V(D)J recombination: how to tame a transposase Vicky L. Brandt David B. Roth Authors’ addresses Vicky L. Brandt 1 , David B. Roth 1,2 1 Program in Molecular Pathogenesis, The Skirball Institute, New York University School of Medicine, New York, NY, USA. 2 Department of Pathology and Cancer Institute, New York University School of Medicine, New York, NY, USA. Correspondence to: David B. Roth NYU School of Medicine, SKI 2 540 First Avenue New York, NY 10016 USA Tel.: þ1 212 263 0945 Fax: þ1 212 263 5711 E-mail: [email protected]Summary: Since the discovery that the recombination-activating gene (RAG) proteins were capable of transposition in vitro, investigators have been trying to uncover instances of transposition in vivo and understand how this transposase has been harnessed to do useful work while being inhibited from causing deleterious chromosome rearrangements. How to preserve the capacity of the recombinase to promote a certain class of rearrangements while curtailing its ability to catalyze others is an inter- esting problem. In this review, we examine the progress that has been made toward understanding the regulatory mechanisms that prohibit transposition in order to formulate a model that takes into account the diverse observations that have been made over the last 15 years. First, we touch on the striking mechanistic similarities between transposition and V(D)J recombination and review evidence suggesting that the RAG pro- teins may be members of the retroviral integrase superfamily. We then dispense with an old theory that certain standard products of V(D)J recombination called signal joints protect against deleterious transpos- ition events. Finally, we discuss the evidence that target capture could serve a regulatory role and close with an analysis of hairpins as preferred targets for RAG-mediated transposition. These novel strategies for harness- ing the RAG transposase not only shed light on V(D)J recombination but also may provide insight into the regulation of other transposases. Introduction Chromosomal translocations have long been recognized to be a hallmark of lymphoid neoplasms (1, 2). Many of these translocations involve the immunoglobulin (Ig) and T-cell receptor (TCR) loci and are thought to arise from aberrant V(D)J recombination events (3). Circumstantial evidence implicating the V(D)J recombinase in these translocations has continued to mount over the past 20 years or so, but the ‘how’ remains elusive. In 1998, both the Gellert (4) and the Schatz (5) laboratories discovered that the recombination- activating gene (RAG) proteins are capable of transposition in vitro, providing an elegant mechanism by which the RAG proteins could cause chromosomal translocations. Yet, it has been difficult to identify unambiguous examples of RAG- mediated transposition in vivo or to pinpoint RAG-mediated transposition events as a cause of actual lymphoid tumors. Immunological Reviews 2004 Vol. 200: 249–260 Printed in Denmark. All rights reserved Copyright ß Blackwell Munksgaard 2004 Immunological Reviews 0105-2896 249
Transcript
V(D)J recombination: how to tame a
transposase
Vicky L. Brandt
David B. Roth
Authors’ addresses
Vicky L. Brandt1, David B. Roth1,2
1Program in Molecular Pathogenesis, The
Skirball Institute, New York University School
of Medicine, New York, NY, USA.2Department of Pathology and Cancer Institute,
Summary: Since the discovery that the recombination-activating gene(RAG) proteins were capable of transposition in vitro, investigators havebeen trying to uncover instances of transposition in vivo and understandhow this transposase has been harnessed to do useful work while beinginhibited from causing deleterious chromosome rearrangements. How topreserve the capacity of the recombinase to promote a certain class ofrearrangements while curtailing its ability to catalyze others is an inter-esting problem. In this review, we examine the progress that has beenmade toward understanding the regulatory mechanisms that prohibittransposition in order to formulate a model that takes into account thediverse observations that have been made over the last 15 years. First, wetouch on the striking mechanistic similarities between transposition andV(D)J recombination and review evidence suggesting that the RAG pro-teins may be members of the retroviral integrase superfamily. We thendispense with an old theory that certain standard products of V(D)Jrecombination called signal joints protect against deleterious transpos-ition events. Finally, we discuss the evidence that target capture couldserve a regulatory role and close with an analysis of hairpins as preferredtargets for RAG-mediated transposition. These novel strategies for harness-ing the RAG transposase not only shed light on V(D)J recombination butalso may provide insight into the regulation of other transposases.
Introduction
Chromosomal translocations have long been recognized to be
a hallmark of lymphoid neoplasms (1, 2). Many of these
translocations involve the immunoglobulin (Ig) and T-cell
receptor (TCR) loci and are thought to arise from aberrant
Fig. 1. The basic steps of V(D)J recombination. The top of thefigure shows a stereotyped immunoglobulin (Ig) or T-cell receptor(TCR) locus with colored boxes representing coding segments. Whiteand black triangles represent 12- and 23-recombination signal sequences(RSSs); one of each type of RSS is recognized and bound by therecombination-activating gene (RAG) proteins (depicted as orangeovals). Although a single RSS can undergo RAG nicking, synapsis isnecessary for RAG cleavage to take place. Once cleavage is completed,the RAG post-cleavage complex remains bound to the signal and
coding ends for an indefinite period, allowing the non-homologousend-joining (NHEJ) factors to perform end processing and joining. Thecoding ends are ligated imprecisely to form a coding joint, which formsthe new antigen receptor gene. The signal ends are ligated to form asignal joint; in deletional recombination (left panel), signal joints areexcised circular products that are lost from the cell during division. Theright panel depicts inversional recombination. The steps are the same,but the orientation of the RSS dictates that the signal joint be retained inthe chromosome to maintain chromosomal integrity.
