Structure-Guided Design of EED Binders Allosterically Inhibiting the Epigenetic Polycomb Repressive Complex 2 (PRC2) Methyltransferase

Article abstract of DOI:10.1021/acs.jmedchem.6b01473

PRC2 is a multisubunit methyltransferase involved in epigenetic regulation of early embryonic development and cell growth. The catalytic subunit EZH2 methylates primarily lysine 27 of histone H3, leading to chromatin compaction and repression of tumor suppressor genes. Inhibiting this activity by small molecules targeting EZH2 was shown to result in antitumor efficacy. Here, we describe the optimization of a chemical series representing a new class of PRC2 inhibitors which acts allosterically via the trimethyllysine pocket of the noncatalytic EED subunit. Deconstruction of a larger and complex screening hit to a simple fragment-sized molecule followed by structure-guided regrowth and careful property modulation were employed to yield compounds which achieve submicromolar inhibition in functional assays and cellular activity. The resulting molecules can serve as a simplified entry point for lead optimization and can be utilized to study this new mechanism of PRC2 inhibition and the associated biology in detail.

Full text of DOI:10.1021/acs.jmedchem.6b01473

Subscriber access provided by UNIV OF REGINA
Article
Structure-Guided Design of EED Binders Allosterically Inhibiting the
Epigenetic Polycomb Repressive Complex 2 (PRC2) Methyltransferase
Andreas Lingel, Martin Sendzik, Ying Huang, Michael David Shultz, John Cantwell, Michael P Dillon,
Xingnian Fu, John Fuller, Tobias Gabriel, X. Justin Gu, Xiangqing Jiang, Ling Li, Fang Liang, Maureen
Margrethe McKenna, Wei Qi, Weijun Rao, Xijun Sheng, Wei Shu, James Sutton, Benjamin Robert Taft,
Long Wang, Jue Zeng, Hailong Zhang, Maya Zhang, Kehao Zhao, Mika K Lindvall, and Dirksen E. Bussiere
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.6b01473 • Publication Date (Web): 19 Dec 2016
Downloaded from http://pubs.acs.org on December 20, 2016
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 48
Journal of Medicinal Chemistry
1
2
3
4
Discovery Chemistry
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
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 48
1
2
3
4
5
6
Structure-Guided Design of EED Binders Allosterically Inhibiting the
Epigenetic Polycomb Repressive Complex 2 (PRC2) Methyltransferase
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
Andreas Lingel,‡¶* Martin Sendzik,‡§ Ying Huang,† Michael D. Shultz,†§ John Cantwell,‡
Michael P. Dillon,‡& Xingnian Fu,† John Fuller,‡ Tobias Gabriel,†¶ Justin Gu,† Xiangqing Jiang,†
Ling Li,† Fang Liang,† Maureen McKenna,‡ Wei Qi,† Weijun Rao,† Xijun Sheng,† Wei Shu,‡
James Sutton,‡& Benjamin Taft,‡ Long Wang,† Jue Zeng,† Hailong Zhang,†¥ Maya Zhang,† Kehao
Zhao,†§ Mika Lindvall,‡ Dirksen E. Bussiere‡#
‡Novartis Institutes for BioMedical Research, 5300 Chiron Way, Emeryville, California 94608,
United States
†Novartis Institutes for BioMedical Research, 2418 Jinke Road, Shanghai 201203, China
Keywords:
Epigenetics, allosteric inhibition, fragment-based lead generation, oncology, EED, PRC2
ACS Paragon Plus Environment

Page 3 of 48
Journal of Medicinal Chemistry
1
2
3
4
Abstract
5
6
7
8
9
PRC2 is a multisubunit methyltransferase involved in epigenetic regulation of early embryonic
development and cell growth. The catalytic subunit EZH2 methylates primarily lysine 27 of
histone H3, leading to chromatin compaction and repression of tumor suppressor genes.
Inhibiting this activity by small molecules targeting EZH2 was shown to result in anti-tumor
efficacy. Here, we describe the optimization of a chemical series representing a new class of
PRC2 inhibitors which acts allosterically via the trimethyllysine pocket of the non-catalytic EED
subunit. Deconstruction of a larger and complex screening hit to a simple fragment-sized
molecule followed by structure-guided regrowth and careful property modulation were employed
to yield compounds which achieve sub-micromolar inhibition in functional assays and cellular
activity. The resulting molecules can serve as a simplified entry point for lead optimization and
can be utilized to study this new mechanism of PRC2 inhibition and the associated biology in
detail.
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
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 48
1
2
3
4
Introduction
5
6
7
8
9
Polycomb Repressive Complex 2 (PRC2) is a conserved methyltransferase playing a major role
in diverse biological processes such as chromatin remodeling and epigenetic silencing.1 The core
of PRC2 is composed of four proteins: EZH2 (or EZH1), EED, Suz12, and RBAP48. The SET
domain of EZH2 contains the S-adenosylmethionine (SAM) dependent methyltransferase
activity which catalyzes the transfer of a methyl group onto the side chain amine of lysine
residues in histone tails, in particular H3K27. The enzyme can accept lysine side chains with
various methylation states (non-, N-mono- and N,N-dimethylated) as substrates, resulting
eventually in N,N,N-trimethylated lysines as a product (H3K27Me3 in case of H3K27
methylation). In addition to EZH2, the presence of EED and Suz12 was shown to be required for
enzymatic activity.2 This was rationalized by structural studies on the entire human PRC23 and
minimal versions of the PRC2 containing only EZH2, EED and truncated Suz12.4–6 These
structures revealed that the three proteins form an intertwined complex in which the catalytic
center is formed by EZH2 but is stabilized by domains of EED and Suz12.
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
Gain-of-function mutations and overexpression of EZH2 are well documented to be involved in
oncogenic transformation, progression of various cancer types, as well as to be associated with
poor prognosis.7–10 Due to the link between EZH2 and cancer, multiple groups have developed
small molecule antagonists to study the disease relevance, biological consequences and efficacy
of inhibiting the enzymatic activity of PRC2.11–18 All low molecular weight inhibitors reported so
far are targeting the EZH2 subunit and are SAM competitive (reviewed for example by Kim et
al.19). It has been shown that such molecules result in reduced H3K27Me3 levels and have an
antiproliferative effect in both cell culture and animal models in EZH2 mutant tumors, and the
first molecules are currently being evaluated in the clinic for various lymphoma indications. A
ACS Paragon Plus Environment

