Rational Design of Selective Small-Molecule Inhibitors for β-Catenin/B-Cell Lymphoma 9 Protein-Protein Interactions

Article abstract of DOI:10.1021/jacs.5b04988

Selective inhibition of α-helix-mediated protein-protein interactions (PPIs) with small organic molecules provides great potential for the discovery of chemical probes and therapeutic agents. Protein Data Bank data mining using the HippDB database indicated that (1) the side chains of hydrophobic projecting hot spots at positions i, i + 3, and i + 7 of an α-helix had few orientations when interacting with the second protein and (2) the hot spot pockets of PPI complexes had different sizes, shapes, and chemical groups when interacting with the same hydrophobic projecting hot spots of α-helix. On the basis of these observations, a small organic molecule, 4′-fluoro-N-phenyl-[1,1′-biphenyl]-3-carboxamide, was designed as a generic scaffold that itself directly mimics the binding mode of the side chains of hydrophobic projecting hot spots at positions i, i + 3, and i + 7 of an α-helix. Convenient decoration of this generic scaffold led to the selective disruption of α-helix-mediated PPIs. A series of small-molecule inhibitors selective for β-catenin/B-cell lymphoma 9 (BCL9) over β-catenin/cadherin PPIs was designed and synthesized. The binding mode of new inhibitors was characterized by site-directed mutagenesis and structure-activity relationship studies. This new class of inhibitors can selectively disrupt β-catenin/BCL9 over β-catenin/cadherin PPIs, suppress the transactivation of canonical Wnt signaling, downregulate the expression of Wnt target genes, and inhibit the growth of Wnt/β-catenin-dependent cancer cells.

Full text of DOI:10.1021/jacs.5b04988

Subscriber access provided by University of South Dakota
Article
Rational Design of Selective Small-Molecule Inhibitors for #-
Catenin/B-Cell Lymphoma 9 Protein–Protein Interactions
Logan R. Hoggard, Yongqiang Zhang, Min Zhang, Vanja Panic, John A. Wisniewski, and Haitao Ji
J. Am. Chem. Soc., Just Accepted Manuscript • DOI: 10.1021/jacs.5b04988 • Publication Date (Web): 09 Sep 2015
Downloaded from http://pubs.acs.org on September 9, 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 the American Chemical Society 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 45
Journal of the American Chemical Society
1
2
3
4
Rational Design of Selective Small-Molecule Inhibitors for ꢀ-Catenin/B-Cell Lymphoma 9
5
6
7
Protein–Protein Interactions
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
Logan R. Hoggard,† Yongqiang Zhang,† Min Zhang, Vanja Panic,
John A. Wisniewski, Haitao Ji*
Department of Chemistry, Center for Cell and Genome Science, University of Utah, Salt Lake
City, Utah 84112–0850
† These two authors contribute equally to this work.
* To whom correspondence should be addressed. Email: markji@chem.utah.edu
Keywords: Protein–protein interactions, Small-molecule inhibitors, Selectivity, Wnt signaling,
ꢀ-Catenin, B-cell lymphoma 9
1
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 2 of 45
1
2
3
4
Abstract
5
6
7
8
Selective inhibition of ꢀ-helix-mediated protein–protein interactions (PPIs) with small
organic molecules provides a great potential for the discovery of chemical probes and therapeutic
agents. A PDB data mining using HippDB indicated that (1) the side chains of hydrophobic
projecting hot spots at positions i, i + 3, and i + 7 of an ꢀ-helix had few orientations when
interacting with the second protein; and (2) the hot spot pockets of PPI complexes had different
sizes, shapes and chemical groups when interacting with the same hydrophobic projecting hot
spots of ꢀ-helix. Based on these observations, a small organic molecule, 4'-fluoro-N-phenyl-
[1,1'-biphenyl]-3-carboxamide, was designed as a generic scaffold that itself directly mimics the
binding mode of the side chains of hydrophobic projecting hot spots at positions i, i + 3, and i +
7 of an ꢀ-helix. Convenient decoration of this generic scaffold led to the selective disruption of
ꢀ-helix-mediated PPIs. A series of small-molecule inhibitors selective for ꢁ-catenin/BCL9 over
ꢁ-catenin/cadherin PPIs was designed and synthesized. The binding mode of new inhibitors was
characterized by site-directed mutagenesis and structure-activity relationship studies. This new
class of inhibitors can selectively disrupt ꢁ-catenin/BCL9 over ꢁ-catenin/cadherin PPIs, suppress
the transactivation of canonical Wnt signaling, downregulate the expression of Wnt target genes,
and inhibit the growth of Wnt/ꢁ-catenin-dependent cancer cells.
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
2
ACS Paragon Plus Environment

