Correlation between recombinase activating gene 1 ubiquitin ligase activity and V(D)J recombination

Abstract

The really interesting new gene (RING) finger ubiquitin ligase domain of the recombinase activating gene 1 (RAG1) V(D)J recombinase protein adopts a standard cross-brace architecture but co-ordinates three zinc ions as opposed to the canonical two. We demonstrated previously that disruption of the conserved zinc co-ordination sites resulted in loss of structural integrity and ubiquitin ligase (E3) activity and interfered with the ability of full-length RAG1 to support recombination. Here we present evidence that amino acids surrounding the third, non-canonical site also contribute to functional interaction with the ubiquitin conjugating (E2) enzyme CDC34, while certain residues on the RING domain’s surface important for interaction between other E2–E3 pairs are less critical to the functional RAG1–CDC34 interaction in this assay. Partial reduction of ubiquitin ligase activity was significantly correlated with reduction in the ability of RAG1 to support recombination of extra-chromosomal substrates (r = 0·805, P = 0·009). While poly-ubiquitin chains could be generated, RAG1 did not promote rapid chain extension following mono-ubiquitylation of substrate, regardless of the E2 enzyme used. No single ubiquitin lysine mutant disrupted the ability of CDC34 to form ubiquitin chains on RAG1, and mass spectrometric analysis of the poly-ubiquitylated products indicated ubiquitin chain linkages through lysines 48 and 11. These data suggest that RAG1 promotes a mono-ubiquitylation reaction that is required for optimal levels of V(D)J recombination.

Introduction

The recombinase activating gene (RAG) products RAG1 and RAG2 form a complex that introduces DNA double-strand breaks during V(D)J recombination, a programmed DNA rearrangement that is crucial to the development of the vertebrate immune system. The minimal ‘core’ regions of RAG1 (amino acids 384–1008 of 1040) and RAG2 (amino acids 1–383 of 527) together are sufficient to promote recombination in intact cells.^1–3 However, in the context of the full-length protein, mutations in regions of RAG1 outside of the core also reduce recombination activity, indicating that they play a vital role in normal immune development.^2–6 More specifically, mutations in the zinc (Zn)-binding really interesting new gene (RING) domain of RAG1 outside of the core have been shown to interfere with recombinase activity.^2,3,5,6 They can also inhibit the proper maturation of B lymphocytes in human patients, leading to the development of Omenn’s syndrome and similar immune disorders.^7,8

Ubiquitylation, the covalent attachment of the highly conserved 76-amino acid ubiquitin protein to target proteins, plays regulatory roles in virtually every biochemical system in eukaryotes.^9–11 Conjugation involves a multi-step cascade whereby the C-terminus of ubiquitin is covalently attached to the epsilon amino groups of targeted lysine residues on specific proteins.¹² This process is initiated by the ubiquitin-activating enzyme (E1), which becomes attached to the ubiquitin via a thiol-ester bond. The ubiquityl moiety is then transferred to the active site cysteine of one of several ubiquitin conjugating (E2) enzymes. The ‘charged’ E2 is brought together with a ubiquitylation target protein with the aid of a ubiquitin ligase (E3), and the ubiquitin is transferred to the target.¹² This process may be reiterated to attach ubiquitin to multiple lysine residues on the target protein. Monoubiquitylation at one or several sites has been shown to alter protein function and/or localization.^13–15 Ubiquitin can also be self-conjugated on internal lysine residues, creating polyubiquitin chains. The most common chain linkage is through residue K48 of ubiquitin, which targets the modified protein for degradation by the 26S proteasome.¹⁶ Ubiquitin conjugation has already been shown to play a regulatory role in V(D)J recombination; RAG2 is ubiquitylated by the skp2-cullin-F-box protein complex and targeted for degradation at the G1/S transition,¹⁷ restricting recombination to the G1 phase of the cell cycle.

While the specificity of the ubiquitylation cascade is dependent on the E3,¹⁸ this ligase also plays other roles. Two major classes of E3s have been identified, so-called homology E6 associated protein C-terminal domain (HECT) domain proteins, which include an active site cysteine, and RING finger proteins, which typically co-ordinate two Zn ions in a cross-brace structure comprising either a C₃HC₄ or a C₃H₂C₂ motif.^19,20 Many RING E3 proteins have been shown to stimulate substrate-independent polyubiquitin chain formation by E2 enzymes. The E3 can promote discharge of the ubiquitin from the charged E2, either to a target or to another ubiquitin.²¹ In those cases that have been examined, initial ubiquitylation of the target is the slowest step, followed by rapid, processive elongation of the ubiquitin chain.²¹ The RING finger domain of RAG1 has been shown to function as a ubiquitin ligase.^22,23 A fragment of RAG1 (amino acids 218–389) spanning the RING, a closely associated Zn finger and an upstream basic region promotes auto-ubiquitylation in a reconstituted system of purified proteins including the E2 CDC34; full-length RAG1 likewise undergoes ubiquitylation in intact cells.²²

