|Home | About | Journals | Submit | Contact Us | Français|
The PHD finger of the RAG2 polypeptide of the RAG1/RAG2 complex binds to the histone H3 modification, trimethylated lysine 4 (H3K4me3), and in some manner increases V(D)J recombination. However, in the absence of biochemical studies of H3K4me3 on purified RAG enzyme activity, the precise role of H3K4me3 has not been clear. Here, we find that H3K4me3 stimulates purified RAG enzymatic activity at both the nicking (2 to 5-fold) and hairpinning (3 to 11-fold) steps of V(D)J recombination. Remarkably, this stimulation can be achieved with free H3K4me3 peptide (in trans). This indicates that H3K4me3 functions via two distinct mechanisms. It not only tethers the RAG enzyme complex to a region of DNA, but it also induces a substantial increase in the catalytic turnover number (kcat) of the RAG complex. The H3K4me3 catalytic stimulation applies to suboptimal cryptic RSS sites located at H3K4me3 peaks that are critical in the inception of human T-cell acute lymphoblastic lymphomas.
In vertebrate immune cells, a diverse antigen receptor gene repertoire is achieved via the V(D)J recombination reaction (Fugmann et al., 2000; Geier and Schlissel, 2006; Jung et al., 2006). Each of the V, D, or J segment to be combined is flanked by a recombination signal sequence (RSS) with moderately conserved heptamer (CACAGTG) and nonamer (ACAAAAACC) sequences separated by non-conserved 12 or 23 base pairs (bp); hence, an RSS with either a 12 or 23 bp spacer is called a 12-RSS or 23-RSS, respectively. Efficient recombination requires one 12-RSS and one 23-RSS, referred to as the 12/23 rule (Fugmann et al., 2000; Gellert, 2002; Steen et al., 1997; Swanson et al., 2004).
V(D)J recombination consists of two phases: the cleavage phase done by the lymphoid-specific RAG1/RAG2 complex, and the rejoining phase done by the nonhomologous DNA end joining (NHEJ) pathway, which is required for the repair of general double-strand DNA breaks (DSBs) (Lieber, 2008; Sekiguchi and Ferguson, 2006; Sekiguchi et al., 1999). RAG1, RAG2 and HMGB1 form a complex (hereafter designated the RAG complex) that stably binds to DNA substrates containing a 12/23 RSS pair (synaptic complex) in the presence of divalent cations (Fugmann et al., 2000; Gellert, 2002; Swanson et al., 2004). The RAG complex binding site on DNA can be discerned by footprinting, which shows RAG protein covering both the heptamer and nonamer and a few base pairs of the adjacent coding end (Swanson, 2004; Swanson and Desiderio, 1998; Swanson and Desiderio, 1999). Cleavage by the RAG complex is catalyzed in two steps: nicking at the 5' end of the signal heptamer, followed by DNA hairpin formation by transesterification (Fugmann et al., 2000; Gellert, 2002; Swanson et al., 2004). Nicking can occur at a single RSS, while efficient hairpin formation occurs between a 12/23-RSS pair. Nicking by the RAG complex at the 5' end of the RSS generates a 3'-OH on the coding flank. Then the RAG complex uses the 3'OH of the coding flank to carry out a nucleophilic attack on the anti-parallel strand to create two DNA hairpins at the coding ends and two blunt signal ends with a 5'P and 3'OH (Roth et al., 1993; Schlissel et al., 1993).
Another important aspect of V(D)J recombination is the transition from the RAG-dependent cleavage phase to the NHEJ-dependent joining phase. The RAG complex remains on the cleaved DNA ends after the double-strand breaks. In vitro experiments using naked DNA indicate that holding of signal ends by RAGs after cleavage is quite tight, whereas holding of coding ends is relatively weak (Agrawal and Schatz, 1997; Jones and Gellert, 2001; Nagawa et al., 2004; Tsai et al., 2002). Consistent with this, in vitro reconstitution of coding joint formation indicates coding end release from the RAG post-cleavage complex, but persistence of RAGs with the signal ends (Lu et al., 2008). These observations raise the possibility that in vivo RAG post-cleavage retention of coding ends occurs largely at the chromatin level. The NHEJ apparatus is required to form the coding joint from the two hairpin coding ends and a signal joint from the two blunt signal ends (Ferguson and Alt, 2001; Lieber et al., 2003).