Brandt & Roth �How to tame a transposase
250 Immunological Reviews 200/2004
The formation of these non-standard products provided the
first clue that the broken DNA ends might remain bound to the
RAG proteins after cleavage (7). We now know this to be the
case: the RAG proteins are thought to maintain both coding
and signal ends in a post-cleavage complex in vitro (8, 9) and in
vivo (3, 10–13). Because the only sign that an open-and-shut
joint has formed is the loss or addition of nucleotides to the
junction, which is likely to occur in only a subset of these
joints, the actual frequency of formation of these non-standard
products is impossible to gauge.
Similarities between V(D)J recombination and
transposition
V(D)J recombination possesses several features characteristic
of ‘cut-and-paste’ transposition (discussed in 14)(Fig. 4). Like
the V(D)J recombinase, transposases recognize short, specific
sequences adjoining the mobile DNA segment and introduce
double-strand breaks between these recognition sequences
and the flanking chromosomal DNA (the ‘donor site’), liberat-
ing the transposable element. Elegant biochemical studies
from the Gellert laboratory (15) established early on that the
second step of the RAG cleavage reaction, hairpin formation,
occurs via a one-step transesterification mechanism that bears
a remarkable resemblance to certain transposition and retro-
viral integration events. Indeed, several cut-and-paste trans-
posons excise themselves via a hairpin intermediate (16, 17).
V(D)J recombination and transposition are also similar in that
the broken donor (or coding) ends in both reactions are
joined by general cellular DNA repair mechanisms. For the
purposes of our discussion here, the salient difference between
the two reactions is that in transposition, the transposon
integrates by transesterification into a new target site, whereas
in V(D)J recombination the signal ends bounding the excised
fragment are circularized via a ligation reaction.
Besides the similar chemistry of transposition and V(D)J
recombination, there are other reasons to suspect a common
mechanistic ancestry. The RSSs resemble inverted repeats at
either end of a transposon, and the close proximity and com-
pact nature of the RAG1 and RAG2 genes suggests that they may
once have had to fit onto a single transposable element (4, 5,
18–20). Furthermore, the sudden appearance of RAG genes at
the level of sharks (no homologous genes have been found in
any lower eukaryotes) could be readily explained by positing
their horizontal transfer from a donor organism (20–22). As
Normal coding andsignal joint formation
Hybrid jointformation
Open-and-shut jointformation
Coding joint Signal joint Hybrid joints Open-and-shut joints
Fig. 3. Standard and non-standard products of V(D)J recombination.The usual products of V(D)J recombination, illustrated at the left, are codingand signal joints. Some percentage of the time, however, the recombinaseforms non-standard products. Hybrid joints, shown in the middle panel,arise when the coding and signal ends switch partners: the coding end thathad been paired with a 12-recombination signal sequence (RSS) becamejoined to the 23-RSS and vice versa. The right panel depicts open-and-shutjoints, which arise when the recombinase cleaves the border of the RSS andcoding flank as normal but fails to complete rearrangement and insteadsimply re-joins the same coding end-RSS pairs. Non-standard junctionssometimes lose and/or gain nucleotides during processing, betraying thefact that the coding segment and its RSSs have undergone cleavage andre-joining. Nucleotide changes are represented here by gray hatch marks.
Repaired donorDNA
Transposon inserted
into target DNA(transesterification)
Transposition V(D)J recombination
HO HO
Coding joint Signal joint(ligation)
OH
Target DNA
OH
Fig. 4. Comparison of V(D)J recombination and transposition. Intransposition, a section of DNA is excised from a donor site and inserted intounrelated target DNA by a transesterification reaction, using the 30 hydroxyl(OH) groups from the transposon to cleave the donor DNA. Similarly, indeletional V(D)J recombination, the recombination-activating gene (RAG)proteins cleave DNA strands and remove the signal ends from thechromosome. A principal difference between the two reactions lies in thefate of the excised fragment: whereas the transposon is inserted into targetDNA, the signal ends are ligated to form a signal joint.
OHNicking
Hairpin formation (cleavage)
Fig. 2. Nicking and hairpin formation. The recombination-activatinggene (RAG) proteins nick one DNA strand precisely between therecombination signal sequence (RSS) and its coding segment, generatinga 30 hydroxyl (OH) that is used to attack the other strand. Thistransesterification reaction cleaves the DNA by forming a hairpin on thecoding end segment and a blunt double-strand break on the signal end.
Brandt & Roth �How to tame a transposase
Immunological Reviews 200/2004 251
compelling as these evolutionary considerations are, however,
one significant question had yet to be answered before the
RAG proteins could be considered a full-fledged transposase:
can they catalyze transposition?
From RAG-mediated hybrid joint formation to RAG-
mediated transposition
The first hint that RAG transesterification could catalyze a
reaction akin to transposition came from analysis of joining
products formed in NHEJ-deficient mice and cell lines, which
are severely defective for the standard products of V(D)J
recombination, coding and signal joints. Hybrid joints form
at normal levels from extrachromosomal substrates transiently
transfected into Ku80-, DNA-PKcs-, and XRCC4-deficient
fibroblasts (23, 24) and at Ig heavy-chain loci in Ku80- and
DNA-PKcs-deficient mice (25, 26). Intriguingly, most of
these hybrid joints retain all nucleotides from both ends, and
many show evidence that the hairpin has been opened close to
but not precisely at the tip (23–25). These observations led us
to suggest that the RAG proteins might be capable of joining
the 30 OH of a signal end to coding end by transesterification
in a reaction that is a form of transposition (23, 25)(Fig. 5).