The structure of the RAG1 RING is unusual in that it co-ordinates three as opposed to two Zn ions.²⁴ The first and second Zn ions are co-ordinated by canonical binding residues, and the third ion is co-ordinated by three non-canonical residues upstream (H270, C266 and H295) and one canonical Zn-binding residue (C293) (Fig. 1a). In addition, the RING forms a single, integrated domain with a downstream C₂H₂ Zn finger^24,25 (not shown). More commonly, RING finger proteins are stabilized by co-ordination of two Zn ions in a cross-brace motif, as depicted for cbl (Fig. 1b).²⁶ This somewhat unusual structure of RAG1 raises the possibility that it may interact with E2 enzymes in a manner distinct from canonical RING proteins such as cbl. We recently demonstrated that disruption of the conserved Zn co-ordination sites resulted in the loss of structural integrity and E3 activity and interfered with the ability of full-length RAG1 to support recombination.⁶ Here we present evidence that the third, non-canonical site also contributes to functional interaction with the E2 enzyme CDC34 and the ability of RAG1 to support recombination in intact cells, while substitution at other sites on the RING–CDC34 interface, predicted from comparison with cbl, had less of an effect. We also found a strong correlation between reduction in ubiquitin ligase activity and recombination. Remarkably for CDC34, the ubiquitin chains formed on RAG1 were linked through both ubiquitin K48 and K11. CDC34 activity in this assay was distributive, and the isolated RING domain inhibited substrate-independent ubiquitylation. These data suggest that the unusual interface between RAG1 and CDC34 does not support rapid ubiquitin chain extension and allows for the formation of atypical CDC34-dependent ubiquitin chains. While it is possible for the RAG1 RING to participate in catalysis of poly-ubiquitin chains, it may be optimized to promote mono-ubiquitylation.

Figure 1.

Diagram of the recombinase activating gene 1 (RAG1) and cbl zinc-binding domains. (a) Zinc-binding residues of the RAG1 really interesting new gene (RING) domain are shown, including the co-ordination of the third zinc ion by the non-canonical binding residues C266, H270 and H295. The first zinc site is co-ordinated by residues C290, C293, C310 and C313. The second zinc site is co-ordinated by residues C305, H307, C325 and C328. (b) The cbl diagram represents a classic RING domain with the first zinc co-ordinated by C381, C384, C401 and C404. The second zinc is co-ordinated by residues C395, H398, C416 and C419.

Materials and methods

Expression plasmids and mutagenesis

pJH548 encodes full-length murine RAG1.² pJMJ029 encodes murine RAG1 amino acids 218–389 (spanning the RING, the associated Zn finger, and upstream basic regions) and pJMJ030 encodes amino acids 264–389 (the RING and the Zn finger).²² Both constructs are fused in frame with an N-terminal Xpress epitope and hexa-histidine (6H) tag. Mutations resulting in single or triple amino acid substitutions (I292A, S327F, E294A, N265A, H270A and I292A/C317A/S327F) were introduced using the QuickChange method (Stratagene, La Jolla, CA) according to the manufacturer’s instructions and confirmed by sequencing (Retrogen, Inc., San Diego, CA). pJMJ040 encodes human CDC34 fused in frame with an N-terminal 6H tag (6H-CDC34).

Protein expression and purification

Purifications of wild-type and mutant forms of RAG1[218–389], RAG1[264–389], and protein kinase tagged ubiquitin (PK-Ubi) were performed as previously described.^6,22 6H-CDC34 was purified by nickel affinity chromatography. All purification steps were carried out on ice or at 4°. BL-21 AI Escherichia coli cells (Invitrogen Corporation, Carlsbad, CA) transformed with pJMJ040 were grown at 37° to an optical density (OD) of 0·5 in Luria-Bertani (LB) supplemented with 100 μg/ml ampicillin; expression was then induced by the addition of 0·2%l-arabinose and allowed to continue for 3 hr. Cells were collected by centrifugation, resuspended in buffer A [20 mm Tris, pH 8·0, 0·5 m NaCl, 10 mm beta mercapto-ethanol (BME), 10% glycerol and 0·01% NP-40] supplemented with 1 mg/ml lysozyme, and incubated for 1 hr on ice. The lysate was cleared by high-speed centrifugation and applied to a Ni-NTA column (Qiagen Inc., Valencia, CA) equilibrated in buffer A. The column was developed with a 10-volume gradient from 0 to 100% buffer B (20 mm Tris, pH 7·4, 0·5 m NaCl, 10 mm BME, 10% glycerol, 0·01% NP-40 and 0·3 m imidazole). Fractions containing 6H-CDC34 were pooled (Frac II) and applied to a 25-ml Superdex 200 (Amersham Biosciences, Piscataway, NJ) column equilibrated in buffer C [20 mm Tris, pH 7·4, 0·5 m NaCl, 10% glycerol and 1 mm dithiothreitol (DTT)] at 0·5 ml/min. Fractions were eluted in buffer C at 0·5 ml/min over 1 column volume. Fractions containing 6H-CDC34 were pooled (Frac III) and dialysed overnight against storage buffer (50mm Tris, pH 8·0, 50 mm NaCl, 10% glycerol and 1 mm DTT). 6H-CDC34 Frac III was flash-frozen in small aliquots and stored at −80°.

Leporine E1, methylated ubiquitin (CH₃-Ubi) and ubiquitin lysine substitution mutants were purchased from Boston Biochem (Boston, MA).

Partial proteolysis

For partial proteolysis, 20 ng of trypsin (Sigma-Aldrich Inc, St Louis, MO) was added directly to 4–5 μg of RAG1[218–289]. Reactions were incubated for 2 hr at 37° and terminated by the addition of 2 mm phenylmethylsulphonyl fluoride (PMSF). Proteolysis products were separated on 18% Tris–glycine gels (Invitrogen Corporation, Carlsbad, CA) and analysed by Western blot using an anti-RAG1 peptide antibody as described previously.⁶

Mass spectrometry analysis

The purified RAG1 sample was lyophilized, resuspended in LDS sample buffer (Invitrogen), and loaded on an Invitrogen 4–12% pre-casted Bis–Tris NuPage gel with 3-(N-Morpholino)propanesulphonic acid, 4-Morpholinepropanesulphonic acid (MOPS) running buffer. The protein components separated by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS-PAGE) were stained with SimpleBlue®. Fifteen gel bands equal in size were cut continuously from the top to the bottom for the entire lane. In-gel tryptic digestion was performed to extract the peptides from these gel bands.²⁷ Each peptide sample was desalted using C18 ZipTip (Millipore, Bedford, MA) and then lyophilized and resuspended in 16 μl of 0·1% trifluoroacetic acid (TFA) for liquid chromatography–mass spectrometry (LC-MS) analysis.

