Novel c-Met inhibitory olive secoiridoid semisynthetic analogs for the control of invasive breast cancer

Article abstract of DOI:10.1016/j.ejmech.2016.04.043

Dysregulated receptor tyrosine kinase c-Met and its ligand HGF is valid and attractive molecular target for therapeutic blockade in cancer. Inspired by the chemical structure of the naturally occurring olive secoiridoid (-)-oleocanthal (1) and its documen

Full text of DOI:10.1016/j.ejmech.2016.04.043

Accepted Manuscript
Novel c-Met inhibitory olive secoiridoid semisynthetic analogs for the control of
invasive breast cancer
Mohamed M. Mohyeldin, Belnaser A. Busnena, Mohamed R. Akl, Ana Maria Dragoi,
James A. Cardelli, Khalid A. El Sayed
PII:
S0223-5234(16)30336-1
10.1016/j.ejmech.2016.04.043
EJMECH 8558
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 28 October 2015
Revised Date: 16 April 2016
Accepted Date: 16 April 2016
Please cite this article as: M.M. Mohyeldin, B.A. Busnena, M.R. Akl, A.M. Dragoi, J.A. Cardelli, K.A. El
Sayed, Novel c-Met inhibitory olive secoiridoid semisynthetic analogs for the control of invasive breast
cancer, European Journal of Medicinal Chemistry (2016), doi: 10.1016/j.ejmech.2016.04.043.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT
Novel c-Met Inhibitory Olive Secoiridoid Semisynthetic Analogs for the Control of
Invasive Breast Cancer
1
1
1
2
Mohamed M. Mohyeldin , Belnaser A. Busnena , Mohamed R. Akl , Ana Maria Dragoi , James
2
1
A. Cardelli , Khalid A. El Sayed *
1
Department of Basic Pharmaceutical Sciences, School of Pharmacy, University of Louisiana at
Monroe, 1800 Bienville Drive, Monroe, Louisiana 71201, USA
2
Department of Microbiology and Feist-Weiller Cancer Center, Louisiana State University
Health Sciences Center, Shreveport, Louisiana, USA
*
To whom Correspondence should be addressed. Telephone 318-342-1725.
Fax: 318-342-1737.
E-mail: elsayed@ulm.edu

ACCEPTED MANUSCRIPT
Abstract
Dysregulated receptor tyrosine kinase c-Met and its ligand HGF is valid and attractive
molecular target for therapeutic blockade in cancer. Inspired by the chemical structure of the
naturally occurring olive secoiridoid (-)-oleocanthal (1) and its documented anticancer activity
against c-Met-dependent malignancies, a previous study reported tyrosol sinapate (4) as a c-Met
inhibitor hit. This study reports additional semisynthetic optimization and SAR of 4 to improve
its selective activity against c-Met-dependent breast cancer by increasing its capacity to inhibit
c-Met phosphorylation. Forty-three compounds (5-47) were synthesized, among which the
novel analog homovanillyl sinapate (HVS-16) was distinguished for its remarkable activity.
HVS-16 substantially impaired c-Met-mediated proliferation, migration, and invasion across
human breast cancer cell lines in two- and three-dimensional culture systems, while similar
treatment doses were found to have effect neither on the non-tumorigenic human mammary
epithelial cell growth nor on the c-Met independent breast cancer cell viability. HVS-16 showed
a dose-dependent inhibition of ligand-mediated c-Met activation in human breast cancer cells.
Docking studies revealed that HVS-16 fits very well inside c-Met crystal structures, satisfying
critical interactions at the ATP binding site. This study identified important structural
pharmacophoric features in HVS-16 and correlated its postulated binding pose with c-Met
kinase assay data that would guide future olive secoiridoid bioisostere lead design. Results
presented herein suggest HVS-16 as a promising c-Met inhibitor validated hit with potential to
control invasive breast malignancies with aberrant c-Met activity.
Keywords
Antimigratory, Antiproliferative, Anti-invasive, Breast cancer, c-Met, Olive secoiridoids,
Homovanillyl sinapate.

ACCEPTED MANUSCRIPT
1
. Introduction
c-Met is the prototypic member of a unique subfamily of receptor tyrosine kinases
RTKs) which also contains the protein kinase Ron. It is the only known high-affinity receptor
(
for its ligand hepatocyte growth factor (HGF), also known as scatter factor [1]. Upon HGF
stimulation, c-Met induces several biological responses including cell proliferation, survival,
differentiation, motility, angiogenesis, and invasion [2, 3]. c-Met dysregulated activation is
involved in the development and progression of multiple tumor types, such as hereditary
papillary renal cell carcinoma [4], lung cancers [5], head and neck cancers [6], breast and
prostate cancers [7, 8] as well as gastric cancer [9]. The overexpression of c-Met and/or HGF
has been associated with a metastatic phenotype and poor prognosis in breast and gastric
cancers [10, 11]. Activation of HGF/c-Met signaling in the tumor microenvironment may
confer resistance to anti-RTK cancer therapies already in clinical use, including EGFR and
BRAF anticancer kinase inhibitors [12, 13]. Given the important role of aberrant HGF/c-Met
signaling in cancer progression and resistance to cancer therapies, this signaling axis has
emerged as one of the most promising therapeutic targets in anticancer drug discovery [7].
Targeting the c-Met kinase domain with small molecules resulted in more than 240 c-Met
inhibitors reaching various clinical stages as cancer treatments [14, 15]. Pfizer’s c-Met/ALK
dual inhibitor crizotinib and Exelixis’ VEGFR2/c-Met multikinase inhibitor cabozantinib are
the first small-molecule c-Met inhibitors approved so far by the FDA for the treatment of late-
stage NSCLC and metastatic medullary thyroid cancer, respectively [15].
The Mediterranean populations have reduced risk incidence for particular types of
cancer, compared to other geographical populations. This may be attributed to the
Mediterranean diet rich in extra-virgin olive oil (EVOO), apart from possible genetic factors

ACCEPTED MANUSCRIPT
[
16]. EVOO is rich in other minor bioactive phenolics, including simple phenols, phenolic
acids, flavonoids, lignans, and secoiridoids [17]. (-)-Oleocanthal (1) is a naturally occurring
secoiridoid from EVOO, which has recently attracted considerable attention due to its various
biological effects in inflammation, Alzheimer’s disease and cancer [18-22]. (-)-Oleocanthal has
been shown to mediate its anticancer effects through the disruption of c-Met kinase-related
pathways [20, 23]. 1 Competitively inhibited the c-Met kinase activation in a biochemical assay
[
19, 20]. In addition, the intracellular mechanisms involved in mediating 1’s c-Met-related
anticancer effects in breast cancer cells and in a mouse model have been characterized [23].
Recently, 1 proved to target Hsp90, a chaperone protein stabilizing a number of proteins
required for tumor growth, including c-Met [24]. Ligstroside aglycone (2) is another olive
secoiridoid, which triggered apoptotic cell death in HER2-overexpressing breast carcinoma and
showed moderate cytotoxicity against a panel of human cancer cell lines [25]. Unlike 1,
ligstroside aglycone did not demonsrate c-Met inhibition despite showing moderate
antimigratory activity against human metastatic breast cancer cells [19].
Chemically, 1 is the elenolic acid ester of the common olive phenolic alcohol tyrosol
(3). Based on the activity of 1 as c-Met inhibitor, the synthesis of different tyrosol ester and
carbamate analogs was reported [19]. Design of these analogs was based on replacing the 1’s
elenolic acid moiety with different bioisosteres, including natural olive-derived phenolic acids,
synthetic heteroaromatic acids, and diverse aromatic carbamates, to develop 1-based c-Met
inhibitors. Among these analogs, tyrosol sinapate (4) showed moderate ATP-competitive c-Met
phosphorylation inhibitory activity in cell-free assays [19]. 4 Demonstrated antimigratory,
antiproliferative, and anti-invasive activities against the highly metastatic human breast cancer
MDA-MB-231 cells consistent with its c-Met inhibitory activity, without being toxic to the non-
tumorigenic human mammary epithelial cells [19]. Therefore, the main objective of this study is

