Azaindoles as Zinc-Binding Small-Molecule Inhibitors of the JAMM Protease CSN5

Article abstract of DOI:10.1002/anie.201608672

CSN5 is the zinc metalloprotease subunit of the COP9 signalosome (CSN), which is an important regulator of cullin-RING E3 ubiquitin ligases (CRLs). CSN5 is responsible for the cleavage of NEDD8 from CRLs, and blocking deconjugation of NEDD8 traps the CRLs in a hyperactive state, thereby leading to auto-ubiquitination and ultimately degradation of the substrate recognition subunits. Herein, we describe the discovery of azaindoles as a new class of CSN5 inhibitors, which interact with the active-site zinc ion of CSN5 through an unprecedented binding mode. The best compounds inhibited CSN5 with nanomolar potency, led to degradation of the substrate recognition subunit Skp2 in cells, and reduced the viability of HCT116 cells.

Full text of DOI:10.1002/anie.201608672

Angewandte
Communications
Chemie
International Edition: DOI: 10.1002/anie.201608672
German Edition: DOI: 10.1002/ange.201608672
Inhibitors
Azaindoles as Zinc-Binding Small-Molecule Inhibitors of the JAMM
Protease CSN5
Eva Altmann,* Paul Erbel, Martin Renatus, Michael Schaefer, Anita Schlierf, Adelaide Druet,
Laurence Kieffer, Mickael Sorge, Keith Pfister, Ulrich Hassiepen, Matthew Jones,
Simon Ruedisser, Daniela Ostermeier, Bruno Martoglio, Anne B. Jefferson, and Jean Quancard
Abstract: CSN5 is the zinc metalloprotease subunit of the
COP9 signalosome (CSN), which is an important regulator of
cullin-RING E3 ubiquitin ligases (CRLs). CSN5 is responsible
for the cleavage of NEDD8 from CRLs, and blocking
deconjugation of NEDD8 traps the CRLs in a hyperactive
state, thereby leading to auto-ubiquitination and ultimately
degradation of the substrate recognition subunits. Herein, we
describe the discovery of azaindoles as a new class of CSN5
inhibitors, which interact with the active-site zinc ion of CSN5
through an unprecedented binding mode. The best compounds
inhibited CSN5 with nanomolar potency, led to degradation of
the substrate recognition subunit Skp2 in cells, and reduced the
viability of HCT116 cells.
requires the structural CSN framework for exerting isopepti-
dase activity.[5] Elevated expression of CSN5 has been found
in breast, thyroid, skin, ovarian, lung, and liver cancers.[6] In
hepatocellular carcinoma cells, knockdown of CSN5 by small-
interfering RNA (siRNA) inhibits cell-cycle progression and
causes strong induction of apoptosis in vitro.[7] The down-
regulation of CSN5 impedes CSN function, as demonstrated
by an accumulation of neddylated cullin 1 (Cul1), and results
in a decrease in the oncogenic F-box protein Skp2. Stabiliza-
tion of the NEDD8–CRL complex through inhibition of
CSN5 could thus represent a novel therapeutic approach for
the treatment of CSN5-dependent cancers.[8,9]
A high-throughput screening of the entire Novartis
compound archive was carried out using a time-resolved
fluorescence energy transfer (TR-FRET)-based protease
activity assay with recombinant CSN and a fluorescence-
labelled NEDD8-modified CRL as a substrate (Table S1 in
the Supporting Information). Among other hits[10] a series of
7-azaindoles, exemplified by compound 1, were identified as
CSN5 inhibitors. In a dose–response experiment, the half
maximal inhibitory concentration (IC50) of 1 was determined
to be 1.4 mm in a fluorescence polarization (FP)-based
protease activity assay.[11] To further
Cullin-RING ligases (CRLs) represent the largest family of
E3 ubiquitin ligases, which are responsible for about 20% of
total cellular protein turnover through the ubiquitin-protea-
some system.[1] Neddylation and deneddylation of CRLs are
integral steps in their assembly as active complexes, regulat-
ing binding and ubiquitination of proteins during cell-cycle
progression, cell growth, and survival. Deneddylation, an
essential step for maintaining optimal levels of CRL activity,[2]
is mediated by the COP9 signalosome (CSN), a protein
complex comprising eight subunits.[3] The deneddylation of
CRLs is required for new substrate binding and to prevent
autoubiquitination and subsequent degradation of the sub-
strate recognition subunits of the CRL.[4] The subunit CSN5,
a JAMM metalloprotease, is responsible for the cleavage of
ubiquitin-like NEDD8 from CRLs. Importantly, CSN5
validate 1 as a CSN5 inhibitor, we
performed binding studies with
recombinant CSN5(2-257).