V(D)J recombination is regulated in a developmental and lineage-specific manner (Geier and Schlissel, 2006; Krangel, 2007; Schlissel, 2004a; Schlissel, 2004b). Accessibility of the RAG complex to an RSS is controlled by chromatin remodeling coupled with certain histone modifications at specific loci (Jung et al., 2006; Roth and Roth, 2000). Transient expression of the RAG complex (RAG1 and 2) in non-lymphoid cells can cleave the RSS on plasmid recombination substrates, but it can not cleave all RSSs in the genome (Bowen and Corcoran, 2008; McMurry and Krangel, 2000; Stanhope et al., 1996). Several studies have demonstrated that recombination substrates completely packaged with histones are suppressed for RAG cleavage, while chromatin remodeling by SWI/SNF enables the RAG complex to cleave the RSS substrates in vitro (Du et al., 2008; Golding et al., 1999; Kwon et al., 1998; Kwon et al., 2000; Stanhope et al., 1996). To date, the antigen receptor loci that undergo V(D)J recombination are known to have chromatin with histone H3 and H4 acetylation, and H3 hypermethylation at lysine 4 (H3K4me3). H3 and H4 acetylation serve to recruit chromatin remodeling enzymes. In contrast, H3K4me3 is recognized by the RAG complex via interaction with the RAG2 plant homeodomain (PHD) finger (Liu et al., 2007; Matthews et al., 2007; Ramón-Maiques et al., 2007). In those studies, it was proposed that interaction between H3K4me3 might tether or stabilize the RAGs at the antigen receptor loci or stimulate the RAGs via release of RAG inhibition. However, studies using purified RAG complex and H3K4me3 peptide were not done to assess these proposed mechanisms (Fig. 1).
The PHD finger is located at the carboxyl terminus of RAG2, and it is outside of the “core” RAG2 (amino acids 1 to 387 out of a total of 524)(West et al., 2005). Together with “core” RAG1 (amino acids 384 to 1008 out of a total of 1040), RAG2 is able to carry out V(D)J recombination on naked DNA in vitro and on plasmid substrates in vivo (Suppl. Fig. S1A) (Sekiguchi et al., 2001). Due to greater difficulty in purification of full-length RAG proteins, most early biochemical work on RAGs was done with core regions. However, some patients with severe combined immunodeficiency (SCID) and Ommen syndrome are found to carry mutations in the RAG2 PHD finger, indicating the essential role of this domain during V(D)J recombination (Baker et al., 2008; Ramón-Maiques et al., 2007; West et al., 2005). Consistent with this, mutations of residues contacting H3K4me3 impair V(D)J recombination on both plasmid substrates and genomic loci (Liu et al., 2007; Matthews et al., 2007; Ramón-Maiques et al., 2007). But, paradoxically, purified proteins of those mutants have normal cleavage activities in vitro (Liu et al., 2007; Matthews et al., 2007; Ramón-Maiques et al., 2007).
Here, we report that the H3K4me3 peptide can stimulate purified RAG complexes containing full-length RAG2, but not core RAG2. The stimulation occurs at both the nicking and the hairpinning steps. The stimulation can be achieved not only when the H3K4me3 is tethered to the RSS (in cis), but also when the H3K4me3 is in trans. Action in trans means that the H3K4me3 peptide is doing more than simply tethering the RAG complex to the RSS. That is, the peptide must be causing a change in the catalytic properties of the RAG complex, and we demonstrate a significantly improved catalytic turnover number (kcat). Though bound H3K4me3 efficiently recruits the RAG complex to the RSS in cis, this is not sufficient to preferentially retain coding ends in the RAG post-cleavage complex, though improved nonspecific DNA binding improves the overall stability of the post-cleavage complex. We also examine here the effect of H3K4me3 on RAG cleavage at the suboptimal cryptic RSSs that are found at sites of chromosomal translocation. We find stimulation of cleavage, consistent with our finding that such chromosomal translocation sites are particularly rich in H3K4me3 density in human hematopoietic precursors and T cells. These latter findings give some measure of the extent to which H3K4me3 contributes to the target site preferences of the RAG complex because the vast majority of potential cryptic sites in these regions are not located within the H3K4me3 peaks and are ignored.
