A small molecule inhibitor of PCSK9 that antagonizes LDL receptor binding via interaction with a cryptic PCSK9 binding groove

Article abstract of DOI:10.1016/j.bmc.2020.115344

Proprotein convertase (PC) subtilisin kexin type 9 (PCSK9) inhibits the clearance of low density lipoprotein (LDL) cholesterol from plasma by directly interacting with the LDL receptor (LDLR). As the interaction promotes elevated plasma LDL cholesterol levels and a predisposition to cardiovascular disease (CVD), it has attracted much interest as a therapeutic target. While anti-PCSK9 monoclonal antibodies have been successful in the treatment of hypercholesteremia by decreasing CVD risk, their high cost and a requirement for injection have prohibited widespread use. The advent of an orally bioavailable small molecule inhibitor of the PCSK9-LDLR interaction is an attractive alternative, however efforts have been tempered as the binding interface is unfavourable for binding by small organic molecules. Despite its challenging nature, we report herein the discovery of compound 3f as a small molecule inhibitor of PCSK9. The kinase inhibitor nilotinib emerged from a computational screen that was applied to identify compounds that may bind to a cryptic groove within PCSK9 and proximal to the LDLR-binding interface. A subsequent in vitro PCSK9-LDLR binding assay established that nilotinib was a bona fide but modest inhibitor of the interaction (IC₅₀ = 9.8 μM). Through multiple rounds of medicinal chemistry, 3f emerged as a lead-like molecule by demonstrating disruption of the PCSK9-LDLR interaction at nanomolar levels in vitro (IC₅₀ = 537 nM) with no inhibitory activity (IC₅₀ > 10 μM) against a small panel of kinases. Compound 3f restored LDL uptake by liver cells at sub-micromolar levels and demonstrated excellent bioavailability when delivered subcutaneously in mice. Most significantly, compound 3f lowered total cholesterol levels in the plasma of wild-type mice, thereby providing proof-of-concept that the notion of a small molecule inhibitor against PCSK9 is therapeutically viable.

Full text of DOI:10.1016/j.bmc.2020.115344

Bioorganic&MedicinalChemistry28(2020)115344
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
A small molecule inhibitor of PCSK9 that antagonizes LDL receptor binding
via interaction with a cryptic PCSK9 binding groove
⁎
Benny J. Evisona, , James T. Palmerb, Gilles Lambertc, Herbert Treutleind, Jun Zenge,
Brice Nativelc, Kévin Chemelloc, Qing Zhuf, Jie Wangf, Yanfen Tengf, Wei Tangf, Yanfeng Xuf,
Anuj Kumar Rathig, Sanjay Kumarg, Alexandra K. Suchowerskaa, Jasneet Parmara, Ian Dixonh,
Graham E. Kellya, James Bonnara
a Nyrada Inc., 828 Pacific Highway, Gordon, New South Wales 2072, Australia
b Pharmaceutical Discovery Consultation, 143-145 Flannery Court, Warrandyte, Victoria 3113, Australia
c Laboratoire Inserm UMR 1188 DéTROI, Université de la Réunion Plateforme CYROI, 2 Rue Maxime Rivière, 97490 Sainte Clotilde, France
d Sanoosa Pty. Ltd., Level 30, 35 Collins St, Melbourne, Victoria 3000, Australia
e MedChemSoft Solutions, 3 Beech Close, Ferntree Gully, Victoria 3156, Australia
f ChemPartner, No. 5 Building, 998 Halei Rd, Zhangjiang Hi-Tech Park, Pudong New Area, Shanghai, China
g Jubilant Chemsys Ltd., B-34, Sector-58, Noida 201301, India
h Altnia Group, 13 Fuchsia St, Blackburn, Victoria 3130, Australia
A R T I C L E I N F O
A B S T R A C T
Keywords:
Proprotein convertase (PC) subtilisin kexin type 9 (PCSK9) inhibits the clearance of low density lipoprotein
(LDL) cholesterol from plasma by directly interacting with the LDL receptor (LDLR). As the interaction
promotes elevated plasma LDL cholesterol levels and a predisposition to cardiovascular disease (CVD), it has
attracted much interest as a therapeutic target. While anti-PCSK9 monoclonal antibodies have been successful
in the treatment of hypercholesteremia by decreasing CVD risk, their high cost and a requirement for in-
jection have prohibited widespread use. The advent of an orally bioavailable small molecule inhibitor of the
PCSK9-LDLR interaction is an attractive alternative, however efforts have been tempered as the binding
interface is unfavourable for binding by small organic molecules. Despite its challenging nature, we report
herein the discovery of compound 3f as a small molecule inhibitor of PCSK9. The kinase inhibitor nilotinib
emerged from a computational screen that was applied to identify compounds that may bind to a cryptic
groove within PCSK9 and proximal to the LDLR-binding interface. A subsequent in vitro PCSK9-LDLR binding
assay established that nilotinib was a bona fide but modest inhibitor of the interaction (IC50 = 9.8 µM).
Through multiple rounds of medicinal chemistry, 3f emerged as a lead-like molecule by demonstrating dis-
ruption of the PCSK9-LDLR interaction at nanomolar levels in vitro (IC50 = 537 nM) with no inhibitory
activity (IC50 > 10 µM) against a small panel of kinases. Compound 3f restored LDL uptake by liver cells at
sub-micromolar levels and demonstrated excellent bioavailability when delivered subcutaneously in mice.
Most significantly, compound 3f lowered total cholesterol levels in the plasma of wild-type mice, thereby
providing proof-of-concept that the notion of a small molecule inhibitor against PCSK9 is therapeutically
viable.
Proprotein convertase (PC) subtilisin kexin
type 9 (PCSK9)
Small molecule
Cardiovascular disease
Low density lipoprotein (LDL)
LDL receptor
LDL cholesterol
⁎ Corresponding author.
