Discovery of N-((3R,4R)-4-Fluoro-1-(6-((3-methoxy-1-methyl-1H-pyrazol-4-yl)amino)-9-methyl-9H-purin-2-yl)pyrrolidine-3-yl)acrylamide (PF-06747775) through Structure-Based Drug Design: A High Affinity Irreversible Inhibitor Targeting Oncogenic EGFR Mutants

Article abstract of DOI:10.1021/acs.jmedchem.6b01894

Mutant epidermal growth factor receptor (EGFR) is a major driver of non-small-cell lung cancer (NSCLC). Marketed first generation inhibitors, such as erlotinib, effect a transient beneficial response in EGFR mutant NSCLC patients before resistance mechani

Full text of DOI:10.1021/acs.jmedchem.6b01894

Subscriber access provided by University of Newcastle, Australia
Article
Discovery of N-((3R,4R)-4-fluoro-1-(6-((3-methoxy-1-methyl-1H-pyrazol-4-
yl)amino)-9-methyl-9H-purin-2-yl)pyrrolidine-3-yl)acrylamide (PF-06747775)
Through Structure-Based Drug Design; A High Affinity Irreversible Inhibitor
Targeting Oncogenic EGFR Mutants With Selectivity Over Wild-Type EGFR.
Simon Planken, Douglas Carl Behenna, Sajiv K Nair, Theodore Otto Johnson, Asako Nagata,
Chau Almaden, Simon Bailey, T. Eric Ballard, Louise Bernier, Hengmiao Cheng, Sujin Cho-
Schultz, Deepak Dalvie, Judith G. Deal, Dac M Dinh, Martin P. Edwards, Rose Ann Ferre, Ketan
S Gajiwala, Michelle D. Hemkens, Robert S Kania, John C. Kath, Jean Matthews, Brion W.
Murray, Sherry Niessen, Suvi T M Orr, Mason Pairish, Neal W Sach, Hong Shen, Manli Shi, James
Solowiej, Khanh Tran, Elaine Tseng, Paolo Vicini, Yuli Wang, Scott L. Weinrich, Ru Zhou, Michael
Zientek, Longqing Liu, Yiqin Luo, Shuibo Xin, Chengyi Zhang, and Jennifer Anne Lafontaine
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.6b01894 • Publication Date (Web): 13 Mar 2017
Downloaded from http://pubs.acs.org on March 13, 2017
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Medicinal Chemistry is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Tseng, Elaine; pfizer PGRD, PDM
Vicini, Paolo; pfizer PGRD, Medicinal Sciences
Wang, Yuli; Pfizer Global Research & Development, Oncology research unit
Weinrich, Scott; Pfizer Global Research & Development, Oncology Research
Unit
Zhou, Ru; pfizer PGRD, Medicinal Sciences
Zientek, Michael; Takeda Pharmaceuticals U.S.A., Inc.,
Liu, Longqing; WuXi AppTec
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Luo, Yiqin; WuXi AppTec
Xin, Shuibo; WuXi App Tec Co., Shanghai, China,
Zhang, Chengyi; WuXi AppTec
Lafontaine, Jennifer; Pfizer Global Research and Development, Medicinal
Chemistry
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Discovery of N-((3R,4R)-4-fluoro-1-(6-((3-methoxy-1-methyl-1H-pyrazol-4-yl)amino)-9-
methyl-9H-purin-2-yl)pyrrolidine-3-yl)acrylamide (PF-06747775) Through Structure-
Based Drug Design; A High Affinity Irreversible Inhibitor Targeting Oncogenic EGFR
Mutants With Selectivity Over Wild-Type EGFR.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
†
†
†
†
†
Simon Planken* , Douglas C. Behenna* , Sajiv K. Nair , Theodore O. Johnson , Asako Nagata ,
‡
†
§
†
†
Chau Almaden , Simon Bailey , T. Eric Ballard , Louise Bernier , Hengmiao Cheng , Sujin Cho-
†
§
†
‡
†
†
Schultz , Deepak Dalvie , Judith G. Deal , Dac M. Dinh , Martin P. Edwards , Rose Ann Ferre ,
†
¥
†
†
†
Ketan S. Gajiwala , Michelle Hemkens , Robert S. Kania , John C. Kath , Jean Matthews , Brion
‡
†
†
†
†
†
W. Murray , Sherry Niessen , Suvi T. M. Orr , Mason Pairish , Neal W. Sach , Hong Shen ,
‡
‡
†
§
§
‡
Manli Shi , James Solowiej , Khanh Tran , Elaine Tseng , Paolo Vicini , Yuli Wang , Scott L.
‡
†
§
Ɣ
Ɣ
Ɣ
Weinrich , Ru Zhou , Michael Zientek , Longqing Liu , Yiqin Luo , Shuibo Xin , Chengyi
Ɣ
†
Zhang and Jennifer Lafontaine .
Departments of Worldwide Medicinal Chemistry, §Pharmacokinetics, Dynamics and Metabolism, ‡Oncology
†
¥
Research Unit, Drug Safety, La Jolla Laboratories, Pfizer Worldwide Research and Development, 10770 Science
Ɣ
Center Drive, San Diego, CA, 92121, United States. Wuxi AppTec, 288 Fute Zhong Road, Waigaoqiao Free Trade
Zone, Shanghai 200131, China.
Abstract: Mutant epidermal growth factor receptor (EGFR) is a major driver of non-small cell
lung cancer (NSCLC). Marketed first generation inhibitors, such as erlotinib, effect a transient
beneficial response in EGFR-mutant NSCLC patients before resistance mechanisms render these
inhibitors ineffective. Secondary oncogenic EGFR mutations account for approximately 50% of
relapses, the most common being the gatekeeper T790M substitution that renders existing
therapies ineffective. The discovery of PF-06459988 (1), an irreversible pyrrolopyrimidine
1
inhibitor of EGFR T790M mutants was recently disclosed . Herein, we describe our continued
1
ACS Paragon Plus Environment

