Solution structure of lysine-free (K0) ubiquitin


  • Tao Huang,

    1. Structural Biophysics Laboratory, Center for Cancer Research, National Cancer Institute, Frederick, Maryland
    Search for more papers by this author
  • Jess Li,

    1. Structural Biophysics Laboratory, Center for Cancer Research, National Cancer Institute, Frederick, Maryland
    Search for more papers by this author
  • R. Andrew Byrd

    Corresponding author
    1. Structural Biophysics Laboratory, Center for Cancer Research, National Cancer Institute, Frederick, Maryland
    • Correspondence to: R. Andrew Byrd; Structural Biophysics Laboratory, Center for Cancer Research, National Cancer Institute, Frederick, MD 21702. E-mail:

    Search for more papers by this author


Lysine-free ubiquitin (K0-Ub) is commonly used to study the ubiquitin-signaling pathway, where it is assumed to have the same structure and function as wild-type ubiquitin (wt-Ub). However, the K0-Ub 15N heteronuclear single quantum correlation NMR spectrum differs significantly from wt-Ub and the melting temperature is depressed by 19°C, raising the question of the structural integrity and equivalence to wt-Ub. The three-dimensional structure of K0-Ub was determined by solution NMR, using chemical shift and residual dipolar coupling data. K0-Ub adopts the same backbone structure as wt-Ub, and all significant chemical shifts can be related to interactions impacted by the K to R mutations.


Protein ubiquitination is a posttranslational modification that regulates almost all aspects of cell life. In this process, ubiquitin (a 76 amino acid protein) is covalently attached to a lysine on the targeted proteins through an isopeptide bond formed between the carboxylate of the C-terminal glycine residue of ubiquitin and a free Nζ of a lysine of the substrate. Subsequently, a lysine residue on the first ubiquitin can covalently link through another isopeptide bond to a second ubiquitin and so on to create a polyubiquitin chain on target proteins. There are seven lysine residues in ubiquitin at positions 6, 11, 27, 29, 33, 48, and 63, which can be used in isopeptide linkages and create many types of polyubiquitinated proteins. These polyubiquitin chains can have different topologies, can vary in length, and have different functional consequences.[1] Functional investigations of ubiquitin pathways often utilize lysine-substituted variants, which have either a single lysine or all lysines substituted with arginine.[2] These variants, particularly lysine-free ubiquitin (K0-Ub), are used to define specific ubiquitin modifications and for detecting ubiquitination of specific substrates.[3-7] In spite of its wide applications in biology and biochemistry, no structural study has been reported for K0-Ub, although structural equivalence between wild-type ubiquitin (wt-Ub) and single K-to-R variants[8, 9] have been reported. As lysine represents almost 10% of the residues of ubiquitin, it is of interest whether the substitution alters the three-dimensional (3D) fold or stability of ubiquitin. Furthermore, as NMR is widely used in ubiquitination studies, including the mutants of ubiquitin[8-15] and conformations of polyubiquitin chains,[8, 16-18] it is important to characterize the NMR spectra of K0-Ub and validate the solution structure. Hence, we have used solution NMR to characterize the 3D structure of K0-Ub and interpret the unexpected and dramatic differences in NMR spectra.

Results and Discussion

NMR spectroscopy of K0-Ub

The 15N-1H heteronuclear single quantum correlation (HSQC) spectrum was collected for K0-Ub and compared with wild-type ubiquitin [Fig. 1(A)]. K0-Ub showed a well-dispersed HSQC spectrum with uniform peak intensity and the expected number of backbone amide peaks, indicating that K0-Ub adopts an ordered 3D structure. However, the HSQC spectrum of K0-Ub is substantially different from wt-Ub, with more than half of the peaks shifting to an extent as to be unreliably traceable from wt-Ub to K0-Ub. As the 15N-1H HSQC spectrum represents the chemical environment of each amino acid residue and is considered a “fingerprint” of the protein structure, the large difference in HSQC spectrum might come from significant differences in the chemical environment and raise the question as to whether K0-Ub adopts a distinct 3D structure.

Figure 1.

