ABSTRACT
The recombination activating gene (RAG) 1 and 2 proteins are required for
initiation of V(D)J recombination
in vivo
and have been shown to be sufficient to introduce DNA double-strand breaks at recombination signal sequences (RSSs) in a cell-free assay
in vitro
. RSSs consist of a highly conserved palindromic heptamer that is separated from
a slightly less conserved A/T-rich nonamer by either a 12 or 23 bp spacer of random sequence. Despite
the high sequence specificity of RAG-mediated cleavage at RSSs, direct binding of the RAG proteins to these
sequences has been difficult to demonstrate by standard methods. Even when this
can be demonstrated, questions about the order of events for an individual RAG-RSS complex will require methods that monitor aspects of the complex
during transitions from one step of the reaction to the next. Here we have used
template-independent DNA polymerase terminal deoxynucleotidyl transferase (TdT) in
order to assess occupancy of the reaction intermediates by the RAG complex
during the reaction. In addition, this approach allows analysis of the
accessibility of end products of a RAG-catalyzed cleavage reaction for N nucleotide addition. The results
indicate that RAG proteins form a long-lived complex with the RSS once the initial nick is generated, because the
3
'
-OH group at the nick remains obstructed for TdT-catalyzed N nucleotide addition. In contrast, the 3
'
-OH group generated at the signal end after completion of the cleavage
reaction can be efficiently tailed by TdT, suggesting that the RAG proteins
disassemble from the signal end after DNA double-strand cleavage has been completed. Therefore, a single RAG complex
maintains occupancy from the first step (nick formation) to the second step
(cleavage). In addition, the results suggest that N region diversity at V(D)J
junctions within rearranged immunoglobulin and T cell receptor gene loci can
only be introduced after the generation of RAG-catalyzed DNA double-strand breaks, i.e. during the DNA end joining phase of the V(D)J
recombination reaction.
In vertebrates the antigen binding domains of immunoglobulin (Ig) and T cell
receptor (TcR) molecules are assembled from numerous variable (V), joining (J)
and sometimes diversity (D) gene segments that are separated in the germline (
1
). These gene segments are assembled in a random, yet site-specific manner during early lymphoid development by a process called
V(D)J recombination, which generates the diversity of antibody and TcR
molecules (
2
,
3
). The coding regions of all functional V, D and J gene segments are flanked by
conserved recombination signal sequences (RSS) that consist of a palindromic
heptamer, which is separated from an A/T-rich nonamer by a random spacer sequence of either 12 or 23 bp length.
These RSSs are located directly adjacent to the coding region sequence. Site-specific recombination of V, D and J gene segments
in vivo
has been found to require the presence of two RSSs of different spacer length,
which is generally referred to as the 12/23 bp rule (
4
).
Initiation of V(D)J recombination requires expression of the two recombination
activating genes
RAG
-
1
and
RAG-2
(
5
,
6
). If either of the two
RAG
genes is non-functional, Ig and TcR loci remain in the germline configuration and mice
carrying these mutations are devoid of peripheral B and T lymphocytes (
7
,
8
). Later stages of a complete V(D)J recombination reaction appear to involve
factors that are also necessary to mediate the repair of DNA double-strand breaks, like the 470 kDa DNA-dependent protein kinase, which appears to be mutated in SCID mice (
9
,
10
), the DNA end binding proteins Ku70 and Ku86 (
11
-
13
) and the recently identified
XRCC-4
gene product (
14
).
Recently, a cell-free assay for the initiation of V(D)J recombination has been developed (
15
) and it has been demonstrated that both RAG proteins alone are sufficient to
catalyze the initial DNA double-strand breaks at RSS (
16
). The cleavage reaction does not require a high energy cofactor and occurs by a
two-step mechanism. First, a single-strand nick is introduced at the border of the heptamer of the RSS
and the adjacent coding end, which is followed by nucleophilic attack of the
in situ
generated 3'-OH group at the nick on the phosphodiester bond of the opposite,
complementary DNA strand (
16
). This second step occurs by a direct transesterification reaction, which does
not involve a covalent protein-DNA intermediate and is therefore reminiscent of retroviral integration (
17
). The end products of a RAG-catalyzed cleavage reaction are a hairpin coding end and a blunted, 5'-phosphorylated signal end retaining the RSS.
