Structural insights into the U-box, a domain associated with multi-ubiquitination

Abstract

The structure of the U-box in the essential Saccharomyces cerevisiae pre-mRNA splicing factor Prp19p has been determined by NMR. The conserved zinc-binding sites supporting the cross-brace arrangement in RING-finger domains are replaced by hydrogen-bonding networks in the U-box. These hydrogen-bonding networks are necessary for the structural stabilization and activity of the U-box. A conservative Val→Ile point mutation in the Prp19p U-box domain leads to pre-mRNA splicing defects in vivo. NMR analysis of this mutant shows that the substitution disrupts structural integrity of the U-box domain. Furthermore, comparison of the Prp19p U-box domain with known RING–E2 complex structures demonstrates that both U-box and RING-fingers contain a conserved interaction surface. Mutagenesis of residues at this interface, while not perturbing the structure of the U-box, abrogates Prp19p function in vivo. These comparative structural and functional analyses imply that the U-box and its associated ubiquitin ligase activity are critical for Prp19p function in vivo.

Ubiquitin (Ub) targeting of proteins for degradation by the proteosome involves poly-ubiquination of substrate proteins via an enzyme cascade consisting of activating (E1), conjugating (E2) and ligating (E3) enzymes. E3 ubiquitin ligases vary widely in size, composition and enzymology, reflecting their regulatory role in substrate recognition¹. HECT and RING finger domain–containing proteins constitute two classes of E3 ligases. HECT domains bind Ub through a thioester bond and transfer Ub directly to substrate. RING finger E3s facilitate the transfer of Ub from the E2 to the substrate, rather than binding Ub directly.

A third class of E3 ubiquitin ligases has been recently identified^2,3. This class of proteins contains a U-box motif, first identified in Saccharomyces cerevisiae Ufd2p. Ufd2p promotes the elongation of poly-ubiquitin chains in a U-box–dependent manner (recently termed an ‘E4’ activity)⁴. An alignment of U-boxes and RING motifs indicated that U-boxes lack the strictly conserved histidine and cysteine Zn²⁺-chelating residues found in RING fingers, but they share a similar pattern of hydrophobic and polar amino acids (Fig. 1a), raising the possibility that they have similar folds⁵.

Fig. 1.

The U-box and RING finger domains share a conserved fold. a, Alignment of selected U-box and RING finger domains. The shadings are yellow for hydrophobic, green for residues important in c-Cbl interaction with its E2 and blue for Zn²⁺-chelating residues in the RING fingers. Hs represents human proteins; Sc, S. cerevisiae. Blue arrows above the alignment indicate β-strands, and the red cylinder indicates the central α-helix. b, Ribbon diagram of the three-dimensional structure of the Prp19p U-box domain, with core hydrophobic residues in yellow. c, Stereo view of the Cα trace of the 20 lowest energy NMR structures. d, Overlay of the structures of the Prp19p U-box (green) and the c-Cbl RING finger (pink) (PDB entry 1FBV). e, Backbone r.m.s. deviations between the Prp19p U-box and three RING fingers using ProSup³⁴. Panels (b–d) were generated using MolMol³⁵.

S. cerevisiae Prp19p is an essential pre-mRNA splicing factor that contains an N-terminal U-box^6,7. Interestingly, a mutation within the Prp19p U-box that results in the substitution of an isoleucine for a conserved valine (prp19-1) leads to pre-mRNA splicing defects and the disruption of several key protein–protein interactions within the spliceosome^8,9. The profound physiological consequences that result from this mutation underscore the importance of the U-box motif for Prp19p function and premRNA splicing. Here we present the solution structure of the Prp19p U-box. In addition, the structural and functional consequences of mutations to different structural components of the Prp19 U-box were examined.

