Specificity of the interaction between UBA domains and ubiquitin
Thomas D. Mueller#, Mariusz Kamionka and Juli Feigon*Department of Chemistry and Biochemistry
405 Hilgard Avenue, P.O. Box 951569
University of California
Los Angeles, CA 90095-1569
*corresponding author; email: [email protected]; phone 310 206 6922; fax 310 825 0982
# present address:
Dept. Physiological Chemistry II
Biocenter, University Wuerzburg
Am Hubland
D-97074 Wuerzburg
Germany
Phone: +49 931 888 4170
Fax: +49 931 888 4113
Email: [email protected]
Running title: Interaction between HHR23A UBA domains and ubiquitin
Submitted 11/24/03; Revised 12/30/03
JBC Papers in Press. Published on January 5, 2004 as Manuscript M312865200
Copyright 2004 by The American Society for Biochemistry and Molecular Biology, Inc.
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
Summary
UBA (ubiquitin associated) domains are found in a large number of proteins with diverse
functions involved in ubiquitination, DNA repair, and signaling pathways. Recent studies have
shown that several UBA domain proteins interact with ubiquitin (Ub), specifically p62, the
phosphotyrosine independent ligand of the SH2 domain of p56lck, HHR23A, a human
nucleotide excision repair protein, and DDI1, another damage inducible protein. NMR chemical
shift mapping reveals that Ub binds specifically but weakly to a conserved hydrophobic epitope
on HHR23A UBA(1) and UBA(2) and that the UBA domains bind on the hydrophobic patch on
the surface of the five-stranded β-sheet of Ub. Models of the UBA(1)-Ub and UBA(2)-Ub
complexes obtained from de novo docking reveal different orientations of the UBA domains on
the Ub surface compared to those obtained by homology modeling with the related CUE
domains, which also bind Ub. Our results suggest that UBA domains may interact with Ub as
well as other proteins in more than one way while utilizing the same binding surface.
Keywords: HHR23A, Rad23, NMR, Ubl, proteasome, structure
2
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
Introduction
The lifespan of proteins inside and outside a cell is tightly regulated by the ubiquitin-proteasome
system, and numerous studies show that protein degradation is tightly interlocked with cell cycle
progression and is therefore an integral part of transduction pathways and other cellular
processes (1-4). For protein degradation by the Ub/proteasome system, the target proteins need
to be tagged with a poly-Ub chain. These covalent complexes are then recognized and degraded
by the 26S proteasome (1,2). The principle mechanism of this covalent modification has been
identified; an enzyme cascade known as E1-E2-E3 is responsible for activation and transfer of
Ub onto the target protein in a linkage specific manner (1,5).
The 26S proteasome is formed by a 20S cylindrical proteolytically active subunit and two
19S regulatory subunits (1,2,6,7). The 19S particles represent the lid of the proteasome and
regulate the access to the proteolysis (8). Although the polyubiquitinated substrate seems to be
recognized by the S5a subunit in the 19S particle (9-11), additional contacts between poly-Ub
chains and parts of the 19S regulatory subunit have been identified (12). Deletion studies indicate
that other polyubiquitin binding sites must exist (13).
Monoubiquitination is not sufficient for targeting proteins to the proteasome, however;
assembly of a poly-Ub chain of at least four Ub moieties is required to create a degradation
signal (11,14,15). Although Ub contains seven lysine residues, they are not used with the same
frequency in poly-Ub chain assembly. The predominant linkages observed are Lys 48-Gly 76
(16), Lys 29-Gly 76 (17) and Lys 63-Gly 76 (18), of which Lys 48-linked chains appear to be
the most frequent degradation signal. Poly-Ub assembly via Lys 29 and Lys 63 is less common,
and chain formation via Lys 63 seems to be involved in non-degradation signal events, e.g.
DNA repair (18).
Although key steps of Ub activation and transfer to a substrate as well as the structure of
the 20S subunit of the proteasome are known, the question of how proteins are targeted to the
proteasome remains unanswered. It is not known if there is an additional mechanism to regulate
the time point of degradation. One possibility is that monoubiquitination leads to a "point of no
3
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
return", which proceeds to substrate destruction in a defined time span. Recently, several groups
reported that proteins containing a UBA motif can bind directly to Ub and/or poly-Ub, leading
to an inhibition of the degradation of target substrates through the proteasome (19-21). UBA
domains are a common motif in a variety of protein families involved in protein degradation, cell
cycle control, or DNA repair. In vitro and in vivo assays have revealed that UBA domains of the
DNA-damage inducible proteins, RAD23 (as well as the fission yeast homologue Rhp23) and
DDI1, as well as proteins with no function in DNA repair, p62 and Mud1p, interact specifically
with Ub. Furthermore, poly-Ub chain formation is inhibited by RAD23 in vitro in a
concentration dependent manner (20,21). In addition, it has been shown that poly-Ub chain
extension stops at a length of three Ub moieties and that the inhibition of chain extension of
RAD23 is specific for Lys48-linked chains (22). As a consequence of the UBA-Ub interaction,
Clarke and co-workers showed that RAD23 and DDI1 are involved in the checkpoint control of
the cell cycle (23). Binding of the UBA domains of RAD23 and DDI1 to the nascent poly-Ub
chain of the Pds1 substrate leads to inhibition of chain extension, which subsequently results in
an increased lifetime of Pds1, which would otherwise be rapidly degraded.
In this report we provide a structural basis for the interaction of UBA domains with
monomeric Ub based on an NMR chemical shift mapping study as well as Ub mutagenesis. Ub
binds specifically to both UBA(1) and UBA(2) of HHR23A. The binding interface for both UBA
domains is almost identical despite the low overall sequence similarity. Both UBA domains bind
to the same region of monomeric Ub, which is also involved in the binding to the proteasome
subunit S5a. Models for the UBA(1)-Ub and UBA(2)-Ub complexes were generated from the
chemical shift mapping data by de novo docking as well as by homology modeling with the
solution structure of the closely related CUE domain in complex with Ub (24). These models
revealed very different orientations of the UBA domains on the surface of Ub. Our results
suggest that UBA domains may interact with Ub as well as other proteins, e.g. the HHR23A
binding proteins HIV-1 Vpr (25), methyladenine DNA glycosylase (MPG) (26), p300/cyclic
AMP responsive element binding (CREB)-protein (27), and peptide:N-glycanase (Png1) (28),
4
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
in more than one way while utilizing the same binding surface.
Experimental proceduresPreparation of proteins
UBA(1) and UBA(2) of HHR23A were prepared as previously described (29,30). 15N-
and 15N-,13C-labeled proteins were prepared by growing cells on M9 minimal medium using
15NH4Cl and 13C6-glucose as the sole nitrogen and carbon sources. For the titration of the UBA
domain proteins bovine Ub (amino acid sequence is identical to human Ub) was purchased from
Sigma and purified by gel filtration. Purity was subsequently checked by SDS-PAGE and
analytical reversed phase HPLC. For the NMR chemical shift mapping of Ub, yeast Ub was
prepared from the expression plasmid pET11c-Ub (gift from A. Varshavsky). The protein was
purified by reversed phase HPLC. Alanine mutants of yeast Ub were generated by site directed
mutagenesis using QuikChange (Stratagene). The cDNA sequence was verified by sequencing
using the DyeDeoxy-Terminator method (PerkinElmer). Ub mutant proteins were prepared
similar to wild type Ub. Protein concentrations were determined using an extinction coefficient
at A280 of 1280 M-1cm-1 for Ub and UBA(2) and 5120 M-1cm-1 for UBA(1), based on their
respective amino acid compositions.