We wondered whether the recently reported H3K4me3 effect on V(D)J recombination in vivo was via tethering of the RAGs to the substrate or due to a direct stimulation of RAG activity. To assess the impact of H3K4me3 binding on RAG activity, we first tethered the 21 amino acid H3 N-terminal tail to a heptamer/nonamer (RSS) DNA substrate. This was done by generating 12 and 23-RSS oligonucleotides with a biotin tag at the 3' end of only one strand, which can be bound with streptavidin protein. Since streptavidin (SA) exists as a tetramer, RSS-bound SA is able to bind up to three more biotin molecules. Therefore, end-labeled 12-RSS was incubated with streptavidin protein, followed by addition of biotin-tagged H3 peptide having unmodified or trimethylated K4, thereby creating peptide-bound RSS via the SA (Fig. 2A). Unbound free biotin-tagged RSS or free biotin-tagged peptide was removed by incubating with streptavidin magnetic beads. Products were analyzed on nondenaturing PAGE to evaluate 12-RSS bound protein complexes. The 60 kD SA efficiently bound to one 58 bp 12-RSS strongly (a minority species in the upper region is likely to be two 12-RSS bound to a single SA tetramer)(Fig. 2B, lane 1 versus 2). Adding unmodified H3 or H3K4me3 to the SA:12-RSS showed a small shift in both the lower and upper bands (Fig. 2B, lane 2 versus 3 & 4), which is due to the 2.7 kDa molecular mass of each peptide. Like the 12-RSS just described, unlabeled 23-RSS was also bound with SA and either unmodified H3 or H3K4me3 in the same manner. Then, these substrates were tested in in vitro cleavage assays with different combinations of core and/or full-length RAG proteins.
Full-length RAG2 tagged with maltose binding protein (MBP) was co-expressed with glutathione-S-transferase (GST)-tagged core RAG1 (384–1008 aa) in 293T cells and affinity-purified with amylose resin. In addition, different combinations of RAGs, MBP-core RAG1/GST-core RAG2 (1–387 a.a.), or MBP-full-length RAG1/GST-core RAG2, were also used (Suppl. Fig. S1A & B). We confirmed that SA or peptide-bound RSS had no effect on cleavage by core RAG1/core RAG2 (hereafter designated c/c) (Fig. 2C, lanes 2, 6, 10 and 14) or full-length RAG1/core RAG2 (hereafter designated f/c)(Fig. 2C, lanes 3, 7, 11 and 15). On the contrary, the reaction with H3K4me3-bound RSS and core RAG1/full-length RAG2 (hereafter designated c/f) showed 3.3-fold stimulation of RSS cleavage (Fig. 2C, lane 16). No effect was seen with unmodified H3, SA, or RSS alone (Fig. 2C, lanes 4, 8 and 12). Therefore, the presence of H3K4me3 associated with an RSS is important to stimulate the cleavage activity of a RAG complex having the full-length RAG2.
We next tested binding of RAGs to RSS-bound H3K4me3. Substrates were prepared in a similar manner as above, and then incubated with c/c or c/f RAGs in the presence of Ca2+ to form RSS-RAG complexes that are unable to cleave DNA (Bergeron et al., 2006). c/f RAGs showed poor RSS binding when the RSS was bound with unmodified H3 peptide (Suppl. Fig. S2, lanes 14 & 15), whereas significant stimulation of RAG binding was seen when the RSS was bound with H3K4me3 peptide (lanes 19 & 20). A species with a distinct mobility is seen in the binding reactions containing c/f RAGs and substrate bound to H3K4me3 peptide (indicated with asterisks in Suppl. Fig. S2, lanes 19 & 20), and the amount of that species relative to the free 12RSS:SA:H3 was substantial (compare lane 20 to 16 for H3K4me3 and lane 15 to 11 for unmodified H3 peptide). These results confirm that H3K4me3 can help to recruit the RAG complex to a target site. Though previous work showed that the RAG2 PHD finger could bind to genomic regions containing H3K4me3, the biochemical data here show that a purified RAG complex including this domain can bind to H3K4me3 and can do so while also acting on a cis RSS.
It was previously shown that the RAG2 PHD finger can recognize free H3K4me3 peptide as well as nucleosomes with the H3K4me3 modification (Liu et al., 2007; Matthews et al., 2007; Ramón-Maiques et al., 2007). Above, we found that RSS-bound H3K4me3 peptide can alter the cleavage activity of RAGs via its binding to the RAG complex. Therefore, we asked whether H3K4me3 must be bound to the RSS or whether free H3K4me3 peptide is sufficient to alter the RAG activity.