Comparison of U-box and RING fingers

A Prp19p fragment containing the U-box (residues 1–73) was produced for structure determination by NMR spectroscopy. NMR resonances were assigned using conventional heteronuclear strategies with ¹³C,¹⁵N-enriched protein¹⁰. Only the N-terminal 54 residues of Prp19p(1–73) form a well-folded structural domain. The threedimensional structure of residues 1–56 of the Prp19p U-box (Fig. 1b) was determined from a total of 859 NOE-derived distance constraints, 14 hydrogen bond constraints and 53 backbone dihedral angle constraints. A family of 50 starting structures calculated by a torsion angle dynamics algorithm was refined by restrained molecular dynamics, and the 20 structures with lowest violation energy were selected (Fig. 1c). The U-box domain is well defined, with an r.m.s. deviation versus the mean of 0.27 Å for the backbone atoms and 0.83 Å for all heavy atoms and 98% of all residues in the allowed regions of the Ramachandran plot.

The Prp19p U-box contains a central α-helix (residues Leu28–Gly36), a single turn of helix (Ser46–Glu49), four β-strands and a hydrophobic core involving Phe23, Leu28, Ile40, Ile47 and Ile50 (Fig. 1b). Three of the β-strands form a mini-antiparallel β-sheet composed of Arg12–Leu15, Thr21–Glu24 and Ile50–Ile53. The fourth β-strand (Asn37–Ile40) is packed against a nonregular loop. This organization is similar to the RING-fold that consists of a central α-helix and a series of loops of variable length interspersed with small β-strands^11–16. Although residues 55–70 lack defined structure as determined by the number and range of observable NOEs, they show a substantial degree of line broadening, which reduces the crosspeak signal-to-noise (S/N) ratio in the NMR spectra. In contrast, residues 71–73 have sharp lines characteristic of a flexible unrestrained peptide-like character in solution. These observations suggest that amino acids 55–70 likely exchange among a number of conformations. The Prp19p-binding sites for pre-mRNA splicing factors Snt309p in S. cerevisiae and Cwf7p in Schizosaccharomyces pombe have been mapped to residues 63–73 (ref. 9). The conformational exchange in this region of Prp19p is likely to be quenched upon binding, which would imply a functional role in finetuning interactions with target proteins.

Because hPrp19p has been shown to have E3 ligase activity¹⁷, we compared the Prp19p U-box structure with that of three E3-type RING fingers: c-Cbl¹⁴, RBX1 (ref. 15) and BRCA1 (ref. 16). The remarkable similarity between the Prp19p U-box and RING domains is illustrated using c-Cbl as a representative RING (Fig. 1d,e).

Stabilization by hydrogen-bond networks

The chelation of Zn²⁺ ions is fundamental to the stabilization of RING domains. Although the U-box lacks the ability to bind Zn²⁺, analysis of the three-dimensional structure of the Prp19p U-box reveals that its fold is stabilized by networks of hydrogen bonds and salt bridges (Fig. 2a) in the same locations as the Zn²⁺- binding sites in RING domains (Fig. 2b). Direct experimental evidence supporting this conclusion was obtained from the slowed solvent-exchange kinetics of protons involved in these networks. In particular, extraordinarily slow exchange was observed for the sulfhydryl and hydroxyl groups of Cys3 and Thr41. The ¹H NMR resonances of these two groups are observed as sharp lines at 6.01 and 5.82 p.p.m, which is rare in proteins and constitutes strong evidence of participation in hydrogen bonds. These observations, combined with the numerous NOE connectivities to the residues involved and the resultant hydrogen bonds in the structure, provide direct experimental confirmation of the hydrogen-bonding networks.

Fig. 2.

U-box stability is mediated by hydrogen-bonding networks. a, Structure of the Prp19p U-box with the hydrogen-bonding networks represented by a green sphere with a radius of 2.7 Å. Sphere centers are located at the sulfhydryl of Cys3 and carboxylate of Asp38. b, Structure of the c-CBL RING finger with zinc ions represented by a red sphere with a van der Waals radius of 1.3 Å. c,d, Residues involved in the first and second hydrogen-bonding networks. Shown is an optimized close-up view of the two networks in the Prp19p U-box, with donor-acceptor distances ≤2.7 Å highlighted by green dotted lines. Residues selected for subsequent mutagenesis are displayed in green. e, Sequence alignment of several U-box proteins. The eight positions of the zinc ligands in RING proteins are indicated below. Residues structurally identified as participating in the two hydrogen-bonding networks are shown in red and blue. f, ¹⁵N-¹H HSQC spectra of Prp19p(1–73) and the C3A, E24A, D38A and T41A mutants.