E-mail address: benny.evison@nyrada.com (B.J. Evison).
https://doi.org/10.1016/j.bmc.2020.115344
Received 23 August 2019; Received in revised form 17 January 2020; Accepted 23 January 2020
Availableonline31January2020
0968-0896/©2020TheAuthor(s).PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
~530 Å2 that is largely flat, featureless and devoid of binding pockets
(Fig. 1B), making it unfavourable for binding by small organic mole-
cules.10–13 Characteristics such as these are more conducive to targeting
via antibodies raised against the protein. Indeed, much effort led to the
development of anti-PCSK9 monoclonal antibodies that have demon-
strated excellent LDL-c lowering activity in human clinical trials when
used in combination with statin-based therapies.9,14,15 Most im-
1. Introduction
Proprotein convertase (PC)1 subtilisin kexin type 9 (PCSK9) is a
member of the PC family of proteins, a group of nine serine proteases
that have been characterised in humans and other organisms.1 The
protein is first synthesised as a zymogen (Fig. 1A) in the ER of cells
mainly in the liver, kidney and intestines.2,3 Following translation, the
signal peptide (residues 1–30) is first excised from the PCSK9 poly-
peptide which then undergoes an autocatalytic event that cleaves the
bond between Gln 152 and Ser 153 of the protein (Fig. 1A and
Supplementary Fig. S1).1,3–5 It is this second cleavage event that yields
the mature protein which consists of a prodomain (residues 30–152),
catalytic domain (residues 153–451) and a C-terminal domain (residues
452–692, Fig. 1 and Supplementary Fig. S1).2–4 Despite its covalent
excision from the protein, the prodomain remains non-covalently as-
sociated with the catalytic domain through hydrogen bonding
(Fig. 1B).2 Importantly, this interaction renders the mature protein
catalytically inactive as the prodomain sterically occludes substrate
entry into the enzyme’s active site (Fig. 1B).2,3,5 Once cleavage is
achieved, PCSK9 is directed through the trans-Golgi network to the
plasma membrane and secreted into the extracellular space.1
portantly, the therapies significantly reduced the risk of CVD9,15
,
thereby validating the inhibition of PCSK9 as therapeutically relevant.
Despite their clinical success, the application of monoclonal anti-
bodies as lipid-lowering therapies has its disadvantages. Monoclonal
antibodies are typically very expensive to purchase and must be ad-
ministered via injection on a regular basis.9,16,17 Moreover, there are
issues with shelf-life and their propensity to elicit immunogenic re-
sponses.17 Various alternative strategies have been devised to address at
least some of these limitations. The application of small interfering
RNAs18
,
locked antisense nucleic acids19 and CRISPR-Cas9-based
technologies20,21 to downregulate PCSK9 protein expression are at
various stages of development, yet these have inherent disadvantages of
their own.
An attractive alternative would be the advent of an orally bioa-
vailable small molecule inhibitor of PCSK9. A small molecular weight
agent directed against PCSK9 is highly sought after as a therapy for
CVD, especially given the low cost and ease of administration asso-
ciated with such agents.17 A handful of small molecule programs di-
rected at the inhibition of PCSK9 have been described. Pfizer has re-
The functional characterisation of PCSK9 emerged from studies in-
volving its original identification. Abifadel et al.6 reported that humans
carrying gain-of-function mutations within the PCSK9 gene were sus-
ceptible to increased levels of low-density lipoprotein cholesterol (LDL-
c), a predominant feature of cardiovascular disease (CVD). Soon after, it
was demonstrated that the converse was also true; individuals bearing
loss-of function mutations in the gene encoding PCSK9 typically dis-
played reduced plasma LDL-c7 and experienced fewer CVD events.8
Functional studies in the ensuing years established that PCSK9 was a
key regulator of LDL metabolism; following its secretion from liver cells
into blood plasma9, PCSK9 engages the EGF(A) domain of the LDL re-
ceptor (LDLR) expressed at the plasma membrane of liver cells.1,10,11
The LDLR-PCSK9 interaction then promotes the internalisation and
subsequent degradation of LDLR within the hepatocyte lysosomal
compartment, thereby reducing hepatocyte LDLR expression at the cell
surface.2,10,11 The pathway culminates in a net decrease in the clear-
ance of plasma LDL-c by hepatocytes and elevated LDL-c plasma le-
vels.10 It was quickly realised that disruption of the LDLR-PCSK9 in-
teraction may be harnessed to therapeutically lower plasma LDL-c and
the risk of CVD in humans.
ported the identification of
a
small molecule PCSK9 anti-
secretagogue16,22,23, however this program has not progressed into the
clinic. Several groups including Genentech have identified small pep-
tides that bind directly to PCSK9 and disrupt the protein’s interaction
with LDLR.10–12 Others including Portola, Shifa and Cadila Healthcare
have not reported any small molecules beyond the discovery stage24
and very few have been reported in the primary literature in detail.17,25
Despite the challenging nature of targeting PCSK9 with small molecular
weight agents, we report herein the identification of a small molecule
inhibitor of PCSK9 that disrupts the PCSK9-LDLR interaction at nano-
molar levels in vitro, restores LDL uptake by liver cells cultured in the
presence of PCSK9 and lowers total cholesterol levels in the plasma of
wild-type mice.
2. Materials and methods
The EGF(A)-interacting site of PCSK9 is a solvent-exposed area of
2.1. Materials
1 4A MS, molecular sieve 4A; Abl, Abelson tyrosine kinase; AMB657286, 4-[4-
(benzyloxy)-3-methylbenzoyl]-3-hydroxy-1-[3-(4-morpholinyl)propyl]-5-(4-
pyridinyl)-1,5-dihydro-2H-pyrrol-2-one; BBFO, broad band fluorine observe;
BBO, broad band observe; Bcr-Abl, breakpoint cluster region Abelson tyrosine
kinase fusion protein; BOC, tert-butyloxycarbonyl; Bodipy FL LDL, boron-di-
pyrromethene labelled low density lipoprotein (LDL) from human plasma; BPO,
benzoyl peroxide; c-Kit, tyrosine protein kinase Kit; CRISPR, clustered regularly
interspaced short palindromic repeats; CVD, cardiovascular disease; CYP450,
cytochrome P450; DCM, dichloromethane; DIBAL-H, diisobutylaluminium hy-
dride; DIEA or DIPEA, N,N-diisopropylethylamine; DMA or DMAC, dimethyla-
cetamide; DMEDA, 1,2-dimethylethylenediamine; DMF, dimethylformamide;
DMSO, dimethyl sulfoxide; EGF(A), epidermal growth factor (A); ELISA,
Enzyme-linked immunosorbent assay; ER, endoplasmic reticulum; ESI, elec-
trospray ionisation; FAM, fluorescein amidite; HATU, O-(7-azabenzotriazol-1-
yl)-N,N,N',N'-tetramethyluronium hexafluorophosphate; HepG2, human hepa-
tocarcinoma cell line; IV, intravenous; LAH; Lithium aluminium hydride; LDL,
low density lipoprotein; LDL-c, low density lipoprotein (LDL) cholesterol; LDLR,
low density lipoprotein (LDL) receptor; NBS, N-bromosuccinimide; NMP, N-
methyl-2-pyrrolidone; PC, proprotein convertase; PCSK9, Proprotein convertase
(PC) subtilisin kexin type 9; PO, oral; PDGFRβ; platelet derived growth factor
receptor β; SAR, structure activity relationship; SC, subcutaneously; Src, Src
tyrosine kinase; Solutol HS15, poly-oxyethylene esters of 12-hydroxystearic
acid; TFA, trifluoracetic acid; THF, tetrahydrofuran; TOTP, Tri-o-tolyl phos-
phate
DMSO, DMAC, Ficoll Paque Plus, BSA and ATP were from Sigma (St
Louis, MO, USA and Saint Quentin-Fallavier, France). Niltonib was
purchased from TRC Inc (Toronto, Canada) while AMB-657286 was
from Ambinter (Greenpharma SAS, Orleans, France). Alirocumab was
from Sanofi (Chilly-Mazarin, France) while evolocumab was from
Amgen (Boulogne-Billancourt, France). Solutol HS15 was from BASF
(Ludwigshafen, Germany) and isoflurane was purchased from Hebei
Yipin Pharmaceutical Co. LTD (Guangdon, China). The tyrosine kinases
Src, PDGFR-β and Abl were purchased from Carna Biosciences (Kobe,
Japan) while cKit was obtained from Millipore (Billerica, MA, USA).
