Content uploaded by Philip Coffino
Author content
All content in this area was uploaded by Philip Coffino
Content may be subject to copyright.
Teachers in Higher Education Institutions of the
MOE to A.M., and from the Bugher Foundation and
from the Specialized Research Fund for the Doctoral
Program of Higher Education of the MOE to Y.-G.C.
Supporting Online Material
www.sciencemag.org/cgi/content/full/306/5693/114/
DC1
Materials and Methods
Figs. S1 to S11
References and Notes
21 May 2004; accepted 17 August 2004
Ubistatins Inhibit Proteasome-
Dependent Degradation by
Binding the Ubiquitin Chain
Rati Verma,
1
Noel R. Peters,
2
Mariapina D’Onofrio,
3
Gregory P. Tochtrop,
2
Kathleen M. Sakamoto,
1,4
Ranjani Varadan,
3
Mingsheng Zhang,
5
Philip Coffino,
5
David Fushman,
3
Raymond J. Deshaies,
1
Randall W. King
2
*
To identify previously unknown small molecules that inhibit cell cycle machin-
ery, we performed a chemical genetic screen in Xenopus extracts. One class of
inhibitors, termed ubistatins, blocked cell cycle progression by inhibiting cyclin
B proteolysis and inhibited degradation of ubiquitinated Sic1 by purified pro-
teasomes. Ubistatins blocked the binding of ubiquitinated substrates to the
proteasome by targeting the ubiquitin-ubiquitin interface of Lys
48
-linked chains.
The same interface is recognized by ubiquitin-chain receptors of the pro-
teasome, indicating that ubistatins act by disrupting a critical protein-protein
interaction in the ubiquitin-proteasome system.
Unbiased chemical genetic screens can iden-
tify small molecules that target unknown pro-
teins or act through unexpected mechanisms
(1). To identify previously unknown compo-
nents or potential drug targets required for
cell division, we screened for small mole-
cules that stabilize cyclin B in Xenopus cell
cycle extracts. Cyclin B degradation regu-
lates exit from mitosis and requires activa-
tion of an E3 ubiquitin ligase called the
anaphase-promoting complex/cyclosome
(APC/C) (2). Because APC/C activation re-
quires mitotic entry, we anticipated that this
screen would identify compounds that stabi-
lized cyclin B indirectly by blocking mitotic
entry as well as compounds that directly
inhibited the cyclin proteolysis machinery.
To monitor APC/C activation, we fused the
destruction-box domain of Xenopus cyclin B1
to luciferase (3) and found that the reporter
protein was degraded in mitotic but not inter-
phase extracts (fig. S1). Proteolysis was sen-
sitive to inhibitors of cyclin-dependent kinases
and the ubiquitin-proteasome system but not
affected by inhibitors of DNA replication or
spindle assembly, as expected in egg extracts
lacking exogenous nuclei (4, 5) (fig. S2).
We developed a miniaturized assay system
(6) and screened 109,113 compounds to
identify 22 inhibitors (Table 1). To distinguish
compounds that blocked mitotic entry from
direct inhibitors of proteolysis, we arrested
extracts in mitosis before addition of the
compound and the reporter protein. Sixteen
compounds lost inhibitory activity under these
conditions (class I, fig. S3), whereas six
compounds (class II, fig. S4) remained inhib-
itory. We next activated proteolysis directly in
interphase extracts by adding the APC/C
activator Cdh1 (Cdc20 homolog 1) (7). Again
we found that only class II compounds re-
Fig. 1. Class IIB compounds inhibit degradation and deubiquitination of
UbSic1 by purified 26S proteasomes. (A) Purified 26S proteasomes were
preincubated in the presence or absence of test compounds. UbSic1 was
then added and assayed for degradation by immunoblotting for Sic1 (3).
Py mock refers to pyridine in which C23 was dissolved. (B) Purified 26S
proteasomes were preincubated with 100 6M epoxomicin in the pres-
ence or absence of 100 6M test compound. UbSic1 was then added and
deubiquitination monitored by immunoblotting for Sic1 (3). (C)
Titration of C92 in deubiquitination assay. (D) Structures of C92 and
C59 (ubistatins A and B).