Even though the isolated enzyme
is not proteolytically active, we specu-
lated that the compound could still
bind to the protein, since the active
site is largely intact.[12] In addition, we
demonstrated that the isolated pro-
tein is capable of binding zinc, as
shown by the 88C increase in the
[*] Dr. E. Altmann, Dr. P. Erbel, Dr. M. Renatus, Dr. M. Schaefer,
Dr. A. Schlierf, A. Druet, L. Kieffer, M. Sorge, K. Pfister,
Dr. U. Hassiepen, M. Jones, Dr. S. Ruedisser, D. Ostermeier,
Dr. B. Martoglio, Dr. J. Quancard
melting temperature of recombinant CSN5(2-257) when Zn2+
is added (Figure S1 in the Supporting Information). We
measured affinity by surface plasmon resonance (SPR) and
found that 1 bound reversibly to CSN5(2-257) with a KD value
of 1.5 mm, which is in line with its potency in the biochemical
assay (Figure 1A).
We then used nuclear magnetic resonance (NMR) spec-
troscopy to further characterize the binding of 1 to CSN5(2-
257). As shown in Figure 2, in the absence of ligand, the
resonances of two methionines (e methyl group) are very
weak and broadened to dynamics. Addition of compound 1 to
the protein leads to chemical-shift changes and an intensity
increase of these methyl signals. Selective 13C labelling of
methionine residues combined with site-directed mutagenesis
Novartis Institutes for BioMedical Research
Novartis Campus, 4002 Basel (Switzerland)
E-mail: eva.altmann@novartis.com
K. Pfister, Dr. A. B. Jefferson
Novartis Institutes for BioMedical Research
4560 Horton Street, Emeryville, CA 94608-2916 (USA)
Dr. B. Martoglio
Present address: Pharma Research and Early Development
Roche Innovation Center Basel, 4070 Basel (Switzerland)
Dr. A. B. Jefferson
Present address: Centers for Therapeutic Innovation, Pfizer Inc.
San Francisco, CA 94158 (USA)
Supporting information and the ORCID identification number(s) for
the author(s) of this article can be found under http://dx.doi.org/10.
1002/anie.201608672.
Angew. Chem. Int. Ed. 2016, 55, 1 – 5
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
1
These are not the final page numbers!

Angewandte
Communications
Chemie
Figure 1. SPR sensograms for the binding of 7-azaindole 1 (A) and 6-
azaindole 2 (B) to CSN5, as measured on a ProteOn GLM sensor chip
coated with CSN5 to a value of 120 RU. The same compound dilution
range (10, 3, 1, 0.3, 0.1 mm) was used in both experiments. The data
demonstrate binding of 1 to CSN5, while no binding is observed for 2.
Scheme 1. a) i.HATU, DMF, RT; ii. AcOH, 708C, 80%. b) 4m HCl in
dioxane, RT, 98%. c) I2, CF3SO3Ag, THF, RT, 54%. d) LiOH, THF/H2O
(3:1), RT, 96%. e) HBTU, DIEA, DMF, RT, 91%. f) Pd(PPh3)2Cl2,
Na2CO3, DME, 908C, 11%. g) Pd(PPh3)2Cl2, Na2CO3, DME, 908C,
96%. h) Bis(pinacolato)diboron, Pd2(dba)3, XPhos, KOAc, dioxane/
H2O (2:1), 1108C, 69%. For details see the Supporting Information.
HATU=N-[(dimethylamino)-1H-1,2,3-triazolo-[4,5-b]pyridin-1-ylmethy-
lene]-N-methylmethanaminium hexafluorophosphate N-oxide,
THF=tetrahydrofuran, HBTU=N,N,N’,N’-tetramethyl-O-(1H-benzo-
triazol-1-yl)uronium hexafluorophosphate, DIEA=N-ethyldiisopropyla-
mine, DMF=N,N-dimethylformamide, DME=1,2-dimethoxyethane,
1,2-dimethoxyethane.
Figure 2. HSQC of 13C-methionine-labelled CSN5(2-257) in the absence
(black) and presence (red) of compound 1. The HSQC experiment was
performed at a concentration of 40 mm CSN5(2-257) and 200 mm of
compound 1.
(see the Supporting Information) and information from
a
preliminary, low-resolution structure of CSN5(2-257)
allowed us to assign the methyl resonances of M78, which is
in close proximity to the active-site zinc, and of M117, which is
part of the flexible a-helical loop region covering the
substrate binding cleft.
