Noncovalent Dimerization of Ubiquitin**
Zhu Liu, Wei-Ping Zhang, Qiong Xing, Xuefeng Ren, Maili Liu, and Chun Tang*
Ubiquitin is a small signaling protein in cells and is highly
conserved throughout the eukaryotes. Ubiquitin interacts
with myriad partner proteins that contain one or more
ubiquitin-binding domains (UBDs). To achieve multivalent
binding with several UBDs, ubiquitins are often covalently
linked by an isopeptide bond between the C-terminal
carboxyl group of one ubiquitin and a primary amine in
another;[1a,b]the two subunits in a di-ubiquitin are referred to
as the proximal unit and the distal unit, respectively. All seven
lysines and the N-terminus of ubiquitin can participate in the
isopeptide bond.In tandem, multiple ubiquitins can be
linked up to form a poly(ubiquitin).
Depending on the site of the linkage, di- and poly(ubi-
quitin)s can display distinct quaternary structures, which may
account for their linkage-specific functions.[1a,b]Among the
linkages, Lys11, Lys48, and Lys63-linked poly(ubiquitin)s are
best characterized: Lys63-linked poly(ubiquitin) is involved
in cellular events such as endocytosis and DNA repair, while
both Lys11- and Lys48-linked poly(ubiquitin)s can signal for
proteosomal degradation.[1a,b]In crystal, Lys48-linked di-
ubiquitin mostly adopts a closed conformation, burying
hydrophobic residues around Ile44 in both subunits;[3a,b,4]
Lys63-linked di-ubiquitin adopts an open extended structur-
e;[5a,b,6]while Lys11-linked di-ubiquitin displays intermediate
Structural heterogeneity has also been observed for di-
and poly(ubiquitin)s with the same linkage. Lys48-linked di-
ubiquitin has been crystallized in multiple forms, one of which
actually adopts an open conformation (Figure S1a),while in
solution at a neutral pH value, approxaimely15% adopts the
open conformation.Although adopting an extended con-
formation in crystals,[5a,b]a highly compact structure was
deduced for a sub-population of Lys63-linked di-ubiquitin
from small-angle X-ray scattering data in solution.For
Lys11-linked di-ubiquitin, chains in a single asymmetric unit
of crystal structure display substantial variations with root
mean square (rms) variations over 6 ? (Figure S1b).
In a quest to understand what affords the multitude of
quaternary structures and to elucidate a linkage–structure
relationship for poly(ubiquitin), we serendipitously discov-
ered that free ubiquitin dimerizes noncovalently in solution.
At increasing protein concentration, a subset of peaks of
ubiquitin shift progressively (Figure 1a and S2), correspond-
ing to residues 8, 13, 44, 45, 46, 49, 67, 68, 70, 71, and 73, which
are located at the b-sheet region of the protein and form a
contiguous surface (Figure 1b). Plotting the chemical shift
values measured at 308 8C over protein concentrations, the
Figure 1. a) A representative region of 2D1H-15N HSQC spectra of15N-
labeled ubiquitin, collected at concentrations from 0.2 mm to 3.3 mm
(rainbow-colored from red to purple, respectively) at 308 8C; b) cartoon
and surface representation of ubiquitin. Residues displaying large
chemical shift changes (Dwmax?25 Hz) are colored red. Lys11, Lys48,
and Lys63 are shown as ball-and-stick representations; c) changes of
chemical shift values over protein concentrations for all perturbed
residues can be globally fitted to a monomer–dimer equilibrium with
KD=(4.9?0.3) mm. Relative to the lowest protein concentration
(0.2 mm), the chemical shift differences are expressed as (dH+dN)1/2,
in which dHand dNare in Hz; d) protein rotational correlation time tc
can be fitted to the same equilibrium. Error bars represent one
[*] Z. Liu, Q. Xing, Prof. M. Liu, Prof. C. Tang
State Key Laboratory of Magnetic Resonance and Atomic and
Molecular Physics, Wuhan Institute of Physics and Mathematics
Chinese Academy of Sciences, Wuhan, Hubei 430071 (China)
Prof. W. P. Zhang
Department of Pharmacology, School of Medicine, Zhejiang
Hangzhou, Zhejiang (China)
Dr. X. Ren
National Institute of Child Health and Human Development
National Institutes of Health, Bethesda (USA)
[**] Funding from the Ministry of Science and Technology of China
(2009CB918600 to M.L.), the Chinese Academy of Sciences (KJCX2-
EW-W05 to C.T.), and the National Natural Sciences Foundation of
China (21073230 to C.T.) is acknowledged. We thank Christian
Griesinger and G. Marius Clore for advice, and Yong Duan for
careful reading of the manuscript.