Each sample (6 μl) was loaded on an Agilent 1100 nano-capillary HPLC system (Agilent Technologies, Palo Alto, CA) with 10-cm integrated μRPLC (reverse phase liquid chromatography)-electrospray ionization (ESI) emitter columns (made in-house), coupled online with a linear ion-trap (IT) mass spectrometer (LTQ XP; ThermoElectron, San Jose, CA) for mRPLC-MS/MS analysis. The integrated μRPLC-ESI emitter columns were made of 75-μm internal diameter (i.d.) fused-silica capillaries (Polymicro Technologies, Phoenix, AZ), which were slurry-packed with 5-μm, 300-Å pore size C-18 silica-bonded stationary reverse phase (RP) particles (Jupiter, Torrance, CA) using a slurry packing pump (Model 1666; Alltech Associates, Deerfield, IL). After sample injection, a 20-min wash with 98% of mobile phase A (0·1% formic acid) was applied and peptides were eluted using a linear gradient of 2% mobile phase B (acetonitrile with 0·1% formiac acid) to 42% mobile phase B within 40 min at a constant flow rate of 200 nl/min. The seven most intense molecular ions in the MS scan were sequentially selected for subsequent collision-induced dissociation (CID) using a normalized collision energy of 35%. The mass spectra were acquired in the mass range of m/z 350–1800. The ion source capillary voltage and temperature were set at 1·5 kV and 200°, respectively. The MS/MS data were searched against a database with only RAG1 and ubiquitin protein sequences using BioWorks interfaced SEQUEST (ThermoElectron) operating on a 10-node Beowulf parallel virtual machine computer cluster (Dell, Inc., Round Rock, TX). Ubiquitination (on lysine, 114·00) and methionine oxidation (15·99) were searched as differential modifications. Only tryptic peptides with up to two missed cleavage sites meeting a specific SEQUEST scoring criterion (delta correlation (ΔC_n) ≥ 0·10 and charge state-dependent cross-correlation (X_corr) ≥ 2·0 for [M + H]¹⁺, ≥ 2·5 for [M + 2H]²⁺ and ≥ 3·0 for [M + 3H]³⁺) were considered as legitimate identifications.

In vitro ubiquitylation assays

Auto-ubiquitylation assays were performed as previously described.²² Briefly, purified wild-type or mutant RAG1 proteins (6 μm as dimer) were combined with E1 enzyme (45 nm), CDC34 (300 nm), PK-Ubi (500 μm) and reaction buffer (50 mm Tris, pH 7·4, 2 mm Mg-ATP, 0·001% Brij and 50 mm NaCl) at 37° for 16 hr unless otherwise indicated. Products were separated on 12% denaturing polyacrylamide gels, blotted and probed with anti-Xpress antibody (Invitrogen Corporation, Carlsbad, CA) (1 : 1000) followed by anti-mouse-horseradish peroxidase (HRP) antibody (Pierce Biotechnology, Rockford, IL) (1 : 5000). In some cases blots were probed with an anti-ubiquitin conjugate antibody (1 : 1000) (BioMol International, Inc., Plymouth Meeting, PA) followed by anti-rabbit-HRP antibody (Pierce Biotechnology, Rockford, IL) (1 : 5000).

Karyopherin alpha 1 (KPNA1) ubiquitylation was performed as described previously.²⁸ Briefly, purified maltose-binding protein (MBP)-tagged KPNA1 (1 μm as monomer), E1 (45 nm), the E2 indicated (0·3 μm), PK-Ubi (500 μm), and RAG1 (4 μm as dimer) were incubated in reaction buffer (50mm Tris, pH 7·4, 0·001% Brij, 2mm Mg-ATP, 50nm NaCl, 0·4mm DTT and 0·02mm ZnCl₂) for 16 hr at room temperature. Products were separated on 4–12% NuPAGE gels (Invitrogen, Carlsbad, CA), transferred to nitrocellulose and subjected to western blot with anti-MBP antibody (1 : 10 000), followed by anti-mouse-HRP conjugate (New England Biolabs, Ipswich, MA) (1 : 10 000).

CDC34 charging assay

CDC34 (1·5 μm) was incubated in the presence of E1 enzyme (7·2 nm) and PK-Ubi (100 μm) in charging buffer (55mm Tris, pH 7·4, 5 mm MgCl₂ and 3 mm ATP) at room temperature for up to 30 min. RAG1[264–389] (6 μm) or RAG1 storage buffer was added at time zero as indicated. Reactions were stopped with 10 mmN-ethylmaleimide and 50 mm ethylenediaminetetraacetic acid (EDTA). The degree of CDC34 charging was assessed by western blot.