The data suggest that inhibitor 1 targets the CSN5
catalytic site. Surprisingly, compound 1 does not contain any
known Zn2+ binding motif,[13] which raised the question of its
mode of action. The close analogue 2, which contains a 6-
azaindole motif, did not inhibit the enzyme and was shown by
SPR not to bind to CSN5 (Figure 1B). This suggests that the
7-azaindole substructure is essential for binding.
Initial co-crystallization trials of 1 with CSN5(2-257) were
unsuccessful, possibly due to the moderate binding affinity of
the compound. We thus engaged in an optimization effort
focusing on improving biochemical potency. Key compounds
3–6 and ent-3 were prepared in the course of this optimiza-
tion. The synthesis of these new azaindole inhibitors followed
the route exemplified for compound 4 in Scheme 1.
Coupling of Boc-morpholine-(S)-3-carboxylic acid 7 with
diamine 8 using HBTU afforded an amide intermediate that
was dehydrated at elevated temperatures to afford benzimi-
dazole 9 (Scheme 1). Boc removal with HCl/dioxane yielded
intermediate 10. Silver triflate assisted C2-iodination of
azaindole 11 provided methyl 3-(2-iodo1-1H-pyrrolo[2,3-
b]pyridin-3-yl)propanoate, which was hydrolyzed with
LiOH to the corresponding acid 12. In the following step,
intermediate 12 was coupled to the morpholino-benzimida-
zole 10 to give 13. Elaboration of 13 to target molecule 4 was
accomplished through a palladium(0)-catalyzed arylation
with 16.
We were able to solve an X-ray co-crystal structure of
inhibitor 4 in complex with CSN5(2-257) at a resolution of
2.2 ꢀ (Figure 3) and refine it to an Rfree of 21.9% (Table S3).
The structure shows that the compound is bound in the active
site of the enzyme and displays an unprecedented binding
mode for a zinc metalloprotease inhibitor. The azaindole
moiety is bound in the catalytic center of CSN5 through a dual
interaction. N7 coordinates to the active-site Zn2+ (N Zn
À
distance of 2.1 ꢀ) in a monodentate fashion, while the NH
forms a tight hydrogen bond with Og of the D151 carboxylate
(Figure 3C). The proximal 5-methyl-phenyl group of the 2-
2
www.angewandte.org
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2016, 55, 1 – 5
These are not the final page numbers!

Angewandte
Communications
Chemie
Figure 4. A ligand-interaction diagram for ligand 4. The ligand-inter-
action plot was generated using the algorithm described by Clark
et al.[14]
.
in an intramolecular edge-to-face stacking interaction
(Figure 4).
The inhibitory activity of azaindole derivatives 3–6 was
determined in a FP-based protease activity assay with
recombinant CSN holoenzyme and PT22-Nedd8-Cul4 as the
substrate. IC50 values for the inhibition of CSN5, as deter-
mined in this assay, are reported in Table 1. A clear improve-
Table 1: Activities of the 7-azaindole derivatives.
Compound
1
3
ent-3
4
5
6
IC50 [mm][a]
1.4
0.1
1.2
0.09
0.18
0.06
[a] Inhibition of CSN5 in a FP-based protease activity assay using PT22-
Nedd8-Cul4 as substrate. Data represent the mean of at least 2
experiments performed in duplicate. Individual data points were within
a 2-fold range with each other. For details, see the Supporting
Information.
ment in potency to an IC50 of 0.1 mm was achieved through
rigidification of the linker moiety. The rigidification was
attained by making the bond connecting the amide nitrogen
to the benzimidazole moiety part of a morpholine ring
(Table 1, compound 3). Specific inhibition of CSN protease
activity by 3 was also confirmed by the 12-fold lower potency
of its enantiomer ent-3. Based on the structure of the
morpholino analogue 3, the effects of attaching small
peripheral substituents were evaluated. Only moderate
changes in potency were observed for most substituents
investigated, for example, a methyl (4) or cyano (5) substitu-
ent at position 5 of the proximal phenyl ring. A slight
improvement in potency relative to 3 was observed with the
doubly methylated compound 6.