Peptides FAM-P2, FAM-P4 and FAM-P22 were acquired from GL
Biochem (Shanghai, China) while Coating reagent #3 was from Perkin
Elmer (Waltham, MA, USA). K2EDTA tubes were from Adamas-Beta
(Shanghai, China) while CircuLex PCSK9-LDLR in vitro binding assay
kits, CircuLex PCSK9 ELISA kits and mutant PCSK9-D374Y were from
CycLex (Nagano, Japan). Allophycocyanin-conjugated antibody against
the human LDLR and an IgG1 isotype control were from R&D Systems
(Lille, France). Cholesterol and Triglyceride FS colorimetric assay kits
were purchased from DiaSys, Thermo-Fisher Scientific (Illkirch,
France). Bodipy FL LDL was acquired from Invitrogen (Invitrogen,
Paisley, UK). RPMI and FBS were from Life Technologies (Saint Aubin,
France).
2

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
Fig. 1. (A) The domain structure of
human PCSK9 in its zymogen form. The
numbering throughout indicates the
position of significant amino acid re-
sidues including D186, H226 and S386,
which form the catalytic triad of
PCSK9. D374Y indicates an amino acid
substitution that has been characterised
as a gain-of-function mutation. SP de-
notes signal peptide. Proteolytic clea-
vage sites are denoted by a scissors
icon. Adapted from1 and3. (B) A three-
dimensional cartoon representation of
the X-ray crystal structure of human
PCSK9 lacking its C-terminal domain.
The prodomain is coloured in red, the
catalytic domain in green and the cat-
alytic residues D186, H226 and S386 in
pink. The region of PCSK9 that directly
interacts with the EGF-(A) domain of
LDLR is shown in blue. Adapted from
PDB file code 4NMX.10
2.2. Methods
prepared. Splitting patterns are designated as s (singlet), d (doublet), t
(triplet), q (quartet), m (multiplet) and br s (broad singlet). The
following solvents, reagents or scientific terminology may be referred
to by their abbreviations.
2.2.1. In silico screening
More than 1100 FDA-approved drugs from the database
BindingDB26 were subjected to a virtual screen by computationally
docking each compound into PCSK9 using the software package Qu-
CBit.27 The structure of PCSK9 described by Lo Surdo et al.28 and re-
presented in PDB29 file code 3P5C was adopted for screening, however
the prodomain (Fig. 1B) was omitted. The binding site of interest within
PCSK9 was defined by amino acid residues His 226, Leu 230, Leu 253,
Thr 264, Leu 286, Leu 289, Ala 290, Leu 297, Asn 317 and Ser 386
(Supplementary Fig. S1). Virtual screening was conducted as described
previously.30
2.2.2.2. Synthesis of (S)-N-(3-((3-aminopiperidin-1-yl)methyl)-5-(4-
methyl-1H-imidazol-1-yl)phenyl-4-methyl-3-phenoxybenzamide
(3f). DIBAL-H (1 M in toluene, 346 mL, 346 mmol) was slowly added
to a solution of 44 (30 g, 115 mmol) in THF (600 mL) at −78 °C and the
temperature of the mixture slowly warmed to −50 °C throughout the
addition. The mixture was then allowed to come to room temperature
and further stirred for 3 h. The reaction was subsequently cooled to
−20 °C, quenched with saturated ammonium chloride solution
(1000 mL) and then stirred for 30 min. A white suspension was
filtered through a celite pad and washed with EtOAc (2 × 1000 mL).
The organic layer was separated and washed with water (500 mL),
dried over anhydrous sodium sulphate and concentrated under vacuo to
give 65 (26 g, 97.1%) as a brown solid. 1H NMR (400 MHz; DMSO-d6):
δ 8.24 (s, 1H), 8.17 (s, 1H), 7.96 (s, 1H), 5.64 (t, 1H), 4.62 (d, 2H).
LCMS: m/z 230.87 (M–H)−, 73.2%.
2.2.2. Chemical synthesis
Small scale synthesis of compounds 1a–1h, 2a–2f and 3a–3f was
performed at ChemPartner (Shanghai, China) as detailed in the
Supplementary Information. A larger preparation of 3f (batch J996)
was synthesised at Jubilant Chemsys (Noida, India) and is reported
herein. A summary of the reaction scheme is provided in Fig. 2. All
compounds were synthesized at a purity of greater than 95%. Data
displaying the characterization of all compounds by 1H NMR and LCMS
is provided in the Supplementary Information.
To a solution of 65 (26 g, 112 mmol) in DMF (364 mL), 1 (32.2 g,
392 mmol), L-proline (6.45 g, 56 mmol), CuI (17.07 g, 89.6 mmol) and
K2CO3 (38.7 g, 280 mmol) were sequentially added at room tempera-
ture under N2. The reaction mixture was then heated at 130 °C for 18 h.
Following the consumption of starting material, the reaction mixture
was cooled to room temperature and water (1000 mL) was added to it.
The reaction mass was then filtered through a celite pad and then
washed with EtOAc (2 × 500 mL). The filtrate was subsequently ex-
2.2.2.1. General methods. The yields reported herein refer to purified
products (unless specified). Analytical TLC was performed on Merck
silica gel 60 F254 aluminium-backed plates. Compounds were visualised
by UV light and/or stained with I2, ninhydrin or potassium
permanganate solution followed by heating. Flash column
chromatography was performed on silica gel. 1H NMR spectra were
recorded on a Bruker 400 MHz, Avance II spectrometer with a 5 mm
DUL (Dual) 13C probe and Bruker 400 MHz, Avance III HD
spectrometer with BBFO (Broad Band Fluorine Observe) probe.