Chemical shift mapping experiments
Proteins for titration experiments were dialyzed against identical buffer (50mM sodium
phosphate pH 6.5, 100mM sodium chloride, 2mM DTT-d10 (Cambridge Isotopes Labeling CIL)
to avoid chemical shift changes due to differences in buffer conditions. NMR samples for
titration studies contained 0.25 to 0.5 mM 15N-labeled protein. Unlabeled protein was added
stepwise up to a final ratio of 1:10. Unlabeled Ub was concentrated as far as possible to
minimize volume changes throughout the titration. The “reverse” titration of Ub with UBA
proteins was performed similar to the experiment described above. A similar mapping study
using 15N-labeled UBA(1) or UBA(2) and the Ubl domain of HHR23A was performed in order
to test the specificity of the UBA domains for Ub. In addition, the UBA domains were tested for
homo- and heterodimerization with NMR mapping experiments. 15N-labeled UBA(1) (0.7mM)
5
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
was mixed with unlabeled UBA(2). All measurements were performed at 27° C using a Bruker
DRX-500 or DRX-600. 2D 1H-15N HSQC experiments with watergate water suppression and
water flip-back pulses were used for monitoring the chemical shift changes. All titration data
sets were processed identical using the software XWINNMR (Bruker). 1H dimensions were
referenced using external TSP. 15N nitrogen and 13C carbon frequencies were referenced using
the gyromagnetic ratios 15N/1H = 0.101329118 and 13C/1H = 0.251449530. The changes of
proton and nitrogen chemical shifts were averaged based on their gyromagnetic ratios using the
following equation: ∆δave. = 0.5x [(|∆δ(1H)|+(0.125x|∆δ(15N)|)].
The chemical shifts of yeast Ub were reassigned (chemical shifts published for human Ub
differ from yeast Ub due to three amino acid changes) using 15N,13C-labeled yeast Ub. A set of
triple-resonance experiments (31) (CBCA(CO)NH, CBCANH, HBHA(CO)NH, H(C)(CO)NH-
TOCSY, CC(CO)NH-TOCSY, HCCH-COSY) was acquired to assign all backbone and side-
chain chemical shifts for yeast Ub under the conditions used for the NMR chemical shift
perturbation studies.
Homology modeling of the UBA-ubiquitin complexes
Model complexes of the UBA(1)/UBA(2) with Ub were built using the structure of the
CUE domain of Cue2 and Ub as a template (24). Initial structures of the complexes were
obtained by fitting the Cα-atoms of the residues in helices 1 and 3 of the CUE and UBA
domains. Close contacts between atoms in the interface were removed by manual remodeling
using the software Quanta98. The structures were then subsequently refined by energy
minimization and short (20 ps) molecular dynamics simulation at 300K (Charmm force field,
Quanta98) using only geometrical energy terms to maintain bond length and angles as well as
van der Waals packing.
De novo docking for the UBA-ubiquitin interaction
The program HADDOCK (32) was applied to dock Ub and the UBA domains of
HHR23A. In this approach, data obtained from chemical shift mapping is transformed into a set
of ambiguous distance restraints that are used together with geometrical and electrostatic
6
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
complementarity to dock the two molecules. Chemical shift mapping of UBA(1), UBA(2), and
Ub and surface accessibility data calculated with the program NACCESS (33) were used to
define ambiguous distance restraints as described. The target distance of these restraints was set
to 2.0 Å; force constants for empirical and experimental restraints were used as suggested by the
default settings. Initially 750 structures for a Ub-UBA complex were generated by docking Ub
and UBA(1)/UBA(2) as rigid bodies using only ambiguous distance restraints, van der Waals
energy and electrostatic energy terms using the program suite ARIA1.2 and CNS (34,35). Of
those, 200 structures with the lowest overall energy were subsequently refined allowing the side
chain conformations of residues within the binding interface to be flexible. Finally, 100 refined
structures with lowest overall energy were then refined using a short molecular dynamics
simulation in explicit solvent to allow for a correct implementation of the electrostatic
contribution. The method was applied to two closely related complexes for which structures were
available, the UIM-2 S5a-HHR23A Ubl and the Cue2 CUE domain-Ub complex, to test for the
reliability of that procedure. For the simulation of the Cue2 CUE domain-Ub interaction, the
distance restraint set was defined based on the residues buried in the complex structure; for the
simulation of the HHR23A Ubl-UIM-2 S5a interaction the same criteria for chemical shift
changes and residue accessibility as used for the Ub-UBA domain modeling were applied for
generating the distance restraint set. The test calculations yielded model structures that were
within 1.5 Å rmsd of the experimentally determined structures. Different cutoff values for the
surface accessibility in the definition of the distance restraint set used for the computational
docking did not lead to altered complex architectures, confirming the computational results are
stable. The influence of the template structures of the individual complex components was
determined by using the Ub structure of the Cue2 CUE-Ub complex for docking to the UBA
domains as well as the structure of free Ub (PDB entry code 1UBI). Again no change in complex
architecture could be observed.
Results
Ubiquitin binds specifically to HHR23A UBA(1) and UBA(2) domains
7
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
High-resolution structures of both UBA domains of HHR23A have been determined,
which show that overall the three-helical bundles are largely identical, in spite of a low (~20%)
sequence identity (29,30,36). The main structural differences are slightly different packing of
side chains in the hydrophobic core and the conformation of the N- and C-termini. Both
domains have an unusually large hydrophobic surface patch which was predicted to be a protein-
protein interface (29,30). Based on the structures and sequence analysis of UBA domains, we
predicted that a conserved portion of this hydrophobic patch would be the binding region for Ub
(29).
In order to map the binding site of Ub on UBA domains, NMR chemical shift
perturbation experiments were performed on both UBA(1) and UBA(2) from HHR23A. Samples
of 15N-labeled UBA(1) and UBA(2) were mixed with bovine Ub, and 2D HSQC spectra were
acquired to monitor the changes in the chemical shifts of the backbone amides induced by the
binding to Ub (Fig. 1 A, B). Changes in chemical shifts are observed for several residues in both
UBA(1) and UBA(2) starting at about 0.6 molar equivalents, indicating the formation of specific
complexes in fast exchange on both the 500 and 600 MHz NMR time scale. Assuming a binary
interaction between the UBA domains and Ub, a KD in the range of 500 to 600 µM can be
determined from non-linear fitting of the titrations with either UBA(1) or UBA(2) (see also Fig.
3A, B). However, a binding constant of about 10 µM was reported for full-length Rad23 (19),
which may indicate cooperative binding of both UBA domains (37). Plots of the chemical shift
changes versus residue (Fig. 1 C, D) show that the absolute magnitude of the chemical shift
changes are similar for UBA(1) and UBA(2).
A recent chemical shift mapping study of the closely related HHR23B reported that the
Ubl and UBA domains interact with each other with a reported KD of ~2 mM, which is ~10-fold
higher than their calculated KD for the HHR23B-UBA interactions of ~300 µM (38). The Ubl
domain is structurally very similar to Ub (39,40) and exhibits a high sequence identity. Weak
binding between the HHR23A Ubl and UBA domains was also detected in the context of the full
length protein as well as isolated domains (41) under slightly different buffer conditions and at
8
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
higher field strength, with a binding affinity at least 10-fold lower compared to the already weak
interaction between the UBA domains and Ub.
Ubiquitin binds to a conserved surface epitope of UBA domains
The results of the chemical shift perturbation study were mapped onto the structures of
UBA(1) and UBA(2) (Fig. 1). Although UBA(1) and UBA(2) have a relatively low sequence
identity, the binding interfaces revealed by the chemical shift mapping are remarkably similar.