Supporting information for this article is available on the WWW
Re-use of this article is permitted in accordance with the Terms and
Conditions set out at http://angewandte.org/open.
Angew. Chem. Int. Ed. 2012, 51, 469–472 ? 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
curves can be globally fit to a monomer–dimer equilibrium
with a dissociation constant KDof (4.9?0.3) mm (Figure 1c).
Thus, the dimer population amounts to 50% at approximately
(0.12 mm for dimeric species) at 1 mm. Backbone
relaxation rates are functions of the protein rotational
correlation time tc, which is in turn determined by the
proportion of the dimer. The afforded tcvalues increase
monotonously with protein concentration and can be fit to the
same monomer–dimer equilibrium (Figure 1d). Further
experimental evidence is provided by analytical ultracentri-
fugation analysis. Sedimentation equilibrium data collected at
different wavelengths and different protein concentrations
deviate from the curves predicted for a pure monomeric
species, and give an apparent molecular weight that is
approximately 20% larger than the actual value. The best
fit is obtained using a monomer–dimer equilibrium model
with a KDof (4.4?1.5) mm (Figure 2). Taken together, free
ubiquitin molecules dimerize noncovalently in solution.
To further characterize such a noncovalent dimer, we used
paramagnetic relaxation enhancement (PRE), an NMR
technique that affords long-range distance information for
lowly populated species.[10a,b]A paramagnetic probe was
attached to one of three engineered surface-exposed cys-
teines, K11C, K48C, and K63C, which are located at the
periphery of the dimer interface mapped by chemical shift
perturbation (Figure 1b). PRE NMR is extremely sensitive to
the lowly populated species,[11a–c]in this case, the transiently
formed noncovalent dimer of free ubiquitin. For a rapidly
exchanging system, the observed PRE
population-weighted average of PRE rates experienced in all
conformational states.Providing that the paramagnetic
center and the nuclei under investigation are closer to each
other in the minor state, the transient species can be
manifested through its disproportionally large contributions
to the observed PRE. Applying the PRE technique, a number
of weak protein oligomers have been previously character-
Thus we measured the intermolecular PRE rates for the
backbone amide protons of
mixed with equimolar unlabeled ubiquitin conjugated with a
maleimide-EDTA-Mn2+probe (Figure S3). The measured
PRE rates arise exclusively from the paramagnetically tagged
protein to the isotopically labeled protein, as a result of an15N
isotope filter in the NMR pulse sequence.With 100 mm
NaCl in the solution, nonspecific protein–protein interactions
would be largely suppressed.[16a,b]Owing to the flexibility of
the paramagnetic tag, the paramagnetic center samples a
rather large conformational space. Yet, the separations
between conjugation sites (K11C, K48C, and K63) are much
larger than the variations of the paramagnetic center (Fig-
ure S3), hence the intermolecular PREs from the three sites
afford complementary observations. Interestingly, the PRE
profiles for K11C, K48C, and K63C sites appear quite similar,
with residues 12–14, 46–49, 71–76 displaying PRE G2rates
greater than 20 s?1(Figure 3). These residues are located at
the periphery of the dimer interface mapped by chemical shift
perturbation (Figure 3 insets), indicating the two types of
NMR data are consistent with one other.
15N-labeled wild-type ubiquitin
To obtain an ensemble structure of the ubiquitin dimer,
we performed rigid-body simulated annealing calculations by
refining against the intermolecular PRE rates. The rigid body
includes only residues 1–71, as the C-terminal residues 72–76
are flexible at ps–ns timescale.At the protein concentration
employed for PRE experiments (1 mm total), 12% of hetero-
dimer was expected, which was used as a scaling factor for
PRE back-calculation. The target function comprises the
Figure 2. Sedimentation equilibrium analysis of ubiquitin: a) on a
0.8 mm sample with a rotor speed of 20000 rpm, and b) on a 0.4 mm
sample with a rotor speed of 32000 rpm. The curves can be globally
fitted to a monomer–dimer equilibrium model, with
KD=(4.4?1.5) mm. The experimental data are shown as red circles,
the fitted curves as black lines. The theoretical sedimentation curves
expected for pure ubiquitin monomers are shown as blue lines.