Spectra comparison between ubiquitin and K0-Ub. (A) 15N-1H HSQC spectra overlay of ubiquitin (blue) and K0-Ub (red). (B) CSP between ubiquitin and K0-Ub. Average CSP of 1H and 15N chemical shifts is calculated by the formula: CSP = [(Δδ2NH + Δδ2N/25)/2]1/2 (ppm), where ΔδNH and ΔδN are the amide proton and nitrogen shift differences for conserved residues between ubiquitin and K0-Ub, respectively. “KR” represents lysine to arginine mutations, “P” represents proline and residues 24 and 53 are missing due to solvent exchange. Hydrogen bonds are connected with gray lines and salt bridges are connected with black lines.

The NMR chemical shifts for the backbone atoms (HN, N, C, Cα, Cβ, Hα, and Hβ) of K0-Ub were assigned by standard triple resonance experiments. For 70 of the 72 observed N-HN peaks in the 15N-1H HSQC spectrum, complete resonance assignments were possible, shown in Supporting Information Figure S1. Residues E24 and G53 were not observable, due to exchange with water, as found in wild-type ubiquitin.[19] The backbone chemical shift perturbation (CSP) was calculated between wt-Ub and K0-Ub, shown in Figure 1(B). The CSP values range from 0 to 0.6 ppm, and the CSP for 33 out of 66 conserved residues are larger than 0.1 ppm. A few residues which have large CSPs are adjacent to lysines, such as T7 (next to K6) and A28 (between K27 and K29), indicating that the lysine to arginine substitution imparts a nearest neighbor effect. In the case of T7, there is an additional effect from a backbone hydrogen bond between T7 and K11. If the strength of this hydrogen bond is impacted by the K11R mutation, then T7 could be expected to have an additional shift. In fact, the amide group of T7 shifts 0.36 ppm. Additionally, some of the lysine side chain NζH3+ groups (e.g., K11, K27, K29, and K33) form extensive hydrogen bonds with backbone carbonyls and salt bridges with side chain carboxylates of Glu and Asp, which were observed from the X-ray crystal structure[20] and solution NMR studies[21] (Supporting Information Table S1). The hydrogen bonds and salt bridges from lysine NζH3+ are summarized in Figure 1(B). Many of the residues with large CSPs are hydrogen bond acceptors from lysine NH3+, or have a secondary interaction with one of the affected residues or one residue adjacent. The lysine to arginine substitutions will alter the distances between donor (NζH3+ becoming Nη1H2 or Nη2H2) and acceptor oxygen atoms, hence perturbing or abolishing the hydrogen bonds and salt bridges. It is also possible to have different rotamers of the arginine sidechain to achieve the hydrogen bond with one of the amino groups. For example, K27 is involved in a salt bridge with D52, which has a large CSP. Also, K27 participates in a backbone hydrogen bond with the I23 carbonyl oxygen. If the K27R mutation alters this hydrogen bond strength or direction slightly, then the CSP of I23 (0.22 ppm) could result. Similarly, K27 Nζ participates in a hydrogen bond to the backbone carbonyl of Q41. The Q41 NH correlation peak is not significantly shifted, but the adjacent R42 exhibits a large CSP (0.39 ppm). Careful examination of the observed CSPs in K0-Ub indicates that the majority of the shifted residues correlates with interactions impacted by the conversion of lysine to arginine or are adjacent to one of these interactions. Hence, the possibility arises that there could be substantive conformational affects due to the simultaneous mutation of all seven lysine residues.