Residues exhibiting the largest chemical shift changes upon binding to Ub, G174 and H192 for
UBA(1) and L330, G331 and E348 for UBA(2), respectively, are in similar locations. A large
cluster of residues in UBA(1) involved in binding is located at the C-terminal end of the first
helix, I170, M171, S172 and the first three residues in the short and highly conserved loop 1,
M173, G174 and Y175 (Fig. 1 A, C, E). Several positions on the third helix of UBA(1) also
show large changes upon binding, specifically H192 and R193 at the N-terminus of helix 3 and
Y197 at the second turn of helix 3. Together, these residues form a consecutive patch of about
520 Å on the surface of UBA(1) (Fig. 1 E). The residues with the largest chemical shift changes,
H192, M173, and G174, are in the center of the epitope.
Residue Y197 also exhibits a significant change in chemical shift upon addition of Ub,
but is located on the “back face” of the binding interface. The change for the amide proton and
nitrogen frequencies of Y197 might be attributed to small structural changes in the hydrophobic
core due to interactions of the side chain of Y197 with R193, which is part of the binding
interface. Residues in the C-terminus of UBA(1) do not exhibit chemical shift changes when Ub
is added (Fig. 1 C) and do not seem to be part of the binding interface. This was surprising, since
the C-terminus is close to the hydrophobic patch involved in binding and is of relatively rigid
nature (29).
Analysis of the binding of UBA(2) to Ub reveals that the general location of the epitope
remains the same compared to UBA(1) (Fig. 1). Identical positions in the C-terminal end of
helix 1 (R326, L327 and A329) as well as the first three residues of the hydrophobic loop 1
(L330, G331 and F332) form a large part of the binding interface to Ub. In addition, residues at
9
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
similar positions at the N-terminus of helix 3 (E348 and N349) are also among the residues with
the highest chemical shift changes upon binding. The structurally equivalent residue of E348 in
UBA(1) is P191, therefore no information about whether or not P191 is involved in the binding
of Ub can be deduced from this analysis. A larger number of residues at the C-terminal end of
helix 3 of UBA(2) are involved in binding than observed for UBA(1). On the last turn of helix 3
residue L356 is affected by the binding, while residues of UBA(1), L199 and T200, do not
exhibit significant changes. Overall, with a binding area of approximately 650 Å2, UBA(2)
appears to have about a 25% larger interface than UBA(1), for which the residues within the
binding epitope yield an area of about 520 Å2.
UBA domains bind on the 5-stranded β-sheet surface of ubiquitin
In order to investigate the binding surface for UBA domains on Ub, NMR chemical shift
mapping was performed under the identical conditions as reported above using yeast Ub. Yeast
Ub differs in three amino acids compared to human Ub, S21P, D24E, and S28A, none of which
is close to the binding site for UBA domains determined in this study. The residue P19 in human
Ub is located in a tight turn in the loop between the second β-strand and the first α-helix. The
two other residues D24 and S28 are located on the α↑helix, facing in the opposite direction of
the determined binding interface. We therefore concluded that these mutations do not interfere
with or modulate the binding of the UBA domains to Ub. However, to confirm that the
recognition process is not influenced by indirect effects we repeated the titration using 15N-
labeled human Ub purchased from VLI research (Mavern) and UBA(2); no differences in the
chemical shift changes compared to the study using yeast Ub were detected (data not shown).
Chemical shift mapping of the amide resonances of Ub as a function of added UBA(1) or
UBA(2) up to a ratio of 1:10 Ub:UBA again revealed complexes in fast exchange on the NMR
timescale. Upon addition of UBA(1) to Ub significant changes in chemical shift are observed for
23 residues (Fig. 2 A, C, E). Most of these residues are located on the β-strands or the
connecting loops of the 5-stranded β-sheet of Ub, forming a consecutive patch. L71 to L73 are
located in the C-terminus close to the Gly-Gly motif required for poly-Ub chain extension (Fig.
10
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
2 E). When Ub was titrated with UBA(2), equivalent residues showing a significant change in
their proton and nitrogen amide frequencies were almost identical with those for UBA(1) (Fig.
2).
These results suggest that both UBA domains are recognized by and bind to Ub using a
similar binding epitope. The binding epitopes of UBA(1) and UBA(2) on Ub overlap perfectly,
and the differences in several of the amino acids, e.g. H192 versus E348 and R193 versus N349,
might account for the chemical shift differences observed for the titration of Ub with either
UBA(1) or UBA(2). However, another possibility is that the two UBA domains are oriented
differently on Ub, as discussed below.
Differential chemical shift mapping of UBA domains
To further address the question how the UBA domains of HHR23A bind to Ub,
especially whether the binding mechanism differs between the two UBA domains, we tried
differential chemical shift mapping (42). Five Ub mutants (L8A, R42A, K48A, H68A, and
R72A) with single amino acid substitutions were chosen for study. All of the mutated residues
have side chains oriented towards the binding interface and show significant chemical shift
changes upon binding to the UBA domains. Differences in the chemical shift changes between a
Ub mutant and wild type Ub should be observed if the residue pair is in close proximity in the
UBA-Ub complex. The binding affinities were determined by non-linear fitting of the chemical
shift changes of three residues located within the binding epitope (I170, G174, and H192 for
UBA(1); G331, E348, and L356 for UBA(2)).
Surprisingly, the mutant Ub L8A exhibits much smaller absolute chemical shift changes
upon binding to UBA(1) and UBA(2), suggesting a lower binding affinity. However, quantitative
analysis of the binding curves (Fig. 3A,B) shows that the affinity of UBA(1) and Ub L8A is not
changed significantly (490µM and 570µM, respectively). Similar small changes in KD (less than
2-fold) were observed for the other Ub mutants (K48A, H68A, R72A). Interestingly, the binding
affinity of the Ub mutant R42A for UBA(1) and UBA(2) seems to be increased 3.5- and 2-fold .
11
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
Comparison of the UBA-Ub titration studies with those of the Ub mutants showed no
significant differences in the course of the chemical shift changes for UBA(2). In contrast, for
UBA(1), the chemical shift pattern of Ub mutants L8A and R42A is different compared with
wild type Ub. For L8A significant changes are observed for residues E169 in helix 1 as well as
R193, A194, and E196 in helix 3 (Fig. 3C) indicating that these residues are in close proximity to
each other in the complex. For Ub mutant R42A, the largest differences are observed for residues
E169 (helix 1), and A194, E196, L199 and T200 located in helix 3 (Fig. 3D). Since the side
chains of L8 and R42 are separated by only 6 Å, similar residues can be affected by both Ub
mutants. Model building of a UBA(1)-Ub complex based on just the two identified residue pairs
is however not possible since two interaction points do not define the orientation of two rigid
bodies unambiguously.
These mutagenesis results are consistent with an analysis of the interaction of the CUE
domain of Cue2 protein using two Ub mutants K48A and H68A (24). Although both mutations
are located in the center of the interacting patch they did not influence the interaction
significantly. Apparently, the binding specificity between Ub and its interaction partners is
achieved by the overall surface topology, which explains why a single point mutation is not able
to destabilize the binding substantially.
Homology modeling of the UBA-Ub interaction
Recently, the structures of a UBA homolog the CUE domain in complex with Ub was
determined by NMR and X-ray crystallography (24,43). Based on the structure of the CUE
domain of Cue2 bound to Ub (24), we generated a model for the interaction of UBA(1) and
UBA(2) with Ub (Fig. 4A, B). The solution structure of the Cue2-Ub complex instead of the
crystal structure of the CUE domain of Vps9 bound to Ub (43) was used for model building due
to the higher similarity between the CUE domain structure of Cue2 and the UBA domains of
HHR23A. The crystal structure of the CUE domain of Vps9p forms a domain-swapped inter-
twined dimer and binding to Ub results in a large conformational change. Furthermore, the
binding affinity of the Cue2 CUE domain to Ub (KD ~150µM) is more similar to the interaction
12
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
of the UBA domains of HHR23A with Ub (KD ~500µM). In contrast the binding of the Vps9p
CUE domain is considerably tighter (KD ~ 20µM). The rmsd for Cα positions of the three
helices of UBA and CUE are 1.6 Å (CUE - UBA(2), 29 Cα positions) and 1.9 Å (CUE -
UBA(1), 26 Cα positions). In the final models 410 Å2 (UBA(1)-Ub) and 440 Å2 (UBA(2)-Ub)
surface area of the UBA domains are buried in the interface. This is in good agreement with the
interface size determined by chemical shift mapping. The nature of the interface is almost
exclusively hydrophobic (>80%).