In one network, the Cys3 side chain is positioned between the backbone carbonyl of Lys8 and side chain carboxylate of Glu24, and other residues involved include Ser6, Arg11 and Ser26 (Fig. 2c). Cys3, Ser6, Glu24 and Ser26 correspond to four Zn²⁺-chelating residues in RING domain proteins (Fig. 2e). In the other hydrogen-bonding network, the Thr41 side chain is positioned near the side chain carboxylate of Asp38, and other residues involved include Ser16, Ser19 and Ser46 (Fig. 2d). Residues Ser19, Asp38 and Thr41 correspond to three of the four Zn²⁺- chelating residues in the corresponding site in RING domains (Fig. 2e). The fourth Zn²⁺ ligand in RING domains corresponds to Prp19p Thr21, which is at most only peripherally involved in the hydrogen-bonding network. Instead, other residues such as Ser16 and Ser46 are involved. Interestingly, the carboxylate side chain of Asp38 is at the center of this site and interacts with almost all of the other side chains, which is reminiscent of the role of the Zn²⁺ ion in the RING domain. The other site in the Prp19 U-box does not have such a well-defined focal point. An additional important feature of the hydrogen-bonding/salt-bridge networks is that these interactions should be dynamic in solution. This is reflected in part by the observation that not all conformers in the structural ensemble have identical hydrogen bond donor-acceptor pairs. Nevertheless, all conformers do have both hydrogen-bonding networks. In summary, we find the U-box is stabilized by a more decentralized (and likely dynamic) set of hydrogen-bonding and salt-bridge interactions than that provided by the Zn²⁺ sites in RING domains.

To test the importance of the hydrogen-bonding networks in U-box stability, four key residues were mutated separately to alanine, and the ¹⁵N-¹H HSQC NMR spectrum of the mutant U-boxes was determined and compared with wild type. Cys3 and Thr41 were selected because it was clear from the slow solvent-exchange kinetics that their side chain functional groups were involved in hydrogen bonding. The other two residues mutated, Glu24 and Asp38, were selected because they interact strongly with Cys3 and Thr41. Asp38 was also mutated because it seemed to play a strategic role at the center of the stabilization network. ¹⁵N-¹H HSQC NMR was used to assay the structure of the mutants because wild type Prp19p(1–73) contains well-dispersed resonances resulting from the variable chemical environment of the amides in a well-folded protein. In contrast, the collapse of resonance dispersion in the spectra for the C3A, E24A, D38A and T41A mutant proteins reveals that disruption of a single hydrogen bond in either network results in significant destabilization of the U-box structure (Fig. 2f).

Residues at the putative E2 interface

The crystal structures of c-Cbl and RBX1 RING domains show that each RING contains a shallow hydrophobic groove formed by residues in the α-helix and two loops^13,14. In c-Cbl, this groove interacts with the E2 enzyme UbcH7 (ref. 13). When the Prp19p U-box domain is structurally aligned with these RING domains, a similar hydrophobic surface is seen that involves Ile5, Tyr31, Pro39 and Ile40 (Fig. 3a,b). Mutations in the U-box–containing proteins CHIP, KIAA0860 protein and Ufd2p that correspond to the Y31A and P39A mutations of Prp19p abolish in vitro E3 activity¹⁷. In addition to these hydrophobic residues, the position corresponding to Prp19p Asp34 in c-Cbl and the U-box protein, UIP5, has been identified as essential for E2 interactions^13,18,19.

Fig. 3.

Mutational analysis of putative E2-interacting residues in the U-box. a, Overlay of the three-dimensional structures of the Prp19p U-box (green) and the c-Cbl RING finger (pink) (PDB entry 1FBV). Residues found in the shallow hydrophobic groove are indicated in green for the U-box and in red for c-Cbl. b, Overlay of the three-dimensional structures of the Prp19p U-box (green) and the Rbx1 E3-type RING finger (yellow) (PDB entry 1LDJ). Residues found in the shallow hydrophobic groove are shown in green for the U-box and in red for Rbx1. ¹⁵N-¹H HSQC spectra of c, Prp19p(1–73) and the d, Y31A; e, D34A; and f, P39A mutants.

