidic residues in RAG1 result in mutant derivatives that can bind the RSS but whose ability to catalyze either of the two chemical steps of V(D)J cleavage (nicking and hairpin formation) is severely impaired. Because both chemical steps are affected by the same mutations, a single active site appears responsible for both reactions. Two independent lines of evidence demonstrate that at least two of these acidic residues are directly involved in coordinating a divalent metal ion: The substitution of Cys for Asp allows rescue of some catalytic function, whereas an alanine substitution is no longer subject to iron-induced hydroxyl radical cleavage. Our results support a model in which the RAG1 protein contains the active site of the V(D)J recombinase and are interpreted in light of predictions about the structure of RAG1.

The RAG1/RAG2 complex constitutes a 3' flap endonuclease: implications for junctional diversity in V(D)J and transpositional recombination

During V(D)J recombination, processing of branched coding end intermediates is essential for generating junctional diversity. Here, we report that the RAG1/ RAG2 recombinase is a 3' flap endonuclease. Substrates of this nuclease activity include various coding end intermediates, suggesting a direct role for RAG1/ RAG2 in generating junctional diversity during V(D)J recombination. Evidence is also provided indicating that site-specific RSS nicking involves RAG1/RAG2-mediated processing of a localized flap-like structure, implying 3' flap nicking in multiple DNA processing reactions. We have also demonstrated that the bacterial transposase Tn10 contains a 3' flap endonuclease activity, suggesting a mechanistic parallel between RAG1/RAG2 and other transposases. Based on these data, we propose that numerous transposases may facilitate genomic evolution by removing single-stranded extensions during the processing of excision site junctions.

Mechanistic basis for coding end sequence effects in the initiation of V(D)J recombination

V(D)J recombination is directed by recombination signal sequences. However, the flanking coding end sequence can markedly affect the frequency of the initiation of V(D)J recombination in vivo. Here we demonstrate that the coding end sequence effect can be qualitatively and quantitatively recapitulated in vitro with purified RAG proteins. We find that coding end sequence specifically affects the nicking step, which is the first biochemical step in RAG-mediated cleavage. The subsequent hairpin formation step is not affected by the coding end sequence. Furthermore, the coding end sequence effect can be ablated by prenicking the substrate, indicating that the coding end effect is specific to the nicking step. In reactions in which both 12- and 23-substrates are present, a suboptimal coding end sequence on one signal can slow down hairpin formation at the partner signal, a result consistent with models in which coordination between the signals occurs at the hairpin formation step. The coding end sequence effect on nicking and the coupling of the 12- and 23-substrates explains how hairpin formation can be rate limiting for some 12/23 pairs, whereas nicking can be rate limiting when low-efficiency coding end sequences are involved.

Biased VH Gene Usage in Early Lineage Human B Cells: Evidence for Preferential Ig Gene Rearrangement in the Absence of Selection

Certain VH genes are predominantly expressed in mature B cells. We hypothesized that several, mutually nonexclusive VH-dependent mechanisms operating at distinct stages during B cell development may be responsible for overrepresentation of these VH genes. In the present study, we have assessed whether one of the mechanisms involves preferential rearrangement at the pro-B cell stage. The frequency of individual VH4 and VH3 genes in rearrangement libraries from FACS-purified human CD34+/CD19+ pro-B and CD34−/CD19+ pre-B cells was assessed. The in-frame and out-of-frame rearrangements from both cell populations were analyzed using a high resolution PAGE system. The frequencies of individual VH gene segments among out-of-frame rearrangements from pro-B cells were determined, because these frequencies should reflect only processes before the translation of the μ-heavy chain and should not be biased by selection mechanisms. Our results demonstrate that, at the pro-B cell stage, the V4–34, V4–39, and V4–59 gene segments are the most frequently rearranged VH4 family genes, and the V3–23 and V3–30 gene segments are the most frequently rearranged VH3 family genes. This finding suggests that the predominant expression of these VH genes in peripheral mature B cells is determined to a significant degree by their preferential rearrangement during V-DJ recombination.