Figure 3. X-ray co-crystal structure of 4 (orange) bound in the active
site of CSN5(2-257) (gray) at 2.2 ꢀ resolution (PDB ID: 5M5Q), refined
to Rfree of 21.9%. A) Overall structure as a cartoon representation with
the ligand shown as a stick representation. The N and C termini of the
polypeptide chain are indicated. B) The cyan mesh represents the
refined 2FoÀFc electron density map (contoured at 1s). The dotted line
depicts a stretch of amino acids (100–108) that are poorly defined by
electron density and were removed from the final model. C) A detailed
interaction pattern for the azaindole core with the active site Zn2+ ion
and D151. Distances are given in ꢀ. All images were generated using
PyMOL (The PyMOL Molecular Graphics System, Version 1.8.2.2,
Schrçdinger, LLC.).
biphenyl substituent fills a lipophilic pocket formed by V77
and T154, whereas the distal phenyl ring is orientated towards
a lipophilic cavity formed by M78 and V77. The spacer and
the morpholino group position the benzimidazole substituent
in a hydrophobic area around P99. The benzimidazole group
and the distal phenyl ring of the biphenyl moiety are involved
We next explored the specificity profiles of compounds 4
and 6 against a panel of seven human matrix metalloproteases
(MMPs; Table S2). None of the MMPs tested was inhibited
significantly by these compounds. It could be that the
different active-site architectures of MMPs[15] compared to
CSN5, namely the coordination of the Zn2+, disfavors binding
Angew. Chem. Int. Ed. 2016, 55, 1 – 5
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
3
These are not the final page numbers!

Angewandte
Communications
Chemie
of the azaindole moiety. As shown in Figure 3C, D151, the
third zinc-coordinating residue in CSN5, plays a key role in
ligand binding and is involved in a hydrogen-bonding
interaction with the NH of the azaindole. In contrast to
JAMM proteases, where the catalytic zinc is coordinated by
two histidines and one aspartate (H138, H140, D151), the
catalytic zinc ion in MMPs is coordinated by three histidine
residues. Consequently, binding interactions with the 6-
azaindole motif are not possible, and in addition, there is
not enough space to accommodate the bicyclic heterocyclic
core in the MMP active site.
We then assessed the effects of our inhibitors in cellular
systems. Treatment of human colon carcinoma derived
HCT116 cells with inhibitors 5 or 6 led to the trapping of
Cul1 in the neddylated state (Figure 5A), thus demonstrating
inhibition of deneddylation. As a consequence, the oncogenic
F-box protein Skp2 was degraded in a dose-dependent
manner. These results mirror the effects observed when
CSN5 is depleted by siRNA.[7] Next we explored the effect of
these inhibitors on the viability of HCT116 cells. Compounds
5 and 6 exhibited potent inhibition of proliferation, as seen in
a CellTiter-Glo assay, with EC50 values of 4.7 and 4.9 mm,
respectively, which confirms the potential of CSN5 inhibitors
as anticancer agents (Figure 5B).
ation in HCT116 cells. While further improvements in
potency are still required to reach a suitable drug candidate,
our study, provides an important proof-of-concept for the
usefulness of CSN5 inhibitors as anticancer agents.
Keywords: azaindoles · CSN5 · ubiquitin ligases · inhibitors ·
metalloproteases
[1] a) M. D. Petroski, R. J. Deshaies, Nat. Rev. Mol. Cell Biol. 2005,
6, 9; b) R. J. Deshaies, C. A. Joazeiro, Annu. Rev. Biochem. 2009,
78, 399.
[2] N. W. Pierce, J. E. Lee, X. Liu, M. J. Sweredoski, R. L. J.
Graham, E. A. Larimore, M. Rome, N. Zheng, B. E. Clurman,
S. Hess, S. Shan, R. J. Deshaies, Cell 2014, 153, 161.
[3] G. M. Lingaraju, R. D. Bunker, S. Cavadini, D. Hess, U.
Hassiepen, M. Renatus, E. S. Fischer, N. H. Thomꢁ, Nature
2014, 512, 161.
[4] a) G. A. Cope, R. J. Deshaies, BMC Biochem. 2006, 7, 1; b) R. J.
Deshaies, E. D. Emberley, A. Saha, Subcell. Biochem. 2010, 54,
41; c) T. Schmaler, W. Dubiel, in Conjugation and deconjugation
of ubiquitin family modifiers (Ed: M. Groettrup), Springer, New
York, 2010, pp. 57 – 68; d) D. M. Duda, D. C. Scott, M. F.
Calabrese, E. S. Zimmerman, N. Zheng, B. A. Schulman, Curr.
Opin. Struct. Biol. 2011, 21, 257; e) E. D. Emberley, R. Mosa-
deghi, R. J. Deshaies, Biol. Chem. 2012, 287, 29679.
[5] A. Echalier, Y. Pan, M. Birol, N. Tarvernier, L. Pintard, F. Hoh,
C. Ebel, N. Galophe, F. X. Claret, C. Dumas, Proc. Natl. Acad.
Sci. USA 2013, 110, 1273.