Page 5 of 48
Journal of Medicinal Chemistry
1
2
different approach to target the PRC2 activity was described by Kim et al.,20 utilizing a stapled
EZH2 peptide which is able to disrupt the EZH2-EED interface, underlining the importance of
the functional interplay between these two proteins.
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
EED is a -propeller protein comprised of seven WD40 repeats and was shown to directly bind
EZH2 by interacting with an alpha helical peptide located at the N-terminus of EZH2 (residues
38-68) via an extended groove on one side of the propeller.21 On the opposite side of EED, a
pocket referred to as the “Me3 pocket” is selectively binding histone peptides which contain a
trimethylated lysine.22,23 An aromatic cage formed by the side chains of residues F97, Y148 and
Y365 was shown to coordinate the positively charged quaternary amine of the trimethylated
lysine, and the side chain of the neighboring W364 interacts with the aliphatic side chain of the
lysine residue (Figure 1a).
Figure 1: Interaction of the H3K27Me3 peptide with EED and associated conformational
changes. (a) The X-ray co-structure of EED with H3K27Me3 peptide (PDB accession code:
3jzg) shows that the positively charged tertiary amine of K27 is located within the aromatic cage
formed by EED residues F97, Y148 and Y365, and that W364 interacts with the hydrophobic
part of the lysine side chain. EED is shown in blue whereas the H3K27Me3 peptide is colored
purple. The protein surface around the peptide ligand is represented by a grey surface. (b)
Compared to the apo EED X-ray structure (PDB accession code: 3jzn, shown in light grey),
W364 undergoes a significant rotation to accommodate the peptide. In addition, the side chain of
R367 rotates and forms a stacking interaction with Y365. Both movements are indicated by a
green arrow.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 48
1
2
3
4
5
6
7
8
9
Comparing the trimethylated lysine bound structure to the apo EED protein structure, it is found
that the Me3 pocket is dynamic in nature and can undergo local conformational changes upon
peptide binding. Specifically, the W364 side chain performs a rotation to accommodate the
peptide ligand and the side chain of R367 moves towards F365 so that the planes of the
guanidinium group and the tyrosine side chain are now oriented in parallel to form a better
stacking interaction (Figure 1b).23 It has also been shown that binding of trimethylated histone
peptides, e.g., H3K27Me3, to the Me3 pocket allosterically activates the methyltransferase
activity of PRC2.22 Since this observation demonstrated that peptide binding to the Me3 pocket
can modulate PRC2 activity, we hypothesized that a low molecular weight compound interacting
with EED might display a similar effect, acting either as an activator, or more interestingly, as an
inhibitor.
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
Herein, we report the stepwise optimization of a small molecule series with such characteristics,
i.e. interacting with the Me3 pocket of EED and allosterically inhibiting PRC2 activity,
representing a new mechanism of inhibition.
Results and discussion
A biochemical high throughput screening campaign to find new inhibitors of PRC2 activity led
to the identification of compound 1 (Figure 2a).24 Comprehensive enzymatic characterization
demonstrated inhibition of PRC2 activity with a single digit micromolar IC50 (2.5 µM), and
interestingly showed that the inhibition is non-competitive with the substrates SAM and
H3K27Me0. Instead, the molecule was found competitive with the H3K27Me3 peptide, which
clearly distinguishes the profile of this compound from previously described PRC2 inhibitors.24
Based on the known interaction of EED with H3K27Me3, we explored if the competitive profile
ACS Paragon Plus Environment

Page 7 of 48
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
of 1 can be explained by a direct interaction with EED. To this end, 1 was titrated onto EED
protein selectively 13C labeled at the terminal methyl group of methionine side chains and
spectral changes assessed by 2D HSQC NMR spectroscopy. Clear, dose-dependent chemical
shift perturbations of several methionine methyl resonances were observed when adding the
compound to EED76-441 (Figure 2b), demonstrating a specific and localized binding event.
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 2: PRC2 inhibitor 1 is binding to the trimethyllysine pocket of EED. (a) Chemical
structure, drawn in the same orientation as shown in c), and selected characteristics of the HTS
hit are shown.24 (b) Binding of 1 to isolated EED protein as detected by 2D NMR. Dose-
dependent chemical shift perturbations of a subset of methionine resonances demonstrate that 1
is directly binding to EED. The spectra are colored according to the compound concentration
added, as indicated in upper right corner. (c) The co-structure of EED with 1 (PDB accession
code: 5H17)24 shows that the compound is located in the Me3 pocket, with the methoxyphenyl
group penetrating deepest and sitting on top of the R367 side chain, and the tertiary amine of 1
being in the aromatic cage formed by F97, Y148 and Y365.
Consistent with low micromolar affinity, the spectral changes are in the slow and intermediate
exchange regime on the NMR time scale. The co-crystal structure of 1 with EED revealed that
the compound binds in the trimethyllysine pocket (Figure 1c),24 which is consistent with the
competitive nature of the inhibitor with respect to H3K27Me3 peptide.
The protein undergoes significant local conformational changes in relationship to the apo- and
peptide bound structures. This enables the compound to occupy space previously not utilized by
the peptide ligand, with the methoxyphenyl moiety penetrating into the pocket and being
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 48
1
2
3
4
5
6
7
8
9
positioned on top of R367, and the positively charged piperidine moiety of the tricyclic core of 1
located within the aromatic cage. The rest of the molecule is mainly solvent exposed and does
not seem to participate in major interactions.
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 high complexity of compound 1 and the associated synthetic challenges in exploring
SAR, our first goal was to investigate the key chemical moieties for EED activity and thereby
identify smaller as well as simpler molecules that were substructures of 1. This approach of
ligand deconstruction25,26 has the potential of identifying “minimal pharmacophores” or minimal
fragments which can be regrown more efficiently to improve potency and other properties in
ways that were not originally accessible from the parent molecule. Here, the crystal structure
enabled deconstruction of 1 in the middle of the tricyclic ring – an unusual approach as ring
systems are usually considered “cores” and are typically not broken – and pointed to a
commercially available compound 2 (Figure 3a) as the more buried and likely more relevant
half of 1. Compound 2 showed direct binding to EED in the 2D NMR assay with a KD of 32 µM
and induced chemical shift perturbations of the same resonances perturbed by 1 (Figure S1 in the
Supporting Information), as well as inhibiting PRC2 activity with an IC50 of 95 µM. Despite the
decrease in biochemical potency, ligand efficiency (LE) and lipophilic efficiency (LipE) were
increased by 0.07 from 0.29 (1) to 0.36 (2) and by 0.52 from 0.70 (1) to 1.22 (2), respectively.
The co-crystal structure of 2 with EED (data and refinement statistics are presented in Table S1
in the Supporting Information) shows that the interactions made with the protein are the same as
the ones observed for the parent compound 1 (Figure 3b) and that the binding pose is completely
retained (Figure 3c). This structural data, combined with the efficiency improvements noted
above, suggests that molecule 1 was successfully deconstructed to a minimal pharmacophore 2
that efficiently engages with the Me3 pocket even though the absolute potency decreased.
ACS Paragon Plus Environment