R EPORTS
www.sciencemag.org SCIENCE VOL 306 1 OCTOBER 2004
117
tained inhibitory activity. We concluded that
class I compounds blocked entry into mitosis
or APC/C activation, whereas class II com-
pounds directly blocked components of the
cyclin degradation machinery. We next exam-
ined whether the inhibitors could block turn-
over of a $-catenin reporter protein (8), a
substrate of the SKP1/cullin/F-box protein
(SCF
$-TRCP
,where"-TRCP is "-transduction
repeat–containing protein) ubiquitin ligase
(Table 1). Three class II compounds (class
IIB) were inhibitory, suggesting these com-
pounds inhibited a protein required for the
degradation of both APC/C and SCF
$-TRCP
substrates. Class IIB compounds did not block
cyclin B ubiquitination or 20S peptidase
activity (9), indicating they did not inhibit E1
or act as conventional proteasome inhibitors.
To understand how class IIB compounds
inhibited proteolysis, we turned to a recon-
stituted system using purified 26S proteasomes
and ubiquitinated Sic1 (UbSic1) (10). Degra-
dation of Sic1 requires its ubiquitination by the
ligase SCF
Cdc4
(11 , 12), after which UbSic1 is
docked to the 19S regulatory particle by a
multi-Ub chain receptor (13). Proteolysis of
UbSic1 requires removal of the multi-Ub
chain, catalyzed by the metalloisopeptidase
Rpn11 (14, 15). The deubiquitinated substrate
is concomitantly translocated into the 20S core
particle, where it is degraded. Two class IIB
molecules, C92 and C59 (Fig. 1D), strongly
inhibited UbSic1 turnover in the reconstituted
system (Fig. 1A). To address whether these
1
Department of Biology, Howard Hughes Medical
Institute (HHMI), California Institute of Technology,
Pasadena, CA 91125, USA.
2
Institute of Chemistry and
Cell Biology and Department of Cell Biology, Harvard
Medical School, 240 Longwood Avenue, Boston, MA
02115, USA.
3
Department of Chemistry and Biochem-
istry, Center for Biomolecular Structure and Orga-
nization, University of Maryland, College Park, MD
20742, USA.
4
Division of Hematology-Oncology,
Mattel Children’s Hospital, Jonsson Comprehensive
Cancer Center, David Geffen School of Medicine at
University of California at Los Angeles (UCLA), 10833
Le Conte Avenue, Los Angeles, CA 90095, USA.
5
Department of Microbiology and Immunology, Uni-
versity of California, San Francisco, 513 Parnassus
Avenue, San Francisco, CA 94143–0414, USA.
*To whom correspondence should be addressed.
E-mail: randy_king@hms.harvard.edu
Table 1. Characterization of compounds in Xenopus extract assays. Results are reported as percent
inhibition (percent stimulation). Compounds (200 6M, except C10 and C92, tested at 100 6M) and
cyclin-luciferase (cyc-luc) were added to interphase extracts and then induced to enter mitosis by
addition of nondegradable cyclin B, or extracts were pretreated with nondegradable cyclin B to allow
entry into mitosis before addition of test compound and cyc-luc. Cdh1 was added to interphase extracts
before addition of compound and cyc-luc. Interphase extracts were treated with recombinant axin to
induce turnover of $-catenin-luciferase. Parentheses indicate those values where stimulation, rather than
inhibition, was observed by addition of compound to the reaction.
Compound
Addition before
mitotic entry
Addition
after mitotic
entry
Cdh1-activated
interphase extract
$-catenin
reporter
protein
Class IA
C77 100 4 (12) 0
C58 100 5 (8) 2
C82 100 0 0 0
C34 100 0 (8) 6
C62 84 0 (8) 0
C61 77 8 (8) 2
C13 75 0 (9) 0
C18 73 4 (7) 0
C25 66 3 (6) 0
C54 54 3 (6) 0
C67 53 3 (8) 3
C40 42 0 (6) 3
Class IB
C39 100 9 (7) 67
C57 100 4 0 60
C51 100 0 0 30
C10 33 0 (4) 21
Class IIA
C1 100 100 35 6
C2 80 50 100 0
C8 70 63 20 0
Class IIB
C23 100 100 100 27
C59 97 100 100 70
C92 60 22 65 21
Fig. 2. C92 inhibits binding of UbSic1 to 26S proteasomes and multi-Ub-chain
receptors by binding to K48-linked multi-Ub chains. (A)Purified26S
proteasomes immobilized on anti-Flag beads were incubated with UbSic1 in
the presence or absence of C92 as described in (3). Beads were then washed
and analyzed by immunoblotting for Sic1. (B) Recombinant Gst-Rpn10 and Gst-
Rad23 were immobilized on glutathione sepharose beads and then incubated
with UbSic1 in the presence or absence of C92 and analyzed as in (A). (C)
Equivalent amounts of Gst, Gst-fusion protein, or multi-Ub chains were
incubated with C92 or C1 and analyzed by native gel electrophoresis (28).