Recombination assays

Signal joint (SJ) and coding joint (CJ) assays were performed as described previously.⁶ Briefly, 1 μg of RAG1 expression plasmid (pJH548) wild type, H270A, E294A, N265A, I292A, S327F or I292A/C317A/S327F and 10 μg of recombination substrate pJH200 or pJH299²⁹ were transfected into RAG1^−/−/λ5^−/− cells.³⁰ After 48 hr, total DNA was harvested and recombination was assessed by polymerase chain reaction (PCR).⁶ PCR products were separated on a 1·8% agarose gel, stained with ethidium bromide and visualized on a Kodak Molecular Image station. Values given represent the average and standard deviation of at least three independent trials. In all cases, total replicated plasmid was also assessed by semiquantitative PCR as described previously.⁶

Quantification and analysis

Agarose gels and western blots were digitally captured using a Kodak Image Station 2000MM for 30 seconds with zero binning and quantified using kodak 1D software, version 3.6.5.k2 (Kodak, Rochester, NY). In all cases, the average and standard deviation were calculated from at least three independent trials. Statistical analysis was performed using free calculation tools available at http://www.physics.csbsju.edu/stats/.

Results

Analysis of residues that contribute to functional RAG1–CDC34 interaction

The interface between CDC34 and RAG1 was predicted based on in silico alignment of the RAG1 RING and cbl-UbcH7 crystal structures obtained using the Protein Data Bank (PDB) co-ordinates available from the National Center for Biotechnology Information (NCBI) structural database.^24,26 Initial manual alignments allowed us to designate 32 residue pairs in equivalent positions in the two structures, and these were aligned using the least square fit protocol of the align program (Shareware).³¹ After computer-adjusted superimposition, the 32 residue pairs could be aligned with a root mean square deviation of only 1.13 angstroms (Fig. 2). This level of alignment indicated that, despite the presence of the third Zn co-ordination site in RAG1, the overall structures of the two proteins were quite similar. Based on this alignment, a variety of substitutions at positions in the predicted interface known to be important in other E2–E3 pairs were generated (I292A, S327F and I292A/C317A/S327F). cbl residue I383 has been shown to be important for interaction with F63 of UbcH7.²⁶ This corresponds to RAG1 amino acid I292. UbcH7 residue A98 interacts with cbl F418, which corresponds to RAG1 S327.²⁶ We also introduced substitutions in surface positions in the region of the non-canonical Zn site (E294A and N265A) and one that disrupted this site (H270A).

Figure 2.

In silico alignment of really interesting new gene (RING) structures. Backbone traces of the recombinase activating gene 1 (RAG1) RING structure with the cbl RING structure determined as a part of the cbl–UbcH7 complex are shown.^24,26 Red, UbcH7; green, cbl; yellow, RAG1; magenta spheres, cbl zinc ions; grey spheres, RAG1 zinc ions. Protein Data Bank (PDB) ID for the cbl–ubcH7 co-structure is 1fbv, PDB ID for RAG1 RING is 1rmd.

The substituted RAG1[218–389] proteins were assayed for their ability to promote auto-ubiquitylation in the presence of CDC34; a representative auto-ubiquitylation assay is shown (Fig. 3a). At the completion of the reaction, the total concentration of ubiquitylated RAG1 exceeded that of CDC34 by about 40-fold, confirming that CDC34 was turning over in this reaction. The percentage of RAG1[218–389] converted to mono-, di-, tri-, and tetra-ubiquitylated products in at least three independent trials was quantified by western blot with an antibody specific for the RAG1 construct. The amino acid substitutions had varying effects on ubiquitylation activity (Fig. 3c). The I292A, S327F and I292A/C317A/S327F substitutions had no significant impact, although we did notice greater prep-to-prep variation among the substitution mutants relative to wild type. There was also a trend towards lower levels of di-, tri- and tetra-ubiquitylation among some of these mutants, but the difference was not significant based on an unpaired Student’s t-test. The substitutions E294A, H270A and N265 produced 3- to 6-fold reductions in activity. These differences were considered significant based on an unpaired Student’s t-test [Fig. 3c; N265A, t = 7·39 (P = 0·001); H270A, t = 7·83 (P = 0·001); E294A, t = 5·12 (P = 0·004)].

Figure 3.

Ubiquitin ligase activity of recombinase activating gene 1 (RAG1) mutants. (a) Sample auto-ubiquitylation assay performed as described in the Materials and methods using RAG1[218–389] wild type (WT) and CDC34 as indicated. Products were analysed by western blot with anti-Xpress antibody followed by anti-mouse-horseradish peroxidase (HRP) antibody. The positions of unmodified (R1), mono- and poly-ubiquitylated RAG1 are indicated. (b) Karyopherin alpha 1 (KPNA1) ubiquitylation assays were performed as described in the Materials and methods using the E2 enzyme UbcH2 or UbcH5a and RAG1[218–389] WT. The percentage of KPNA1 converted to mono- (Ubi₁), di- (Ubi₂), tri- (Ubi₃) and tetra-ubiquitylated (Ubi₄) species was determined by western blot for each sample. The average and standard deviation for three independent trials are shown. (c) Ubiquitin ligase assays were performed as described in (a) using the WT or various mutants of RAG1[218–389]. The percentage of RAG1 converted to mono- (Ubi₁), di- (Ubi₂), tri- (Ubi₃) and tetra-ubiquitylated (Ubi₄) species was determined by western blot for each sample. The average and standard deviation for at least three independent trials are shown; the significance of the decrease in auto-ubiquitylation relative to WT was determined by Student’s t-test (*P < 0·05; **P < 0·01). I/C/S, triple mutant I292A/C317A/S327F. (d) Partial proteolysis was performed with trypsin as described in the Materials and methods. Products were separated by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS-PAGE) on an 18% gel and analysed by western Blot as described previously.⁶ Open arrows, undigested RAG1[218–389]; filled arrows, cleavage product spanning amino acids 254–377. Please note that extraneous lanes have been excised between the I/C/S and S327F samples and between H270A and N265A.