Page 9 of 48
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: Deconstruction of 1 yields compound 2 which retains binding mode and critical
interactions. (a) Chemical structures, drawn in the same orientation as shown in c), and selected
characteristics of 1 and 2 are shown. (b) The co-structure of EED with 2 (PDB accession code:
5U5K) shows that compound 2 is located in the same position in the Me3 pocket where the
parent compound 1 was found. The piperidine amine is positively charged and located inside the
aromatic cage, whereas the methoxyphenyl group is stacking with R367. A structured water
bridges interactions with N194 and Y365. (c) Overlaying the co-structures of EED with 1 and 2
shows that the bound state of 1 (green) and 2 (orange) are highly similar and equivalent atoms
are in almost identical positions.
The methoxyphenyl moiety is deeply buried inside the protein and is forming a pi-positive
charge stacking interaction with the guanidinium group of R367. We refer to this part of the
binding site as the “deep pocket”. The side chain of R367 is involved in a similar interaction with
Y365 in the apo-structure of the protein, and an outward movement of Y365 provides the
required space to accommodate the compound (Figure 4a).
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 48
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 4: Conformational changes observed upon binding of compound 2. Colors are
corresponding to Figure 1 and Figure 2. (a) Overlay of the apo EED X-ray structure (PDB
accession code:3jzn) with the X-ray co-structure of EED with compound 2 (PDB accession code:
5U5K). Y365 swings out which results in the formation of the “deep pocket” occupied by the
methoxyphenyl moiety of compound 2, whereas R367 rotates towards the compound to form a
pi-positive charge stacking interaction. Both movements are indicated by green arrows. (b)
Overlay of the X-ray co-structure of EED with H3K27Me3 peptide (PDB accession code: 3jzg)
with the X-ray co-structure of EED with compound 2 (PDB accession code: 5U5K). W364
rotates out to a location which is occupied by the peptide backbone in the peptide complex. The
large movement of Y365 results in a relocation of the aromatic cage, so that the positively
charged piperidine moiety of 2 is shifted compared to the tertiary amine of lysine 27 of the
H3K27Me3 peptide. Both movements are indicated by green arrows.
The ligand moiety in the deep pocket is entirely surrounded by protein, with some space
available in the ortho-position next to the meta-methoxy group, but not in the para-position. The
piperidine ring is located in the aromatic cage formed by Y365, Y148 and F97. Interestingly, the
charge-bearing nitrogen in the piperidine ring is positioned in the same location as the quaternary
amine of the peptide ligand methylated lysine. This suggests that the small molecule ligand is
recapitulating the cation-pi interaction of aromatic cages (Figure 4b). The two tyrosine side
chains are oriented parallel to the piperidine ring, whereas the phenylalanine is oriented
perpendicular. As seen for 1, the methylene linker between the methoxyphenyl group and the
ACS Paragon Plus Environment

Page 11 of 48
Journal of Medicinal Chemistry
1
2
3
4
5
6
piperidine in compound 2 provides a turn to the molecule and enables the placement of the two
moieties in the respective subpockets (Figure 3b).
7
8
9
Consistent with of our overall drug discovery efforts on EED,27 we pursued a stepwise strategy
in modifying the compound starting in the deep pocket and then moving towards the aromatic
cage with the goal of simultaneously improving potency, LE and LipE values.
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
Potency enhancement through modifications of the moiety in the deep pocket
Based on the observation that the methoxyphenyl group is stacking with the guanidinium group
of the R367 side chain and considering the associated attractive forces between the electron-rich
aromatic system and the delocalized positive charge, we initially investigated the effect of the
methoxy group. Removing the methoxy group resulted in a compound without detectable
potency (3, Table 1), while substituting the ortho-position (4) led to similar potency. Moving the
methoxy group to the para position was accompanied by significant potency loss (5, Table 1).
These results are consistent with structural analysis of available space obtained from the co-
crystal structure, as described above.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 48
1
2
3
4
Table 1: Molecules of the amine series: Initial optimization of the moiety in the deep pocket
5
6
7
8
9
Compound
2
R1
IC50 [µM]
LE1
cLogP
LipE1
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
95
0.36
2.8
1.22
> 500
146
> 500
119
95
< 0.35
0.35
2.9
2.8
2.8
2.9
3.4
3.4
2.8
< 0.40
1.04
3
4
5
6
7
8
9
< 0.30
0.33
< 0.50
1.02
0.34
0.62
40
0.37
1.00
35
0.38
1.66
1LE and LipE values are calculated according to LE=1.36*pIC50/HAC and LipE=pIC50-cLogP, respectively.
Introduction of fluorine next to the meta-methoxy group at the para-position (6) did not
significantly alter the potency, indicating that fluorine is small enough to be tolerated in this
location. While we hypothesized that the fluorine might be able direct the neighboring methoxy
ACS Paragon Plus Environment