Chemical shifts (δ) are expressed in parts per million (ppm) with
reference to the deuterated solvent peak in which the sample was
tracted with ethyl acetate (2
× 500 mL), washed with brine
(2 × 500 mL), dried (Na2SO4) and concentrated under vacuo to give
crude residue. The residue was triturated by EtOAc (3 × 250 mL) to
give 66 (12 g, 45.9%) as a light brown solid. 1H NMR (400 MHz; DMSO-
d6): δ 8.34–8.30 (m, 2H), 8.12 (s, 1H), 8.01 (s, 1H), 7.64 (bs, 1H), 5.64
(t, 1H), 4.67 (d, 2H), 2.17 (s, 3H). LCMS: m/z 234.1 (M + H)+, 93.9%.
Et3N (28 mL, 205 mmol) was subsequently added to a solution of 66
3

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
Fig. 2. The synthetic route used to generate compound 3f.
(12 g, 51.4 mmol) in DCM (240 mL) and stirred at room temperature for
15 min. Methane sulfonylchloride (7.95 mL, 102 mmol) was then
slowly added at 0 °C and the reaction stirred at room temperature for a
further 3 h. The progress of the reaction was monitored by TLC and
after consumption of starting material, the reaction mixture was
quenched with water (500 mL) and extracted with DCM (2 × 250 mL).
The combined organic layer was washed with brine (200 mL), dried
anhydrous with sodium sulphate and concentrated under vacuo to give
67 (11 g, crude, 93.6%) as a brown gummy liquid. The crude pre-
paration was used in the next step without further purification.
K2CO3 (12.2 g, 88.3 mmol) was added to a solution of tert-butyl (S)-
piperidin-3-ylcarbamate (7.07 g, 35.3 mmol) in dry DMF (68.2 mL) and
stirred for 10 min prior to the addition of 67 (11 g, crude). The resulting
mixture was stirred at room temperature for 18 h, and the progress of
reaction was monitored by TLC which showed consumption of starting
material. The reaction mixture was then diluted with water (500 mL)
and extracted with EtOAc (2 × 500 mL). The combined organic layer
was washed with brine solution (200 mL), dried with sodium sulphate
and concentrated under vacuo to give a crude residue. The residue was
purified by combi flash using 3% MeOH/DCM to give 35 (11 g, 58.6%)
as a yellow solid. 1H NMR (400 MHz; DMSO-d6): δ 8.40–8.30 (m, 2H),
8.05 (s, 1H), 8.01 (s, 1H), 7.64 (s, 1H), 6.74 (d, J = 7.2 Hz, 1H),
3.70–3.60 (m, 2H), 3.52–3.40 (m, 1H), 2.80–2.70 (m, 1H), 2.71–2.60
(m, 1H), 2.17 (s, 3H), 2.02–1.82 (m, 2H), 1.72–1.60 (m, 2H), 1.50–1.40
(m, 2H), 1.38 (s, 9H). LCMS: m/z 416.24 (M + H)+, 86.5%.
1.72–1.60 (m, 3H), 1.20–1.40 (m, 2H), 1.35 (s, 9H). LCMS: m/z 386.25
(M + H)+, 85.4%.
To a solution of 68 (15.0 g, 87.9 mmol), 69 (77.3 mL, 87.9 mmol),
Cu-powder (2.79 g, 43.9 mmol), CuI (8.37 g, 43.9 mmol) and K2CO3
(24.3 g, 175 mmol) were mixed in a steel vessel and stirred for 2 days at
230 °C. Progress of the reaction mixture was monitored by TLC and
LCMS. The resulting mixture was diluted with EtOAc (250 mL), 1 N
NaOH (500 mL) and stirred for 1 h at room temperature. The aqueous
layer was extracted with EtOAc (3 × 250 mL) and then acidified with
1 N HCl solution (pH 2) and extracted with EtOAc (3 × 250 mL). The
organic layer was washed with brine solution (1 × 250 mL), dried with
anhydrous sodium sulphate and then concentrated under vacuo to give
58 (8.0 g, 39.8%) as a white solid. 1H NMR (400 MHz; DMSO-d6): 12.96
(s, 1H), 7.64 (d, 1H), 7.46–7.38 (m, 3H), 7.30 (s, 1H), 7.15 (t, 1H), 6.97
(d, 2H), 2.27 (s, 3H).
HATU (13.7 g, 36.1 mmol) and DIPEA (16.78 mL, 96.3 mmol) were
sequentially added to a solution of 58 (5.5 g, 24 mmol) in dry DMF
(110 mL) at room temperature. After 15 min stirring at this tempera-
ture, 36 (10 g, 26.0 mmol) was added. The reaction mixture was stirred
at room temperature for 18 h and progress of reaction mixture was
monitored by TLC which showed consumption of starting material. The
resulting mixture was diluted with water (300 mL) and then extracted
with ethyl acetate (2 × 500 mL). The organic layer was washed with
brine solution (8 × 250 mL), dried over anhydrous sodium sulphate
and concentrated under vacuo to give crude residue. The residue was
purified by combi flash using 2% MeOH/DCM to give 64 (8.7 g, 60.6%)
as a brown solid. LCMS: m/z 596.39 (M + H)+, 98.06%.
To a stirred solution of 35 (11 g, 26.4 mmol) in EtOH:H2O (6:1,
385 mL), ammonium chloride (8.48 g, 158.8 mmol) and Fe-powder
(13.29 g, 238 mmol) were added. The reaction mixture was stirred at
70 °C for 3 h and progress of reaction mixture was monitored by TLC
which showed consumption of starting material. The reaction mixture
was then cooled to room temperature and filtered through a celite pad
followed by a wash with EtOH (250 mL). The filtrate was concentrated
to dryness and then diluted with DCM (1000 mL) and water (500 mL).
The collected organic layer was dried over anhydrous sodium sulphate
and concentrated under vacuo to give 36 (10 g, 98.03%) as a yellow
solid. 1H NMR (400 MHz; DMSO-d6): δ 7.89 (s, 1H), 7.22 (s, 1H),
6.62–6.50 (m, 2H), 6.47 (s, 1H), 5.33 (bs, 2H), 3.52–3.36 (m, 3H),
2.82–2.72 (m, 1H), 2.70–2.60 (m, 1H), 2.13 (s, 3H), 1.90–1.82 (m, 1H),
To a solution of 64 (8.7 g, 14.60 mmol) in 1,4-dioxane (435 mL),
HCl (g) was purged at 0–25 °C for 2 h. The reaction mixture was con-
centrated under vacuo to give crude material which was a slurry in
EtOAc (2 × 250 mL) and filtered. The solid material was dissolved in
water (150 mL), acetonitrile (50 mL) and lyophilized to give 3f. HCl
(8.2 g, 93.07%). 1H NMR (400 MHz; DMSO-d6 with D2O): δ 9.35 (s,
1H), 8.08 (s, 1H), 7.94 (s, 1H), 7.82 (s, 1H), 7.74–7.72 (d, 1H), 7.64 (s,
1H), 7.51–7.49 (d, 1H), 7.42–7.36 (m, 2H), 7.14–7.10 (t, 1H),
6.94–6.92 (d, 2H), 4.32–4.30 (d, 2H), 3.43 (bs, 2H), 3.27 (bs, 1H), 2.86
(bs, 2H), 2.32 (s, 3H), 2.25 (s, 3H), 1.98–1.90 (m, 2H), 1.76–1.73 (m,
1H), 1.51 (m, 1H). LCMS: m/z 495.26 (M + H)+, 98.47%.
