H3k4me3 Induces Allosteric Conformational Changes in the DNA-Binding and Catalytic Regions of the V(D)J Recombinase

H3K4me3 induces allosteric conformational changes in the DNA-binding and catalytic regions of the V(D)J recombinase

John Bettridgea,b, Chan Hyun Nac,d, Akhilesh Pandeyc,d, and Stephen Desiderioa,b,1

aDepartment of Molecular Biology and Genetics, The Johns Hopkins University School of Medicine, Baltimore, MD 21205; bInstitute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD 21205; cMcKusick-Nathans Institute of Genetic Medicine, The Johns Hopkins University School of Medicine, Baltimore, MD 21205; and dDepartment of Biological Chemistry, The Johns Hopkins University School of Medicine, Baltimore, MD 21205

Edited by Frederick W. Alt, Boston Children’s Hospital and Howard Hughes Medical Institute, Harvard Medical School, Boston, MA, and approved December 27, 2016 (received for review September 26, 2016)

V(D)J recombination is initiated by the recombination-activating cleft and participates in formation of the active site (11). Although gene (RAG) recombinase, consisting of RAG-1 and RAG-2 subunits. residues 387 through 527 of RAG-2 are dispensable for DNA The susceptibility of gene segments to cleavage by RAG is associated cleavage in vitro, this region supports several regulatory functions with histone modifications characteristic of active chromatin, in- in vivo, including binding of RAG-2 to H3K4me3 (7, 8, 12). This cluding trimethylation of histone H3 at lysine 4 (H3K4me3). Binding function is mediated by a noncanonical plant homeodomain (PHD) of H3K4me3 by a plant homeodomain (PHD) in RAG-2 stimulates finger that spans residues 415 through 487 (12, 13). Engagement of substrate binding and catalysis, which are functions of RAG-1. This H3K4me3 by the PHD finger promotes recombination in vivo (7, 8) has suggested an allosteric mechanism in which information re- and synthetic peptides bearing the H3K4me3 modification stimu- late cleavage of RSS substrates by RAG in vitro (14–16), consistent garding occupancy of the RAG-2 PHD is transmitted to RAG-1. To with the interpretation that H3K4me3 is an allosteric activator of determine whether the conformational distribution of RAG is altered the V(D)J recombinase. by H3K4me3, we mapped changes in solvent accessibility of cysteine This interpretation was reinforced by the identification within thiols by differential isotopic chemical footprinting. Binding of RAG-2 of an autoregulatory region, the presence of which was H3K4me3 to the RAG-2 PHD induces conformational changes in revealed by second site mutations that rescue the activity of RAG-1 within a DNA-binding domain and in the ZnH2 domain, which RAG-2 lacking a functional PHD finger (15). Disruption of this acts as a scaffold for the catalytic center. Thus, engagement of autoregulatory region is associated with constitutive increases in RSS H3K4me3 by the RAG-2 PHD is associated with dynamic conforma- binding affinity, catalytic rate, and recombination frequency, thus tional changes in RAG-1, consistent with allosteric control by mimicking the stimulatory effects of H3K4me3 (15). These obser- active chromatin. vations support a model in which RAG activity is suppressed by an autoinhibitory domain whose action is relieved by active chromatin. DNA recombination | genomic plasticity | allosteric control | epigenetic Allosteric activation is usually accompanied by a change in the modification | immune development distribution of accessible protein conformations in the ligand- bound state (17). We asked whether the conformational distribution of RAG is altered upon binding of H3K4me3 to the RAG-2 ll forms of DNA processing—replication, transcription, re- — PHD finger. To quantify effects of H3K4me3 on the conformations Acombination, and repair use allosteric regulation, often as of RAG-1 and RAG-2 we carried out differential isotopic chemical a basis for molecular discrimination but also to establish a se- footprinting of solvent-exposed cysteine thiols, in combination with quence of interactions or to bias the outcome of a reaction. In many instances the allosteric ligand is a specific DNA structure, as Significance in Cre-mediated recombination, in which the Holliday junction intermediate effects allosteric conformational changes that switch active and inactive Cre monomers (1). In other instances the al- Accessibility of antigen receptor gene segments to recombination- losteric ligand is a DNA-bound protein array, as in λ integration, activating gene (RAG), the V(D)J recombinase, is correlated with which is driven to completion by a flanking DNA-protein array that marks of active chromatin. One mark, histone H3 at lysine 4 biases the conformation of λ-integrase (2). (H3K4me3), binds to a plant homeodomain (PHD) in the RAG-2 V(D)J recombination, the process by which antigen receptor subunit; mutations that abolish binding of the PHD to H3K4me3 genes are assembled, is also subject to allosteric control, but in this also impair V(D)J recombination. Engagement of H3K4me3 by case the allosteric ligand is a specific chromatin mark rather than a RAG-2 enhances substrate binding and catalysis, which are func- DNA structure. V(D)J recombination is initiated by recombination- tions of RAG-1. We show that H3K4me3 acts through the PHD to activating gene (RAG)-1 and RAG-2, which together cleave DNA induce conformational changes in an autoinhibitory domain re- at recombination signal sequences (RSSs) flanking the participating siding within RAG-2 as well as in substrate-binding and catalytic gene segments (3). There are two classes of RSS, termed 12-RSS regions of RAG-1. Our data suggest that H3K4me3 promotes and 23-RSS, in which heptamer and nonamer elements are sepa- displacement of the autoinhibitory domain as well as concomitant rated by spacers of 12 bp or 23 bp; physiological DNA cleavage conformational alterations in RAG-1 that favor increased substrate requires the pairing of a 12-RSS with a 23-RSS (3). V(D)J re- affinity and catalytic rate. combination acts in an ordered, locus-specific fashion during lymphoid development. The accessibility of gene segments to V(D)J Author contributions: J.B., A.P., and S.D. designed research; J.B., C.H.N., and S.D. per- recombination is positively correlated with transcription at the formed research; J.B. and A.P. contributed new reagents/analytic tools; J.B., C.H.N., and unrearranged locus and with histone modifications characteristic of S.D. analyzed data; and J.B., C.H.N., and S.D. wrote the paper. active chromatin, including hypermethylation of histone H3 at ly- The authors declare no conflict of interest. sine 4 (H3K4me3) (4–10). This article is a PNAS Direct Submission. RAG-1 and RAG-2 are 1,040 and 527 aa residues long, re- Freely available online through the PNAS open access option. spectively. The catalytic core and DNA-binding functions are largely 1To whom correspondence should be addressed. Email: [email protected]. contained within RAG-1 (11). RAG-2, which is also essential for This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. DNA cleavage activity, comprises part of a putative DNA-binding 1073/pnas.1615727114/-/DCSupplemental.