To address the structural ramifications of mutations in the putative E2-binding region, we constructed Y31A, D34A and P39A mutants of the Prp19p U-box. The ¹⁵N-¹H HSQC NMR spectrum of the mutant U-boxes was measured and compared with wild type (Fig. 3c–f). In the cases of Y31A and D34A, the U-box structure is not greatly perturbed (Fig. 3d,e), implying that these mutations will be useful to directly assess the Prp19p interaction with its presumed E2 ligand. In contrast, Prp19p P39A does not have a defined tertiary structure (Fig. 3f). U-box mutants corresponding to residues Asp34 and Pro39 have been used to study U-box–E2 interactions^17,19. In the case of the proline mutant, our studies show that mutations at this position in the U-box most likely disrupt its structure rather than interfere specifically with E2 interactions.

Prp19p requires the U-box for function

The prp19-1 point mutation encoding a V14I substitution²⁰ was identified in a genetic screen for splicing mutants²¹. Val14 is a part of the well-packed hydrophobic core of the U-box (Fig. 4a). Thus, we suspected that this mutation might cause a structural defect in the protein. Consequently, NMR analysis was carried out on the mutant protein. As with the mutants involving the hydrogen-bonding network, the ¹⁵N-¹H HSQC spectrum of Prp19-1p (1–73) (Fig. 4b) shows limited dispersion and decreased linewidth compared with the wild type protein, indicating loss of well-defined tertiary structure. Detailed analysis of the U-box structure revealed Val14 is so well packed that extension of the side chain by one methylene group at this position, converting Val to Ile, is sufficient for destabilizing the central hydrophobic core. Indeed, the F23G mutation of the central core residue also disrupts U-box structure (data not shown) and Prp19p function (Table 1). These results show that the defects found in prp19-1 are caused by U-box destabilization.

Fig. 4.

The U-box domain is required for Prp19p function. a, Ribbon diagram of the U-box structure. The Val14 side chain is red, and the side chains of Phe23, Leu28, Ile40, Ile47 and Ile50 are yellow. Val14 has a central role in the protein core, forming extensive contacts with Phe23 and Leu28 in the hydrophobic core, as well as with the aliphatic portions of Lys25 and Glu52, which are solvent exposed. b, ¹⁵N-¹H HSQC spectrum of Prp19-1p(1–73). Ubiquitination assays preformed with in vitro transcribed and translated c, Prp19p; d, Prp19-1p; or e, Prp19p(64–504). Controls lacking recombinant E1, E2 (Ubc3) and/or Prp19 proteins are indicated in each panel with a minus sign. The reaction mixtures were resolved under reducing conditions by SDS-PAGE, and the separated proteins were subjected to immunoblot analysis with antibodies to ubiquitin (upper panels) or visualized directly by autoradiography (lower panels).

Table 1.

Consequence of mutations in vivo

	Consequence in vitro	Area of mutation	prp19Δ in vivo rescue
Prp19	Positive control	Wild type	+
VECTOR	Negative control	−	−
Prp19(64–504)	Negative control	U-box delete	−
F23G	Misfolded	Hydrophobic core	−
C3A	Misfolded	Hydrogen bond	−
E24A	Misfolded	Hydrogen bond	−
D38A	Misfolded	Hydrogen bond	−
T41A	Misfolded	Hydrogen bond	−
Y31A	Well folded	Putative E2 interface	−
D34A	Well folded	Putative E2 interface	−
P39A	Misfolded	Putative E2 interface	−

Six mammalian U-box–containing proteins, including hPrp19p, have been tested and found to show E3 ubiquitin ligase activity in vitro^17,22,23. Similar to hPrp19p, S. cerevisiae Prp19p promoted the ligation of ubiquitin to proteins in an ubiquitin ligase reaction (Fig. 4c), whereas no ubiquitination products were observed in the presence of a mock reaction or in the absence of recombinant E1 and E2 enzymes (Fig. 4c). To determine whether the U-box is essential for this activity, we tested Prp19-1p, which contains the V14I substitution, and a U-box deletion mutant, Prp19p(64–504). These mutations abolished activity (Fig. 4d,e).