4

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
2.2.3. In vitro PCSK9-LDLR binding assay
human PCSK9, the protein was initially biotinylated using an EZ-link
Compounds were screened for their inhibitory activity against the
PCSK9-LDLR interaction using a CircuLex PCSK9-LDLR in vitro binding
assay kit according to the manufacturer’s instructions. Briefly, stock
solutions of test compounds were prepared as 10 mM stocks in DMSO
and then serially diluted in 1× reaction buffer by dispensing each di-
lution into the wells of a microplate pre-coated with the LDLR-EGF-AB
domain. Recombinant human His-tagged PCSK9 (7 ng) was subse-
quently added to the reaction mix (100 µL reaction volume) and then
incubated at room temperature while shaking at 300 rpm for 3 h before
washing thoroughly (4×) with wash buffer. A biotinylated anti-His tag
monoclonal antibody (1× 100 µL of conjugate dilution buffer) was
added to each well and the microplate incubated for a further 1 h at
room temperature while shaking at 300 rpm. After washing each well
again 4× with wash buffer, HRP-conjugated streptavidin (100 µL) was
added and the microplate incubated at room temperature for 20 min,
shaking at 300 rpm. The samples were subjected to a final thorough
wash (4×) with wash buffer before the addition of substrate reagent
(100 µL) to each well. The microplate was then shaken at 300 rpm for a
further 15 min at room temperature and the reaction terminated by the
addition of stop solution (100 µL). The absorbance of each sample was
measured at 450 nm and 540 nm using an EnSpire Multimode Plate
Reader (Perkin Elmer, Waltham, MA, USA). The percent inhibition of
binding (y-axis) was plotted as a function of compound concentration
(x-axis) and the IC50 determined for each compound.
NHS-PEG4-biotin reagent (Thermo-Fisher Scientific) and then desalted
using
a Zeba spin desalting column (Thermo-Fisher Scientific).
Biotinylated PCSK9 (27 µg/mL) was subsequently loaded onto a Dip
and Read Super Streptavidin (SSA) biosensor followed by a blocking
step with biocytin. Compound 3f (0.2–25.6 µM) was permitted to as-
sociate with captured PCSK9 for 2 min and then allowed to dissociate
for a further 10 min.
2.2.6. LDL HepG2 cell-uptake assay
Human liver HepG2 cells (ChemPartner, Shanghai, China) were
seeded in 96-well cluster plates at 2 × 105 cells/mL (100 µL) and al-
lowed to attach overnight. The gain-of-function PCSK9-D374Y mutant
(2 μg/mL) was added, along with test compounds as indicated. Each
sample was incubated for 16 h, whereupon the medium was replaced
with fresh medium containing 10 μg/mL Bodipy FL LDL, and the
samples were incubated for a further 4 h. Samples were then washed
using warm PBS and LDL uptake was quantified on a EnVision
Multimode Plate Reader (Perkin Elmer, Waltham, MA, USA) using ex-
citation and emission wavelengths of 485 and 530 nm, respectively.
2.2.7. Ex vivo LDLR surface expression on primary human lymphocytes
Human peripheral blood mononuclear cells (PBMCs) were isolated
from healthy human volunteers under protocols approved by the in-
stitutional ethics committee of Université de la Réunion using Ficoll
Paque Plus. Human PBMCs were subsequently frozen at −80 °C in
RPMI culture medium containing 70% FBS and 10% DMSO until use.
Freshly thawed PBMCs were seeded into flat bottom 96-well plates
(2 × 105 cells per well) in RPMI containing 10 mM HEPES, 1 mM so-
dium pyruvate and 0.5% FCS for 2 h at 37 °C. The culture medium was
subsequently supplemented with or without 600 ng/mL of recombinant
PCSK9-D374Y for 4 h in the presence or absence of increasing con-
centrations of compound 3f or alirocumab. Cells were subsequently
washed with ice-cold PBS containing 1% BSA and then incubated with
an allophycocyanin-conjugated antibody against human LDLR or an
IgG1 isotype control for 20 min at room temperature away from light.
Cell samples were washed twice in ice-cold PBS-1% BSA and once in
ice-cold PBS prior to analysis on a CytoFLEX flow cytometer (Beckman
Coulter, Pasadena, CA). Forward- versus side-scatter gates were set to
include only viable lymphocytes and a minimum of 5 × 103 lympho-
cytes were analysed for LDLR expression. The level of LDLR on the
lymphocyte surface for each treatment condition was expressed as a %
of the vehicle-only control that was unexposed to PCSK9-D374Y.
2.2.4. In vitro kinase assays
All compounds were initially prepared as 10 mM stocks in DMSO
and subsequently diluted in DMSO to 50× their final concentrations.
Ten µL of each DMSO dilution was transferred to a well in a fresh 96-
well cluster plate containing 90 µL 1× Kinase buffer (50 mM HEPES pH
7.5, 10 mM MgCl2, 2 mM DTT and 0.01% Brij-35) and then mixed for
10 min. Five µL of each compound dilution was subsequently added to a
384-well plate in duplicate. Ten µL of enzyme solution containing either
Src, cKit, PDGFR-β or Abl (final concentrations 0.6, 6, 10 or 0.9 nM,
respectively) in 1× Kinase buffer was then added to each well and the
mix incubated at room temperature for 10 min. To initiate each reac-
tion, 10 µL of peptide solution containing FAM-labelled peptide (final
concentrations 3 µM FAM-P2 for Abl, 3 µM FAM-P22 for cKit and
PDGFR-β or 3 µM FAM-P4 for Src) and ATP (final concentrations 36, 6,
40 and 12 µM for Src, cKit, PDGFR-β or Abl, respectively) in 1× Kinase
buffer was added to each well. All reactions were incubated at 28 °C for
1 h and then terminated by the addition of 25 µL stop buffer (100 mM
HEPES pH 7.5, 50 mM EDTA, 0.2% Coating Reagent #3 and 0.015%
Brij-35). All samples were then subjected to analysis using an EZ Reader
Ⅱ (Caliper Life Sciences, Waltham, MA, USA) (downstream voltages
−500 V, upstream voltages −2250 V, base pressure −0.5 PSI, screen
pressure −1.2 PSI) to read conversion values. Conversion values were
transformed into % inhibition of kinase activity using the formula:
2.2.8. Pharmacokinetic studies
All pharmacokinetic studies were reviewed and approved by the
Institutional Animal Care and Use Committee of ChemPartner
(Shanghai, China). For 5-in-1 cassette dosing studies, male CD1 mice
(31–32 g, 6–7 weeks) were acquired from LC Laboratory Animal Co.