Page 13 of 48
Journal of Medicinal Chemistry
1
2
3
4
5
6
group into an orientation consistent with the conformation observed in the crystal structure of 2,
the lack of improvement might be a result of competing electronic effects.
7
8
9
Following the previous observations, we next explored the possibility of filling the observed
space by introducing a second ring fused to the phenyl ring at the ortho- and meta-position, while
keeping oxygen containing groups. Introduction of benzofuran or benzodioxole moieties, linking
the substituents of the methoxy compounds in the ortho- and meta-position, demonstrated that a
moderate but consistent two-fold increase can be achieved (7, 8, and 9). The two benzofuran
regioisomers (7, 8) behave consistent with the methoxyphenyl compounds (3, 4) with the meta-
O-substitution slightly preferred. The benzodioxole (9) displays potency comparable to 8 (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
Modifications of the moiety localized within the aromatic cage
The secondary amine of the piperidine ring in 2 is positively charged (with a measured pKa of
9.9) and positioned in the aromatic cage, similar to the trimethylated lysine side chain in the
EED-H3K27Me3 co-crystal structure. Additionally, the amine participates in a hydrogen bond
network with a structured water and polar groups of the side chains of surrounding amino acids
(Figure 3b). Thus, the interactions within this subpocket are expected to be significant
contributors to potency and could provide a great potential for optimization. Prospective
optimization of cation-pi interactions and pi-pi interactions is challenging using conventional
computational methods employed for compound assessment, such as docking with scoring
functions and structure-based affinity predictions. Therefore, we decided to apply an
experimental approach and synthesized molecules with moieties sampling different possible
concepts of interaction with the cage, while retaining the moieties in deep pocket. We evaluated
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 48
1
2
3
4
three concepts: a point charge, a delocalized charge combined with an aliphatic pi system, and a
5
6
delocalized charge combined with an aromatic system (Figure 5).
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 5: Design concepts for different moieties located in the aromatic cage. The side chain of
F97, Y184 and Y365, forming the aromatic cage of EED, are shown. An example chemical
structure of a moiety representing the respective concept is shown inside the cage with the area
of the blue color representing the localization of the charge.
Concept 1 was explored by the HTS hit and its deconstructed analog 2. Our design efforts
towards concept 2, combining a delocalized charge with a pi stacking interaction, yielded
amidines and guanidines with different substitution patterns and ring closures. One example is
the guanidinium containing compound 10 (Table 2), which has a measured pKa of >11 and
shows a two-fold improvement in IC50 compared to the parent compound 2, despite an
anticipated dehydration penalty. Next, we combined the new guanidinium containing moiety in
the aromatic cage with moieties previously characterized in the deep pocket (Table 2). A
combination with the benzofuran group resulted in compounds 11 and 12, which had single digit
micromolar potency in the biochemical assay, showing that improvements arising from changes
in the two sub-pockets were not simply additive in nature. To evaluate if the improvement
observed going from the piperidine to the guanidinium was due to the charge or potential pi-pi
interactions, we tested the corresponding urea analog (13) and found that the potency dropped by
ACS Paragon Plus Environment

Page 15 of 48
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
approximately 80 fold, demonstrating the importance of the positive charge in the aromatic cage
in the context of this series. An improvement to single digit micromolar potency was also
observed for the previously identified benzodioxole analog (14, Table 2, Figure 6a). In addition
to the enhancements in potency, we also observed improvement in efficiency parameters, in
particular LipE as a result of the guanidinium group (e.g., a LipE of 3.51 of compound 14
compared to 1.66 of compound 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
Table 2: Molecules of the guanidinium series
G-401 ELISA
IC50 [µM]
Compound
IC50 [µM]
LE1
cLogP
LipE1
43
0.33
1.9
2.47
n.d.
10
2.8
2.1
174
3.9
0.40
0.41
0.27
0.39
2.5
2.5
1.7
1.9
3.05
3.18
2.06
3.51
2.5
2.5
n.d.
1.6
11
12
13
14
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 48
1
2
3
4
5
1LE and LipE values are calculated according to LE=1.36*pIC50/HAC and LipE=pIC50-cLogP, respectively.
A co-structure of EED and 14 (Figure 6b, Table S1 in the Supporting Information), combining
6
7
8
9
improvements in the two sub-pockets, shows that the benzodioxole moiety is placed on the R367
side chain and that the position of the phenyl ring is completely retained with respect to 2
(Figure 6c).
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 6: Co-structure of 14 with EED (PDB accession code: 5U5T). (a) Chemical structure and
selected characteristics of 14 are shown. (b) The guanidinium group of 14 interacts directly with
the side chains of N194 and E238 and via a structured water with E238. (c) The overlay of the
co-structures of EED with 2 (orange) and 14 (magenta) shows that the compounds adopt a highly
similar binding mode.
The charged guanidinium group is replacing a water molecule and is positioned to interact
directly with the side chains of N194 and E238. In addition, a water molecule bridges an
interaction to the second oxygen of E238.
Having reached the single digit micromolar potency level in the functional assay with
compounds 11, 12, and 14, we decided to test these compounds for activity in a cell-based assay
using a proximal marker. The respective assay reads out the cellular methylation status of K27 of
histone 3 in G-401 cells, which are rhabdoid tumor cells of kidney origin known to be dependent
on PRC2 and therefore sensitive to PRC2 inhibitors.15,28 The detection of H3K27Me3 is
ACS Paragon Plus Environment