We termed this reaction ‘RAG-mediated joining’ to distinguish
this reaction from ‘true’ transposition, which by definition
involves capture of a free target molecule (at that time, the
signal end was thought to attack the coding end that was
already present in the RAG post-cleavage complex, but recent
work calls this assumption into question; see below). In vitro
experiments soon confirmed that purified truncated RAGproteins
could indeed catalyze formation of non-standard joints (27).
The key question then became whether the RAG proteins
can capture targets other than hairpin coding ends. The coding
ends might be considered merely a target of convenience, if
they remain trapped in close proximity to the signal ends in
the post-cleavage complex. The answer came through closer
examination of what had been an unexplained product on a
gel that appeared in early in vitro V(D)J reactions using core
RAG proteins. [The full-length proteins have been difficult to
use in purified form for biochemical studies, so until recently,
in vitro experiments relied on truncated or ‘core’ RAG proteins
that retain apparently normal catalytic activity. Possible func-
tions of some of the ‘dispensable’ regions of the proteins are
discussed below; other potential functions are discussed in
(28)]. Within a day of each other, the Gellert (4) and Schatz
(5) laboratories published evidence that this product resulted
from a bona fide transposition event: the RAG proteins can
integrate signal ends into random targets, at least in the test
tube. No preference for particular target sequences was noted,
except for a slight preference for G-C base pairs on both sides
of the integration site (4, 5).
The discovery that the RAG proteins can catalyze transpos-
ition was exciting for two reasons. First, it supported earlier
speculations about the evolutionary origin of the vertebrate
immune system. Second, it suggested interesting new models
for RAG-mediated oncogenic chromosome rearrangements (4,
14). In addition to straightforward scenarios such as gener-
ation of chromosomal translocations by single-ended transpos-
ition (a reaction that occurs readily in vitro) and insertional
mutagenesis via standard double-ended transposition events,
Gellert and colleagues (4) suggested that the branched inter-
mediates formed by transposition might be resolved by further
RAG cleavage, fragmenting the chromosome and potentially
leading to translocations that lack RSS elements and thus
leavenomolecular fingerprints that betray recombinase involve-
ment (see figure 4 in ref. 14). Subsequent work revealed
that the RAG proteins are indeed able to resolve branched
intermediates of transposition in this way (29), and chromo-
some translocations that can be explained by such a mechan-
ism have been discussed (4). The inability to prove or disprove
its origin in a recombinase-mediated event leaves open the
possibility that RAG-mediated transposition may be more
common than we realize. Nevertheless, Hiom et al. (4) suggest
A Hairpin formation C Transposition
B Hybrid joint formation
OH
OH
OH
HO
HO
HO
Fig. 5. Different outcomes for recombination-activating gene (RAG)-mediated transesterification reactions. (A) Standard hairpin formationas it occurs in the context of V(D)J recombination: RAG nicking liberatesa 30 hydroxyl (OH) that then attacks the opposite strand, forming ahairpin and effecting cleavage. (B) This reaction can operate essentially inreverse: a 30 OH group on a signal end can attack a hairpin coding end,leading to the formation of a hybrid (or open-and-shut) joint. (C) The 30
OH groups of either a signal end fragment or a transposon can be used toinsert the donor fragment into target DNA.
Brandt & Roth �How to tame a transposase
252 Immunological Reviews 200/2004
that translocation by such a mechanism is likely to be rare
because most signal ends are ultimately incorporated into
signal joints. More recent work from our laboratory calls this
view of signal joints as a protective mechanism into question
(see next section).
Shortly after the discovery that the RAG proteins could act as
an authentic transposase, additional mechanistic similarities
between the RAG proteins and the transposase/retroviral inte-
grase superfamily came to light. Members of this superfamily
employ three acidic amino acid residues, two aspartates and a
glutamate (the so-called ‘DDE motif’) to coordinate catalytic-
ally active divalent metal ions in the active site (30–32).
Because DNA cleavage by the RAG proteins is dependent
upon the presence of divalent metal ions, it was reasonable
to suppose that the RAG proteins might also employ some
combination of acidic residues. In the absence of structural
information and because sequence comparisons with other
transposases yielded little useful information, we decided to
undertake a comprehensive site-directed mutagenesis approach,
mutating all 109 evolutionarily conserved D and E residues
from core RAG1 and RAG2 (33). This approach, while labor
intensive, had the singular advantage of testing all potential
catalytic acidic residues in both proteins and would thus enable
discovery of an active site shared between the two proteins or a
metal coordination scheme that employs a different combin-
ation of acidic residues, such as that used by the TnsA protein of
the Tn7 transposase (34). We identified three catalytically
defective mutants affecting positions D600, D708, and E962
in murine RAG1. Biochemical experiments provided evidence
that at least one of these residues, D708, contacts the catalytic
divalent metal ion (33). Further confirmation that the two D
residues contact the catalytic metal was provided by elegant
iron-induced OH radical protein cleavage experiments (35).