Page 3 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
efforts to achieve potency across EGFR oncogenic mutations and improved kinome selectivity,
resulting in the discovery of clinical candidate PF-06747775 (21), which provides potent EGFR
activity against the four common mutants (exon 19 deletion (Del), L858R, and double mutants
T790M/L858R and T790M/Del), selectivity over wild-type EGFR, and desirable ADME
properties. Compound 21 is currently being evaluated in Phase-I clinical trials of mutant EGFR
driven NSCLC.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Introduction:
Activating mutations in the Epidermal Growth Factor Receptor (EGFR) are one of the major
2
drivers of non-small cell lung cancer (NSCLC) . While reversible first generation EGFR
inhibitors are initially an effective treatment, it is well documented that a drug resistant
gatekeeper T790M mutation accounts for ~50% of disease progression in patients treated with
2a, 3
first generation agents . Irreversible second generation EGFR inhibitors, which covalently
modify C797 in the ATP binding site of EGFR, are more potent against the T790M
4
mutants . These agents, however, are dose limited by their exceptional potency against wild-
5
type (WT) EGFR, which drives mechanism-based toxicity . Efforts to overcome this limitation
led to the emergence of irreversible third generation EGFR inhibitors that provide effective
potency against both drug resistant T790M mutants and the original activating mutants while
6
sparing activity against WT EGFR .
The recently disclosed third generation EGFR inhibitor PF-06459988 (compound 1, Figure 1)
was discovered by de novo structure-based drug design with a strategic focus on improving
1
inhibitor potency by increasing reversible binding affinity . Implementation of this strategy
enabled the intrinsic reactivity of the inhibitor’s warhead functionality to be minimized in
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
parallel as a means to reduce the probability of any safety risks associated with an irreversible
7
inhibitor mode of action . Compound 1 provided low nM biochemical and cellular potency
against both EGFR T790M resistant mutants (T790M/L858R and T790M/Del), excellent
selectivity for the drug resistant mutants over WT EGFR and in vitro and in vivo ADME
characteristics well suited for in vivo efficacy studies.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
HN
Cl
N
HN
N
O
O
N
N N
O
1
EGFR
EGFR
IC50 13 nM
L858R_T790M
IC50 7 nM
Del_T790M
EGFR L858R IC50 21 nM
EGFR Del IC50 140 nM
EGFR WT IC50 5110 nM
Figure 1: 1 a previously reported T790M mutant inhibitor that has exceptional selectivity over WT EGFR. The 1.9
Å co-crystal structure (PDB ID 5HG7) with EGFR T790M/L858R protein shows the ligand covalently modifying
Cys797 of the EGFR protein as designed. The core also forms the classic donor-acceptor-donor motif to the kinase
hinge region. Also highlighted is the gatekeeper residue, Met790. IC data is from cell-based assays.
50
Furthermore, by virtue of its weakly electrophilic acrylamide warhead, the broad proteome
reactivity of 1 was significantly reduced relative to previous covalent EGFR inhibitors
8
possessing more reactive warhead groups . While 1 possessed characteristics desirable in a
clinical candidate, its potency against the two original activating mutants of EGFR (L858R and
the exon19_deletion (Del)) was mixed. Recognizing the physicochemical property limitations of
the pyrrolopyrimidine series in improving Del potency without compromising the excellent
3
ACS Paragon Plus Environment

Page 5 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
attributes of 1 (substitution of the halogen with an aromatic ring led to improved Del but with an
increase in clearance due to high logD), an alternative chemical series was pursued. Herein, we
describe our efforts applying the principles of affinity-driven potency and low intrinsic
electrophile reactivity to i) increase Del potency while retaining potency against the T790M
mutants and selectivity over the WT EGFR); ii) improve broad protein kinase selectivity across
the kinome and iii) maintain excellent ADME properties suitable for once daily oral dosing in
patients.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Results and Discussion:
To identify suitable chemical leads, a directed screening approach was applied. The lead
generation screening strategy was to screen known reversible kinase inhibitors from the Pfizer
compound collections, identify hits that were selective for T790M over the WT EGFR, and then
append a warhead onto selected scaffolds (as informed by computational modeling using a
6
d
T790M EGFR co-crystal structure ). From screening a diverse kinase targeted library set of
approximately 20,000 compounds, a purine containing lead was identified, compound 2 (Figure
9
2
) . 2 exhibited EGFR T790M/L858R double mutant potency with an IC of 78 nM and WT
5
0
selectivity of 76 fold in a biochemical kinase assay. However despite acceptable passive
-6
permeability (1.9x10 cm/s), when tested in a cellular context (NSCLC derived H1975 line
expressing T790M/L858R double mutant (DM)), 2 was inactive even at 10 µM, illustrating the
1
opportunity for covalent inhibitor modality. As in our previous work , increasing potency by
est
optimizing non-covalent binding affinity, using an estimated Ki assay (K ) to assess progress,
i
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
and incorporating weakly reactive aliphatic warheads (as measured using a glutathione assay)
1
0
were the focus of the optimization program .
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
8
N 9
N
7
N
HN
6
N NH
2
OH
2
(racemic)
T790M_L858R
WT/DM Ratio: 76x
T790M H1975 IC50 : > 10mM
N
IC50 78 nM
N
H
pEGFR
Figure 2. Modeled docking of screen hit 2 into the EGFR T790M/L858R protein (PDB ID 5HG7). Note core hinge
interactions including polarized C(8) C-H bond. The close proximity of the C(2) vector to target Cys797 indicated
the best vector on which to append an electrophile.
To determine how to best incorporate a suitable electrophile, binding of 2 was modeled using
T790M/L858R EGFR crystal structure. The predicted binding mode (Figure 2) involved a
tridentate hydrogen bonding interaction to the kinase hinge region, consisting of two formal
hydrogen bonds (with the core N(7) acting as an acceptor and the aniline NH as a donor), and a
polarized C‒H bond at the C(8)-position acting as a second hydrogen bond donor (interacting
with the carbonyl of hinge residue GLN791). The N(9) isopropyl group was in close proximity to
the gatekeeper methionine residue which was predicted to be driving selectivity over WT EGFR,
and the C(2) valinol group appeared to be in closest proximity to the target Cys797 residue
5
ACS Paragon Plus Environment