This reaction can occur at an isolated RSS when it is carried out in the
presence of Mn
2+
as divalent cation. Some limited degree of concerted cutting at presumably
synapsed RSSs with different spacer lengths (i.e. following the 12/23 bp rule)
is favored in the presence of Mg
2+
(
18
,
19
). The degree of concerted cleavage at two signals is markedly lower than
cleavage in crude extracts (
19
). Therefore, for the current study we have focused on the use of purified RAGs
in the cleavage of a single RSS using Mn
2+
as the metal cofactor. Despite the strong sequence specificity of the RAG-mediated cleavage reaction, documentation of binding of the RAG proteins
to RSSs by conventional methods, including gel mobility shift assays or DNA-protein crosslinking experiments, has been difficult thus far (see
Discussion). One possible explanation for this is the rapid on-off rates for RAG-1 and RAG-2 binding at the RSSs. This prompted us to analyze the
accessibility of the reaction intermediate and end products of the RAG cleavage
reaction for N nucleotide addition by the enzyme terminal deoxynucleotidyl
transferase (TdT) in order to indirectly assess whether a more stable complex
might form between RAG proteins and target DNA at any point during the cleavage
reaction.
Here we show that the internal 3'-OH group of the nicked intermediate, which is generated during the
first step of a RAG-catalyzed cleavage reaction and which remains unconverted for at least 3 h
during the assay, is resistant to N nucleotide addition by TdT, indicating that
the RAG proteins form a complex with the substrate DNA that persists after the
single-strand cleavage step (nicking) has occurred and until completion of the
strand transfer reaction (hairpin formation).
Recombinant baculoviridae for expression of maltose binding protein (MBP)-tagged, truncated mouse RAG-1 and RAG-2 proteins were kindly provided by Drs Dik van Gent and Martin
Gellert (NIH). MBP-tagged RAG proteins were expressed in Hi-5 insect cells (Invitrogen, San Diego, CA) grown at 28oC in either suspension cultures on a shaker set to 100 r.p.m.
or as adherent cells in 150 mm dishes in Ex-Cell 400 serum-free insect medium (JRH Biosciences, Woodland, CA). Expression of
recombinant RAG proteins was generally carried out by co-infecting adherent Hi-5 cells for 48-72 h with recombinant RAG-1 and RAG-2 baculovirus stocks at ratios between 1:3 and
1:5 with a multiplicity of infection ranging between 5 and 10.
DNA oligomers were synthesized by the Protein and Nucleic Acid Chemistry
Laboratory (PNACL) at Washington University School of Medicine (St Louis, MO).
Crude oligomers were gel purified by dentauring PAGE and isolated by a standard
crush and soak procedure. Aliquots of 10 pmol were used for 5'-labeling with [[gamma]-
32
P]ATP, which were subsequently annealed to 10 pmol cold complementary DNA by
heating to 96oC in 25 mM Tris-HCl, pH 8.0, 100 mM NaCl for 10 min, followed by slow cooling from
65oC to room temperature over a time period of 1-2 h. Samples of 1 [mu]l (0.1 pmol) annealed substrate were used in the standard
cleavage reaction.
Primer sequences.
Sequences for substrate B (79mer substrate): UG078 (top strand), 5'-ATCAGGATGTGGTGATCCACAGTGTGATCCCTCCTCACAAAAACCGCAGGTCTTCAGTT-3'; UG079 (bottom strand), 5'-AACTGAAGACCTGCGGTTTTTGTGAGGAGGGATCACACTGTGGATCACCACATCCTGAT-3'. Sequences for substrate C
(79mer substrate), UG211 (top strand): 5'-ATCAGGATGTGGTACAGTGTGATCCCTCCTCACAAAAACCGCAGGTCTTCAGTT-3'; UG079 (bottom strand), 5'-AACTGAAGACCTGCGGTTTTTGTGAGGAGGGATCACACTGTGCATCACCACATCCTGAT-3'. Sequences for substrate A
(50mer) and its pre-nicked derivatives: UG071 (top strand), 5'-TTGCATCGCATGCCTCCACAGTGATCATCCTCGAGACAAAAACCTGCAAC-3', UG072 (bottom strand), 5'-GTTGCAGGTTTTTGTCTCGAGGATGATCACTGTGGAGGCATGCGATGCAA-3'; UG073 (top
strand/coding end), 5'-TTGCATCGCATGCCTC-3'; UG074 (top strand/signal end), 5'-CACAGTGATCATCCTCGAGACAAAAACCTGCAAC-3'; UG075 (top strand/coding
end, gap -1), 5'-TTGCATCGCATGCCT-3'; UG076 (top strand/coding end, flap +1), 5'-TTGCATCGCATGCCTCA-3'; UG077 (top
strand/coding end, flap +5), 5'-TTGCATCGCATGCCTCAAGTG-3'.