Page 17 of 48
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
performed by ELISA, and we refer to the assay as the G-401 ELISA assay. We found that the
compounds were able to inhibit H3K27 methylation in cells with an IC50 of approximately 2 µM
(Table 2), demonstrating that compounds inhibiting PRC2 through binding to EED result in the
same cellular readout as previously described for EZH2 inhibitors. Interestingly, the IC50 values
observed in the biochemical and cellular assays are comparable for several of these compounds,
despite limited permeability (see below), which could be a result of their allosteric mechanism of
inhibition.
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
Property improvement
While we achieved cellular activity with molecules of the guanidinium series, the guanidinium
group is generally not preferred because the strong positive charge often negatively impacts cell
permeability. Indeed, when comparing Caco-2 data of representative examples of the piperidine
series (8, pKa of 9.8) and the guanidinium series (12, where the very close analog 14 has a pKa
of >11), we observe a 15-fold drop in permeability from 34.3 to 2.3 10-6 cm/s (Table S2 in the
Supporting Information). Therefore, we next explored design concept 3 which retained a positive
charge – now delocalized in an aromatic pi system (Figure 5) – while lowering the pKa value.
Such chemical moieties were more challenging to identify by informatics, because pKa
predictions for such less frequently used systems may not be reliable and databases of pKa
measurements do not typically annotate where the ionization center is located. Searching of the
BioLoom database29 for measured pKa values of 8 or greater, combined with substructure
searching for aromatic atoms and ascending sorting by molecular weight, identified top scoring
entries where a basic pKa was typically in the context of an aromatic ring, including 4-
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 48
1
2
3
4
5
6
aminopyridine (average pKa of 9.1) and 2-aminoimidazole (pKa of 9.2 with 4,5-dimethyl
substitution).
7
8
9
We decided to explore the 2-aminoimidazole moiety (15, 16, Table 3), as it might allow
simultaneously improving the physico-chemical properties of the series by mitigating the high
pKa, decreasing the number of H-bond donors and rotatable bonds, as well as PSA, while
keeping all observed interactions. Compound 16 has a measured pKa of 9.2 – significantly lower
than the corresponding guanidinium (14) with pKa of >11 – showing that it retains the positive
charge and is therefore a suitable moiety. Accordingly, molecules of the aminoimidazole series
show good permeability, as exemplified by compound 15 (16.8 compared to 2.32 10-6 cm/s for
compound 12, and 25.6 10-6 cm/s for compound 16, Table S2 in the Supporting Information). In
addition, the IC50 of these aminoimidazole analogs approaches the 1 µM mark and they showed
cellular activity (Table 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
Table 3: Molecules of the aminoimidazole series.
G-401 ELISA
IC50 [µM]
Compound
15
R1
R2
IC50 [µM]
LE1
cLogP
LipE1
H
2.7
0.38
3.0
2.57
12.4
H
1.3
0.40
2.5
3.39
1.0
16
ACS Paragon Plus Environment

Page 19 of 48
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
isopropyl
isopropyl
isopropyl
0.43
7.9
0.38
0.33
0.35
3.6
3.8
3.9
2.77
1.30
1.99
2.0
n.d.
2.9
17
18
19
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
1.3
1LE and LipE values are calculated according to LE=1.36*pIC50/HAC and LipE=pIC50-cLogP, respectively.
The co-structure of 16 (Table S1 in the Supporting Information) confirmed that the
aminoimidazole group is replacing the guanidinium group and that the amine is engaged in
water-mediated interactions with N194 and E238 (Figure 7b). Interestingly, direct binding to
N194 and E238 was lost and previously observed water-mediated interactions were regained.
Figure 7: Co-structures of EED binders with improved properties (PDB accession codes: 5U62
and 5U5H). (a) Chemical structure and selected characteristics of 16 are shown. (b) The 2-
aminoimidazole group in 16 acts as an effective replacement for the guanidinium group in 14
with improved properties. (c) Chemical structure and selected characteristics of 19 are shown.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 48
1
2
3
4
5
(d) The binding mode of the ortho-fluoro meta-methoxyphenyl moiety of 19 is similar to the one
observed for the benzodioxole group. The fluorine points towards the side chain amide of N194.
6
7
8
9
To evaluate growth opportunities on the imidazole ring, we introduced aliphatic groups (e.g.,
isopropyl, 17) at the 4-position, achieving sub-µM activity in the functional assay (Table 3) and
an IC50 of 2 µM in the cell-based G-401 ELISA assay.
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
Encouraged by this result, we tested compound 17 also in a cell proliferation assay monitoring
the effect of the compound on the growth of Pfeiffer cells, which are B lymphocytes and EZH2
mutant, making them sensitive to PRC2 inhibitors.11,12 Using 17, we observed a dose-dependent
anti-proliferative effect with an EC50 of 3.4 µM, demonstrating that compounds targeting the
EED Me3 pocket not only lead to a reduction of the cellular H3K27 methylation level, but also
have a direct effect on cell growth, as previously seen for EZH2 inhibitors.
In an additional optimization step, we re-investigated the moiety in the deep pocket in the
presence of the aminoimidazole ring system to mitigate potential liabilities presented by the
aromatic system with high electron density. Based on the available structural information, we
introduced fluorine at the ortho-position, which could make a possible interaction with
neighboring backbone or side chain amide groups (e.g., N194),30 and prepared corresponding
analogs (18, 19, Table 3). Compound 18, which contains only an ortho-substituted fluorine on
the phenyl ring, retains single digit micromolar affinity, demonstrating approximately 100-fold
increase in potency compared to the early phenyl/piperidine-derivative 3. We generally observed
a two-fold boost in potency when introducing the ortho-fluorine. The ortho-fluoro meta-methoxy
substituted compound 19 (Figure 7c) displays biochemical potency at the 1 µM level, an IC50 of
2.9 µM in the G-401 ELISA assay and an EC50 in the cell proliferation assay of 4.4 µM.
ACS Paragon Plus Environment