Figure 3. Intermolecular1H transverse PRE G2 rates measured on an
equimolar mixture (0.5 mm each) of15N-labeled wild-type ubiquitin
and unlabeled ubiquitin mutant (K11C, K48C, or K63C) conjugated
with a maleimide-EDTA-Mn2+probe. Back-calculated PRE rates for
residues 1–71 are shown as blue lines. Insets: residues with observed
PRE values >20 s?1are colored purple on protein surface. Error bars
represent one standard deviation.
? 2012 Wiley-VCH Verlag GmbH & Co. KGaA, WeinheimAngew. Chem. Int. Ed. 2012, 51, 469–472
intermolecular PRE restraints for all three tagging sites, van
der Waals repulsive term and a weak radius-of-gyration term
applied to the entire dimer and to the dimer interface mapped
by chemical shift perturbation.While keeping one subunit
fixed, the other subunit is allowed to rotate and translate.
Good agreement between observed and calculated PRE
values is only achieved with multiple-conformer representa-
tion for the ubiquitin dimer. The overall PRE Q-factor is
above 0.5 with three-conformer representation and levels off
with 10 or more conformers (Figure 4a). Using a 10-confor-
mer representation, the PRE Q-factors are 0.173, 0.191, 0.160,
and 0.176 for all three sites, K11C, K48C, and K63C,
respectively (Figure 4b), and the PRE profiles can be
mostly reproduced (Figure 3a–c).
Upon forming a noncovalent dimer, the two ubiquitin
molecules bury on average a solvent-accessible area of
(601.8?336.5) ?, which encompasses residues 4–12, 42–51,
and 62–71, all located at the b-sheet region of the protein
(Figure 4c). The two subunits in the dimer are not confined to
a single configuration but rather adopt an array of relative
orientations (Figure 4d and S4). Ensemble distributions
comprising a wide range of conformers have also been
observed for other weak oligomers,[14a,b]which may be
inherent to all weakly interacting systems.Importantly,
the ensemble structure of ubiquitin dimers can be cross-
validated against the intermolecular PRE data measured
using a different paramagnetic probe conjugated at the same
three sites (Figure S5a-b). Optimizing only the positions of
nitroxide spin radical probes while fixing the dimer structure
obtained from EDTA–Mn2+PRE data, the back-calculated
PRE values agree well with the observed values using the
second paramagnetic probe (Figure S5c-e).
To the best of our knowledge, this is the first report
showing the noncovalent dimerization of free ubiquitin; all
previous work was focused on covalently linked poly(ubi-
quitin) or ubiquitin complexes with various UBDs. Non-
covalent dimerization of free ubiquitin has several implica-
tions. First, a covalent linkage would promote noncovalent
interactions. Although the concentration of free ubiquitin is
only approximately 10 mm in mammalian cells,[1a]the protein
mostly exists in covalently linked forms, affording a much
higher effective concentration. For Lys48-linked di-ubiquitin,
given that the linker consists of the C-terminal tail of the
proximal unit (residues 72–76) and Lys48 side chain of the
distal unit, the effective protein concentration for noncova-
lent dimerization is as high as (83.7?36.7) mm (Figure S6).
As extrapolated from the dimerization equilibrium for the
free ubiquitin, (83.9?1.9)% of the di-ubiquitin should be
committed to closed conformation with the b-sheet region in
contact, which is consistent with earlier experimental
Secondly, a covalent linkage may restrict the relative
movement for the two adjacent subunits in a poly(ubiquitin),
and select a subset of conformers from the ensemble structure
of the non-covalent dimer. Indeed, the crystal structure of
Lys48-linked di-ubiquitin falls into the boundary of the
atomic probability map delineated by the noncovalent
dimer (Figure 5a). The noncovalent dimer, however, encom-
passes more interfacial residues than the Lys-48 linked
Figure 4. Ensemble structure of the noncovalent ubiquitin dimer;
a) the PRE Q factors for all three tagging sites as a function of the
number of conformers in the ensemble; b) correlations between
observed and calculated PRE values, error bars represent one standard
deviation; c) decrease in the relative solvent exposure upon dimeriza-
tion; residues that are completely buried in the free form (solvent
accessible area <10 ?) are denoted with a gray bar; d) reweighed
atomic probability map plotted at 20% threshold for the distribution
of the15N-labeled ubiquitin (gray meshes) relative to the unlabeled,
paramagnetically tagged ubiquitin (purple surface). The two perspec-
tives are related by an 1808 8 rotation. The dimer interface is colored in
red, encompassing residues 4–12, 42–51, and 62–71.