ACCEPTED MANUSCRIPT
to subject the initial hit 4 to extensive structure-activity relationship (SAR) studies via
optimization of its sinapic acid and tyrosol moieties as well as the ester linker.
2
. Results and Discussion
2
.1. Chemistry
Thirty-seven new 7-9, 11-14 (Scheme 1), 16, 18-39 and 41-47 (Scheme 2) and six
known analogs 5, 6, 10, 15 (Scheme 1), 17 and 40 (Scheme 2) were synthesized. The esters 5-9
and 13-47 have been prepared by convenient synthesis using the highly chemoselective
Mitsunobu esterification reaction (Scheme SI1). Sinapic acid amidation reaction of tyramine
has been used to synthesize analog 10 (Scheme SI2). Furthermore, condensation reaction of
tyramine with sinapaldehyde has been used to generate the imine 11 (Scheme SI3). In a similar
fashion, the 1,2,4-oxadiazole analog 12 has been synthesized via condensation reactions
involving sinapic acid and hydroxyl-benzamidine oxime (Scheme SI4). Most reaction yields
were in the range of 50-60%. Concerning the known analogs, tyrosol ferulate (5) was first
isolated as a natural product from Heracleum lanatum Michs. in 1982 [26]. Tyrosol coumarate
(6) was isolated from Sargentodoxa cuneata in 1988 [27]. Both 5 and 6 have shown weak
cytotoxicity against the oral cancer cell lines SAS and OEC-M1 (IC >100 µM) [28]. Tyrosol
5
0
sinapamide (10) was first isolated from the branches of Porcelia macrocarpa and showed
moderate cytotoxicity against human myeloid leukemia HL-60 and cervical cancer HeLa cell
lines [29-31]. Hydroxytyrosol ferulate (15) was patented in 2007 for dermatological care
purpose [32]. Homovanillyl ferulate (17) has been only reported as a natural product from the
wood of Alnus hirsuta in 1972 [33]. Finally, coniferyl ferulate (40) was isolated from the aerial
parts of Coreopsis longula in 1985 [34]. Recently, 40 demonstrated a strong inhibition of
human placental glutathione S-transferase (GST) activity, with an IC value of 0.3ꢀꢁM, and
5
0

ACCEPTED MANUSCRIPT
therefore proposed as an antitumor adjuvant therapy to improve drug sensitization and reverse
resistance [35]. Re-synthesizing these known compounds aimed to assess their c-Met inhibitory
and anticancer activities against MDA-MB-231 cells. The identity of known compounds was
1
13
confirmed by comparison of their H and C NMR data (Tables SI1-4, SI9, SI12-15 and SI20)
with literature [26, 27, 31, 34].
1
13
The structures of the new compounds were confirmed by H NMR, C NMR, and MS
spectra. For instance, the HRESIMS data of 7 showed a molecular ion peak at m/z 345.1336
-
1
13
[
M-H] , suggesting a molecular formula C H O and possible ester analog of 3. H and C
19 22 6
NMR data indicated that 7 is 4-hydroxyphenethyl 3-(4-hydroxy-3,5-dimethoxyphenyl)-
1
propanoate (Tables SI1 and SI12). The H NMR and PENDANT data of 7 showed typical
patterns of aromatic signals for para-substituted phenethyl moiety, in addition to aromatic and
1
methoxy signals for the sinapinic acid moiety. The H NMR data of 7 also revealed the
downfield shifting of the methylene H -1' to δH 4.24, compared to that of its parent tyrosol
2
moiety 3 (∆δH +0.54), suggesting possible esterification of the primary alcohol C-1' (Table SI1).
Esterification was further confirmed in PENDANT spectrum through the characteristic
downfield ester carbonyl C-1 (δC 173.1, Table SI12), in addition to the downfield shift of C-1'
(
ꢂδ +4.20, Table SI12), and the upfield shift of C-2' (ꢂδ -4.20). Additionally, esterification
C
C
3
was supported by a J-HMBC correlation of the methylene triplet H -1' with the ester carbonyl
2
1
13
C-1, confirming the identity of 7 as tyrosol sinapinate. The H and C NMR data of compounds
-47 (Tables SI1-22) as well as their HRESIMS data were used to confirm their identity in a
similar fashion.
8
2
2
.2. Biological Evaluation and Structure-Activity Relationship (SAR)
.2.1. Z'-LYTE™ c-Met kinase assay

ACCEPTED MANUSCRIPT
The initial screening cascade used the Life Sciences’ cell-free Z'-LYTE™ enzymatic
assay as the screening assessment for designed compounds. Because all c-Met-mediated
functions are dependent on its kinase activity [36], the in vitro ability of synthesized analogs 5-
4
7 to inhibit c-Met phosphorylation (activation) was directly tested on the purified kinase
domain of c-Met (amino acids 956–1390), that was in vitro phosphorylated to achieve the
highest level of intrinsic kinase activity. In this experiment, Z′-LYTE™ Tyr6 peptide was used
as a substrate; thus, the changes in its phosphorylation can directly reflect the c-Met kinase
activity. Meanwhile, 1 and 4 were used as positive controls for activity comparison. Table 1
shows the concentration that resulted in 50% c-Met phosphorylation inhibition (IC values) for
5
0
all compounds. The calculated IC of 1 in this assay was 5.2 µM, which was consistent with its
5
0
reported value 4.8 µM, validating the results of this study [19, 20]. The SAR study was based
upon investigating three moieties of the preliminary hit 4; the sinapic acid and tyrosol moieties
and their ester linker. The SAR study started with exploring the importance of methoxy groups
in the sinapic acid moiety of 4 for c-Met inhibition by synthesizing analogs 5 and 6. Tyrosol
ferulate (5) has only one methoxy substituent in ortho-position to the C-7 phenolic hydroxyl
group. Analog 5 was able to inhibit c-Met kinase phosphorylation by nearly the same extent as
the parent tyrosol sinapate (4) with an IC50 value of 12.9 µM. However, omitting both 4’s
methoxy groups, as represented by tyrosol coumarate (6) significantly reduced the activity to an
IC value of 20.2 µM (Table 1). Therefore, the presence of at least one methoxy group next to
5
0
the phenolic hydroxyl in the acid part of 4 is essential for the kinase inhibitory activity. The
importance of the free C-7 phenolic hydroxyl group for c-Met inhibition was previously
investigated via the remarkable activity difference between 4 and its C-7 methyl ether, which
highlighted the hydrogen bond donating role (HBD) of the acid part's phenolic hydroxyl for
binding [19]. Bioisosteric replacement of this group with an amino group did not show a