Page 3 of 45
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
Efficient targeting of protein–protein interactions (PPIs) with small organic molecules is
a significant challenge.1 Hot spots at PPI interfaces provide a starting point for designing PPI
inhibitors.2 However, how to achieve an efficient conversion of hot spot knowledge to a small-
molecule inhibitor is an unsolved task.3 For PPIs, the protruding hot spot of the ligand protein
packs against the hot spot pocket of the target protein. The hot spot pocket of the target protein
tends to undergo different induced-fit rearrangements when forming the complex structures with
different protein partners and small-molecule ligands.2,4 Structure-based design of PPI inhibitors
based only on the hot spot pocket(s) of the target protein can be challenging. We decided to
analyze the orientations and chemical properties of the projecting hot spots of the ligand protein
in the PPI structures. Small organic molecules (called fragments) were then designed to mimic
the binding mode of projecting hot spots and used as the starting points for fragment-based
inhibitor optimization. The role of ꢀ-helical peptides in mediating PPIs is well demonstrated.5
For many ꢀ-helix-mediated PPIs, the side chains of the projecting hot spots of an ꢀ-helix appear
at the same face and define the interacting surface for PPIs. We aimed to design a generic
scaffold that itself can directly mimic the binding mode of the side chains of hydrophobic
projecting hot spots at positions i, i + 3, and i + 7 of an ꢀ-helix. This generalized scaffold was
then decorated for the selective disruption of specific protein–protein interfaces. Herein, we
report one such scaffold and it's application to generate selective inhibitors for the ꢁ-catenin/B-
cell lymphoma 9 (BCL9) PPI, a key downstream effector of canonical Wnt signaling.
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
The canonical Wnt signaling pathway plays a critical role in embryonic development, cell
proliferation, and tissue homeostasis.6 ꢁ-Catenin is the primary mediator of this signaling
pathway. The hyperactivation of canonical Wnt signaling leads to an accumulation of ꢁ-catenin
in the cell nucleus. Nuclear ꢁ-catenin forms a supercomplex with T-cell factor (Tcf)/lymphoid
3
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 4 of 45
1
2
3
4
5
6
7
8
9
enhancer-binding factor (Lef), BCL9/B9L (BCL9-like, a BCL9 paralogue), and CREB (cAMP
response element-binding protein)-binding protein (CBP)/p300, etc. to activate the transcription
of a number of ꢀ-catenin target genes including AXIN2, LGR5, cyclin D1, c-myc, and LEF1,
which further promote cancer cell growth, migration, resistance to current drugs, and evasion
from apoptosis. More importantly, canonical Wnt signaling is aberrantly overactivated in cancer
stem cells, which drives cancer growth, seeds metastases, and causes cancer recurrence after
remission. Significant efforts have been made to discover inhibitors for the canonical Wnt
signaling pathway.7 The formation of the ꢀ-catenin/BCL9 complex in the cell nucleus is the
penultimate step of this signaling pathway. The transcriptional overactivation of canonical Wnt
signaling is dependent on the formation of the ꢀ-catenin/BCL9 complex, as evident by the
upregulation of BCL9 and B9L in cancer cells8–11 and the results from BCL9 or B9L
knockdown8–13 and dominant negative BCL9 or B9L8,9 experiments. The crystallographic
analysis reveals that the homology domain 2 of BCL9 (residues S352–F374) adopts an ꢁ-helical
structure to interact with the first armadillo repeat of ꢀ-catenin.14 Stapled BCL9 L351–F374 ꢁ-
helical peptides were designed to inhibit ꢀ-catenin/BCL9 PPIs.15,16 Among them, SAH-BCL9B
can pass the cell membrane, bind to ꢀ-catenin, disrupt ꢀ-catenin/BCL9 PPIs, and selectively
suppress the transcription of Wnt target genes. This stapled peptide also inhibited cancer cell
growth, angiogenesis, and metastasis without overt damage to normal tissues in the mouse
xenograft models for colorectal carcinoma and multiple myeloma. The immunogenic effect and
the in vivo stability of this stapled ꢁ-helical peptide have not yet been reported. High-throughput
screening (HTS) identified a small organic molecule, carnosic acid, that can disrupt the ꢀ-
catenin/BCL9 PPI, inhibit ꢀ-catenin-dependent transcription, and destabilize activated ꢀ-
catenin.17 Carnosic acid is a natural antioxidant and associated with many biological activities.
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
ACS Paragon Plus Environment

Page 5 of 45
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
Its catechol substructure readily reacts with protein nucleophiles after oxidation and has been
recognized as a substructure of pan assay interference compounds (PAINS).18,19 This compound
can be problematic in inhibitor optimization and biological studies. Therefore, there is an urgent
need to develop class specific small-molecule inhibitors for the ꢀ-catenin/BCL9 PPI.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Results and Discussion
Rational design of a small-molecule scaffold to mimic the binding mode of the side chains
of hydrophobic hot spots at positions i, i + 3, and i + 7 of an ꢀ-helix.
The helical PPI database, HippDB,5,20–22 was interrogated to retrieve the protein–protein
complexes that (1) are predicted to have hydrophobic projecting hot spots at positions i, i + 3,
and i + 7, and (2) have concave pocket(s) on the second protein to accommodate projecting hot
spots. The analyzed hydrophobic residues of ꢁ-helices were leucine (L), isoleucine (I),
methionine (M), valine (V), phenylalanine (F), and tryptophan (W). Out of 27,746 PPI entries in
HippDB (on January 15, 2014), 733 PPI structures that meet these two criteria were retrieved
and further superimposed by backbone heavy atoms at positions i, i + 3, and i + 7 of ꢁ-helices to
minimize the root mean square deviations (rmsd). It was observed that the side chains of the
projecting hot spots of ꢁ-helices at the same position (i , i + 3, or i + 7) tended to be placed in the
limited spatial areas when forming the protein–protein complexes. Two main groups of the
spatial areas were identified for each of the three projecting hot spots, as shown in Figures 1A–
1C and Supporting Information Figure S1. Orientations I and II for hot spot i point away from
and towards i + 3, respectively. Orientations I and II for hot spot i + 3 point towards i and i + 7;
and orientations I and II for hot spot i + 7 point towards and away from i + 3. It was also noted
that a small set of side chains of projecting hot spots F and W were not located in the areas for
5
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 6 of 45
1
2
3
4
5
6
7
8
9
orientations I and II, and the detailed analysis will be reported separately. A small-molecule
scaffold, 4'-fluoro-N-phenyl-[1,1'-biphenyl]-3-carboxamide (compound 1, Figure 1D), was then
designed by a bioisostere-based fragment hopping protocol23,24 to mimic the positioning of
hydrophobic side chains of projecting hot spots with consideration of the sizes of hot spot
pockets of the target protein, as well as the biocompatibility of newly-generated ligand
structures.24–26 Three rotatable single bonds in 1 produce at least 16 low-energy conformations
and allow a good coverage of the conformational space as described in Note S1. The extensive
HippDB data mining indicated that 1 mimicked 86% PPIs that had such hydrophobic projecting
hot spots from ꢀ-helices and the concave pockets from the target proteins (631 out of 733
structures, Table S1). Two side chain patterns that were difficult for scaffold 1 to mimic are
orientation I (i)-orientation I (i + 3)-orientation II (i + 7) and orientation I (i)-orientation II (i +
3)-orientation II (i + 7). Eighteen protein–protein complexes of these 631 structures, which
disruption potentially have important biomedical applications, were selected as the examples.
Four examples are shown in Figures 1E–1H, and 14 other examples are described in Figure S2.
Based on the structure of 1, substituents RA–RC in Figure 1D can be introduced to trigger
inhibitor selectivity because the structural features that differentiate ꢀ-helix-mediated PPIs can
reside in the protein–protein interfaces adjacent to the hot spot pockets, as observed by a
comparison of 18 protein–protein complexes in Figures 1E–1H and Figure S2.
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
6
ACS Paragon Plus Environment