Page 21 of 48
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
In order to assess if these more advanced compounds of the series still act by the same
mechanism observed for the original screening hit,24 we performed equivalent enzymatic studies
with compounds 16 and 19 (Figure S2 in the Supporting Information). These experiments
showed a non-competitive profile with SAM and H3K27Me0, but a competitive profile with
H3K27Me3, confirming the expected mechanism of inhibition.
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
Solving the X-ray co-structure of EED with 19 (Figure 7d, Table S1 in the Supporting
Information) showed that the location of the methoxyphenyl group is retained and that the
fluorine atom is pointing towards N194, indeed making a positive interaction with the side chain
amide as designed, with a distance of 3.6 Å to the amide carbon atom. While we haven’t
explored the 4-position of the aminoimidazole moiety further thus far, the isopropyl group is
pointing towards the solvent, indicating that this vector is available for further optimization in
the future.
Conclusions
Although inhibition of PRC2 activity was shown to result in a decrease of H3K27 methylation
levels as well as inhibition of cell proliferation and in vivo efficacy in lymphoma mouse models,
so far EZH2-directed molecules are the only class of small molecule inhibitors described in the
literature. While these molecules are currently being evaluated in the clinic, recent studies report
target mutations in WT or mutant EZH2 to occur after prolonged dosing, resulting in resistance
to EZH2 inhibitors.31,32 Because all currently known inhibitors act via the same mechanism, this
raises the potential that such and similar mutations could affect their ultimate efficacy.
Therefore, inhibitors acting by a different mode of inhibition are highly valuable, due to the
different resistance profile and potential synergistic effect with known inhibitors.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 48
1
2
3
4
5
6
7
8
9
In this study, we have described a small molecule series binding to the Me3 pocket of the PRC2
subunit EED and leading to inhibition of the methyltransferase activity. This mode of inhibition
is allosteric in nature, in contrast to the orthosteric, SAM-competitive EZH2 inhibitors.
Interestingly, binding of the natural H3K27Me3 peptide ligand to the Me3 pocket allosterically
activates PRC2 activity, and this agonistic effect was hypothesized to facilitate spreading of the
repressive mark along the chromatin.22 In contrast, the described compounds act antagonistically
through the same pocket. We believe that the inhibition is a result of two mechanisms: First,
blocking the Me3 pocket with an inhibitor prevents proper localization of PRC2 to chromatin
with existing H3K37Me3 marks. Second, interaction of the compound with the Me3 pocket is
directly inhibiting the methyltransferase activity by an allosteric mechanism. By analyzing the
available crystal structures of EED in isolation, we were not able to decipher the molecular basis
for this phenomenon. However, the recently reported crystallographic systems containing the
PRC2 core components EZH2, EED and Suz124,5 will help to explain the molecular basis of this
new mechanism.
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
Our chemistry strategy involved deconstruction of a HTS hit (1) to a substructure representing
half of the parent molecule (2). While the screening hit was a structurally complex molecule and
synthetically challenging to access, the resulting fragment featured greatly reduced complexity as
well as higher synthetic tractability and therefore represented a better starting point for chemistry
optimization. Biophysical and structural validation confirmed a retained binding mode and
fragment-based drug discovery principles confirmed success in deconstruction: for even though
potency was reduced, ligand efficiency and lipophilic efficiency increased (Figure S3 in the
Supporting Information). The less complex starting point enabled a stepwise optimization in a
way that would not have been possible from the more complex tricyclic HTS hit and yielded an
ACS Paragon Plus Environment

Page 23 of 48
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
optimized fragment with further increased ligand efficiency parameters (14, Figure S3 in the
Supporting Information). The optimized fragment allowed us to validate activity in cell-based
assays, demonstrating that a similar outcome observed for EZH2 inhibitors can be achieved with
an EED binder which provided confidence to pursue this approach and to start improving
properties. Replacing the guaninidinium group with the aminoimidazole moiety significantly
improved permeability while maintaining good efficiency and cell-based activity. We also
established a vector available for property modulation and potency improvement, and introduced
a moiety in the deep part of the pocket that is more drug-like without significant potency loss
(19). Together, these changes led to a decrease in efficiency parameters, a phenomenon often
observed when going beyond the initial optimization phase towards lead optimization, but to a
simpler, tractable and more efficient molecule when compared to the HTS hit. The resulting
molecules can be further optimized and can serve as tool compounds to study the biology of
EED-mediated inhibition of PRC2 activity.
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
Experimental Section
Complete information on material and methods can be found in the Supporting Information.
Protein expression and purification. PRC2 (containing human EZH2, Suz12, EED, RBAP48
and AEBP2 proteins) used in the biochemical assay was expressed in insect cells and purified as
described previously.15 Human EED protein (residues 76–441, Uniprot ID O75530) was
produced in E. coli. BL-21 (DE3) CodonPlus cells which were transformed with a pDEST14
(Life Technologies) plasmid containing the His6-TEV-EED76-441 construct. Unlabeled protein
was expressed in cells grown in LB medium, whereas selectively methionine-13C labeled
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 48
1
2
3
4
5
6
7
8
9
protein for NMR studies was produced in minimal media supplemented with 250 mg/L L-
methionine-(methyl-13C) and various amounts of the other 19 amino acids. Proteins were
purified by IMAC using a Ni column followed by size exclusion chromatography.
NMR binding assay. 2D [13C, 1H]-HSQC experiments were recorded on a Bruker NMR
spectrometer at 600 MHz 1H Larmor frequency equipped with a 5 mm QCI cryogenic probe at
296 K. Samples contained 40 µM selectively methionine-13C-methyl labeled EED76-441 protein
in 25 mM d-Tris pH 8, 150 mM sodium chloride, 2 mM d-DTT, and 10% D2O. KD values were
determined by titrating EED protein with increasing amounts of ligand and following the
resulting chemical shift perturbations.
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
X-ray crystallography. The crystal co-structures of EED76-441 with compounds 2 and 19 were
determined using crystals grown in a precipitant composed of 3–3.5 M sodium formate22,23 using
the vapor diffusion method. Crystals grew in 3–5 days at 18 ºC and harvested and soaked in
defined drops consisting of 30 µL of precipitant with 1–2 mM of compound (typically
solubilized in DMSO) for 24–48 h. Crystals were cryopreserved for data collection using a
cryosolution consisting of 3.5 M sodium formate, 1–2 mM of compound, and 30% (v/v)
glycerol.
The crystal co-structures of EED with compounds 14 and 16 were determined by using a
PEG/salt combination as a precipitant. Briefly, EED was incubated with 10 mM -nicotinamide
adenine dinucleotide hydrate, 2 mM of a tightly binding proprietary compound, and 0.5 mM of a
synthesized peptide which comprises the helix on EZH2 which interacts with EED.21 The
crystals were grown using the vapor diffusion method. 1 µL of the protein mixture was
combined with 1 µL of a precipitant comprised of 20% (w/v) PEG3350, 0.2 M potassium iodide,
ACS Paragon Plus Environment