In the UBA-Ub models the axis of helix 1 is oriented in an angle of 45° to the β-strand 5
of Ub, helix 3 is running at an angle of 50° across the β5, the helical axis of helix 2 is running
almost in parallel with β5 (Fig. 4A). Hydrophobic residues in loop 1 (G174-Y175 UBA(1) and
G331-F332 UBA(2)) of the UBA domain interact with the β3-β4 loop residues of Ub (Fig. 4A).
The N-terminus of helix 3 of the UBA domains (P191-H192 UBA(1) and E348-N349 UBA(2))
is close to the C-terminus of β-strand 5 and the β1-β2 loop of Ub. The important hydrophobic
triad of Ub, residues L8, I44, and V70, is buried in the interface. In the Ub-UBA(1) complex, L8
of Ub is surrounded by UBA(1) residues E169, I170 (helix 1), P191, and H192 (loop 2) (Fig.
4A). In the complex of Ub-UBA(2), residue L8 of Ub has van der Waals contacts to residues of
UBA(2) occupying similar positions in the three-helical bundle (A323, L327 (helix 1), and E348
(loop 2)) (see Fig. 4B). I44 of Ub is packed against M173 (helix 1), Y175 (loop 1), and V195
(helix 3) of UBA(1); in Ub-UBA(2) residues L330 (helix 1), A352, and L356 (helix 3) of
UBA(2) are close to I44 of Ub. V70 of Ub either contacts H192 (loop 2) of UBA(1), or residues
A323, L327 (helix 1) and E348 (loop2) of UBA(2).
The differences in the environment of Ub V70 bound to UBA(1) vs. UBA(2) are due to
slightly different interhelical angles of the three-helical bundle, which lead to helix 1 of UBA(1)
pointing further away from the binding interface. Although different residues of both UBA
domains interact with L8, I44, and V70 of Ub, the hydrophobic nature in the contact area is
preserved. Only a very small number of possible intermolecular hydrogen bonds can be
identified in the binding interfaces of the model complexes: 5 for Ub-UBA(1) (Ub-UBA: K6-
13
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
E169, G47-Y175, G47-M173, H68-E169, L71-H192) and 2 for Ub-UBA(2) (Ub-UBA: T7-
R326, H68-A329). Similarly, in the complex of Cue2 CUE domain and Ub only two hydrogen
bonds are observed. The dearth of polar interactions in addition to the relatively small interface
area (~500 Å2) might explain the relatively low binding affinity between the monomeric CUE
domain and Ub as well as for UBA domains and Ub.
De novo docking of the ubiquitin-UBA domain complexes
In addition to the homology modeling, we performed a de novo docking using the
program HADDOCK and employing data from our chemical shift mapping (32). The chemical
shift changes of UBA(1), UBA(2), and Ub were used together with surface accessibility data to
define ambiguous distance restraints between the two molecules. The method relies on
geometrical and electrostatic complementarity of the binding epitopes of the interacting
molecules, and it has been applied successfully to other complexes (32). Much to our surprise,
the results of the de novo docking of the Ub-UBA domain complexes revealed completely
different complex structures compared to those obtained by homology modeling (Fig. 4 and 5).
Additionally, the architecture of the complexes resulting from de novo docking for the Ub-
UBA(1) and Ub-UBA(2) interaction (Fig. 5 A,B) is also different, indicating that the UBA
domains of HHR23A might interact with Ub differently.
A structural alignment of the Ub molecules of the two models of UBA(2)-Ub (homology
model vs. de novo docking) shows the large difference. In the complex of UBA(2)-Ub obtained
by de novo docking, helices 1 and 3 of UBA(2) run almost parallel to β-strand 5 of Ub, and
UBA(2) helix 2 and Ub β-strand 5 are oriented at an angle of about 20° (Fig. 5B) relative to
each other. In the homology model (and hence for the template complex Cue2 CUE domain-Ub)
,helices 1 and 3 of the UBA domain and β-strand 5 of Ub share an angle of almost 45°; and the
helical axis of helix 2 of UBA(2) runs parallel to β-strand 5 of Ub (Fig. 4B). Consequently, the
de novo docking model can be transformed into the homology model by a rotation of about 45°
in counter clockwise direction. A large positional movement is observed for the residues in the
GFP-loop of UBA(2), with the Cα-positions of L330, G331, and F332 differing by 7 to 8 Å
14
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
between the two modeling approaches. The positional change of the residues in the loop 2 (E348
and N349) and helix 3 (L356) with respect to the Ub binding interface is smaller (distances for
Cα positions: E348 3 Å; N349 4 Å, and L356 4 Å). Consequently a different residue pairing is
observed in the interface of the de novo docking model. L8 of Ub is surrounded by R326, L327,
L330, and E348 of UBA(2), I44 of Ub is in van der Waals contacts with residues A352, N353,
and L356 of UBA(2) and V70 of Ub is in close proximity to N349 and A352 of UBA(2). In the
complex of the Cue2 CUE domain and Ub (and thus in the homology model of UBA(2)-Ub), the
residues located at the N-terminus of the helix 1 are part of the binding interface, which is not
the case for the de novo docking. In the homology model, A323 and R326 both contact residues
L8 and T9 of Ub, whereas for the de novo docking model, A323 shares no contacts with any
residues of Ub.
De novo docking of UBA(1) and Ub resulted in a model complex that differs from the
homology model to an even greater extent. The helical axes of helix 1, 2, and 3 of UBA(1) in the
two models differ by about 55°, 70°, and 100°, respectively. In order to transform both models
into each other a rotation of almost 90° is required. In the UBA(2)-Ub models the largest
positional discrepancies between the two different models are observed for the GFP-loop, while
for the UBA(1)-Ub models the major positional changes are for helix 3. Residue H192 is located
on top of the β1-β2 loop of Ub and contacts K6, L8, and G10, whereas in the homology model
H192 is positioned between the β-strands 5 and 3 and contacts residues R42, V70, and L71. This
clearly shows that the residue paring is quite different for these two models. In both models the
side chain of M173 of UBA(1) makes comparable contacts and is located in a hydrophobic
pocket formed by H68 and I44 of Ub. In the de novo docking model Y175 of UBA(1) is in close
proximity to R42 of Ub possibly making hydrogen bonds between the hydroxyl group of Y175
and the guanidinium group of R42. In contrast in the homology model Y175 is in van der Waals
contact with G47 of Ub with a possible hydrogen bond between the hydroxyl group Y175 and
the backbone carbonyl of G47 (Fig. 5B). Analysis of the differences between UBA(1)-Ub and
UBA(2)-Ub obtained by de novo docking suggests that the residue pairing might be quite
15
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
different although the residue positions in the UBA domain architecture involved in the binding
are conserved.
