The Prp19 U-box Crystal Structure Suggests a Common Dimeric Architecture for a Class of Oligomeric E3 Ubiquitin Ligases ^†,‡

Protein expression and purification

S. cerevisiae Prp19 constructs were expressed in E. coli as a 6xHIS tag fusion protein from pET15b. Protein was produced and purified using established protocols for Prp19(1-73) (19). Briefly, proteins were expressed using BL21(DE3) cells (Novagen) in Luria-Broth (LB). For seleno-methionine labeled protein, protein was expressed using 834(DE3) cells (Novagen) in M9 minimal media supplemented with seleno-methionine. Proteins were purified using a Ni²⁺-nitrilotriacetic acid column (Qiagen) with an imidazole gradient from 25-500 mM. The 6xHIS-tag was cleaved by incubation with thrombin (1 U/1 mg protein) for 12 h at 4°C for U-box constructs and 36 h at 4°C for Prp19(1-133). Protein was then purified using anion exchange over a MonoQ column (Amersham-Pharmacia) with a gradient of NaCl from 10-500 mM.

Strains and media

S. cerevisiae strains used in this study were grown either in synthetic minimal medium with the appropriate nutritional supplements or yeast extract-peptone-extrose (26). Transformations were performed by the lithium acetate method (27). Haploid strain containing a temperature-sensitive copy of PRP19 (prp19-1) (KGY1811) transformed with pRS415 plasmids containing either wild-type PRP19 or prp19-L15E. The ability of prp19-L15E to rescue growth at the restrictive temperature was scored by spotting serial dilutions and incubating them at permissive (25°C) or restrictive (36°C) temperature for 3 days. prp19::HIS3 ura3-52 leu2-Δ1 haploid strain carrying a URA3-selectable vector expressing wild-type PRP19 transformed with either PRP19 cDNA or prp19-L15E under control of the GAL1 promoter in a LEU2-based vector. Ura⁺ Leu⁺ transformants were serially diluted on plates containing or not 5-flourootic acid and uracil to score the ability of prp19-L15 to rescue growth of the prp19Δ strain after 3 days.

Analytical ultracentrifugation

Sedimentation velocity experiments were conducted with an Optima XLA (Beckman-Coulter, Fullerton, CA), with a 4-hole An60Ti rotor using double sector cells with charcoal-filled Epon centerpieces (pathlength 1.2 cm) and quartz windows. The experiments were conducted at 25°C at a speed of 40,000 rpm and concentration profiles measured at 280 nm. The velocity scans were analyzed with the program Sedfit (version 8.7) (28) using 300 scans collected approximately 2 minutes apart. Size distributions were determined for a confidence level of p = 0.95, a resolution of n = 300 and sedimentation coefficients between 0.1 and 20 s.

Crystallization

Crystals of the U-box domain, Prp19(1-58), were prepared in hanging-drop vapor-diffusion experiments with a 1:1 mixture of protein at 15 mgs/ml and 33% Poly-Ethylene Glycol (PEG) 4000, 75 mM MgCl₂, 0.1 M Tris pH 8.5, 1 mM DTT at 4°C. Seleno-methionine labeled protein crystals were obtained in 38% PEG 4000, 75 mM MgCl₂, 0.1 M Tris pH 8.5, 1 mM DTT at 4°C. Prior to data collection, crystals were serially transferred to mother liquor containing 5, 10, 15% ethylene glycol and flash frozen. Diffraction data statistics are summarized in Table I.

Crystal structure determination

Data were collected at the Center for Advanced Microstructures and Devices (CAMD) using beamline PX. Selenomethionine MAD data were collected to 2.5 Å with wavelength selection based on a fluorescence energy scan. The data were processed using DENZO/SCALEPACK (29). SOLVE was used to locate all six (one per subunit) expected selenium sites (30). Native crystals of the Prp19 U-box dimer diffracted to 1.5 Å. Phase extension and initial model building was accomplished using RESOLVE (31). No averaging of the multiple subunits in the asymmetric unit was employed during refinement. Additional model building and refinement was accomplished using XtalView (32) and Refmac5 (33). Final refinement statistics are summarized in Table 1. The atomic coordinates and structure factors have been deposited in the PDB (PDB ID code 2BAY).

Dimer interfaces were analyzed using the Protein-Protein interaction server (http://www.biochem.ucl.ac.uk/bsm/PP/server/).