The role of components of recombination signal sequences in immunoglobulin gene segment usage: a V81x model

It has long been appreciated that some immunoglobulin (and T-cell receptor) gene segments are used much more frequently than others. The VHsegment V81x is a particularly striking case of overusage. Its usage varies with the stage of B-cell development and with the strain of mice, but it is always high in B cell progenitors. We have found that the coding sequence and the recombination signal sequences (RSS) are identical in five mouse strains, including CAST/Ei, a strain derived from the species Mus castaneus. Thus, the strain differences cannot be attributed to sequences within V81x itself. V81x RSS mediated recombination at rates significantly higher than another VHRSS. Although the V81x nonamer differs at one base pair from the consensus sequence, an RSS with this nonamer and a consensus heptamer recombines as well as the consensus RSS. When the V81x spacer is replaced by that of VA1, the frequency of recombination decreases by approximately 5-fold; thus, the contribution of variation in natural spacers to variability in VHusage in vivo is likely to be more than has been previously appreciated. Furthermore, the contribution of the heptamer and nonamer to differential VHusage in our assay is correlated inversely with their conservation throughout the VHlocus.

Detection of RAG protein-V(D)J recombination signal interactions near the site of DNA cleavage by UV cross-linking

V(D)J recombination is initiated by double-strand cleavage at recombination signal sequences (RSSs). DNA cleavage is mediated by the RAG1 and RAG2 proteins. Recent experiments describing RAG protein-RSS complexes, while defining the interaction of RAG1 with the nonamer, have not assigned contacts immediately adjacent to the site of DNA cleavage to either RAG polypeptide. Here we use UV cross-linking to define sequence- and site-specific interactions between RAG1 protein and both the heptamer element of the RSS and the coding flank DNA. Hence, RAG1-DNA contacts span the site of cleavage. We also detect cross-linking of RAG2 protein to some of the same nucleotides that cross-link to RAG1, indicating that, in the binding complex, both RAG proteins are in close proximity to the site of cleavage. These results suggest how the heptamer element, the recognition surface essential for DNA cleavage, is recognized by the RAG proteins and have implications for the stoichiometry and active site organization of the RAG1-RAG2-RSS complex.

The complete nucleotide sequence of the human immunoglobulin heavy chain variable region locus

The complete nucleotide sequence of the 957-kb DNA of the human immunoglobulin heavy chain variable (VH) region locus was determined and 43 novel VH segments were identified. The region contains 123 VH segments classifiable into seven different families, of which 79 are pseudogenes. Of the 44 VH segments with an open reading frame, 39 are expressed as heavy chain proteins and 1 as mRNA, while the remaining 4 are not found in immunoglobulin cDNAs. Combinatorial diversity of VH region was calculated to be approximately 6,000. Conservation of the promoter and recombination signal sequences was observed to be higher in functional VH segments than in pseudogenes. Phylogenetic analysis of 114 VH segments clearly showed clustering of the VH segments of each family. However, an independent branch in the tree contained a single VH, V4-44.1P, sharing similar levels of homology to human VH families and to those of other vertebrates. Comparison between different copies of homologous units that appear repeatedly across the locus clearly demonstrates that dynamic DNA reorganization of the locus took place at least eight times between 133 and 10 million years ago. One nonimmunoglobulin gene of unknown function was identified in the intergenic region.

Antigen receptor engagement turns off the V(D)J recombination machinery in human tonsil B cells

The germinal center (GC) is an anatomic compartment found in peripheral lymphoid organs, wherein B cells undergo clonal expansion, somatic mutation, switch recombination, and reactivate immunoglobulin gene V(D)J recombination. As a result of somatic mutation, some GC B cells develop higher affinity antibodies, whereas others suffer mutations that decrease affinity, and still others may become self-reactive. It has been proposed that secondary V(D)J rearrangements in GCs might rescue B cells whose receptors are damaged by somatic mutations. Here we present evidence that mature human tonsil B cells coexpress conventional light chains and recombination associated genes, and that they extinguish recombination activating gene and terminal deoxynucleotidyl transferase expression when their receptors are cross-linked. Thus, the response of the recombinase to receptor engagement in peripheral B cells is the opposite of the response in developing B cells to the same stimulus. These observations suggest that receptor revision is a mechanism for receptor diversification that is turned off when antigen receptors are cross-linked by the cognate antigen.

