|Home | About | Journals | Submit | Contact Us | Français|
Ubiquitination involves the covalent attachment of the ubiquitin C-terminus to the lysine sidechain of a substrate protein by an isopeptide bond. The modification can comprise a single ubiquitin moiety or a chain of ubiquitin molecules joined by isopeptide bonds between the C-terminus of one ubiquitin with one of the seven lysine residues in the next ubiquitin. Modification of substrate proteins with Lys63-linked polyubiquitin plays a key non-degradative signaling role in many biological processes including DNA repair and NF-κB activation, whereas substrates modified by lysine-48 (Lys48) linked chains are targeted to the proteasome for degradation. The distinct signaling properties of alternatively linked ubiquitin chains presumably stems from structural differences that can be distinguished by effector proteins. We have determined the crystal structure of Lys63 tetraubiquitin at a resolution of 1.96 Å and performed Small Angle X-ray scattering (SAXS) experiments and molecular dynamics (MD) simulations to probe the conformation of Lys63 tetraubiquitin in solution. The chain adopts a highly extended conformation in the crystal, in contrast with the compact globular fold of Lys48 Ub4. Small Angle X-ray scattering (SAXS) experiments show that the tetraubiquitin chain is dynamic in solution, adopting an ensemble of conformations that are more compact than the extended form in the crystal. The results of these studies provide a basis for understanding the differences in the behavior and recognition of Lys63 polyubiquitin chains.
Protein modification by covalent conjugation of one or more ubiquitin molecules, referred as ubiquitination (or ubiquitylation), has been shown to play important regulatory roles in many cellular processes including protein degradation, transcription, cell-cycle progression, viral infection and the immune response1,2. This conjugation involves a cascade of enzymatic steps that results in the covalent attachment of the C-terminal glycine residue of ubiquitin to the ε-NH2 of a lysine sidechain in the substrate protein3. Ubiquitin itself can also be ubiquitinated at any of its seven lysine residues 4, or at its amino-terminus 5, giving rise to different poly-ubiquitin chains.
Polyubiquitin chains of different linkage types, as distinguished by the lysine residue to which each successive ubiquitin C-terminus is attached, have distinct biological functions. The best understood among these are the Lys48-linked ubiquitin chains, which target proteins for degradation by the 26S proteasome 6. Lys63-linked ubiquitin chains, on the other hand, play non-degradative roles in different signaling pathways 7, notably NFκB transcription activation 8 and the DNA damage response 9, 10. Other types of chain linkages have also been identified in cells 11, including branched and mixed-linkage chains 12, 13. Although the cellular roles of these “atypical” ubiquitin chains are yet to be studied in detail, they have been implicated in a few biological processes 14. For example, K33/K29 mixed chains have been shown to regulate AMPK-related kinase activation while K29-linked poly-Ubs have been shown to play role in lysosomal degradation of proteins 15, 16.
The distinct roles played by different polyubiquitin chains indicate that effector molecules can distinguish among chains of differing linkage types. For example, the NZF domain of TAB2 binds specifically to Lys63-linked chains 17, the ataxin3 UIM repeat preferentially binds to Lys48-linked chains 18, and the UBAN motif of NEMO recognizes linear head-to-tail ubiquitin chains 19. A thorough understanding of the molecular details of ubiquitin-mediated signaling therefore requires elucidation of the structures of different poly-ubiquitin chains and also their complexes with cognate partners.
The crystal structure of Lys48-linked tetraubiquitin determined at neutral pH showed it to adopt a compact globular conformation 20. That conformation could not be adopted by Lys63-linked tetraubiquitin as judged by the relation between the Lys63 side chains and the ubiquitin C-termini. Indeed, NMR studies 21, 22 and an x-ray crystal structure 17 of Lys63-linked diubiquitin show an open structure with limited contacts between ubiquitin monomers, as compared with the closed conformation of Lys48-linked diubiquitin.