MBP-tagged RAG proteins were isolated from insect cells essentially as
described by McBlane
et al
. (
16
). In brief, infected insect cells were sonicated for 2 min on ice in binding
buffer (25 mM Tris-HCl, pH 7.5, 500 mM NaCl, 20 mM imidazole, 2 mM 2-mercaptoethanol), cell debris was removed by centrifugation at 12
000 r.p.m., 4oC, 15 min and cleared supernatants were incubated with Ni-NTA agarose beads (Qiagen, Hilden, Germany) in batch mode at 4oC by end-over-end rotation for 1 h. The beads were washed three
times with binding buffer and RAG proteins were batch-eluted with binding buffer containing 1000 mM imidazole. The eluate was 5-fold diluted in amylose resin binding buffer (25 mM Tris-HCl, pH 8.0, 500 mM NaCl, 0.25% Tween 20, 2 mM
dithiothreitol) and was then incubated with amylose resin (NEB Biolabs,
Beverly, MA) in batch mode at 4oC by end-over-end rotation for 1 h. The amylose beads were washed twice with
binding buffer and twice with binding buffer without Tween 20 and RAG proteins
were eluted with 25 mM Tris-HCl, pH 8.0, 500 mM NaCl, 2 mM dithiothreitol, 20 mM maltose. Proteins
were finally dialyzed for 3 h at 4oC against 25 mM Tris-HCl, pH 8.0, 150 mM KCl, 10% glycerol, 2 mM dithiothreitol. Protein
samples were analyzed by SDS-PAGE followed by SYPRO-orangetm (Molecular Probes, Eugene, OR) staining and visualized on
either a regular UV light box or using a STORM Phosphorimager equipped with
ImageQuant[ordf] software (version 1.1) (Molecular Dynamics, Mountain View, CA). Proteins
were further visualized by silver staining of the protein gels using a BioRad
silver stain kit (BioRad, Hercules, CA) according to the recommendations of the
manufacturer. Based on SYPRO-orange and silver staining, all RAG protein samples had a purity of >95%.
RAG protein-catalyzed cleavage reactions were generally carried out in 25 mM MOPS-KOH, 7.5 mM Tris-HCl, 10 mM NaCl, 30 mM KCl, 60 mM potassium glutamate, 2.4
mM dithiothreitol, 1 mM MnCl
2
, 2% glycerol, pH 7.2 (including the components from the dialysis buffer and
from the buffer used to anneal the substrates) for 3 h at 37oC with 100 fmol labeled DNA substrate and recombinant RAG-1 and RAG-2 in the range 50-100 fmol. Cleavage and hairpin activity was not
detectable if RAG-1 or RAG-2 were used individually.
TdT tailing assays were carried out in the same buffer (whenever TdT tailing was
performed in the absence of RAG proteins, the reaction volume was adjusted with
dialysis buffer to achieve identical ionic conditions) supplemented with 10 [mu]M dNTPs and 2 [mu]M ddNTPs for 3 h at 37oC in a 10 [mu]l volume. The dideoxynucleotides were added in order to limit
the degree of N nucleotide addition at 3'-OH groups. TdT was obtained from Boehringer Mannheim Biochemicals.
Cleavage and/or TdT tailing reactions were stopped on ice by addition of 10 [mu]l formamide, 1 [mu]l 100 mM EDTA and 1 [mu]l 1% SDS. Samples were denatured at 99oC for 10 min, cooled on ice and immediately fractionated by 20%
denaturing PAGE containing 8 M urea using a regular sequencing unit. Specific
signals were visualized by overnight autoradiography at -80oC and/or phosphorimager analysis. Quantification of signals was
carried out using the ImageQuant 1.1 software package (Molecular Dynamics).