1904–1909 | PNAS | February 21, 2017 | vol. 114 | no. 8 www.pnas.org/cgi/doi/10.1073/pnas.1615727114 Downloaded by guest on October 1, 2021 mass spectrometry (18). Strikingly, binding of H3K4me3 to RAG-2 RAG-2 (14, 15). In addition to full-length, wild-type RAG-2, we is accompanied by robust increases in the solvent accessibility of used RAG-2(W453A), in which the PHD finger cannot bind RAG-1 within the dimerization and DNA-binding domain (DDBD) H3K4me3 (7). These RAG-2 variants, tagged at the amino termi- and in the ZnH2 domain, which acts as a scaffold for the catalytic nus with MBP, were coexpressed individually with maltose binding center (11). These H3K4me3-induced changes in solvent accessi- protein (MBP)-tagged cR1ct and RAG tetramers were purified by bility are abolished by mutation of the RAG-2 PHD finger. Our a protocol that removes endogenous H3K4me3 (15) (Fig. S1A). observations provide direct evidence that engagement of H3K4me3 TheactivefractionsoftheseRAGpreparations, as determined by by the RAG-2 PHD finger is associated with discrete changes in the burst kinetics (Fig. S1 B and C)rangedfrom5.3to5.4%(Fig. S1D), conformational distribution of RAG-1, consistent with the stimula- consistent with earlier observations (15). tory effects of H3K4me3 on RSS binding and cleavage by RAG. Equivalent amounts of each active RAG tetramer were assayed in vitro for coupled cleavage of a radiolabeled 12-RSS in the Results and Discussion presence of unlabeled 23-RSS substrate and increasing amounts of Information indicating engagement of H3K4me3 by the RAG-2 a histone H3-derived peptide containing trimethylated lysine 4 PHD finger must be communicated within the RAG tetramer in (H3K4me3) or unmethylated lysine 4 (H3K4me0). As previously such a way that substrate affinity and catalytic rate are increased. reported (15), the activity of wild-type RAG was stimulated in a An allosteric ligand can be thought to act on an ensemble of protein dose-dependent fashion by H3K4me3, but not by H3K4me0, native states by inducing a shift in the energetic distribution of those whereas the RAG-2(W453A) mutation abolished responsiveness states and their associated protein conformations (17). In accor- to H3K4me3 (Fig. S2 A and B). We proceeded to use these dance with this view, binding of H3K4me3 to the RAG-2 PHD preparations in chemical footprinting and partial proteolysis assays. finger may be accompanied by a shift in the distribution of RAG native states, so as to favor conformations associated with increased Mapping Solvent Accessibility of RAG Cysteine Thiols by Differential DNA binding and catalysis. We sought to obtain physical evidence Isotopic Chemical Footprinting. The chemical footprinting procedure for the induction of such conformational changes by H3K4me3. is outlined in Fig. 1A. Two protein conformations are depicted, one Conformational states are expected to differ with respect to the in the absence (Fig. 1A, Top Left) and one in the presence (Fig. 1A, detailed accessibility of amino acid side chains to solvent (19). Sol- Top Right) of bound H3K4me3; these single conformations are vent-accessible cysteine side chains are susceptible to alkylation by meant to represent populations of conformations whose distribu- monobromobimane (mBBr), which reacts with free thiols to produce tions differ. RAG preparations were preincubated in the presence a covalent adduct (18). We used this reaction, in combination with of H3K4me3 or H3K4me0 and then subjected to a 1-min pulse mass spectrometry, to determine the effect of H3K4me3 binding on of alkylation with light mBBr ([1H]mBBr). After alkylation by the distribution of solvent-exposed cysteine thiols in RAG. [1H]mBBr was quenched, RAG was denatured in guanidinium The version of RAG-1 used here, cR1ct, lacks the amino-ter- chloride (GdmCl) and unmodified cysteine thiols were alkylated minal noncore region and is more soluble than the wild-type pro- with hexadeuterated mBBr-d6 ([2H]mBBr). Light alkylation is tein but retains responsiveness to H3K4me3 in complexes with represented in Fig. 1A by red plain R and heavy alkylation by red

A SH C + H3K4me3 H3K4me3 GSGLQPAVC467LAIR z = 2

HS [1H]mBBr 1 pulse SR

GdmCl, RS [2H]mBBr H3K4me0 120 chase

S Fig. 1. Mapping H3K4me3-induced conformational D changes in RAG by pulse alkylation mass spectrometry. H3K4me3 478 S VNTFLSC SQYHK (A) Experimental scheme. Binding of H3K4me3 is pro- RS z = 3 Trypsin, posed to result in a shift in the distribution of RAG LC-MS conformations, accompanied by a change in the acces- H3K4me0 H3K4me3 sibility of some cysteine thiols to solvent. Solvent accessibility is probed by pulse alkylation with [1H]mBBr. Following quenching of the first alkylation reaction and