(D) C92 and C59 interact specifically with K48-linked Ub on native gels. Ub (16
6M), K48-linked di-Ub (8 6M), or tetra-Ub chains (8 6M) were preincubated
with a twofold molar excess (mono-Ub and di-Ub) or equivalent amounts
(tetra-Ub) of test compounds before being resolved on native gels as in (C).
Tetra K29Ub, K48Ub, and K63Ub refer to tetraubiquitin chains with ubiquitin
linked via K29, K48, or K63. MW refers to molecular weight standards.
R EPORTS
1 OCTOBER 2004 VOL 306 SCIENCE www.sciencemag.org
118
compounds acted upstream or downstream of
Rpn11 isopeptidase, we treated proteasomes
with the 20S inhibitor epoxomicin, which
results in Rpn11-dependent substrate de-
ubiquitination (14, 16) and accumulation of
deubiquitinated Sic1 within the 20S chamber
(13). This reaction was completely blocked
by C92 (Fig. 1B), with a median inhibitory
concentration (IC
50
) of about 400 nM (Fig.
1C). C59, which is structurally related to C92,
also inhibited deubiquitination of UbSic1 (IC
50
0
1 6M), whereas C23 inhibited marginally
(fig. S5). Thus C92 and C59 potently
blocked proteolysis at or upstream of the
essential isopeptidase-dependent step.
Selective recognition of the multi-Ub chain
by the 26S proteasome is the first step in
UbSic1 degradation (13). C92 strongly inhi-
bited binding of UbSic1 to purified 26S pro-
teasomes (Fig. 2A), suggesting that it inhibited
UbSic1 turnover by blocking the first step in
the degradation process. The multi-Ub chain
receptors Rad23 and Rpn10 serve a redundant
role in targeting UbSic1 to the proteasome and
sustaining its degradation (13). In the absence
of the Ub-binding activities of Rpn10 and
Rad23, UbSic1 is not recruited, deubiquiti-
nated, or degraded by purified 26S protea-
somes. We thus tested whether C92 could
interfere with binding of UbSic1 to recombi-
nant Rpn10 and Rad23. C92 abolished bind-
ing of UbSic1 to both proteins (Fig. 2B), even
though these receptors use distinct domains
Ethe Ub-interaction motif (UIM) and the Ub-
associated (UBA) domain, respectively^ to
bind ubiquitin chains (17). C59 also abrogated
binding of UbSic1 to Rpn10, whereas other
compounds were without effect (fig. S5).
To distinguish whether C92 inhibited pro-
teolysis by binding to proteasome receptor
proteins or to the Ub chain on Sic1, we ex-
ploited the negative charge of C92 to deter-
mine whether compound binding induced a
mobility shift of the target proteins upon
fractionation on a native polyacrylamide
gel. C92 was preincubated with recombinant
Rpn10, Rad23, or a mixture of Ub chains
containing two to seven Ub molecules. The
mobility of the multi-Ub chains, but not Gst-
Rpn10 or Gst-Rad23, was altered by incuba-
tion with C92, suggesting that C92 bound Ub
chains (Fig. 2C). Ubiquitin molecules can be
linked to each other in vivo through different
internal lysines, including K29, K48, and
K63 (18). The K48-linked chain is the
principal targeting signal in proteolysis,
whereas K63-linked chains are implicated
in enzyme regulation (19). Whereas C92 and
C59 efficiently shifted the native gel mobil-
ity of K48-linked ubiquitin chains, they had
little or no effect on K29- or K63-linked
chains (Fig. 2D). Because C92 and C59 bind
to ubiquitin chains and block interactions
with proteasome-associated receptors with-
out affecting 26S assembly or peptidase
activity (fig. S6), we refer to these com-
pounds as ubistatin A and B, respectively.