Oligomerization state of the Prp19 U-box

The U-box containing protein Prp19 contains three recognized domains: an N-terminal U-box, a central coiled-coil, and a C-terminal WD40 repeat β-propeller domain (Figure 1A). Recently, we demonstrated using electron microscopy and analytical ultra-centrifugation that Prp19 exists as a tetramer, with the coiled-coil domain serving as the primary oligomerization domain (34). The EM studies also revealed that, in the absence of additional binding partners the Prp19 WD40 repeat domains are flexibly tethered to the tetrameric core of the molecule (34). The U-box domain was not directly detectible by EM due to its small size. Thus, the status of the U-box domain in the context of the Prp19 tetramer remained unclear.

To obtain insight into the structure of the U-box within the context of the Prp19 tetramer, a construct was characterized, Prp19(1-133), which contains both the N-terminal U-box and central coiled-coil domain. The oligomerization state of this construct was determined by analytical ultracentrifugation. The major peak (75%) exhibited a sedimentation value (s) of 2.7 with a frictional ratio of 1.9. This is consistent with a tetrameric state of 66kDa (Figure 1B) and reaffirms our previous finding that the coiled-coil domain is both necessary and sufficient for Prp19 tetramerization (34). In addition to the tetramer, a small amount of monomer (<10%, s=1.1) as well as higher order species (∼15%) were detected in the analytical ultracentrifugation analysis (Figure 1B). The frictional ratio indicates an elongated molecule, consistent with the presence of an extended coiled coil domain linked to the globular U-box (34).

CD and NMR spectroscopy were then used to characterize the secondary and tertiary structure of Prp19(1-133), as compared to the intact Prp19 tetramer and the isolated U-box. The minima in the CD spectrum at 208 and 222 nm indicate Prp19(1-133) is a well folded protein with mainly helical secondary structure (Figure 1C), as expected based on the previous U-box structure and the predicted coiled-coil region. Analysis of the ¹⁵N-¹H HSQC NMR spectrum of Prp19(1-133) reveals extensive line broadening with only a few broad resonances discernable (Figure 1D), which is nearly identical to the spectrum of the intact Prp19 tetramer. This result is insightful because it strongly implies that in the context of the tetrameric coiled-coil domain, the U-box domains do not exist as flexibly tethered independent domains, but rather, Prp19(1-133) tumbles as a single unit. If the U-box domains were flexibly tethered to the core coiled-coil tetramer, narrow signals would be observed in the NMR spectrum corresponding closely to those of the isolated U-box domain, as has been seen for example in the small independent RPA32C domain of the ∼50 kDa RPA14/32 complex and the C-terminal domains of the L7/L12 stalk proteins in the intact ribosome (35, 36).

To obtain deeper insight into the state of the U-box in the context of the Prp19 tetramer, we used analytical ultracentrifugation to examine the propensity of the isolated U-box domain to self-associate. This analysis shows that the Prp19 U-box exists in equilibrium between two states, which are assigned to a monomer (s=1.4) and a dimer (s=2.1) in a ratio of 7:3 (Figure 1E) with a frictional ratio of 1.3. Thus, although weak, the U-box does have an intrinsic propensity to dimerize. Importantly, we have observed that the equilibrium is shifted towards the dimer by increasing ionic strength and protein concentration. These results have implications for the U-box, indicating that the prosperity to dimerize will be enhanced significantly in the context of intact Prp19 due to the close spatial proximity of pairs of U-boxes on either side of the tetramer (Fig 1F). This can be understood by considering that tethering of the U-box via a short (∼10 residue) linker to the tetrameric coiled-coil domain results in a very high local concentration of U-boxes, which shifts the monomer-dimer equilibrium strongly towards the dimeric state.

To summarize, the EM, AU, CD and NMR data, taken together, strongly support our contention that the Prp19 U-box exists in a dimeric state in the context of intact Prp19. This conclusion is fully consistent with previous yeast two hybrid evidence that the U-box domain of Prp19 is involved in protein self-association (37).

