E. Bouffard, B.W. Zaro, M.M. Dix et al.
Tetrahedron Letters xxx (xxxx) xxx
N
R1
N
N
NH
R3
N
NH
N
N
MeO
R2
O
O
N
H
N
H
N
AZD9291
NH
N
R1 =
R2 = H, OMe
R3 = H or
N
N
N
N
O
N
N
N
O
N
Fig. 1. Structure of AZD9291 and library compounds.
magnesium bromide, on the 4-position of the 2,4-dichloropyrimi-
dine. The second chlorine was then substituted by various
nitroanilines bearing either a hydrogen or a methoxy group as R2
and either a hydrogen or a fluorine in the 4 position. In the latter
case, the fluorine was substituted by different amines as R3 groups.
The nitro group was reduced into an amine, using hydrogen and
palladium on carbon or iron powder and ammonium chloride, in
order to introduce the Michael acceptor moiety (Fig. 2).
acceptor. After elimination of the chlorine with triethylamine in
acetonitrile, the final compounds were obtained (Fig. 3f).
Using this synthetic route, the following compounds were
obtained (Fig. 4).
The in situ activity of the various analogs was evaluated using
the NSCLC lung cancer line H1975, which harbors a T790M-EGFR
variant, by competitive activity-based protein profiling (ABPP)
[3]. Cells were first treated with an inhibitor analog or DMSO vehi-
cle (1 and 10 mM). After 2 h, the cells were treated with the alkyne-
bearing AZD9291 probe (Fig. 5). After 4 h, cells were lysed, and
membrane and soluble proteomes isolated and subject to treat-
ment with rhodamine-azide by copper(I)-catalyzed azide-alkyne
cycloaddition [7], as described previously [4]. Probe-labeled pro-
teins were then separated by SDS-PAGE and visualized by in-gel
fluorescence scanning (Fig. 6). In this competitive ABPP experi-
ment, a compound is considered inhibitory against T790M-EGFR
or off-targets if it blocks the AZD9291 probe labeling of these pro-
teins at 1 or 10 mM test concentrations (reflected in a loss of in-gel
fluorescence signal for these proteins).
These studies demonstrated that replacing the diamino group
with other basic amines had variable impact on EGFR and cathep-
sin reactivity, but did not furnish analogs with superior selectivity.
In contrast, an analog where the diamino group was removed along
with removal of the N-methyl of the indole unit (EB037) main-
tained T790M-EGFR engagement at 1 and 10 mM while showing
reduced reactivity with cathepsins. In contrast, analog EB017,
which also lacked the diamino group, but maintained the N-methyl
indole lost interactions with both EGFR and cathepsins. We should
also note that other off-targets of AZD9291 were retained with
EB037 (e.g., ~100 kDa soluble protein).
The introduction of the acrylamide group in the final step
proved to be more tedious than expected for some of the analogs.
With only one equivalent of acryloyl chloride, the reaction was
incomplete and the starting material and the product could not
be separated (Fig. 3a). The use of a slight excess of acryloyl chloride
lead to a mixture of product and a by-product bearing 2 acrylamide
groups, also non-separable (Fig. 3b). The use of acrylic acid and a
coupling agent, T3P, gave the product with non-separable
impurities. The same issue occurred when using acrylic acid and
triphenylphosphite (Fig. 3c). To avoid the formation of the bis-
acrylamide compound, the synthesis route was modified to protect
the secondary amine with a Boc group. However, during the depro-
tection of the Boc group, the acrylamide was partially cleaved lead-
ing to a mixture of products that were not separable (Fig. 3d). We
also used tert-butyl acrylate and ethyl acrylate in an attempt to
achieve selective introduction on the primary amine, but no reac-
tion was observed (Fig. 3e). We then decided to use 2 equivalents
of acryloyl chloride to introduce the acrylamide and hydrolyze the
acrylamide on the guanidine-like nitrogen. Several acidic
conditions were tested and a solution of HCl 3 M in AcOEt was
determined to be optimal, despite the formation of another
by-product resulting from the addition of a chlorine on the Michael
2
Bouffard, Elise