ACCEPTED MANUSCRIPT
significant improvement in the activity as represented by analog 38, which has an IC value of
5
0
1
2.2 µM, suggesting the need for only one HBD group at the C-7 position for best c-Met
2
,3
binding (Table 1). Reducing the sinapic acid’s double bond ꢂ in 4, as in analog 7, abolished
the kinase inhibitory activity. Similarly, substituting the α-carbon of the olefinic double bond
with cyano or sulfhydryl groups significantly reduced the activity as in analogs 8 and 9. The
2
,3
influence of changing the geometrical isomerism (E/Z isomers) around the olefinic ꢂ
has
been also studied through the synthesis of the Z-analogs 34 and 36 and comparing their c-Met
inhibitory activities with the corresponding E-analogs 33 and 35 (Table 1). The E-isomer, in all
analogs, proved to be significantly more potent at the ATP binding pocket of c-Met compared
to the Z-isomer, which clearly indicated that the c-Met inhibitory activity of this class resides in
the E-isomer. Concerning the ester linker, ester analogs of 3 were previously shown to be more
active than carbamate and sulfonyl carbamate analogs, suggesting the preference of the ester
pharmacophore for better c-Met inhibitory activity [19]. However, the critical role of the ester
functionality in 4 has been further explored by using different bioisosteres, including the amide
1
0, the imine 11, and the 1,2,4-oxadiazole moiety 12. The lack of c-Met kinase inhibitory
activity of 10-12 confirmed the important role of the ester functionality at the ATP binding
pocket (Table 1). The ester moiety was hypothesized not to contribute any binding role, based
on previous findings, yet it provided the proper molecular size and conformation for adequate
binding [19, 20]. The ester linker must have been required to maintain the bioactive U-shaped
binding mode at the Met’s ATP binding pocket characteristic for type-I binding and thus, it is
indispensable for the activity. Therefore, ferulic and/or sinapic acid moieties were maintained as
appropriate scaffolds for c-Met inhibition while optimizing the alcohol part. The alcohol moiety
optimization began with determining the activity role of the tyrosol’s C-6' phenolic hydroxyl
group of 4 via its bioisosteric replacement with an amino group in analogs 18 and 19 and with a

ACCEPTED MANUSCRIPT
methoxy group in analogs 24 and 26. Compounds 18 and 19 exhibited more than 2-fold
increase in the c-Met kinase inhibitory activity versus 4, with IC values of 5.2 and 6.1 µM,
5
0
respectively. On the other hand, replacing the C-6' phenolic hydroxyl group with a methoxy
group in 24 and 26 caused a drastic reduction in the activity, with IC >20.0 µM (Table 1). This
5
0
clearly highlighted the importance of the HBD role of C-6' phenolic hydroxyl group in 4 for
maintaining potency, and suggests that an additional HBD at this position, as represented by an
amino group, may offer better binding affinity at the ATP binding pocket, improving the overall
bioactivity. This finding augmented previous preliminary binding mode study of 4 in three
highly resolved c-Met kinase domain crystal structures where the tyrosol’s C-6' phenolic
hydroxyl group exhibited HBD interactions with the hinge region’s essential amino acids
Met1160 or Pro1158 of the three c-Met crystal structures [19]. The chain extension/shortening
as a classical medicinal chemistry optimization strategy was adopted in order to investigate the
optimal length of the alkyl side chain linker in the tyrosol moiety of 4 for efficient c-Met
inhibition. Increasing the chain length by one carbon atom, as represented by analogs 20, 21,
and 35 showed a slight improvement in the c-Met inhibitory activity compared to 4, with IC50
values of 8.2, 9.1 and 8.0 µM, respectively (Table 1). Decreasing the length of the alkyl side
chain linker by one carbon atom, as represented by analog 32, demonstrated a nearly similar
pattern of activity enhancement, with an IC value of 8.3 µM (Table 1). However, shortening
5
0
the chain length by an additional carbon atom represented by analogs 22 and 23 completely
diminished the c-Met inhibitory activity (Table 1). This clearly indicates the important role of
the distance limit between the two aromatic rings A and B of the alcohol and the acid moieties
for optimal c-Met inhibition. The results suggested that retaining six to eight chemical bonds i.e.
five to seven atoms in the linker could afford a proper spatial distance between the phenyl rings
A and B and thus maintain the c-Met inhibition. Based on these results, 4-hydroxybenzyl

ACCEPTED MANUSCRIPT
alcohol and 3-(4-hydroxyphenyl)-1-propanol in addition to tyrosol were considered as potential
scaffolds appropriate for further optimization of the alcohol moiety. In order to explore the
effect of replacing the tyrosol’s phenolic hydroxyl group with a catechol, the common olive
phenolic hydroxytyrosol was used in the esterification reactions, as represented by analogs 13-
1
5 and 37. Replacing the tyrosol moiety in 4, 5 and 38 with hydroxytyrosol, as shown in
analogs 13, 15 and 37, improved the c-Met inhibitory activity by at least 2-folds, demonstrating
IC values of 5.8, 7.8 and 6.9 µM, respectively (Table 1). The binding role of this C-5' phenolic
5
0
hydroxyl group was investigated by replacing it with an amino group as in analog 28.
Surprisingly, 28 was much less active (~2-fold) than 13, with an IC50 value of 11.4 µM,
suggesting no HBD role for C-5' OH in 13 for the c-Met inhibition (Table 1). Instead, C-5'
oxygen in 13 is likely to play a better hydrogen bond acceptor (HBA) role for c-Met binding
when compared to the aniline nitrogen in 28. The lone pair of the 28’s nitrogen atom can
interact with the neighboring π-system of the benzene ring to form various resonance structures
and hence it is less likely to take part in a hydrogen bond (HB). In contrast, the C-5' oxygen in
1
3 has an additional lone pair of electrons which are not involved in resonance and more likely
to play a significant HBA role at the ATP binding pocket, improving the overall activity as
compared to 28. In order to gain further insight on this position, the C-5' phenolic hydroxyl
group in 13, 15 and 37 was replaced with a methoxy group in analogs 16, 17 and 33.
Interestingly, these analogs exhibited nearly six-fold enhancement in the c-Met kinase
inhibitory activity, with IC values of 1.0, 1.3 and 1.2 µM, respectively (Table 1, Figure 1).
5
0
This clearly indicated the preference of a small hydrophobic methoxy group, retaining the HBA
property, over the C-5' phenolic hydroxyl group in ring A. The binding affinity enhancement of
1
6, 17 and 33 might be attributed to the ability of the C-5' methoxy group to exert hydrophobic
interactions at the ATP binding site of the c-Met protein, since both the methoxy and hydroxyl

