A Minimal β-Lactone Fragment for Selective β5c or β5i Proteasome Inhibitors

Article abstract of DOI:10.1002/anie.201502931

Broad-spectrum proteasome inhibitors are applied as anticancer drugs, whereas selective blockage of the immunoproteasome represents a promising therapeutic rationale for autoimmune diseases. We here aimed at identifying minimal structural elements that confer β5c or β5i selectivity on proteasome inhibitors. Based on the natural product belactosin C, we synthesized two β-lactones featuring a dimethoxybenzyl moiety and either a methylpropyl (pseudo-isoleucin) or an isopropyl (pseudo-valine) P1 side chain. Although the two compounds differ only by one methyl group, the isoleucine analogue is six times more potent for β5i (IC₅₀=14 nM) than the valine counterpart. Cell culture experiments demonstrate the cell-permeability of the compounds and X-ray crystallography data highlight them as minimal fragments that occupy primed and non-primed pockets of the active sites of the proteasome. Together, these results qualify β-lactones as a promising lead-structure motif for potent nonpeptidic proteasome inhibitors with diverse pharmaceutical applications.

Full text of DOI:10.1002/anie.201502931

.
Angewandte
Communications
International Edition: DOI: 10.1002/anie.201502931
German Edition: DOI: 10.1002/ange.201502931
Proteasome Inhibitors
A Minimal b-Lactone Fragment for Selective b5c or b5i Proteasome
Inhibitors**
Michael Groll,* Vadim S. Korotkov, Eva M. Huber, Armin de Meijere, and Antje Ludwig*
Abstract: Broad-spectrum proteasome inhibitors are applied
as anticancer drugs, whereas selective blockage of the immu-
noproteasome represents a promising therapeutic rationale for
autoimmune diseases. We here aimed at identifying minimal
structural elements that confer b5c or b5i selectivity on
proteasome inhibitors. Based on the natural product belacto-
sin C, we synthesized two b-lactones featuring a dimethoxy-
benzyl moiety and either a methylpropyl (pseudo-isoleucin) or
an isopropyl (pseudo-valine) P1 side chain. Although the two
compounds differ only by one methyl group, the isoleucine
analogue is six times more potent for b5i (IC50 = 14 nm) than
the valine counterpart. Cell culture experiments demonstrate
the cell-permeability of the compounds and X-ray crystallog-
raphy data highlight them as minimal fragments that occupy
primed and non-primed pockets of the active sites of the
proteasome. Together, these results qualify b-lactones as
a promising lead-structure motif for potent nonpeptidic
proteasome inhibitors with diverse pharmaceutical applica-
tions.
of each of the CPs and peptide bond cleavage by all of the
active sites follows a universal principle. It is thus the unique
chemical nature of the specificity (S) pockets that gives rise to
distinct substrate preferences through interaction with the
side chains of the ligand (P sites).[3] As a rule, the cleavage
specificities of the three proteasomal active sites are assigned
by using chromogenic screening substrates and relate to the
preferred P1 amino acid. In this way, subunit b1 (caspase-like;
C-L) was determined to cleave after acidic residues, b2
(trypsin-like; T-L) after basic residues, and b5 (chymotrypsin-
like; ChT-L) after non-polar residues.[4]
Increased expression levels of the iCP or its active
subunits b1i, b2i, and b5i have recently been associated with
the development and progression of neurodegenerative
diseases,[5] autoimmune disorders,[6] and certain types of
cancers.[7] ONX 0914, a b5i-selective peptidic a’,b’-epoxyke-
tone inhibitor, has been shown to suppress autoreactive
immune responses in mouse models of rheumatoid arthritis,
lupus, and experimental encephalomyelitis by preventing the
production of interleukins and the release of interferon-g.[8]
Crystallographic data for the murine cCP and iCP identified
the differently sized S1 pockets in b5c and b5i, formed by
unique orientations of Met45, as the major determinant of the
enhanced affinity of ONX 0914 for b5i. Based on the
observation that ligands dock via their P1 residue to the
active site, compounds with large P1 residues were supposed
to favour b5i, whereas small P1 side chains were suggested to
be more appropriate for b5c.[9]
Successive drug design efforts have succeeded in improv-
ing existing b5i inhibitors by exchanging their P1-phenyl-