We next tested the ability of ubistatins to
block proteolysis of ornithine decarboxylase
(ODC), whose degradation does not require
ubiquitin (20). Whereas a 30-fold molar ex-
cess of ubistatin A over the substrate strongly
inhibited UbSic1 degradation by purified
yeast proteasomes (Fig. 1A), a 100-fold molar
excess of ubistatin A over the substrate had
no effect on degradation of radiolabeled
ODC by purified rat proteasomes (fig. S7).
Ubistatin B marginally inhibited ODC turn-
over at this concentration (12%). In contrast,
a 20-fold molar excess of cold ODC inhibited
degradation of labeled ODC by 43% under
the same conditions. These data indicate that
ubistatins at low concentrations preferentially
inhibit the degradation of ubiquitin-dependent
substrates. Inhibition of ODC turnover by
high concentrations of ubistatins, especially
ubistatin B (fig. S7), may reflect either
nonspecific activity or specific inhibition of
a targeting mechanism shared by ubiquitin-
dependent and ubiquitin-independent sub-
strates of the proteasome (20).
On the basis of the selectivity of ubi-
statin A for binding K48-linked chains and
inhibiting the ubiquitin-dependent turnover
of Sic1 but not the ubiquitin-independent
turnover of ODC, we tested the effect of
ubistatin A on protein degradation within
intact mammalian cells. Because the neg-
ative charge on ubistatin A precluded ef-
ficient membrane permeation, we introduced
the compound into cells by microinjection
and monitored degradation of an androgen
receptor–green fluorescent protein (AR-GFP)
fusion protein by fluorescence microscopy.
Microinjection of a synthetic compound (protac,
proteolysis-targeting chimeric molecule), which
recruits AR-GFP to SCF
$-TRCP
, induces rapid
proteasome-dependent turnover of AR-GFP
(21). Microinjection of 100 nM ubistatin A
into mammalian cells inhibited the Protac-
induced degradation of AR-GFP as effi-
ciently as 100 nM epoxomicin (fig. S8),
demonstrating that ubistatin A is an effective
Fig. 3. Ubistatin A
binding to K48-linked
di-Ub induces site-
specific perturbations
in NMR spectra for
both Ub domains. (A)
Backbone NH chem-
ical shift perturbation,
%&, and percent signal
attenuation caused by
ubistatin A binding as
a function of residue
number for the distal
(left) and the proxi-
mal (right) domains.
Ub units are called
‘‘distal’’ and ‘‘proxi-
mal’’ to reflect their
location in the chain
relative to the free C
terminus. The dia-
gram (top) depicts
the location of the
G76-K48 isopeptide
bond between the
two Ub domains. As-
terisks indicate res-
idues that showed
significant signal at-
tenuation that could
not be accurately
quantified because of
signal overlap. (B)
Mapping of the per-
turbed sites on the
surface of di-Ub. The
distal and proximal
domains are shown
in surface representa-
tion and colored blue
and green, respec-
tively; the perturbed
sites on these domains are colored yellow and red and correspond to residues with %& 9 0.075
parts per million and/or signal attenuation greater than 50%. Numbers indicate surface location of
the hydrophobic patch and some basic residues along with G76 (distal) and the side chain of K48
(proximal).
R EPORTS
www.sciencemag.org SCIENCE VOL 306 1 OCTOBER 2004
119
inhibitor of ubiquitin-dependent degradation
in multiple experimental settings.
The specificity of ubistatin A for K48-
linked ubiquitin chains suggested that it might
bind at the Ub-Ub interface, which is well
defined in K48-linked chains but is not present
in K63-linked di-ubiquitin (Ub
2
)(22). We
performed nuclear magnetic resonance
(NMR) titration studies of K48-linked Ub
2
by using a segmental labeling strategy (23).