Crystal structure of the Prp19 U-box dimer

With the objective of producing crystals of the dimeric state of the Prp19 U-box, crystallization trials were set up with a relatively high protein concentration of 15 mg/mL. Crystals were obtained in a relatively rapid manner which diffracted to high resolution. Initial phases were derived from MAD (multi-wavelength anomalous diffraction) experiments using SeMet-substituted protein and the structure was determined to 1.5 Å, with three dimers in the asymmetric unit (Table I, Figure 2A). The U-box domain adopts the typical U-box fold and all copies in the asymmetric unit are superimposable with a backbone r.m.s.d. of 0.28 Å (19, 20). The dimer subunits overlay well with the previously reported NMR structure of the Prp19 U-box in the monomeric state (backbone r.m.s.d. of 0.94 Å), with all significant differences between the crystal and solution structures at the dimer interface. The major structural difference between the structure of the monomer and the dimer is in the conformation of a loop region immediately adjacent to the third β-strand (residues 53-57), which is involved in critical intermolecular contacts (Figure 2A).

The dimer interface of the Prp19 U-box is formed by residues contributed from the N-terminal loop and the three β-strands. The overall buried surface area is 695 Ų/subunit with a 69% hydrophobic and 31% hydrophilic character, and a high degree of shape complimentarity (S_c=0.75) (38). The hydrophobic character of the dimer interface is due to residues Leu15, Ile22, Val51, and Ile53 (Figure 2B, C). Importantly, all four positions are conserved long chain hydrophobics (Val, Ile, Leu) in Prp19 from yeast to man.

Functional analysis of the Prp19 U-box dimer

In order to assess the role of U-box dimerization in Prp19 function, mutations were made to the residues identified at the dimer interface (Figure 2A), including L15, V51, and I53. In order to select the best candidate for the more complex biological assays, biophysical analyses of these mutations were performed first to determine how effective these mutations were at disrupting the ability of the Prp19 U-box to self associate. Not unexpectedly, the most significant effect was from mutation of leucine 15, the hydrophobic residue at the center of the dimer interface. Analytical ultra-centrifugation analysis of the L15E mutant provided a single peak with an s value of 1.4; indicating it exists exclusively in a monomeric state. The frictional ratio of 1.3 is similar to that observed for the peak of the wild-type monomer. Thus, there is no evidence of the monomer-dimer equilibrium seen for the wild-type protein (Figure 2D, cf. Figure 1E). To ensure that the mutant protein adopts a stable, native-like fold an ¹⁵N-¹H HSQC spectrum was acquired for the L15E U-box construct (Figure 2E). The spectrum of the mutant was found to be nearly identical to that of the wild-type U-box, indicating that the structure was not perturbed by the mutation. Consequently, the L15E mutant was selected for analysis of the effect on Prp19 biological activity arising from destabilizing U-box dimerization.

Prp19 is an essential protein, and thus required for viability. Consequently, the functional significance of Prp19 mutations could be determined in vivo by introducing the L15E mutation into the full-length protein. Two complimentary approaches were were utilized. The first involved assessing the ability of the mutant protein to complement the prp19 temperature sensitive allele, prp19-1, at the restrictive temperature (Figure 2F, upper panels). The second approach involved complementation of the null allele of prp19 by a conventional plasmid shuffle approach (Figure 2F, lower panels). The results for L15E in both assays revealed that this single mutation, which disrupts U-box dimerization, is sufficient to compromise cell viability. The origin of this effect is presumably through disruption of the ligase activity of Prp19, or perhaps through destabilization of the NTC complex as observed for the Prp19-1 mutation (19). Taken together, the physical and functional effects of dimer interface mutations underscore the physiological relevance of the Prp19 U-box dimer and imply that dimerization of the Prp19 U-box is necessary for its function in vivo.

Stabilization centers in the U-box domain

As already noted, the binding of two zinc atoms are essential for the stability of the RING-finger domains. The high-resolution crystal structure of the Prp19 U-box provides the first detailed view of the series of hydrogen bonds and salt bridges that serve to stabilize the U-box fold in place of the essential zinc atoms in the RING-finger domains (19). Both protein functional groups and tightly bound water molecules are observed in the U-box networks (Figure 3A). Several useful insights have been obtained by comparing to the corresponding zinc-mediated stabilization centers in the crystal structure of the Rag1 RING-finger (Figure 3B) (39).

The C-terminal stabilization center in the U-box domain involves four core residues (Ser16, Ser19, Asp38, Thr41) with the non-conserved Glu43 peripherally involved (Figure 3C). There are no detectable solvent molecules in this network. The overall arrangement is very similar to that of the corresponding zinc-mediated stabilization center seen in RING-finger domains (Figure 3D). In the case of Prp19, Asp38 appears to be central to the interaction network and closely mimics the zinc atom at the center of the RING-finger stabilization center.