LTD and were allowed access to food and water ad libitum throughout
the in-life phase of the study. Mice were dosed intravenously (IV) with a
cassette of 0.4 mg/kg 1b, 1c, 1d, 1g and 3a in 5% DMAC, 5% Solutol
HS15 and 90% saline. At designated time points (0.083, 0.25, 0.5, 1, 2,
4, 8 and 24 h), mice were restrained manually when approximately
20 µL of blood was collected in K2EDTA tubes via the facial vein for
serial bleeding or cardiac puncture (under anaesthesia with isoflurane)
for terminal bleeding. Each blood sample was stored on wet ice until
20 µL of each sample was diluted with 60 µL deionised H2O. Each
sample was then vortexed well and then stored at −80 °C until analysis.
Samples were assayed for their levels of 1b, 1c, 1d, 1g and 3a by LC/
MS/MS using an API 4000 triple Quadruple system (Applied
Biosciences, Foster City, CA USA) operating in a positive electrospray
ionisation mode.
% Inhibition = [(MA X)/(MA MI)] × 100%
where MA = conversion value of DMSO only controls, MI = conversion
value of no enzyme controls and X = conversion value at any given
compound dose. IC50 values were then calculated by plotting dose–r-
esponse curves using the XLfit application in Excel software.
2.2.5. Biolayer interferometry binding analyses
The interaction of evolocumab and compound 3f with recombinant
human PCSK9 was analysed by biolayer interferometry using an Octet®
RED96 system (ForteBio, Pall Life Sciences, Menlo Park, CA, USA).
Evolocumab (1.25 µg/mL) was initially loaded onto a Dip and Read
Anti-hIgG Fc Capture (AHC) biosensor in 1× kinetic buffer.
Recombinant human PCSK9 (1.56– 50 nM) was subsequently allowed
to associate with captured evolocumab in 1× assay buffer (1× kinetic
buffer, 0.1% glycerol) for 5 min and then permitted to dissociate for a
further 10 min. For analyses involving compound 3f and recombinant
A second 5-in-1 cassette dosing study was conducted as described
above with the following modifications: 2a, 3c, 3d, 3e and 3f were
analysed, CD1 mice were 28–29 g and 7–8 weeks of age and 20 µL of
5

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
each blood sample was diluted with 40 µL deionised H2O in preparation
for analysis.
parallel, a separate cohort was dosed with evolocumab (10 mg/kg in
100% saline) once every 5 days. An n of 5 mice was used for each
treatment group. On days 1, 5 and 10 of treatment, blood samples were
collected from each animal via the retro-orbital plexus in a standard
Eppendorf tube containing 1 µL 0.5 M EDTA on ice. Samples were
subjected to centrifugation at 2500 g, the plasma collected and then
stored at −20 °C until analysis. On day 15, mice were humanely eu-
thanised and blood collected and processed as described above. The
abdominal cavity of each animal was dissected and visually inspected
for signs of toxicity. All plasma samples were subsequently assayed for
their total cholesterol and triglyceride levels using Cholesterol and
Triglyceride FS colorimetric assay kits, respectively. Plasma samples
were also analysed for their PCSK9 levels using a CircuLex mouse
PCSK9 ELISA kit according to the manufacturer’s instructions.
For a dedicated pharmacokinetic study of compound 3f, male
C57BL/6 mice (19–21 g) were purchased from LC Laboratory Animal
Co. LTD and were given ad libitum access to food and water throughout
the in-life phase of the study. Mice were dosed either IV with 5 mg/kg
3f in 85% saline, 10% Solutol HS15 and 5% DMSO, orally (PO) with
20 mg/kg 3f in 1% methylcellulose in H2O or subcutaneously (SC) with
20 mg/kg 3f in 85% saline, 10% Solutol HS15 and 5% DMSO. The
animals were restrained manually at designated time points (0.033,
0.083, 0.25, 0.5, 1, 2, 4, 7 and 24 h) when approximately 110 µL of
blood was collected in K2EDTA tubes via retro orbital puncture for
semi-serial bleeding or cardiac puncture (under anaesthesia with iso-
flurane) for terminal bleeding. Blood samples were placed on ice and
centrifuged (2000 g, 5 min at 4℃) within 15 min of collection to obtain
plasma samples which were subsequently stored at approximately
−70 °C until analysis. Plasma samples were assayed for their levels of
3f by LC/MS/MS using an API 5500 Qtriple system (Applied
Biosciences) operating in a positive electrospray ionisation mode.
3. Results and discussion
3.1. The identification of a small organic molecule that disrupts the PCSK9-
LDLR interaction
2.2.9. In vivo efficacy studies
A major challenge associated with targeting the PCSK9-LDLR in-
teraction with small molecules pertains to the relatively flat surfaces of
the two interfaces involved (Fig. 1B and5). With this in mind, the cat-
alytic domain of PCSK9 (Fig. 3A and PDB code 3P5C) was visually
probed in the absence of the prodomain for sites proximal to the LDLR-
binding interface that may accommodate the binding of a small mole-
cular weight agent. It was surmised that an appropriately shaped small
molecule binding within such a site may sufficiently disrupt the PCSK9-
LDLR interface to inhibit association of the two proteins. On close in-
spection of the catalytic domain, it was noted that a groove normally
occupied by the C-terminus of the prodomain and defined by residues
The in vivo research protocol #2018040410477054 was approved
by the animal use and care committee (CYROI). Female C57BL/6 wild-
type and PCSK9-/- knockout mice (8–16 weeks) were from Jackson
Laboratory (Bar Harbor, NE, USA) and were housed in an accredited
animal facility (20–25 °C, 50–80% humidity, 12 h light/dark cycles
with access to standard chow and water ad libitum). Prior to dosing,
mice acclimatised for 7 days with wild-type and PCSK9−/− knockout
mice caged separately (5 animals per cage). Mice were dosed SC daily
for 14 days with either 3f (3.28 mg/kg/day or 16.4 mg/kg/day in 5%
DMSO and 95% saline) or vehicle only (5% DMSO and 95% saline). In
B
A
Kinase binding motif
80
C
60
40
20
0
10-7
10-6
10-5
10-4
[Nilotinib] M
Fig. 3. Nilotinib is a small molecule inhibitor of PCSK9. (A) A molecular surface representation of the predicted interactions of nilotinib with the groove of PCSK9
leading into the enzyme’s active site. The model was generated following a virtual screen of the BindingDB database26 for compounds anticipated to interact with
PCSK9 in the binding groove as defined in the Materials and Methods. Nilotinib is represented as a stick model with the carbon skeleton coloured in cyan while the
molecular surface of the catalytic domain of PCSK9 is depicted in green. Amino acid residues that line the putative binding site are depicted in copper orange while
the EGF-(A) LDLR-binding region of the PCSK9 is shown in blue. The residue highlighted in yellow is D212 and may provide a key interaction with analogues of
nilotinib. The C-terminal domain of PCSK9 has been omitted for the purposes of clarity. (B) The chemical structure of nilotinib, emphasising one of its kinase binding
motifs. (C) The inhibition of PCSK9 binding with LDLR by nilotinib shown as a function of drug concentration. Recombinant human His-tagged PCSK9 was mixed for
3 h at room temperature with varying concentrations of nilotinib (as designated) in microplates coated with the LDLR-EGF-AB domain and 1 × reaction buffer. The
PCSK9-LDLR interaction was subsequently detected using a colorimetric-based assay as detailed in the Materials and Methods. Panel (C) is a single example of seven
independent experiments, with the error bars representing the S.D. of duplicates within this example.