alanine moiety for a cyclohexyl residue.[10] However, all of
these currently available proteasome inhibitors feature a pep-
tide-based backbone that only targets the non-primed sites of
the substrate binding channel shaped by the subunits b5 and
b6. Considering that subunit b6 is unchanged between cCP
and iCP, pronounced interactions of the ligand with b6 are
supposed to hamper b5i selectivity. We therefore aimed at
developing small nonpeptidic proteasome inhibitors that
target both the primed and non-primed sites solely of subunit
b5. As a scaffold, we used the bisbenzyl-protected homobe-
lactosin C (1; Figure 1A), a derivative of the natural product
b-lactone belactosin C from Streptomyces sp. UCK14.[11]
Unlike the family of b-lactone-g-lactam inhibitors, including
the microbial metabolite omuralide (clasto-lactacystin, 2;
Figure 1A),[4a,12] crystallographic data demonstrated that
1 only binds to b5, even at concentrations as high as
20 mm.[13] Notably and in contrast to 2, belactosin C and its
derivatives occupy both the S1 specificity pocket and the
previously neglected primed substrate binding channel,
including the adjacent subunit b4.[13,14] In order to address
T
he eukaryotic 20S proteasome (core particle, CP), a multi-
catalytic protease of 720 kDa that is responsible for non-
lysosomal protein degradation, is a key player in many
cellular processes, including cell cycle progression and the
immune response.[1] Whereas yeast harbors one 20S protea-
some (yCP), there are three individual CP types in verte-
brates and they differ only in the composition of their
proteolytically active subunits: the constitutive proteasome
(cCP), the immunoproteasome (iCP), and the thymoprotea-
some (tCP).[2] Protein breakdown occurs in the central lumen
[*] Prof. Dr. M. Groll, Dr. V. S. Korotkov, Dr. E. M. Huber
Center for Integrated Protein Science at the Department Chemie
Technische Universität München
Lichtenbergstrasse 4, 85748 Garching (Germany)
E-mail: michael.groll@tum.de
Dr. A. Ludwig
CharitØ Universitätsmedizin Berlin CCM
Medizinische Klinik für Kardiologie und Angiologie
CharitØplatz 1, 10117 Berlin (Germany)
E-mail: antje.ludwig@charite.de
Prof. Dr. A. de Meijere
Institut für Organische und Biomolekulare Chemie der
Georg-August-Universität Gçttingen
Tammannstrasse 2, 37077 Gçttingen (Germany)
[**] This work was supported by the Deutsche Forschungsgemeinschaft
(DFG, grant GR1861/10-1 to M.G.). A. Kçnig is acknowledged for
excellent technical support. We are grateful to the staff of the
beamline BW6, DESY, Hamburg (Germany).
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.201502931.
7810
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 7810 –7814

Angewandte
Chemie
Figure 1. A) Structures of bisbenzylated homobelactosin C (1), omuralide (2), and the analyzed b-lactones 3–5. B) Crystal structure of yCP:3 at the
yb5 active site (PDB ID 4Z1L). The 2FOÀFC electron density map (blue mesh, contoured at 1.0 s) is shown for 3 and the active site Thr1; dotted
lines mark hydrogen bonds to the oxyanion hole Gly47NH and to Thr1NH2. Met45 forms the bottom of the S1 pocket and determines the affinity
and specificity of inhibitor binding. C) The superposition of 1,[13] 3, and MG132[16] at yb5 shows that the peptidic compound MG132 solely binds
to the non-primed sites S1–S3, whereas homobelactosin C (1) and its derivative (3) occupy the S1 pocket (the P1 site of the ligand is marked by
a dotted half circle) and the primed pockets. D) The superposition of yeast and mouse b5 subunits reveals the distinct conformations of Met45.[9]
The black arrow marks the reorientation of Met45 upon binding of 3 to yb5. The Ile-like P1 site of 3 (marked by a dotted half circle) undergoes
hydrophobic interactions with all Met45 conformations and hence does not discriminate between b5c and b5i. E) Connolly surface representations
of mouse b5c and b5i active sites with 3 modelled by the superposition of yb5:3. Positive and negative electrostatic surface potentials are
contoured from À30 kTeÀ1 (red) to 30 kTeÀ1 (blue). For clarity, residues 47 and 48 of b5 have been removed. The S1 pockets are indicted by
a white line to illustrate their different sizes in b5c and b5i (see also Figure S1C). It should be noted that binding of the P1-Ile side chain to b5c
requires the reorientation of Met45 as in yb5 in panel D to avoid clashing.