Page 25 of 48
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
and 0.1M Tris-HCl, pH 8.5, on a cover slip which was suspended over a reservoir comprised of
0.5 mL of precipitant at 18 ºC and sealed. The crystals grew in 4–6 days at 18 ºC and then
harvested and soaked in defined drops consisting of 30 µL of precipitant and 2 mM of compound
for 24 h. Crystals were cryopreserved for data collection using a cryosolution consisting of 30%
PEG 400 (v/v), 20% (w/v) PEG 3350, 0.2 M potassium iodide, and 0.1M Tris-HCl, pH 8.5.
Structures were solved by molecular replacement using the structure of a previously determined
apo EED structure with or without the EZH2 peptide as warranted. The coordinates were refined
to convergence using COOT,33 BUSTER,34and the PHENIX35 suite of programs. Data collection
and refinement statistics for all structures are given in Table S1 in the Supporting Information.
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
Biochemical assay. Inhibition of PRC2 activity was measured by quantification of the co-
product SAH by LC-MS, as described.15,36 Briefly, PRC2 (10 nM) was incubated with co-factor
SAM (1 µM, at Km) and substrate histone H3[21–44, K27Me0] peptide (1.5 µM, at Km) as well
as test compound at various concentrations (with a top concentration of either 50 mM or 10 mM
and three-fold serial dilutions to a total of 12 points) in assay buffer (20 mM Tris-HCl, pH 8.0,
0.01% Triton X-100, 0.5 mM DTT and 0.1% BSA). After incubation of the reaction for 2 h at
r.t., the reactions were stopped by adding 2.5% TFA and 320 nM SAH-d4. The amount of SAH
produced from the reactions was measured using an API4000 liquid chromatography/tandem
mass spectrometry (LC/MS/MS) system. SAH-d4 was used as an internal standard (IS) for SAH
detection and normalization. The plot of SAH peak area/IS peak area versus SAH concentration
was used to generate the normalization factor of SAH. The production of SAH from the
enzymatic reaction was derived from the standard curve of SAH. The detection limit of our
system for SAH is approximately 1 to 2 nM, and the range is linear up to 400 nM. All
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 48
1
2
3
4
5
6
7
8
9
measurements were run in duplicates and IC50 values were highly reproducible. The reported
values are qualified IC50 values and not absolute IC50 numbers, since 50% inhibition was not
always achieved with the test compounds in the early phase of the optimization. In addition, we
made a general observation that the maximum achieved inhibition varies and doesn’t consistently
reach 100%, which could be a result of the allosteric mechanism.
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
Cell-based assays. Cells were maintained in a humidified incubator at 37 °C, 5% CO2 (v/v).
The effect of the compounds on the cellular H3K27 methylation status was assessed in G-401
cells (ATCC CRL-1441). Cells were seeded in 384 well plates and cultured for 24 h before
treatment with serial dilution of compounds. A DMSO control containing the same volume
percentage of DMSO was also included. After incubation for 48 h, cells were washed with
phosphate based buffer (PBS, pH 7.4) twice and lysed by adding lysis buffer (0.4 N HCl; 45 µl
per well). After gentle agitation for 30 min at 4 °C, neutralization buffer (0.5 M sodium
phosphate dibasic, pH 12.5, Protease Inhibitor Cocktail, 1 mM DTT; 36 µl per well) was added
to the wells. The plate was agitated again, followed by transfer of a 10–15 µL aliquot of the cell
lysates to a 384 well ELISA plate. After overnight incubation at 4 °C, the wells were washed
with TBST 5 times, and TBST was completely removed after each wash using a paper towel.
After blocking with blocking buffer (TBST with 5% BSA) for 1 h at r.t., the primary antibody
was added in blocking buffer, and the plate was incubated for 1 h at r.t. After washing again,
ECL substrate was added and the plates were centrifuged at 2,000 rpm for 2 min. The signal was
read using a PerkinElmer Envision Reader. Percentage inhibition was calculated against the
DMSO control after normalization of the H3K27Me3 signal to the H3 signal for individual
ACS Paragon Plus Environment