Since the results of the de novo docking for UBA(1)-Ub and UBA(2)-Ub deviate from
the models obtained by homology modeling, we tested whether the differences might be an
artifact of the docking procedure. Since all 100 refined structures of both complexes cluster into
one single structure family exhibiting an rmsd within the cluster of less than 0.8 Å, the observed
complex architecture obtained by the docking simulation seems to be very stable. Different sets
of distance restraints for the docking did not lead to altered complex architectures. We also tested
the docking for two other complexes, the Cue2 CUE domain Ub interaction (24) and the
complex of HHR23A Ubl domain and UIM-2 of S5a (39). For both simulations, structure data,
and for the latter, structure and chemical shift mapping data, were available. The docking of the
Cue2 CUE domain and Ub reproduced the experimental structure of Kang et al. (24) with an
rmsd of less than 1 Å for the Cα atoms and about 1.3 Å for all heavy atoms. In the docking
simulation of S5a UIM-2 and the Ubl domain of HHR23A, the final structures obtained had an
rmsd of 1.1 Å and 1.8 Å for the Cα-atoms and all heavy atoms, respectively, compared with the
experimental structure. These simulations confirm that the setup is able to reproduce structures
of two experimentally determined complexes that have similar binding mechanism and chemistry
(mainly hydrophobic interaction).
Discussion
Is the hydrophobic knob a general ubiquitin binding site for UBA domains?
The chemical shift mapping study presented here reveals a structural basis for the
recognition and binding of UBA domains of HHR23A to Ub. The hydrophobic surface patches
that were predicted to be the site of protein-protein interactions for the UBA domains (29)
comprise a large portion of the binding epitope for Ub. In addition, some charged and polar
residues appear to be important based on the chemical shift mapping. These results agree very
well with a mapping study of HHR23B UBA domains published very recently by Choi and
coworkers (38), except for the equivalent residue to HHR23A UBA(1) 197. Despite the
16
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
relatively low binding affinity of 500 µM, the interaction of the UBA domains with Ub is
specific. Furthermore, binding between HHR23A/B UBA and Ubl domains, is an order of
magnitude weaker (38,41).
The amino acid sequence identity of UBA domains is about 25%, which is too low to
propose a common biological function or even a common binding mechanism. However, if the
analysis is limited just to the binding interface determined in this study, the degree of similarity
of the binding epitope is as high as for the hydrophobic core of the UBA domains (Fig. 6). The
binding epitope of Ub on UBA(1) and UBA(2) is comprised of 9 and 12 residues, respectively.
The high degree of conservation for the residues located on the surface (similarity for binding
region >50%, similarity within hydrophobic core ~60%) suggests that these are involved in a
common binding interface (29). The chemical shift mapping study of UBA(1) and UBA(2)
presented here confirms the prediction that these residues in the hydrophobic patch are part of a
binding epitope to Ub. Despite a similar location on UBA(1) and UBA(2), however, we also note
differences in these epitopes. In UBA(2) the second leucine residue L356 of the highly conserved
double leucine motif clearly exhibits changes in its chemical shift upon binding to Ub, whereas
the same residue position in UBA(1), L199, is not affected. In addition, the number of residues
on helix 3 affected in the titration with Ub is different, with six residues of UBA(2) involved in
the binding epitope but only three residues of UBA(1).
The five-stranded β-sheet of ubiquitin is an universal binding site for ubiquitin-interacting
proteins
Despite the differences in the binding epitopes determined for the UBA domains of
HHR23A, the position of the interface on Ub is practically identical for both UBA domains. This
region has been identified for the interaction with several other protein domains indicating that
the hydrophobic surface patch on the five-stranded β-sheet of Ub is probably a general protein-
protein interface that can facilitate various interactions. At least six ubiquitin-interacting
domains have been described so far (44). The structures of five domains have been determined
and the site of their interaction with Ub has been mapped. The three-dimensional structures of
17
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
these domains, UEV (ubiquitin E2 enzyme variant) (45), NZF (novel zinc finger domain) (46),
UIM (ubiquitin-interacting motif) (39,47,48), UBA (29,30,36), and CUE (24,43), vary greatly in
architecture and size, being either a single helix (UIM), pure β-strand structure (NFZ), three
helical bundles (CUE, UBA) or mixed alpha-beta structure (UEV). Despite this large variety, all
domains interact with Ub via the same hydrophobic patch on the five-stranded β-sheet of Ub.
Comparing the binding to Ub for all domains reveals only very minor differences in the location
of the binding sites. The center of the binding site, the hydrophobic patch around residue I44 of
Ub, seems to be identical for all ubiquitin-interacting domains so far, although residues close to
that patch have been mapped to different biological functions (49). Very little structural data for
the interaction of monomeric Ub or ubiquitin-like domains in complex with an ubiquitin-
interacting domain is available so far (24,39,43), probably due to the low to moderate binding
affinities for such interactions (KD ~10 to 500µM) (44). A comparison of the binding mechanism
of a single α-helix with an Ub-like domain (complex of HHR23A Ubl-UIM-2 of S5a) with
that of a three-helical bundle bound to Ub (complexes of Cue2 CUE domain and Vps9p CUE
with Ub) shows that complexes of Ub and Ub-interacting domains can adopt different
architectures. The single α-helix of the UIM motif of S5a binds on top of β-strand 5 of
HHR23A Ubl, with the axes of the α-helix of the UIM motif and of β-strand 5 of Ub running
anti-parallel (39). In contrast, the three-helical bundle of the CUE domain of Cue2 binds via
helices 1 and 3, which run across β-strands 1, 3, and 5 at an angle of about 30° (24). In the case
of the CUE domain of Vps9p, the interaction is even more complex. In the X-ray crystal
structure of the complex of the CUE domain of Vps9p with Ub, the CUE domain forms a
domain-swapped dimer, in which helices 1 and 3 interact with Ub in a similar manner to that
observed for the Cue2 CUE domain-Ub complex, but here helix 2 has additional contacts with
residues of Ub (43). Despite the differences in the architecture of the complexes, the interacting
amino acids are conserved, with only hydrophobic amino acids taking part in the interaction. The
absence of hydrogen bonds in the center of the interface probably explains the limited specificity
of Ub, since there is no requirement to maintain the geometry of hydrogen bonding acceptors or
18
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
donors, and therefore specificity for a binding partner is only generated by geometrical
restrictions for the interacting hydrophobic side chains. It is also interesting to note, that the
binding affinity between Ub and Ub-interacting domains of known structures correlates with the
size of the interface. The Cue2 CUE domain Ub interface measures 450 Å2 and has a binding
affinity of about 150 µM (24), while the S5a UIM-2 HHR23A Ubl complex has a buried surface
area of roughly 600 Å2 (39) and a KD ~10 µM. The interface between the CUE domain of
Vps9p and Ub measures 520 Å2 (due to additional interacting residues in the second helix)
resulting in an affinity of 20 µM (43).
A model for the interaction of UBA domains and ubiquitin
One very surprising result of the modeling of the UBA-Ub interaction presented in this
study is the large differences between the models obtained by homology modeling and de novo
docking using the program HADDOCK. The homology models were built using the NMR
structure of the Cue2 CUE-ubiquitin complex, with the CUE domain being replaced by the UBA
domains of HHR23A. Residues of the Cue2 CUE domain that interact with residues of the five-
stranded β-sheet of Ub are either conserved or replaced by homologous amino acids (Fig. 6) in
the UBA domains of HHR23A. However, docking of UBA(2) to Ub using the program
HADDOCK resulted in a model with the three-helix bundle rotated clockwise by about 45°. For
the de novo docked complex involving UBA(1) the three-helix bundle is rotated by almost 90°
but in counterclockwise direction (Fig. 4 and 5). Thus, not only do the de novo docking results
differ from the homology modeling approach, but the de novo docking even suggests different
complex architectures for UBA(1) and UBA(2) bound to Ub. The differences in the complex of
the two CUE domains of Cue2 and Vps9p with Ub provide experimental support for the idea that
the interface of Ub allows for the binding of different helical bundle geometries. The CUE
domain of Vps9p, being a domain-swapped dimer, binds to Ub not only via helices 1 and 3, but
also via additional residues in helix 2 which likely contribute to affinity and specificity (43).
