ARTICLES
Targeting Bcr–Abl by combining allosteric
with ATP-binding-site inhibitors
Jianming Zhang
1
*, Francisco J. Adria´n
2
*, Wolfgang Jahnke
3
, Sandra W. Cowan-Jacob
3
, Allen G. Li
2
,
Roxana E. Iacob
4
, Taebo Sim
1,5
, John Powers
6
, Christine Dierks
2
, Fangxian Sun
2
, Gui-Rong Guo
2
, Qiang Ding
2
,
Barun Okram
7
, Yongmun Choi
1
, Amy Wojciechowski
1
, Xianming Deng
1
, Guoxun Liu
2
, Gabriele Fendrich
3
,
Andre´ Strauss
3
, Navratna Vajpai
8
, Stephan Grzesiek
8
, Tove Tuntland
2
, Yi Liu
2
, Badry Bursulaya
2
,
Mohammad Azam
6
, Paul W. Manley
3
, John R. Engen
4
, George Q. Daley
6
, Markus Warmuth
9
& Nathanael S. Gray
1
In an effort to find new pharmacological modalities to overcome resistance to ATP-binding-site inhibitors of Bcr–Abl, we
recently reported the discovery of GNF-2, a selective allosteric Bcr–Abl inhibitor. Here, using solution NMR, X-ray
crystallography, mutagenesis and hydrogen exchange mass spectrometry, we show that GNF-2 binds to the
myristate-binding site of Abl, leading to changes in the structural dynamics of the ATP-binding site. GNF-5, an analogue of
GNF-2 with improved pharmacokinetic properties, when used in combination with the ATP-competitive inhibitors imatinib
or nilotinib, suppressed the emergence of resistance mutations in vitro, displayed additive inhibitory activity in biochemical
and cellular assays against T315I mutant human Bcr–Abl and displayed in vivo efficacy against this recalcitrant mutant in a
murine bone-marrow transplantation model. These results show that therapeutically relevant inhibition of Bcr–Abl activity
can be achieved with inhibitors that bind to the myristate-binding site and that combining allosteric and ATP-competitive
inhibitors can overcome resistance to either agent alone.
Chronic myelogenous leukaemia (CML) is a haematological malig-
nancy caused by a chromosomal rearrangement that generates a
fusion protein, Bcr–Abl, with deregulated tyrosine kinase activity.
Although clinical remission is usually achieved in early-stage disease
with the drug imatinib, which targets the ATP-binding site,
advanced-stage patients may relapse as a result of the emergence of
clones expressing inhibitor-resistant forms of Bcr–Abl. Two recently
approved drugs, nilotinib (AMN107)
1
and dasatinib (BMS-
354825)
2,3
, address most of the imatinib resistance mutations except
the ‘gatekeeper’ T315I mutation, which is situated in the middle of
the ATP-binding cleft
4–6
.
GNF-2 is a highly selective non-ATP competitive inhibitor of onco-
genic Bcr–Abl activity (half-maximal inhibitory concentration (IC
50
)
0.14mM)
7
. Using NMR spectroscopy and X-ray crystallography, we
identify the myristoyl pocket located near the carboxy terminus of the
Abl kinase domain as the precise binding site of GNF-2 to Bcr–Abl. By
selecting for Bcr–Abl alleles resistant to GNF-2 in vitro, we identify
residues both within and outside of the myristate cleft that are
required for drug efficacy. Simultaneous binding to Bcr–Abl of a
myristoyl mimic and an ATP-competitive inhibitor decreases the
appearance of resistance-conferring mutations and results in the
inhibition of both wild-type and T315I Bcr–Abl kinase activity and
cell growth. Hydrogen-exchange mass spectrometry demonstrates
that binding of GNF-5 to the myristate pocket results in alterations
to the conformational dynamics of the ATP-binding site and provides
a possible mechanism for allosteric communication between these
sites.
GNF-2 binds to the C-terminal myristate pocket of Abl
GNF-2 had previously been suggested to bind in the myristate-binding
pocket of Abl, on the basis of the observation that engineered mutations
located at the entrance (A337N) and rear (A344L) of the myristoyl cleft
conferred resistance to GNF-2 but not to imatinib
7
. To establish the
GNF-2-binding site by an independent biophysical method, we used
solution NMR
8,9
on the Abl/imatinib/GNF-2 complex
10,11
,todemon-
strate that GNF-2 induces chemical shift changes that cluster around the
myristate-binding pocket (Fig. 1a). No significant perturbations in
chemical shift were observed for the ATP pocket, indicating that
GNF-2 does not interfere with imatinib for binding at the ATP-binding
site. Myristic acid was found to induce qualitatively the same pattern of
perturbations in chemical shift (Fig. 1b), providing additional evidence
that GNF-2 and myristate share the same binding site.
Crystal structure of Abl/imatinib/GNF-2 complex. The binding of
GNF-2 to the myristoyl pocket of Abl was further confirmed by X-ray
crystallography. The structure of the Abl/imatinib/GNF-2 complex
was obtained by soaking crystals of Abl/imatinib/myristate, obtained
as described in ref. 10, in an excess of GNF-2. As judged by the shape
of the electron density (Supplementary Fig. 1), GNF-2 replaces the
myristoylated peptide in the crystals. There are two molecules in the
asymmetric unit, and the myristate-binding site is fully occupied by
*These authors contributed equally to this work.
1
Dana-Farber Cancer Institute, Harvard Medical School, Department of Cancer Biology and Department of Biological Chemistry and Molecular Pharmacology, 250 Longwood Avenue,
Seeley G. Mudd Building 628, Boston, Massachusetts 02115, USA.
2
Genomics Institute of the Novartis Research Foundation, Department of Chemistry, 10675 John Jay Hopkins Drive,
San Diego, California 92121, USA.
3
Novartis Institutes for Biomedical Research, CH-4056 Basel, Switzerland.
4
The Barnett Institute of Chemical & Biological Analysis and Department
of Chemistry & Chemical Biology, Northeastern University, Boston, Massachusetts 02115, USA.
5
Life Sciences Research Division, Korea Institute of Science and Technology 39-1,
Hawolgok-dong, Seongbuk-gu, Seoul, 136-791, Korea.
6
Division of Pediatric Hematology/Oncology, Children’s Hospital and Dana-Farber Cancer Institute; Division of Hematology,
Brigham and Women’s Hospital; Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School; Howard Hughes Medical Institute; Boston,
Massachusetts 02115, USA.
7
Department of Chemistry and the Skaggs Institute for Chemical Biology, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla,
California 92037, USA.
8
Biozentrum, University of Basel, Klingelbergstrasse 70, CH-4056 Basel, Switzerland.
9
Novartis Institutes for BioMedical Research, Inc., 250 Massachusetts
Avenue, Cambridge, Massachusetts 02139, USA.
Vol 463 | 28 January 2010 | doi:10.1038/nature08675
501
Macmillan Publishers Limited. All rights reserved©2010