The pattern of ubiquitylation products suggested that ubiquitin chains were not preferentially extended in this reaction. One current model for the creation of poly-ubiquitin chains suggests that the addition of the first ubiquitin to substrate is rate limiting, and that the subsequent extension of the ubiquitin chain is far more rapid;²¹ alternatively, pre-formed chains may be transferred to substrate.³² The poly-ubiquitylation reaction is not truly processive, because the E2 must dissociate after the addition of each ubiquitin in order to be re-charged by E1.³³ Instead, the rate of chain extension is faster, presumably as a result of more favourable kinetic factors. If this is true, the pattern of ubiquitylated products should vary depending on the relative kinetic parameters. To a first approximation, the pattern can be predicted by the formula: where RAG1, RAG1−Ubi₁ and RAG1−Ubi_n are unmodified, mono- and greater-than-mono-ubiquitylated RAG1, respectively, and and are the Michaelis constants for addition of the first ubiquitin and subsequent ubiquitins, respectively (equivalent formulas can be derived for k_cat and velocity). For example, if and 1% of RAG1 has undergone ubiquitylation, the expected ratio of [RAG1−Ubi₁]:[RAG1−Ubi_n] is 99 : 1, but if is 100-fold higher than , the expected ratio is about 1 : 1. The prediction is independent of the overall rate of the reaction, but depends only on the relative rates of the two phases and the percentage of RAG1 (or other target) that has been modified. In our reaction, the mono-ubiquitylated species is predominant even late in the reaction, and the pattern is consistent with that predicted by . This indicates that RAG1 does not promote very rapid ubiquitin chain elongation by CDC34 following initial, slow mono-ubiquitylation. This prediction also held when samples were taken at earlier time-points when a smaller percentage of RAG1 had been modified (data not shown). We also confirmed that CDC34 was capable of promoting rapid substrate-independent chain elongation (see below), indicating that there was no defect in our CDC34 preparation.

We recently demonstrated that RAG1 specifically promotes ubiquitylation of KPNA1 in a manner dependent on the E2 enzyme UbcH2 or UbcH5a.²⁸ To determine whether the lack of rapid ubiquitin chain extension was unique to auto-ubiquitylation by RAG1 or RAG1–CDC34 interaction, we repeated the analysis using UbcH2 or UbcH5a as the E2 and KPNA1 as the substrate. RAG1 was held in excess of KPNA1, such that formation of the RAG1–KPNA1 complex would not be rate limiting in KPNA1 ubiquitylation. In theory, this condition should favour poly-ubiquitylation. We again observed a pattern of ubiquitylated products consistent with regardless of the E2 used (Fig. 3b).

We recently identified a trypsin-resistant fragment spanning amino acids 234–377 of RAG1 and established that amino acid substitutions such as C325Y that disrupt the folded structure of the RING produce relatively little 234–377 product.⁶ We used the partial trypsinolysis assay to check for gross misfolding in our panel of RAG1 mutants (Fig. 3d). In each case, at least 50% of the total protein, including those RAG1 mutants that had decreased ubiquitin ligase activity, was converted to a trypsin-resistant product (Fig. 3d). Thus, while these amino acid substitutions affect the ubiquitin ligase activity of RAG1, they do so without markedly disrupting the overall fold of the protein.

Analysis of ubiquitin chain linkage promoted by RAG1[218–389]

We confirmed our previous finding that the multiply ubiquitylated products observed in our assay were indeed the result of ubiquitin chains attached to a single RAG1 lysine residue, despite our observation that RAG1 has no preference for extension of ubiquitin chains. Ubiquitin ligase assays were conducted in the presence of methylated ubiquitin, which can be conjugated to a target protein but cannot form ubiquitin chains. As has been reported previously,²² multiply ubiquitylated products were detected when PK-Ubi was present (Fig. 4a, lane 2), while CH₃-Ubi produced primarily mono-ubiquitylated product (about 95% of total) and several much less abundant multiply ubiquitylated species (Fig. 4a, lane 3).

Figure 4.

Activity of various ubiquitin lysine mutants. (a) Auto-ubiquitylation assays were performed as described in the Materials and methods, with protein kinase tagged ubiquitin (PK-Ubi) or methylated ubiquitin [(CH₃)-Ubi] as indicated. Products were analysed by western blot with anti-Xpress antibody followed by anti-mouse-horseradish peroxidase (HRP) antibody. (b) The recombinase activating gene 1 (RAG1)[218–389] ubiquitylation assay was performed with PK-Ubi; products were separated on a 4–12% NuPAGE gel in 3-(N-Morpholino)propanesulphonic acid, 4-Morpholinepropanesulphonic acid (MOPS) buffer (Invitrogen) alongside Mark 12 molecular weight (MW) standards (Invitrogen), stained with SimpleBlue and cut into slices as indicated (J1–J15). (c) Assays were performed as in (a) with RAG1[218–389] and various ubiquitin lysine mutants as indicated. R1₂, RAG1 dimer.

To determine the linkage of these ubiquitin chains, assays were performed in the presence of various ubiquitin lysine substitution mutants (Fig. 4c). In each of these ubiquitin mutants, a different lysine was substituted with arginine. Thus, poly-ubiquitin chains could form through any lysine residue except the one that had been mutated. CDC34 was able to promote formation of di- and tri-ubiquitin chains on RAG1[218–389] with each of the point mutants except for ubiquitin K29R (Fig. 4c, lane 6). The mono-ubiquitylated band in this lane was also light, suggesting that the inactivity of ubiquitin K29R was probably attributable to low specific activity of this particular protein preparation. Consistent with this, mass spectrometric analysis, discussed in detail below, indicated that no ubiquitin K29 chains could be detected in the reaction. These data suggest that no single lysine is absolutely required for the formation of ubiquitin chains in this reaction.