6 denaturation of RAG in guanidinium chloride (GdmCl), rt rt remaining unmodified cysteine thiols are alkylated with CP,L 2 CP,H H3K4me0 [ H]mBBr. RAG is then fragmented with trypsin and peptides resolved by LC-MS. For each cysteine-containing

peptide, total light or heavy alkylation (Σ CP,L and Σ CP,H, B E respectively) is determined by summing the correspond- H3K4me3 AVC431LTLFLLALR H3K4me3 ALHC796DIGNAAE ing peak areas over two dimensions (red arrows): iso- z = 2 z = 2 topologues and retention time (rt). (B–E) Representative spectra of four mBBr-modified peptides from RAG-1. Percent relative intensity is plotted as a function of mass: charge ratio. Spectra obtained from H3K4me3-treated samples are shown above the midline in red; spectra obtained from H3K4me30-treated samples are reflected 6 6 below the midline in gray. Sequences of the corre- H3K4me0 H3K4me0 sponding peptides and their charges (z) are given at

upper right. The heavy-modified peak is separated from BIOCHEMISTRY the light-modified peak by 6 Da (Δ6).

Bettridge et al. PNAS | February 21, 2017 | vol. 114 | no. 8 | 1905 Downloaded by guest on October 1, 2021 boldface R. The modified RAG subunits were digested with relative accessibility, LC/(LC + HC), of each detectable cysteine trypsin and the products were analyzed by liquid chromatography residue in the presence of H3K4me3 (Fig. 2A, red bars) was com- tandem mass spectrometry (LC-MS/MS). pared with its relative accessibility in the presence of H3K4me0 Idealized mass spectrograms are diagrammed at the bottom of (Fig. 2A, gray bars). Significant increases in accessibility of cysteine Fig. 1A, where the x axes denote m/z,they axes denote chro- to alkylation by mBBr were observed at four positions: 431, 467, matographic retention time, and the z axes denote relative peak 478, and 796. No significant decreases in relative accessibility were intensity. The mass spectra on the left depict the theoretical dis- associated with H3K4me3 treatment. The effect of H3K4me3 on tribution of 13C isotopologues for a tryptic peptide in which a accessibility is depicted alternatively in Fig. 2B as the difference specific cysteine residue, CP, has been alkylated by mBBr-d6. The between the mean relative accessibility measured in the presence of 13 mass spectra on the right indicate the Cisotopologuesofthe H3K4me3 ([LC/(LC + HC)]me3) and that measured in the presence same tryptic peptide, in which CP has been alkylated by light mBBr of H3K4me0 ([LC/(LC + HC)]me0). These results indicate that or mBBr-d6. Such a pattern would be expected if H3K4me3 were several cysteine thiols in RAG-1 exhibit increased susceptibility to to alter the conformational distribution of RAG so as to increase alkylation by mBBr in the presence of H3K4me3. the accessibility of residue CP to solvent. H3K4me3 exerts its allosteric effects on DNA binding and Tryptic peptides containing cysteine residues alkylated by light catalysis through an interaction with the PHD finger of RAG-2. mBBr were identified by a mass shift of 190.20 Da relative to the To determine whether the effects of H3K4me3 on alkylation of unmodified form; peptides modified by mBBr-d6 exhibited the RAG-1 are similarly mediated we performed differential foot- expected mass shift of 6.05 relative to their light modified coun- printing with RAG complexes containing the RAG-2 W453A terparts. Differences in the patterns of mBBr alkylation obtained mutation, which disrupts binding of the PHD finger to H3K4me3. in the presence of H3K4me3 or H3K4me0 were observed for The W453A mutation abolished the ability of H3K4me3 to enhance several cysteine-containing peptides from RAG-1 (Fig. 1 B–E). alkylation by mBBr at each of the positions where enhanced alkyl- For example, a tryptic peptide spanning residue C431 exhibited ation was observed in the wild-type complex (Fig. 2 C and D). Taken robust alkylation by light mBBr in the presence of H3K4me3 (Fig. together these observations indicate that engagement of H3K4me3 1B, Upper, red) but was nearly refractory to alkylation by light by the PHD finger of RAG-2 is communicated to the RAG-1 mBBr in the presence of H3K4me0 (Fig. 1B, Lower,gray).The subunit as an alteration in the distribution of RAG-1 conformations. extent of alkylation by heavy mBBr following denaturation of The active fraction of the RAG preparations used above was only RAG was similar in the presence of H3K4me0 or H3K4me3 (Fig. about 5%. This was likely a result of the sonication used to remove 1B). Mass spectra corresponding to peptides spanning C467, H3K4me3. By substituting a mild DNase treatment for sonication, C478, and C796 of RAG-1 also exhibited differences between the we were able to obtain RAG free from detectable H3K4me3 and H3K4me3 and H3K4me0 spectra with respect to alkylation by residual DNase activity (Fig. S4 A and B). The activity of this light mBBr (Fig. 1 C–E). preparation was about 24% (Fig. S4 C and D). The DNase-treated Differential footprinting with mBBr and mBBr-d6 provided a ro- preparation exhibited significant H3K4me3-induced increases in bust means to quantify the extent of pulse alkylation in samples that mBBr accessibility at three of the four sites whose accessibility was had been treated with H3K4me3 or H3K4me0. This was achieved in three steps. First, spectra representing all detectable peptides spanning a particular cysteine residue CP were binned. Second, the peak intensities for all tryptic peptides containing CP were summed over ABRAG-1-mBBr, RAG-2 wt RAG-1-mBBr, RAG-2 wt 13 retention time and over the C isotopologues corresponding to 0.30 heavy or light alkylation (ΣCP.H and ΣCP.L, respectively; Fig. 1A, 0.25 Bottom). Third, the relative accessibility of each modified cysteine 0.20 0.15 residue to light mBBr was defined as LC/(LC + HC), where L is ΣCP.L Σ 0.10 and H is CP.H. Three independent pulse alkylation-mass spec- 0.05 trometry assays were carried out for wild-type RAG and the mutant 0.00 RAG preparation. -0.05 431 467 478 599 656 727 796 867 907 The relative accessibility of RAG-1 cysteine residues to pulse Cysteine residue alkylation in a native protein preparation incubated with the H3K4me0 control showed good correlation with the computed CDRAG-1-mBBr, RAG-2(W453A) RAG-1-mBBr, RAG-2(W453A) solvent-accessible area of cysteine residues in two available RAG 0.30 structures (11, 20) (Fig. S3 and Table S1). The accessibility of RAG- 0.25 1 C431 to pulse alkylation was more closely correlated with its sol- 0.20 0.15 vent-accessible area in the electron microscopic structure (20) than 0.10 in the crystallographic structure (11), suggesting that the former 0.05 structure may more closely approximate the predominant solution 0.00 – -0.05 conformation of the NBD in the absence of H3K4me3 (Fig. S3 A D). 431 467 478 599 656 727 796 867 907 Alkylation of RAG-2 cysteine residues could only be compared with Cysteine residue the solvent-accessible areas computed from the crystallographic structure, because the electron microscopic structure used zebrafish Fig. 2. Binding of H3K4me3 to the PHD of RAG-2 induces localized con- RAG-2, which is highly divergent from murine RAG-2. The only formational changes in RAG-1. (A) Relative accessibility [L/(L + H)] of RAG-1 large discrepancy we observed between accessibility to pulse alkyl- cysteine thiols to pulse alkylation by light mBBr in the presence of wild-type ation and computed effective surface area was at C350 of RAG-2, RAG-2 and H3K4me3 (red) or H3K4me0 (gray). Residues are numbered below. which was inaccessible to mBBr in the native state but exposed in Values indicate mean and SEM of three trials. Statistical significance was determined by ANOVA. *P < 0.05; **P < 0.01; ***P < 0.001. (B)Differ- the crystal structure (Fig. S3 E and F). This difference is likely + − + explained by the fact that the X-ray structure was determined using ential mean relative accessibilities ([L/(L H)]me3 [L/(L H)]me0) of RAG-1 cys- a truncated form of RAG-2 in which the carboxyl-terminal residues teine thiols in wild-type RAG complexes pulse alkylated in the presence of H3K4me3 or H3K4me0. Residues are numbered below. (C) Relative accessi- beyond C350 were disordered, whereas our pulse alkylation assays bility [L/(L + H)] of RAG-1 cysteine thiols to pulse alkylation by light mBBr in the were performed with full-length RAG-2. presence of RAG-2(W453A) and H3K4me3 (red) or H3K4me0 (gray). Residues are numbered below. Values indicate mean and SEM of three trials. Statistical sig- H3K4me3 Binding to the RAG-2 PHD Finger Is Associated with Solvent nificance was determined byANOVAandindicatedasinA.(D) Differential mean Accessibility Changes in RAG-1. For RAG-1 in complex with wild- relative accessibilities ([L/(L + H)]me3 − [L/(L + H)]me0) of RAG-1 cysteine thiols type RAG-2, peptides spanning 10 of the 15 cysteine residues in in complexes containing RAG-2(W453A), pulse alkylated in the presence of cR1ct were consistently detectable. For each of these peptides, the H3K4me3 or H3K4me0.