Page 7 of 45
Journal of the American Chemical Society
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 1. Design of a small-molecule scaffold that itself can directly mimic the binding mode of the side
chains of hydrophobic projecting hot spots at positions i, i + 3, and i + 7 of an ꢀ-helix. (A) Fifty-seven PPI
structures that have projecting hot spots LꢀꢀLꢀꢀꢀL are rmsd superimposed based on the backbone heavy
atoms at positions i, i + 3, and i + 7. These 57 PPI structures are listed in Table S1. (B) Twenty-five PPI
structures that have are projecting hot spots LꢀꢀIꢀꢀꢀL, as shown in Table S1, are rmsd superimposed based
on the backbone heavy atoms at positions i, i + 3, and i + 7. (C) Twenty PPI structures that have projecting
hot spots LꢀꢀXꢀꢀꢀW (X = L, I, V, and F, Table S1) are rmsd superimposed based on the backbone heavy
atoms at positions i, i + 3, and i + 7. (D) Chemical structure of compound 1, 4'-fluoro-N-phenyl-[1,1'-
biphenyl]-3-carboxamide. The distances between the centroids of each aromatic system were indicated in Å.
The rotatable single bonds are indicated red. MW, molecular weight; tPSA, topological polar surface area;
ClogP, calculated lipophilic coefficient; HBD, H-bond donor; and HBA, H-bond acceptor. (E) Compound 1
7
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 8 of 45
1
2
3
4
5
6
7
8
9
mimics the positioning of the side chains of L366, I369, and L373 of human BCL9 (chain C) and is docked
to human ꢀ-catenin (chain A). PDB id, 2GL7.14 (F) Compound 1 mimics the side chain positioning of L141,
I144, and L148 of mouse puma BH3 domain (chain C) and is docked into mouse prosurvival protein A1
(chain A). PDB id, 2VOF.27 (G) Compound 1 mimics the side chain positioning of I23, W26, and F30 of
protein N from Granada virus nucleocapsid (chain C) and is docked to protein N of Granada virus
nucleocapsid (chain D). PDB id, 4J4X.28 (H) Compound 1 mimics the side chain positioning of V361, M364,
and V368 of the calcineurin catalytic subunit (chain A) and is docked to the cacineurin regulatory subunit
(chain B). PDB id, 4F0Z.29
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
8
ACS Paragon Plus Environment

Page 9 of 45
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
Design of ꢀ-catenin/BCL9 inhibitors based on scaffold 1. Crystallographic and biochemical
analyses revealed two hot regions at the ꢀ-catenin/BCL9 interface (Figure 2A):9,14,30–33 Hot
region 1: D162, E163, and D164 of human ꢀ-catenin form an acidic knob30 and interact with
H358 and R359 of human BCL9. The D162A mutation of ꢀ-catenin reduced its binding with
BCL9.30 ꢀ-Catenin D164A abrogated the interaction with BCL9 or B9L in vitro31 and BCL9-
dependent Wnt transcription in vivo32. Mutation of H358 or R359 of BCL9 to alanine
significantly reduced its interaction with ꢀ-catenin.14 The 358HRE360/358AKQ360 mutation of
BCL9 completely disrupted its binding with ꢀ-catenin.9 Hot region 2: L366, I369, and L373 of
BCL9 interact with a hydrophobic pocket that is lined with L159, V167, L160, A171, M174,
L178, A149, A152, and L156 of ꢀ-catenin. BCL9 L366K9, L373A9, or L366A/I369A14
prevented its binding with ꢀ-catenin in the pulldown experiment. A similar result was also
observed in the cell-based study. BCL9 L366K or B9L L411K could not co-immunoprecipitate
with ꢀ-catenin.9 BCL9 L366A, I369A, or L373A had no observable inhibition of wild-type ꢀ-
catenin/wild-type BCL9 PPIs in the fluorescence polarization (FP) competitive inhibition
assay.33 The pulldown experiment showed that ꢀ-catenin L159A or L178A bound with BCL9,
however, a double mutant, ꢀ-catenin L156A/L159A, abolished its interaction with BCL9.14 Our
site-directed mutagenesis experiments demonstrated that ꢀ-catenin L156S or L159S reduced the
binding affinity with BCL9, while ꢀ-catenin D145A, E155A, and D145A/E155A had no affect
on ꢀ-catenin/BCL9 PPIs. ꢀ-Catenin double mutants, L156S/L159S and L156S/L178S, abolished
their binding with BCL9 (The results are shown in Figures S3 and S4).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
9
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 10 of 45
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 2. Design of small-molecule ꢀ-catenin/BCL9 inhibitors. (A) Hot regions 1 and 2 for the ꢀ-
catenin/BCL9 PPI. (B) An overlay of the side chains of BCL9 L366, I369, and L373 with 4'-fluoro-N-
phenyl-[1,1'-biphenyl]-3-carboxamide. (C) Chemical structures of 2–16. (D) AutoDock result of 10 with ꢀ-
catenin. A stick model of this docking result is shown in Figure S5. (E) AlphaScreen Ki values of 2–16. Each
set of data was expressed as mean ± standard deviation (n = 3). Details are shown in Figure S6.
10
ACS Paragon Plus Environment