Dual role of RAG2 in V(D)J recombination: catalysis and regulation of ordered Ig gene assembly

Immunoglobulin genes are assembled during lymphoid development by a series of site-specific rearrangements that are tightly regulated to ensure that functional antibodies are generated in B (but not T) cells and that a unique receptor is present on each cell. Because a common V(D)J recombinase comprising RAG1 and RAG2 proteins is used for both B- and T-cell antigen receptor assembly, lineage-specific rearrangement must be modulated through differential access to sites of recombination. We show here that the C-terminus of the RAG2 protein, although dispensable for the basic recombination reaction and for Ig heavy chain DH to JH joining, is essential for efficient VH to DJH rearrangement at the IgH locus. Thus, the RAG2 protein plays a dual role in V(D)J recombination, acting both in catalysis of the reaction and in governing access to particular loci.

Distinct roles of RAG1 and RAG2 in binding the V(D)J recombination signal sequences

The RAG1 and RAG2 proteins initiate V(D)J recombination by introducing double-strand breaks at the border between a recombination signal sequence (RSS) and a coding segment. To understand the distinct functions of RAG1 and RAG2 in signal recognition, we have compared the DNA binding activities of RAG1 alone and RAG1 plus RAG2 by gel retardation and footprinting analyses. RAG1 exhibits only a three- to fivefold preference for binding DNA containing an RSS over random sequence DNA. Although direct binding of RAG2 by itself was not detected, the presence of both RAG1 and RAG2 results in the formation of a RAG1-RAG2-DNA complex which is more stable and more specific than the RAG1-DNA complex and is active in V(D)J cleavage. These results suggest that biologically effective discrimination between an RSS and nonspecific sequences requires both RAG1 and RAG2. Unlike the binding of RAG1 plus RAG2, RAG1 can bind to DNA in the absence of a divalent metal ion and does not require the presence of coding flank sequence. Footprinting of the RAG1-RAG2 complex with 1, 10-phenanthroline-copper and dimethyl sulfate protection reveal that both the heptamer and the nonamer are involved. The nonamer is protected, with extensive protein contacts within the minor groove. Conversely, the heptamer is rendered more accessible to chemical attack, suggesting that binding of RAG1 plus RAG2 distorts the DNA near the coding/signal border.

Functional analysis of coordinated cleavage in V(D)J recombination

V(D)J recombination in vivo requires a pair of signals with distinct spacer elements of 12 and 23 bp that separate conserved heptamer and nonamer motifs. Cleavage in vitro by the RAG1 and RAG2 proteins can occur at individual signals when the reaction buffer contains Mn2+, but cleavage is restricted to substrates containing two signals when Mg2+ is the divalent cation. By using a novel V(D)J cleavage substrate, we show that while the RAG proteins alone establish a moderate preference for a 12/23 pair versus a 12/12 pair, a much stricter dependence of cleavage on the 12/23 signal pair is produced by the inclusion of HMG1 and competitor double-stranded DNA. The competitor DNA serves to inhibit the cleavage of substrates carrying a 12/12 or 23/23 pair, as well as the cutting at individual signals in 12/23 substrates. We show that a 23/33 pair is more efficiently recombined than a 12/33 pair, suggesting that the 12/23 rule can be generalized to a requirement for spacers that differ from each other by a single helical turn. Furthermore, we suggest that a fixed spatial orientation of signals is required for cleavage. In general, the same signal variants that can be cleaved singly can function under conditions in which a signal pair is required. However, a chemically modified substrate with one noncleavable signal enables us to show that formation of a functional cleavage complex is mechanistically separable from the cleavage reaction itself and that although cleavage requires a pair of signals, cutting does not have to occur simultaneously at both. The implications of these results are discussed with respect to the mechanism of V(D)J recombination and the generation of chromosomal translocations.

V(D)J recombination signal recognition: distinct, overlapping DNA-protein contacts in complexes containing RAG1 with and without RAG2

Protein interactions with V(D)J recombination signal sequences (RSSs) were mapped in complexes containing RAG1 with (M1/2) or without (M1) RAG2. In both complexes, RAG interactions with the DNA backbone are biased toward one side of the helix; nonamer contacts resemble those of Hin with hixL. In the M1 complex, DNA contacts are centered on the nonamer. In the M1/2 complex, protein-RSS interactions extend through the spacer and into the nonamer-proximal portion of the heptamer. Chemical modifications near the heptamer-coding junction are overrepresented in the M1/2 complex, providing evidence for perturbation of DNA structure in this region. Thus, while RAG1 alone can bind the nonamer, RAG2 is required for heptamer occupancy.