solely subunit b5 and to identify structural elements that
discriminate between b5c and b5i, we investigated a series of
belactosin C derivatives.[14,15] In the course of these inves-
tigations, we found a minimal scaffold that is suitable for high-
affinity binding to b5. These fragments feature a N-(3,5-
dimethoxybenzyl)aminocarbonyl side chain at the C2 posi-
tion of either a (3S)-((1S)-methylpropyl)- (isoleucine mimetic
3; Figure 1A) or a (3S)-(isopropyl)-4-oxooxetane-(2R)-car-
boxamide (valine mimetic 4; Figure 1A) moiety.[15]
X-ray crystallographic analysis of 3 in complex with the
yCP was performed according to our standard protocols[17]
and yielded data to 3.0 Š resolution with an Rfree of 22.9%
(see Table S1 in the Supporting Information). The electron
density map displayed the inhibitor covalently bound to the
N-terminal Thr1Og of all of the active sites owing to the high
ligand concentrations used for crystal soaking (Figure 1B and
Figure S1A in the Supporting Information). Interestingly,
superposition of 3 bound to yb1, yb2, and yb5 shows that the
ligand adopts a uniform orientation in each substrate binding
channel (Figure S1B). The isoleucine P1 side chain protrudes
into the S1 specificity pocket and causes a 1 Š shift of Met45,
whereas the C2-OH group of the ligand, generated through
nucleophilic attack of Thr1Og on the b-lactone ring, forms
a strong hydrogen bond to the carbonyl oxygen of residue 19
(2.7 Š), as observed for homobelactosin C.[13,18] The 3,5-
dimethoxybenzyl group points towards the primed substrate
binding channel and is only stabilized by hydrophobic
interactions with Ca and Cb of Ser129, a strictly conserved
residue involved in catalysis.[13] In conclusion, the crystallo-
graphic investigations of 3 provide first insight into the primed
substrate binding channels of the C-L and T-L active sites.
The common binding mode of 3 with all three active sites
implies that the subunit selectivity of compounds mainly
depends on their interactions with the primed and non-
primed specificity pockets of the individual binding clefts. The
described minimal scaffold required to orient ligands to the
Angew. Chem. Int. Ed. 2015, 54, 7810 –7814
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
7811

.
Angewandte
Communications
Figure 2. A) Dose–response curves for 3- and 4-mediated inhibition of the ChT-L activities of human cCP and iCP. Experiments were performed in
triplicate and normalized to a DMSO-treated control. B) Inhibition profiles of 3 and 4 in lysates of HUVEC (cCP) and monocytic THP-
1 (predominantly iCP) cells. Values are given as the mean percentage of residual activityÆstandard error of the mean (SEM).
primed site of all active b subunits is now amenable to future
combinatorial chemistry to create selective compounds for
individual active sites.