1906 | www.pnas.org/cgi/doi/10.1073/pnas.1615727114 Bettridge et al. Downloaded by guest on October 1, 2021 altered in the sonicated preparation: C467, C478, and C796; in within the PHD finger. The apparent increase in solvent accessibility distinction, the DNase-treated preparation showed no significant of these residues is consistent with a destabilizing effect of the increase in mBBr accessibility at C431 (Fig. S4 E and F). C431 lies W453A mutation on the PHD finger (Fig. 3 C and D). within the nonamer-binding domain, one of two DNA-binding re- We considered the possibility that the alterations in cysteine al- gions in RAG-1. Enhanced accessibility of C431 to mBBr in the less kylation induced by H3K4me3 might result from changes in the as- active preparation is not likely to result from denaturation of the sociation of RAG-1 with RAG-2, rather than alterations in the nonamer-binding domain by sonication, because the basal accessi- conformation of tetrameric RAG. Three lines of evidence argue bility of C431 to mBBr in the sonicated preparation is similar to that against this interpretation. First, no H3K4me3-induced changes in seen in the nuclease-treated preparation. It remains possible that the cysteine accessibility were observed in RAG-2 at its interface with conformations available to the sonicated preparation differ in detail RAG-1 as structurally defined (11, 20). Second, the active fraction of from those available to the nuclease-treated preparation. Alterna- wild-type RAG was similar in the presence of H3K4me0 or tively, or in addition, the two preparations may differ with respect H3K4me3 (Fig. S5). Third, electrophoretic mobility shift assays typ- to the presence of any copurifying material that might interact ically resolve two discrete RAG–DNA complexes, each containing a with the nonamer-binding domain. In either instance the confor- RAG-1 dimer and one or two RAG-2 subunits (21); H3K4me3 did mational alteration at C431 is less robust than those observed at not alter the relative yields of these complexes (Fig. S6). These ob- C467, C478, and C796. The ability of the two RAG preparations servations suggest that under the conditions of these assays the in- to undergo similar alterations at these three sites suggests that troduction of H3K4me3 is unaccompanied by an alteration in the some portion of inactive RAG retains the ability to undergo amount of RAG-1 associated with RAG-2. conformational change in response to H3K4me3. H3K4me3-Induced Conformational Changes in RAG-1 and RAG-2 H3K4me3-Induced Changes in mBBr Accessibility Not Detected in the Mapped by Limited Proteolysis and LC-MS/MS. We used limited RAG-2 Core Region. Peptides spanning 11 of the 17 cysteine residues proteolysis in combination with LC-MS/MS to confirm that in full-length, wild-type RAG-2 were consistently detected. Of these H3K4me3 binding to the RAG-2 PHD finger is associated with cysteine residues, eight lie within the RAG-2 core region and three conformational changes in RAG. We chose thermolysin for this within the PHD finger; residues from the highly acidic auto- purpose because of its relatively low specificity, which increases the inhibitory region were inconsistently detected. Residues C114, C191, likelihood of identifying cleavage sites whose accessibility was altered and, to a lesser extent, C350 were pulse-alkylated by light mBBr in upon binding of H3K4me3. Use of thermolysin was expected to the absence of H3K4me3; no significant changes in the extent of promote detection of peptides from regions with a paucity of tryptic alkylation at these or other cysteine residues were observed in the cleavage sites, in particular the RAG-2 autoinhibitory domain. presence of H3K4me3 (Fig. 3 A and B). The W453A mutation had RAG complexes containing wild-type RAG-2 or RAG- no effect on the accessibility of RAG-2 core residues to mBBr in the 2(W453A) were incubated with thermolysin for various times in the presence of either H3K4me3 or H3K4me0 but was associated with a presence of H3K4me3 or H3K4me0. H3K4me3 enhanced the large increase in basal alkylation of C446 and C478, which reside production of fragments migrating at about 85 kDa and 60 kDa (Fig. 4A, arrows). In contrast, the proteolytic pattern observed for the mutant RAG complex in the presence of H3K4me3 was similar to the pattern obtained in the presence of H3K4me0 (Fig. 4A), ABRAG-2-mBBr, RAG-2 wt RAG-2-mBBr,G ,G RAG-2 wt indicating that the effect of H3K4me3 on proteolysis was de- 0.30 0.25 pendent on an interaction with the PHD finger. 0.20 The fragments whose production was stimulated by H3K4me3 0.15 were excised, digested with Lys-C, and identified by LC-MS/MS. 0.10 Sites of thermolysin cleavage were determined by comparison with 0.05 reference spectra obtained for RAG digested with Lys-C alone. 0.00 -0.05 The 85-kDa band was produced by thermolysin cleavage of RAG-2 41 67 78 114 178 191 287 350 446 458 478 at two neighboring sites, F384 and F386 (Fig. 4 A and B). The Cysteine residue resulting amino-terminal fragments, including the MBP moiety, have calculated molecular weights of 84.49 and 84.74 kDa, in CDRAG-2-mBBr, RAG-2(W453A) RAG-2-mBBr, RAG-2(W453A) 0.30 good agreement with their electrophoretic mobilities. The 60-kDa 0.25 species was produced by thermolysin cleavage of RAG-1 at I496 0.20 and F497 (Fig. 4 A and B), which is predicted to yield carboxyl- 0.15 terminal fragments of 62.21 and 62.32 kDa, again consistent with 0.10 0.05 the observed mobilities. 0.00 -0.05 Localization of H3K4me3-Induced Conformational Changes to DNA- 41 67 78 114 178 191 287 350 446 478 Cysteine residue Binding, Catalytic, and Autoinhibitory Domains of RAG. RAG-1 contains six discrete structural domains (11, 22). Of the three cysteine Fig. 3. Accessibility of RAG-2 cysteine thiols to mBBr in the presence of residues of RAG-1 that consistently exhibit increased solvent H3K4me3 or H3K4me0. (A) Relative accessibility [L/(L + H)] of wild-type accessibility upon H3K4me3 binding, two, C467 and C478, lie in RAG-2 cysteine thiols to pulse alkylation of RAG tetramers by light mBBr in the DDBD (Fig. 5 A and C). The third residue, C796, resides in the the presence of H3K4me3 (red) or H3K4me0 (gray). Residues are numbered ZnH2 domain (Fig. 5 B and D), which acts as a scaffold for the below. Values indicate mean and SEM of three trials. No significant differ- catalytic region (11). A fourth residue, C431, exhibited inconsistent ences (ANOVA) were observed at the cysteine residues detectable by LC-MS. increases in solvent accessibility and lies in the nonamer binding (B) Differential mean relative accessibilities ([L/(L + H)]me3 − [L/(L + H)]me0) of domain (NBD) (Fig. 5 B and E). The sites in RAG-1 at which RAG-2 cysteine thiols in wild-type RAG complexes pulse alkylated in the pres- C H3K4me3 enhances thermolysin cleavage, I496 and F497, also ence of H3K4me3 or H3K4me0. Residues are numbered below. ( )Relative reside in the DDBD, providing additional evidence for an allosteric accessibility [L/(L + H)] of RAG-2(W453A) cysteine thiols to pulse alkylation of RAG tetramers by light mBBr in the presence of H3K4me3 (red) or H3K4me0 conformational change in this domain (Fig. 4B). In mutant mice expressing core-RAG-2, the RSSs most commonly associated with (gray). Residues are numbered below. Values indicate mean and SEM of three β trials. No significant differences (ANOVA) were observed at the cysteine residues VH and V segments exhibit selective impairment of rearrange- detectable by LC-MS. (D) Differential mean relative accessibilities ([L/(L + H)]me3 − ment (23). The ability of the RAG-2 PHD to modulate the con-