It was previously found that substitution at RAG1 K233 eliminated ubiquitylation of the protein,²² suggesting that this residue was the primary site of modification. Tandem mass spectrometric analysis was used to confirm the site of modification on RAG1 and the linkage of the ubiquitin chains. Poly-ubiquitylated RAG1[218–389] products were separated on a denaturing polyacrylamide gel, and proteins from the various gel slices were subjected to in-gel digestion with trypsin (Fig. 4b). The last three amino acids of ubiquitin are RGG, so trypsinized peptides that span a ubiquitylated residue are tagged with a di-glycyl moiety. Of the 20 lysines in the 218–389 region, 13 were represented within trypsin fragments, indicating good sequence coverage of the protein had been obtained. Five RAG1[218–389] lysine residues were found to be modified by ubiquitylation: K257, K233, K380, K388 and K377 (Table 1). The lysine in the Xpress epitope upstream of RAG1 also underwent modification (data not shown). This measurement is not quantitative, so no conclusion can be drawn from the relative abundance of the peptides. However, an additional observation suggested that K257 and K233 were the primary sites of modification. These products were found in gel slice J11 at relatively low molecular weight (MW) (approximately 36 kD; Fig. 4b), indicating that, when RAG1[218–389] was mono-ubiquitylated, the modification occurred only at K257 or K233. The remaining ubiquitylated RAG1 peptides were not seen until gel slice J8 (approximately 50 kD; Fig. 4b). This result suggests that K257 and K233 were preferentially modified early in the reaction, and the other sites were only modified later, after extended incubation. All of the modified lysines occurred either in the basic region upstream or in the short extension following the RING. None of the four lysine residues within the RING itself was found to be modified. Also, of the nine lysine residues in the basic region, only modification of K233 and K257 was detected, both of which are 100% conserved among RAG1 proteins. Thus, while modification was not confined exclusively to one lysine residue as our earlier mutagenesis data suggested, there was some preference for two conserved lysines, K233 and K257, in the basic region upstream of the RING domain.

Table 1.

Mass spectrometry analysis of ubiquitylated recombinase activating gene 1 (RAG1)[218–389]

	Frequency	Position
RAG1-GG fragments
R.VSSK*EVLK.K	9	K257
R.HQPNVQLSK*K.L	7	K233
K.ESK*ETLVHINK.G	5	K380
K.ETLVHINK*G-	3	K388
KYNHHVSSHK*ESK.E	2	K377
Ubi-GG fragments
R.LIFAGK*QLEDGR.T	41	K48
K.TLTGK*TITLEVEPSDTIENVK.A	4	K11

Ubiquitylated peptides were identified by their di-glycyl moiety. The top panel shows RAG1 trypsinized peptides found to be modified by ubiquitin, and the bottom panel indicates fragments from ubiquitin chain linkage. The amount of the particular fragment detected in this sample and the residue where ubiquitylation occurred (*) are indicated.

Ubi, ubiquitin.

The predominant ubiquitin lysine residue linking the ubiquitin chains generated in this assay was K48, the position most commonly used in CDC34-dependent chains.^21,34 However, K11 ubiquitin–ubiquitin linkages were also observed. As CDC34 is the only E2 enzyme present in this reaction, it must also be responsible for these linkages. Thus RAG1[218–389] was capable of promoting CDC34-dependent ubiquitin chains linked through both K48 and K11. The ability to use both of these residues explains why no single ubiquitin mutant was able to completely disrupt poly-ubiquitylation of RAG1[218–389]. These data indicate that interaction between RAG1 and CDC34 does not constrain CDC34 to K48-dependent ubiquitin chain extension.

RAG1 limits CDC34 activity

It has been shown for other RING finger proteins that the ligase can stimulate both modification of target and rapid extension of substrate-independent ubiquitin chains.^35,36 Both RAG1[218–389], which includes the RING and the site of auto-ubiquitylation, and RAG1[264–389], which includes only the RING, appeared to limit substrate-independent ubiquitylation by CDC34. Ubiquitylation assays were performed essentially as described above, but products were probed with anti-ubiquitin conjugate antibody (note that this produces a stronger signal for poly-ubiquitin species). Substrate-dependent and substrate-independent ubiquitylation were thus assessed simultaneously. Unless otherwise noted, all components were added at time zero and reactions were incubated for 16 hr. In the absence of CDC34, some short, E2-independent ubiquitin chains can be formed during the 16 hr incubation (Fig. 5a, lanes 1 and 6). CDC34 generated high-molecular-weight (HMW) poly-ubiquitin chains in a pattern consistent with rapid, preferential chain extension (Fig. 5a, lane 8), and a large fraction of the free ubiquitin was converted to HMW product. These chains were too numerous and varied to be separated as discrete bands, but their signal distribution peaked at around 150 kD, corresponding to an average chain length of 12–15 PK-Ubi.

Figure 5.

Substrate-dependent and -independent ubiquitylation. (a) Ubiquitin ligase assays (16 hr) were performed as described in the Materials and methods using recombinase activating gene 1 (RAG1)[218–389], CDC34 and RAG1[264–389] as indicated. Products were analysed by western blot with anti-ubiquitin conjugate antibody followed by anti-rabbit-horseradish peroxidase (HRP) antibody. (b) Staging scheme for reactions shown in (a) lanes 2 and 3. (c) CDC34 charging assays were performed as described in the Materials and methods in the presence of RAG1[264–389] or RAG1 storage buffer as indicated. The percentage CDC34 charged was determined by western blot. The average and standard deviation for three independent trials are shown. R1, RAG1.