Well-defined site-specific perturbations were
observed in the resonances of the backbone
amides of both Ub units in Ub
2
(Fig. 3),
indicating that the hydrophobic patch residues
L8, I44, V70 (24), and neighboring sites
(including basic residues K6, K11, R42, H68,
and R72) experienced alterations in their
molecular environment upon binding of ubis-
tatin A. The same hydrophobic patch is
involved in the formation of the interdomain
interface in Ub
2
(23, 25) and mediates the
binding of ubiquitin to multiple proteins
containing CUE (coupling of ubiquitin con-
jugation to ER degradation), UBA, and UIM
domains (17). At the high concentrations of
compound used in the NMR titration experi-
ments, ubistatin A induced a similar pattern of
chemical shift perturbations in monomeric
ubiquitin, suggesting that the effect of ubi-
statin A on Ub
2
arises from its direct binding
to the hydrophobic patch and the basic res-
idues around it. The same sites are perturbed
when ubistatin A binds tetra-Ub chains (26).
Although there is intense interest in devel-
oping drugs for defined molecular targets, it is
often difficult to know a priori which proteins
can be most effectively targeted with small
molecules. Our study demonstrates that chem-
ical genetic screens in complex biochemical
systems such as Xenopus extracts can identify
small-molecule inhibitors that act through
unexpected mechanisms. Although target iden-
tification remains challenging, our work high-
lights the value of reconstituted biochemical
systems to illuminate the mechanism of action
of inhibitors discovered in unbiased screens.
The recent approval of the 20S proteasome
inhibitor Velcade (Millenium Pharmaceuticals,
Cambridge, MA) for treatment of relapsed
multiple myeloma (27) has suggested that the
ubiquitin-proteasome system is an attractive
target for cancer drug development. The iden-
tification of ubistatins indicates that the ubiq-
uitin chain itself provides another potential
opportunity for pharmacological intervention
in this important pathway.
References and Notes
1. T. U. Mayer, Trends Cell Biol. 13, 270 (2003).
2. J. M. Peters, Mol. Cell 9, 931 (2002).
3. Materials and methods are available as supporting
material on Science Online.
4. M. Dasso, J. W. Newport, Cell 61, 811 (1990).
5. J. Minshull, H. Sun, N. K. Tonks, A. W. Murray, Cell
79, 475 (1994).
6. L. A. Walling, N. R. Peters, E. J. Horn, R. W. King, J.
Cell. Biochem. S37, 7 (2001).
7. C. M. Pfleger, M. W. Kirschner, Genes Dev. 14, 655
(2000).
8. A. Salic, E. Lee, L. Mayer, M. W. Kirschner, Mol. Cell 5,
523 (2000).
9. N. Peters, R. W. King, unpublished data.
10. R. Verma, H. McDonald, J. R. Yates 3rd, R. J. Deshaies,
Mol. Cell 8, 439 (2001).
11. D. Skowyra et al., Science 284, 662 (1999).
12. J. H. Seol et al., Genes Dev. 13, 1614 (1999).
13. R. Verma, R. Oania, J. Graumann, R. J. Deshaies, Cell
118, 99 (2004).
14. R. Verma et al., Science 298, 611 (2002); published
online 15 August 2002; 10.1126/science.1075898.
15. T. Yao, R. E. Cohen, Nature 419, 403 (2002).
16. L. Meng et al., Proc. Natl. Acad. Sci. U.S.A. 96, 10403
(1999).
17. R. Hartmann-Petersen, M. Seeger, C. Gordon, Trends
Biochem. Sci. 28, 26 (2003).
18. J. Peng et al., Nat. Biotechnol. 21, 921 (2003).
19. C. M. Pickart, Cell 116, 181 (2004).
20. M. Zhang, C. M. Pickart, P. Coffino, EMBO J. 22, 1488
(2003).
21. K. M. Sakamoto et al., Mol. Cell. Proteomics 2, 1350
(2003).
22. R. Varadan et al., J. Biol. Chem. 279, 7055 (2004).
23. R. Varadan, O. Walker, C. Pickart, D. Fushman, J. Mol.
Biol. 324, 637 (2002).
24. Single-letter abbreviations for the amino acid resi-
dues are as follows: H, His; I, Ile; K, Lys; L, Leu; R, Arg;
and V, Val.
25. W. J. Cook, L. C. Jeffrey, M. Carson, Z. Chen, C. M.
Pickart, J. Biol. Chem. 267, 16467 (1992).