[L/(L + H)]me0) of RAG- 2(W453A) cysteine thiols in RAG complexes pulse alkylated formation of the DDBD may provide a rationale for altered RSS BIOCHEMISTRY in the presence of H3K4me3 or H3K4me0. recognition in core-RAG-2 mice.

Bettridge et al. PNAS | February 21, 2017 | vol. 114 | no. 8 | 1907 Downloaded by guest on October 1, 2021 cRAG-1ct mimics the stimulatory effects of H3K4me3 on RSS binding and A RAG-2 wt RAG-2(W453A) 303030303030H3K4me(n) catalysis of DNA cleavage (15). In this regard it has been suggested 5 5 10 10 15 15 5 5 10 10 15 15 time (24) that the RAG-2 noncore region stabilizes the association of 170 cR1ct 130 RAG-1 with RAG-2 through an interaction with the DDBD of R2 RAG-2 F384, F386 100 RAG-1. A speculative notion, which remains to be tested, is that 70 the acidic autoinhibitory domain enjoys an association with the RAG-1 I496, F497 50 DDBD through charge–charge interactions, impairing RSS 40 binding until inhibition is relieved by H3K4me3. Here we favor the conservative interpretation that H3K4me3-in- 30 T duced changes in solvent accessibility represent changes in RAG 25 conformation induced by occupancy of the PHD finger. A narrower B interpretation is that changes in solvent accessibility reflect exposure 384 391 459 515 588 719 791 962 1008 of residues that are occluded by the PHD finger in the absence of RAG-1 NBD DDBD PreR RNH ZnC2 ZnH2 CTD H3K4me3. Although this interpretation is consistent with the avail-

C431 C478 C796 able data, it seems unlikely that that all of these solvent accessibility C467 I496 F497 changes occur at sites of direct contact with the PHD finger, because 1 352 405 414 487 527 they occur in at least two separate regions of the RAG-1 surface: the RAG-2 K K K K K K ID PHD D DDBD and the ZnH2 domain. F384 F386 Allosteric Regulation by Epigenetic Marks. Other enzymes in addition Fig. 4. H3K4me3 binding to the RAG-2 PHD exposes thermolysin cleavage to RAG are subject to control by chromatin modifications. The sites in a DNA-binding domain of RAG-1 and the autoinhibitory domain of histone methyltransferase polycomb repressive complex (PRC)2 is RAG-2. (A) RAG tetramers containing cRAG-1ct in complex with wild-type activated by binding of H3K27me3 to a WD40 propeller in the RAG-2 or RAG-2(W453A) were incubated with thermolysin in the presence of EED subunit (25), whereas the histone acetyltransferase Tip60 is H3K4me3 or H3K4me0. H3K4me(n), H3K4 methylation state (3 or 0). Time stimulated by an interaction between its chromodomain and of incubation is indicated in minutes. cR1ct and R2 indicate the positions of intact cRAG-1ct and RAG-2, respectively. Red arrows indicate the positions of H3K9me3 (26). The DNA methyltransferase Dnmt3a, like RAG, is cleavage products whose excision from wild-type RAG-2–containing com- activated through a PHD finger, although in the case of Dnmt3a plexes is stimulated by H3K4me3; sites of thermolysin cleavage as de- the allosteric ligand is H3K4me0 (27). In none of these instances is termined by LC-MS are indicated. (B) Sites of H3K4me3-stimulated alkylation it known how information regarding occupancy of the allosteric (green triangles) and proteolysis (red triangles) are superimposed on dia- regulatory site is propagated within the protein. grams of RAG-1 (Upper) and RAG-2 (Lower). CTD, carboxyl-terminal domain; Evidence presented here indicates that a signal initiated by D, degradation domain; ID, autoinhibitory domain; K, Kelch-like domain; binding of H3K4me3 to the RAG-2 PHD is propagated to the PreR, pre-RNaseH domain; RNH, RNaseH domain; ZnC2, ZnC2 domain; and RAG-2 autoinhibitory domain as well as to the DNA-binding and ZnH2, ZnH2 domain. Residues at domain boundaries are numbered above. catalytic regions of RAG-1, but the pathways by which this signal is transmitted are unknown. Specifically, it is unclear whether allosteric signal transmission involves direct contact between the F384 and F386 lie within the acidic autoinhibitory domain of RAG-2 PHD and the catalytic or DNA-binding domains of RAG-2 (Fig. 4B), which is not contained within the available RAG-1. Because available structures of RAG do not include the structures (11, 20, 22). The ability of H3K4me3 to promote pro- PHD or autoinhibitory regions of RAG-2 (11, 20), it is not yet teolysis at these sites indicates that allosteric activation of RAG-2 is known where these domains lie relative to the sites of H3K4me3- accompanied by increased accessibility of the autoinhibitory do- induced conformational change in RAG-1. main to thermolysin. This change in conformation may reflect disruption of an inhibitory interaction that suppresses RAG ac- Implications for Coupling of V(D)J Recombination to Transcription. tivity, because mutation of the RAG-2 autoinhibitory domain Initiation of V(D)J recombination is coupled to the lineage- and