End-labeled 12-RSS and unlabeled 23-RSS oligonucleotide substrates were incubated with different combinations of RAGs in the absence or presence of various concentrations of the unmodified or trimethylated H3K4. Reaction products were separated by denaturing PAGE. The reactions with c/f RAGs and H3K4me3 showed a 4-fold increase in the generation of hairpin product, whereas there was no difference in reactions with unmodified peptide (Fig. 3, lanes 25–28 versus lanes 20–24). Reactions with either c/c or f/c RAG complexes were not altered in their nicking or hairpinning in the presence of both peptides (Fig. 3, lanes 2–19). These results indicate (a) that binding between the RAG2 C-terminus and H3K4me3 causes RAG conformational changes that modify its catalytic activity, independent of any tethering (Fig. 1B and C), and (b) that free H3K4me3 peptide has the ability to stimulate RAG activity, indicating that H3K4me3 can act in trans to stimulate RAG cleavage. These points were not demonstrated in the previous studies (Liu et al., 2007; Matthews et al., 2007; Ramón-Maiques et al., 2007).
We also tested RSS binding of the RAG complex in the presence of free H3K4me3 peptide. End-labeled 12-RSS was incubated with c/c or c/f RAGs on ice to form RSS-RAG complexes, and the products were separated on nondenaturing PAGE (Suppl. Fig. S3). These binding studies were done in the presence of a 10-fold excess of unlabeled nonspecific blunt duplex DNA. We see that the total amount RSS binding by c/c RAG and c/f RAG is similar (Suppl. Fig. S3, lanes 2 and 11). The free H3 unmodified peptide did not alter the binding of c/c or c/f RAG complexes (Suppl. Fig. S3, lanes 2–6 and lanes 11–15). But the free H3K4me3 peptide did specifically improve the RSS binding of the c/f RAG and not the c/c RAG complexes (Suppl. Fig. S3, lane 11 versus 16–19 and lane 2 versus 7–10). Therefore, the RAG complex binding to free H3K4me3 peptide does enhance the substrate recognition, in addition to activating cleavage activity of the RAG complex (see below and Discussion).
It is known that a point mutation in the PHD finger can interfere with H3Km4me3 binding to full-length RAG2 protein, but no effect on biochemical activity or purified RAG complexes was tested (Liu et al., 2007; Matthews et al., 2007; Ramón-Maiques et al., 2007). Using this same point mutation, we purified active c/f RAG complexes (Suppl. Fig. S1B). The mutant RAG complexes had very similar nicking and hairpinning activity as wild type c/f RAG complexes, but could no longer be stimulated by the H3K4me3 peptide (Fig. 4A). This demonstrates the specificity of the H3K4me3 enzymatic effects for the PHD finger of RAG2.
Previous work had shown that H3K4me2 has partial binding to the PHD finger of RAG2, but somewhat less than H3K4me3 (Liu et al., 2007; Matthews et al., 2007; Ramón-Maiques et al., 2007). H3K4me1 was much weaker, and unmodified H3 and other peptides did not bind. When we test nicking and hairpinning activity of the c/f RAG complexes, we find that the enzymology very nicely reflects the known binding observations (Fig. 4B and Suppl. Fig. S4). H3K4me3 > H3K4me2 > H3K4me1 >> unmodified H3 is the relative strength of enzymatic activity stimulation.
Because free H3K4me3 peptide is sufficient to alter the cleavage activity of the RAG complex, we were interested in knowing which steps during the RAG cleavage are stimulated. Cleavage by the RAG complex is catalyzed in two steps: nicking at the 5' end of the signal heptamer, followed by DNA hairpin formation involving the anti-parallel strand via a transesterification mechanism (Fugmann et al., 2000; Gellert, 2002; Steen et al., 1997; Swanson et al., 2004). To analyze nicking, time course experiments with end-labeled 12-RSS and c/f RAG complex were done (Fig. 5A). Robust nicking was observed in the first few minutes in the presence of H3K4me3 (Fig. 5A, lanes 16, 17) and full-length substrate was clearly consumed, compared to the reaction with unmodified H3 peptide or no peptide (Fig. 5A, lanes 15–21 versus 1–14). In addition, hairpin formation was detected at earlier time points in the reaction with H3K4me3. Nicking by c/f RAGs was accelerated upon binding to H3K4me3 at the early time points, and the stimulation of nicking is about 2-fold at the 2 min point (Fig. 5B). The stimulation at later time points is less because nicked products are converted into hairpin products.
We did the same experiment using an end-labeled 23-RSS as substrate. Addition of H3K4me3 stimulated nicking of the 23-RSS by c/f RAGs at the early time points by 4.6-fold at 1 min and 2.3-fold at 2 min, respectively (Fig. 5C and D). Hairpin products can be also detected at earlier time points for those reactions.