The effect of Me2+ cofactors at the initial stages of V(D)J recombination

V(D)J site-specific recombination mediates the somatic assembly of the antigen receptor gene segments. This process is initiated by the recombination activating proteins RAG1 and RAG2, which recognize the recombination signal sequences (RSS) and cleave the DNA at the coding/RSS junction. In this study, we show that RAG1 and RAG2 have the ability to directly interact in solution before binding to the DNA. RAG1 forms a homodimer, which leads to the appearance of two distinct RAG1.RAG2 complexes bound to DNA. To investigate the properties of the two RAG1.RAG2 complexes in the presence of different Me2+ cofactors, we established an in vitro Mg2+-based cleavage reaction on a single RSS. Using this system, we found that Mg2+ confers a specific pattern of DNA binding and cleavage. In contrast, Mn2+ allows aberrant binding of RAG1.RAG2 to single-stranded RSS and permits cleavage independent of binding to the nonamer. To determine the contribution of Me2+ ions at the early stages of V(D)J recombination, we analyzed specific DNA recognition and cleavage by RAG1.RAG2 on phosphorothioated substrates. These experiments revealed that Me2+ ions directly coordinate the binding of RAG1.RAG2 to the RSS DNA.

Assembly of a 12/23 paired signal complex: a critical control point in V(D)J recombination

The 12/23 rule requires that V(D)J recombination only occurs between recombination signals with 12 and 23 base pair spacers. We show that the 12/23 rule is established prior to DNA cleavage, by the formation of a synaptic complex containing both 12-spacer and 23-spacer signals. The RAG1 and RAG2 proteins, as well as the DNA bending protein HMG1, are needed for efficient formation of this complex. We show further that the synaptic complex is the functional complex for coupled cleavage. After cleavage, all four broken DNA ends remain associated with the RAG proteins in a postcleavage synaptic complex, whose existence helps to explain the known role of RAG1 and RAG2 in the subsequent end-joining events that complete V(D)J recombination.

Transposase makes critical contacts with, and is stimulated by, single-stranded DNA at the P element termini in vitro

P elements transpose by a cut-and-paste mechanism. Donor DNA cleavage mediated by transposase generates 17 nucleotide (nt) 3' single-strand extensions at the P element termini which, when present on oligonucleotide substrates, stimulate both the strand-transfer and disintegration reactions in vitro. A significant amount of the strand-transfer products are the result of double-ended integration. Chemical DNA modification-interference experiments indicate that during the strand-transfer reaction, P element transposase contacts regions of the substrate DNA that include the transposase binding site and the duplex portion of the 31 bp inverted repeat, as well as regions of the terminal 17 nt single-stranded DNA. Together these data suggest that the P element transposase protein contains two DNA-binding sites and that the active oligomeric form of the transposase protein is at least a dimer.

RAG-1 and RAG-2-dependent assembly of functional complexes with V(D)J recombination substrates in solution

V(D)J recombination is initiated by RAG-1 and RAG-2, which introduce double-strand DNA breaks at recombination signal sequences (RSSs) of antigen receptor gene segments to produce signal ends, terminating in blunt, double-strand breaks, and coding ends, terminating in DNA hairpins. While the formation of RAG-RSS complexes has been documented, observations regarding the individual contributions of RAG-1 and RAG-2 to RSS recognition are in conflict. Here we describe an assay for formation and maintenance of functional RAG-RSS complexes in the course of the DNA cleavage reaction. Under conditions of in vitro cleavage, the RAG proteins sequester intact substrate DNA in a stable complex which is formed prior to strand scission. The cleavage reaction subsequently proceeds through nicking and hairpin formation without dissociation of substrate. Notably, the presence of both RAG-1 and RAG-2 is essential for formation of stable, functional complexes with substrate DNA under conditions of the sequestration assay. Two classes of substrate mutation are distinguished by their effects on RAG-mediated DNA cleavage in vitro. A mutation of the first class, residing within the RSS nonamer and associated with coordinate impairment of nicking and hairpin formation, greatly reduces the stability of RAG association with intact substrate DNA. In contrast, a mutation of the second class, lying within the RSS heptamer and associated with selective abolition of hairpin formation, has little or no effect on the half-life of the RAG-substrate complex.