ACCEPTED MANUSCRIPT
can act as HBA groups but the superior activity of the methoxy group at this position suggested
additional hydrophobic interactions with the c-Met kinase. Subsequently, a strategy of adding a
methoxy group at the same position, next to the phenolic hydroxyl group, in both 4-
hydroxybenzyl alcohol and 3-(4-hydroxyphenyl)-1-propanol was adopted to probe possible
activity improvement. Accordingly, 3-(4-hydroxyphenyl)-1-propanol was replaced with 4-(3-
hydroxypropyl)-2-methoxyphenol in the esterification reactions as represented by analogs 29
and 30. Adding a methoxy group in 29 and 30 did not improve the activity as would have been
expected, compared to the unsubstituted analogs 20 and 21 (Table 1). This might be due to the
involvement of the methoxy group in unfavorable interaction(s) or steric clashes with the target
receptor or its inability to demonstrate good fitting within the hydrophobic space as a result of
the alkyl side chain elongation by one carbon atom. In order to solve this problem, a structural
rigidification strategy, via introducing a double bond in the alkyl side chain of the alcohol part,
was adopted to confer a favorable effect on the entropic (ꢂS) component of the binding affinity
without adversely affecting the enthalpy (ꢂH) component, if nearly the correct conformer can
be locked at the ATP binding pocket via this strategy. However, this approach was not
2
',3'
successful, as the ꢂ
unsaturation in the alkyl side chain of 4-(3-hydroxypropyl)-2-
methoxyphenol totally eliminated the activity, as seen in analogs 31 and 40 (Table 1).
Encouraged by the promising c-Met inhibitory activity of analog 32, compared to the parent 4, a
methoxy group was sought to be placed next to the phenolic OH in ring A to gain potency,
which afforded compound 41. The methoxy substitution only resulted in approximately 4-fold
increase in the activity with the 5-atom linker between the phenyl rings A and B (41 versus 32)
compared to a 14-fold enhancement in the potency for the 6-atom linker (16 versus 4, Table 1).
A possible explanation could be that the five-atom linker is neither enough to provide the
optimal distance to form an intermolecular HB from a probable donor amino acid in the ATP

ACCEPTED MANUSCRIPT
binding pocket to the C-4′-OCH in 41 nor to perfectly accommodate this methoxy into the
3
hydrophobic space of c-Met kinase. Hence, the 6-atom linker provided by the 2-phenylethanol
(tyrosol) ester series was optimum and might furnish novel promising c-Met inhibitors.
Therefore, the active homovanillyl alcohol moiety in 16 was maintained and an additional
substitution with a methoxy group at C-7' position of the phenyl ring A was attempted.
However, the additional C-7′-OCH substituent in 42 and 43 significantly eroded the potency,
3
resulting in a 3-fold loss of activity when compared to 16 and 17 (Table 1). This might be due
to a steric factor at the C-7′ position or the involvement of the methoxy group in unfavorable
interaction(s) with the target receptor. At this point, the PDB deposited crystal structure of
crizotinib (PDB code: 2WGJ) was obtained as an ATP-competitive c-Met inhibitor [37].
Crizotinib structure (48) was analyzed and identified its 2-aminopyridine moiety bound to the c-
Met's hinge region via HB interaction of the pyridine nitrogen with the Met1160 amide NH. On
the basis of this observation, a strategy of pharmacophore hybridization was adopted in an
attempt to gain potency via introducing the typical kinase hinge binding aminopyridyl group,
similar to that of 48, to the six-atom and five-atom linker pharmacophores as represented by
analogs 44-47. Interestingly, compounds 44 and 45, with six-atom linker, were almost
equipotent to 16 and 17, against c-Met kinase enzyme, respectively, suggesting the ability of the
2
-aminopyridine moiety to make up for the lost activity due to the absence of small
hydrophobic methoxy group at C-5' in the pyridyl moiety, possibly by offering additional
binding pharmacophore, C-6' amino group, in addition to the electron-poor pyridine ring which
might be advantageous to the potency (Table 1). On the other hand, the five-atom linker analogs
4
6 and 47 were less active (>2-folds) than 39 and 41, with IC values of 5.8 and 5.3 µM,
50
respectively (Table 1). This could be attributed to the inability of the five-atom linker to provide
the optimal distance to form an intermolecular HB between the ATP binding pocket amino

ACCEPTED MANUSCRIPT
acids with the amino pyridine nucleus of 46 and 47, along with the lack of potential
hydrophobic interactions due to the absence of C-5' methoxy group and thus negatively
influencing the activity. Taken together, the SAR presented herein afforded nine analogs (16,
1
7, 33, 39, and 41-45) with better c-Met inhibitory activity than oleocanthal (1, Figure 1, Table
). Apparently, HVS-16 proved to be the most potent inhibitor in this cell-free assay, being able
1
to inhibit c-Met phosphorylation induced by the addition of ATP in a dose-dependent manner
and demonstrating nearly fourteen-fold enhancement in the activity versus the parent 4 and five-
fold activity improvement compared to 1 (Figure 1, Table 1). Thus, HVS-16 can be considered
as a potential c-Met inhibitor hit appropriate for further validation.
2
.2.2. HGF-induced cell proliferation MTT assay
The growth inhibitory effect of analogs 1-47 on the c-Met-dependent, triple negative
human breast carcinoma (TNBC) MDA-MB-231 and MDA-MB-468 cells was assessed using
MTT assay. HGF (40 ng/mL) was used in the growth media as a mitogen to induce cell
proliferation through activating the c-Met receptor. In this assay, at least four different
concentrations per each analog were tested and used to assess the concentration that resulted in
5
0% cell growth inhibition (IC ) (Table 2). Oleocanthal (1) and tyrosol sinapate (4) were used
50
as standard positive controls. The parent 4 as well as the natural olive phenolics 2 and 3 showed
moderate to high micromolar antiproliferative activity against both breast cancer cell lines
(Table 2). In general, nearly all olive secoiridoid semisynthetic bioisosteres demonstrated better
antiproliferative activity than the parent 4 as well as the natural phenolics 2 and 3, highlighting
not only the importance of the moieties of tested analogs that belong to acids and alcohols used
for the activity, compared to 3, but also the significance of introducing particular substituents at
the two aromatic rings A and B of the alcohol and the acid moieties, compared to 4, as well as

ACCEPTED MANUSCRIPT
the impact of optimizing the distance limit between these two rings. The antiproliferative
activity level of all synthesized analogs towards MDA-MB-468 cell line is parallel to the one
observed for MDA-MB-231 cells and thus the discussion mainly focused on the latter cell line.
Analogs 16, 20, 21, 24-26, 28, 37, 39, 41, 44 and 45 showed the most promising
antiproliferative activity in this assay with IC values <10 µM (Table 2). Obviously, 21, 39 and
5
0
4
1 were among the most active compounds, inhibiting cell viability with IC values of 4.4, 4.2
50
and 6.3 µM, respectively, and demonstrating thirteen to nineteen-fold enhancement in the
activity, compared to 4 (IC 79.5 µM) and more than two to four-fold improvement versus 1
5
0
(
IC 15 µM, Table 2, Figure 2). This clearly indicates the significant impact of extending the
50
tyrosol alkyl side chain linker in 21 as well as chain shortening in 39 and 41 on improving the
antiproliferative activity. HVS-16, the most potent c-Met phosphorylation inhibitor in the Z-
LYTE™ assay, was also the most active analog in cell proliferation assay. HVS-16
significantly inhibited MDA-MB-231 and MDA-MB-468 cell growth in a dose-dependent
manner as compared to vehicle-treated cells, with IC values of 3.8 and 6 µM, respectively,
5
0
exhibiting nearly twenty-one-fold improvement in the antiproliferative activity versus 4, parallel
to its c-Met kinase inhibitory potential (Table 2). Generally, the antiproliferative SAR observed
in this assay is parallel to the one discussed above for the c-Met inhibitory activity in the Z-
LYTE™ kinase assay, with only few exceptions to be addressed. Firstly, although the
semisynthetic esters 39 and 41 were almost equipotent to HVS-16 in cell proliferation assay,
they were 2-fold less active than HVS-16 in the Z-LYTE™ kinase assay suggesting other
molecular target(s), in addition to c-Met, for their antiproliferative activity (Tables 1 and 2).
Secondly, analogs 20 and 21 were active against the breast cancer MDA-MB-231 cells, with
IC values of 7.6 and 4.4 µM, respectively, whereas only marginally active against c-Met in the
5
0
Z-LYTE™ assay, with IC50 values of 8.2 and 9.1 µM, respectively (Tables 1 and 2). This