Given that H3K4me3 can stimulate the c/f RAG complex in trans, we wondered whether this is due to improved Km of the RAG complex for the RSS (i.e., heptamer/nonamer) or due to increased catalytic turnover rate (kcat). As we have previously done for the c/c RAG complexes (Yu and Lieber, 1999; Yu and Lieber, 2000), we did initial rate kinetic analysis for the c/f RAG complex, but now in the presence of either unmodified H3 or H3K4me3 (Fig. 6). Initial rates were done at a range of substrate concentrations (Fig. 6A & B). Burst kinetics were done at various substrate concentrations to establish that 3.4% of the c/f RAG complexes were catalytically active (similar to c/c RAG complexes (Fig. 6C)(Yu and Lieber, 1999; Yu and Lieber, 2000)).
Best-fits of the initial rate curves indicate that the Km for the c/f RAG complex is siimilar to the Km for the c/c RAGs that we published previously (Km = 58 nM for c/f RAG with unmodified H3 and Km = 40 nM for c/f RAG with H3K4me3)(Fig. 6D). Hence, the H3K4me3 stimulation is due only to a small (<1.5-fold) improvement in Km. However, the kcat increases from 0.81 min−1 when H3K4 is present (similar to c/c RAG complexes (Yu and Lieber, 1999; Yu and Lieber, 2000)) to a kcat of 2.82 min−1 with H3K4me3. Hence, the primary improvement in catalytic function for the RAG complex (c/f) comes from the improved catalytic (kcat) caused by H3K4me3 binding in trans.
We next asked whether H3K4me3 stimulates hairpin formation by RAGs. We performed time course experiments with c/f RAGs using end-labeled pre-nicked 12-RSS and unlabeled pre-nicked 23-RSS oligonucleotides. Although hairpin products in the reaction with or without unmodified H3 peptide are almost undetectable in the first 2 min, addition of H3K4me3 significantly stimulated hairpin formation at those time points (Fig. 7A lanes 1–3, 9–11 versus lanes 17–19). Moreover, robust hairpin products were also observed at the late time points. Overall stimulation of hairpinning by H3K4me3 is between 3.3 and 11.4-fold, with the largest stimulation in the initial rate of the reaction (~10-fold)(Fig. 7B).
We also examined the relative efficiency of hairpin formation between pre-nicked 12/23-RSS versus pre-nicked 12/12-RSS in the presence of H3K4me3. We find that H3K4me3 also stimulated hairpin formation for a 12/12-RSS pair, indicating that H3K4me3 does not contribute to enforcement of the 12/23 rule (Suppl. Fig. S5).
After the RAG complex completes coupled cleavage, four DNA ends, consisting of two hairpinned coding ends and two blunt signal ends, are generated. The RAG post-cleavage complex (PCC) consisting of either the two signal ends alone or two signal and two coding ends has been described in in vitro experiments (Jones and Gellert, 2001; Tsai et al., 2002). We examined this issue by capturing the complex directly in an in vitro V(D)J recombination assay. We found that the PCC retained the signal ends quite tightly, but retention of coding ends was much weaker using RSS substrates (data not shown). We wondered if H3K4me3 bound coding ends might increase RAG holding after cleavage.
In order to assess this issue, we utilized GST affinity pull-down assays to capture the PCC (Suppl. Fig. S6A). Because it is important to capture the RAG proteins tagged with GST efficiently in this assay, our GST-cRAG1/MBP-fRAG2 protein complex was affinity purified using MBP from cell lysates, and it was never exposed to glutathione during RAG purification. Unmodified H3 or H3K4me3 bound 12- and 23-RSS substrates were incubated with c/f RAGs and cleaved (Suppl. Fig. S6A). Four percent of the cleavage reaction was saved and designated as `input' prior to the purification procedure. The remaining reactions were followed by pull-down with magnetic beads covalently linked with glutathione. Beads were magnetically collected at the bottom of the tube and gently washed to remove nonspecific DNA associated with the beads (Suppl. Fig. S6B). All fractions were denatured with SDS/EDTA and associated DNA fragments were analyzed by denaturing PAGE. The level of the cleaved fragments was higher when the peptide is methylated at K4 (Suppl. Fig. S6C, lane 3,4). GST protein was used as a negative control in the pull-down assay. About half the cleaved fragments were detected in the unbound fraction because the GST-tagged RAG complex could not be completely captured by the GST beads, and some of the RAG-RSS complexes were released (Suppl. Fig. S6C,,lanes 7 and 8). In the secondary washes, no DNA was detected (Suppl. Fig. S6C, lanes 13–16). In the bound fractions, approximately 2–3 % of nonspecific binding of intact substrates was detected in the reaction with GST (Suppl. Fig. S6C, lanes 17 and 18).