Page 27 of 48
Journal of Medicinal Chemistry
1
2
3
4
5
6
samples. The data were then fit to a dose response curve to get the IC50 values of the test
compounds.
7
8
9
Compound-dependent inhibition of cell proliferation was assessed in exponentially growing
Pfeiffer cells (ATCC CRL-2632) in 384 well plates. After 6 days of compound incubation,
CellTiter-Glo reagent (CTG reagent, Promega G7573) was used for cell growth measurement.
The luciferase signal was read using a PerkinElmer Envision Reader, the percentage inhibition
was calculated against a DMSO control and IC50 values determined by fitting the inhibition
curves.
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
Determination of pKa values. Calculated pKa values were determined with the program MoKa
(Molecular Discovery). Experimental pKa values were determined by the capillary
electrophoresis method and potentiometric method, as described previously.37
Caco-2 assay. Caco-2 permeablity measurements were performed on 21 day Caco2 cell cultures.
Plates were washed of cell media by aspirating the full volume (75 µL apical, 250 µL
basolateral) and replacing it with warmed Caco2 Transport Buffer three times. The compound
source plate was prepared in a 2 mL deep well block. 4 µL of each compound (at 4 mM in
DMSO) was added to each well, followed by the addition of 1.6 mL of transport buffer to each
well resulting in a compound concentration of 10 µM. After aspirating the full volume of buffer
in the apical transwell, compounds are dosed in triplicate in the forward (A-B) and reverse (B-A)
directions. For time point 0 (t0), 10 µL is aspirated from each donor well and added to 30 µL of
transport buffer in the analytical plate The dosed transwell plates were then incubated at 37 ºC
with humidity and shaking at 180–240 rpm for 2 h. After incubation, 40 µL aliquots from
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 48
1
2
3
4
5
6
7
8
9
receiver wells and 10 µL aliquots from apical wells (for recovery calculation) were collected
from all wells of the transwell plate. Next, 40 µL of mobile phase (80% water, 20% methanol,
0.3 ng/mL glubiride (internal standard)) was added to all wells of the analytical plate. Each well
of the analytical plate was analyzed by LC/MS/MS (Waters UPLC system combined with an
ABI 4000 Qtrap Triple Quad Mass Spectrometer) running in MRM mode for compound and 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
internal standard. The apparent flux (Papp) for each well is calculated as follows: Papp
=
(dQ/dt)*(1/C0)*(1/A) where dQ/dt is the transport rate, C0 is the compound concentration at t0,
and A is the monolayer surface area. The efflux ratio is calculated by dividing the Papp B-A by
the Papp A-B.
Chemistry. Complete experimental methods (synthesis and characterization) for each target
compound and intermediates are in the Supporting Information. Purification of the final
compounds was carried out either using prepacked silica gel cartridges (Analogix, Biotage or
1
ISCO) or a reverse phase C18 column. H and 13C NMR spectra were recorded on Bruker
spectrometers. NMR chemical shifts (δ) are quoted in parts per million [ppm] and referenced to
the residual solvent peak. The purity of all compounds was determined to be >95% by analytical
HPLC.
1
3-(3-methoxybenzyl)piperidine (2). Purchased from BioBlocks, #BP039-1. Yellow Powder. H
NMR (400 MHz, DMSO-d6): δ = 8.90 (d, J = 110.5 Hz, 2H), 7.20 (t, J = 7.8 Hz, 1H), 6.74 (m, J
= 7.6, 6.7 Hz, 3H), 3.71 (s, 3H), 3.32 (s, 2H), 3.14 (d, J = 11.5 Hz, 1H), 3.00 (d, J = 11.3 Hz,
1H), 2.81–2.59 (m, 1H), 1.96 (s, 1H), 1.82–1.46 (m, 4H), 1.15 (q, J = 11.7, 11.1 Hz, 1H). 13C
NMR (100 MHz, DMSO-d6): δ = 159.7, 140.9, 129.8, 121.6, 115.1, 111.9, 55.4, 47.9, 43.7, 35.0,
28.6, 22.0. LC-MS (m/z): 206.0 [M+H]+.
ACS Paragon Plus Environment

Page 29 of 48
Journal of Medicinal Chemistry
1
2
3
4
5
6
3-benzylpiperidine (3), 3-(2-methoxybenzyl)piperidine (4), and 3-(4-methoxybenzyl)
piperidine (5) are commercially available.
7
8
9
3-(4-Fluoro-3-methoxybenzyl)piperidine (6). A solution of tert-butyl 3-(4-fluoro-3-
methoxybenzyl)piperidine-1-carboxylate (compound 6-1 in the Supporting Information , 30 mg,
0.093 mmol) in 4M HCl/dioxane (0.5 mL) was stirred at r.t. for 3 h. After removing the solvent
under reduced pressure, the residue was purified with HPLC to yield compound 6 (13.7 mg, 42%
yield). 1H NMR (400 MHz, MeOD-d4): δ = 7.05–7.00 (m, 2H), 6.79 (br. s., 1H), 3.85 (s, 3H),
3.26–3.23 (m, 1H), 2.89 (br. s., 1H), 2.77–2.54 (m, 3H), 2.05 (br. s., 1H), 1.97–1.77 (m, 2H),
1.70–1.67 (m, 1H), 1.42–1.22 (m, 1H). LC-MS (m/z): 224.1 [M+H]+.
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
3-(Benzofuran-7-ylmethyl)piperidine (7). HCl/dioxane (4M, 0.2 mL) was added to a solution
of tert-butyl 3-(benzofuran-7-ylmethyl)piperidine-1-carboxylate (compound 7-1 in the
Supporting Information, 21 mg, 0.067 mmol) in DCM. The mixture was stirred at r.t. for 3 h.
The mixture was evaporated to dryness to give compound 7 as its HCl salt (18.2 mg,
quantitative). 1H NMR (400 MHz, MeOD-d4): δ = 7.85–7.72 (m, 1H), 7.51 (d, J = 7.53 Hz, 1H),
7.25–7.16 (m, 1H), 7.16–7.09 (m, 1H), 6.90–6.78 (m, 1H), 3.23 (d, J = 11.80 Hz, 1H), 3.05–2.84
(m, 3H), 2.76 (t, J = 12.30 Hz, 1H), 2.36–2.20 (m, 1H), 1.98–1.82 (m, 2H), 1.80–1.64 (m, 1H),
1.45–1.25 (m, 1H). LC-MS (m/z): 216.2 [M+H]+.
3-(Benzofuran-4-ylmethyl)piperidine (8). To a solution of tert-butyl 3-(benzofuran-4-
ylmethyl)piperidine-1-carboxylate (compound 8-1 in the Supporting Information, 57 mg,
0.181mmol) in DCM was added HCl/dioxane (4M, 0.2 mL). The mixture was stirred at r.t. for 3
h. The mixture was concentrated under reduced pressure to give compound 8 as its HCl salt (47.8
mg, quantitative). 1H NMR (400 MHz, MeOD-d4): δ = 7.85–7.70 (m, 1H), 7.38 (d, J = 8.28 Hz,
1H), 7.24 (t, J = 7.78 Hz, 1H), 7.06 (d, J = 7.28 Hz, 1H), 6.98 (s, 1H), 3.27–3.37 (m, 1H), 3.24
ACS Paragon Plus Environment