A C E ZnH2

ZnC2 C478

RNH Fig. 5. Localization of H3K4me3-dependent confor- C478 C431 C431 mational changes to sites of DNA binding and catalysis in RAG-1. (A) Locations of C467 and C478 in the C467 CTD DDBD. The structure (11) was depicted from coordi- C467 DDBD PreR nates deposited in the Protein Data Bank (PDB) under C478 ID code 4WWX. Only the RAG-1 chains are shown. NBD C467 Domains are labeled as defined in Fig. 4B.(B)The RAG-1 dimer in A is rotated horizontally ∼180° to B D display the locations of C431 and C796 in the NBD and ZnH2 domain, respectively, of RAG-1. Domains ZnC2 ZnH2 D708 D600 are labeled as defined in Fig. 4B.(C) H3K4me3-induced C796 increases in solvent accessibility within the RAG-1 DDBD. C796 C467 and C478 lie in two helical bundles that comprise a portion of the RSS heptamer binding site. (D)TheRAG-1 RNH catalytic center. The residues that comprise the catalytic motif—D600, D708, and E962—are highlighted in red. PreR CTD DDBD C796 lies within one of the helices of the ZnH2 domain, which bridges E962 to the remainder of the NBD E962 catalytic center. (E) H3K4me3-induced increase in C431 solvent accessibility at C431, which lies at the NBD dimerization interface.

1908 | www.pnas.org/cgi/doi/10.1073/pnas.1615727114 Bettridge et al. Downloaded by guest on October 1, 2021 developmental-stage-specific activation of transcription at specific complex dissociates from active RAG, allowing capture of a antigen receptor loci. This coupling underlies the ordered rear- downstream, cryptic RSS in convergent orientation; these cryptic rangement of heavy- and light-chain genes during lymphocyte de- cleavage sites show a significant overlap with H3K4me3 (31). velopment and provides a means by which Ig and T-cell receptor Although the mechanism underlying such sampling by RAG is expression are restricted to the B- and T-lymphoid lineages, re- unknown, theoretical and experimental considerations (32, 33) spectively. Transcriptional activity at antigen receptor loci is posi- suggest that RAG may assume distinct conformations in sam- tively correlated with the local deposition of H3K4me3 (28), and pling mode and in its various RSS binding modes. If so, this the ability of H3K4me3 to serve as an allosteric activator of RAG suggests that H3K4me3 may modulate the transition from sam- suggests a molecular mechanism by which transcription and re- pling to RSS capture by inducing a conformational shift in RAG combination are coupled. H3K4me3 in itself is insufficient, how- that increases affinity for the RSS. ever, to establish locus specificity, because this epigenetic mark is enriched near all active promoters (29). Indeed, the distribution of Materials and Methods RAG-2 in chromatin is positively correlated with the density of Cell culture, expression constructs, purification of RAG protein, oligonu- H3K4me3 (30). The required specificity could, in principle, be cleotide substrates, assays for coupled DNA cleavage, LC-MS, limited pro- provided by combined recognition of H3K4me3 and one or more teolysis, and accessible surface area calculations are described in SI Materials RSSs, although the participation of an additional ligand distinct and Methods. from H3K4me3 remains possible. Results presented here suggest a sequential process of site specification in which the binding of Pulse Alkylation. Pulse alkylation was based on a published procedure (18). H3K4me3 to the RAG-2 PHD induces conformational changes in The protocol is provided in SI Materials and Methods. the DNA-binding domains of RAG-1 that confer increased affinity for RSSs. The RAG heterotetramer contains two subunits each of MS Data Analysis and Quantification. MS/MS spectra were searched against a RAG-1 and RAG-2. It is not known whether both PHD fingers customized database using the SEQUEST search algorithm (Proteome Discoverer must be occupied by H3K4me3 to effect allosteric activation, nor is 1.4; Thermo Scientific). Cysteine alkylation was quantified using a custom pro- it known whether the stimulatory effects of H3K4me3 on RSS gram. Details of the analysis are given in SI Materials and Methods. binding and catalysis are exerted in cis or in trans,withrespect to RAG-1. Answers to these questions will be critical to an un- ACKNOWLEDGMENTS. We thank Sandra Gabelli of the Johns Hopkins De- derstanding of how synapse formation and cleavage occur in the partment of Biophysics and Biophysical Chemistry for assistance with solvent accessibility calculations and our colleagues in the Johns Hopkins Department context of active chromatin. of Molecular Biology and Genetics for stimulating discussions. This work was Observations of off-target RAG activity have suggested a supported by Grant R01CA160256 (to S.D.), the Johns Hopkins Center for sampling mechanism in which one canonical RSS in a paired Proteomics Discovery, and shared instrumentation Grant S10OD021844.