GNF-2 in one and partly in the other. GNF-2 binds in an extended
conformation in the myristate pocket with the CF
3
group buried at the
same depth as the final two carbons of the myristate ligand (Fig. 2).
There is a favourable, but probably weak, polar interaction between
one fluorine atom and the main chain of L340 (similar to that
observed between nilotinib and D381 of Abl)
12
, and there are water-
mediated hydrogen bonds, but no direct hydrogen bonds with the
protein. As expected, most of the interactions between GNF-2 and
the protein are hydrophobic. As discussed below, mutation of three
residues near the mouth of the myristate-binding site (C464Y, P465S
and V506L) is found to cause resistance to the binding of GNF-2,
presumably for steric reasons. The overall structure of the Abl kinase
domain complexed to GNF-2 is similar to that of the myristate com-
plex (Fig. 2b), with some small differences that probably result from
crystal contacts, but no changes in the ATP-binding site.
GNF-2 analogue structure–activity relations
After the identification of GNF-2 as a lead compound, a systematic
evaluation of the structural features necessary to impart function as a
cellular Bcr–Abl inhibitor were investigated through the synthesis of
more than 200 analogues and rationalized in the context of binding
to the myristate-binding site (X.D. and N.S.G., unpublished observa-
tions, and Supplementary Fig. 2a). The structure–activity relations
are fully consistent with the conformation of GNF-2 observed in the
crystal structure: the trifluoromethoxy group of GNF-2 can only be
accommodated at the para position; the aniline NH is required
because of the formation of a water-mediated hydrogen bond to
the backbone carbonyls of A433 and E462; water-mediated hydrogen
bonds between the carboxamide of GNF-2 and Abl confer enhanced
inhibitory activity; and the extended compound conformation is
required to fit the cylindrical binding cavity.
Drug combinations reduce emergence of resistant mutants
We sought to investigate the frequency with which Bcr–Abl-dependent
Ba/F3 cells would become resistant to combinations of GNF-2 and
imatinib in comparison with each compound alone. The number of
resistant clones that emerged as a result of continuous exposure to 1mM
imatinib was decreased by at least 90% when cells were treated for up to
21 days with 1 mM imatinib combined with 5 or 10mM GNF-2 (Fig. 3a).
These results demonstrate that combinations of GNF-2 and imatinib
can cooperate to suppress the emergence of resistance mutations.
Identification of the Bcr–Abl mutants resistant to GNF-2. To dis-
cover the full complement of Bcr–Abl mutants that induce resistance
to GNF-2 we performed two types of selection. In the first, Bcr–Abl-
transformed Ba/F3 cells were cultured in the presence of increasing
concentrations of GNF-2 to allow cells to evolve drug resistance as
described previously for the ATP-competitive inhibitor PD166326
(ref. 13). In the second approach, Bcr–Abl was randomly mutated in
Escherichia coli and the mutant clones were expressed in Ba/F3 cells,
which were then grown in the presence of inhibitor
14
. These screens
resulted in the identification of a total of 306 mutants, 163 (12 sites)
from the first and 143 (22 sites) from the second (Supplementary Fig. 3).
More than 80% of the resistant colonies contained Bcr–Abl mutations
clustered in the myristate-binding pocket or the SH2 and SH3 domains
(Supplementary Fig. 3). This is in contrast to ATP-competitive inhibi-