In order to gain insight into the structure and topology adopted by longer Lys63-linked polyubiquitin chains, we determined the 1.96 Å resolution crystal structure of Lys63-linked tetraubiquitin. The crystals contain one tetra-ubiquitin chain per asymmetric unit and show a highly extended linear arrangement of ubiquitin monomers. The extended arrangement showed no inter-subunit interactions, in agreement with NMR results. The extended conformation of the Lys63 tetraubiquitin in the crystal suggests that the chain would likely adopt other conformations in solution. To further explore this conformational flexibility of the Lys63-linked tetraubiquitin chains in solution, we carried out small angle X-ray scattering measurements (SAXS) of Lys63 tetraubiquitin and used molecular dynamics simulations to deduce the ensemble of structures adopted by Lys63 tetraubiquitin in solution. We find that Lys63 tetraubiquitin adopts conformations in solution that are more compact than the extended form seen in the crystal, yet do not exhibit any specific inter-subunit interactions. These studies provide a detailed understanding of the structure and dynamics of Lys63 tetraubiquitin and how these chains could be recognized in the cell.
Lys63-linked tetraubiquitin chain was synthesized using two ubiquitin mutants (K48R-K63R and Ub-D77) following Pickart and Raasi 23. All the enzymes required for this synthesis viz., E1 (human), Ubc13/MMS2 (Yeast) and Yuh1 (yeast) were also expressed and purified as recombinant proteins in Escherichia coli. After the chain of required length was synthesized, it was subjected to a final step of purification using size-exclusion chromatography in Superdex 75 prep-grade 26/60 column. The protein was then dialyzed in 10 mM Tris-HCl buffer (pH 7.6) containing 25 mM NaCl, overnight at 4°C. It was then concentrated using Vivaspin concentrators and stored at −80°C in small aliquots.
Crystals of Lys63 Ub4 were initially obtained at 20°C in hanging drop setups from sparse matrix screening with buffer 1 from Wizard II (Emerald Biosciences). The condition was then optimized to get fewer and better crystals. The best crystals were obtained using 4–6% PEG 3000 + 100 mM Na-acetate (pH 4.6–5.0) + 10 μM Zn acetate. However all the crystals were found to diffract poorly (best 3.5Å) and also showed very high mosaicity and B-factor after data processing. The diffraction improved substantially upon dehydration of the crystals by step-wise increase of PEG 3000 to 25% over a period of 4–5 days and a 1.95Å dataset was collected from frozen crystals at 100 K at APS, Argonne, IL at the fixed wavelength beamline 14-BMC.
The data was processed in the space group P213 using the program HKL2000 24. The structure was solved by molecular replacement using the program MOLREP 25 from CCP4 Suite 26. The initial model was subjected to iterative building and refinement cycles using COOT 27 and REFMAC5 28. All the side chains were built first followed by the completion of the chain. Iso-peptide bonds were created towards the end with addition of water molecules.
SAXS data were collected at the SIBYLS beamline (beamline 12.3.1) at the Advanced Light Source in Lawrence Berkeley Labs. Data were collected from the sample at concentrations of 3, 6 and 9 mg/ml and no concentration dependence was observed. The sample buffer contained 10 mM Tris-HCl (pH 7.6), 50 mM NaCl and 0.5 mM EDTA at 10°C. Five and half second exposures were merged to create a final scattering profile. P(r) functions were determined from the scattering profiles using GNOM 29. Scattering profiles were calculated from crystal structures using CRYSOL 30. The MES program was run as described 31 with 3 independent runs per radius of gyration ranging from 20 to 60 Å at 1 Å steps.
Crystals of Lys63 tetraubiquitin formed in spacegroup P213 and contain one chain per asymmetric unit. Crystals obtained after initial rounds of optimization typically diffracted to 3.5–4.0 Å resolution. However, the diffraction improved significantly upon gradual dehydration of the crystals with step-wise increases of the precipitant concentration, yielding crystals diffracting to 1.96 Å resolution. The structure was solved by molecular replacement using the structure of monomeric ubiquitin 32 as the search model. Data collection and refinement statistics are presented in Table 1.
The crystal structure shows a linear arrangement of the Lys63-linked ubiquitin monomers (Figure 1a & b) that is starkly different from the compact globular fold adopted by Lys48-linked tetra-ubiquitin (Figure 1c). Instead, the monomers in the Lys63-linked chain are joined by well-ordered linkages but do not contact each other. The absence of interactions between the individual ubiquitin monomers of the Lys63 Ub4 agrees with the previously reported NMR results, where no stable inter-subunit interactions could be observed in Lys63-linked di- and tetra-ubiquitin chains 21, 22 as well as with the extended structure of crystallized di-ubiquitin 17. The extended conformation of the tetraubiquitin chain and the absence of intra-chain contacts that could stabilize the observed conformation suggested that Ly63-linked tetraubiquitin in solution would likely be flexible and adopt other conformations in addition to that seen in the crystal.