1. Guo F, Gopaul DN, van Duyne GD (1997) Structure of Cre recombinase com- 18. Chen YT, Collins TR, Guan Z, Chen VB, Hsieh TS (2012) Probing conformational plexed with DNA in a site-specific recombination synapse. Nature 389(6646): changes in human DNA topoisomerase IIα by pulsed alkylation mass spectrometry. 40–46. J Biol Chem 287(30):25660–25668. 2. Biswas T, et al. (2005) A structural basis for allosteric control of DNA recombination by 19. Marsh JA, Teichmann SA (2011) Relative solvent accessible surface area predicts lambda integrase. Nature 435(7045):1059–1066. protein conformational changes upon binding. Structure 19(6):859–867. 3. Schatz DG, Swanson PC (2011) V(D)J recombination: Mechanisms of initiation. Annu 20. Ru H, et al. (2015) Molecular mechanism of V(D)J recombination from synaptic RAG1- Rev Genet 45:167–202. RAG2 complex structures. Cell 163(5):1138–1152. 4. Chakraborty T, et al. (2007) Repeat organization and epigenetic regulation of the DH- 21. Swanson PC (2002) A RAG-1/RAG-2 tetramer supports 12/23-regulated synapsis, Cmu domain of the immunoglobulin heavy-chain gene locus. Mol Cell 27(5):842–850. cleavage, and transposition of V(D)J recombination signals. Mol Cell Biol 22(22): 5. Goldmit M, et al. (2005) Epigenetic ontogeny of the Igk locus during B cell develop- 7790–7801. ment. Nat Immunol 6(2):198–203. 22. Yin FF, et al. (2009) Structure of the RAG1 nonamer binding domain with DNA reveals 6. Jung D, Giallourakis C, Mostoslavsky R, Alt FW (2006) Mechanism and control of V(D)J a dimer that mediates DNA synapsis. Nat Struct Mol Biol 16(5):499–508. recombination at the immunoglobulin heavy chain locus. Annu Rev Immunol 24: 23. Liang HE, et al. (2002) The “dispensable” portion of RAG2 is necessary for efficient 541–570. V-to-DJ rearrangement during B and T cell development. Immunity 17(5):639–651. 7. Liu Y, Subrahmanyam R, Chakraborty T, Sen R, Desiderio S (2007) A plant homeo- 24. Byrum JN, et al. (2015) An interdomain boundary in RAG1 facilitates cooperative domain in RAG-2 that binds Hypermethylated lysine 4 of histone H3 is necessary for binding to RAG2 in formation of the V(D)J recombinase complex. Protein Sci 24(5): efficient antigen-receptor-gene rearrangement. Immunity 27(4):561–571. 861–873. 8. Matthews AG, et al. (2007) RAG2 PHD finger couples histone H3 lysine 4 trimethy- 25. Margueron R, et al. (2009) Role of the polycomb protein EED in the propagation of lation with V(D)J recombination. Nature 450(7172):1106–1110. repressive histone marks. Nature 461(7265):762–767. 9. Morshead KB, Ciccone DN, Taverna SD, Allis CD, Oettinger MA (2003) Antigen re- 26. Sun Y, et al. (2009) Histone H3 methylation links DNA damage detection to activation ceptor loci poised for V(D)J rearrangement are broadly associated with BRG1 and of the tumour suppressor Tip60. Nat Cell Biol 11(11):1376–1382. flanked by peaks of histone H3 dimethylated at lysine 4. Proc Natl Acad Sci USA 27. Li BZ, et al. (2011) Histone tails regulate DNA methylation by allosterically activating 100(20):11577–11582. de novo methyltransferase. Cell Res 21(8):1172–1181. 10. Subrahmanyam R, et al. (2012) Localized epigenetic changes induced by DH re- 28. Perkins EJ, Kee BL, Ramsden DA (2004) Histone 3 lysine 4 methylation during the pre- combination restricts recombinase to DJH junctions. Nat Immunol 13(12):1205–1212. B to immature B-cell transition. Nucleic Acids Res 32(6):1942–1947. 11. Kim MS, Lapkouski M, Yang W, Gellert M (2015) Crystal structure of the V(D)J re- 29. Barski A, et al. (2007) High-resolution profiling of histone methylations in the human combinase RAG1-RAG2. Nature 518(7540):507–511. genome. Cell 129(4):823–837. 12. Ramón-Maiques S, et al. (2007) The plant homeodomain finger of RAG2 recognizes 30. Ji Y, et al. (2010) Promoters, enhancers, and transcription target RAG1 binding during histone H3 methylated at both lysine-4 and arginine-2. Proc Natl Acad Sci USA V(D)J recombination. J Exp Med 207(13):2809–2816. 104(48):18993–18998. 31. Hu J, et al. (2015) Chromosomal loop domains direct the recombination of antigen 13. Callebaut I, Mornon JP (1998) The V(D)J recombination activating protein RAG2 receptor genes. Cell 163(4):947–959. consists of a six-bladed propeller and a PHD fingerlike domain, as revealed by se- 32. Marcovitz A, Levy Y (2013) Weak frustration regulates sliding and binding kinetics on quence analysis. Cell Mol Life Sci 54(8):880–891. rugged protein-DNA landscapes. J Phys Chem B 117(42):13005–13014. 14. Grundy GJ, Yang W, Gellert M (2010) Autoinhibition of DNA cleavage mediated by 33. Melero R, et al. (2011) Electron microscopy studies on the quaternary structure of p53 RAG1 and RAG2 is overcome by an epigenetic signal in V(D)J recombination. Proc Natl reveal different binding modes for p53 tetramers in complex with DNA. Proc Natl Acad Sci USA 107(52):22487–22492. Acad Sci USA 108(2):557–562. 15. Lu C, Ward A, Bettridge J, Liu Y, Desiderio S (2015) An autoregulatory mechanism 34. Bergeron S, Anderson DK, Swanson PC (2006) RAG and HMGB1 proteins: Purification imposes allosteric control on the V(D)J recombinase by histone H3 methylation. Cell and biochemical analysis of recombination signal complexes. Methods Enzymol 408: Reports 10(1):29–38. 511–528. 16. Shimazaki N, Tsai AG, Lieber MR (2009) H3K4me3 stimulates the V(D)J RAG complex 35. Shevchenko A, Tomas H, Havlis J, Olsen JV, Mann M (2006) In-gel digestion for for both nicking and hairpinning in trans in addition to tethering in cis: Implications mass spectrometric characterization of proteins and proteomes. Nat Protoc 1(6): for translocations. Mol Cell 34(5):535–544. 2856–2860. 17. Motlagh HN, Wrabl JO, Li J, Hilser VJ (2014) The ensemble nature of allostery. Nature 36. Niedermeyer TH, Strohalm M (2012) mMass as a software tool for the annotation of 508(7496):331–339. cyclic peptide tandem mass spectra. PLoS One 7(9):e44913. BIOCHEMISTRY

Bettridge et al. PNAS | February 21, 2017 | vol. 114 | no. 8 | 1909 Downloaded by guest on October 1, 2021