Page 7 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
offering a vector for warhead incorporation. SBDD suggested a pyrrolidine linker was suitable to
place an acrylamide warhead in close proximity to the target cysteine. Key examples are shown
in Table 1.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
N
N
N
R1
N
R2
d
H1975
A549
Rel
est
Compound
ID
K
i
k
obs/I
-1 -1
WT/
DM
R1
R2
IC50
IC50
LogD
GSH
a
a
(nM)
(M s )
c
c
(nM)
(nM)
T
1/2
H
H
H
H
N
NH
N
3
953
3
Ent-3
4
32
29
280
19
387
13
> 35
15
2.5
ND
> 15
> 22
b
N
N
(1 µM)
O
N
NH
N
2
58
164
> 10,000
1.8
N
N
b
(5 µM)
O
N
N
N
N
1
1,886
NH
9
(0.1 µM)
277
2.2
N
N
O
b
N
5
3
NH
Ent-4
161
ND
ND
652
-
2.0
2.3
ND
N
N
b
(5 µM)
O
H
N
4452
5
36
52
13
> 18
NH
b
(1 µM)
N
N
O
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
H
N
N
N
9
7
Ent-5
183
ND
ND
805
-
2.7
2.2
ND
NH
b
(
5 µM)
N
N
O
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
H
N
H
2041
H
6
85
49
16
> 15
N
b
(
1 µM)
N
N
O
est
Table 1: Summary Early Purine Examples. a) Ki and k /I derived as previously published; b) Inhibitor
obs
concentration at which the k /I is calculated c) H1975 and A549 IC represent activities against T790M/L858R
obs
50
and WT EGFR respectively; d) Rel GSH T1/2 represents the intrinsic warhead reactivity expressed as the fold-
change in half-life of parent compound in a glutathione stability assay relative to that of canertinib (t1/2 ca: 21 min).
As shown in Table 1, the pyrrolidine-based linkers contained within 3 and 4 showed promising
est
early data with Ki values approaching our 10 nM goal. Preference for the 3-acrylamido
pyrrolidine warhead was observed, based both on higher affinity (as determined by extrapolated
est
K ) and a significant increase in the rate of covalent modification (measured k /I) observed
i
obs
with 4 compared to 3. An enantio-preference was observed for both pyrrolidine based linkers, as
illustrated when comparing enantiomers of compound 3 and 4. These observations were in
agreement with modeling predictions that placed the exocyclic (S)-acrylamide in closer
proximity to the target Cys797 (see Figure 3).
7
ACS Paragon Plus Environment

Page 9 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3. Modeled docking of simplified compounds containing pyrrolidine linked warheads into the EGFR
T790M/L858R protein (PDB ID 5HG7). Distance of highlighted CYS797 to the acrylamide is shown. A and B)
linker from 4 showing “pseudo-axial” and “pseudo-equatorial” warhead conformations respectively; C) isomeric
pyrroldine linker from 3.
Together with its very low intrinsic reactivity (relative GSH half-life of 22-fold), high potency in
both biochemical and H1975 cell assays (9 nM and 19 nM, respectively), a high k /I value
obs
-1 -1
approaching 12,000 M s and 15 fold selectivity over WT EGFR, the 3-amino pyrrolidine
derivative contained in 4 represented the best linker motif, and formed the basis for further
design optimization.
1
1
Anilines are known potential toxicophores and the high molecular weight, high logD and basic
nature of the bis-aniline fragment contained in 4 (ACD calc pKa 7.9) constrained the molecule’s
physiochemical profile and likely contributed to observed liabilities including ion-channel
1
2
activity, poor permeability and poor LipE . Thus, an early optimization goal was to replace the
aniline fragment with a pyrazole moiety (as shown in Table 1, compounds 5, 6), leveraging
1
knowledge derived from program to discover 1 . Compound 5 with a 3-amino pyrrolidine linker
est
afforded activity with a Ki of 36 nM and an IC of 52 nM in the H1975 cell assay, while
50
8
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
-1 -1
retaining WT selectivity and low reactivity. However, with a relatively low k /I of 4452 M s ,
obs
1
3
more potency could be obtained by improving alignment of the warhead to the cysteine . The
bicyclic pyrrolidine linker of 6 also afforded activity but offered no benefit over the simpler
linker of compound 5.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Modeling of lead compound 5 suggested that the acrylamide would prefer to adopt a less
favored, “pseudo-axial” conformation in order to place the warhead proximal to Cys797 for
facile covalent modification (Figure 3a). As a result of the incurred strain energy penalty
(
calculated at 1.3 kcal/mol in solution phase), it was hypothesized that a potency gain would be
realized if the molecule could be modified to favor this conformation. Replacement of a
pyrrolidine hydrogen with a fluorine is known to impart a gauche effect, whereby an adjacent
1
4
electronegative substituent prefers to be in a gauche orientation . Addition of fluorine onto the
-position of the pyrrolidine ring of compound 5 was calculated to result in a slight preference
by 0.1 kcal/mol) for the 3,4-anti conformation. In fact, fluorine introduction to give 7 resulted in
4
(
a significant potency gain from both an affinity and inactivation rate perspective (Table 2), with
-1 -1
the k /I increased to 62,982 M s . This reflected the less costly accessibility of the anti
obs
conformation and the improved orientation and proximity of the warhead relative to the cysteine
for 7 compared to 5.
N
N
N
HN
N
R
N
N
est
d
e
Compound
R
K
i
k
obs/I
H1975
A549
WT/DM LogD
LipE
Rel
9
ACS Paragon Plus Environment

Page 11 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
a
-1 -1
a
ID
(nM)
(M s )
IC50
IC50
GSH
c
c
(
nM)
(nM)
T
1/2
N
4
,452
NH
5
36
52
652
13
26
-
2.3
2.6
2.5
5.1
6.1
4.3
> 18
> 18
> 18
b
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(1 µM)
O
N
62,982
(0.05
NH
7
2
7
178
F
O
b
µM)
N
2
44
>10
uM
NH
8
152
713
F
b
(
5 µM)
O
est
Table 2: Effect of Fluorine Addition to Linker; a) Ki and k /I derived as previously published. b) Inhibitor
obs
concentration at which the k /I is calculated. c) H1975 and A549 IC represent activities against T790M/L858R
obs
50
est
and WT EGFR respectively; d) LipE calculated from Ki and experimental logD; e) Rel GSH T1/2 represents the
intrinsic warhead reactivity expressed as the fold-change in half-life of parent compound in a glutathione stability
assay relative to that of canertinib (t1/2 ca: 21 mins).
A 1.5 Å cocrystal of compound 7 with EGFR L858R/T790M (with additional surface mutation
1
5
V948R, distal from the ligand binding site, Figure 4 ) was obtained which confirmed the
predicted anti conformation of the fluorine and the acrylamide. A small molecule X-ray structure
also later confirmed the low energy state (see supporting information for small molecule crystal
structure of 21, Figures SF1-SF3). In addition, as predicted from docking experiments, the fluoro
substituent of 7 is within 3.6 Å of the G-loop Phe723 and able to interact via a dipole-dipole
1
6
17
interaction ; the cumulative result is a profound 10-fold increase in potency, and a lipE
increase of 0.8 relative to the des-fluoro analog. A high enantiomeric preference was observed as
1
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
the distomer (compound 8) was more than 60 fold less potent against the T790M/L858R double
mutant.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 4. 1.5 Å Cocrystal Structure (PDB ID 5UG8) of compound 7 with EGFR_L858R/T790M Mutant (only
the ligand binding site is shown). Note anti conformation of fluorine and warhead linker and apparent fluorine
interaction with Phe723.
1
In the previously disclosed pyrrolopyrimidine work , addition of fluorines into warhead linkers
was accompanied by an increase in intrinsic reactivity via lowering of the acrylamide LUMO
through the inductive effect of the fluorine, as illustrated in Table 3 when comparing the
compound pairs 9 - 10, and 11 - 12.
HN
HN
N
N
Cl
NH
N
N
N
HN
N
O
HN
N
R
HN
N
NH
R
N
N
R
N
N
N
NH
N
N
N
O
O
O
N
1
1
ACS Paragon Plus Environment

Page 13 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
R=H
R=F
(10)
R=H
(11)
R=F
(12)
R=H
(13)
R=F
(14)
(
9)
T790M/L858R
3
2
1
4
3
5
105
14
56
9
est
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Ki (nM)
Rel GSH T1/2
4
22
(fold change)
Table 3: Fluorine Introduction into the Warhead Linker. Fluorine introduction in pyrrolopyrimidine and early
purine examples showed an increase in the relative intrinsic reactivity of the warhead. Rel GSH T1/2 represents the
intrinsic warhead reactivity expressed as the fold-change in half-life of parent compound in a glutathione stability
assay relative to that of canertinib (t1/2 ca: 21 mins).
Because a minimally reactive warhead was sought, this observed increase in intrinsic reactivity
precluded advancement of these compounds. Similar results were seen in the purine series with
the same fluorinated pyrrolidine linkers 13 and 14 (Table 3) where fluorine introduction led to an
increase in reactivity, with the GSH relative half-life dropping from 14 to 9 fold. However,
unlike these examples, it was predicted that the anti conformation of the 3-acrylamido-4-
fluoropyrrolidine of 7 would not significantly increase the reactivity of the warhead (Table 2)
when the potential for orbital overlap between the C-F bond and the nitrogen of the acrylamide
was considered. In examples where an increase in reactivity was observed, it was possible to
align the σ* of the C-F bond with the nitrogen lone pair of the acrylamide, which would lower
the LUMO energy and increase the reactivity of the warhead (Figure 5a).
1
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Figure 5: a) Activating Fluorine: Overlap between warhead nitrogen NB-orbital and σ* orbital of C-F bond of 14
increases the reactivity of the Michael acceptor. b) Neutral Fluorine: Orthogonal orbitals have minimal interaction
which resulted in little effect on intrinsic reactivity of 7.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
In the case of compound 7, the orthogonal orientation between the warhead nitrogen non-
bonding (NB) orbital and the C-F σ* orbital reduces interaction with the electrons of the LUMO;
as a result the fluorine would not be predicted to adversely impact the intrinsic reactivity (Figure
5
b). Upon testing in the glutathione based reactivity assay, fluorine introduction was found not to
exhibit a measurable difference in intrinsic reactivity when comparing the relative GSH
reactivity half-lives between 5 and 7 (Table 2). Thus, fluorine addition was found to have a
profound benefit by i) anti conformational preference, ii) a dramatic increase in potency, iii)
increased interaction with the protein, and iv) no negative impact on the intrinsic reactivity.
Compound 7 had moderate clearance with an exceptional EGFR mutant profile, possessing
single digit nanomolar potency against all 4 mutants and 20 fold selectivity over WT EGFR.
Thus, having optimized the warhead linker group, the series was evaluated for selectivity against
the kinome.
Kinase Selectivity: The kinome selectivity for the simple pyrrolidine-linker compound 5 was
moderate, with a hit-rate of 25% when assessed against a diverse 38 membered kinome panel
(where a hit is defined as > 50% inhibition in biochemical screening assay at 1 µM at Km
concentrations of ATP, see supporting information Table ST2). This is approximately the same
for 1. While addition of fluorine was beneficial from a potency and efficiency perspective, this
change had a deleterious effect on the kinome selectivity with the hit-rate increasing to 43%. One
possible explanation for the increase in off-target kinase activity is the observed interaction
between the fluorine and Phe723. This Phe is well conserved throughout the kinome (67% of
1
3
ACS Paragon Plus Environment

Page 15 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
kinases possess a Phe at this location) and specific interaction with it could result in an increase
in kinase promiscuity. To improve upon this profile, a selectivity pocket afforded by hinge
residue Leu792 (Figure 6) was targeted. In many kinases (59% of the kinome), this hinge residue
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(
H5) is either a larger Phe or Tyr, offering a strategy for selectivity by building a steric clash
with those larger residues. Filling this pocket has been shown to achieve kinase selectivity for
1
8
19
other kinase programs such as ALK , MPS-1 and LRRK2 which both possess a leucine
residue at the corresponding hinge position.
Figure 6: Co-Crystal Structure of 16 (1.3 Å, PDB ID 5UG9). The kinase selectivity pocket that is afforded via
hinge residue Leu792 is illustrated. Many kinases (59%) have a large Phe or Tyr residue in this position, precluding
pyrazole substitution. Filling this pocket with a methoxy group afforded an improved kinase off-target profile
without significant detrimental impact on in vitro human microsomal intrinsic clearance.
Placing small groups (methyl, methoxy, ethyl) in this region afforded compounds 15-17 (Table
4
). Addition of methyl (15), methoxy (16) or ethyl (17) all improved kinase selectivity, with
compound 16 providing the best combined potency and lipE, with minimal impact on the in vitro
microsomal intrinsic clearance or potency. Compound 16 showed activity against only two
1
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
kinases outside the Tec family both of which possess a leucine residue at the H5 hinge region,
thus circumventing the selectivity strategy (see supporting information Table ST2).
H1975
A549
IC50
Kinome
est
Compound
ID
Ki
kobs/I
WT/
DM
hLM Clint
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
d
Structure
IC50
LogD LipE
hitrate
a
-1 -1 a
(µl/min/mg) e
(nM)
(M S )
c
c
f
(nM)
(nM)
(%)
N
N
6
2,982
N
HN
N
7
N
N
2
(0.05 µM)
7
178
26
2.6
6.1
25
58
F
N
b
O
HN
N
N
N
HN
5699
N
N
1
5
17
26
1,450
56
2.6
5.2
30
13
F
b
N
(1 µM)
N
O
HN
N
N
N
16
HN
N
28,437
N
O
N
8
5
7
160
649
32
89
2.6
3.5
5.5
4.2
32
81
13
F
b
N
(0.1 µM)
O
HN
N
N
N
6
415
HN
N
1
7
N
19
8
N
F
b
(
1 µM)
N
O
HN
est
Table 4: in vitro Profiles for Pyrazoles Designed to Improve Kinase Selectivity. a) Ki and k /I derived as
obs
previously published. b) Inhibitor concentration at which the k /I is calculated. c) H1975 and A549 IC represent
obs
50
est
activities against T790M/L858R and WT EGFR respectively. d) LipE calculated from Ki and experimental logD.
e) hLM: Human Liver Microsomal intrinsic clearance. f) Kinase hitrate expressed as % kinases that showed >50%
inhibition at 1 µM inhibitor at K levels ATP from a diverse 38 membered panel.
m
1
5
ACS Paragon Plus Environment

Page 17 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Single Mutant Potency: One of the primary objectives for this work was to improve the potency
against the Del mutation. Current clinical T790M inhibitors are least potent against this mutant
form compared to other single (L858R) and double T790M mutants, and this was also seen in
our previous work using the 5-chloro-pyrrolopyrimidine scaffold, where 1 was 11 fold less
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
potent against Del than T790M/L858R (PC9 and H1975 cell lines, respectively) . Plotting double
mutant (T790M/L858R, H1975) vs single mutant (Del, PC9) cell potency highlights the
differentiation between the pyrrolopyrimidine and the purine scaffolds (Figure 7). In general, the
purine series (shown in blue) provided equivalent potency or better for Del mutation, in contrast
to the pyrrolopyrimidine series (shown in orange) where a 10-fold loss was generally observed.
1
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 7: Single Mutant vs. Double Mutant Potency. Plot showing series potencies against the T790M/L858R
double mutant (H1975) and the Del (PC9) single mutant highlighting the superior Del mutant coverage of the purine
series (blue) over the previous work’s pyrrolopyrimidine series (orange).
When assessed against the single mutations (L858R and Del) and double mutations
(T790M/L858R and T790M/Del), the purine series exhibited higher potency against all key
mutant isoforms with more than 15-fold better potency against Del than our previously disclosed
compound 1. Single digit nM potencies were observed across each of the four mutant EGFR
NSCLC cell lines for 16 (Table 5).
1
7
ACS Paragon Plus Environment

Page 19 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
H1975
PC9-DRH
PC9
(Del)
140
H3255
(L858R)
21
A549
(WT)
5110
WT
ratio*
36
Compound
(
T790M/L858R) (T790M/Del)
1
7
13
7
7
2
4
1
3
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
9
16
2
2
9
178
1450
160
20
1
1
1
5
6
7
26
5
56
2
32
7
ND
ND
649
89
Table 5: pEGFR Potency Data (IC , nM). The purine series example 16 displayed a flatter potency profile across
50
the EGFR mutants relative to compound 1 from previous work where the Del mutation was hardest to cover. (*ratio
expressed using lowest EGFR mutant potency. ND not determined)
N(9) Substituent and EGFR Wild-Type Selectivity: With broad kinase selectivity and the
desired EGFR mutant profile in hand, efforts were then focused on investigating the purine N(9)
substituent and its impact on WT EGFR selectivity. The N(9) group was considered a driver for
i
WT selectivity because of its close proximity to the gatekeeper residue. Analysis of N(9) Pr
purine co-crystal structures showed that in order to bind in the active site of WT EGFR, rotation
i
of the isopropyl group was necessary to avoid an electrostatic clash between the lipophilic Pr
group and the polar threonine gatekeeper. This was confirmed with a co-crystal of reversible
acetate analog 18 bound in WT EGFR (Figure 8).
1
8
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
N
N
N
HN
N
NH
N
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
N
N
O
1
8
Compound
Ki* 161 nM
i
Figure 8: N(9) Pr Rotation on Binding to WT EGFR. Co-crystal structures of compound 18 in EGFR
L858R/T790M (left, 1.3 Å resolution, 5UGA) and WT EGFR (right, 2.1 Å resolution, 5UGB) illustrated the
i
i
orientation of the Pr group relative to the gatekeeper residue: note the rotation of the N(9) Pr when bound in the
WT pocket. *Measured K in T790M/L858R double mutant.
i
Whereas in the L858R/T790M structure, the two isopropyl methyl groups of the ligand orient
themselves toward the gatekeeper residue with productive VDW interactions, Thr790 in the WT
t
structure precludes this both sterically and electrostatically. The tert-butyl ( Bu) substituted
analog was predicted to have increased WT selectivity, as modeling predicted that the larger
substitution removes the possibility of the N(9) group to offer a hydrogen toward Thr790. Due to
the critical nature of this N(9)-gatekeeper interaction, ethyl and methyl analogs were also
synthesized for comparison (see PDB ID 5UGC). Data for key examples are shown in Table 6,
together with selected ADME properties.
H1975
A549
IC50
RRCK
est
Compound
ID
K
i
WT/
DM
LipE
LipE
(WT)
HLM Clint
hHep
(µl/min/MM)
-6
Structure
IC50
LogD
(10
cm/sec)
(nM)
(H1975)
(µl/min/mg)
(nM)
(nM)
1
9
ACS Paragon Plus Environment

Page 21 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
N
N
N
1
6
HN
N
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
N
8
5
160
32
2.6
5.7
4.2
18
31
23
N
O
F
N
O
HN
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
N
N
N
N
N
HN
N
N
1
9
O
18
17
673
40
3.3
4.5
2.9
16
65
ND
F
N
N
O
HN
N
N
HN
N
2
0
N
N
O
F
6
8
83
11
2.3
5.8
4.9
12
9
5.6
N
O
HN
N
N
HN
N
2
1
N
N
O
F
13
12
307
26
2.0
5.9
4.5
9
8
3
N
O
HN
NH
N
N
HN
N
N
O
ND
675
750
1
1.8
4.4
4.3
2
8
ND
2
2
F
N
N
O
HN
Table 6: Purine N(9) variation. The larger tert-butyl group afforded the highest selectivity margin over WT but at
est
the expense of higher clearance and lower LipE. Ki derived as previously published. H1975 and A549 IC50
represent activities against T790M/L858R and WT EGFR respectively. LipE calculated from IC50 and experimental
logD. HLM: human liver microsomal clearance. RRCK: permeability as measured using low-efflux MDCK-LE cell
line. hHep: human hepatocyte intrinsic clearance.
2
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
t
Introduction of the Bu N(9) group (19) afforded an incremental increase in the WT/DM ratio,
but concomitant increase in clearance driven by the extra lipophilicity prevented progression.
t
Generally across a number of purines, Bu N(9) did result in the best WT selectivity when
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
t
compared with other alkyl groups ( Bu > iPr ≈ Me > Et as seen in the pairwise plot, Figure 9),
i
with Pr and Me N(9) being approximately equal, and Et being the least selective over WT.
Figure 9: Matched Pairwise Plot by N(9) Substituent. Plot illustrated cellular pEGFR WT selectivity seen by
varying N(9) substituent size; the larger tert-butyl group consistently afforded the highest selectivity. The methyl
substituent examples go against the trend and in many cases improved WT selectivity relative to the corresponding
ethyl or isopropyl analog.
2
1
ACS Paragon Plus Environment

Page 23 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
t
When considered from a LipE perspective, the Bu N(9) analog 19 had a much lower LipE of 2.9
compared to the methyl analog 21 (LipE 4.5) vs. WT EGFR for reasons discussed above. The
ethyl analog 20 was able to bind with the ethyl group orthogonal to the core and is buried in a
hydrophobic region of the G-loop, making no steric clash with the WT or T790M proteins, hence
WT selectivity is reduced (supporting information, Figure SF4). The biochemical and cellular
potency of methyl N(9) compound 21 was within two-fold of the isopropyl analog 16. However,
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
1 (most likely because of the reduction in logD) possessed greatly improved in vitro clearance
2
1
-6
(hLM 8 µL/min/mg and hHEP 3 µL/min/MM), high permeability (RRCK 9 x10 cm/sec) and
increased LipE. Additionally, this was accompanied by a loss of CYP450 time-dependent
inactivation (TDI) that was observed for examples with higher logD (data not shown). With the
improved physiochemical profile (MW 415 g/mol, tPSA 115, pKa 3.53, logD 2.0) and resultant
superior ADME properties, excellent potency and 26-fold selectivity over the WT, the N(9)
2
2
methyl analog 21 (PF-06747775) was deemed to have the best overall balance of properties
and was selected for full characterization.
Pre-Clinical Evaluation Overview of 21.
-1 -1
The covalent inhibitory potency (kinact/K 350,000 – 610,000 M s ) of 21 was high for both
i
double mutant proteins, significant for single activating mutant proteins (kinact/K 68,000 –
i
-1 -1
-1 -1
1
03,000 M s ), and weak for WT EGFR (k /K = 12,000 M s ) (full characterization details
inact i
included in supporting information, Table ST5). Cellular potency assessment of 21 against the
EGFR single mutants (Del and L858R) and the double mutants (T790M/Del and T790M/L858R)
revealed single digit nanomolar IC s for all but the T790M/L858R isoform which was had an
5
0
IC50 of 12 nM (Table 7). Selectivity for inhibition of the four activating EGFR mutants vs. WT
was 26 fold when compared to the least active EGFR mutant (T790M/L858R H1975). Compared
2
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
1
to our previously disclosed compound 1 , the mutant EGFR isoform inhibition potency was
6
f
improved for both single mutants. Other clinical T790M inhibitors Tagrisso (AZD9291 , 23),
6
g
and rociletinib (CO1686 , 24) (from Astra Zeneca and Clovis, respectively) are included for
comparative reference. When the WT selectivity was assessed against the weakest activating
mutant activity, 21 was nearly ten-fold better than 23 (26- vs 3-fold selective) and slightly better
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
23
than 24 (26- vs 20-fold selective). Also shown is first generation quinazoline inhibitor erlotinib
25), which despite being potent against the single EGFR mutants lacks potency against the
(
T790M double mutants. Inclusion of a warhead onto a typical quinazoline scaffold (26) rescues
the T790M activity, but with no selectivity observed over WT.
L858R/T790M
IC50 (nM)
Del/T790M
IC50 (nM)
L858R
Del IC50
(nM)
WT IC50
(nM)
Ratio
(WT / least
potent
IC50 (nM)
Compound ID
Structure
H1975
PC9-DRH
H3255
PC9
A549
target)*
N
N
HN
N
21
N
N
F
12
3
4
5
307
26
36
O
O
HN
N
N
HN
Cl
N
1
HN
N
O
O
N
1
1
3
5
7
21
36
140
5,100
N
N
O
N
HN
N
N
O
2
3 (AZD9291)
O
N
13
56
152
3
N
H
(AstraZeneca)
N
F
F
F
N
HN
N
NH
2
4 (CO1686)
Clovis)
O
1
6
9
46
79
1598
20
NH
(
N
N
O
O
2
3
ACS Paragon Plus Environment

Page 25 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
HN
N
O
O
O
O
N
9003
9736
ND
10
11
11
1
2
5 (Erlotinib)
F
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
O
HN
N
Cl
26
HN
O
7
1
1
6
1
N
Table 7: in vitro profile comparison of pEGFR IC50 (nM) in NSCLC cell lines for 21 and clinical
EGFR_T790M inhibitors. *Cell ratio expressed as WT / least potent target, except for Erlotinib which is
considered on single mutant data.
Kinase selectivity of 21 was screened using enzyme assays against a total of 273 kinases
constituting approximately 50% of the human catalytic protein kinome (supporting information,
Table ST3). Under assay conditions of ATP concentration equal to K with 1 mM of 21, 15
m
kinases of the 273 tested were inhibited > 50% (supporting information, Table ST4). These
1
5 kinases were tested in follow up enzyme assays where conditions were modified by
increasing the ATP concentration to simulate cellular ATP levels. Under these more
physiologically-relevant conditions, 7 kinases showed > 50% inhibition (ERBB4, BTK, BMX,
TXK, TEC, ROS, and ACK/TNK2). It is notable that of these 7 kinases, the first 5 listed have a
cysteine residue in the same position as that of EGFR-Cys797, the residue that is targeted for
covalent modification by 21, and hence these kinases are subject to time-dependent irreversible
inhibition. Only two of the 6 tested non-cysteine kinases (FER and ACK/TNK2) were inhibited
with moderate potency, albeit with a selectivity margin of > 9-fold as compared to EGFR mutant
target cellular potencies of 21.
To extend this selectivity analysis to the more physiologically-relevant setting of intact cells,
selected kinases were tested for inhibition by 21 in the cellular context (supporting information,
Table ST4). Four tested kinases (BTK, JAK3, ROS, and BLK) showed weak inhibition in the
2
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
cellular context. Collectively, the kinase selectivity analysis of 21 using standard enzymatic and
cellular formats identified 7 kinases of the 273 tested (ERBB4, ERBB2, BMX, TXK, TEC, FER,
and ACK/TNK2) with the potential for inhibition by 21, yielding a high selectivity index of
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
9
7%.
Proteome Reactivity Assessment: In order to assess the total cellular proteome reactivity of 21,
(and to compare relative to 1), an alkynylated variant was synthesized (at a position that
produced a minimal change in terms of target protein reactivity and specificity, designated here
2
4
P_21, figure 10) . The physiochemical properties of this chemical probe including its cellular
IC against WT and double mutant EGFR, k /I and GSH reactivity closely matched that of the
5
0
obs
corresponding parent molecules making this probe an adequate surrogate of its parent inhibitor
(supporting information, Table ST6). The generation and characterization of the chemical probe
1
for 1, called here P_1, has been previously described . The reactivity of these two probes was
compared to that of a chemical probe based on 26 (P_AQ – see Table 7), a quinazoline-based
covalent second generation pan-ERBB2 inhibitor, which contains a similar Michael acceptor
acrylamide. Initial experiments were performed in the biologically relevant H1975 lung cancer
cells, as this line expresses the double mutant L858R T790M form of EGFR. The in situ
reactivity of these three probes was assessed in a competitive mode by gel-based analysis (Figure
1
0). We were gratified by the lack of proteome-wide reactivity observed with the chemical
probe P_21 and P_1 compared to P_AQ in both the particulate and soluble proteomes. In fact,
for P_21 the only protein that reacts and competes with parent inhibitor 21, would be predicted
to be EGFR (marked with *). Finally, P_21 displays even less proteome reactivity that P_1
2
5
ACS Paragon Plus Environment

Page 27 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
under these conditions highlighting the exquisite selectivity and reactivity of this third-
generation T790M inhibitor.
HN
N
F
Cl
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
N
N
N
O
HN
N
Cl
HN
N
O
O
N
HN
N
N
HN
O
F
N
O
N N
O
HN
N N
O
P_AQ
P_1
P_21
Figure 10: Competitive in situ labeling profiles of 1 µM of each chemical probe P_1 and P_21, vs. a
representative aminoquinazoline probe, P_AQ. H1975 cells were treated 1 µM parent inhibitor (26, 1, 21) or an
equal amount of DMSO for 2 hours. Next, 1 µM of the corresponding chemical probe is added for 3 hours.
Representative proteome labeling profiles are also generated with 100 nM of each chemical probe for 3 hours. After
labeling, cells are harvested, lysed and separated into particulate (left) and soluble (right) proteomes. Equal amounts
2
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
of protein lysate were reacted with azide-rhodamine and covalently labeled proteins were revealed by gel-based
fluorescence (gels are shown in grayscale). * Highlights a protein in the particulate proteome predicted to be EGFR
by molecular weight that is fully competed with parent inhibitor.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Microsomal and hepatic in vitro clearance was species dependent, with mouse and rat showing
high clearance and dog and human showing low clearance (Table 8), with little or no inhibition
of cytochrome P450 isoforms. Blood stability was high across all species, and 21 was found to
distribute preferentially into plasma relative to whole blood in all species except dog (blood to
plasma ratio ranged from 0.5-1.2). Plasma protein binding was relatively low. Metabolite
identification using hepatocytes and microsomes revealed no active metabolites were observed.
Glutathione adducts and demethylation were the major observed pathways (supporting
information, Figure SF5 and Table ST8).
Mouse
74
Rat
141
183
237
Dog Human
Microsomal Clint,app (mL/min/kg)
Hepatic Clint,app (mL/min/kg)
Blood Half Life (mins)
3
< 6
3
196
17
306
279
348
0
.20-
Protein Binding (fu)
0.22
0.75
0.27
0
.49*
Human Cytochrome P450 inhibition
<
30% @ 10 µM
(3A4, 2D6, 2C8, 2C9, 1A2)
CL (mL/min/kg)
53
49
0.66
0.28
11
12
0.94
1.3
66
-
-
-
-
p
Vdss (L/kg)
1.48
0.56
60
T (hours)
1/2
F %
2
7
ACS Paragon Plus Environment

Page 29 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Table 8: Additional key in vitro and single–dose PK summary for 21. *Fu in mice was found to be strain dependent.
Intravenous clearance, Vdss and T1/2 data assessed for mouse (female Nu/Nu), rat (male Wistar-Han) and dog (male
beagle) using a 1 mg/kg solution. Bioavailability (F %) determined via dosing a 1 mg/kg crystalline suspension in
mouse, 30 mg/kg in rat and 3 mg/kg in dog. Half-life based on the terminal β-phase.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Following i.v. administration at 1 mg/kg to mice, rats and dogs, 21 exhibited a moderate to high
plasma clearance, and low steady state volume of distribution (Vd ), resulting in a short plasma
ss
half-life (CL , Table 8). After oral dosing in preclinical species, 21 was rapidly absorbed with
p
low to moderate oral bioavailability, which is believed to be driven by first pass metabolism
(based on rat portal vein studies). In preclinical species, hepatic clearance was predominant
when compared to non-hepatic clearance. The route of clearance was discerned by scaling the
hepatic blood clearance utilizing the CL obtained from non-clinical species hepatocytes, the
int
blood-to-plasma ratio, and the free fraction in plasma as opposed to the clearance observed in
intravenous studies. In all non-clinical species tested, the clearance scaling method from
hepatocytes predicted approximately 2 fold or less than what had been observed in vivo for 21.
Hence in vitro human hepatocyte scaling was deemed sufficient for the human clearance
prediction. In humans, the major route of clearance of 21 is likely to be the hepatic metabolic
clearance, which includes oxidative metabolism and glutathione conjugation via glutathione S-
2
5
th
th
transferase . The human clearance was predicted to be approximately 1/20 to 1/30 of hepatic
blood flow and the predicted terminal half-life range was 12 to 13 hours. Oral bioavailability was
predicted to be moderate (~80%) with a Tmax of ~4 hours. Details of the full evaluation of 21 in
vivo using pharmacokinetic/pharmacodynamic (PK/PD) and tumor growth inhibition (TGI)
studies will be disclosed in a separate publication.
2
8
ACS Paragon Plus Environment