As expected, accumulation of signal ends was observed in the bound fraction for both unmodified and methylated peptide-bound RSS (Suppl. Fig. S6C, lane 19, 20, and the quantitation of the recovery of DNA fragments in the PCC is summarized in Suppl. Fig. S6D). On the other hand, the amount of hairpin (H) coding ends associated with RAGs was much less than that of signal ends, indicating release of coding ends from the PCC. The ratio of the H3K4me3 or unmodified H3 bound-hairpin coding ends to the signal ends held in the PCC is 31 and 26 %, respectively, indicating that having H3K4me3 on the hairpin coding end does not specifically improve RAG holding of coding ends after cleavage relative to signal ends (Suppl. Fig. S6E).
Though H3K4me3 did not specifically increase coding end binding relative to the signal end number, it is interesting that H3K4me3 improves overall binding by the c/f RAG complex (Suppl. Fig. 6C, lanes 4, 8, & 20 versus 3, 7 & 19). Importantly, the H3K4me3 of sequence-independent binding is not seen for c/c RAGs (Suppl. Fig. 6F). Therefore, via the PHD finger, H3K4me3 induces an increase in RAG sequence-independent DNA binding. This is likely to be important for stabilizing the PCC. (The improved binding seen above in Suppl. Fig. S3 due to H3K4me3 is likely also due to the sequence-independent effects described here.)
The SCL-SIL deletions are the most common chromosomal rearrangement in T-cell acute lymphoblastic leukemia (ALL) and display all the characteristics of V(D)J recombination (Swerlow et al., 2008). However, neither the SCL nor the SIL cryptic RSS recombines efficiently on transfected plasmids ex vivo (Marculescu et al., 2002; Raghavan et al., 2001). In vitro, they exhibit minimal nicking and even lower hairpinning with core RAGs (Zhang and Swanson, 2008).
The genome-wide H3K4me3 distributions in both human T cells and hematopoietic stem cells have been reported (Barski et al., 2007; Cui et al., 2009). Using the resource files provided on the corresponding database, we found that chromosome breakpoints which cause certain subtypes of lymphoma and leukemia are actually located directly within distinctive peaks of activating histone modifications, including H3K4me3 (Suppl. Fig. S7A, see also Suppl Fig. S8) (Barski et al., 2007). In one remaining translocation hotspot (Hox11), the cryptic RSSs (cRSS) is within 1500 bp of the H3K4me3 peak, which is still relatively close (7 nucleosomes away).
In light of the proximity of these cRSSs to H3K4me3 regions, we were interested in knowing whether H3K4me3 also stimulated the weak or cryptic RSSs found at these breakpoints (Suppl. Fig. S7D). We tested SIL and SCL sequences in in vitro RAG cleavage assays. End-labeled single SIL or SCL oligonucleotide substrates were incubated with c/c or c/f RAGs to be nicked in the presence of free unmodified H3 or H3K4me3 peptide. Adding H3K4me3 indeed stimulated nicking at the SIL (Suppl. Fig.S7B, left panel, lanes 3, 6, 7). The SCL was a relatively poor signal, but its nicking was also improved by the H3K4me3 peptide (Suppl. Fig. S7B, right panel, lanes 3, 6, 7).
We next examined hairpin formation for the labeled SIL or SCL with a cold RSS partner. Hairpin formation on the SIL was greatly enhanced by H3K4me3 with either a 12- or 23-RSS partner (Suppl. Fig. S7C, top panel, lanes 5, 8, 9). For the reaction with the SCL substrate, the signal was much weaker than that of SIL, but its hairpin products were clearly increased in the presence of H3K4me3 and 12-RSS partner (Suppl. Fig. S7C, bottom panel in darker image inset). These biochemical observations and the chromatin immunoprecipitation data from the database (Barski et al., 2007; Cui et al., 2009) implicate the H3K4me3 modifications at these chromosome breakpoints in the misrecognition by RAGs in these T-ALL chromosomal translocations/deletions.