A much different picture emerges for the N-terminal stabilization center of the Prp19 U-box, which includes five core residues (Cys3, Ser6, Arg12, Glu24, Ser26) and two high occupancy water molecules (Figure 3E). The backbone trace of this region of Prp19 is virtually superimposable on the RING-finger domain of Rag1 (Figure 3E, 3F). Despite the remarkable similarity of the backbone conformation, the U-box center differs markedly from that of a RING-finger (Figure 3F). In particular, the residues involved in stabilizing the U-box are more widely distributed than the four zinc chelating residues in the RING finger domain and many more side-chain to backbone amide hydrogen-bonds are apparent. In addition, the U-box center includes high occupancy solvent molecules, something that is not found in the RING-finger structure.

Interestingly, Cys3 and Ser6 in Prp19 are structurally homologous to two of the RING-finger zinc chelating ligands. Although there is no U-box atom that directly corresponds to the RING-finger zinc atom, Glu-24 is located near the center of the extended stabilization site and makes contacts with both solvent molecules and Ser26. The latter is located one residue shifted from the corresponding zinc chelating residue in the RING finger domain. This positional shift appears to result in an expansion of the U-box center and the inclusion of the tightly bound water and Arg12. While static pictures of the stabilization centers provide significant insight, analysis of conformational variability within individual molecules and between the different copies in the unit cell reveal these inter-residue interactions are dynamic. For example, differences in the side-chain conformations of Arg12, Glu24, and Ser26 are evident when comparing different molecules in the asymmetric unit.

Stabilization of RING-finger like domains

The high resolution structures now available for U-box and RING-finger domains provide a basis for analysis of other related domains. In particular, the comparison of residues now shown to contribute to domain stabilization is highly informative. Consider, for example, the Miz-finger domain, which is related to U-box and RING-finger domains but is contained in proteins (e.g. the Siz and PIAS families) that function as E3 ligases in SUMOlation as opposed to ubiquitination pathways (40). Alignment of the U-box domain of Prp19, the RING-finger domain of Rag1, and several Miz-finger domains provides tantalizing clues to the architecture and structural basis for stabilization of the Miz-finger domain, the structure of which has not been determined. This analysis strongly suggests that the Miz-finger has a unique composition (Figure 4A). The Miz-finger domain likely contains a canonical RING-like stabilization center (red) combined with a Prp19 U-box-like stabilization center (aqua) (Figure 4A). Thus, the Miz-finger domain is likely a hybrid U-box/RING-finger domain with N-terminal stabilization via a network of hydrogen bonds and salt bridges, and C-terminal stabilization via zinc chelation (Figure 4B). Since the Miz-finger domains are involved in SUMO-signaling, it is intriguing to consider the biochemical and biological implications of the use of these distinct, yet architecturally similar, domains in parallel cascades. In summary, the addition of the Prp19 U-box to the collection of high resolution structures of RING-finger domains enhances the ability to predict the molecular architecture of RING-finger/U-box-related domains.

Model of the Prp19 tetramer

The determination of the structure of the U-box dimer enables construction of a complete model for the intact Prp19 tetramer: a central tetrameric coiled-coil, with two U-box dimers and four flexibly attached WD40 repeat domains (Figure 5A). E3 ligases require two essential functional domains: an E2 recruiting domain (e.g. U-box) and a substrate recognition domain (e.g. WD40 repeat domain) (Figure 5B). It is commonly held that the mode of action of the E3 ligase is to bring the E2 and substrate into close proximity, thereby catalyzing the transfer of the activated ubiquitin (11, 41). Utilization of complex quaternary structure for positioning of the E2 and substrate has been proposed to be a common feature of complex E3 ubiquitin ligases such as BRCA1/BARD1 and the multicomponent SCF and APC ligases. For example, the βTrcP1 SCF complex has been proposed to structurally enforce the precise structural positioning of E2 and substrate required for catalysis through the spatial organization of the E2 recruiting RING finger and the substrate recruiting WD40 repeats (11). In any single Prp19 polypeptide chain (protomer), the U-box and WD40 repeat domains are spatially remote. The model of the quaternary structure of Prp19 reveals the U-box domain from one Prp19 protomer is in fact in close proximity to a WD40 repeat domain from one of the other protomers (Figure 5C). It is likely that the molecular scaffold created by the Prp19 tetrameric architecture is important for orienting the E2-recruiting U-box dimer with respect to both WD40 repeats and also to other members of the NTC splicing complex (Figure 5C). Clearly, further structural analysis of Prp19 in complex with substrate and its E2 is required to elucidate the mechanism of action and extent of coordination of the oligomer subunits.