Page 11 of 45
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
An overlay of scaffold 1 with BCL9 L366, I369, and L373 is shown in Figures 1E and
2B. To verify this design, compounds 2–16 in Figure 2C were synthesized (The synthetic routes
for 2–16 are shown in Schemes S1–S6). Compounds 2–5 were designed to mimic the binding
mode of the side chains of L366 and I369. The AlphaScreen assay34 showed that these
compounds inhibited the ꢀ-catenin/BCL9 PPI at a low mM range (Figure 2E). A 4-aminoethoxy
or 4-aminopropyl side chain was introduced to 5 with an attempt to form charge-charge and H-
bond interactions with E155. The AlphaScreen assay showed that the resulting compounds, 6
and 7, were 5-fold more potent than 5. The one-carbon-shorter derivative, 8, and the hydroxyl
derivative, 9, were less potent than 7 probably due to weakened electrostatic interactions.
Compounds 10–13 were designed to explore the pocket adjacent to ꢀ-catenin L178. The para
and meta positions of the terminal benzene ring of the biphenyl-3-amino moiety can tolerate a
fluorine substitution. Compounds 14–16 were designed to optimize the interactions with residues
L159, L160, V167, and A171. The results showed that 4-fluoro and 3,4-difluoro derivatives, 10
and 14, exhibited higher potency than 15. On the other hand, the 3-thionyl derivative, 16, had a
similar potency as 10. An AutoDock35 docking result of 10 is shown in Figure 2D.
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
Structure-activity relationship-based hit optimization. Compounds 17–29 in Figure 3A were
designed to generate more potent ꢀ-catenin/BCL9 inhibitors (The synthetic routes for 17–29 are
shown in Schemes S7–S11). A second aminoethoxy group was introduced to 10 and 14 to
produce additional charge-charge and H-bond interactions with ꢀ-catenin D145. As shown in
Figure 3C, compounds 17 and 18 were more potent than 10 and 14. Compounds 19 and 20 were
designed to explore the role of ꢀ-catenin D145 and E155 in inhibitor binding. Compound 19 was
predicted to have H-bonds but no charge-charge interactions with D145 and E155, while 20 lost
both H-bond and charge-charge interactions. The AlphaScreen assay showed that 19 and 20 were
11
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 12 of 45
1
2
3
4
5
6
7
8
9
6- and 65-fold less potent than 18, respectively. The (S)-pyrrolidin-3-yloxy group of 21 and 22
was designed to replace the aminoethoxy group of 17 and 18 with the attempt to increase the
cellular permeability. These two compounds were 2- to 3-fold more potent than 17 and 18,
respectively, likely due to the conformational constraint of the five-membered ring. Compound
22 exhibited a Ki value of 2.1 ± 0.41 µM and was more potent than carnosic acid in the parallel
assay. Both AutoDock and Glide36 docking studies predicted the same binding mode for 22, as
shown in Figure 3B. The substitution of the pyrrolidine ring of 22 with an aziridine ring led to a
decrease in inhibitory activity. Derivative 24 has the 3,4-difluoro substituent on both terminal
benzene rings, exhibiting slightly lower potency than 21. The replacement of the 3,4-
difluorophenyl group of 22 with a 2-naphthyl or methyl group drastically decreases its inhibitory
potency, projecting the size of the hot spot pocket. The other three stereoisomers of 22 were also
synthesized. The AlphaScreen assay of 27–29 indicated that these three stereoisomers had the
similar inhibitory activities as 22. This result is consistent with that predicted from the AutoDock
studies because D145 and E155 are located on the surface of the ligand binding pocket and can
access to the pyrrolidine nitrogen atoms of all studied stereoisomers.
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
12
ACS Paragon Plus Environment

Page 13 of 45
Journal of the American Chemical Society
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. Optimization of new ꢀ-catenin/BCL9 inhibitors. (A) Chemical structures of 17–30. (B) AutoDock
result of 22 with ꢀ-catenin. A stick model of this docking result is shown in Figure S7. (C) AlphaScreen Ki
values of 17–30 and carnosic acid for the ꢀ-catenin/BCL9 PPI and the AlphaScreen Ki values of 18, 21–24,
27–30 and carnosic acid (CA) for the ꢀ-catenin/E-cadherin PPIs. Each set of data was expressed as mean ±
standard deviation (n = 3). Details are shown in Figures S8 and S9. (D) ITC results of 22 with wild-type and
mutant human ꢀ-catenin (residues 138–686). Details are in Figure S10. The Ki value of 22 from the
AlphaScreen assay (2.1 ± 0.41 µM) is 6.3-fold higher than the Kd value from the ITC experiment (0.33 ±
0.0010 µM). The high concentrations of ꢀ-catenin (7.5–10 µM) and small-molecule inhibitors (75–100 µM)
used in the ITC studies might also contribute to the lower ITC Kd value. (E) AlphaScreen competitive
inhibition assay results of 18, 21, and 22 for the disruption of wild-type ꢀ-catenin/wild-type BCL9 and
13
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 14 of 45
1
2
3
4
mutant ꢀ-catenin/wild-type BCL9 PPIs. Each set of data was expressed as mean ± standard deviation (n = 3).
5
6
7
Details are shown in Figure S11.
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
14
ACS Paragon Plus Environment