26. D. Fushman, unpublished data.
27. J. Adams, Nat. Rev. Cancer 4, 349 (2004).
28. R. Verma et al., Mol. Biol. Cell 11, 3425 (2000).
29. We thank the Developmental Therapeutics Program,
National Cancer Institute, for providing access to
compound collections, C. Pickart for tetraubiquitin
chains of defined linkages, A. Salic for recombinant
Axin and $-catenin-luciferase, and C. Sawyers for AR-
GFP. G.T. is supported by NIH National Research Service
Award GM068276. K.M.S. is supported by a UCLA
Specialized Programs of Research Excellence in Prostate
Cancer Development Research Seed Grant (P50
CA92131), U.S. Department of Defense (DAMD17-03-
1-0220), and NIH (R21CA108545). P.C. is supported by
NIH R01 GM-45335. D.F. is supported by NIH grant
GM65334. R.J.D. is supported by HHMI and the Susan
G. Komen Breast Cancer Foundation (DISS0201703).
R.W.K. is supported by the NIH (CA78048 and
GM66492), the McKenzie Family Foundation, and the
Harvard-Armenise Foundation and is a Damon Runyon
Scholar. Screening facilities at the Harvard Institute of
Chemistry and Cell Biology were supported by grants
from the Keck Foundation, Merck and Company, and
Merck KGaA. R.J.D. is a founder and paid consultant of
Proteolix, which is negotiating with Caltech and
Harvard to license a patent related to ubistatin.
Molecular interaction data have been deposited in the
Biomolecular Interaction Network Database with
accession codes 151787 to 151791.
Supporting Online Material
www.sciencemag.org/cgi/content/full/306/5693/117/
DC1
Materials and Methods
Figs. S1 to S8
Table S1
1 June 2004; accepted 6 August 2004
Regulation of Cytokine Receptors
by Golgi N-Glycan Processing
and Endocytosis
Emily A. Partridge,
1,3
Christine Le Roy,
1
Gianni M. Di Guglielmo,
1
Judy Pawling,
1
Pam Cheung,
1,2
Maria Granovsky,
1,2
Ivan R. Nabi,
4
Jeffrey L. Wrana,
1,2
James W. Dennis
1,2,3
*
The Golgi enzyme "1,6 N-acetylglucosaminyltransferase V (Mgat5) is up-
regulated in carcinomas and promotes the substitution of N-glycan with poly
N-acetyllactosamine, the preferred ligand for galectin-3 (Gal-3). Here, we
report that expression of Mgat5 sensitized mouse cells to multiple cytokines.
Gal-3 cross-linked Mgat5-modified N-glycans on epidermal growth factor and
transforming growth factor–" receptors at the cell surface and delayed their
removal by constitutive endocytosis. Mgat5 expression in mammary
carcinoma was rate limiting for cytokine signaling and consequently for
epithelial-mesenchymal transition, cell motility, and tumor metastasis. Mgat5
also promoted cytokine-mediated leukocyte signaling, phagocytosis, and
extravasation in vivo. Thus, conditional regulation of N-glycan processing
drives synchronous modification of cytokine receptors, which balances their
surface retention against loss via endocytosis.
Co-translational modification of proteins in
the endoplasmic reticulum by N-glycosylation
facilitates their folding and is essential in
single-cell eukaryotes. Metazoans have addi-
tional Golgi enzymes that trim and remodel
the N-glycans, producing complex-type N-
glycans on glycoproteins destined for the
cell surface. Mammalian development re-
quires complex-type N-glycans containing
N-acetyllactosamine antennae, because their
complete absence in Mgat1-deficient em-
bryos is lethal (1, 2). Deficiencies in N-
acetylglucosaminyltransferase II and V (Mgat2
and Mgat5) acting downstream of Mgat1
reduce the content of N-acetyllactosamine,
and mutations in these loci result in viable
mice with a number of tissue defects (3, 4). N-
glycan processing generates ligands for vari-
ous mammalian lectins, but the consequences
of these interactions are poorly understood.
The galectin family of N-acetyllactosamine-
binding lectins has been implicated in cell
R EPORTS
1 OCTOBER 2004 VOL 306 SCIENCE www.sciencemag.org
120