Figure 5. Comparison between the crystal structure of Lys48-linked di-
ubiquitin and the ensemble structure of ubiquitin noncovalent dimer;
a) the proximal unit (purple surface) of the di-ubiquitin crystal struc-
ture[3a]is superimposed to one subunit in the ensemble structure of
the noncovalent dimer; the distal unit is shown as blue cartoon. The
noncovalent dimer is represented the same way as in Figure 3;
b) colored in orange, a large portion of the noncovalent dimer interface
becomes exposed in Lys48-linked di-ubiquitin. The covalent dimer
interface is colored in red.
Angew. Chem. Int. Ed. 2012, 51, 469–472? 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
covalent dimer—residues 4–7, 10–12, and 62–66 that are part
of the dimer interface in the former become solvent-exposed
in the latter (Figure 5b). Interestingly, these residues are also
involved in interactions with certain UBDs[20a–d]and may
permit the initial latching in respective binding processes.
Even though noncovalent ubiquitin dimerization appears
compatible with every covalent linkage (Figure S4), it has
been proposed that di-ubiquitins with Lys29, Lys33, Lys63 and
head-to-tail linkages predominantly exist in open conforma-
tion with little inter-subunit contacts.A possible explan-
ation is that steric hindrance around the amine group in the
distal unit prevents the two linked subunits from coming
together. Notwithstanding, it is likely that di-ubiquitin can
fluctuate between open and closed conformations, albeit with
different relative proportions for different linkage (Fig-
ure 6a). When in closed state, two adjacent subunits would
still adopt multiple relative orientations (Figure 6b), but
more restricted than the noncovalent dimer. Upon binding to
a specific UBD, each ubiquitin subunit is molded to a
particular tertiary structure, either by an induced-fit or
conformational selection mechanism. As such, fluctuations
among the ensemble members of ubiquitin dimer represent
dynamics at the quaternary structure level that complements
protein dynamics at the tertiary level.
Ubiquitin has been a favorite model system for NMR and
biophysics method development.[23a–c]Yet, the protein has
always been assumed monomeric in solution. As the last
implication but not the least, the presence of a noncovalent
dimer for free ubiquitin, although minor, needs to be taken
into account, especially when quantitating small differences.
Received: September 1, 2011
Revised: October 21, 2011
Published online: November 23, 2011
paramagnetic relaxation enhancement ·
protein–protein interactions · ubiquitin
Keywords: dimerization · NMR spectroscopy ·
 a) L. Hicke, H. L. Schubert, C. P. Hill, Nat. Rev. Mol. Cell. Biol.
2005, 6, 610–621; b) I. Dikic, S. Wakatsuki, K. J. Walters, Nat.
Rev. Mol. Cell. Biol. 2009, 10, 659–671.
 D. Komander, Biochem. Soc. Trans. 2009, 37, 937–953.
 a) W. J. Cook, L. C. Jeffrey, M. Carson, Z. Chen, C. M. Pickart,
J. Biol. Chem. 1992, 267, 16467–16471; b) C. L. Phillips, J.
Thrower, C. M. Pickart, C. P. Hill, Acta Crystallogr. Sect. D 2001,
 W. J. Cook, L. C. Jeffrey, E. Kasperek, C. M. Pickart, J. Mol.
Biol. 1994, 236, 601–609.
 a) D. Komander, F. Reyes-Turcu, J. D. Licchesi, P. Odenwaelder,
K. D. Wilkinson, D. Barford, EMBO Rep. 2009, 10, 466–473;
b) S. D. Weeks, K. C. Grasty, L. Hernandez-Cuebas, P. J. Loll,
Proteins Struct. Funct. Genet. 2009, 77, 753–759.
 A. B. Datta, G. L. Hura, C. Wolberger, J. Mol. Biol. 2009, 392,
 A. Bremm, S. M. Freund, D. Komander, Nat. Struct. Mol. Biol.
2010, 17, 939–947.