Page 15 of 45
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
Isothermal titration calorimetry (ITC) and site-directed mutagenesis studies. The ITC study
showed that 22 bound wild-type ꢀ-catenin but not BCL9 (Figure S10). The trend of the Kd values
of 10 and 22 from the ITC studies were consistent with that observed from the AlphaScreen
assay. The Kd value of 22 with ꢀ-catenin D145A/E155A was 9.6-fold higher than that with wild-
type ꢀ-catenin (Figure 3D), implicating the importance of the side chains of D145 and E155 for
inhibitor binding. The Kd values of 22 with ꢀ-catenin L159S and L156S/L178S were 0.92 ±
0.033 and 1.6 ± 0.045 µM, respectively, further validating the proposed binding mode. Since
neither D145A nor E155A affects the ꢀ-catenin/BCL9 PPI, as demonstrated by the AlphaScreen
competitive binding and fluorescence anisotropy binding assays (Figures S3 and S4), the
AlphaScreen competitive inhibition assays were performed to evaluate the effects of these
mutations to the Ki values of 18, 21, and 22. As shown in Figure 3E, the Ki values of 22 for ꢀ-
catenin D145A/BCL9 and ꢀ-catenin E155A/BCL9 PPIs were 13 ± 1.7 µM and 13 ± 1.8 µM,
respectively, indicating that the side chain carboxylic groups of D145 and E155 of ꢀ-catenin
were important for inhibit potency. The deletion of both side chain carboxylic groups of D145
and E155 led to a further reduction of the inhibitory activity. The Ki value of 22 was 97 ± 7.8 µM
for ꢀ-catenin D145A/E155A double mutant/BCL9 PPIs. The same trend of the Ki value changes
was also observed for 18 and 21 (Figure 3E).
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
Inhibitor selectivity studies. ꢀ-Catenin has two functions in cells. One is the interaction with
Tcf/Lef, BCL9/B9L, CBP/p300, etc. in the cell nucleus to culminate canonical Wnt signaling.
The second is the interaction with cadherin to fulfill its function for cell–cell adhesion. The
crystallographic analyses reveal that the interface of the ꢀ-catenin/BCL9 PPI was also used to
bind region V of E-cadherin.14,37 Two key residues of murine E-cadherin, F871 and L874 (PDB
id, 1I7W), occupy the same positions as human BCL9 L366 and I369 (PDB id, 2GL7) and
15
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 16 of 45
1
2
3
4
5
6
7
8
9
project to the same hot spot pocket of ꢀ-catenin (Figure S12). This overlap presents a potential
risk that ꢀ-catenin/BCL9 inhibitors might disrupt ꢀ-catenin mediated cell–cell adhesion, although
the ꢀ-catenin/E-cadherin PPI has a much larger contacting surface than the ꢀ-catenin/BCL9 PPI.
The AlphaScreen selectivity assay was used to evaluate the selectivity of 18, 21–24 and 27–29
between ꢀ-catenin/BCL9 and ꢀ-catenin/E-cadherin PPIs. As shown in Figure 3C, all tested ꢀ-
catenin/BCL9 inhibitors can selectively target ꢀ-catenin/BCL9 over ꢀ-catenin/E-cadherin PPIs.
Among them, compound 22 exhibited 125-fold selectivity and was more selective than carnosic
acid in the parallel assay. It should be noted that the Kd value for the ꢀ-catenin/BCL9 PPI (280–
710 nM)14,34 is higher than that for the ꢀ-catenin/E-cadherin PPI (41–82 nM).34,38–40 The inhibitor
selectivity can partially be owing to the weaker interaction between ꢀ-catenin and BCL9. Noting
that hot region 1 in Figure 2A is used to bind BCL9 but not E-cadherin, compound 30 was
designed and synthesized to form charge-charge and H-bond interactions with D164 (Figure S13
and Scheme S12). The AlphaScreen assay showed that this compound exhibited a comparable
inhibitory potency for ꢀ-catenin/BCL9 PPIs as 22. However, it exhibited much higher selectivity
for ꢀ-catenin/BCL9 over ꢀ-catenin/E-cadherin PPIs, underscoring the future direction for
inhibitor optimization.
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
The derived Ki values of small-molecule inhibitors are much lower for the ꢀ-
catenin/BCL9 PPI than for the ꢀ-catenin/E-cadherin PPIs. These results seemly contradict with
the definition of the Ki, for which a competitive inhibitor should have the same Ki value
regardless of the identities and concentrations of the competing ligands. We believe that this
difference in the Ki values of PPI inhibitors comes from different contacting surface areas and
hot regions between protein–protein complexes. For example, the crystallographic analysis
indicated that the surface area for the ꢀ-catenin/E-cadherin PPI was extensive and had five
16
ACS Paragon Plus Environment

Page 17 of 45
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
regions.37 Only region V is overlapped with that for the ꢀ-catenin/BCL9 PPI.14 The mechanism
of the partial inhibition by the concentrations of the inhibitors used in the biochemical assays is
likely different between ꢀ-catenin/BCL9 and ꢀ-catenin/E-cadherin PPIs, resulting in the different
Ki values. More analysis will be required to fully understand this difference. The experimental
IC50 values are also provided in Figure S9 for comparison.
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
Cellular uptake and the inhibition of the ꢀ-catenin/BCL9/Tcf transcriptional activity. To
determine whether new ꢀ-catenin/BCL9 inhibitors can be taken up by of Wnt/ꢀ-catenin
dependent cancer cells, an HPLC/MS analysis were conducted to measure the concentration of
22 in HCT 116 colon cancer cells. HCT116 cells were first treated with 2 µM of 22 for 24 h. The
intracellular 22 was then successfully separated from the other cellular components by HPLC
(Figure 4A) and characterized by MS (Figure S14). A further quantification concluded that the
concentration of 22 in HCT116 cells was 1.66 ± 0.0286 µM (83% cellular retention ratio, Figure
S15). It is noted that this cellular uptake assay relies on extraction of lager numbers of cells (1.7
× 106) and cannot determine the concentrations of the inhibitor in specific subcellular
compartments. To determine whether new ꢀ-catenin/BCL9 inhibitors can inhibit the
transactivation of canonical Wnt signaling, the Wnt-responsive luciferase reporter assays were
performed with Wnt-activated human embryonic kidney cell 293 (HEK293), colorectal cancer
cell SW480, and triple-negative breast cancer cell MDA-MB-231. As shown in Figure 4B,
compounds 10 and 21–23 can effectively inhibit the TOPFlash luciferase (luciferase reporter
with wild-type Tcf4 binding sites) activity without reducing absolute Renilla values (internal
controls) and affecting the FOPFlash luciferase (luciferase reporter with mutant Tcf4 binding
sites) activity.
17
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 18 of 45
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. Cellular uptake and the inhibition of Wnt activation by new ꢀ-catenin/BCL9 inhibitors.
(A) HPLC-UV chromatograms of the studied samples. HCT116 colon cancer cells were used.
The HPLC retention time of 22 was 15.8 min. The MS data of pure and intracellular 22 are
18
ACS Paragon Plus Environment

Page 19 of 45
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
shown in Figure S14. Details to derive the intracellular concentration of 22 are shown in Figure
S15 and described in the Methods section. (B) Physical properties and the Wnt-responsive
luciferase reporter assay results for 10, 21–23, and carnosic acid. Details are in Figure S16. Each
set of data is expressed as mean ± standard deviation (n = 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
19
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 20 of 45
1
2
3
4
5
6
7
8
9
Inhibition of the expression of Wnt/ꢀ-catenin target genes. AXIN2, LGR5, Cyclin D1, and
LEF1 are target genes for the canonical Wnt signaling pathway. AXIN2 is a specific target gene
for this signaling pathway. LGR5 is a marker for intestinal stem cells. Cyclin D1 and LEF1 are
upregulated in many cancer cells and promote tumorigenesis. Real-time quantitative PCR
experiments were performed to evaluate the effects of new ꢀ-catenin/BCL9 inhibitors on the
mRNA level of these Wnt target genes. As shown in Figure 5A, compound 22 down-regulated
the transcription of AXIN2, LGR5, LEF1, and cyclin D1 in dose-dependent manners in MDA-
MB-231 cells. More than 50% of mRNA expression was inhibited at the doses of 2 and 4 µM of
22. The dose-dependent inhibition of mRNA expression of Wnt target genes was also observed
for SW480 cells (Figure S17). The protein expression levels of cyclin D1, c-myc, the active form
of ꢀ-catenin (ABC)41, and total ꢀ-catenin in SW480 cells were examined by Western blot
analysis (Figure 5B). The protein expression levels of cyclin D1 and c-myc were significantly
reduced after the treatment of 22. Compound 22 reduced activated ꢀ-catenin in the cell nucleus
but had no effect on E-cadherin-bound ꢀ-catenin, indicating that 22 does not inhibit the upstream
sites of canonical Wnt signaling. The dose-dependent inhibition of mRNA and protein
expression was also observed for 23, as shown in Figure S18.
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
20
ACS Paragon Plus Environment