We also note that the globular regions of each of the four ubiquitin monomers in the Lys63 tetraubiquitin chain alternate between two conformations that differ in the loop region comprising residues 6 through 12. Inclusion of this region during superposition of the globular region of the ubiquitin monomers (residues 1 through 70) increase the root mean squre deviation to ~0.85 Å from ~0.4 Å. The reason for this difference remains unclear and could not be attributed to the differences in crystal packing environments of each monomer.
Ubiquitin has a conserved hydrophobic patch comprising residues Leu8, Ile44 and Val70 that has been shown to be important for function 33. These residues mediate the interaction of ubiquitin with E2 enzymes 34 and also with ubiquitin binding domains (reviewed in 35). The importance of these hydrophobic residues in the Lys48-linked ubiquitin chain recognition has been studied in detail 36, 37, 38. In addition, the hydrophobic patch has also been implicated in interactions with Lys63 linked chain interactions 39 and is shown to participate in the linear diubiquitin recognition by NEMO 40. In contrast with Lys48 tetraubiquitin, in which the hydrophobic patches from two adjacent monomers coalesce to form a single surface (Figure 1c), the hydrophobic patches in the crystal structure of Lys63 tetraubiquitin are exposed in all four ubiquitin monomers (Figure 1b). We also note that the hydrophobic patches of Lys63 monomers do not lie on the same surface of the molecule, but are rotated roughly 90 degrees relative to one another in successive monomers. This particular arrangement is probably favored in the crystal, where the molecules are conformationally “locked” in one particular orientation due to crystal packing. However, we note that one of the solution structures presented by Varadan et al.22 does also show this orthogonal positioning of the ubiquitin hydrophobic surface in successive monomers along with alternate arrangements. Given the presumed flexibility in the linkages that join the ubiquitin moieties in Lys63 chain, the individual monomers would probably be free to undergo rotational movements that would change the relative orientation of these patches.
Despite the strong tendency of ubiquitin to associate via the hydrophobic patches, we were surprised to find that these patches do not take part in crystal packing of the Lys63 Ub4 chains. Instead, the crystal contacts are formed by the interaction of an ubiquitin monomer in one chain with the isopeptide linkage of another chain in the crystal (Figure 2). This crystal packing interaction could potentially reflect a mode of interaction between polyubiquitin chains, or of ubiquitin-like domains with Lys-63-linked chains. Several proteins including IKKα and IKKβ have been found to contain Ubiquitin like domains (UBLs). These UBL domains could possibly interact with Lys63 chains in a manner similar to that seen in the crystal. However, to our knowledge, no such interactions have been reported to exist in the literature.
Since the extended conformation of the Lys63 tetraubiquitin is clearly influenced by lattice contacts, we used small angle X-ray scattering (SAXS) combined with molecular dynamics (MD) simulations to explore the structure of Lys63 tetraubiquitin in solution. SAXS is a powerful tool to study the conformation of macromolecules in solution, particularly when a high resolution crystal structure is available 41. We measured the x-ray scattering profile of Lys63 tetraubiquitin in solution (Fig. 3a) and derived the pair distribution function, P(r), from the scattering profile to determine the distribution of interatomic vectors in the tetraubiquitin chain (Fig. 3b). As shown in Figure 3a, the measured x-ray scattering profile does not fit the theoretically calculated profiles from the crystal structures of either Lys63 tetraubiquitin (this study) or of Lys48 tetraubiquitin 20. This indicates that Lys63 tetraubiquitin adopts a distinct conformation in solution that is more compact than that seen in the crystal.