This result is what one would expect if these residues participate
in a DDE metal-binding motif. Crystal structures of the HIV
and avian sarcoma virus (ASV) integrases in the presence of
divalent metal ions suggest that the carboxylates of the two
aspartates are critical for positioning a metal ion to facilitate
cleavage of the scissile phosphodiester bond (36–39). The
catalytic core of RAG1 thus bears at least a superficial resem-
blance to that of the DDE motif-containing retroviral integrase
superfamily members. This resemblance subsequently received
further support from secondary structure predictions, which
identified a region of RAG1 encompassing D600 and D708
that bears some similarities to members of the retroviral inte-
grase superfamily (40). As expected for a member of a DDE
motif, E962 is clearly required for catalytic activity but not for
DNA binding. There is as yet, however, no direct evidence that
this residue in RAG1 contacts a metal ion (33, 35). Indeed, such
evidence has been difficult to come by for many proteins in the
retroviral integrase superfamily (32, 36–39). A recent structure
of the Tn5 transposase clearly shows the aspartic acid coordinat-
ing one of two catalytic Mn2þ ions (41); it is quite possible that
E962 in RAG1 also serves to contact a second metal ion. None-
theless, structural characterization of the RAG proteins will be
necessary in order to place them definitively within the DDE-
containing retroviral integrase superfamily.
Do signal joints protect against unwanted
rearrangements?
The principal difference between V(D)J recombination and cut-
and-paste transposition lies in the fate of the excised fragment:
rather than being inserted into target DNA, the RSSs are ligated to
form a signal joint, which ismost often an excised circular product
that is lost during cell division. At some antigen receptor loci, the
orientation of the RSS necessitates inversional recombination in
which the signal joint is retained on the chromosome to preserve
chromosomal integrity (42) (Fig. 1), but the result is the same. The
signal ends are joined to one another rather than to target DNA.
Until recently, it was reasonable to assume that the raison
d’etre for signal joint formation was to prevent the potentially
reactive OH groups from being inserted into target DNA (4,
14, 43). The signal joint (at least in its most common, excised
form) has no other known function, and joining certainly
seems like a handy way to neutralize the reactive 30 OH
groups, which could engage in either transposition or illegit-
imate joining events. So convincing was this rationale that no
one questioned it for years, despite a curious observation that
suggested the truth might be a bit more complex.
Signal ends are much more abundant than coding ends in
the thymocytes and bone marrow cells of wildtype mice
(44–47). The Gellert (48), Roth (44), and Schatz (47) laboratories
all proposed that signal ends must have considerably longer
lifespans than coding ends, although none could explain why
this would be the case, especially in light of the fact that coding
joint formation entails a greater number of steps (and DNA
repair factors) than signal joint formation. Nonetheless, signal
ends persist until RAG expression is downregulated (47, 48), at
which time signal joint formation takes precedence. These in vivo
observations, along with the fact that the RAG proteins remain
tightly bound to signal ends after cleavage in vitro (8, 49),
seemed to support the longer lifespan model.
There is, of course, another obvious possibility: signal joints
might be re-cleaved by the RAG proteins. They do, after all,
contain an appropriate pair of RSSs. Others had shown that
Brandt & Roth �How to tame a transposase
Immunological Reviews 200/2004 253
signal joints can provide a single RSS for generating rare
secondary rearrangements involving a second RSS located else-
where (50, 51), and there is no reason one of the RSSs in a
signal joint should not be available to the recombinase. But
cleavage between the two RSSs in the same signal joint poses
two problems. First, because the binding footprint of the RAG
proteins covers a few nucleotides of the coding flank (52),
steric constraints should prevent two RAG complexes from
binding simultaneously to both RSSs. Second, because synapsis
of a 12/23 RSS pair is required for hairpin formation (6), it
was difficult to imagine that signal joints, which of course
have no nucleotides between the RSSs, could be cleaved via the
normal mechanism [a distance between RSSs shorter than 40
theless, we decided to explore the possibility that signal joints
might serve as substrates for RAG-mediated cleavage (55).
What we found was rather unexpected. Not only were signal
joints readily re-cleaved by the RAG proteins in vivo and in vitro,
but they are cleaved by a novel nick–nick mechanism in
which the RAG proteins bound to one RSS introduce a nick
on one strand, while another RAG complex bound to the other
RSSs nicks the other strand, bypassing hairpin formation and
forming two blunt signal ends (55). To substantiate these find-
ings, we tested two RAG1 mutants that can nick but are incap-
able of hairpin formation. Despite being severely defective for
cleavage of standard plasmid substrates, these two mutants
efficiently cleaved signal joints, providing strong evidence that
cleavage was occurring through a nick–nick pathway (Fig. 6).
Furthermore, we found that signal joints are excellent substrates
for transposition in vitro, with both ends produced by signal joint
cleavage being integrated into the target (55). In retrospect,
these findings should not have been so surprising, because it
had long been appreciated that the RAG proteins could readily
nick a single RSS.
Signal joint cleavage explains three otherwise puzzling phe-
nomena. First, signal ends are much more abundant than
coding ends in normal lymphocyte precursors, even though
both are produced by the same cleavage event. Signal joint
cleavage is a more parsimonious explanation for this relative
overabundance than the previously postulated longer lifespan
model. Second, signal joints do not accumulate to appreciable
levels until after RAG expression is downregulated. It had been
suggested that the stable complexes formed by the RAG pro-
teins on signal ends might block signal joint formation (8, 47,
48, 56). Our laboratory’s in vivo work with mutant RAG1 and
RAG2 proteins, however, demonstrated that, far from block-
ing joining, the RAG proteins actually form a necessary scaf-
fold for proper joining of the signal ends and coding ends
(10–12). Ongoing re-cleavage of signal joints provides a sim-
ple explanation for the observation that coding joints accumu-
late in the presence of high levels of RAG proteins while signal
joints do not. Third, mutations that inhibit joining have a
differential effect on signal and coding end levels. Mutations
that block coding joint formation greatly increase coding end
levels (more than 100 fold), but levels of signal ends are not
appreciably raised by mutations that block signal joint forma-
tion (44, 46, 56). This paradox is elegantly resolved by signal
joint re-cleavage, because we now appreciate that signal end
levels depend upon the rate of formation, joining, and
re-cleavage.