Page 21 of 45
Journal of the American Chemical Society
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 5. Further cell-based studies of new ꢀ-catenin/BCL9 inhibitors. (A) Quantitative real-time PCR study
to determine the changes of mRNA expression of AXIN2, LGR5, LEF1, and cyclic D in MDA-MB-231 cells
in response to different concentrations of 22. Each set of data is expressed as mean ± standard deviation (n =
3). (B) Western blot analysis to monitor the change of protein expression of cyclin D1, c-myc, ABC, and
total ꢀ-catenin in response to different concentrations of 22 in SW480 cells. ꢀ-Tubulin was used as an
internal reference. Experiments were performed in duplicate. (C) Co-immunoprecipitation experiments to
evaluate the selectivity of 22 for ꢀ-catenin/BCL9 over ꢀ-catenin/cadherin PPIs in HCT116 cells. IP,
immunoprecipitation; IB, immunoblotting; input, 10% amount of cell lysate. Experiments were performed in
duplicate. (D) MTs assay to monitor the effects of 21–23, and carnosic acid (CA) on the growth of SW480,
HCT116, HT29, MDA-MB-231, MDA-MB-436, A549, and MCF10A cells. Each set of data is expressed as
mean ± standard deviation (n = 3). (E) Inhibition of the anchorage-independent growth of 22 in HCT116
cells. Details are in Figure S19.
21
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 22 of 45
1
2
3
4
5
6
7
8
9
Cell-based inhibitor selectivity and the inhibition of viability for Wnt/ꢀ-catenin-dependent
cancer cells. Co-immunoprecipitation experiments were performed to evaluate inhibitor
selectivity in a cellular context. As shown in Figure 5C, compound 22 inhibited the ꢀ-
catenin/BCL9 PPI in a dose-dependent manner. A parallel experiment indicated that 22 had no
effect on the ꢀ-catenin/E-cadherin PPI at the concentrations that were sufficient to inhibit ꢀ-
catenin/BCL9 PPIs. The MTs cell viability assays were performed to assess the effect of ꢀ-
catenin/BCL9 inhibitors on the growth of colorectal cancer cells, SW480, HCT116, and HT29,
and triple-negative breast cancer cells, MDA-MB-231 and MDA-MB-436, all of which have
hyperactivated Wnt signaling (Figure 5D). The MTs assay results showed that 21–23 inhibited
cell growth in dose-dependent manners. Compounds 21 and 22 were more potent than carnosic
acid. Compounds 21 and 22 exhibited higher cytotoxicity for Wnt signaling-activated cells over
Wnt signaling-latent cells, such as lung adenocarcinoma cell line A549 and normal mammary
epithelial cell line MCF10A. As shown in Figure 5E, compound 22 inhibited the anchorage-
independent growth of SW480 cells in a dose-dependent manner.
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
Comparison with the related studies. Several approaches, such as hydrocarbon-stapled ꢁ-
helices,42 covalent hydrogen bond mimetics,43 and helical foldamers using ꢀ-peptides,44 have
been reported to stabilize a short peptide into a helical conformation for the inhibition of native
PPIs. Terphenyl derivatives are small organic molecules that can mimic one face of an ꢁ-helical
peptide.45 The substituents on the terphenyl scaffold can occupy the area where the side chains of
ꢁ-helical residues i, i + 3/4, and i + 7 are located. Some terphenyl derivatives have been reported
as new inhibitors for Bcl-xL/Bak46,47 and MDM2/p5348 PPIs. A number of other small molecules
that can mimic ꢁ-helical residues at positions i, i + 3/4, and i + 7 were also reported including
terpyridines,49 pyridazines,50 dipiperazinobenzenes,51 oxopiperazines,52 terphthalamides,53
22
ACS Paragon Plus Environment