tors such as imatinib, PD166321 and AP23464 (refs 13–15), for which
most resistance mutations cluster adjacent to the kinase catalytic site.
To validate the functional relevance of these mutations, we engineered
individual mutant Bcr–Abl-transformed Ba/F3 cells for nine of the most
frequently isolated GNF-2-resistant mutations and for the T315I ‘gate-
keeper’ mutation. These selected mutations located in the SH3 domain
(P112S), the SH3–SH2 domain linker (Y128D), the SH2 domain
(Y139C), the SH2–kinase-domain linker (S229P), the ATP-binding site
(T315I) and adjacent to the myristate-binding site (C464Y, P465S,
F497L, E505K and Y506L) were introduced individually into Bcr–Abl
by site-directed mutagenesis. Of these, only the T315I substitution has
previously been reported to confer resistance to imatinib
16,17
. GNF-2 had
an IC
50
against all ten mutants that was elevated 5–50-fold relative to that
against wild-type Bcr–Abl-transformed Ba/F3 cells (Fig. 3b). The three
most frequently recovered mutations (60% of the total) were located in
close proximity to the myristate-binding site (C464Y, P465S and E505K)
and were shown to confer complete resistance to GNF-2 up to a con-
centration of 10 mM (Fig. 3c).
To examine how the mutations affected the ability of GNF-2
to inhibit Bcr–Abl-mediated signalling, we examined Bcr–Abl
autophosphorylation and phosphorylation of a downstream substrate,
STAT5, after treatment with inhibitor (Supplementary Fig. 4a). At a
concentration of 10mM, GNF-2 could inhibit the phosphorylation of
Bcr–Abl and STAT5 in all mutants except the three myristate-site
mutations (E505K, P465S and C646Y) and the ‘gatekeeper’ T315I
mutation.
Mutations in the myristate pocket interfere with GNF-2 binding.
We tested the ability of mutant Bcr–Abl proteins, obtained from
crude cell lysates, to bind to a GNF-2 affinity resin (Supplementary
Fig. 4b, c)
7
. These experiments revealed that only the mutations
located in the myristate-binding site (C464Y, P465S and E505K)
ab
122.0
125.0
124.0
123.0
122.0
125.0
124.0
123.0
8.0 7.0 p.p.m. 7.0 p.p.m.8.0
Figure 1 | NMR spectroscopy provides evidence for GNF-2 binding to the
C-terminal myristate pocket of Abl. a, Top: a heteronuclear single-quantum
coherence spectrum of the Abl/imatinib complex with (red) and without
(black) GNF-2, showing chemical-shift changes induced by ligand binding
(the y-axis scale is in p.p.m.). Bottom: mapping of the chemical-shift changes
to the structure of the Abl/imatinib complex (PDB accession 1OPK; ref. 10)
identifies the myristate pocket as the GNF-2-binding site. The size of the
spheres is proportional to the magnitude of the chemical shift changes. b,As
in a except that myristic acid (blue in top panel) was used instead of GNF-2.
a E462
L510
P465
L429
L340
L341
HOH
HOH
HOH

E

I

I

H

F
I–I loop
Y435
C464
E505
V468
V506
F497
A337
A344
A413
I432
F493
I502
ba
Figure 2 | Crystal structure of GNF-2 bound to the Abl myristoyl pocket.
a, Abl kinase is indicated in green (helices are indicated by transparent
cylinders), with the bent part of the I-helix in yellow, GNF-2 resistance
mutations in pink, and GNF-2 carbons in cyan. Hydrogen bonding and
other polar interactions are indicated by dotted red lines. b, Superposition of
the Abl/imatinib/myristate (white), Abl/imatinib/GNF-2 (green and yellow)
and Abl-imatinib (red) structures. GNF-2 is coloured in cyan, and myristic
acid in magenta.
ARTICLES NATURE | Vol 463 | 28 January 2010
502
Macmillan Publishers Limited. All rights reserved©2010