In seminal work, it was shown that the RAG complex can bind to H3K4me3 peptide in vivo and in vitro (Liu et al., 2007; Matthews et al., 2007; Ramón-Maiques et al., 2007). Moreover, forms of the RAG complex that can bind to H3K4me3 are able to recombine extrachromosomal substrates more efficiently (Liu et al., 2007; Matthews et al., 2007). An important next step in understanding the interplay between chromatin and the RAG complex is to determine how H3K4me3 influences RAG biochemical function. Direct enzymatic assessments of the effects of H3K4me3 on the RAG binding, nicking, hairpinning and postcleavage activities are essential to determine what impact H3K4me3 is having on the RAG complex.
Not surprisingly, we show here H3K4me3 does provide an additional `tether' for the RAG complex to the substrate beyond the RAG binding to the heptamer/nonamer. Hence, mechanism 1 of Figure 1 does apply. This would seem to be an obvious extension of the previous work and to some extent it is. But the previous work did not determine whether the H3K4me3 binding could apply to an active RAG complex and one that was enzymatically active on an RSS.
More importantly, the studies here also demonstrate that mechanism 2 applies. H3K4me3 stimulates both the RAG nicking and hairpin formation steps. Saturating amounts of H3K4me3 peptide stimulate equally well in cis (when the H3K4me3 peptide is tethered to the RSS) or in trans (free in solution). The equivalent stimulation via cis or trans indicates that the H3K4me3 is not merely functioning by mechanism 1, because the trans form of the H3K4me3 cannot possibly serve a tethering function (i.e., no mechanism 1 function). Hence, there must be a RAG conformational change induced by H3K4me3 that improves the catalytic function of the RAG complex (i.e., mechanism 2) in a manner that improves the RAG catalytic activity for both the nicking and hairpin formation steps. We show that the basis of the improved catalytic activity is a primarily an improved catalytic turnover number (kcat) by the RAG complex when H3K4me3 is present in trans as compared to H3K4.
An equimolar amount of c/c RAG complex is no more active than the c/f RAG complex in our studies (Suppl. Fig. S1C), and the kcat and Km for c/c RAG complexes are not significantly different from those of c/f RAG complexes (Yu and Lieber, 2000). Therefore, we see no biochemical evidence that the full-length form of purified RAG2 is inhibitory or that the H3K4me3 is relieving this inhibition, and this is consistent with what was seen previously (Elkin et al., 2005). The apparent in vivo inhibitory effect seen previously (Matthews et al., 2007), as pointed out by those authors, may be due to other proteins that bind to the RAG2 PHD finger within the cell unless H3K4me3 is present to disrupt that inhibitory protein. Alternatively, in vivo specific activity differences of RAG complexes could also explain the impression that there is some type of anti-repression.
Given our results, both mechanisms 1 and 2 clearly apply to the role of H3K4me3. That is, H3K4me3 in cis (on a nearby histone octamer) serves as an independent binding site to stabilize the RAG complex at the nearby RSS binding site (mechanism 1; see Fig. 1). In addition, H3K4me3 improves the catalytic turnover number of the RAG complex (mechanism 2), presumably by causing a conformational change of the RAG complex active site.
When H3K4me3 is present, we note that the c/f RAG complex pulls down all DNA with substantially improved efficiency (Suppl. Fig. S6C & F). [This does not occur with c/c RAGs or with H3K4 on c/f RAGs.] While this effect is not specific to the coding ends, it would still help the RAG complex transiently hold on to the coding ends, after the hairpin formation step. Perhaps this is one way in which release of the coding ends is minimized. Sequence-independent holding of the phosphate backbone at any distance (i.e., nonspecific binding) might be the optimal mode for the RAG complex to hold onto the coding ends, given that the coding ends must undergo hairpin opening by the Artemis:DNA-PKcs complex in cis, action by polymerases (TdT, as well as polymerase mu and lambda), and ligation by the XLF:XRCC4:ligase IV complex. A sequence-independent RAG interaction anywhere in the backbone of the coding end would still permit these other enzymes to act on the coding end terminus. In addition, it is consistent with the fact that the RAG complex binds to all coding ends, regardless of their DNA sequence (Fugmann et al., 2000; Gellert, 2002; Steen et al., 1997; Swanson et al., 2004).