Does signal joint cleavage have pathological consequences?
In an analysis of both healthy individuals and individuals with
T-cell acute lymphoblastic leukemia (T-ALL), Marculescu and
colleagues (57) found that the translocation junctions in the
T-ALL patients derive from ongoing recombination involving
the signal joint. Thus, signal joints can constitute unstable
elements with oncogenic potential and do not necessarily
provide the protection that we assumed. So what does prevent
the RAG proteins from transposing signal ends?
Modulating transposase activity by controlling target
capture
It was recognized that selection of the target DNA molecule
would provide an ideal regulatory control point, because
blocking target capture should prevent transposition without
Nick
Nick
Fig. 6. A novel nick–nick pathway for cleaving signal joints. Signaljoints are cleaved by the recombination-activating gene (RAG) proteins ina manner that obviates synapsis and hairpin formation. The RAG proteins,depicted together as an orange oval, nick the 50 of each recombinationsignal sequence (RSS), either simultaneously or sequentially. Curvedarrows indicate where nicking occurs.
Brandt & Roth �How to tame a transposase
254 Immunological Reviews 200/2004
affecting V(D)J recombination (58). To begin to dissect this
key step, Neiditch and colleagues (58) identified a target-
capture complex containing RAG proteins, signal ends, and a
target DNA molecule and showed that its formation depends
upon the presence of active-site DDE residues presumed to
coordinate catalytic metal ions. This finding raised an import-
ant question: where does target DNA bind? In the case of
Tn10, whose behavior parallels that of the RAG transposase
in many respects, stable target interactions require cleavage to
liberate the transposon ends from the flanking DNA. This
activity suggests that the target DNA binds to a region of the
Tn10 protein that also binds the flanking DNA prior to cleav-
age (59). In contrast, Neiditch and colleagues (58) found
that the RAG transposase is capable of committing to a par-
ticular target molecule either before or after RSS cleavage, i.e.
target interactions do not require RSS cleavage. The ability of
the RAG–RSS complex to select a target prior to RSS cleavage
implies that target binding does not entail the release of the
coding ends. As Neiditch et al. (58) noted, however, these
results do not rule out the possibility that RAG monomers
not involved in donor cleavage might bind to target. Indeed,
it is quite likely that the active form of the RAG transposase
contains at least a dimer of RAG1 (60–62). Thus, the ability of
the transposase to commit to a target prior to RSS cleavage
does not conflict with the proposal that target binding is
carried out by active site regions responsible for binding to
Fig. 7. A mechanism for preventing transposition into an exogenous
DNA target. Observations of non-standard products in mice and cell linesdeficient in non-homologous end-joining (NHEJ) indicate that hybridjoints and open-and-shut joints can be formed by the recombination-activating gene (RAG) proteins alone, without the help of the NHEJrepair machinery. The RAG proteins can perform transesterification on anexisting hairpin, opening it and re-joining it to a signal end (Fig. 5). Theoriginal model for open-and-shut joint formation is depicted on the left:the coding end is joined to its original partner recombination signalsequence (RSS), while both coding and signal ends are retained withinthe post-cleavage complex. This ‘stable post-cleavage complex model’makes it difficult to explain how a hybrid joint would arise, as the signalends must be flipped. In our model of RAG-mediated transposition intohairpin targets, however, the isomerization step is not a problem. Lee et al.(63) demonstrated that the RAG proteins exhibit a strong preference forhairpin targets. Thus, that small fraction of RAG complexes that arepredisposed to transpose will find that the nearest favored target happensto be a hairpinned coding end. This favored target accounts for theformation of hybrid and open-and-shut joints and may help explain theapparently low frequency of transposition into other targets in vivo.
Brandt & Roth �How to tame a transposase
256 Immunological Reviews 200/2004
more efficiently than their full-length counterparts. The authors
suggested that our polymerase chain reaction (PCR) assay,
which detects DH-to-JH4 deletional hybrid joints, would not
detect products bearing large deletions and that the products
we detected might represent ‘highly selected products of
attempted D-to-JH joints that deleted back to the 50 D RS’
(76). Although it is true that our assay would not detect highly
deleted products, we were nonetheless readily able to detect
high levels of undeleted products in at least seven individual
mice, which shows that formation of these joints is not
affected by lack of Ku80 or DNA-PKcs (25, 26). Sequence
analysis of the products (29 junctions from three individual
animals) revealed precise joining of intact signal ends to
undeleted coding ends. We think it unlikely that random
deletion of coding ends would chew back the entire D seg-
ment precisely to the RSSs in each instance, yielding junctions
that mimic perfect hybrid joints. Production of these joints by
alternative end joining rather than RAG transesterification
seems even more unlikely, in light of the fact that signal joints
almost always form with nucleotide loss in these same Ku80-
deficient mice (25). It is also not clear in what sense these
hybrid joints are ‘highly selected’. These non-coding products
are under no obvious immunological selection, and selection
for chromosomal rejoining would not necessarily favor such
precise joints. Furthermore, the DH–JH4 coding joints
amplified in the same PCR reaction (and probed with the
same JH probe) serve as an internal control; unlike the hybrid
joints, these junctions are severely diminished in the NHEJ
mutants (25, 26). With these observations in mind, the most
parsimonious interpretation is that the precise hybrid
joints formed at the Ig heavy-chain locus in these animals
are generated by RAG-mediated joining/transposition.