[6] a) M.-H. Lee, R. Zhao, L. Phan, S.-C. J. Yeung, Cell Cycle 2011,
10, 3057.
[7] Y.-H. Lee, A. D. Judge, D. Seo, M. Kitade, L. E. Gomez-Quiroz,
T. Ishikawa, J. B. Andersen, B.-K. Kim, J. U. Marquardt, C.
Raggi, I. Avital, E. A. Conner, I. MacLachlan, V. M. Factor, S. S.
Thorgeirsson, Oncogene 2011, 30, 4175.
[8] T. J. Shackleford, F. X. Claret, Cell Div. 2010, 5, 26.
[9] J. R. Skaar, J. K. Pagan, M. Pagano, Nat. Rev. Drug Discovery
2014, 13, 889.
[10] A. Schlierf, E. Altmann, J. Quancard, A. B. Jefferson, R.
Assenberg, M. Renatus, M. Jones, U. Hassiepen, M. Schaefer,
M. Kiffe, A. Weiss, C. Wiesmann, R. Sedrani, J. Eder, B.
Martoglio, Nat. Commun. 2016, 7, 13166.
Figure 5. A) Biological effects of the CSN5 inhibitors. An immunoblot
is shown for Cul1 and Skp2 after treatment of HCT116 cells with
compounds 5 and 6 at the indicated concentrations. B) Inhibition of
cell proliferation with CSN5 inhibitors 5 and 6. Individual dose–
response curves for HCT116 cells treated with 5 (blue points) and 6
(red triangles), n=2, results are shown as the meanÆSD. The
experimental data were fitted using the nonlinear regression program
Origin 8.5.0 SR1 (OriginLab Corporation, Northampton, MA). For
details, see the Supporting Information.
[11] J. R. Lakowicz, Principles of Fluorescence Spectroscopy,
Springer, New York, 2006, p. 1.
[12] S. Cavadini, E. S. Fischer, R. D. Bunker, A. Potenza, G. M.
Lingaraju, K. N. Goldie, W. I. Mohamed, M. Faty, G. Petzold,
R. E. J. Beckwith, R. B. Tichkule, U. Hassiepen, W. Abdulrah-
man, R. S. Pantelic, S. Matsumoto, K. Sugasawa, H. Stahlberg,
N. H. Thomꢁ, Nature 2016, 531, 598.
In summary we have discovered a novel class of active-
site-directed, small-molecule inhibitors of the JAMM pro-
tease CSN5. The X-ray co-crystal structure of one these
compounds with CSN5 revealed an unprecedented binding
mode, with the 7-azaindole core coordinating the active-site
Zn2+ ion through N7, and the indole NH engaging in
a hydrogen bond with the D151 side-chain carboxylate. The
optimized inhibitors are selective for CSN5 over other human
metalloproteases. They recapitulate the effects observed for
CSN5 knockdown by siRNA and they inhibit cell prolifer-
[13] K. Kawai, N. Nagata, Eur. J. Med. Chem. 2012, 51, 271.
[14] A. M. Clark, P. Labute, M. Santavy, J. Chem. Inf. Model. 2006,
46, 1107.
[15] C. Tallant, A. Marrero, F. X. Gomis-Rꢂth, Biochem. Biophys.
Acta Mol. Cell Res. 2010, 1803, 20.
Manuscript received: September 5, 2016
Revised: October 27, 2016
Final Article published: && &&, &&&&
4
www.angewandte.org
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2016, 55, 1 – 5
These are not the final page numbers!

Angewandte
Communications
Chemie
Communications
Inhibitors
CSN5 is the zinc metalloprotease subunit
of the COP9 signalosome, which is an
important regulator of cullin-RING E3
ubiquitin ligases. A new class of CSN5
inhibitors (orange) was discovered and
shown to interact with the active-site zinc
ion in an unprecedented binding mode.
Optimized nanomolar CSN5 inhibitors
result in decreased levels of the onco-
protein Skp2 and reduce the viability of
HCT116 cells.
E. Altmann,* P. Erbel, M. Renatus,
M. Schaefer, A. Schlierf, A. Druet,
L. Kieffer, M. Sorge, K. Pfister,
U. Hassiepen, M. Jones, S. Ruedisser,
D. Ostermeier, B. Martoglio,
A. B. Jefferson,
J. Quancard
&&&& — &&&&
Azaindoles as Zinc-Binding Small-
Molecule Inhibitors of the JAMM
Protease CSN5
Angew. Chem. Int. Ed. 2016, 55, 1 – 5
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
5
These are not the final page numbers!