6

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
His 226, Leu 230, Leu 253, Thr 264, Leu 286, Leu 289, Ala 290, Leu
297, Asn 317 and Ser 386 (Fig. 3A and Supplementary Fig. S1) may
potentially satisfy these requirements. In an effort to identify a com-
pound of this nature, an in silico screen of the FDA-approved small
molecule collection in BindingDB was deployed.26 The screen yielded
10 theoretical candidates which included nilotinib (Fig. 3B). Fig. 3A
depicts a computational docking pose of nilotinib bound within the
nominal binding groove of PCSK9 (with some residues omitted for
clarity) and is proximal to the EGF(A)-binding surface and catalytic
triad of the protein. Nilotinib was sourced from a commercial vendor
and subsequently assayed for inhibition of the PCSK9-LDLR interaction
using an in vitro binding assay. The assay confirmed that nilotinib was a
bona fide inhibitor of the PCSK9-LDLR interaction with an assay IC50 of
approximately 15.9 µM (Fig. 3C). The result was robust with the drug
demonstrating an average assay IC50 of 9.8 µM, a value that was de-
rived from seven independent experiments (IC50 range 3.9 – 15.9 µM,
Table 1).
3.2. Medicinal chemistry campaign
3.2.1. Series 1
Given that nilotinib is a marketed Bcr-Abl inhibitor (Tasigna®) and
possessed unwanted kinase activity, the next phase of discovery focused
on minimizing off-target kinase inhibition while maintaining or im-
proving activity in the PCSK9-LDLR binding assay. In the first round of
optimization, removal of the kinase binding moiety (Fig. 3B) and early
substitutions of the trifluoromethyl group of nilotinib (Fig. 3B) sharply
reduced kinase inhibitory activity against a small panel of kinases re-
presented by Src, c-Kit, Abl and PDGFR-β (Table 1), while maintaining
or increasing potency in the PCSK9-LDLR binding assay (Table 1). It
should be noted that PDGFR-β, c-Kit and Abl were selected as they
represent a sub-set of kinases particularly susceptible to inhibition by
nilotinib31 and are therefore ideal proxies for analysing any undesirable
kinase inhibitory activity retained by compounds generated in this
campaign. While Src is relatively impervious to inhibition by
Table 1
The chemical structures of nilotinib, compounds 1a–1h and associated IC50 (µM) values as determined by an in vitro PCSK9-LDLR binding assay and in vitro c-Kit, Abl,
PDGFR-β or Src kinase assays. Each assay was performed once unless otherwise specified as described in the Materials and Methods. ID indicates identifier.
ID
Structure
PCSK9-LDLR IC50 (µM)
Kinase assay IC50 (µM)
c-Kit
Abl
PDGFR-β
0.025
Src
1.9
Nilotinib
9.8 (range 3.9–15.9, n = 7)
0.158
0.0012
N
H
N
N
N
N
N
N
N
N
N
H
O
O
O
N
F
F
F
1a
1b
8.4
0.592
1.49
0.334
7.95
2.48
1.51
> 10
4.36
N
H
N
N
N
H
F
F
F
0.491
N
H
N
N
N
H
N
N
H
1c
0.758
2.88
> 10
> 10
> 10
N
H
N
N
N
N
N
H
O
N
N
1d
1e
0.714
11.3
2.57
2.77
> 10
> 10
5.42
> 10
> 10
> 10
N
H
N
N
N
N
H
N
O
HO
O
1f
9.28
13.4
0.347
6
> 10
> 10
> 10
> 10
> 10
> 10
N
H
N
N
N
N
N
H
O
HO
O
1 g
1 h
7.41
0.302
0.142
> 10
> 10
N
H
N
N
N
N
N
H
O
7

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
nilotinib31, it was selected as a representative of the Src family of tyr-
osine kinases. Inhibition of members of this family by any given com-
pound can generate clinically deleterious side-effects.32 Exchange of the
pyridine ring present in nilotinib for an unsubstituted phenyl (1a–1h,
Table 1) consistently reduced kinase inhibitory activity across the
series, with certain examples showing only micromolar inhibition of c-
Kit. At the same time, the most potent antagonists of the target pro-
tein–protein interaction bore basic substitutions off the right-hand
phenyl ring meta to the imidazole (1b–1d, Table 1), all of which
achieved sub-micromolar activity. Acidic or non-basic hydrophilic
groups (1e–1g, Table 1) were effectively no more potent than nilotinib
itself, although they were consistently less active against Abl, PDGFR-β
and Src kinases. Abl-inhibitory activity returned in the case of the
neutrally substituted example 1h (Table 1).
rationalised that the introduction of a fluorine atom at position X of
Fig. 4A may endow greater metabolic stability by inhibiting possible
aromatic ring metabolism. Second, modelling of compound 1d’s basic
piperazinomethylene moiety indicated a potential interaction through
the acidic side-chain of Asp 212 of PCSK9 that could be exploited fur-
ther in this series (Figs. 3A, 4B and Supplementary Fig. S1). Third, only
modest variations around the imidazole group (Fig. 4A) were probed as
modelling suggested that any substituent off the right hand 5-mem-
bered heterocycle larger than a methyl group would not favour binding
(data not shown). Table 2 indicates the results of this series.
The introduction of a fluoro substituent at the para position of the
left-hand phenyl ring (e.g. compound 2a) was insignificant in the
context of PCSK9-LDLR inhibition (Table 2). That some inhibition of
certain kinases at sub-10 μM levels remained (compounds 2a, 2c–2f)
may be attributable to other pharmacophores (Table 2). There are
multiple examples of morpholine- and aminopyrimidine-based hinge-
binding motifs33 so it is not possible to rule out such orientations of
compounds in this series for interactions with c-Kit, for example
(Table 2).