Supporting Information). In vitro activity assays in cell lysates
revealed that 3 and 4 dose-dependently inhibit proteasome-
mediated protein degradation and predominantly affect ChT-
L activity (Figure 3A). As expected, the CP inhibition results
in a breakdown in overall proteasomal function, thereby
causing an accumulation of a destabilized GFP reporter in ear
fibroblasts of the UbiquitinG76V-GFP1-mouse (Figure 3B,
upper panel) and an increase in polyubiquitylated proteins
(Figure 3B, lower panel). We quantified the cytotoxicity by
treating HeLa cells with 500 nm and 1 mm of 3 and 4 in an
MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium
bromide) cell proliferation assay for 24 h. Remarkably, cell
viability was twice as high upon the application of 3 and 4
compared to the standard proteasome inhibitor MG132
(Figure 3C), a significant and important feature for the
future development of selective CP-type drugs for non-cancer
applications such as inflammation or autoimmune diseases.
Moreover, we performed flow cytometry cell cycle analysis
with HeLa cells incubated with both compounds at a concen-
tration of 500 nm for 24 h (see Supporting Information).
Distinct sub-G1 peaks (corresponding to apoptotic cells) were
observed after treatment with 3 (8.91 Æ 0.44% of all cells), 4
(10.03 Æ 1.08%), and MG132 (19.68 Æ 2.04%). The increased
numbers of cells in the G2/M phase demonstrate that both
compounds induce cell cycle arrest in HeLa cells (control:
12.31 Æ 0.84%, 3: 23.78 Æ 0.49% and 4: 23.36 Æ 2.62% of all
non-apoptotic cells), albeit to a much less pronounced degree
than MG132 (40.36 Æ 2.87%). In addition to 3 and 4, we
tested the b-lactones (2R, 3S)-5 and (2S, 3R)-5[14,15] in our
biological assays. Both compounds feature an additional
methyl group at the C2 position of compound 3, which was
derived from the natural product Salinosporamide A (mar-
izomib),[19] a second generation proteasome inhibitor cur-
rently being evaluated in clinical trials. Surprisingly, the
Next, we characterized the belactosin derivatives 3 and 4,
which only differ in a single P1-methyl group, for their
inhibitory potency towards human cCP and iCP. Indeed, 3 and
4 dose-dependently block the ChT-L activity of b5c and b5i
owing to their apolar moieties (Figure 2A). Although both
compounds feature no P2, P3, or P4 sites, they are highly
effective. Interestingly, the half maximal inhibitory concen-
tration (IC50) values for 3 are quite similar for b5c (21.35 nm)
and b5i (14.37 nm), whereas 4 turned out to be more selective
for b5c (26.87 nm) than for b5i (83.62 nm). Hence, the single
P1-methyl group that distinguishes 3 from 4 is decisive in
terms of subunit selectivity. Additionally we performed
inhibition assays with lysates from human umbilical vein
endothelial cells (HUVEC, which contain predominantly
cCP) and THP-1 cells (a human monocytic leukemia cell line
containing predominantly iCP). Increasing concentrations of
3 blocked ChT-L activity in lysates of both cell types to
a similar degree, while 4 was more potent in HUVEC lysates
(Figure 2B). Although the tested hydrophobic P1 residues
are accepted by both b5c and b5i, the herein presented results
prove that the ChT-L activity of b5i demands more bulky
amino acids compared to b5c and disfavors smaller ones like
Val.[9] The peculiar conformation of Met45 in subunit b5i
enlarges the S1 pocket and cannot properly stabilize the
smaller P1-Val side chain of 4 by van der Waals contacts
(Figure 1D,E).
In addition, we investigated the effect of 3 and 4 on
proteasome activity in living mammalian cells. HeLa cells
were treated for 4 h with varying concentrations of 3, 4 and
the standard proteasome inhibitor Z-Leu-Leu-Leu-al
(MG132); DMSO was used as the solvent control (see the
7812
www.angewandte.org
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 7810 –7814

Angewandte
Chemie
Figure 3. A) Inhibition of the ChT-L, T-L, and C-L activities by 3 and 4 compared to MG132 in HeLa cells. Relative activities are expressed as
percentages normalized to a DMSO-treated control. B) UbiquitinG76V-GFP1 in ear fibroblasts of transgenic mice and ubiquitylated proteins
markedly accumulate in the presence of increasing concentrations of 3, 4, and MG132 compared to DMSO alone. C) Staining was performed by
western blot with anti-GFP and anti-ubiquitin (anti-Ubi) antibodies, respectively. Amido black staining of western blot membranes served as
a control for equal protein loading (LC). C) The viability of HeLa cells exposed to 3, 4, or MG132 was assessed by a MTT assay. Cell cycle related
changes (number of cells in G1, S and G2/M phases) and apoptotic cells (number of cells in sub-G1) were detected by flow cytometry analysis of
the DNA content and normalized to a DMSO-treated control. For panels (A) and (C) values are given as the means of three independent
experimentsÆSEM.