Recombinant core RAG-1 (amino acids 384-1004) and RAG-2 (amino acids 1-387) proteins were expressed in Hi-5 insect cells as MBP fusion proteins (with the
MBP tag fused to the N-terminus) containing an additional His
9
tag and three myc epitopes at the C-terminus for immunodetection and further purification (
16
). After two affinity purification steps using Ni-NTA-agarose and amylose resin (see Materials and Methods), RAG
proteins were obtained that were >95% pure as judged by SYPRO-orangetm protein staining (data not shown). In order to obtain active RAG
proteins, insect cells had to be co-infected with recombinant RAG-1 and RAG-2 baculovirus stocks at ratios ranging between 1:5 and 1:3
(data not shown).
The efficiency of endonucleolytic cleavage at recombination signal sequences by
these RAG proteins was strongly dependent on the coding end sequence of the
substrate if the reaction was carried out in the presence of Mn
2+
as divalent cation (
17
,
20
). To illustrate this finding, Figure
1
shows the kinetics of a cleavage reaction for two different substrates that
differ by only 1 bp at the terminal position of the coding end, i.e. directly
adjacent to the palindromic heptamer of the RSS. If the top strand of a
cleavage substrate is labeled at the 5'-end, the nicked intermediate of the cleavage reaction can be
detected by denaturing polyacrylamide gel electrophoresis as a 17 nt DNA
fragment, which can be discriminated from the covalent hairpin coding end
migrating at twice the size, i.e. 34 nt (Fig.
1
). The hairpin results from nucleophilic attack of the newly generated free 3'-OH at the nick on the phosphodiester bond of the opposite
complementary DNA strand, thereby leading to a double-strand break at the border of the RSS with the coding end (
17
).
One way to detect whether the intermediates are stably bound by RAG proteins
throughout the cleavage reaction is to investigate to what extent the
in situ
generated internal 3'-OH group at the initial nick might be accessible for N nucleotide
addition catalyzed by TdT. In order to demonstrate that a 3'-OH group at a nick can serve as a substrate for TdT, three DNA
oligomers were designed, 5'-phosphorylated and annealed to generate the nicked intermediate
structure. The sequence of this substrate was the same as the RSS substrate,
displaying a slow rate of hairpin formation. To confirm that the three DNA
oligomers would anneal in the predicted fashion (i.e. form the nicked double-stranded DNA molecule), the bottom strand was 5'-labeled, annealed with the two complementary oligomers and
digested with restriction enzymes that would cut upstream (
Sph
I and
Fok
I) and downstream (
Mbo
I and
Xho
I) of the internal nick. In each case >95% of the substrate was specifically
cleaved by these restriction enzymes (data not shown), confirming that the DNA
oligomers did anneal in the predicted fashion to generate the nicked double-stranded DNA substrate.
TdT-catalyzed N nucleotide addition at such an internal nick can be detected
if the DNA oligomer upstream of the nick is labeled at its 5'-end. As shown in Figure
2
, N nucleotide addition was readily detectable at the internal nick, beginning
at TdT concentrations as low as 0.2 U/1 [mu]l reaction (20 U/ml) and increasing in intensity as the TdT concentration
was raised to 25 U/10 [mu]l reaction (2500 U/ml), a concentration at which >98% of the nicked DNA
substrate could be tailed by TdT. All TdT tailing reactions were carried out in
the presence of RAG cleavage buffer (see Materials and Methods), with Mn
2+
as the divalent cation in order to maintain reaction conditions identical to
those chosen for RAG-mediated cleavage reactions. This degree of N nucleotide tailing was
identical to that observed with a blunt-ended DNA substrate (data not shown).
We analyzed the extent to which N nucleotides could be added by TdT to either
the nicked intermediate, the newly generated blunted signal end or the 3'-OH group at the distal end of the hairpin. All these products
became detectable during a RAG-catalyzed cleavage reaction at comparable levels if the two different
substrates with either a propensity for nicking or hairpin formation were used.
Titration of various amounts of TdT into the RAG-catalyzed cleavage reaction demonstrates that N nucleotides can be added
to the signal end after completion of the cleavage reaction (Fig.
3
, right panel), as evidenced by a similar degree of TdT tailing observed at the
3'-OH group of the hairpin (Fig.
3
, middle panel), which should be readily accessible for TdT since it is located
17 bp from the RSS. However, no N nucleotide addition was detectable at the 3'-OH group that results from single-strand nicking at the RSS as the first step of a RAG-catalyzed cleavage reaction (see Fig.