We note, in addition, that the differential chemical shift mapping employing Ub mutants
showed differences for UBA(1) and UBA(2). These might indicate that both UBA domains bind
19
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
to Ub by a different binding mechanism, supporting the results obtained by the de novo docking
procedure. Analysis of the distribution of charged residues surrounding the hydrophobic patch,
which is in the center of the binding interface for CUE and UBA domains, clearly shows that the
electrostatic potentials are distinct for these domains. Biological data suggest differences in
functions of the CUE and UBA domain (50-52), and sequence comparison clearly distinguishes
between the CUE and UBA family (53). All CUE domains identified so far bind to monomeric
Ub with low to moderate affinity (20 to 160 µM), and some but not all CUE domains seem to
bind to poly-Ub in addition (52). It was not reported whether their binding affinity for poly-Ub
is higher than for monomeric Ub; however data from Shih et al. (52) suggests that at least the
CUE domain of Vps9 has no preference for long poly-Ub chains over short poly-Ub chains.
This binding preference is probably required for the maintenance of monoubiquitination (52),
which is in turn a signal for trafficking and receptor endocytosis. However for a more
quantitative analysis the binding constants of several CUE domains for mono- and poly-Ub
have to be determined.
The biological function of the UBA domain is, on the other hand, still in debate
(20,22,23,37,54). Several groups have reported that UBA domains bind to monomeric Ub as well
as to poly-Ub, although the data is contradictory in some cases (19,54). Binding to monomeric
Ub was associated with inhibition of further extension of the nascent Ub chain which results in
an inhibition of the degradation of a substrate (20-23). The binding to poly-Ub was explained
with a possible shuttle function for the transport of a substrate to the proteasome (37,54-57).
Although the physiological “Ub-target” of the UBA domains of RAD23 and other proteins has
not been determined, all in vitro binding experiments have yielded an at least 1000-fold greater
affinity of the UBA domains for tetra-Ub compared to mono-Ub (22,54). Therefore it is very
likely that in the presence of polyubiquitinated substrates, the main binding partner of UBA
domains might be these poly-Ub chains rather than monoubiquitinated substrates or free Ub
unless the concentration of polyubiquitinated substrates is very low. The molecular nature of the
tighter binding of UBA to poly-Ub is not yet clear; however, structure analysis (58) as well as
20
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
mutagenesis data on poly-Ub (59) suggest that they do not form plain linear poly-protein chains
like pearls on a string, but rather adopt globular structures. Hence additional contacts between
the linked Ub moieties and possibly other epitope(s) of the UBA domain might lead to the
increase in affinity in comparison to monomeric Ub.
Since UBA domains are often associated together with Ub-like domains in modular
proteins, the UBA domains could act as poly-Ub “receptors” whereas the Ubl domain might
interact directly with the proteasome. However such a shuttle mechanism has not yet been
confirmed in vivo. The differences for UBA and CUE domains in their biological function as
well as in the probable binding target, monomeric Ub versus poly-Ub, make it plausible that the
binding mechanisms of CUE and UBA domains do not need to be identical.
Interestingly, our de novo model was recently at least partially confirmed by results by
Walters and co-workers (41). Based on chemical shift differences between the isolated Ubl and
UBA domains and the domains in the context of the full length HHR23A protein, it was
concluded that in full-length HHR23A the Ubl domain interacts in a dynamic fashion with the
individual UBA domains and this interaction has a 1:1 stoichiometry, i.e. Ubl is exchanging
between one or the other UBA domains. No interdomain NOEs were observed, consistent with
very weak binding. Using residual dipolar couplings measurements, Walters and coworkers tried
to define the relative orientation of the domains in the modular protein (41). Although the
coordinates for their models of the interaction between Ubl and the UBA domains are not
available, using the figures of their publication we find a striking similarity between their models
for Ubl-UBA interaction and our Ub-UBA de novo models. Similar to the interaction with Ubl,
an identical surface epitope of Ub is involved in the binding to both UBA(1) and UBA(2)
domains. Helix 1 of either UBA(1) (residues 191-199) or UBA(2) (residues 348-356) contacts
the five-stranded β-sheet of either Ub or Ubl domain. For UBA(1) the relative orientation of the
helix in the model is the same as observed for HHR23A Ubl-UBA(1). For UBA(2), the
orientation of helix 1 is rotated by 180° in our de novo model compared to the results of Walters
et al. (41). In contrast to our chemical shift mapping studies and those performed by Ryu et al. on
21
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
HHR23B UBA domains (38), Walters and co-workers (41) propose that only one helix of the
UBA domains contributes to the interaction with the Ubl domain. We find that the binding of the
UBA domains to Ub includes residues from both helices 1 and 3. This difference might explain
the much lower affinity of UBA domains for Ubl (KD ~2 mM) (38) than for Ub (KD ~300 to
500 µM). In summary, these results indicate that the interaction between Ub and UBA domains
might be different from the binding to Ub found for the structurally homologous CUE domain in
solution. However, further experimental data is required to understand how UBA domains
interact with Ub on a molecular level and whether the binding mechanism is different from the
CUE domains.
Finally, we note that RAD23 has been described as a binding partner through its UBA(2)
domain for various proteins involved in DNA repair and cell cycle control, e.g. HIV-1 Vpr (25),
methyladenine DNA glycosylase (MPG) (26), p300/cyclic AMP responsive element binding
(CREB)-protein (27), and peptide:N-glycanase (Png1) (28). Since the UBA-binding epitopes of
these proteins exhibit very different structures, the UBA domain of RAD23 must be able to bind
to various structural architectures. Therefore the binding mechanism of UBA domains is also
likely to vary for the diverse interacting proteins.
Acknowledgements
We thank Dr. Alexander Varshavsky for the gift of yeast ubiquitin expression plasmid
pET11c-Ub, Dr. Carina Johansson, Darian Cash and Nathan Cho for performing some of the
NMR experiments, and Evan Feinstein for manuscript and figure preparation. This work was
supported by NIH grant AI43190 to I.S.Y. Chen and J.F. The coordinates for the complex
models are available on request.
22
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
References
1. Ciechanover, A. (1998) EMBO J 17, 7151-7160.
2. Hershko, A., and Ciechanover, A. (1998) Annu Rev Biochem 67, 425-479.
3. Pickart, C. M. (2000) Trends Biochem Sci 25, 544-548.
4. von Arnim, A. G. (2001) Sci STKE 2001, E2.
5. Koegl, M., Hoppe, T., Schlenker, S., Ulrich, H. D., Mayer, T. U., and Jentsch, S. (1999)
Cell 96, 635-644.
6. Groll, M., Ditzel, L., Lowe, J., Stock, D., Bochtler, M., Bartunik, H. D., and Huber, R.
(1997) Nature 386, 463-471.
7. Lowe, J., Stock, D., Jap, B., Zwickl, P., Baumeister, W., and Huber, R. (1995) Science
268, 533-539.
8. Rubin, D. M., and Finley, D. (1995) Curr Biol 5, 854-858.
9. Deveraux, Q., Jensen, C., and Rechsteiner, M. (1995) J Biol Chem 270, 23726-23729.
10. Deveraux, Q., van Nocker, S., Mahaffey, D., Vierstra, R., and Rechsteiner, M. (1995) J
Biol Chem 270, 29660-29663.
11. van Nocker, S., Deveraux, Q., Rechsteiner, M., and Vierstra, R. D. (1996) Proc Natl
Acad Sci U S A 93, 856-860.
12. Lam, Y. A., Lawson, T. G., Velayutham, M., Zweier, J. L., and Pickart, C. M. (2002)
Nature 416, 763-767.