Page 23 of 45
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
benzoylureas,54 benzamides and trisbenzamides,55–57 trispyridylamides,58,59 pyrrolopyrimidines,60
and triazines61. All of these ꢀ-helix mimetics use rigid or pre-organized scaffolds to overlap with
the backbone of ꢀ-helices and then position three substituents to the area where the side chains of
residues i, i + 3/4, and i + 7 are located.62 Projecting hot spots at positions i, i + 3/4, and i + 7 of
ꢀ-helices could be the primary driving force for many other PPIs and protein folding. The
installation of functional groups to trigger inhibitor selectivity/specificity and maintain drug-like
properties has not been an easy task for the above mentioned ꢀ-helix mimetics with few
successful examples reported.48,63 HTS identified that nutlin64 and benzodiazepinedione65 can
mimic F19 (i), W23 (i + 4), and L26 (i + 7) of an ꢀ-helix of p53 for the disruption of the
MDM2/p53 PPI. In a separated study, oxindole was designed to mimic the side chain of W23 of
p53, and two hydrophobic substituents were introduced to the spiro-oxindole core to mimic the
binding mode of the side chains of F19 and L26 of p53.66 All of these compounds were design
for the inhibition of MDM2/p53 PPIs and have not been generalized to the other PPI interfaces.
An anthraquinone scaffold was designed to mimic the sagittal axis of an ꢀ-helix. The substituents
on this scaffold were then used to mimic the binding mode of projecting hot spots L62 (i), I65 (i
+ 3), and D67 (i + 5) of Bim BH3 domain for the dual inhibition of the Bcl-2/Bim and Mcl-
1/Bim PPIs.67 The quinone substructure in the anthraquinone scaffold can be reactive and has
also been recognized as a PAINS substructure.18,19 Semi-rigid scaffolds that can express hot spot
side chains have been used to mimic the backbones of ꢀ-helices and explore key orientations at
the PPI interfaces to generate new inhibitors for unexplored PPIs.68 This study highlighted the
importance to mimic key side chains at the protein–protein interfaces.
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 PDB data mining based on the HippDB indicated that (1) the side chains of
hydrophobic projecting hot spots at positions i, i + 3, and i + 7 of an ꢀ-helix were placed to
23
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 24 of 45
1
2
3
4
5
6
7
8
9
common spatial areas when they interacted with the second protein; and (2) the hot spot pockets
of different PPI complexes had different sizes, shapes and chemical groups when they interacted
with the same hydrophobic projecting hot spots at positions i, i + 3, and i + 7 of ꢀ-helix. Based
on these observations, 4'-fluoro-N-phenyl-[1,1'-biphenyl]-3-carboxamide, was designed as a
generic scaffold that itself directly mimics the binding mode of the side chains of hydrophobic
projecting hot spots at positions i, i + 3, and i + 7 of an ꢀ-helix. A decoration of this generic
scaffold could lead to small-molecule inhibitors that selectively disrupt ꢀ-helix-mediated PPIs. A
series of selective small-molecule inhibitors for ꢁ-catenin/BCL9 PPIs including 22 was
developed based on scaffold 1. It is worth noting that scaffold 1 does not contain the exact
structures of ꢀ-helical side chains. Instead, it is a binding mode mimicry and serves as a starting
point for the design of selective small-molecule inhibitors for specific PPIs.
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
Conclusion
The discovery of a selective inhibitor for helix-mediated PPIs is challenging. The side
chains of hydrophobic projecting hot spots of an ꢀ-helix are oriented to the limited spatial areas
when it interacts with the second protein. This observation provides an opportunity to design
small-molecule inhibitors to mimic the binding mode of the side chains of ꢀ-helix hot spots and
generate selective PPI inhibitors with high ligand efficiency. In this study, compound 1 in Figure
1D was design as a generic scaffold that simultaneously mimic the hydrophobic interaction of
the side chains of hot spots at positions i, i + 3, and i + 7 of an ꢀ-helix. A PDB data mining
indicated that this scaffold can mimic 86% 733 PPIs in the HippDB that have predicted
hydrophobic projecting hot spots with this pattern and the concave pockets from the interacting
protein. This scaffold enables the design of selective small-molecule inhibitors for helix-
24
ACS Paragon Plus Environment

Page 25 of 45
Journal of the American Chemical Society
1
2
3
4
5
6
mediated PPI targets, because the sites that trigger PPI selectivity often exist adjacent to the hot
spot pockets where the projecting hot spots interact.
7
8
9
The ꢀ-catenin/BCL9 PPI is a key downstream effector of canonical Wnt signaling.
Previous crystallographic and biochemical studies revealed that BCL9 S352–F374 adopts an ꢁ-
helix structure to interact with the first armadillo repeat of ꢀ-catenin with L366 (i), I369 (i + 3),
and L373 (i + 7) of BCL9 as the projecting hot spots. Based on the structure of compound 1, a
small-molecule inhibitor for ꢀ-catenin/BCL9 PPIs, compound 22 in Figure 3A, was designed and
synthesized. In the design of 22, two positively charged pyrrolidino groups and a substituted
phenyl ring are introduced to scaffold 1 to trigger inhibitor selectivity between ꢁ-helix-mediated
PPIs. The binding mode of 22 was then evaluated by structure-activity relationship and site-
directed mutagenesis studies using two orthogonal assays, AlphaScreen and ITC. The cell-based
studies demonstrated that 22 can inhibit the transactivation of canonical Wnt signaling,
selectively disrupt ꢀ-catenin/BCL9 over ꢀ-catenin/cadherin PPIs in biochemical and cell-based
assays, downregulate Wnt target genes, and inhibit the growth of cancer cells with
hyperactivated Wnt signaling. The HPLC/MS analysis showed that 22 was effectively taken up
by HCT116 cells. The inhibitory activities of 22 and 23 (a structurally similar but less potent
analogue) are consistent in the biochemical and cell-based assays, indicating a specific inhibition
of canonical Wnt signaling. All of the above data indicate that 22 represents the first class-
specific small-molecule inhibitor for the ꢀ-catenin/BCL9 PPI reported to date. The further
optimization based on 22 will lead to potent and selective chemical probes for the ꢀ-
catenin/BCL9 PPI and potentially a new class of drug candidates for the treatment of advanced
cancers. Compared to scaffold 1, the two positively charged pyrrolidino rings and the additional
phenyl ring in 22 likely resulted in this compound and its derivatives specific ꢀ-catenin/BCL9
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
25
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 26 of 45
1
2
3
4
5
6
7
8
9
inhibitors in the cell-based studies. Future studies will be the optimization of inhibitors, the
determination of the ꢀ-catenin structure in complex with a small-molecule inhibitor (no such a
structure has been reported yet), the application of scaffold 1 to the other PPI interfaces in which
the ꢁ-helical partners also have hydrophobic projecting hot spots at positions i, i + 3, and i +7,
and the design of new inhibitor scaffolds that can simultaneously mimic the binding mode of
three projecting hot spots of ꢁ-helices.
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
Methods
Protein structure for computer modeling. The crystallographic coordinates for human ꢀ-
catenin (PDB id, 2GL7, 2.60 Å resolution, Rcryst = 0.223) were obtained from the Research
Collaboratory for Structural Bioinformatics (RCSB) protein data bank (PDB). The preparation of
the crystal structure and molecular modeling were achieved with the commercially available
Schrodinger
(http://www.schrodinger.com/),
BIOVIA
Discovery
Studio
3.0
(http://accelrys.com/), and SYBYL X2.0 (http://www.tripos.com/) software packages. The
missing side chains of ꢀ-catenin were added in SYBYL X2.0. The protonation states of the
residues were set to pH 7.0 when adding hydrogen. The AMBER 7 force field 99 and the
AMBER FF99 charges within SYBYL X2.0 were used to optimize the orientation of hydrogen
atoms and the missing side chains of the protein. After the protein complex was optimized.
Chains B (Tcf4), C (BCL9), D (the second monomer of ꢀ-catenin), E (the second monomer of
Tcf4), F (the second monomer of BCL9), and solvent molecules were removed, leaving only one
monomer of ꢀ-catenin for further calculation. The residues in the BCL9 L366/I369/L373 binding
site of ꢀ-catenin include D144–A146, L148, A149, A152, I153, E155–L160, D162–A171,
M174, V175, Q177, L178, K180, K181, A183, S184, A187, I188, M194, and I198.
26
ACS Paragon Plus Environment

