Angewandte
Chemie
philes (1-CA, 1-FA, 1-AA, 1-EO, 1-AZ) displayed a prefer-
ence for b5i (Table ST1 and Figure S1 in the Supporting
Information). The most potent electrophile in the screen was
1
-CA, which showed substantial activity against b5i (IC =
50
1
.24 mm, Figure 2). This is in agreement with studies that
propose a-chloroacetamides for sustained targeting of non-
[10,15]
catalytic cysteines.
Despite the unselective peptide back-
bone of 1-CA, it displayed nine-fold selectivity for b5i (b5c/
b5i = 9), which is comparable to that of ONX 0914 (b5c/b5i
[
5]
ꢀ
10). Importantly, 1-CA was inactive against the subunits
b1c, b1i, b2c, and b2i (IC > 100 mm, Table ST2). The more
50
stable 1-FA, as well as the unreactive congener 1-PA, showed
significantly decreased IC50 values compared to 1-CA, and
both blocked the activity of b5i (IC = 36.25 mm and 24.23 mm,
50
respectively) and b5c (IC = 43.84 mm and 29.05 mm, respec-
50
tively) to the same extent (b5c/b5i = 1.2, Figure 2 and
Table ST1). These findings indicate that the b5i binding
affinity originates from the a-chloroacetamide electrophile
forming a covalent thioether with Cys48 (Scheme S2).
Next, we aimed to assess the covalent binding mode of 1-
CA by X-ray analysis. Since mammalian iCPs are challenging
to crystallize, we mimicked the S4 pocket of b5i by replacing
Gly48 of the yeast proteasome subunit yb5 with Cys48 in
a plasmid-shuffling procedure (Figure S3a). Subsequent crys-
tallization and structure elucidation of the yb5G48C mutant
yCP (2.8 resolution, Rfree = 20.1%, PDB ID: 5CGF,
Table ST4) revealed an orientation of Cys48 identical to
that observed in iCP from mouse. In addition, elucidation of
the yb5G48C:ONX 0914 complex structure (2.8 resolution,
Rfree = 20.6%, PDB ID: 5CGI) showed a conformation of the
ligand analogous to that observed in mb5i (Figure S4, S5).
Strikingly, soaking of yb5G48C yCP crystals with 1-CA
followed by X-ray analysis (2.9 resolution, Rfree = 23.1%,
PDB ID: 5CGG) displayed the ligand exclusively bound to
the mutant yb5 subunit. 1-CA occupied the substrate binding
channel by adopting an antiparallel b-sheet in a similar
manner to known inhibitors that are based on decarboxylated
Figure 2. Schematic representation of the substrate binding channel of
b5i with the specificity pockets S1–S4 and Cys48 (underlined). The
CFZ-derived decarboxylated peptides 1 contain distinct P4 side-chain
electrophiles (R, gray): a-chloroacetamide (1-CA), a-fluoroacetamide
(1-FA), acrylamide (1-AA), vinyl sulfonamide (1-VS), epoxide (1-EO),
and aziridine (1-AZ). Corresponding non-reactive controls: propiona-
mide (1-PA), ethylsulfonamide (1-EA), cyclopropanamide (1-CP). A
complete list of compounds, including half-maximal inhibitory concen-
tration (IC ) values, can be found in the supporting information
5
0
(Tables ST1–3). The lower panel shows the a-chloroacetamides 1-CA–
4
-CA with their corresponding P2 and P3 residues and the control, 1-
PA. The in vitro IC50 values were determined by using purified human
iCP or cCP. [a] A high IC50 b5c/b5i ratio indicates selectivity for b5i.
Cys48 participates in forming the substrate binding channel
of b5i by partially shaping the S2 and S4 pockets. According to
structural superpositions, it is accessible via the P4 side chains
of tetrapeptides (Figure 1). Consequently, we initiated our
inhibitor design by exchanging the P4 residue of CFZ with l-
[13,14]
peptides.
In fact, the structure revealed continuous
electron density connecting the acetamide function of the
P4 side chain of 1-CA to the thiol group of the introduced
Cys48 (Figure 3a). This linkage confirms a covalent mode of
action and explains the nine-fold selectivity of 1-CA for b5i.
In contrast, soaking of wild-type yCP crystals as a model for
cCP showed empty yb5 substrate channels, thus emphasizing
the importance of Cys48 for 1-CA binding.
2
,3-diaminopropionic acid (Dap). Dap suits the steric require-
ments of S4 and allows the late-stage introduction of electro-
philes owing to its side-chain amino function (Figure 2).
The CFZ-inspired peptide backbone was prepared by
solid-phase peptide synthesis using the Fmoc strategy and was
C-terminally capped with the previously described 4-methyl-
Based on these results, we optimized the peptidic back-
bone to improve b5i selectivity. As a starting point, we used
the peptide composition of ONX 0914 as a molecular blue-
print and generated 2-CA (Figure 2 and Table ST2). Unex-
pectedly, 2-CA showed decreased potency against human b5i
(IC = 6.65 mm) and did not bind to the yb5G48C mutant in
[
13,14]
benzyl amine.
In the final step, we introduced various
electrophiles by utilizing the corresponding acid chlorides, N-
hydroxysuccinimide esters, or carboxylic acids in amide
coupling reactions. This straightforward synthesis was used
to generate a set of decarboxylated peptides with diverse side-
chain electrophiles that were shown to be suitable for
targeting soft thiol nucleophiles (1-CA, 1-FA, 1-AA, 1-VS,
50
soaking experiments. To understand this drop in potency, we
compared the binding mode of ONX 0914 with that of 1-CA
and found pronounced differences: the structure of ONX
0914 bound to yb5G48C revealed a distinct orientation of the
P2-TyrOMe that facilitates attractive sulfur–arene interac-
[
9]
1
-EO, and 1-AZ; Figure 2). As controls, we prepared their
unreactive congeners 1-PA, 1-EA, and 1-CP (Figure 2).
Our initial screening efforts using human iCP and cCP
showed that sulfonamide compounds (1-VS, 1-EA) are slightly
selective for b5c, whereas amide-bond-connected electro-
[16,17]
tions with Cys48.
In contrast, the P2-Phe of 1-CA is
displaced by the P4 side chain, which covalently binds to
Cys48, thereby restricting the S2 pocket. To probe the
Angew. Chem. Int. Ed. 2015, 54, 15888 –15891
ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim