Noncovalent Mutant Selective Epidermal Growth Factor Receptor Inhibitors: A Lead Optimization Case Study

Article abstract of DOI:10.1021/acs.jmedchem.5b01412

Because of their increased activity against activating mutants, first-generation epidermal growth factor receptor (EGFR) kinase inhibitors have had remarkable success in treating non-small-cell lung cancer (NSCLC) patients, but acquired resistance, through a secondary mutation of the gatekeeper residue, means that clinical responses only last for 8-14 months. Addressing this unmet medical need requires agents that can target both of the most common double mutants: T790M/L858R (TMLR) and T790M/del(746-750) (TMdel). Herein we describe how a noncovalent double mutant selective lead compound was optimized using a strategy focused on the structure-guided increase in potency without added lipophilicity or reduction of three-dimensional character. Following successive rounds of design and synthesis it was discovered that cis-fluoro substitution on 4-hydroxy- and 4-methoxypiperidinyl groups provided synergistic, substantial, and specific potency gain through direct interaction with the enzyme and/or effects on the proximal ligand oxygen atom. Further development of the fluorohydroxypiperidine series resulted in the identification of a pair of diastereomers that showed 50-fold enzyme and cell based selectivity for T790M mutants over wild-type EGFR (wtEGFR) in vitro and pathway knock-down in an in vivo xenograft model.

Full text of DOI:10.1021/acs.jmedchem.5b01412

Subscriber access provided by NEW YORK MED COLL
Article
Noncovalent Mutant Selective Epidermal Growth Factor
Receptor Inhibitors: A Lead Optimization Case History
Robert Andrew Heald, Bryan K Chan, 2. Callie Bryan, Charles Eigenbrot, Christine Yu, Daniel Burdick,
Emily J. Hanan, Emily Chan, Gabriele Schaefer, Hank La, Hans Purkey, Jamie Knight, Ivana Yen, Philip
Jackson, Krista Bowman, Kyle Mortara, Michael Lainchbury, Lily Shao, Richard Elliott, Eileen Seward,
Shiva Malek, Stephen Schmidt, Steve Sideris, Timothy P Heffron, Kuen Yeap, Yuan Chen, and Sam Mann
J. Med. Chem., Just Accepted Manuscript • Publication Date (Web): 10 Oct 2015
Downloaded from http://pubs.acs.org on October 11, 2015
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 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Noncovalent Mutant Selective Epidermal Growth
Factor Receptor Inhibitors: A Lead Optimization
Case Study
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
,
∆
†
†
†
⊥
Robert Heald,* Krista K. Bowman, Marian C. Bryan, Daniel Burdick, Bryan Chan, Emily

§
◊
‡
∆
†
Chan, Yuan Chen, Saundra Clausen, Charles Eigenbrot, Richard Elliott, Emily J. Hanan,
∆
∆
§
∆
◊
∆
Philip Jackson, Jamie Knight, Hank La, Michael Lainchbury, Shiva Malek, Sam Mann,

†
#
◊
⊥
Mark Merchant, Kyle Mortara, Hans Purkey, Gabriele Schaefer, Stephen Schmidt, Eileen
∆
◊

†
∆
◊
‡
Seward, Steve Sideris, Lily Shao, Shumei Wang, Kuen Yeap, Ivana Yen, Christine Yu,
†
Timothy P. Heffron
∆
Argenta, Early Discovery Charles River, 7/9 Spire Green Centre, Flex Meadow, Harlow, Essex
†
‡
§
CM19 5TR, United Kingdom. Departments of Discovery Chemistry, Structural Biology, Drug
◊
#
Metabolism and Pharmacokinetics, Biochemical and Cellular Pharmacology, Molecular

⊥
Oncology, Research Oncology and Protein Expression Genentech Inc., 1 DNA Way, South San
Francisco, California, 94080, United States.
Due to their increased activity against activating mutants, first-generation epidermal growth
factor receptor (EGFR) kinase inhibitors have had remarkable success in treating non-small cell
lung cancer (NSCLC) patients, but acquired resistance, through a secondary mutation of the
gatekeeper residue, means that clinical responses only last for 8-14 months. Addressing this
ACS Paragon Plus Environment
1

Journal of Medicinal Chemistry
Page 2 of 60
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
unmet medical need requires agents which can target both of the most common double mutants:
T790M/L858R (TMLR) and T790M/del(746-750) (TMdel). Herein we describe how a
noncovalent double-mutant selective lead compound was optimized using a strategy focused on
the structure-guided increase in potency without added lipophilicity or reduction of 3-
dimensional character. Following successive rounds of design and synthesis it was discovered
that cis-fluoro substitution on 4-hydroxy and 4-methoxy piperidinyl groups provided synergistic,
substantial, and specific potency gain either through direct interaction with the enzyme and/or
effects on the proximal ligand oxygen atom. Further development of the fluoro-hydroxy
piperidine series resulted in the identification of a pair of diastereomers which showed 50-fold
enzyme and cell based selectivity for T790M mutants over wild-type EGFR (wtEGFR) in vitro
and pathway knock-down in an in vivo xenograft model.
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
GRAPHICAL ABSTRACT
Gatekeeper: Thr/Met790
Identical wtEGFR (cyan) and T790M/ L858R (gray) binding modes,
>
100-fold enzyme and cell selectivity.
ACS Paragon Plus Environment
2

Page 3 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
INTRODUCTION
Activating mutations of the receptor tyrosine kinase epidermal growth factor receptor (EGFR),
most commonly the point mutation L858R or deletions within exon 19, increase EGFR-driven
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
1-4
cell proliferation and survival leading to oncogene addiction in non-small cell lung cancer
5
(NSCLC) bearing these mutations. Successful treatment of EGFR-mutated NSCLC with
6
-10
erlotinib or gefitinib
tyrosine-kinase inhibitor).
mutation of the gatekeeper residue threonine-790 to methionine (T790M).
maintains catalytic function of the enzyme but reduces the activity of inhibitors gefitinib and
is followed in most cases by acquired resistance to the respective TKI
1
1-13
(
In particular, around 60% of patients acquire a secondary
1
2-16
This mutation
1
7
erlotinib through occlusion of part of the binding site similarly to the T315I gatekeeper residue
1
2, 13, 18-20
resistance mutation in Abl following treatment with TKIs in CML.
A second
mechanism, in which the T790M containing mutants have an increased affinity for ATP,
resulting in reduced cellular potency for the ATP-competitive inhibitors has also been
2
1
discovered.
The predominant strategy employed by medicinal chemists to overcome these resistance
mechanisms has been the use of covalent inhibitors which are able to bond with the poorly
conserved Cys797 within the EGFR active site, providing both potency and kinase-selectivity.
Clinical results with these second and third-generation compounds have been mixed. Clinical
efficacy of second-generation EGFR inhibitors containing reactive groups was limited by
associated skin rash and gastrointestinal toxicity, possibly because of their potency against wild-
2
2, 23
type EGFR (wtEGFR).
These results and additional reports of acquired resistance to one
2
4, 25
such covalent inhibitor via the T790M mutation
led to concerns that drug levels sufficient to
inhibit T790M mutant forms of EGFR may not be achieved with these agents. Recent results
ACS Paragon Plus Environment
3

Journal of Medicinal Chemistry
Page 4 of 60
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
(
(
disclosed after the conclusion of this work) for third-generation covalent inhibitors AZD9291
mereletinib) and CO-1686 (rociletinib), which have been designed to be selective for T790M
containing EGFR mutants over wtEGFR, indicate promising efficacy and tolerability with this
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
26-33
approach.
However, as with first and second-generation compounds, resistance has emerged:
there has been a recent report of C797S mutation or loss of the T790M mutation in cell-free
3
4
plasma DNA samples from patients who have developed resistance and a second separate
3
5
report of C797S mutation in biopsy samples from a single patient. Additionally, studies with
third-generation resistant cell lines have shown that the allelic context of the activating
gatekeeper and C797S mutations affects the sensitivity to the three generations of inhibitors with
3
6
no EGFR TKIs alone or in combination able to suppress activity when the mutations are in cis.
These data suggest that there is a need for drugs which do not rely on covalent reaction with
Cys797 for potency or selectivity. The dual ALK/EGFR inhibitor AP26113 (brigatinib) is one of
3
7
the few noncovalent EGFR T790M selective inhibitors to be resported although the primary
focus of clinical investigation appears to be in ALK (anaplastic lymphoma kinase) driven lung
38
cancers.
3
9
Our previous publication describes the first stage of our work to identify noncovalent
mutant-selective inhibitors of the major resistant mutants of EGFR: T790M/L858R (TMLR) and
T790M/del(746-750) (TMdel), with high selectivity over wtEGFR.
Using structure-based
design following an HTS campaign a novel series of pyrimidinyl-pyridines which offer
selectivity for inhibition of TMLR and TMdel over wtEGFR and kinases in general were
discovered. This selectivity is due to an unusual binding mode (1, Figure 1) where the
pyrimidine ring sits underneath the mutated Met790 gatekeeper residue. Methionine-aryl
interactions are frequently observed in protein structures with the methionine sulfur often found
ACS Paragon Plus Environment
4

Page 5 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
40
to be positioned in the same plane as the aryl ring. Additionally, the binding sites of adenine
rings often feature methionine residues with the sulfur above the plane of the aromatic ring at a
distance of ~4Å, as exemplified in the complex of catechol-O-methyltransferase with its cofactor
S-adenosylmethionine (PDB code 2cl5). The Met790 gatekeeper residue of the TMLR enzyme
and pyrimidine ring of 1 are in a similar juxtaposition (Figure 1c). The pyrimidine ring is also
held in place through a hydrogen bond with the non-conserved Thr854 residue. This residue, or
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
4
1
a corresponding hydrogen bond donor is unusual in combination with a methionine gatekeeper.
The methoxy piperidine, which was found to be key to selectivity and stabilization of this
binding mode, fills the lipophilic space between Met790 and Lys745 placing the ether oxygen, a
weak hydrogen-bond acceptor, within hydrogen-bonding distance with the conserved catalytic
lysine (Figure 1). Successful X-ray refinement places the methoxyl oxygen in the preferred
equatorial conformation, while the methoxyl methyl carbon is less strongly indicated. The
aminopyridine core makes two additional hydrogen-bonding interactions with the hinge, with the
pyrimidine CH also within close contact with the carbonyl oxygen of Gln791. Finally, the N-
substituent on the ring fused to the pyridine (imidazole in the case of 1) fills the lipophilic ribose
pocket formed by Val726, Leu844 and Cys797. A cysteine residue in this position in the ribose
pocket is unusual in kinases so although the interactions of this group are lipophilic this may still
contribute to selectivity.
ACS Paragon Plus Environment
5

Journal of Medicinal Chemistry
Page 6 of 60
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
Lys745
Met790
Val726
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
Leu844
Cys797
b
a
Met790
c
Figure 1. Representations of the X-ray structure of 1 complexed with TMLR (PDB code 5C8K).
Some residues removed for clarity. a) Chemical structure of 1 and interaction map with TMLR,
purple arrows indicate hydrogen bonds; b) overview of 1 in the binding site, the surface shown is
colored by electrostatic potential, hydrogen bonds indicated by dashed gray lines; c) close-in
image showing the position of the pyrimidine and piperidine rings relative to the mutated
gatekeeper residue, inter-atom distances for the methionine gatekeeper sulfur and heavy atoms of
the pyrimidine ring and piperidine and shown by dashed purple lines, values in Å.
It is generally accepted that with increasing lipophilicity (LogP) the risk of promiscuous
42,43
binding and toxicity is increased, and metabolic stability and solubility are reduced.
Similar
4
4
relationships have been postulated to exist with reduced 3-dimensional character and increased
ACS Paragon Plus Environment
6

Page 7 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
45
aromatic ring count. These risk factors were anticipated to be a particular challenge in our
optimization program for a number of reasons:
1
. The EGFR ATP binding-site is surrounded by mostly lipophilic residues, reflected in the
highly lipophilic nature of optimized compounds erlotinib and gefitinib (clogP 3.8, 4.3;
measured Log D 3.1, 3.6 respectively). In T790M EGFR lipophilicity is increased by
substitution of the only H-bonding residue used by quinazoline inhibitors (excluding
hinge interactions) T790 with methionine.
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. The binding site occupied by our lead compounds is flatter than in the case of anilino-
quinazoline EGFR inhibitors where the aryl ring sits out of plane and is orientated
towards the gatekeeper pocket (see pdb entry 1M17).
3
. Achieving cellular potency and selectivity is made more difficult by the low K [ATP] of
M
46
T790M mutant EGFR (TMLR = 1.3 µM, TMdel = 2.1 µM ), compared to many other
kinases. Literature values for K [ATP] vary significantly although it is generally
M
accepted that the activating mutations have increased K
M
compared to wtEGFR and
T790M kinases.
There is much discussion in the medicinal chemistry literature about the correlation of
lipophilic ligand efficiency (LLE) with “compound quality” and the use of this and similar
4
7-50
metrics in lead optimization programs.
Due to the particular challenges envisaged with the
optimization of our series of T790M EGFR inhibitors and centered on the thermodynamic basis
51
for the use of lipophilic efficiency in optimizing enthalpic contributions to potency our strategy
was to analyze potency in the context of measured LogD values. This was in order to avoid
hydrophobic-effect driven potency increases which may well be non-selective and lead to
ACS Paragon Plus Environment
7

Journal of Medicinal Chemistry
Page 8 of 60
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
ADMET issues. In concert with this we sought to retain, or add to, the 3-dimensional character
present in lead compounds. Finally, we made structural changes which removed metabolic soft
spots identified in in vitro and in vivo studies focusing on increasing in vitro metabolic stability
and lowering unbound in vivo clearance.
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
RESULTS AND DISCUSSION
The azabenzimidazole sub-series was chosen over other 5,6- and 6,6- systems due to superior
physicochemical properties, chiefly solubility. Initially we carried out a survey of the effect of
the azabenzimidazole N-substituent changes on enzyme affinity (Table 1). The TMLR enzyme
assay alone was used primarily for optimization as a very strong correlation was found for
TMLR vs TMdel affinity. Generally, the potency of these analogues correlated with
lipophilicity, which could be expected based on the lipophilic nature of the ribose pocket.
Ligand efficiency was retained in comparison with the unsubstituted analogue 2. The isopropyl
group appeared near-optimal at this stage and was retained in the preparation of further
analogues due to the combination of good potency and LE but without excessive lipophilicity
(
LogD) and the absence of a chiral center.
Table 1. TMLR enzyme potency of azabenzimidazole N-substituent analogues. Ent 1 and Ent 2
denote separated single unknown stereoisomers.
H
N
N
N
N
N
N
N
O
1
R
ACS Paragon Plus Environment
8

Page 9 of 60
Journal of Medicinal Chemistry
1
2
3
4
TMLR
Kiapp
(nM)
1
TMLR
LLE/LE
compd
R
a
LogD
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
2
4.8/0.36
20
2.9
0
1
2
3
4
H
4.6/0.37
4.3/0.35
329
201
1.9
2.4
5 3
6
7
8
4
4.8/0.36
4.9/0.38
87
40
2.3
2.5
9
0
1
2
3
4
5
6
5
7 6
4.5/0.36
48
14
2.8
2.8
8
9
0
1
2
7
5
4
.1/0.38
.8/0.38
(
racemic)
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
8
17
2.9
(
ent 1)
9
4.8/0.38
5.0/0.32
4.3/0.34
4.3/0.35
22
2.9
1.8
2.8
3.4
(ent 2)
1
1
0
1
149
71
O
F
F
F
5 12
18
6
7
8
9
0
ACS Paragon Plus Environment
9

Journal of Medicinal Chemistry
Page 10 of 60
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
13
5.0/0.33
4.9/0.35
76
27
2.1
2.7
O
F
14
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
F
F
a
LLE was calculated using measured LogD; Ligand efficiency, LE = (G)/N where N is the
number of non-hydrogen atoms. Wavy line indicates the point of attachment to the imidazole N.
See experimental section for details of assay errors and controls.
Compound 5 was profiled fully and found to be labile in liver microsomes and hepatocytes
across species (human and rat liver microsomal Clhep 15 and 44 mL/min/kg respectively) and
also in the absence of co-factor (NADPH) in human liver microsomes (Clhep 9 mL/min/kg).
Incubation of a number of compounds in this series in human liver cytosol in the absence of co-
factor (NADPH) showed disappearance of parent and the formation of an oxidative (M+16)
metabolite. These data indicated likely involvement of non-CYP mediated oxidation through an
enzyme such as aldehyde oxidase (AO) / xanthine oxidase (XO). AO and XO are known to
2
52
metabolize various heterocycles, mainly at CH next to sp ring nitrogen atoms.
From the
binding mode of this series we anticipated that only the azabenzimidazole 2-position could
tolerate substitution with a methyl group without loss of activity. Encouragingly, 15, the 2-
methyl analogue of 5 (Table 2) showed a modest increase in potency and >10-fold selectivity for
inhibition of EGFR auto-phosphorylation in the TMLR-expressing cell line H1975 over the
wtEGFR cell line H292. Furthermore, 15 showed no loss of parent on incubation with cytosol as
well as improved stability in human hepatocytes. In vivo clearance in rat was in line with that
53
extrapolated from in vitro data suggesting that in vivo clearance is mainly driven by hepatic
metabolism and that PK for the series could be optimized through analysis of in vitro metabolic
stability data. The major metabolites identified following incubation of 15 in human liver
ACS Paragon Plus Environment
1
0

Page 11 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
microsomes were demethylation (M1, 26%), hydroxylation plus glucoronidation (M2, 10%) and
hydroxylation plus glucoronidation of M1 (M3, 9%). The in vitro potency, ADME and rat PK
profile of 15 are shown in Table 2.
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
Due to the clear metabolic liability of the methoxy piperidine and the probable weak hydrogen-
bonding ability of the ether oxygen the piperidine substituent was targeted for optimization in the
next round of synthesis. A number of simple analogues were prepared, focusing on removing
the metabolic soft spot and including potentially stronger hydrogen bond acceptors in
approximately the same position as the methoxy group. Data for four close analogues is shown
in Table 3.
Table 2. In vitro potency, ADME and PK profile of 15.
H
N
N
N
N
N
N
N
O
15
TMLR Kiapp (nM)
19
wtEGFR Kiapp (nM)
875
46
WT/TMLR Kiapp
p-EGFR/proliferation H1975 IC50/EC50 (µM)
p-EGFR H292 IC50 (µM)
Kinetic solubility (µM)
0.882/ 6.7
>10
87
cLogP (measured LogD)
3.7 (2.4)
83/40/36/26/14
a
Microsomal Clhep m/r/c/d/h (mL/min/kg)
ACS Paragon Plus Environment
1
1

Journal of Medicinal Chemistry
Page 12 of 60
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
a
Hepatocyte Clhep m/r/c/d/h (mL/min/kg)
77/35/33/21/5
MDCK Papp A:B (x10-6 cm/s)
29
40/620
30
b
Rat IV (5 mg/kg) Cl
/Cl
p u
(mL/min/kg)
F (10 mg/kg) %
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
a
b
m: mouse; r: rat, d: dog, c: cynomolgus monkey h: human; Cl
p
: plasma clearance,
Cl
u
:unbound plasma clearance (Cl
p
/F
u
where F is the fraction unbound in plasma). Papp AB:
apparent permeability, apical to basolateral; F: oral bioavailability. See experimental section for
details of assay errors and controls.
u
Table 3. In vitro potency, LogD, and human in vitro metabolic stability for simple piperidine
analogues.
H
N
N
N
N
N
N
N
2
R
TMLR
Kiapp
HLM/HH Clhep
(mL/min/kg)
2
a
compd
R
cLogP LogD
LLE
(
nM)
1
1
1
5
6
7
OMe
OH
19
3.7
3.3
2.7
2.2
2.4
1.8
1.6
1.7
5.3
5.7
5.6
5.1
14/5
3/14
7/2
34
SO
2
Me
64
c
1
8
CONH
2
151
8/-
a
b
LLE was calculated using measured LogD; HLM: human liver microsomes, HH: human
hepatocytes; 2-H azabenzimidazole scaffold. See experimental section for details of assay
errors and controls.
c
The hydroxy piperidine analogue 16 was of comparable potency to 15, and was much more
stable in HLM (3 cf. 14 mL/min/kg). However, lower stability was observed in hepatocytes
ACS Paragon Plus Environment
1
2

Page 13 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
presumably due to Phase II conjugation of the hydroxyl group of 16. The piperidine sulfone
analogue 17 was ~3-fold less potent but was more stable in both microsomes and hepatocytes.
The amide 18 lost >5-fold potency relative to the methyl ether 15. The LLE values based on
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
54
measured LogD indicate that both 16 and 17 represent a marginal improvement in
potency/lipophilicity balance. An X-ray structure of 17 in complex with the TMLR enzyme
(
Figure 2a) showed Lys745 within hydrogen bonding distance of one of the sulfone oxygen
atoms, with this group in the more energetically favored equatorial conformation. The hydrogen
bond acceptor shifted ~1.4 Å compared to the methoxy piperidine analogue (Figure 2h) and the
lysine moves accordingly to accommodate this. The methyl group of the sulfone is also
positioned within interacting distance with Phe723 and Val726 (3.0 and 3.1 Å respectively).
Perhaps one reason for the lack of potency improvement with this substitution is that one of the
sulfone oxygen atoms is redundant making no clear contribution to binding.
ACS Paragon Plus Environment
1
3

Journal of Medicinal Chemistry
Page 14 of 60
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
Lys745
b
a
Val726
Lys745
Val726
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
Phe723
Asp855
Lys745
c
d
Phe723
Phe723
Asn842
Asn842
Lys745
e
f
Phe723
Lys745
g
h
Lys745
Met790
Phe723
ACS Paragon Plus Environment
1
4

Page 15 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Figure 2. Representations of the X-ray structures of various piperidine-replacement analogues
complexed with TMLR showing close-in images of the pyrimidine 2-substituent and key
interacting residues. Hydrogen bonds are indicated by dashed gray lines. a) 17 (PDB code
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
5
C8M). The sulfone is in the more stable equatorial conformation, hydrogen bond shown
between Lys745 and one of the oxygen atoms of the sulfone, Phe723 is 3-3.2 Å from the methyl
of the sulfone; b) 23 (PDB code 5C8N). The methoxy group is flipped to the axial conformation
but is still able to interact with Lys745 through a hydrogen bond, the pendant amino group is
able to make a salt bridge with Asp855; c) 24 (PDB code 5CAL). Image shown has the pendant
group in the “open” conformation where there is no hydrogen bond with Lys745, the amide NH
2
is within hydrogen-bonding distance with Asn842, one of the methyl groups is positioned 3.7 Å
from Phe723; d) hybrid fit of 24 to the X-ray electron density map, electron density shown in
green e) 27 (green carbons) (PDB code 5CAN) in comparison with the ligand conformation of
2
4 (light gray carbons). The amide NH is positioned so as to enable hydrogen bonding with
2
Asn842 in both structures; f) 29 (PDB code 5CAO). The ligand bound conformation is possibly
stabilized by an intramolecular hydrogen bond between the sulfone and aminopyrimidine NH; g)
30 (PDB code 5CAP). The O-linked analogue of 29, where there is no intramolecular hydrogen
bond with the sulfone, binds with the sulfone more out of plane but still interacting with Lys745;
h) overlay of the X-ray structures of various pyrimidine substituent analogues (17, cyan; 29,
green and 30, gray) showing the conservation in the positioning of the hydrogen bond acceptor
and Lys745 in the complexes of these analogues.
4
,4-Disubstituted quaternary piperidines 19-23 were designed to provide a steric block to
metabolism of the methoxy piperidine or phase-II conjugation of the hydroxyl piperidine.
Additionally these analogues also provided a vector for accessing polar interactions in particular
ACS Paragon Plus Environment
1
5

Journal of Medicinal Chemistry
Page 16 of 60
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
with Asp855 and Asn842. As shown in Table 4, two of these analogues, 21 and 23, showed
retention of potency with increased LLE indicating that the added polar groups, although sub-
optimal, were achieving a positive interaction with the protein. Gratifyingly, compounds 21 and
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
2
3 also have improved HLM stability relative to compound 15 (Clhep = 2 and 6 respectively
compared to 14 mL/min/kg).
Table 4. In vitro potency, LogD, and human in vitro metabolic stability for quaternary piperidine
analogues.
H
N
N
N
O
N
N
N
N
2
R
b
TMLR
Kiapp
HLM/HH Clhep
2
a
compd
R
cLogP LogD
LLE
(
mL/min/kg)
(
nM)
492
77
1
2
9
0
CH
2
N(Me)
OMe
2
3.7
3.6
3.0
3.5
3.2
1.5
2.4
1.7
2.0
0.1
4.8
4.7
5.9
5.3
8/-
14/-
6/10
9/-
CH
2
21
22
CH
2
OH
OH
24
CH CH
2
2
50
b
2
3
CH
2
CH
2
NH
2
28
7.5
2/-
a
b
LLE was calculated using measured LogD HLM: human liver microsomes, HH: human
hepatocytes. See experimental section for details of assay errors and controls.
The X-ray structure of 23 with TMLR (Figure 2b) showed that the basic primary amine is
positioned within an appropriate distance to make a salt-bridge with Asp855. There is no
potency gain with this interaction possibly due to an increased de-solvation penalty,
ACS Paragon Plus Environment
1
6

Page 17 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
displacement of water which usually occupies this position and the fact that Asp855 is already
part of a favorable hydrogen-bonding network in the apo enzyme. The X-ray structure also
showed that 23 binds with the piperidine methoxy axial, now the preferred conformation due to
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
55
the larger A-value of the ethyl amino group, but the hydrogen bond with Lys745 is maintained.
A diverse library of >200 analogs varying the 2-substituent of the pyrimidine was prepared using
commercially available primary and secondary amines. The amines were triaged using
5
6
calculated property filters in an empirically-driven approach to identify a replacement for the
methoxypiperidine moiety which conferred improved metabolic stability and comparable
potency. The vast majority of these analogues were substantially less potent than the methoxy
piperidine (data not shown) with only one compound showing activity < 30 nM, 24, Table 5.
This analogue was significantly more polar than 15 (LogD lowered by 0.8 units), devoid of an
obviously metabolically labile group on the amino pyrimidine, and had improved stability in
liver microsomes and hepatocytes.
Table 5. In vitro potency, LogD, and human in vitro metabolic stability for library analogue 24
and related compounds.
NH
S
O
N
O
O
O
NH2
NH2
NH2
H N
2
O
24
25
26
27
H
N
N
N
O
H
N
N
O
O
O
N
S
S
OH
N
NH2
O
O
3
R
28
29
30
31
ACS Paragon Plus Environment
1
7

Journal of Medicinal Chemistry
Page 18 of 60
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
b
TMLR
Kiapp
(nM)
wtEGFR
Kiapp
(nM)
HLM/HH Clhep
3
a
compd
R
cLogP LogD
LLE
(
mL/min/kg)
2
2
2
2
2
2
3
4
5
6
7
8
9
0
Me
Me
H
22
658
196
124
196
38
390
>2500
>440
681
2.4
3.0
2.1
2.3
3.3
2.0
2.1
2.8
1.5
1.9
1.6
1.6
2.2
1.4
1.6
2.0
6.2
4.3
5.1
5.3
4.5
6.0
5.9
4.5
5/3
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
4/-
Me
Me
Me
Me
Me
4/5
5/<1
5/6
4/1
5/2
>1500
1820
2100
>2500
52
31
323
a
b
LLE calculated using measured LogD
hepatocytes. See experimental section for details of assay errors and controls.
HLM: human liver microsomes, HH: human
The bound conformation of 24 as determined by X-ray showed the N-linked side chain in an
“
extended” conformation (Figure 2c), but an alternate fit to the electron density showed a
“closed” conformation (Figure 2d). In the extended conformation the amide NH is able to
hydrogen bond with Asn842 and the carbonyl appears redundant. In the closed conformation the
amide carbonyl is positioned within hydrogen-bonding distance with Lys745 and the amide NH
2
2
makes no discernable interaction. It is likely that the closed conformation of 24 is the preferred
small molecule conformation due to an intramolecular hydrogen bond between the carbonyl and
NH and the Thorpe-Ingold effect, but both bound conformations may contribute to potency.
The S- and O- linked compounds 25 and 26 are likely to exist in the extended ligand
conformation in solution due to the absence of a strong intramolecular H-bond and the
propensity for O- and S- linked side chains to sit out of plane with the pyrimidine core. Both
analogues are less potent than 24 suggesting that the closed conformation predominates in this
compound and provides greater potency. Supporting this, docking of 24 in the receptor from its
ACS Paragon Plus Environment
1
8

Page 19 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
own X-ray using GLIDE provides a closed binding conformation as the highest scoring result.
Competition between the two binding conformations was further illuminated following the
serendipitous synthesis (see synthesis section below for details) of 27. This compound is only 3-
fold less potent than the ring open analogue and cannot adopt the closed binding conformation,
but less selective. The X-ray structure (Fig. 2e) showed that the pyrrolidine amide binds
analogously to 24 in the open conformation, making a hydrogen bond with Asn859 but not
Lys745, in this case in a low energy small molecule conformation. Compound 27 also retained
good in vitro stability in human liver microsomes and hepatocytes. As 24 suffered from
chemical stability issues (hydrolysis of the amide to acid under acidic conditions, hypothesised to
proceed via a cyclic intermediate formed through reaction on the pyrimidine) alkene-linked and
methyl sulfone analogues were prepared. The alkenyl analogue 28 shows reduced potency,
potentially due to disruption of the ideal bound conformation- the intramolecular hydrogen bond
is lost and allylic strain will favor positioning of the amide in the conformation shown in the
table. Consistent with the removal of metabolic soft spots and low LogD, 28 shows good in vitro
metabolic stability. The sulfone analogue 29 and ether-linked variant 30 both show reasonable
potency, good in vitro metabolic stability and low LogD. X-ray structures of the two compounds
are shown in Figures 2f and 2g. In both cases one of the sulfone oxygen atoms is able to make a
hydrogen bond with Lys745 although the conformation of the linkers is quite different, reflecting
the difference in lowest energy ligand conformation. Interestingly, while having equivalent
TMLR potency, sulfones 29 and 30 are also in a different position than in the piperidine sulfone
analogue 17 (Figure 2h).
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
Although none of the structurally-diverse methoxy piperidine replacements shown above
increased potency, the methoxy piperidine metabolic soft spot was successfully removed with
ACS Paragon Plus Environment
1
9

Journal of Medicinal Chemistry
Page 20 of 60
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
retention of potency and selectivity over wtEGFR. When tested against a panel of 220 kinases at
1 M compounds 17, 29, and 30 retained selectivity compared with 2 whilst 27 was considerably
more promiscuous (see Supporting Info Table S1 page S2). The reduced TMLR/wt and kinase
selectivity of 27 could possibly be due to the change in interactions within the binding site.
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
Table 6. In vitro and in vivo rat clearance for a selection of analogues.
a
b
c
d
RLM Clhep
RH Clhep
IV dose
(mg/kg)
0.5
Rat Cl
p
(Cl )
u
Calce.
pKa
compd
(
mL/min/kg) (mL/min/kg)
(mL/min/kg)
165 (934)
250 (909)
158 (707)
53 (156)
1
2
7
4
26
16
22
19
10
24
10
21
9
3.6
3.5
3.4
0.8
0.5
29
30
31
1.0
0.5
9
0.5
50 (200)
1.3
a
b
c
d
RLM: Rat liver microsomes; RH: rat hepatocytes; Cl
p
: plasma clearance; Cl
u
: unbound
e
plasma clearance (Cl
u
= Cl
p
/F
u
where F is the fraction unbound in plasma); Calculated using
u
ACD Percepta ® (GALAS) using a 2-substituted pyrimidine sub-structure. See experimental
section for details of assay errors and controls.
However, in vivo rat PK of five diverse analogues with improved microsomal and hepatocyte
stability showed high clearance for those which retained TMLR potency (Table 6), with total
clearance in excess of liver blood flow for 17, 24 and 29. Chemical instability due to amide
hydrolysis was noted for 24 but the apparent extra-hepatic clearance of 17 and 29 is not so easily
explained. Both total and unbound plasma clearance were lower for analogues 30 and 31 which
have reduced TMLR potency. It was noted that 30 and 31 have significantly lower calculated
pKa of the pyrimidine nitrogen indicative of reduced electron density. The empirical observation
that increased in vivo metabolic stability correlates with reduced electron density on the
pyrimidine seems logical as the reduced electron density of the heterocycle could slow ring
ACS Paragon Plus Environment
2
0

Page 21 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
oxidation as well as glucuronidation of the hinge-binding NH. This trend was shown to some
extent in the hepatocyte stability data where 30 and 31 are more stable than 17 and 29. Based on
these findings, principally the in vivo data, fluoro piperidine analogues were conceived in order
to retain the potency of the methoxy or hydroxyl piperidine but reduce electron density on the
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
5
7
pyrimidine in order to improve metabolic stability.
It has been proposed that equatorial 3-
5
8
fluoro and 3,3-difluoro substitutions have the greatest effect on piperidine pKa, and thus,
presumably, pyrimidine electron density in the case of our TMLR inhibitors.
Table 7. In vitro potency and physicochemical properties of fluoropiperidine analogues. All
analogues are racemic, relative stereochemistry shown.
H
N
N
N
N
N
N
N
O
3
R
5
R
6
R
4
R
b
p
c
H292
p-EGFR
IC50
Rat Cl
H1975
p-
TMLR wtEGFR
(Cl )
u
4
5
6
a
EGFR IC50
(Prolif.)
(µM)
compd
R3
R
R
R
Kiapp
Kiapp
LogD LLE
(
mL/min/kg)
(
nM)
(nM)
(Prolif.)
(
µM)
2
Me
Me
H
H
F
F
F
F
F
H
H
H
H
H
H
F
20
19
1200
400
2.3
2.3
2.3
2.1
1.5
1.6
2.8
5.4
5.4
6.3
6.3
6.3
6.2
5.3
0.88 (7.0)
2.36 (6.5)
0.25 (0.56)
0.29 (0.88)
1.0 (1.6)
--
32
33
CH
Me
CH OH Me
Me
2
OH Me
--
Me
2.7
4.3
13
1080
150
>10
>10
--
3
3
3
3
4
5
6
7
2
107 (584)
90 (545)
151 (382)
H
H
910
CH
2
2
OH
16
>0.53
249
2.53 (>10)
1.0 (2.38)
>10
>10
CH
OH Me
7.4
ACS Paragon Plus Environment
2
1

Journal of Medicinal Chemistry
Page 22 of 60
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
3
3
4
8
9
0
Me
Me
H
F
H
H
F
F
F
9.5
44
57
>403
1840
540
2.6
2.5
2.1
5.4
4.8
5.1
1.04 (1.7)
--
>10
Me
--
CH
2
OH Me
--
--
a
b
c
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
LLE calculated using measured LogD; Cl
clearance (Cl = Cl /F where F
p
: plasma clearance; Cl
u
: unbound plasma
u
p
u
u
is the fraction unbound in plasma). See experimental section for
details of assay errors and controls.
On profiling the fluoro piperidine analogues (data shown in Table 7) it was immediately clear
that, aside from any potential impact on PK, both 3,3-di-fluoro (37 and 38) and cis mono-fluoro
(33-36) substitution provided a significant increase in potency against TMLR but not wtEGFR in
the biochemical, p-EGFR auoto-phosphorylation and cell proliferation assays. Profiling against
panel of 220 kinases at 1 M confirmed that cis-fluoro substitution only increased potency
against TMLR with no change in the inhibition of other kinases compared to the des-fluoro
analogue: compound 33 inhibited 25 kinases 40-80% and 10 kinases >80% with the same pattern
of inhibition as 2 (see Supporting Information page S2). Conversely, trans mono-fluoro (39 and
4
0) substitution reduced potency. Synthesis of unsubsituted and 3-fluoropiperidine analogues
showed that fluorine substitution alone had no effect on potency (data not shown) further
demonstrating the synergy of the cis OMe/OH and F substituents. Further analysis revealed that
the cis-mono fluoro substitution increased potency without an increase in measured LogD
whereas the increase in potency of the di-fluoro analogues was almost certainly lipophilicity-
driven. This is illustrated in a plot of pKi vs LogD (Figure 3) where both pairs of cis-mono
fluoro piperidine analogues are differentiated.
ACS Paragon Plus Environment
2
2

Page 23 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
LLE 6.5
LLE 6
LLE 5.5
LLE 5
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. TMLR pKi vs measured LogD for hydroxyl and methoxy fluoro piperidine
analogues. Lines of equal LLE, calculated using measured LogD, shown. Points annotated with
compound numbers. Increased LLE is only seen for cis-monofluoro analogues. Green: cis-
monofluoro; blue: des-fluoro; yellow: gem-difluoro; orange: trans-monofluoro.
An X-ray structure obtained from soaking diastereomeric mixture 33 with TMLR
preferentially yielded crystals of the 3R-4S enantiomer (Figure 4). In this structure the fluorine
was in an axial conformation and within a short distance of the methionine gatekeeper as well as
Lys745, possibly positively interacting with either or both of these residues due to the ability of
5
9
fluorine to form polar, hydrophobic or multipolar interactions. It has been shown that this
1
9
amphiphilic character can be correlated to the F NMR chemical shift due to the influence of the
6
0
19
local environment of fluorine on the character of the C-F bond.
F NMR shows a large
dispersion, >300 ppm, depending on the number of oxygen, sulfur, nitrogen and other halogen
atoms in the α, β, and  positions. High shielding (shift > -160 ppm) has been shown to strongly
6
0
correlate with proximity to NH side-chains in PDB structures.
The hydrogen bond formed by
6
1
fluorine may well be very weak but it has been postulated that CHF is a better hydrogen
6
2 19
bonding group than CF
2
.
F NMR for the piperidinyl pyrimidines (see Supporting Info page
S91) showed that for the hydroxyl piperidines the cis-fluorine is more shielded (electron rich)
1
9
than the trans-fluorine ( F  -200 and -188 ppm respectively). It has been discovered that when
ACS Paragon Plus Environment
2
3

Journal of Medicinal Chemistry
Page 24 of 60
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
within 3 Å of an oxygen atom in a low energy conformation, fluorine substitution for hydrogen
6
3
can lead to a lowering, or no increase, in LogD despite the addition of a lipophilic atom.
Indeed it is the case that there is no significant change in LogD between direct pairs 2 and 33, 32
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
19
and 34. The increased shielding observed in the F NMR of the cis-fluoro piperidine is also
possible confirmation of the O-F “interaction”. Although this effect is not fully understood it has
been proposed that fluorine-induced polarization of the proximal oxygen atom could lead to
6
4
stronger H-bonds involving this atom. In cis-fluoro hydroxy and methoxy piperidines on our
scaffold this polarization would increase the hydrogen bond acceptor ability of the
ether/hydroxyl oxygen, strengthening the energy of interaction with Lys745. In trans-fluoro
hydroxyl and methoxy piperidines the F…O distance is also <3 Å but the fluorine is positioned
equatorially and unable to make optimal lipophilic contacts. Fluorine substitution can also affect
piperidine conformation. When positioned on a piperidine ring fluorine prefers to be in the axial
6
5
conformation, but this conformational preference has only been studied in the case of
protonated piperidines where there can be an H-F hydrogen bond. For un-protonated piperidines
steric factors (A 1,3-strain) would still favor the equatorial conformation, albeit marginally. An
additional conformational factor is the gauche effect- the orbital interaction between the  C–H
bond and the * C–F bond, hyperconjugation- which is maximized when the C-H and C-F bonds
are anti-periplanar. Further discussion of the influence of fluoro substitution is presented below.
ACS Paragon Plus Environment
2
4

Page 25 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Lys745
Met790
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 4. Representation of the X-ray structure of 33 (PDB code 5CAQ) showing the
interactions of the fluoro methoxy piperidine with key residues Met790 and Lys745. Inter-atom
distances indicated by dashed purple lines with values in Å. The fluorine is positioned axially
and within 3.5 Å of Met790 and 2.9 Å of the NH of Lys745. The F-O distance is 2.8 Å. 33 was
soaked as mixture of enantiomers, the 3R, 4S enantiomer crystallizing preferentially.
Compared to the des-fluoro analogue 15, no change of rat in vivo clearance was seen with the
cis-fluoro hydroxy or fluoro methoxy analogues (34: 1 mg/kg IV Cl
p
(Cl
u
) 107 (584) mL/min/kg;
35: 5 mg/kg IV Cl (Cl ) 90 (545) mL/min/kg; 36: 0.5 mg/kg IV Cl
u
p
(Cl ) 151 (382) mL/min/kg).
u
Even if the hypothesis of the correlation of lower pyrimidine pKa and electron density were true,
the fluorines in 34, 35 and 36 are axial and so may have a negligible electron withdrawing effect
(see above). In the absence of data for compounds in Table 7 it would be assumed that
metabolism of 34 is driven by de-methylation, and 35-36 driven by Phase II conjugation, and
perhaps this is the case, but in vitro stability data for 34, 35 and 36 still under-predicted in vivo
6
6
clearance. The trans analogues, where the fluorine would be equatorial, were not profiled and
ACS Paragon Plus Environment
2
5

Journal of Medicinal Chemistry
Page 26 of 60
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
for difluoro compounds the increase in lipophilicity would probably counteract any benefit from
reduced electron density.
Due to the TMLR-specific increase in potency without added lipophilicity, the methoxy and
hydroxyl cis mono-fluoro piperidine substituents were combined with small changes to the
azabenzimidazole 1- and 2- substituents focusing on modifications which would block
metabolism of the 2-methyl group and increase lipophilic interactions in the ribose pocket
without introducing further metabolic soft spots. Thus, combinations were made with the
trifluor-isopropyl group (as in 14) together with methyl, hydroxymethyl and 2-hydroxyethyl
groups at the azabenzimidazole 2-position. The rational for making these changes was that the
hydroxymethyl 2-substituent was the active metabolite of the 2-methyl compounds and steric
bulk was added both to block oxidation and glucuronidation of the hydroxyl compounds.
Additionally, the change to trifluor-isopropyl in 14 is LLE neutral- potency and LogD are
increased to the same extent. The most potent combination analogues prepared, 41 and 42
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
(
profiled initially as mixtures of diastereomers, cis OH/F and OMe/F) are shown in Table 8. All
possible stereoisomers of 41 and 42 were prepared but none was more potent (data not shown).
Compared to earlier direct analogues (42 with 36; 41 with 34) a ~10-fold increase in cellular
potency and 3.5 to 6.5 -fold increase in affinity is apparent, improvements which come without
any significant increase in measured LogD (~0.2 units), whereas cLogP is increased ~0.6 units.
The increase in potency against TMLR was shown to be accompanied by retention of kinase
selectivity: when profiled against a panel of 308 kinases at 1 M compounds 42a and 42b (the
individual diastereomers of 42) showed >80% inhibition of only 14 individual kinases (see
Supporting Information table S2 page S3). The analogues 41 and 42 were also tested against the
isolated TMdel enzyme and in a TMdel expressing cell line: erlotinib-resistant PC9 cells (PC9
ACS Paragon Plus Environment
2
6

Page 27 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
(
ER)). These cells are grown in the presence of a high concentration of erlotinib in order to
cause and maintain the T790M resistance mutation. Potency in the biochemical assay against the
TMdel enzyme is significantly weaker than TMLR, a trend which was not evident in earlier
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
39
analogues.
However, cell potency, which is obviously more important, is actually better
against PC9(ER) than H1975. Increased proliferation potency in the PC9(ER) cells could be due
to greater sensitivity to inhibition of the EGFR pathway, but this should not affect pEGFR EC50
The low cell drop-off from TMdel Ki to PC9(ER) pEGFR EC50, given the low K
.
M
[ATP] of this
enzyme, is curious and is perhaps related to the protein construct used in the biochemical assay.
Table 8. Profiles of fluoro-hydroxy and fluoro-methoxy combination analogues.
H
N
N
N
N
N
N
N
O
CF3
OH
F
4
R
(
Piperidine stereochemistry relative, other stereochemistry absolute)
4
4
4
1 (R = Me)
42 (R = H)
TMLR Kiapp (nM)
TMdel Kiapp (nM)
wt EGFR Kiapp (nM)
LogD
1.1
7.0
2.5
6.0
122
2.2
313
1.8
pEGFR H1975 IC50 (nM)
H1975 prolif. EC50 (nM)
pEGFR PC9(ER) IC50 (nM)
PC9(ER) prolif. EC50 (nM)
pEGFR H292 IC50 (nM)
21
261
112
27
900
81
90
170
>10,000
>10,000
ACS Paragon Plus Environment
2
7

Journal of Medicinal Chemistry
Page 28 of 60
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
H292 prolif EC50 (nM)
7,800
>10,000
7/11/73/18/19
4/13/68/4/19
71 (348)
a
Microsomal Clhep h/r/m/d/c (mL/min/kg)
15/28/81/24/36
10/25/75/10/30
23 (334)
a
Hepatocyte Clhep h/r/m/d/c (mL/min/kg)
b
Rat IV (5 mg/kg) Cl
/Cl
p u
(mL/min/kg)
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 ½ (h), Vss (L/kg)
Rat %F
1.7, 3.7
0.71, 3.5
c
d
85
29
a
b
m: mouse; r: rat, d: dog, c: cynomolgus monkey h: human; Cl
unbound plasma clearance (Cl /F where F
p
: plasma clearance, Cl
u
:
c
p
u
u
is the fraction unbound in plasma) ; formulated in
d
1
0% DMSO/10% Cremophor EL in saline; formulated in 10% ethanol/60% PEG100/30% SQ
water. See experimental section for details of assay errors and controls.
Compared with published covalent and noncovalent TMLR inhibitors, compounds 41 and 42
show substantially increased selectivity in cell p-EGFR and proliferation assays (see Supporting
Information Table S5 page S11). Compound 41 is more potent than AP26113, AZD9291 and
WZ4002 in the p-EGFR assay, but ~10-fold less potent than the covalent analogues in
proliferation assays reflecting the different mechanism of action.
The individual diastereoisomers making up 41 were prepared and profiled and found to have
essentially identical in vitro potency (see Supporting Information Table S4 page S11). X-ray
structures with TMLR showed that the fluorine of 41a (3R, 4S) is in the same position as with
earlier analogue 33 but in the diastereoisomer 41b (3S, 4R) (Figure 5a and 5b) the position is
transposed. Here the fluorine is still within close proximity of the piperidine ether oxygen but
not the catalytic lysine or the sulfur of Met790. Instead, the fluorine is within 3.3 Å of Val726
and 3.6 Å of the carbon atom of the methyl of Met790, possibly making lipophilic contacts with
these residues. These structures also revealed the interactions of the azabenzimidazole 1- and 2-
substituents. The hydroxyl group at the azabenzimidazole 2-position is able to make a water-
ACS Paragon Plus Environment
2
8

Page 29 of 60
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
mediated interaction with Asp800 and the backbone NH of Cys797. One of the fluorines of the
trifluoromethyl has the appropriate vector and distance relative to the base of the carbonyl of
6
7
Leu718 to make a positive interaction and the trifluoromethyl group itself fills the lipophilic
ribose pocket. Additionally, the trifluoromethyl is within contact distance (3.2 Å) of the
piperidine, potentially making the bound conformation of the molecule more favorable. A
structure of 41a in wtEGFR was also obtained (Figure 6). Despite the >100-fold lower potency
against the wt enzyme, the binding mode and position of all residues (except the gatekeeper) is
essentially identical indicating that the potency difference arises from the difference in
interactions with the gatekeeper residues.
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
9
For the diastereomeric mixture 42 F NMR shifts of -72 ppm (CF
3
) and -198 ppm (CHF) are
seen consistent with a de-shielded CF fluorine environment and a shielded piperidine CHF,
3
correlated in the literature with hydrophobic/multipolar and polar interactions respectively.
However, the hypothesis that the observed potency increase is due to hydrogen bonding of the
fluorine does not fit well with the rest of the data. The X-ray data, the equivalent potency of the
two diastereomeric pairs 41a/41b and 42a/42b (vide infra) and the specific increase in TMLR
potency over wtEGFR suggests that fluoro substitution increases potency by enhancing the
hydrogen-bonding capability of the ether oxygen and, more importantly, making lipophilic
contacts with the methionine gatekeeper and other neighboring lipophilic residues.
ACS Paragon Plus Environment
2
9