A general architecture for dimerization?

While the number of characterized oligomeric E3 ubiquitin ligases continues to grow, it remains to be seen if common architectures are utilized to orient both E2 and substrate recruiting domains. Does the Prp19 U-box dimer reveal a common architecture for orienting and positioning the E2 recruiting domain in an E3 ligase? Dimerization through RING-finger domains in E3 ligases has been examined in two previous studies of Rag1 (39) and BRCA1/BARD1 (21). Comparison of the topology of Prp19 (Figure 6A) with BRCA1 (Figure 6B) and Rag1 (Figure 6C) is instructive. The Prp19 U-box and BRCA1 RING-finger dimers have conserved overall topologies with a similar interface (Figure 6A, B). In both cases, the dimer interface is formed by the face of the three β-strands and additional residues from the N-terminus. Importantly, for BRCA1/BARD1 the dimer interface encompasses residues N- and C-terminal to the RING-finger, which come together to form a four-helix bundle. The formation of the stable four-helix bundle provides the driving force for dimerization, which in turn results in steric constraints that orient the two RING-fingers. Thus, we believe the four-helix bundle in BRCA1/BARD1 serves in a role similar to the four helix coiled-coil in the Prp19 tetramer. In the case of the Rag1 homodimer, a helix from each subunit contributes to the dimer interface, which orients the RING-finger domains in-trans with a separation of ∼15-20 Å. In all three cases, oligomerization positions the two U-box or RING-fingers in a very similar relative orientation. Thus, it appears there is a common overall architecture positioning the E2 recruiting domains in these U-box/RING-finger oligomeric E3 ligases.

If our hypothesis is correct, this architecture should be present in a range of oligomeric E3 ligases. One such E3 ligase is Mdm2, an important regulator of p53 that belongs to the RING-finger family of E3 ligases that has been shown to form RING-finger-mediated homodimers and also heterodimers with MdmX (22, 42). In fact, there are striking similarities between Mdm2, MdmX and Prp19 with respect to conservation of the hydrophobic residues critical for formation of the U-box dimer interface (Figure 6C): Mdm2 contains four long-chain hydrophobic residues in these positions and in MdmX three of the four positions are also hydrophobic residues. Thus, it is quite likely that Mdm2 and MdmX utilize a similar dimerization interface and overall architecture as that found in Prp19.

Functional implications of U-box/RING-finger dimerization

Many essential RING-finger and U-box domain proteins including Mdm2, MdmX, PML, Parkin, Ufd2, and CHIP have been shown to form complex oligomeric E3 ligases (22-25, 43). Consequently, it is highly relevant to consider the possible functional implications of the dimeric state of U-box/RING-finger domains. The potential for recruitment of multiple E2 enzymes carrying activated ubiquitin molecules is clearly very important. Prp19 has two identical E2 binding surfaces on opposite sides of the tetramer and the U-box dimer structure revealed that they are distinct from the dimer interface and therefore both accessible (9, 19). Heterodimerization may offer a means for functional spectificity, enabling the recruitment of two different E2 enzymes. Indeed, BRCA1 has been shown to selectively recruit its E2, Ubc5, in the context of the BRCA1/BARD1 heterodimer (44). The ability to recruit multiple and/or different E2 enzymes may play a role in differentiating the type of ubiquitin-like modifications catalyzed by multi-functional E3 ligases. Mdm2, for instance, has been shown to function in either ubiquitin mediated signaling through recruitment of Ubc5, or in NEDD8 signaling through recruitment of Ubc12 (45). In addition, the ability to recruit more than one E2 to the site of ubiquitination may provide a means for the efficient formation of poly-ubiquitin chains or for the formation of specific branched ubiquitin chains. For example, Ufd2 and CHIP, both U-box domain proteins, can function independently in catalyzing the addition of short ubiquitin chains to UNC-45, whereas the hetero-oligomer is able to rapidly and efficiently poly-ubiquitinate this substrate (43). Thus, the recruitment of multiple E2 enzymes may be critical for the E4 ligase activity observed for Ufd2 (43, 46, 47).