diastereomers (2R, 3S)-5 and (2S, 3R)-5 did not inhibit the
proteasome in HeLa cells and were inactive in the MTTassay
(Figure S2). These results prove the robustness of the
presented data and demonstrate that a single methyl group
can be decisive not only for subunit selectivity but also for the
entire biological activity of a compound.
[1] E. M. Huber, M. Groll, Angew. Chem. Int. Ed. 2012, 51, 8708 –
8720; Angew. Chem. 2012, 124, 8838 – 8850.
[2] S. Murata, K. Sasaki, T. Kishimoto, S. Niwa, H. Hayashi, Y.
Takahama, K. Tanaka, Science 2007, 316, 1349 – 1353.
[3] L. Borissenko, M. Groll, Chem. Rev. 2007, 107, 687 – 717.
[4] a) M. Groll, L. Ditzel, J. Lçwe, D. Stock, M. Bochtler, H. D.
Bartunik, R. Huber, Nature 1997, 386, 463 – 471; b) M. Orlowski,
C. Cardozo, C. Michaud, Biochemistry 1993, 32, 1563 – 1572.
[5] a) M. Mishto, E. Bellavista, A. Santoro, A. Stolzing, C. Ligorio,
B. Nacmias, L. Spazzafumo, M. Chiappelli, F. Licastro, S. Sorbi,
A. Pession, T. Ohm, T. Grune, C. Franceschi, Neurobiol. Aging
2006, 27, 54 – 66; b) M. Diaz-Hernandez, F. Hernandez, E.
Martin-Aparicio, P. Gomez-Ramos, M. A. Moran, J. G. Castano,
I. Ferrer, J. Avila, J. J. Lucas, J. Neurosci. 2003, 23, 11653 – 11661.
[6] a) A. Visekruna, N. Slavova, S. Dullat, J. Grone, A. J. Kroesen,
J. P. Ritz, H. J. Buhr, U. Steinhoff, Int. J. Colorectal. Dis. 2009, 24,
1133 – 1139; b) L. R. Fitzpatrick, V. Khare, J. S. Small, W. A.
Koltun, Dig. Dis. Sci. 2006, 51, 1269 – 1276.
In conclusion, we identified minimal structural elements
that distinguish between b5c and b5i. Using the b-lactone
scaffold as a streamlined proteasome inhibition motif, we
synthesized two belactosin derivatives with valine- and
isoleucine-like P1 side chains that exhibit exquisite potency
in the low nanomolar range. We found that a single P1-methyl
group creates a six-fold difference in potency for b5i.
Selective inhibition can thus be achieved solely through the
P1 site, a result of crucial importance for the future design of
proteasome inhibitors. Furthermore, the reduced cytotoxicity
compared to currently applied peptide proteasome inhibitors
qualify these b-lactones as promising lead structures for
applications in inflammatory diseases. In particular, derivati-
zation of the dimethoxybenzyl side chain offers a great
opportunity to extend these scaffolds towards the previously
neglected primed substrate binding channel of b5c and b5i, as
well as the other proteasomal active sites.
[7] Y. K. Ho, P. Bargagna-Mohan, M. Wehenkel, R. Mohan, K. B.
Kim, Chem. Biol. 2007, 14, 419 – 430.
[8] a) T. Muchamuel, M. Basler, M. A. Aujay, E. Suzuki, K. W.
Kalim, C. Lauer, C. Sylvain, E. R. Ring, J. Shields, J. Jiang, P.
Shwonek, F. Parlati, S. D. Demo, M. K. Bennett, C. J. Kirk, M.