 M. L. Matsumoto, K. E. Wickliffe, K. C. Dong, C. Yu, I.
Bosanac, D. Bustos, L. Phu, D. S. Kirkpatrick, S. G. Hymowitz,
M. Rape, R. F. Kelley, V. M. Dixit, Mol. Cell 2010, 39, 477–484.
 R. Varadan, O. Walker, C. Pickart, D. Fushman, J. Mol. Biol.
2002, 324, 637–647.
 a) G. M. Clore, C. Tang, J. Iwahara, Curr. Opin. Struct. Biol.
2007, 17, 603–616; b) G. M. Clore, J. Iwahara, Chem. Rev. 2009,
 a) J. Iwahara, G. M. Clore, Nature 2006, 440, 1227–1230; b) C.
Tang, C. D. Schwieters, G. M. Clore, Nature 2007, 449, 1078–
1082; c) P. H. Keizers, M. Ubbink, Prog. Nucl. Magn. Reson.
Spectrosc. 2011, 58, 88–96.
 D. Yu, A. N. Volkov, C. Tang, J. Am. Chem. Soc. 2009, 131,
 C. Tang, J. Iwahara, G. M. Clore, Nature 2006, 444, 383–386.
 a) C. Tang, J. M. Louis, A. Aniana, J. Y. Suh, G. M. Clore, Nature
2008, 455, 693–696; b) C. Tang, R. Ghirlando, G. M. Clore,
J. Am. Chem. Soc. 2008, 130, 4048–4056.
 J. Iwahara, C. Tang, G. M. Clore, J. Magn. Reson. 2007, 184, 185–
 a) J. Y. Suh, C. Tang, G. M. Clore, J. Am. Chem. Soc. 2007, 129,
12954–12955; b) N. L. Fawzi, M. Doucleff, J. Y. Suh, G. M.
Clore, Proc. Natl. Acad. Sci. USA 2010, 107, 1379–1384.
 O. F. Lange, N. A. Lakomek, C. Fares, G. F. Schroder, K. F.
Walter, S. Becker, J. Meiler, H. Grubmuller, C. Griesinger, B. L.
de Groot, Science 2008, 320, 1471–1475.
 C. Tang, G. M. Clore, J. Biomol. NMR 2006, 36, 37–44.
 M. Ubbink, FEBS Lett. 2009, 583, 1060–1066.
 a) M. Hu, P. Li, L. Song, P. D. Jeffrey, T. A. Chenova, K. D.
Wilkinson, R. E. Cohen, Y. Shi, EMBO J. 2005, 24, 3747–3756;
b) S. Lee, Y. C. Tsai, R. Mattera, W. J. Smith, M. S. Kostelansky,
A. M. Weissman, J. S. Bonifacino, J. H. Hurley, Nat. Struct. Mol.
Biol. 2006, 13, 264–271; c) M. Renatus, S. G. Parrado, A.
D?Arcy, U. Eidhoff, B. Gerhartz, U. Hassiepen, B. Pierrat, R.
Riedl, D. Vinzenz, S. Worpenberg, M. Kroemer, Structure 2006,
14, 1293–1302; d) P. Peschard, G. Kozlov, T. Lin, I. A. Mirza,
A. M. Berghuis, S. Lipkowitz, M. Park, K. Gehring, Mol. Cell
2007, 27, 474–485.
 D. Fushman, O. Walker, J. Mol. Biol. 2010, 395, 803–814.
 a) T. Wlodarski, B. Zagrovic, Proc. Natl. Acad. Sci. USA 2009,
106, 19346–19351; b) D. Long, R. Bruschweiler, PLoS Comput.
Biol. 2011, 7, e1002035.
 a) N. Tjandra, A. Bax, Science 1997, 278, 1111–1114; b) K.
Lindorff-Larsen, R. B. Best, M. A. Depristo, C. M. Dobson, M.
Vendruscolo, Nature 2005, 433, 128–132; c) N. V. Nucci, M. S.
Pometun, A. J. Wand, Nat. Struct. Mol. Biol. 2011, 18, 245–249.
Figure 6. a) Scheme proposed for quaternary dynamics of di- or
poly(ubiquitin); a) with an isopeptide linkage, the two adjacent ubiq-
uitin subunits can either adopt open or closed conformations; b) when
in a closed state, the two subunits can fluctuate among various
? 2012 Wiley-VCH Verlag GmbH & Co. KGaA, WeinheimAngew. Chem. Int. Ed. 2012, 51, 469–472