When RAG1[218–389] was added to the reaction, the pattern of ubiquitylated products was altered (Fig. 5a, lane 4). The primary products were discrete bands of MWs consistent with mono- to penta-ubiquitylated RAG1[218–389] (34·7, 46·0, 57·3, 68·6, 79·9, and 91·2 kD). Keeping in mind that penta-ubiquitylated RAG1 will produce a much stronger signal than the mono-ubiquitylated species, the relative abundance of these products was consistent with what we observed previously. Only a very faint smear of HMW products was evident, indicating that CDC34 did not promote robust substrate-independent ubiquitylation in this reaction. The further addition of RAG1[264–389] to this reaction drastically curtailed ubiquitylation of RAG1[218–389] (Fig. 5a, lane 3), indicating that RAG1[264–389] was able to compete with RAG1[218–389] for binding to CDC34. This was true even though RAG1[264–389] was added 4hr after the start of the reaction, at which point RAG1[218–389] had already undergone mono-ubiquitylation (Fig. 5a, cf. lanes 2 and 3; the staging scheme is shown in Fig. 5b). This again indicates that CDC34 did not bind to and preferentially extend a ubiquitin chain that had already been initiated. Unlike RAG1[218–389], the addition of RAG1[264–389] did not produce any unique products. This is consistent with our published observation that RAG1[264–389] does not undergo ubiquitylation²² and our finding that the primary sites of ubiquitylation are not present in this construct (Table 1). Like RAG1[218–389], RAG1[264–389] limited substrate-independent poly-ubiquitylation by CDC34 (Fig 5, cf. lanes 7 and 8), and both RAG1 constructs appeared to substantially reduce the level of CDC34 auto-ubiquitylation (Fig. 5, cf. lanes 2–5 and 7 with 8).

Part of this effect may be attributable to reduced CDC34 charging. Neither protein forms a complex with uncharged CDC34 sufficiently stable to be detected by gel filtration or co-immunoprecipitation, although a very weak interaction can be detected in a BacterioMatch (Stratagene, Cedar creek, TX) two-hybrid assay (data not shown). We determined that RAG1[264–389] could act as an inhibitor in a CDC34 charging assay in which we measured the rate of formation of CDC34-ubiquitin thioester catalysed by E1 enzyme (Fig. 5b); analysis of RAG1[218–389] in this manner was complicated by the fact that RAG1[218–389] is also a substrate for ubiquitylation, so it is not included here. RAG1[264–389] reduced the initial rate of CDC34 charging by fourfold to fivefold. While this confirms that the RAG1 RING can bind to uncharged CDC34, it is not sufficient to explain the complete lack of detectable substrate-independent ubiquitin chain formation in the presence of RAG1[264–389] after 16 hr of incubation. One model to explain this data is that both RAG1 constructs can bind to ubiquitin-charged CDC34, but dissociation of this complex is contingent upon substrate-dependent discharge of ubiquitin. Overall, these data are consistent with our earlier hypothesis that the RAG1 RING does not promote rapid, preferential extension of poly-ubiquitin chains.

Recombination activity is correlated with ubiquitin ligase activity

We demonstrated previously that substitutions that eliminate ubiquitin ligase activity also diminish the ability of RAG1 to form both coding joints and signal joints.⁶ These data indicated that RAG1 ubiquitin ligase activity was required for robust recombination when RAG1 was present at relatively low levels in the cell. The new panel of substitution mutants was analysed for the ability to support signal and coding joint formation. Recombination was performed under conditions where RAG1 is limiting. Substitution mutations were transferred to the low copy number expression plasmid pJH548, and only 1 μg of expression plasmid was used, 10-fold lower than our standard assay. The average and standard deviation from at least three independent transfections are shown. Control amplifications indicated no differences in the amount of replicated recombination template present in each reaction (data not shown). Each of the mutants that demonstrated a significant reduction in E3 activity also showed a partial impairment in the ability to support signal joint formation (Fig. 6a); these were significant based on unpaired Student’s t-test [N265A, t = 19·8 (P < 0·001); H270A, t = 23·5 (P < 0·001); E294A, t = 7·48 (P = 0·002)]. A significant partial recombination defect was also detected for the triple substitution mutant even though this protein demonstrated no significant effect on ubiquitin ligase activity in our assay [I/C/S, t = 8·42 (P = 0·001)]. This may indicate that our ubiquitin ligase assay was not sensitive enough to detect subtle defects in activity and/or that the RING domain has a separate function in recombination that is partially disrupted by these substitutions. There was a significant correlation between the degree of defect in ubiquitin ligase activity and the decrease in signal joint formation [Fig. 6b; r = 0·805 (P = 0·009)]. We saw a similar trend in coding joint formation (Fig. 6c).

Figure 6.

Recombination activity of recombinase activating gene 1 (RAG1) mutants. Signal joint (SJ) (a) and coding joint (CJ) (c) assays were performed as described in the Materials and methods, with RAG1 wild type (WT), N265A, H270A, I292A, E294A, S327F or I292A/C317A/S327F (I/C/S) as indicated. Recombination was assessed by polymerase chain reaction (PCR); the average and standard deviation were calculated from three independent trials. The significance of recombination impairment relative to WT was assessed by Student’s t-test (*P < 0·05; ***P ≤ 0·001). (b) Correlation between deficits in recombination and E3 activity was assessed by regression analysis. The data points indicated by small arrows are from previously published work;⁶ remaining data points are from this paper.