Solution structure of K0-Ub

The secondary structure of K0-Ub was determined from chemical shift assignments using PECAN[22] and TALOS+[23] for K0-Ub and compared with ubiquitin (PDB ID: 1UBQ or 1D3Z). Both PECAN and TALOS+ reveal that K0-Ub adopts the same secondary structure as ubiquitin, with average deviations of phi and psi angles 3.4° and 5.9° from wt-Ub, respectively (Supporting Information Fig. S2). The 3D structure of K0-Ub was calculated using the chemical shift-based Rosetta (CS-Rosetta) method,[24] with the addition of HN-N residual dipolar coupling (RDC) data. All assigned HN, N, C, Cα, Cβ, and Hα chemical shifts plus 52 N-HN RDCs, representing a random selection of 80% of complete RDC data, were used in the calculations. The remaining 13 RDCs (20% of the data) were used for cross-validation. The RDC data significantly improved the convergence of the ensemble of structures, shown in Figure 2(A). The ensemble of the 20 lowest-energy structures [Fig. 2(B)] exhibited excellent Ramachandran statistics and had an overall backbone root mean squared deviation (RMSD) of 0.30 Å in the structured region (Table 1). The best model, with the lowest Rosetta energy, is in good agreement with the RDC values withheld from the Rosetta calculation, and the RDC Q-factor is 0.20, determined using PALES.[25] The lowest energy model of K0-Ub has a backbone RMSD of 0.62 Å compared to the crystal structure of wt-Ub (PDB: 1UBQ), and the superposition of K0-Ub and wt-Ub is shown in Figure 2(C). Interestingly, 65 RDCs measured for K0-Ub fit very well to the solution structure of wt-Ub (PDB: 1D3Z), exhibiting a Q-factor of 0.15. These results confirm that K0-Ub adopts the same 3D structure as wild-type ubiquitin.

Figure 2.

Solution structure of K0-Ub compared with wt-Ub. (A) Plot of rescored Rosetta all-atom energy versus Cα RMSD relative to the lowest-energy model (bold dot on vertical axis). (B) 20 lowest-energy CS-Rosetta models of K0-Ub superimposed. (C) Superposition of best K0-Ub model (green) with ubiquitin (orange; PDB: 1UBQ).

Table 1. Structural Statistics for the Ensemble of 20 Calculated Structures of K0-Ub
RMSD valuesAllOrderedaSelectedb
  1. a

    Residues with sum of phi and psi order parameters >1.8. Ordered residue ranges: 2–51, 54–71.

  2. b

    Residues selected based on: Dihedral angle order parameter, with S(phi) + S(psi) ≥ 1.8. Selected residue ranges: 2–51, 54–71.

  3. c

    With respect to mean and standard deviation for a set of 252 X-ray structures <500 residues, of resolution ≤ 1.80 Å, R-factor ≤ 0.25 and R-free ≤ 0.28; a positive value indicates a “better” score.

 All backbone atoms1.4 Å0.3 Å0.3 Å
 All heavy atoms1.6 Å0.8 Å0.8 Å
Structure quality factors—overall statistics
 Mean scoreSDZ-scorec
 Procheck G-factora (phi/psi only)−0.09N/A−0.04
 Procheck G-factora (all dihedral angles)0.25N/A1.48
 Verify 3D0.440.0336−0.32
 ProsaII (−ve)0.910.03851.08
 MolProbity clashscore4.061.75870.83
Ramachandran plot summary from procheckb
 Most favored region95.5%  
 Allowed region4.5%  
 Generously allowed region0.0%  
 Disallowed region0.0%  

The sensitivity of the 15N-1H HSQC spectrum to the hydrogen bonding and salt-bridge network within a protein has been observed previously,[26, 27] where the structure remains unchanged but the 15N-1H chemical shifts are significantly affected. Despite the large number of chemical shift differences (50% of ubiquitin residues have CSP values larger than 0.1 ppm), K0-Ub retains the same structure as wild type. This situation might seem surprising in light of the widely held view that chemical shift affects imply either ligand binding or structural changes; rather, the data suggest that interpretation of amide HN-N CSP data include consideration of hydrogen bonding perturbations. We have also observed a small population, generally <10–15%, of K0-Ub molecules with altered 15N-1H chemical shifts, but the same C, Cα, Cβ, Hα, and Hβ chemical shifts. This occurs for many peaks in the HSQC spectrum. Assignment of these signals and analysis of the secondary structure from TALOS+ shows that the minor population adopts the same fold as the major population and these shifts represent alternate conformations in the hydrogen bonding and salt-bridge network, which occur in mostly random members of the molecular ensemble and not as a single minor species. See Supporting Information for further discussion.

Thermal stability of K0-Ub

The thermal stability of K0-Ub, assessed by the melting temperature (Tm), was examined using differential scanning calorimetry (DSC; Supporting Information Fig. S3). K0-Ub is a fairly thermostable protein, exhibiting a Tm of 71.9°C; however, this value is 19°C lower than that of wt-Ub (Tm = 90.6°C). The decreased Tm of K0-Ub is consistent with the loss or perturbation of hydrogen bonds and salt bridges that occur in the substitution of lysine to arginine. Nevertheless, K0-Ub remains a stable, folded analog of wt-Ub and should not represent any difficulties or substantively altered behavior in biochemical experiments.