In human T-cell ALL, some of the common translocations involve recombination of oncogenes with the T-cell antigen receptor loci such that the oncogene becomes expressed in T cells when it is not normally expressed, such as for the Hox11, Ttg-2 (also called LMO2), and Ttg-1 translocations (Swerlow et al., 2008). About 7% of T-cell ALL involves a translocation of the LMO2 gene (also called Ttg-2); <1% involves Ttg-1; and 5 to 10% involves Hox11. In human T-cell ALL, 25 to 30% involve the interstitial deletion between SIL and SCL, resulting in the inappropriate expression of the tal-1 (SCL) gene, which is not normally expressed in T cells (Swerlow et al., 2008).
We previously used replicating extrachromosomal substrates to assay the relative strength of the cRSS sites associated with all of these oncogene translocations (Raghavan et al., 2001). If one assigns consensus heptamer/nonamer RSSs a strength of 20,000, then the relative strengths of these cryptic RSSs are 760 for LMO2; 38 for Ttg-1; and 27 for SIL. HOX11 and SCL did not give measurable recombination levels, but were screened to levels less than 1. These replicating minichromosomes are only 6kb in size and over half of that length is the SV40 large T transcription unit (Gauss et al., 1998; Gauss and Lieber, 1996). We have not measured H3K4me3 on these substrates, but the very similar murine substrates are rich in H3K4me3 (Liu et al., 2007; Matthews et al., 2007; Ramón-Maiques et al., 2007), as are other replicating human minichromosomes (Okitsu and Hsieh, 2007).
Biochemical study of core RAG complex nicking at these same sites demonstrated levels of nicking that were similar to ours (Zhang and Swanson, 2008). However, those authors did not observe very much hairpinning for these cRSS sites. Our finding of increased nicking and hairpinning illustrates the importance of H3K4me3 stimulation for the c/f RAG complex even at suboptimal cryptic sites (Suppl. Fig. S7B & C).
In the current study, we searched the resource database kindly provided by Dr. Keji Zhao as part of his published study of T cell chromatin modifications (see Supplemental Discussion). We found that the cryptic RSS sites are ones that are located directly within (for SCL and SIL, as well as Hox11 and Ttg-1) or very near (<1500 bp for LMO2) peaks of H3K4me3 (Suppl. Fig. S7A and S8) at levels that parallel those that regulate the antigen receptor loci in those T cells (Suppl. Fig. S9). Our biochemical studies here show that cryptic RSS sites that are very weak for hairpinning can be stimulated by H3K4me3 to more efficient levels of hairpin formation. Hence, our findings of H3K4me3 stimulation of nicking and hairpinning for weak cryptic RSS sites and the proximity of H3K4me3 to these cryptic RSS sites in vivo provides a circumstantial but compelling case for the importance of H3K4me3 at specific translocation sites. Over 1000 potential cryptic heptamers are located in the SIL/SCL deletion zone, but the only ones that were actually recombined by the RAG complex in the interstitial deletions of human T-acute lymphoblastic lymphoma are the ones that are within the two distinct H3K4me3 peaks.
(See Supplemental text)
Construction of expression vector for the glutathione S-transferase (GST) fused core region of mouse RAG1 or RAG2 was described elsewhere (Yu and Lieber, 1999; Yu and Lieber, 2000). DNA fragment of maltose binding protein (MBP) fused core region or full-length mouse RAG1 or RAG2, kind gift from Dr. Patrick Swanson (Creighton U.), was inserted in the pEBB vector under control of the EF-1α promoter (Bergeron et al., 2006). An amino acid substitution in RAG2 was created to make the W453R mutant, and this was introduced using QuickChange method (Stratagene, La Jolla, CA) according to the manufacturer's instructions and confirmed by sequencing. The mammalian expression vectors for the fusion proteins for maltose binding protein (MBP) and the core region of mouse RAG1 (amino acids 384 to 1008), full-length of mouse RAG1, or full-length of mouse RAG2 were co-transected into 293T cells with the vector for the fusion protein for glutathione S-transferase (GST) and the core region of mouse RAG2 (amino acids 1 to 383), or mouse RAG1, respectively. Proteins were purified over an amylase resin (New England Biolab) column as described previously (Bergeron et al., 2006). Recombinant HMGB1 protein was expressed in bacteria and purified as described previously (West and Lieber, 1998).
We thank members of the Lieber lab, Dr. Patrick Swanson (full-length murine RAG1 and RAG2 cDNA), and Dr. Chih-Lin Hsieh. This work was supported by NIH grants to M.R.L. and postdoctoral fellowships from the Uehara Memorial Foundation and Japan Society for the Promotion Science to N.S.