13. Glickman, M. H., Rubin, D. M., Fu, H., Larsen, C. N., Coux, O., Wefes, I., Pfeifer, G.,
Cjeka, Z., Vierstra, R., Baumeister, W., Fried, V., and Finley, D. (1999) Mol Biol Rep
26, 21-28.
14. Beal, R., Deveraux, Q., Xia, G., Rechsteiner, M., and Pickart, C. (1996) Proc Natl Acad
Sci U S A 93, 861-866.
15. Thrower, J. S., Hoffman, L., Rechsteiner, M., and Pickart, C. M. (2000) EMBO J 19, 94-
102.
23
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
16. Chau, V., Tobias, J. W., Bachmair, A., Marriott, D., Ecker, D. J., Gonda, D. K., and
Varshavsky, A. (1989) Science 243, 1576-1583.
17. Mastrandrea, L. D., You, J., Niles, E. G., and Pickart, C. M. (1999) J Biol Chem 274,
27299-27306.
18. Spence, J., Sadis, S., Haas, A. L., and Finley, D. (1995) Mol Cell Biol 15, 1265-1273.
19. Bertolaet, B. L., Clarke, D. J., Wolff, M., Watson, M. H., Henze, M., Divita, G., and
Reed, S. I. (2001) Nat Struct Biol 8, 417-422.
20. Chen, L., Shinde, U., Ortolan, T. G., and Madura, K. (2001) EMBO Rep 24, 24.
21. Ortolan, T. G., Tongaonkar, P., Lambertson, D., Chen, L., Schauber, C., and Madura, K.
(2000) Nat Cell Biol 2, 601-608.
22. Raasi, S., and Pickart, C. M. (2003) J Biol Chem 278, 8951-8959.
23. Clarke, D. J., Mondesert, G., Segal, M., Bertolaet, B. L., Jensen, S., Wolff, M., Henze,
M., and Reed, S. I. (2001) Mol Cell Biol 21, 1997-2007.
24. Kang, R. S., Daniels, C. M., Francis, S. A., Shih, S. C., Salerno, W. J., Hicke, L., and
Radhakrishnan, I. (2003) Cell 113, 621-630.
25. Withers-Ward, E. S., Jowett, J. B., Stewart, S. A., Xie, Y. M., Garfinkel, A., Shibagaki,
Y., Chow, S. A., Shah, N., Hanaoka, F., Sawitz, D. G., Armstrong, R. W., Souza, L. M.,
and Chen, I. S. (1997) J Virol 71, 9732-9742.
26. Miao, F., Bouziane, M., Dammann, R., Masutani, C., Hanaoka, F., Pfeifer, G., and
O’Connor, T. R. (2000) J Biol Chem 275, 28433-28438.
27. Zhu, Q., Wani, G., Wani, M. A., and Wani, A. A. (2001) Cancer Res 61, 64-70.
28. Suzuki, T., Park, H., Kwofie, M. A., and Lennarz, W. J. (2001) J Biol Chem 276, 21601-
21607.
29. Mueller, T. D., and Feigon, J. (2002) J Mol Biol 319, 1243-1255.
30. Withers-Ward, E. S., Mueller, T. D., Chen, I. S., and Feigon, J. (2000) Biochemistry 39,
14103-14112.
31. Cavanagh, J., Palmer, A. G., Fairbrothers, W., and Skelton, N. (1996) Protein NMR
24
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
spectroscopy: Principles and Practice, Academic Press, San Diego.
32. Dominguez, C., Boelens, R., and Bonvin, A. M. (2003) J Am Chem Soc 125, 1731-1737
33. Hubbard, S. J., and Thornton, J. M. (1993), London.
34. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-Kunstleve,
R. W., Jiang, J. S., Kuszewski, J., Nilges, M., Pannu, N. S., Read, R. J., Rice, L. M.,
Simonson, T., and Warren, G. L. (1998) Acta Crystallogr D Biol Crystallogr 54 (Pt 5),
905-921.
35. Linge, J. P., O’Donoghue, S. I., and Nilges, M. (2001) Methods Enzymol 339, 71-90
36. Dieckmann, T., Withers-Ward, E. S., Jarosinski, M. A., Liu, C. F., Chen, I. S., and
Feigon, J. (1998) Nat Struct Biol 5, 1042-1047.
37. Rao, H., and Sastry, A. (2002) J Biol Chem 277, 11691-11695.
38. Ryu, K. S., Lee, K. J., Bae, S. H., Kim, B. K., Kim, K. A., and Choi, B. S. (2003) J Biol
Chem 278, 36621-36627.
39. Mueller, T. D., and Feigon, J. (2003) EMBO J 22, 4634-4645.
40. Walters, K. J., Kleijnen, M. F., Goh, A. M., Wagner, G., and Howley, P. M. (2002)
Biochemistry 41, 1767-1777.
41. Walters, K. J., Lech, P. J., Goh, A. M., Wang, Q., and Howley, P. M. (2003) Proc Natl
Acad Sci U S A 100, 12694-12699.
42. Medek, A., Hajduk, P. J., Mack, J., and Fesik, S. W. (2000) J Am Chem Soc 122, 1241-
1242.
43. Prag, G., Misra, S., Jones, E. A., Ghirlando, R., Davies, B. A., Horazdovsky, B. F., and
Hurley, J. H. (2003) Cell 113, 609-620.
44. Schnell, J. D., and Hicke, L. (2003) J Biol Chem 278, 35857-35860.
45. Pornillos, O., Alam, S. L., Rich, R. L., Myszka, D. G., Davis, D. R., and Sundquist, W. I.
(2002) EMBO J 21, 2397-2406.
46. Wang, B., Alam, S. L., Meyer, H. H., Payne, M., Stemmler, T. L., Davis, D. R., and
Sundquist, W. I. (2003) J Biol Chem 278, 20225-20234.
25
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
47. Fisher, R. D., Wang, B., Alam, S. L., Higginson, D. S., Robinson, H., Sundquist, W. I.,
and Hill, C. P. (2003) J Biol Chem 278, 28976-28984.
48. Shekhtman, A., and Cowburn, D. (2002) Biochem Biophys Res Commun 296, 1222-
1227.
49. Sloper-Mould, K. E., Jemc, J. C., Pickart, C. M., and Hicke, L. (2001) J Biol Chem 276,
30483-30489.
50. Madura, K. (2002) Cell Cycle 1, 235-244.
51. Davies, B. A., Topp, J. D., Sfeir, A. J., Katzmann, D. J., Carney, D. S., Tall, G. G.,
Friedberg, A. S., Deng, L., Chen, Z., and Horazdovsky, B. F. (2003) J Biol Chem 278,
19826-19833.
52. Shih, S. C., Prag, G., Francis, S. A., Sutanto, M. A., Hurley, J. H., and Hicke, L. (2003)
EMBO J 22, 1273-1281.
53. Ponting, C. P. (2000) Biochem J 351 Pt 2, 527-535.
54. Wilkinson, C. R., Seeger, M., Hartmann-Petersen, R., Stone, M., Wallace, M., Semple,
C., and Gordon, C. (2001) Nat Cell Biol 3, 939-943.
55. Elsasser, S., Gali, R. R., Schwickart, M., Larsen, C. N., Leggett, D. S., Muller, B., Feng,
M. T., Tubing, F., Dittmar, G. A., and Finley, D. (2002) Nat Cell Biol 4, 725-730.
56. Funakoshi, M., Sasaki, T., Nishimoto, T., and Kobayashi, H. (2002) Proc Natl Acad Sci
USA 99, 745-750.
57. Chen, L., and Madura, K. (2002) Mol Cell Biol 22, 4902-4913.
58. Phillips, C. L., Thrower, J., Pickart, C. M., and Hill, C. P. (2001) Acta Crystallogr D Biol
Crystallogr 57, 341-344.