3
, left panel). In fact, at least 90% of the 3'-OH groups of the nick in the reaction intermediate were completely
resistant to TdT tailing at concentrations of up to 5 U/10 [mu]l reaction, which resulted in >95% tailing in the case of a control nicked
substrate (Fig.
2
).
If the nicked intermediate is not susceptible to TdT-catalyzed addition because of RAG complex occupancy, then phenol
extraction would be expected to reverse this. This was tested by incubating the
RAG core proteins with the substrate to generate the nicked species and a small
amount of hairpin product (Fig.
4
). Half of the reaction mix was phenol extracted and half was not. The TdT
tailing assay shows that while the phenol-extracted nicked species could readily be tailed, the 3'-OH is protected in the unextracted reaction (Fig.
5
).
Recombination signal sequences that flank coding regions of TcR and Ig gene
segments are highly conserved during vertebrate evolution. Although they are an
indispensable requirement for the recognition and site-specific cleavage catalyzed by RAG proteins, direct binding of RAG
proteins to these sequences has not yet been reported by standard methods.
Recently, RAG-1 has been shown to interact with the nonamer (
24
,
25
) and, while this manuscript was under review, RAG-1 and RAG-2 have been shown to form a specific complex with a complete single
signal sequence when stabilized by glutaraldehyde (
26
). We have used an indirect approach to assess RAG protein-DNA interaction by using the template-independent DNA polymerase TdT in order to probe the intermediates
and end products of the RAG cleavage reaction for accessibility to N nucleotide
tailing. In order to be able to compare the extent of TdT tailing on either the
reaction intermediates or the end products, we took advantage of the finding
that in cleavage buffer with Mn
2+
as divalent cation, different coding ends strongly affect the outcome of the
RAG-catalyzed cleavage reaction. Hence, different substrates would yield
comparable levels of either intermediate or end products.
The results from these analyses are that the nicked intermediate that is formed
as a first step of a RAG-mediated cleavage reaction is resistant to N nucleotide tailing by TdT
(despite the fact that the same structure is a good substrate for TdT under
identical assay conditions). Some of our substrates result in substantial
conversion to double-strand cleavage product (15-20% in Fig.
5
) and yet the nicked intermediate remains completely protected by RAGs, here as
well as when conversion to cleavage product is much lower. In contrast, the 3'-OH group generated at the signal end after completion of the
reaction is readily accessible for TdT tailing. These results indicate that RAG
proteins form a persistent complex with the substrate DNA once the reaction has
been initiated by nicking at the heptamer/coding end border.
It should be noted that our analysis provides information only after the nicking
step. We cannot infer anything about binding of RAG proteins prior to this step
because the modifying agent, TdT, is only capable of adding nucleotides after
the nick is created.
Recent results have provided information on the V(D)J reaction pathway (
15
,
17
,
18
) and our results allow additional detail to be included in such a reaction
scheme for a single recombination signal (Fig.
6
). None of these steps is ATP dependent (
15
,
17
,
18
). The initial binding step is reversible. The strand cleavage step (Fig.
6
, step II) is depicted as irreversible; a reverse reaction would represent the
formation of a phosphodiester bond in the absence of an energy source. Once the
nicked intermediate is formed (Fig.
6
, structure C), our results indicate that TdT has no significant access to the 3'-OH of the nick. TdT access can only be detected at the signal end
after the hairpin is formed (Fig.
6
, structure D). This indicates that RAG complex dissociation from the nicked
intermediate (Fig.
6
, structure C) to generate an unbound nick (Fig.
6
, structure E) can only occur to a relatively small extent and this must be
below the detection limit of the TdT assay used here. The fact that an unbound
nicked intermediate (E) can be converted to hairpin product (D) in the presence
of RAG proteins indicates the an equilibrium between C and E exists, as shown
here and by McBlane
et al
. (
16
). However, our data indicate that the equilibrium between structures C and E is
shifted heavily in favor of C. Hence, the pathway from A -> B -> C -> D, without diversion to E, is likely to be the favored one.
We would like to thank Drs Dik van Gent and Martin Gellert for the generous gift
of the baculovirus stocks that were required for expression of the maltose
binding protein tagged RAG-1 and RAG-2 proteins. We would further thank them for communication of results
prior to publication and their interest in the progress of this study. U.G. is
a post-doctoral fellow of the Boehringer Mannheim Foundation.
REFERENCES
Return