One interesting potential explanation for the difference
between our observations in mice and the Alt group’s obser-
vations in transfected fibroblasts is that this phenomenon
might be locus-specific. It should be noted, however, that,
using extrachromosomal substrates, we demonstrated that
full-length RAG proteins can generate hybrid joints in NHEJ-
deficient cells (24). Thus, while we do not dispute the possi-
bility that core RAG proteins might form hybrid joints more
efficiently than the full-length proteins, our work clearly pre-
dicts that full-length RAGs should catalyze robust hybrid joint
formation in vitro. So far, two of three studies support our
model revealing no defect in the ability of full-length RAG1
and RAG2 proteins to catalyze hybrid joint formation in vitro
(64, 65). Only one study found that full-length RAG2
diminishes hybrid joint formation (66); the reason for this
discrepancy remains unresolved. Although consistent with our
findings at the IgH locus in NHEJ-deficient mice, these obser-
vations have led the Oettinger (64) and Swanson (65) groups
to postulate that other factors besides full-length RAGs might
be responsible for the apparent reduction in RAG-mediated
hybrid joints observed in fibroblasts transiently transfected
with full-length as opposed to core proteins. This interesting
possibility remains to be explored.
Could non-standard joints protect against transposition
in vivo?
The ability of the RAG proteins to generate non-standard
junctions (open-and-shut and hybrid joints) by a transpos-
itional mechanism raises the interesting possibility that these
products, while immunologically irrelevant, might provide an
important line of defense against dangerous transposition
events by encouraging the RAG transposase to generate harm-
less transposition products (63). We have considered two
alternative scenarios. In the first case, the decision to transpose
is made at an early step, while the coding ends remain asso-
ciated with the signal ends in the post-cleavage complex. If, as
we have postulated, the coding ends occupy the target-binding
site, a ‘transposition’ event can occur with rejoining of the
signal ends to the hairpin coding ends to produce open-and-
shut joints (Fig. 7). These events would provide a perfectly safe
outlet for the recombinase’s transpositional tendencies and
have the advantage of not allowing capture of an exogenous
target DNA molecule (63). This proposal resembles a model
put forth by Elkin et al. (64), who suggested that full-length
RAG2 may stabilize the bound hairpin coding ends in the RAG
complex, preventing target capture. In their model, however,
protection would be lost as soon as coding joints form. In our
model, the protection conferred by forming open-and-shut
joints would be permanent.
In the second scenario (Fig. 7), the preference of the RAG
transposase for capture of hairpin ends could allow non-
standard junctions to play a protective role even after the
coding ends have dissociated from the post-cleavage complex,
provided that they remain unjoined and available for capture
(63). According to this model, a RAG-signal end complex
inclined to transpose would be able to capture (or re-capture)
preferred targets, the hairpin coding ends, and affect transpos-
ition to form either an open-and-shut joint or a hybrid joint (if
the RSS partners of the coding ends are swapped). This model
provides a simple explanation for the formation of hybrid
Brandt & Roth �How to tame a transposase
Immunological Reviews 200/2004 257
joints, which, if they form in the context of the post-cleavage
complex, would require some sort of isomerization of the
complex to swap the ends prior to cleavage.
Are non-standard junctions formed frequently enough to
serve a protective role? Only a small fraction (0.1–5%) of
the signal end complexes formed in vitro, even with truncated
RAG proteins, undergo transposition (4, 8). Hybrid joints are
quite frequently generated from extrachromosomal substrates
by full-length RAG proteins (up to approximately 30% of
products) (7), and a substantial fraction of these have struc-
tures compatible with RAG-mediated joining (transposition)
(7, 23, 24). Hybrid joints are also found, although less com-
monly, at endogenous antigen receptor loci. Deletional hybrid
joints, which require formation of only a single junction, are
readily detected at the Ig heavy-chain and TCR � loci (25, 77).
Inversional rearrangements that form two hybrid joints, not
surprisingly, are less common, but they have also been
detected at the Ig heavy-chain locus (78, 79). Of course,
only a subset of open-and-shut junctions can be detected
(those that alter the nucleotide sequence at the junction);
even so, these are readily formed by full-length RAG proteins
from extrachromosomal substrates (approximately 1% of the
frequency of standard junctions) (53) and at endogenous
antigen receptor loci (up to 19% at TCR �) (80). Even if
only a fraction of these events are generated by a transpos-
itional mechanism (we expect some proportion to form by
end joining), these numbers seem high enough to make a
significant contribution to protecting the organism against
deleterious transpositional rearrangements.
The mechanism described above in essence uses a ‘safe’
transposition event to defend against transposition to an
exogenous target DNA. Gellert and colleagues (29) recently
articulated another defense mechanism that capitalizes on
features inherent in transposases: the ability to reverse the
reaction through disintegration. This reaction occurs quite
efficiently in vitro at high Mgþ concentrations; it will be inter-
esting to determine whether disintegration occurs frequently
in vivo.