Page 27 of 45
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
Fragment hopping to design 4'-fluoro-N-phenyl-[1,1'-biphenyl]-3-carboxamide as a
binding mode mimicry for hydrophobic side chains of ꢀ-helical hot spots i, i + 3, and i + 7.
The step-wise protocol for fragment hopping was described previously.23,24 The crystallographic
coordinates for human ꢀ-catenin (PDB id, 2GL7) were used in fragment design. The side chains
of three projecting hot spots of BCL9, L366, I369, and L371 were identified as the critical
binding elements to mimic. The BCL9 L366/I369/L373 binding site of ꢀ-catenin was defined as
the hot spot pocket to accommodate the designed fragment. The hydrophobic side chains that
defined the hot spot pocket were from L148, A149, A152, L156, L159, L160, V167, K170,
A171, M174, and L178 of ꢀ-catenin. The basic fragment library23 and the bioisostere library23
were interrogated to identify new fragments that can match the proposed critical binding
elements.
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
The binding modes of new fragments were generated by fragment docking using
AutoDock 4.2. Only the polar hydrogen atoms were kept on the protein structure, and the
Kollman united atom charges were assigned. The grid maps were calculated using AutoGrid with
the grid spacing of 0.375 Å. For the fragment docking any atoms within 6 Å from the above
critical binding elements were used to define the grid box, which resulted in two pockets: one
pocket included the side chains of A152, L156, L159, L160, V167, K170, A171, and M174 of ꢀ-
catenin, and the side chain carboxylic oxygens of E155 and D162 of ꢀ-catenin; and the second
+
pocket included the side chains of L148, A149, A152, M174, and L178, the side chain NH3 of
K181, the side chain OH of S184, the backbone NH of A183, the side chain carboxylic oxygen
atoms of D145 and S184, and the backbone carbonyl oxygen of L148 of ꢀ-catenin. Docking was
performed using the Lamarckian genetic algorithm, and the pseudo-Solis and Wets method was
applied for the local search. Each docking experiment was performed for 50 times, yielding 50
27
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 28 of 45
1
2
3
4
5
6
7
8
9
docked conformations. The other settings were the standard default parameters. The results of
the docking experiments were evaluated by the auxiliary clustering analysis and the visual
inspection to match the proposed binding elements. The binding poses of the fragments that
matched the proposed binding elements were stored in a SYBYL molecular database.
The distance between each fragment of two pockets was measured, and the linkers in the
linker library23 were merged to generate the ligand structure. The synthetic accessibility of the
generated compounds was evaluated with the assistance of SciFinder. The generated ligand
scaffolds were then docked in the binding site and visually inspected. If the newly generated
ligand scaffold did not match the proposed critical binding elements, the ligand scaffold was
rejected, and new structures were constructed by repeating the above steps. If the binding mode
of the new ligand scaffold matched the proposed critical binding elements, they would be kept
for further evaluation. The rules for metabolic stability23 and the toxicophore library23 were used
to evaluate biocompatibility of the acceptable ligand scaffolds.
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
Ligand docking using Glide 5.8. The Glide grid box was defined to include all the residues of
the BCL9 L366/I369/L373 binding site. The default parameters were used to generate the
receptor grid. The standard precision (SP) mode was used in ligand docking. The ligand scaling
factor was set to 0.5 for the atoms with the partial charges lower than 0.15. The number of poses
per ligand for the initial phase of docking was increased to 10,000. The 1,000 best poses per
ligand were kept for energy minimization with a maximum number of the minimization steps of
5,000. A maximum of 100,000 ligand poses per docking run and 50 poses per ligand were
collected. Up to 100 poses per ligand were kept for the post-docking minimization. The default
settings were used for the remaining parameters.
28
ACS Paragon Plus Environment

Page 29 of 45
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
Ligand docking using AutoDock 4.2. The three-dimensional (3D) structures of the ligands were
built, and the partial atomic charges were calculated using the Gasteiger–Marsili method. The
rotatable bonds in the ligands were defined using AutoTors, which also united the nonpolar
hydrogens and partial atomic charges to the bonded carbon atoms. The grid maps were
calculated using AutoGrid. The AutoDock area was defined to include all of the residues in the
BCL9 L366/I369/L373 binding site, and the grid spacing was set to 0.375 Å. Docking was
performed using the Lamarckian genetic algorithm, and the pseudo-Solis and Wets method was
applied for the local search. Each docking experiment was performed 100 times, yielding 100
docked conformations. The other settings were the default parameters. All of the ligands
followed the same docking protocol. The results of the docking experiments were evaluated by
the auxiliary clustering analysis and the visual inspection to match the proposed critical binding
elements.
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
Synthesis of compounds 2–30. The synthetic procedures for 2–30 are shown in the supporting
information.
Protein expression and purification. ꢀ-Catenin and its mutants (residues 138–686) were cloned
into a pET-28b vector carrying a C-terminal 6 × histidine (Novagen), and transformed into E. coli
BL21 DE3 (Novagen). Cells were cultured in LB medium with 30 µg/mL kanamycin until the
OD600 was approximately 0.8, and then protein expression was induced with 400 µM of IPTG at
20 °C overnight. Cells were lysed by sonication. The proteins were purified by Ni-NTA affinity
chromatography (30210, Qiagen) and dialyzed against a buffer containing 20 mM of Tris (pH
8.5), 100 mM NaCl, 10% glycerol, and 3 mM DTT. The purity of ꢀ-catenin was greater than 95%
as determined by SDS-PAGE gel analysis. Native non-denaturing gel electrophoresis was
performed to confirm the homogeneity of the purified proteins. Thermal-shift assay was
29
ACS Paragon Plus Environment