59. Beal, R. E., Toscano-Cantaffa, D., Young, P., Rechsteiner, M., and Pickart, C. M. (1998)
Biochemistry 37, 2925-2934.
26
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
Figure Captions
Figure 1: Chemical shift mapping of UBA(1) and UBA(2). (A and B) 500 MHz 1H-15N 2D
HSQC spectra of UBA(1) and UBA(2) free and bound to Ub. (A) 13C-,15N-labeled UBA(1)
free (black contour levels) and with 1:10 Ub (red) and (B) 15N-labeled UBA(2) free (black) and
with 1:10 Ub (red) at 27 ºC. Blue contours are used for side chain amide protons. (C and D)
Chemical shift change vs. sequence of UBA(1) and UBA(2). Graph representing the average
chemical shift change of the amide proton and nitrogen ∆δave of (C) UBA(1) and (D) UBA(2)
upon addition of 10 equivalent Ub. The dashed line indicate the threshold chosen for the color
coding used in Figure 1 E and F. Chemical shift changes of ∆δave smaller than 0.2 p.p.m. were
considered insignificant. (E and F) Binding interface of Ub on UBA(1) and UBA(2). Residues of
(E) UBA(1) and (F) UBA(2) shifting by more than 0.05 p.p.m. upon addition of Ub are marked
in orange. A representation of the molecular surface is shown on the left and a ribbon sketch is
shown on the right.
Figure 2: Chemical shift mapping of Ub. (A and B) 1H-15N HSQC of free Ub and Ub bound to
UBA(1) and UBA(2). 13C-,15N-labeled Ub in free conformation (black) and in complex with
(A) UBA(1) and (B) UBA(2) (red contour levels, ratio 1:4). Blue contour levels mark side chain
amide resonances. (C and D) Chemical shift change vs. sequence of Ub bound to UBA(1) and
UBA(2). The average chemical shift change (combined amide proton and amide nitrogen
chemical shift as for Figure 1 C and D) is shown for Ub bound to (C) UBA(1) and (D) UBA(2).
The dashed line indicates the threshold of chemical shift change used for Figures 2 E and F. (E
and F) Binding area of UBA(1) and UBA(2) on Ub. The changes of amide proton and nitrogen
chemical shift are displayed on the surface (left) and on a ribbon diagram (right) of human Ub as
classified in Figure 2 C, D. Residues are marked in orange according to the chemical shift
change (∆δave) on Ub upon addition of (E) UBA(1) and (F) UBA(2). For Ub binding to UBA(1)
(E) large changes (> 0.1 ppm.) for the chemical shifts are found for the residues L8, K11 (β1 -
β2 loop), I44 (β3), G47, K48, Q49 (β4), H68, L69 (β5), R72 and L73 (C-terminus). Residues
exhibiting smaller changes are V5, T7, G10, T12 (β1-β2 loop), I13 (β2), R42, L43 (β3), F45,
27
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
A46 (β4), L50, Q62 (β4-β5 loop), L67 (β5) and L71 (C-terminus). For Ub binding to UBA(2),
residues L8, T9, G10, K11 (β1-β2 loop), I13, T14 (β2), R42, I44 (β3), G47, K48, Q49 (β4),
G53 (β4-β5 loop), H68, L69, V70 (β5), L71, R72 and L73 (C-terminus) of Ub show significant
chemical shift changes.
Figure 3: Differential chemical shift mapping of the Ub-UBA interaction. (A) Analysis of the
binding curve determined from chemical shift mapping of the Ub-UBA(1) interaction. The
average chemical shift change of the residue G174 of UBA(1) upon binding to Ub is plotted
against the concentration of Ub or Ub mutant. (B) As in (A) but for the interaction of Ub and
UBA(2). The average chemical shift change of residue G331 was used for analysis. (C)
Comparison of the chemical shift changes upon addition of Ub/Ub L8A to HHR23A UBA(1).
The boxes mark the residues for which a comparison of the chemical shift change of the Ub (red)
and Ub L8A (green) titration yielded a difference of more than 3-fold. The according residues
are annotated. (D) As in figure (C) but comparing the titration of Ub (red) and Ub R42A (green).
Figure 4: Homology models for the UBA(1)-Ub and UBA(2)-Ub interaction. (A) Model for the
interaction of HHR23A UBA(1) and Ub based on the structure of the complex of Cue2 CUE
domain and Ub, in (B) for the interaction of UBA(2) and Ub. Left panel: Ribbon sketch with the
structure of UBA(1) (A) and UBA(2) (B) displayed by a color ramp blue to red (N-terminus to
C-terminus). The helical axes of the UBA domains are displayed by dashed black lines, as a
reference for the orientation the axis of β-strand 5 of Ub is shown as dashed red line. Middle
panel: Interface between UBA(1) (A) or UBA(2) (B) and Ub, the surface of Ub is shown in atom
colors. Residues of UBA(1) or UBA(2) in direct contact with Ub are shown and labeled. Right
panel: As for middle panel but residues of Ub (grey) in direct contact with residues of UBA(1)
(A) or UBA(2) (B) (dark grey) are shown and labeled.
Figure 5: De novo docking models for the UBA(1)-Ub and UBA(2)-Ub interaction. Similar to
Fig. 4, the interaction of (A) UBA(1) and (B) UBA(2) with Ub is shown. The complexes were
obtained by computational docking using ambiguous restraints based on the chemical shift
28
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
mapping results. Left panel: The helical axes of the UBA domains (UBA(1): in A; UBA(2) in B)
are represented by dashed black lines, the orientation of the UBA domain in respect to Ub can be
determined from the axis of β-strand 5 of Ub shown as red dashed line. Middle panel: The
interface between UBA(1) (A) and UBA(2) (B) and Ub. The surface of Ub is shown in atom
color; the residues of either UBA domain in contact with Ub are shown and labeled. Right panel:
As for middle panel, but residues of Ub in direct contact with UBA(1) (A) and UBA(2) (B) are
shown in grey and labeled.
Figure 6: Sequence alignment of UBA and CUE domain proteins. The sequences of several UBA
domains (SWISS-PROT entry codes are indicated) and CUE domains were aligned based on a
structural superposition of HHR23A UBA(1) and Cue2 CUE domain. The helices are marked by
boxes, the arrows indicate residues which either exhibit changes of their chemical shift upon
binding to Ub (upper panel, UBA domain) or which are buried in the complex (lower panel,
Cue2 CUE-Ub). The amino acids are color coded using green for hydrophobic residues
(A,F,I,M,L,V,Y,W), red for negatively, blue for positively charged residues (H,K,R), and orange
for polar amino acids (N,Q,S,T). The consensus sequence for UBA domains was taken from the
SMART database (http://www.embl-heidelberg.de/smart), the symbols for the amino acid
grouping are as follows: l = aliphatic (I, L, V); . = any; a = aromatic; c = charged; h =
hydrophobic (A, C, F, G, H, I, K, L, M, R, T, V, W, Y); p = polar; + = positively charged; - =
negatively charged; s = small (A, C, D, G, N, P, S, T, V); u = tiny (A, G, S) and t = turnlike (A,
C, D, E, G, H, K, N, Q, R, S, T).
29
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
Thomas D. Mueller, Mariusz Kamionka and Juli FeigonSpecificity of the interaction between UBA domains and ubiquitin
published online January 5, 2004J. Biol. Chem.
10.1074/jbc.M312865200Access the most updated version of this article at doi:
Alerts:
When a correction for this article is posted•
When this article is cited•
to choose from all of JBC's e-mail alertsClick here
by guest on June 11, 2020http://w
ww
.jbc.org/D
ownloaded from
Top Related