Conclusion
The V(D)J recombinase is clearly functionally related to trans-
posases, and few doubt that the RAG proteins and their
associated RSSs are derived from an ancient transposable ele-
ment. This presents an interesting problem for the immune
system, which wants to put the transposasc to good use
without enabling authentic transposition into exogenous
targets — a dangerous process for the organism as a whole.
Only vestiges of transpositional activity remain, which can be
seen in the test tube under appropriate conditions (4, 5) yet
very rarely in vivo (71) [or perhaps more commonly in vivo in
the form of specialized products, open-and-shut and hybrid
joints (23, 25)]. One strategy common to many transposase
systems is to carefully regulate the expression of the transpo-
sase to prevent high levels of activity, but this regulation is
not an option for the immune system because the RAG
proteins are critical to lymphocyte development. This predica-
ment has forced the RAG proteins and the vertebrate immune
system to take what may be novel strategies to allow transposon
excision but restrain transposition. Some of these strategies
should provide insights not only into V(D)J recombination
but also into mechanisms that regulate other transposase
systems.
Note
As this manuscript went to press, Raghavan et al. (Nature
2004,428:88–93) published results suggesting that RAG
recognition of altered DNA structures might be responsible
for some chromosomal translocations. The authors examined a
common breakpoint sequence and found that RAG proteins
can nick one strand of the non-B DNA in vitro. This dovetails
nicely with the earlier work of Lee et al. (63) showing that the
RAG proteins can target hairpin, cruciform, and other altered
DNA structures for transposition and suggesting that such
altered strucutres may induce RAG activity that may lead to
translocations.
References
1. Tycko B, Sklar J. Chromosomal translocations
in lymphoid neoplasia: a reappraisal of the
recombinase model. Cancer Cells 1990;2:1–8.
2. Vanasse GJ, Concannon P, Willerford DM.
Regulated genomic instability and neoplasia
in the lymphoid lineage. Blood
1999;94:3997–4010.
3. Roth DB. Restraining the V(D)J recombinase.
Nat Rev Immunol 2003;3:656–666.
4. Hiom K, Melek M, Gellert M. DNA
transposition by the RAG1 and RAG2
proteins: a possible source of oncogenic
translocations. Cell 1998;94:463–470.
5. Agrawal A, Eastman QM, Schatz DG.
Transposition mediated by RAG1 and RAG2
and its implications for the evolution of the
immune system. Nature 1998;394:744–751.
6. van Gent DC, Ramsden DA, Gellert M. The
RAG1 and RAG2 proteins establish the 12/23
rule in V(D)J recombination. Cell
1996;85:107–113.
7. Lewis SM, Hesse JE, Mizuuchi K, Gellert M.
Novel strand exchanges in V(D)J
recombination. Cell 1988;55:1099–1107.
Brandt & Roth �How to tame a transposase
258 Immunological Reviews 200/2004
8. Agrawal A, Schatz DG. RAG1 and RAG2 form
a stable postcleavage synaptic complex with
DNA containing signal ends in V(D)J
recombination. Cell 1997;89:43–53.
9. Hiom K, Gellert M. Assembly of a 12/23
paired signal complex: a critical control point
in V(D)J recombination. Mol Cell
1998;1:1011–1019.
10. Qiu JX, Kale SB, Yarnell Schultz H, Roth DB.
Separation-of-function mutants reveal critical
roles for RAG2 in both the cleavage and
joining steps of V(D)J recombination. Mol
Cell 2001;7:77–87.
11. Yarnall Schultz H, Landree MA, Qiu JX, Kale
SB, Roth DB. Joining-deficient RAG1 mutants
block V(D)J recombination in vivo and
hairpin opening in vitro. Mol Cell
2001;7:65–75.
12. Huye LE, Purugganan MM, Jiang MM, Roth
DB. 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. Mol Cell Biol
2002;22:3460–3473.
13. Tsai CL, Drejer AH, Schatz DG. Evidence of a
critical architectural function for the RAG
proteins in end processing, protection, and
joining in V(D)J recombination. Genes Dev
2002;16:1934–1949.
14. Roth DB, Craig NL. VDJ recombination: a
transposase goes to work. Cell 1998;94:
411–414.
15. van Gent DC, Mizuuchi K, Gellert M.
Similarities between initiation of V(D)J
recombination and retroviral integration.
Science 1996;271:1592–1594.
16. Kennedy AK, Guhathakurta A, Kleckner N,
Haniford DB. Tn10 transposition via a DNA
hairpin intermediate. Cell 1998;95:125–134.
17. Bhasin A, Goryshin IY, Reznikoff WS. Hairpin
formation in Tn5 transposition. J Biol Chem
1999;274:37021–37029.
18. Sakano H, Huppi K, Heinrich G, Tonegawa S.
Sequences at the somatic recombination sites
of immunoglobulin light-chain genes. Nature
1979;280:288–294.
19. Oettinger MA, Schatz DG, Gorka C, Baltimore D.
RAG-1 and RAG-2, adjacent genes that
synergistically activate V(D)J recombination.
Science 1990;248:1517–1523.
20. Thompson CB. New insights into V(D)J
recombination and its role in the evolution of the
immune system. Immunity 1995;3:531–539.
21. Roth DB. From lymphocytes to sharks: V(D)J
recombinase moves to the germline. Genome
Biol 2000;1:reviews1014.1–1014.4.