The interatomic vectors calculated in the pair distribution function (Fig. 3b) show that the maximum dimension of Lys63 tetraubiquitin in solution (Dmax) is 118±2Å with a radius of gyration (Rg) 32.3 Å, which is considerably smaller than the Dmax (154 Å) and Rg (44 Å) values calculated from the crystal structure. This apparent reduction of Dmax and radius of gyration indicates that Lys63-linked tetraubiquitin is more compact in solution than in the crystal. At the same time, the peaks in the experimental P(r) also show that Lys63 tetraubiquitin adopts a much more extended structure than the crystal structure of Lys48 tetraubiquitin (Figs. 3a & b). Interestingly, the second maximum in P(r) of ~41Å corresponding to the average distance between any two consecutive ubiquitin monomers within the Lys63 polyubiquitin chain closely matches the calculated value from the crystal structure (Fig. 3b). However, the third maximum, which corresponds to the distances between first and third ubiquitin monomer or between the second and fourth, is considerably shorter in the observed profile (63Å) as compared to the one calculated from the crystal structure. The combination of a large Dmax but minor third and fourth maxima suggested that the Lys63 ubiquitin chains most likely exist as an ensemble of conformations at larger length scales.
Further analysis of the solution conformation of Lys63 tetraubiquitin was done using the program Minimum Ensemble Search (MES) 31. This program finely samples the conformational space of the macromolecule using molecular dynamics (MD) simulations at high temperatures, where only covalent bonds and van der Waals interactions are taken into consideration. A genetic algorithm then identifies the minimal ensemble from all these conformations that best fits the data. The result shows that no single conformation from the MD simulations adequately explains the data. Instead, an optimal fit could be obtained with an ensemble of a minimum of three conformations (Figure 3c & 4). Inclusion of more conformations in the ensemble did not lead to any significant improvement in the fit, as measured by chi-squared values (data not shown). The first two dominant conformations (represented by 69% and 17% in the ensemble) (Figure 4a & b) preserve the overall linearity observed in the crystal structure but are notably more compact than the conformation in the crystal. The ensemble also contains a minor population (13% in the ensemble) (Figure 4c) that is highly compact as compared to the crystal structure.
Though the resolution of SAXS data cannot be used to determine the inter-subunit orientations accurately, we note that, in all of these MES conformations, the conserved hydrophobic interaction surface (colored red) of the ubiquitin monomers do not interact with each other and are either completely or partially exposed. This probably suggests that Lys63 chains are likely to be found in solution conformations that leave the hydrophobic patches exposed and would therefore be readily accessible for interaction with cognate proteins.
In summary, our results show that Lys63-linked polyubiquitin forms a relatively open conformation, in agreement with the existing biochemical and biophysical data. The elongated structures of the K63 tetraubiquitin, stabilized by a criss-crossed arrangement of chains within the crystal lattice, do not belong to any closed symmetry group. Thus, addition of more ubiquitin monomers to either the proximal or distal end of the chain can occur without requiring any rearrangement of the structure. Although the highly extended crystal structure of Lys63 tetraubiquitin agrees with predicted models based on diubiquitin structure 17, the longer chain is notably more collapsed in solution.
Nevertheless, none of the representative solution conformations or the crystal structure buries the conserved hydrophobic patch of any of the ubiquitin monomers. This indicates that these patches remain available for interactions with other proteins.
The requirement of an ensemble instead of a single conformation to explain the SAXS data point towards the fact that Lys63 ubiquitin chains with four or more monomers do not form a single well-defined structure in solution. Instead, tetraubiquitin and presumably longer K63 chains exist in a dynamic equilibrium of many different conformations. Upon binding to a cognate protein, Lys63 chains would probably adopt one specific topology depending on the binding partner. This fluidity in the structure of Lys63 polyubiquitin may be important for its biology, where these chains are recognized by proteins with potentially different structures in different signaling pathways.
Moreover, the conformational space sampled by the free Lys63 polyubiquitin can be restricted upon attachment to substrate proteins, each of which could have a unique influence on the Lys63 polyubiquitin chain behavior. All these considered together point towards a highly plastic nature of Lys63-linked polyubiquitin mediated signaling mechanisms, which could account for the diverse roles played by this type of polyubiquitin modification in multiple signaling pathways.
We thank Michal Hammel from LBNL for help with the implementation of MES. Supported by the Howard Hughes Medical Institute (A.B.D and C.W.) and NCI SBDR grant # CA92584 (G.H.).
Coordinates and structure factors have been deposited in the Protein Data Bank with accession number 3HM3.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.