V(D)J recombination: in vitro coding joint formation

Antigen receptor genes are assembled through a mechanism known as V(D)J recombination, which involves two different joining reactions: signal and coding joining. Formation of these joints is essential for antigen receptor assembly as well as maintaining chromosomal integrity. Here we report on a cell-free system for coding joint formation using deletion and inversion recombination substrates. In vitro coding joint formation requires RAG1, RAG2, and heat-labile factors present in the nuclear extract of nonlymphoid cells. Both inversion- and deletion-mediated coding joint reactions produce diverse coding joints, with deletions and P nucleotide addition. We also show that deletion-mediated coding joint formation follows the 12/23 rule and requires the catalytic subunit of DNA-dependent protein kinase.

Nicking is asynchronous and stimulated by synapsis in 12/23 rule-regulated V(D)J cleavage

The first step in DNA cleavage at V(D)J recombination signals by RAG1 and RAG2 is creation of a nick at the heptamer/coding flank border. Under proper conditions in vitro the second step, hairpin formation, requires two signals with spacers of 12 and 23 bp, a restriction referred to as the 12/23 rule. Under these conditions hairpin formation occurs at the two signals at or near the same time. In contrast, we find that under the same conditions nicking occurs at isolated signals and hence is not subject to the 12/23 rule. With two signals the nicking events are not concerted and the signal with a 12 bp spacer is usually nicked first. However, the extent and rate of nicking at a given signal are diminished by mutations of the other signal. The appearance of DNA nicked at both signals is stimulated by more than an order of magnitude by the ability of the signals to synapse, indicating that synapsis accelerates nicking and often precedes it. These observations allow formulation of a more complete model of catalysis of DNA cleavage and how the 12/23 rule is enforced.

Disruption of the Multiple Tumor Suppressor Gene MTS1/p16INK4a/CDKN2 by Illegitimate V(D)J Recombinase Activity in T-Cell Acute Lymphoblastic Leukemias

We have recently shown that the multiple tumor suppressor gene 1 (MTS1 ) encoding the p16INK4a and p19ARF cell-cycle inhibitors is inactivated by deletion or disruption in most human T-cell acute lymphoblastic leukemias (T-ALLs), representing the most frequent genetic event thus far described in this disease. To analyze the mechanism of these chromosomal events, we used cloning, sequencing, and/or polymerase chain reaction mapping to study 15 rearrangements occurring in the MTS1 locus. We found that these breakpoints occur in two clusters (MTS1bcrα and MTS1bcrβ ). The three rearrangements occurring in MTS1bcrα correspond to a recurrent recombination juxtaposing 5′ MTS2 exon 1 and 5′ MTS1 exon 1α sequences. Breakpoints for 10 of 12 rearrangements within MTS1bcrβ are located at a polymorphic (CA) repeat, suggesting that this sequence might play a role in the clustering. For both MTS1bcrα and MTS1bcrβ, sequence analyses and PCR mapping experiments show that the tightly clustered breakpoints are located in the vicinity of heptamers whose sequence is similar to those involved in the V(D)J recombination. Moreover, short deletions, GC-rich random nucleotide additions, and clone-specific junctional sequences are present in all cases, further suggesting that the rearrangements are due to illegitimate V(D)J recombinase activity. These data are the first demonstration that a tumor suppressor gene can be inactivated by the V(D)J recombinational mechanism. Moreover, they reinforce the view that this process, normally required for T-cell differentiation, plays a crucial role in the pathogenesis of T-ALL.

Drosophila P-element transposase is a novel site-specific endonuclease

We developed in vitro assays to study the first step of the P-element transposition reaction: donor DNA cleavage. We found that P-element transposase required both 5' and 3' P-element termini for efficient DNA cleavage to occur, suggesting that a synaptic complex forms prior to cleavage. Transposase made a staggered cleavage at the P-element termini that is novel for all known site-specific endonucleases: the 3' cleavage site is at the end of the P-element, whereas the 5' cleavage site is 17 bp within the P-element 31-bp inverted repeats. The P-element termini were protected from exonucleolytic degradation following the cleavage reaction, suggesting that a stable protein complex remains bound to the element termini after cleavage. These data are consistent with a cut-and-paste mechanism for P-element transposition and may explain why P elements predominantly excise imprecisely in vivo.