22. Hansen JD, McBlane JF. Recombination-
activating genes, transposition, and the
lymphoid-specific combinatorial immune
system: a common evolutionary connection.
Curr Top Microbiol Immunol 2000;248:
111–135.
23. Han J-O, Steen SB, Roth DB. Ku86 is not
required for protection of signal ends or for
formation of nonstandard V(D)J
recombination products. Mol Cell Biol
1997;17:2226–2234.
24. Han J-O, Erskine LA, Purugganan MM,
Stamato TD, Roth DB. V(D)J recombination
intermediates and non-standard products in
XRCC4-deficient cells. Nucleic Acids Res
1998;26:3769–3775.
25. Bogue MA, Wang C, Zhu C, Roth DB. V(D)J
recombination in Ku86-deficient mice:
distinct effects on coding, signal, and hybrid
joint formation. Immunity 1997;7:37–47.
26. Bogue MA, Jhappan C, Roth DB. Analysis of
variable (diversity) joining recombination in
DNA-dependent protein kinase (DNA-PK)-
deficient mice reveals DNA-PK-independent
pathways for both signal and coding joint
formation. Proc Natl Acad Sci USA 1998;
95:15559–15564.
27. Melek M, Gellert M, van Gent DC. Rejoining
of DNA by the RAG1 and RAG2 proteins.
Science 1998;280:301–303.
28. Steen SB, Han J-O, Mundy C, Oettinger MA,
Roth DB. Roles of the ‘‘dispensable’’ portions
of RAG-1 and RAG-2 in V(D)J recombination.
Mol Cell Biol 1999;19:3010–3017.
29. Melek M, Gellert M. RAG1/2-mediated
resolution of transposition intermediates:
two pathways and possible consequences.
Cell 2000;101:625–633.
30. Kulkosky J, Jones KS, Katz RA, Mack JPG,
Skalka AM. Residues critical for retroviral
integrative recombination in a region that is
highly conserved among retroviral/
retrotransposon integrases and bacterial
insertion sequence transposases. Mol Cell Biol
1992;12:2331–2338.
31. Grindley NDF, Leschziner AE. DNA
transposition: from a black box to a color
monitor. Cell 1995;83:1063–1066.
32. Rice P, Mizuuchi K. Structure of the
bacteriophage Mu transposase core: a
common structural motif for DNA
transposition and retroviral integration.
Cell 1995;82:209–220.
33. Landree MA, Wibbenmeyer JA, Roth DB.
Mutational analysis of RAG-1 and RAG-2
identifies three active site amino acids in
RAG-1 critical for both cleavage steps of
V(D)J recombination. Genes Dev 1999;13:
3059–3069.
34. Hickman AB, Li Y, Mathew SV, May EW,
Craig NL, Dyda F. Unexpected structural
diversity in DNA recombination: the
restriction endonuclease connection. Mol Cell
2000;5:1025–1034.
35. Kim DR, Dai Y, Mundy CL, Yang W,
Oettinger MA. Mutations of acidic residues in
RAG1 define the active site of the V(D)J
recombinase. Genes Dev 1999;13:3070–3080.
36. Bujacz G, et al. High-resolution structure of
the catalytic domain of avian sarcoma virus
integrase. J Mol Biol 1995;253:333–346.
37. Bujacz G, et al. The catalytic domain of avian
sarcoma virus integrase: conformation of the
active-site residues in the presence of divalent
cations. Structure 1996;4:89–96.
38. Goldgur Y, Dyda F, Hickman AB, Jenkins TM,
Craigie R. Three new structures of the core
domain of HIV-1 integrase: an active site that
binds magnesium. Proc Natl Acad Sci USA
1998;95:9150–9154.
39. Maignan S, Guilloteau J-P, Zhou-Liu Q,
Clement-Mella C, Mikol V. Crystal structures
of the catalytic domain of HIV-1 integrase
free and complexed with its metal cofactor:
high level of similarity of the active site with
other viral integrases. J Mol Biol
1998;282:359–368.
40. Fugmann SD, Villey IJ, Ptaszek LM, Schatz DG.
Identification of two catalytic residues in
RAG1 that define a single active site within the
RAG1/RAG2 protein complex. Mol Cell 2000;
5:97–107.
41. Lovell S, Goryshin IY, Reznikoff WR,
Rayment I. Two-metal active site binding of a
Tn5 transposase synaptic complex. Nat Struct
Biol 2002;9:278–281.
42. Lewis SM. The mechanism of V(D)J joining:
lessons from molecular immunological and
comparative analyses. Adv Immunol 1994;
56:27–150.
43. Market E, Papavasiliou FN. V(D)J
recombination and the evolution of the
adaptive immune system. PLos Biol
2003;1:E16.
44. Roth DB, Menetski JP, Nakajima PB, Bosma MJ,
Gellert M. V(D)J recombination: broken DNA
molecules with covalently sealed (hairpin)
coding ends in scid mouse thymocytes. Cell
1992;70:983–991.
45. Schlissel M, Constantinescu A, Morrow T,
Baxter M, Peng A. Double-strand signal
sequence breaks in V(D)J recombination are
blunt, 50-phosphorylated RAG-dependent,
and cell cycle regulated. Genes Dev 1993;
7:2520–2532.
46. Zhu C, Roth DB. Characterization of coding
ends in thymocytes of scid mice: implications
for the mechanism of V(D)J recombination.
Immunity 1995;2:101–112.
47. Livak F, Schatz DG. T-cell receptor alpha locus