To determine whether the E3 ligase activity of Prp19p corresponds to its function in vivo, U-box mutations and truncations were introduced into the full-length PRP19 gene. These prp19 mutants, along with wild type PRP19, were then assayed for their ability to complement the null allele of prp19 (PRP19 is an essential gene) by a conventional plasmid shuffle approach. As expected, deletion of the entire U-box of Prp19p disrupted protein function in vivo (Table 1). Also, mutations in the hydrogen-bonding networks that destabilize the protein structure caused loss of function (Table 1). Interestingly, the D34A and Y31A mutants, which are predicted to interrupt the interaction with an E2, were also unable to rescue prp19Δ (Table 1). Because these mutations do not alter U-box structure (Fig. 3d,e) but do alter the interaction surface conserved between RING-finger and U-box proteins, these single amino acid substitutions likely eliminate the ability of the U-box to interact with its ligand, presumably an E2 enzyme.

The U-box and ubiquitination

U-box proteins are far less abundant than RING finger proteins in most eukaryotes, with only two encoded by the S. cerevisiae genome. The U-box proteins CHIP and S. cerevisiae Ufd2p clearly function in vivo by stimulating multi-ubiquitination in conjunction with specific E2 proteins^4,22,24,25. Interestingly, it was recently reported that CHIP, like Ufd2p, is able to perform as an E4 ligase in addition to displaying in vitro E3 ligase activity²⁵. This suggests that E4 activity may be a common feature of U-box proteins. If so, the structural analysis of the U-box provided here will allow the basis of E4 activity to be examined in more detail.

Ubiquitination plays various roles in the cell. Although many poly-ubiquitinated proteins are rapidly degraded by the 26S proteosome, ubiquitination can create a regulatory, rather than a proteolytic, signal¹. Prp19p is essential for pre-mRNA splicing and is present in a complex that contains many proteins^9,26. Thus far, however, its biochemical role in pre-mRNA splicing and that of the proteins with which it interacts are unknown. The combined functional and structural analyses presented here strongly suggest that Prp19p acts as an ubiquitin ligase in vivo, setting the foundation to explore the role of ubiquitination in pre-mRNA splicing.

Methods

Expression and purification of Prp19p(1–73) fragments for NMR analysis

Recombinant Prp19p(1–73) fused to a N-terminal His₆-tag and its indicated mutants were generated by PCR sitedirected mutagenesis and expressed in Escherichia coli BL21 (pLysS) (Novagen). All mutants were confirmed by DNA sequencing. Samples enriched in ¹³C and ¹⁵N were produced by growth in M9 minimal media with ¹³C-glucose and ¹⁵NH₄Cl as the sole carbon and nitrogen sources. Cells were lysed in 20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA and 5 mM dithiothreitol (DTT), and proteins were purified using Ni²⁺-nitrilotriacetic acid (Ni²⁺-NTA) columns (Qiagen) following the manufacturer’s instructions. The His₆-tag was cleaved by incubation with thrombin (25 U per −20 mg protein) for 4 h at room temperature. The Prp19p(1–73) proteins were further purified using a Mono-Q 10/10 column (20 mM sodium phosphate, pH 7.0, elution 10–300 mM NaCl) (Amersham Pharmacia Biotech). NMR spectra were recorded on 1 mM protein samples in 20 mM sodium phosphate, pH 7.0, 20 mM NaCl and 5 mM DTT.

NMR spectroscopy and resonance assignments

NMR experiments were performed on Bruker Avance spectrometers operating at 600 and 800 MHz proton frequencies. Experiments were carried out at 30 °C for Prp19p(1–73) and 25 °C for the mutants. Sequencespecific backbone assignments were made using ¹⁵N,¹³C-labeled Prp19p(1–73) using standard triple-resonance experiments¹⁰. Side chain assignments were obtained from 2D ¹H COSY and TOCSY as well as 3D HCCH-TOCSY spectra.