ACCEPTED MANUSCRIPT
clearly indicates that the chain extension strategy adopted in 20 and 21 was able to improve the
antiproliferative activity, only partially through inhibition of c-Met activation. A possible
explanation could be that alkyl side chain elongation by an additional carbon might allow the
phenyl ring A of the alcohol moiety to be involved in π-π stacking with a molecular target
receptor other than c-Met, thus anchoring the molecule in a pose that allows for better ligand-
receptor interaction, and hence better antiproliferative activity. Similarly, analog 28, with an
amino group at the C-5' of the alcohol part, was only 2-fold less potent than HVS-16 in cell
proliferation assay, with an IC value of 8.6 µM, while being much less active (~11-folds) than
5
0
HVS-16 against c-Met in the kinase assay, with an IC value of 11.4 µM (Tables 1 and 2). This
5
0
may suggest the ability of the C-5' amino group of 28 to demonstrate strong HB interactions to
molecular target(s) involved in cell proliferation, other than c-Met, offering better binding
affinity and improving the overall activity of this compound in cell proliferation. Finally,
despite the fact that replacing the C-6' phenolic hydroxyl group of the tyrosol part with a
methoxy group as well as changing the geometrical isomerism (E/Z isomers) around the olefinic
double bond of the acid part proved detrimental to c-Met inhibitory activity in the Z-LYTE™
assay, the semisynthetic esters 24-26 were still able to demonstrate promising antiproliferative
activity against MDA-MB-231 cells, with IC50 values of 7.3, 8.2 and 9.2 µM, respectively,
suggesting molecular target(s) different from c-Met for their antiproliferative activity (Table 2).
The effect of various doses of 21, 39, and 41 on the viability of MDA-MB-231 cells after 72 h
culture period is demonstrated in Figure 2.
2
.2.3. HGF-induced cell migration assay
Activation of the HGF/c-Met axis promotes cell migration, which contribute to the
metastatic characteristics of malignant cells [23]. The in vitro wound-healing assay (WHA) is a

ACCEPTED MANUSCRIPT
simple method to study directional cell migration in two dimensions [23, 38]. Thus, compounds
1
-47 were evaluated for their antimigratory effect in the scratch WHA using the highly
metastatic MDA-MB-231 cells. HGF (40 ng/mL) was used as a mitogen to induce cell
migration via c-Met activation. The parent 4 and the natural phenolics 2 and 3 showed moderate
to high ꢁM antimigratory activity (Table 2). Generally, all semisynthetic analogs excluding 5, 7,
8
, 10 and 34 showed better antimigratory activity than their parent 4 (Table 2). Compared to
their effect on cell proliferation, nearly all compounds were more potent as MDA-MB-231 cell
migration inhibitors, with most analogs showing IC values <15 µM (Table 2). Compounds 11,
5
0
1
3, 15-17, 20, 21, 33, 35, 37, 39, 44 and 45 showed the most promising antimigratory activity
with IC values <5 µM (Table 2). Figure 3A shows the effect of the most active analogs HVS-
5
0
1
6, 44 and 45 on cell migration across the wound inflicted in the MDA-MB-231 cell monolayer
compared to the vehicle control as well as 10 µM dose of 1 and 20 µM dose of 4 as positive
controls [19]. Apparently, HVS-16 was the most active compound, significantly suppressing
HGF-induced cell migration in a dose-dependent manner with an IC50 value of 2.5 µM,
demonstrating thirteen-fold enhancement in the activity, compared to 4 (IC50 33.5µM) and
three-fold improvement versus 1 (IC 7.5 µM) (Table 2, Figure 3). It is interesting to note that
5
0
HVS-16 was the most active c-Met phosphorylation inhibitor in the Z-LYTE™ assay,
suggesting a strong correlation between the results of both assays. Similarly, analogs 17, 33, 44
and 45 were able to inhibit cell migration with IC50 values of 4.6, 4.1, 3.6 and 3.9 µM,
respectively, parallel to their c-Met kinase inhibitory activities (Table 2, Figure 3). However,
there was an exception related to the C-5' substitution at the phenyl ring A with HBD
substituents in analogs 13, 15, 28 and 37. Compounds 13, 15 and 37, with C-5' hydroxyl
substituent, demonstrated comparable antimigratory activities to 17 and 33, with IC values of
5
0
4
.2, 4.5 and 4.6 µM, respectively, while being five to seven-fold less active than 17 and 33

ACCEPTED MANUSCRIPT
against c-Met in the kinase assay (Tables 1 and 2). Additionally, analog 28, with C-5' amino
substituent, was still able to show promising antimigratory activity, with an IC value of 5.2
5
0
µM, despite being nine-fold less active than 17 in the Z-LYTE™ assay (Tables 1 and 2).This
illustrates the important role of the HBD substituents at C-5' position for optimal antimigratory
activity, but only marginally through inhibition of c-Met activation. In a similar fashion,
analogs 20, 21 and 35 were almost equipotent to 33 in the WHA, with IC values of 4.8, 4.1
5
0
and 4.3 µM, respectively, while being much less active (~8-fold) than 33 in the kinase assay
Tables 1 and 2). Therefore, analogous to cell proliferation, the chain extension in these analogs
(
maintained the antimigratory activity, in spite of the significant drop in c-Met phosphorylation
inhibition, compared to 33. A significant difference in the antimigratory activity between 33
2
,3
versus 34 was observed upon changing the acid part’s ꢂ geometry (E/Z isomers), consistent
with the results of the Z-LYTE™ assay (Tables 1 and 2). However, changing the geometry in
analogs with 7-atom linker between the phenyl rings A and B (35 versus 36) did not exhibit the
same drop in antimigratory activity similar to that observed in the 6-atom linker analogs (33
versus 34) (Table 2). This clearly augmented our earlier observation that different molecular
targets, other than c-Met, are dictating the migratory potential of these semisynthetic esters with
7
-atom linker. In contrast, the 5-atom linker analogs 39 and 41 were 2-fold less active than
HVS-16 in WHA, with IC values of 4.0 and 5.4 µM, respectively, parallel to their c-Met
5
0
kinase inhibitory activities (Table 2). Surprisingly, the imine 11 and the 1,2,4-oxadiazole analog
2 were active in this cell migration assay, with IC values of 4.6 and 16.3 µM, respectively,
1
5
0
whereas were nearly inactive in both Z-LYTE™ and cell proliferation assays (Tables 1 and 2).
Compounds 11 and 12 appeared to enhance only the antimigratory activity, compared to the
parent analog 4, by offering better binding affinity to migration targets, other than c-Met, and
hence, improving the overall activity. Meanwhile, analogs 14, 22-27, 31 and 40 demonstrated

ACCEPTED MANUSCRIPT
promising antimigratory activities, with almost no corresponding c-Met inhibitory potential in
the kinase assay, suggesting other molecular target(s) for their antimigratory activity (Table 2).
2
.2.4. HGF-induced cell invasion assay
During the complicated process of cancer metastasis, the invasion of cancer cells is
among the most important and serious steps. The new HVS-16, 21 and 39, in addition to the
parent 4, were further evaluated for their ability to inhibit the invasiveness of the aggressive
®
MDA-MB-231 cells using the Cultrex BME cell invasion assay (Table 3, Figure 4) [38, 39].
Oleocanthal (1) was used as a positive control and its IC50 was calculated for activity
comparison (Table 3). HGF (40 ng/mL) was used as a mitogen to induce cell invasion via c-Met
activation. Compound selection for testing in this assay was based on the overall performance in
previous assays. HVS-16 was the most active in Z-LYTE™ as well as cell proliferation and
migration assays while analogs 21 and 39 were among the most potent in cell proliferation and
migration assays (Tables 1 and 2). The three analogs also demonstrated varying activity levels
of c-Met inhibition in the cell-free assay, so that a strong correlation between their anti-invasive
and c-Met inhibitory activities can be established (Table 1). Four different subtoxic
concentrations (1, 5, 10 and 20 µM) per each analog were tested and used to determine IC50
values (Table 3, Figure 4).The parent 4 showed moderate ꢁM anti-invasive activity, with an
IC50 value of 19.8 µM (Table 3). Generally, the three tested analogs HVS-16, 21, and 39
demonstrated better anti-invasive activity profile than their parents 1 and 4 (Table 3, Figure 4).
This clearly highlighted the significance of introducing small hydrophobic methoxy
substituents, retaining the HBA property, at the aromatic ring A in HVS-16 and 39, compared
to 4, and the impact of optimizing the distance limit between the aromatic rings A and B, as
represented by 21 and 39. HVS-16 was the most active hit, significantly and dose-dependently

ACCEPTED MANUSCRIPT
decreasing the level of HGF-mediated cell invasion, with an IC value of 2.7 µM. HVS-16
5
0
demonstrated seven-fold enhancement in anti-invasive activity, compared to 1 and 4, parallel to
its c-Met inhibitory activity (Table 3, Figure 4). Analog 39 was 3-fold less active than HVS-16
in cell invasion assay, with an IC value of 7.1 µM, consistent with their activity levels in the
5
0
c-Met kinase assay (Table 3, Figure 4). Similarly, 21 exhibited a slight improvement in the anti-
invasive activity compared to 4, with an IC value of 14.2 µM, perfectly matching its c-Met
5
0
inhibitory activity (Table 3, Figure 4). There is an obvious strong correlation between the
results of both Z-LYTE™ and cell invasion assays, reinforcing the hypothesis that c-Met is the
primary target for the anti-invasive effects of these tested analogs.
2
.2.5. HGF-induced 3D spheroid disaggregation model
It is well-known that cultures grown as 3D spheroids more accurately mimic the
natural tumor microenvironment and thus, more closely recapitulate the in vivo response to
drugs, compared to the traditional 2D monolayer cultures [40]. Therefore, the aim of this study
was to characterize the effects of HVS-16, 39, and 41 on an implemented 3D spheroid culture
of the breast cancer cell line MDA-MB-231, stimulated with HGF, in an attempt to assess its in
vitro sensitivity to these analogs versus tumor cells grown as 2D monolayer cultures, which
might give a better prediction of the in vivo profile of this class of compounds. After spheroid
formation, the drug responsiveness was evaluated by measuring the extent of spheroid growth
in response to either DMSO as vehicle control or increasing concentrations of HVS-16 (1, 3 and
1
0 µM) or a single dose of either 39 or 41 (10 µM), for a treatment period of 72 h. Additionally,
has been tested at 20 ꢁM for activity comparison. Images were captured every 4 h using the
1
Incucyte real time imaging platform (Figures SI63 and SI64). The three tested analogs
demonstrated much better activity than 1, consistent with their activity pattern in MDA-MB-231

ACCEPTED MANUSCRIPT
2
D monolayer cultures (Figure 5). HVS-16 significantly reduced HGF-induced spheroid growth
in a dose dependent fashion, with 80% inhibition at 10 µM dose (Figure 5). Interestingly, 39
and 41 were also able to reduce spheroid growth by about 80% at 10 µM, parallel to their
antiproliferative activities in MTT assay, despite being 2-fold less active than HVS-16 in the Z-
LYTE™ kinase assay, suggesting other molecular target(s), in addition to c-Met, for their
activity. The results of this assay clearly demonstrated the ability of tested analogs to block
HGF-driven growth not only in 2D but also in 3D cultures of the human triple negative breast
cancer (TNBC), which may in turn suggest a promising in vivo profile for these analogs.
2
.2.6. Selective cytotoxic activity evaluation
The selective cytotoxicity of the new olive secoiridoid bioisosteres was assessed in the
MTT assay using the immortalized non-tumorigenic human mammary epithelial cell line
MCF10A (Figure 6, Table SI23). The analogs HVS-16, 20, 21, 24-26, 28, 37, 39, 41, 44 and 45
showing the most promising activity in MTT proliferation assay, with IC values <10 µM,
50
were selected for testing in this assay at different concentrations per each analog and the IC50
values were determined for proper selectivity assessment (Table SI23). All tested compounds
were non-toxic to MCF10A cells, compared to the vehicle-treated control group, at
concentrations several-fold higher than their IC values in cell proliferation assay, suggesting
5
0
their good selectivity towards malignant cells (Figure 6, Table SI23). For instance, HVS-16, the
most active c-Met phosphorylation inhibitor in the Z-LYTE™ assay, was shown to have no
remarkable effect on the viability and growth of non-tumorigenic MCF10A mammary epithelial
cells up to 80 µM in culture (IC = 226.4 µM, Table SI23), while it exhibited IC values of 3.8
5
0
50
and 2.5 µM in proliferation and migration assays, respectively, against the MDA-MB-231 cells
Figure 6, Table 2). These results suggested the excellent selectivity of HVS-16 towards breast
(

ACCEPTED MANUSCRIPT
cancerous cells, which renders this analog a potential candidate for the control of invasive
breast cancer.
2
.2.7. Biological characterization of HVS-16
The enzymatic and cellular results suggested the potential of homovanillyl sinapate
(HVS-16, Scheme 2) as the most potent synthesized c-Met inhibitory hit (Table 1, Figure 1).
HVS-16 also showed the most potent anticancer activity in different cell-based assays, without
discernable effects on the non-tumorigenic human mammary epithelial cells (Figures 3-6).
Taken together, HVS-16 was considered appropriate for further validation of its c-Met
inhibitory activity at the molecular level. In order to evaluate its off-target effects, HVS-16 was
tested against the c-Met-independent T-47D human breast cancer cells. This particular cell line
represents a completely different breast cancer phenotype, compared to the TNBC phenotype
used throughout the study. ERα, one of the most important targets in human breast cancer
therapy, is expressed in T-47D cells, whereas the TNBC cells lack the expression of ERα due to
epigenetic silencing [41]. On the other hand, c-Met is overexpressed in the TNBC MDA-MB-
2
31 and MDA-MB-468 cells, while it is absent in T-47D cells [42].
.2.7.1. Effect of HVS-16 on c-Met phosphorylation
2
The human TNBC MDA-MB-231 cell line was chosen to assess the effect of HVS-16
on HGF-induced c-Met phosphorylation using Western blot analysis, to further validate the
initial biochemical data. Phospho-c-Met refers to the phosphorylation of the kinase domain at
Y1234/1235. Cells were exposed to different doses (1, 5 and 10 µM) of HVS-16 and the
expression and phosphorylation levels of c-Met protein in these cell lysates were then
determined (Figure 7A). Results demonstrated significant dose-dependent inhibition of HGF-
induced c-Met phosphorylation after treatment with HVS-16 for 72 h, consistent with its

ACCEPTED MANUSCRIPT
observed cell-free activity and matching the antiproliferative effects (Figure 7). Meanwhile,
HVS-16 treatment did not affect the total c-Met levels at the tested concentrations in vitro
(Figure 7A).
2
.2.7.2. Evaluation of HVS-16 off-target effects
The growth inhibitory effect of HVS-16 on human breast cancer cells with different c-
Met status was assessed using MTT assay. The antiproliferative effects of various doses of
HVS-16 on the HGF-mediated growth of the Met-dependent MDA-MB-231, MDA-MB-468,
and the Met-independent T-47D breast cancer cell lines, after 72 h culture period are shown in
Figure 7B. Treatment with HVS-16 caused a dose-dependent suppression of HGF-induced
proliferation of human breast cancer cells from lines MDA-MB-231 and MDA-MB-468,
compared to their vehicle-treated control groups (Figure 7B). Interestingly, larger
concentrations of HVS-16 were required to significantly abolish the cell viability of both cell
lines when grown in HGF-free media after 72 h (data not shown). The IC values for HVS-16
5
0
treatment in HGF-supplemented media were 3.8 and 6 µM in MDA-MB-231 and MDA-MB-
68 breast cancer cells, respectively (Figure 7B). However, the IC50 values for HVS-16
4
treatment in HGF-free media were >40 µM for both cell lines (data not shown). This finding
indicated that HVS-16 treatment is only dependent on the presence of HGF, confirming its
proposed molecular mechanism as a direct inhibitor of the HGF/c-Met signaling. Importantly,
HVS-16 lacked activity against the Met-independent T-47D breast cancer cells, being several-
fold less active in this cell line compared to the c-Met expressing breast cancer cells (IC =
5
0
1
09.8 µM, Figures 7B and SI65). This result confirmed that the differential sensitivity of the
human breast cancer cells to HVS-16 is linked to its inhibitory effects on c-Met signaling.

ACCEPTED MANUSCRIPT
The in vitro assays clearly indicated that HVS-16 significantly and specifically inhibited
HGF-dependent cell growth across c-Met-expressing TNBC cell lines. TNBC is characterized
by a lack of the expression of the ER, progesterone receptor and HER-2 [43]. It accounts for
about 17% of all breast cancers, representing an aggressive clinical behavior and is generally
associated with poor prognosis; thus chemotherapy remains the only systemic treatment option
available for these patients [43]. Accordingly, there is a need to develop new treatments for this
aggressive subtype, which currently lack targeted therapy. The inhibition of mammary cancer
cell growth was associated with the ability of HVS-16 to disrupt c-Met receptor activation in
response to HGF in MDA-MB-231 cells. Collectively, these data support the fact that c-Met
inhibition can be the primary therapeutic target for the antitumor effects of HVS-16 in vitro.
2
.3. HVS-16 Binding Mode at the c-Met Kinase Domain
The c-Met’s ATP binding site includes: (1) Hinge region: Met1160 and Pro1158.
Interaction at this site is highly characteristic for all compounds targeting the ATP binding site
in kinase domains; (2) Central hydrophobic region; (3) Two smaller hydrophobic subpockets;
and (4) c-Met activation loop (Asp1222-Lys1245) [44]. The central hydrophobic region is
usually defined by three important amino acids Lys1110, Val1092, and Met1211. The first
hydrophobic subpocket is aligned by Ile1084, Ala1108, and Leu1157 while the second
hydrophobic subpocket is defined by Phe1089. There are basically two classes of c-Met
inhibitors, ATP competitive and ATP non-competitive inhibitors. About 20 c-Met crystal
structures have been disclosed to date, whether in complex with a ligand or not, revealing two
distinct binding modes for the ATP competitive inhibitors [2]. The ATP competitive inhibitors
are further divided into two classes; class I inhibitors, which are relatively selective and assume
an approximate U-shape geometry within the c-Met’s ATP-binding site through interactions

ACCEPTED MANUSCRIPT
with the hinge region’s Met1160 and the activation loop residue Tyr1230, while class II
inhibitors are less selective, adopting more extended orientation [44]. Overall, a good binding
affinity for any c-Met inhibitor hit is hypothesized to be through interactions with at least one of
Asp1222, Phe1223, or Tyr1230 at the activation loop and either Pro1158 or Met1160 at the
hinge region, along with a good fitting within one or more of the aforementioned hydrophobic
regions [2].
A detailed in silico docking analysis was performed to further elucidate the binding
mode of HVS-16 at the c-Met’s ATP binding pocket. Docking studies have been carried out
separately on different c-Met crystal structures to minimize false positive results due to
conformational variations. Four highly resolved c-Met kinase domain crystal structures (PDB
codes: 3CE3, 3F82, 3U6I, and 4XYF) at a resolution ≥ 2 Aº were used to investigate HVS-16
possible binding modes within the catalytic domain of non-phosphorylated c-Met using
Schrödinger software (Figures 8, SI66-67 and SI71-72). In some crystal structures, HVS-16
assumed a shallow U-shaped binding mode at the c-Met kinase domain with partial wrapping
around Met1211, which was the typical conformation of class I c-Met inhibitors [2]. HVS-16
adopted an extended conformation similar to class II c-Met inhibitors in other crystal structures.
Adopting either conformation could be due to different shape and size of the binding cavity
existing in various c-Met crystal structures after the originally co-crystallized ligand is removed.
The visualization of the docked pose of HVS-16 emphasized its complete shape fitting
at the ATP binding pocket of c-Met kinase domain (Figures 8E and SI72). The two phenolic
hydroxyl groups on the aromatic rings A and B in HVS-16 were virtually proposed to be the
main binding and anchoring pharmacophoric groups at the c-Met kinase domain (Figures 8B,
8
C, SI66 and SI67). Detailed examination of HVS-16 binding pose revealed that the tyrosol’s

ACCEPTED MANUSCRIPT
C-6' phenolic hydroxyl group is either participating in a critical single HB with the backbone of
Met1160 or forming a pair of bidentate HB interactions with the backbone of both Met1160 and
Pro1158 in the hinge region of the four studied c-Met crystal structures (Figures 8B, 8C, SI66
and SI67). Alternatively, the C-6'-O-methyl ether analogs 24 and 26, which lack free C-6'
hydroxyl group, failed to satisfy such critical HB interactions within the hinge region and
subsequently showed poor activity in the c-Met biochemical assay (Table 1). This docking
result clearly corroborated with the SAR data, which suggested the need for at least one HBD
group at the C-6' position. In addition, the aromatic C-7 phenolic hydroxyl group of the sinapic
acid moiety formed a strong HB interaction either with the backbone carbonyl oxygen of
Glu1127 in crystal structures 3F82 and 3CE3 (Figures 8B and SI67), or with Arg1208 in crystal
structure 4XYF (Figure 8C), thus hindering the participation of these residues in catalysis. This
HB appeared to impart a substantial portion of the binding affinity, as proven by the activity
lack of the previously reported C-7-O-methyl ether analogs [19]. The C-7 amino group in
analog 33 appeared to maintain this critical HB interaction with the kinase domain, and thus
retained the activity in the c-Met biochemical assay (Table 1). The docking results supported
the SAR data, which suggested the need for a HBD group at the C-7 position. However, there
was an exception related to the lack of this C-7 HB interaction to the phenolic hydroxyl group
of HVS-16 in crystal structure 3U6I (Figure SI66). Instead, the aromatic C-6 methoxy oxygen
of sinapic acid was uniquely engaged in a HB interaction with the backbone amide hydrogen of
Asp1222 at the activation loop’s DFG motif of this crystal structure (Figure SI66). Additionally,
the same C-6 methoxy group was buried within a hydrophobic pocket lined by Phe1223 (DFG
motif), Met1131, Val1155 and Ala1221 in both crystal structures 3U6I and 4XYF (Figures SI66
and 8C). Despite C-6 methoxy group lacked HB interactions in crystal structures 3F82 and
3
CE3, it was able to occupy a hydrophobic pocket defined by Ala1127, Ile1145, Val1155, and

ACCEPTED MANUSCRIPT
Phe1124 (Figures 8B and SI67). Ile1145 demarcates the back of this pocket, which referred to
as the Ile1145 pocket or the C-helix pocket [45]. This pocket has only been reported in a subset
of kinases and is present due to the uncharacteristic position of the C-helix, which is further
away from the ATP binding site than it is in most kinases [36, 44]. The aromatic C-8 methoxy
group of sinapic acid was solvent exposed in all c-Met crystal structures and therefore should
not have a significant impact on c-Met inhibition. These interactions underscore the importance
of at least one methoxy group in the sinapic acid part of HVS-16 for binding and subsequent
kinase inhibitory capacity, which is consistent with the generated SAR using the cell-free assay.
Interestingly, these observations could explain the ability of analog 17 to retain the c-Met
inhibitory activity, compared to HVS-16, as well as the significant reduction in the potency of
6
, lacking both methoxy groups, versus 4 in the kinase assay (Table 1). In the lipophilic back
pocket of the activation loop, the sinapic acid’s aromatic ring B is sandwiched between
Met1211 and Tyr1230, forming a strong π-π stacking interaction with Tyr1230, thus hindering
its autophosphorylation necessary for c-Met activation and stabilizing the inhibitory
conformation of the activation loop (PDB 4XYF, Figure 8C). The E-configuration of the
sinapate moiety is presumably essential to maintain the π-π stacking interaction of ring B with
Tyr1230 as well as the HB interaction of the C-7 phenolic OH and this might be the reason that
the E-configuration was critical for the c-Met inhibition in the cell-free assay. The terminal 2’-
phenethyl part in HVS-16 occupied a deep hydrophobic pocket defined by Tyr1159, Leu1157,
Ala1108, Val1092 and Leu1140 in all c-Met crystal structures (Figure 8). The lower c-Met
inhibitory activity of the benzyl derivatives 39 and 41 and the 3’-phenylpropyl analogs 20 and
2
1 in the cell-free assay, relative to HVS-16, suggested that the phenethyl moiety might be
optimal for fitting into this hydrophobic pocket.

ACCEPTED MANUSCRIPT
The key modification that led to improved potency against c-Met involved substitution
of the C-5' of 4’s ring A with a methoxy group in HVS-16. This change resulted in 14-fold
activity increase in the kinase assay (Table 1). Docking studies further justified this activity
enhancement by showing the C-5' methoxy group, despite lacking HB interactions, is buried in
the binding pocket and exerting hydrophobic interactions with the side chains of Ile1084,
Phe1089, Val1092, and Leu1140 at the c-Met kinase hydrophobic sub-pockets (Figures 8B, 8C,
SI66 and SI67). Typically, it resides between Ile1084 and the residue at the tip of the
phosphate-binding loop (P-loop), Phe1089. These hydrophobic interactions might explain, at
least in part, the significant activity improvement impact of this methoxy group in HVS-16,
compared to the unsubstituted analog 4, as well as analogs 13 and 15 with hydroxyl group at
this position (Table 1). However, the c-Met inhibitory activity of HVS-16 is still in the low
micromolar level, perhaps because the methoxy group is not sufficiently bulky enough to fill the
hydrophobic pocket or lacks the potential to engage in π-electron interactions with the aromatic
residues comprising the pocket. The ester moiety appeared not to contribute any binding role,
yet it provided the proper molecular size and U-shaped conformation alignment necessary for
adequate fitting at the binding pocket (Figures 8, SI66 and SI67). Although the six-atom linker
tethered the two aromatic rings A and B did not have direct binding role, it ultimately played a
crucial role in properly aligning HVS-16 binding pharmacophores at the c-Met kinase domain
and thus, maintained the critical HB interactions with both the hinge region and the activation
loop (Figures 8, SI66 and SI67).
To validate the docking results, the original ligand of each c-Met crystal structure used
was docked into its ATP binding pocket by using the same parameters which have been used
for docking HVS-16 (Figures 8D, SI68-70). The bound conformation of the co-crystal ligand
was generated with a good root mean square displacement (RMSD) of 0.2 Å, showing the

ACCEPTED MANUSCRIPT
robustness of the docking protocols. Figure 8D shows the docking model of HVS-16 in the co-
crystal structure of 49 bound to the c-Met kinase (PDB code: 3F82). HVS-16 almost overlaid
the original co-crystallized ligand 49, demonstrating the same critical interaction with the hinge
region’s Met1160 [46]. However, 49 engaged in two HB interactions with the backbone amide
hydrogen of Asp1222 and the side chain of Lys1110 while HVS-16 participated in a single HB
interaction with Glu1127 (Figure 8D). Despite most parts of the two molecules overplayed
perfectly, the terminal 4-fluoro-phenyl ring in 49 extended to occupy a deep hydrophobic
pocket defined by Phe1134, Leu1195, and Phe1200 unlike the aromatic C-8 sinapate methoxy
group in HVS-16, which adopted a different conformation and missed this pocket (Figure 8D).
These observations might explain, at least in part, the moderate activity level of HVS-16 in the
c-Met biochemical assay compared to 49 [46]. The docking studies suggested the good fitting of
HVS-16 in the c-Met kinase. The binding mode of HVS-16 within the catalytic c-Met domain
is consistent with the kinase assay data, suggesting the potential of HVS-16 as a promising c-
Met inhibitory hit.
3
. Conclusion
The present study validated the discovery of HVS-16 as a novel small-molecule c-Met
inhibitory hit inspired by the olive secoiridoid oleocanthal with therapeutic potential for the
control of c-Met-dependent breast cancer. The structural simplicity and synthetic feasibility of
the olive secoiridoid analogs render them appropriate candidates for future preclinical
optimization. The in vitro cellular potency of HVS-16 in 2D and 3D breast cancer cultures was
well-correlated with its cell-free c-Met phosphorylation inhibitory activity coupled with high
selectivity to cancerous cells. The novel structure of HVS-16 provides new opportunities to
chemically improve the binding affinity, as shown by docking studies. Collectively, the present