We have examined the structure of K0-Ub by solution NMR methods and demonstrated that K0-Ub retains the same secondary and tertiary structure as wild-type ubiquitin. This is similar to findings for single K-to-R mutants[8, 9] and for cases that examined hydrophobic core residues,[8, 9, 15] showing that the hydrophobic core of ubiquitin could be redesigned without altering the 3D structure. The structural equivalence is consistent with our observations that K0-Ub is readily (and equivalently to wt-Ub) loaded by the E1 ubiquitin activating enzyme onto E2 conjugating enzymes.[28] The chemical shift assignments and solution structure obtained in these studies enables confident, NMR-based monitoring of ubiquitination and conformational preferences of K0-Ub in general mechanistic studies of the ubiquitination pathway.

Materials and Methods

Protein expression and purification

The coding sequence of K0-Ub in a pET vector was the kind gift of Dr. Allan M. Weissman. Uniformly 15N- or 15N/13C-labeled K0-Ub were expressed in Escherichia coli (BL21DE) cells grown on M9 minimal medium enriched with 0.1% 15NHCl4 alone or together with 0.3% uniformly 13C-labeled d-glucose. Protein expression was induced by adding 1 mM isopropyl β-d−1-thiogalactopyranoside at OD600 ∼ 0.8 for 3 h at 37°C. E. coli cells were harvested and lysed in 50 mM Tris, pH 7.4 using a microfluidizer. Soluble cell lysate were passed over a Q-sepharose column (GE Healthcare), the flow-through containing K0-Ub was collected and concentrated. K0-Ub was further purified using a Superdex S75 column in 50 mM Tris, pH 7.2 (GE Healthcare).

NMR spectroscopy

NMR samples were prepared in 50 mM Tris, pH 7.2 and experiments were performed at 25°C on a Bruker Avance-III 700 MHz spectrometer, equipped with a triple-resonance cryogenic probe. The protein samples were 0.7 mM for resonance assignment experiments. The assignment of the backbone resonances of K0-Ub were accomplished by recording and analyzing HNCACB, CBCA(CO)NH, HNCO, and HBHACONH triple resonance datasets.[29] Data processing and analysis were performed using the nmrPipe[30] and Sparky[31] software packages, respectively. Backbone HN-N RDCs were measured using an in-phase anti-phase (IPAP) experiment[32] with 15N K0-Ub aligned in 4% (w/v) C12E5 polyethylene glycol/n-hexanol media.[33]

Structure calculation

Structures of K0-Ub were calculated using chemical shift-based Rosetta (Rosetta 3.4) using the standard protocols24,34 and incorporating backbone N, HN, C, Cα, Hα, and Cβ chemical shifts together with 80% N-H RDC data (52 RDCs randomly selected from total 65 RDCs in structured region from 1 to 71). The calculations generated 10,000 all-atom models, and the 3000 low-energy models were extracted and further rescored against chemical shifts. The final 20 lowest energy structures were chosen as representative of the calculation. The remaining 20% of the RDCs (13 RDCs, not used in calculation) were used to cross-validate the final models by PALES, using the alignment tensor fitted from the 52 RDCs that were used in calculation.[25] Validation of the structure was performed using protein structure validation suite (PSVS) (

Data deposition

The coordinates of the solution structure have been deposited in the PDB as entry 2MI8. The chemical shift assignments and RDC data have been deposited in the BioMagResBank as entry 19670.

Differential scanning calorimetry

The thermal stabilities of wt-Ub and K0-Ub were measured with a VP-DSC MicroCalorimeter (MicroCal, GE Healthcare). The concentrations of the two proteins were 1.2 mg/mL in 50 mM Tris (pH 7.2). The heating rate used was 1°C/min, and the scanning was performed from 25 to 100°C.


The authors thank Marzena A. Dyba and Sergey G. Tarasov for assistance with DSC measurements and Yang Shen, Nikolaos Sgourakis, Oliver Lange for help with CS-Rosetta calculation. None of the authors have any conflict of interest regarding the material or data presented in this publication.