Proton-proton distance constraints were derived from three NOESY spectra: a 2D homonuclear ¹H NOESY spectrum (τ_m = 100 ms), a 3D ¹⁵N-resolved [¹H, ¹H] NOESY (τ_m = 100 ms) and a 3D ¹³C-resolved [¹H, ¹H] NOESY (τ_m = 100 ms). Data was processed using XWINNMR (Bruker) and analyzed using XEASY²⁷. The solution structure of Prp19p(1–73) was determined using NOE-derived distance constraints and torsion angle constraints derived from chemical shifts using TALOS²⁸. Torsion angle dynamics calculations were performed using DYANA²⁹, followed by restrained molecular dynamics using AMBER³⁰ following established protocols^31–33. A summary of structural statistics is provided in Table 2.

Table 2.

Structural statistics¹ of the Prp19p U-box

Restraints for calculation
Total NOE restraints	859
Intraresidue	225
Sequential	213
Medium Range	160
Long range	261
Hydrogen bonds	14
Dihedral angle constraints	53
Constraint violations (mean and s.d.)
Distance constraints ≤0.2 Å	0.4 ± 0.5
Dihedral angle constraint violations ≤5°	0.9 ± 0.7
Maximum distance constraint violation	0.19 ± 0.03
Maximum dihedral constraint violation	6.3 ± 3.2
AMBER2 energies (kcal mol−1)
Constraint energy	17.8 ± 1.5
Total energy	−2,117 ± 12
Ramachandran statistics (%)3
Most favored	77
Additionally allowed	20
Generously allowed	1
Disallowed	2
R.m.s. deviation from mean structure (Å)
Backbone atoms	0.27
All heavy atoms	0.83

¹Structure statistics refers to an ensemble of 20 structures with the lowest energy from 50 calculated structures.

²See ref. 30.

³Statistics for the mean coordinates after restrained molecular dynamics determined with PROCHECK-NMR³⁶.

Ubiquitination assays

The in vitro ubiquitination assay was performed as described¹⁷ with some modifications. In brief, empty pSK(+), pSK(+)PRP19 (pKG1781), pSK(+)prp19-1 (pKG177) or pSK(+)PRP19(64–504) (pKG550) were translated in vitro in the presence of [³⁵S]-Trans label (ICN Pharmaceuticals) with the use of the T_NT-Coupled Reticulocyte Lysate System (Promega). In vitro translated, ³⁵S-labeled Prp19p, ³⁵S-labeled Prp19p-1, ³⁵S-labeled Prp19p (64–504) or mock reaction (3 µl each) was combined with 0.1 µg His₆-E1, 1 µg His₆-Ubc3, 0.5 units phosphocreatine kinase, 1 µg ubiquitin (Sigma), 25 mM Tris-HCl, pH 7.5, 120 mM NaCl, 2 mM ATP, 1 mM MgCl₂, 0.3 mM DTT and 1 mM creatine phosphate (20 µl total reaction volume). Reactions were incubated for 2 h at room temperature and terminated by the addition of SDS sample buffer. One portion of the sample was resolved on 10% SDS-PAGE gels, transferred to a PVDF membrane, and subjected to immunoblot analysis with a rabbit polyclonal antibody to ubiquitin (Sigma) at a 1:100 dilution. After western anaylsis, blots were exposed to PhosphorImager screens and visualized using ImageQuant (version 3.3) (Molecular Dynamics). Another portion was resolved on 10% SDS-PAGE gels, and the proteins were visualized by fluorography.

Strains and media

A prp19:HIS3 ura3-52 leu2-Δ haploid strain carrying a URA3-selectable vector expressing wild type PRP19 was transformed by a standard lithium acetate method with PRP19 cDNA and mutants under control of the GAL1 promoter in a LEU2-based vector. Ura⁺ Leu⁺ transformants were streaked to plates containing 5-fluorourotic acid to score the ability of prp19 mutations to rescue growth of the prp19Δ strain.

Coordinates

Coordinates for the ensemble of structures have been deposited at the Protein Databank (accession code 1N87). The NMR chemical shifts have been deposited in the BMRB (accession code 5594).