Groettrup, Nat. Med. 2009, 15, 781 – 787; b) M. Basler, S. Mundt,
T. Muchamuel, C. Moll, J. Jiang, M. Groettrup, C. J. Kirk,
EMBO Mol. Med. 2014, 6, 226 – 238; c) H. T. Ichikawa, T.
Conley, T. Muchamuel, J. Jiang, S. Lee, T. Owen, J. Barnard, S.
Nevarez, B. I. Goldman, C. J. Kirk, R. J. Looney, J. H. Anolik,
Arthritis Rheum. 2011, 64, 493 – 503.
Keywords: cytotoxicity · immunoproteasome · inhibitors ·
lead structures · b-lactones
[9] E. M. Huber, M. Basler, R. Schwab, W. Heinemeyer, C. J. Kirk,
M. Groettrup, M. Groll, Cell 2012, 148, 727 – 738.
[10] G. de Bruin, E. M. Huber, B. T. Xin, E. J. van Rooden, K. Al-
Ayed, K. B. Kim, A. F. Kisselev, C. Driessen, M. van der Stelt,
How to cite: Angew. Chem. Int. Ed. 2015, 54, 7810–7814
Angew. Chem. 2015, 127, 7921–7925
Angew. Chem. Int. Ed. 2015, 54, 7810 –7814
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
7813

.
Angewandte
Communications
G. A. van der Marel, M. Groll, H. S. Overkleeft, J. Med. Chem.
2014, 57, 6197 – 6209.
[11] a) A. Asai, A. Hasegawa, K. Ochiai, Y. Yamashita, T. Mizukami,
J. Antibiot. 2000, 53, 81 – 83; b) O. V. Larionov, A. de Meijere,
Org Lett. 2004, 6, 2153 – 2156.
[12] a) G. Fenteany, R. F. Standaert, W. S. Lane, S. Choi, E. J. Corey,
S. L. Schreiber, Science 1995, 268, 726 – 731; b) M. Groll, E. P.
Balskus, E. N. Jacobsen, J. Am. Chem. Soc. 2008, 130, 14981 –
14983; c) A. List, E. Zeiler, N. Gallastegui, M. Rusch, C.
Hedberg, S. A. Sieber, M. Groll, Angew. Chem. Int. Ed. 2014, 53,
571 – 574; Angew. Chem. 2014, 126, 582 – 585.
[15] A. de Meijere, V. S. Korotkov, A. V. Lygin, O. V. Larionov, V. V.
Sokolov, T. Graef, M. Es-Sayed, Org. Biomol. Chem. 2012, 10,
6363 – 6374.
[16] M. L. Stein, H. Cui, P. Beck, C. Dubiella, C. Voss, A. Kruger, B.
Schmidt, M. Groll, Angew. Chem. Int. Ed. 2014, 53, 1679 – 1683;
Angew. Chem. 2014, 126, 1705 – 1709.
[17] N. Gallastegui, M. Groll, Methods Mol. Biol. 2012, 832, 373 – 390.
[18] S. Kawamura, Y. Unno, A. List, A. Mizuno, M. Tanaka, T.
Sasaki, M. Arisawa, A. Asai, M. Groll, S. Shuto, J. Med. Chem.
2013, 56, 3689 – 3700.
[19] a) M. Groll, B. C. Potts, Curr. Top. Med. Chem. 2011, 11, 2850 –
2878; b) M. Groll, R. Huber, B. C. Potts, J. Am. Chem. Soc. 2006,
128, 5136 – 5141.
[13] M. Groll, O. V. Larionov, R. Huber, A. de Meijere, Proc. Natl.
Acad. Sci. USA 2006, 103, 4576 – 4579.
[14] V. S. Korotkov, A. Ludwig, O. V. Larionov, A. V. Lygin, M. Groll,
A. de Meijere, Org. Biomol. Chem. 2011, 9, 7791 – 7798.
Received: March 30, 2015
Published online: May 14, 2015
7814
www.angewandte.org
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 7810 –7814