The consistency of these recombination differences across multiple trials indicates that they are not a result of random fluctuations in transfection efficiency. However, we cannot rule out the possibility that subtle differences in protein expression levels may influence recombination. We attempted to assess expression by western blot. Both the wild type and several mutants that demonstrated a significant reduction in recombination activity were analysed, but even after a long exposure under conditions optimized to reduce background, expression from pJH548 was only marginally detectable above background (data not shown); this was true for the wild-type and mutant proteins tested. We have previously shown that mutations that eliminate E3 activity do not result in significant changes in protein levels when expressed from pJMJ071.⁶ Future work will be required to determine whether this is the case when RAG1 is expressed at very low levels.

Discussion

Ubiquitin conjugation has been shown to be involved in the regulation of many important cellular functions.³⁷ The suitability of this system for a diverse array of regulatory needs is based in part on the large family of ubiquitin ligases that can designate individual targets and modulate the activity of ubiquitin conjugating enzymes. This diversity allows the ubiquitylation process associated with each individual target to be tailored such that it will be designated for degradation in the 26S proteasome, remodelling, further post-translation, etc., as appropriate. By far the most common activity described for ubiquitin ligases that interact with CDC34 is promotion of K48-linked poly-ubiquitin chains,^21,34 leading to proteasomal destruction. Our exploration of the functional interaction between the RAG1 ubiquitin ligase domain and CDC34 suggests an alternate path that does not include rapid ubiquitin chain synthesis or constrain CDC34 to the promotion of ubiquitin K48 linkages.

The RAG1–CDC34 interface required for RAG1 auto-ubiquitylation is atypical

RAG1 is a member of the RING family of proteins, and its RING domain adopts a C₃HC₄ cross-brace structure. Most RING proteins, such as cbl and BRCA1 (breast cancer 1, early onset), form a pocket that creates the binding site for E2–E3 interaction.^26,38 The RAG1 RING domain is unique in that it co-ordinates a third Zn within the cross-brace using residues just upstream of the typical motif.²⁴ While in silico alignment indicated that the co-ordination of the third, non-canonical Zn had only a minimal impact on the overall fold of the domain, several questions were raised which we address herein. (i) Is co-ordination of the third Zn required to maintain the functional interface between RAG1 and CDC34? (ii) Do subtle alterations to the protein surface brought about by the positioning of this third Zn play an important role for interaction with CDC34? (iii) How does this unusual interface affect CDC34 activity? We created and screened a panel of mutants that had amino acid substitutions that either corresponded to positions known to be important for other E2–E3 pairs, or co-ordination of the third Zn ion, or were located on the protein surface in the region of this third ion. The mutations that have been shown to affect functional interaction within other E2–E3 pairs had relatively little effect in our reconstituted RAG1 auto-ubiquitylation assay, although it is clearly possible that they were more defective under intracellular conditions. An exception to this rule is our previous finding the RAG1 P326G mutation is unable to support auto-ubiquitylation;⁶ this site corresponds to an important residue in other E2–E3 pairs.²⁶ Residues that co-ordinate the non-canonical Zn ion did contribute to functional interaction of RAG1 and CDC34, as did those that were positioned on the surface of the protein in this region. Thus the RAG1 CDC34 interface is composed of both standard and unique elements, as would be expected from its distinct structure.

The unusual interface promotes atypical CDC34 behaviour

RAG1 RING domain-promoted CDC34 behaviour was atypical in two ways. First, CDC34-synthesized poly-ubiquitin chains in these assays were not dependent on ubiquitin K48. Mass spectrometry analysis indicated that, while K48 was the most common linkage, K11 was also frequently used. This finding is highly unusual, as it has been reported in the literature that CDC34 is structurally constrained to a ubiquitin K48-dependent pathway.²¹ Secondly, both the pattern of RAG1 auto-ubiquitylation and our competition experiments indicated that chain elongation was not rapid. We did observe rapid production of RAG1-independent/substrate-independent poly-ubiquitin chains by CDC34. This activity was inhibited by RAG1[218–389] and RAG1[264–389]. We hypothesize that the RAG1 RING domain binds to charged CDC34, but prevents substrate-independent discharge. When a suitable substrate such as the RAG1 basic region is present, the ubiquityl moiety is transferred and CDC34 is released. If no target is present, CDC34 remains bound and is unable to participate in substrate-independent chain synthesis. Furthermore, the RAG1–CDC34 interface does not switch to a ‘processive’ mode that rapidly extends K48-linked ubiquitin chains. Instead, the distributive discharge is more haphazard, resulting in transfer sometimes to ubiquitin K48, sometimes to K11, and sometimes to other lysine residues within the target. These observations suggest that the primary role of the RAG1 RING may be in promotion of mono-ubiquitylation, either of itself or of another target.

Physiological significance

The ability of full-length RAG1 mutants to support recombination of extra-chromosomal substrates correlates with their ubiquitin ligase activity. Previously we demonstrated that the E3-deficient substitution P326G supports only marginal recombination activity when RAG1 is expressed at low levels in the cell.⁶ The mutants presented herein with the most severe E3 defects also demonstrated the most dramatic reduction in recombination, although in no cases were they as defective as P326G in either E3 activity or recombination. These data are consistent with our hypothesis that the RAG1 ubiquitin ligase plays a role in promoting or activating recombination, perhaps via auto-ubiquitylation or ubiquitylation of KPNA1.²⁸ We have also found previously that full-length RAG1 undergoes ubiquitylation in the cell.²² The addition of proteasome inhibitors did not stabilize the ubiquitylated RAG1 species, suggesting they might represent multiple mono-ubiquitylations or non-K48-linked chains. Our current finding that RAG1 does not stimulate rapid poly-ubiquitylation is consistent with this. In the future it will be important to isolate RAG1-dependent ubiquitylated proteins from the cell to examine their pattern of modification.

Disclosures

The authors have no financial conflicts of interest.