3.2.2. Series 2
With this encouraging early SAR in hand, the next phase of ex-
ploration sought to probe changes in three disparate portions of the
molecule to promote affinity for PCSK9, minimise anti-kinase activity
and improve on any perceived ADME liabilities (Fig. 4A). First, it was
Consistent with the modelling of compound 1d and a putative
A
B
Fig. 4. (A) A molecular scaffold showing where changes to the molecule were designed to afford increased affinity for PCSK9 along with decreased off-target
selectivity and improved drug-like properties. (B) A molecular model of compound 1d docked into the putative binding site of PCSK9. In this model, the basic
piperazinomethylene moiety of compound 1d is proximal to the acidic side-chain of Asp 212 of PCSK9 (indicated in yellow), potentially enabling a stabilising
interaction to occur (indicated by a light blue double-headed arrow).
8

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
Table 2
The chemical structures of compounds 2a–2f. The IC50 value (µM) of each compound was evaluated in either an in vitro PCSK9-LDLR binding assay or an in vitro c-Kit,
Abl, PDGFR-β or Src kinase assay. Each assay was conducted once as described in the Materials and Methods. ID indicates identifier. An asterisk denotes an IC50 value
that was obtained over a suboptimal assay dynamic range, due perhaps to physical compound characteristics (e.g. solubility).
ID
Structure
PCSK9-LDLR IC50 (µM)
Kinase assay IC50 (µM)
c-Kit
1.96
Abl
PDGFR-β
2.04
Src
2a
0.428*
> 10
> 10
> 10
> 10
2b
2c
14
> 10
9.87
> 10
> 10
> 10
> 10
2.57
2d
2e
40.1
10.7
0.79
3.1
> 10
> 10
> 10
5.52
> 10
> 10
2f
3.29*
0.093
> 10
> 10
> 10
interaction of its piperazinomethylene moiety with Asp 212 of PCSK9
shown in Fig. 4B, the requirement for a distal basic group extending
from the right-hand phenyl ring to achieve potency was clear from this
exercise. Like the potency displayed by piperazine-bearing compounds
1b–1d (Table 1), compound 2a displayed sub-µM inhibition of the
PCSK9-LDLR interaction (Table 2), most likely because of the basic
nature of its aminopiperidine group. In contrast, compounds 2b–2d,
while retaining the piperidine group of 2a yet decorated with various
appendages from the six-membered ring, lost substantial potency re-
lative to 2a in the binding assay (Table 2). Compound 2c,
(IC50 = 2.57 μM, Table 2) was a possible exception, perhaps due to a
non-ionic hydrogen bond interaction between the Asp 212 side-chain of
PCSK9 (Fig. 4B) and 2c’s amide group, but this was not verified.
Substitution of the imidazole in 1c with a methylisoxazole group
having similar geometry yet different charge potential cost approxi-
mately a log in potency in the PCSK9-LDLR binding assay (compound
2e, Table 2), pointing to the need for the basic nitrogen. The flanking
methyl group was retained not so much for physical interaction but to
minimize the potential for heme binding/CYP450 enzyme interaction.
Removal of the imidazole from 1c altogether (2f, Table 2) served to
reduce PCSK9-LDLR binding somewhat, and reintroduced a degree of
kinase inhibition, perhaps via the aforementioned aminopyrimidine
type pharmacophore.33
as truncating the left-hand side of the molecule, to lower molecular
weight, to decrease the chance of alternative kinase binding modes, and
to enhance novelty (Fig. 4A). With this in mind, we retained the me-
thylimidazolyl-phenyl group and basic moieties on the right-hand side
of the molecule that were crucial for maintaining sub-µM binding
against PCSK9-LDLR (Table 3). This series clearly demonstrated the
benefit of smaller left-hand groups in avoiding off-target (kinase) in-
hibition (Table 3). While PCSK9-LDLR binding was substantially re-
duced in the case of the toluamide 3b relative to the comparator 1c, this
compound showed no sub-10 μM kinase inhibition in the small panel
(Table 3). The aryl ketone 3e behaved similarly, though with improved
target binding (Table 3).
Replacement of the aminopyrimidine linker with aryl ethers (3c, 3d,
and 3f, Table 3) illustrated the most favourable potency and selectivity
numbers thus far. Noteworthy was the isoquinoline ether 3d, which
yielded the most potent small molecule inhibition of PCSK9-LDLR yet
observed (IC50 = 78 nM, Table 3), albeit across a narrow dynamic
range that could have been limited by physical compound character-
istics. The selectivity was also less for 3d against the kinase panel
(Table 3), perhaps due to kinase hinge binding via the isoquinoline
nitrogen.
3.3. A summary of the medicinal chemistry campaign
3.2.3. Series 3
Collectively, the results indicate that potency against the PCSK9-
LDLR interaction is likely driven both by favourable basic groups
The next phase of optimization explored core modifications as well
9

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
Table 3
PCSK9 readily associated with evolocumab in a concentration-depen-
dent manner at low doses (1.56–50 nM) of the protein, consistent with a
very tight interaction (KD = 8.0 pM,35). The interaction was also highly
stable, with no clear loss of PCSK9 binding detectable in the first 10 min
of dissociation (Fig. 5A). Next, biotinylated PCSK9 was loaded onto a
Super Streptavidin (SSA) biosensor and then exposed to increasing
concentrations of compound 3f. While no binding was evident at 0.2
and 0.4 μM 3f (Fig. 5B), a clear increase in biolayer interferometry shift
was notable at 0.8 and 1.6 μM 3f (Fig. 5B), indicating direct binding by
compound 3f to PCSK9. It is important to note that non-specific binding
by the compound was eliminated by subtracting background signal.
Moreover, investigations are ongoing to further confirm this pre-
liminary data.
The structures of compounds 3a–3f. Compounds were assayed for inhibition of
the LDLR-PCSK9 interaction using an in vitro binding assay as described in the
Materials and Methods. In parallel, each compound was also evaluated for in-
hibition of c-Kit, Abl, PDGFR-β or Src activity using an in vitro kinase assay as
detailed in the Materials and Methods section. Values represent the assay IC50
(µM) of each compound obtained from a single experiment. ID indicates iden-
tifier. An asterisk denotes that the IC50 value was obtained over a suboptimal
assay dynamic range as described in the legend of Table 2.
ID Structure
PCSK9-
Kinase assay IC50 (µM)
LDLR
IC50 (µM) c-Kit
Abl PDGFR-β Src
3a
13.4
0.618 > 10 > 10
> 10
> 10
3.5. Compounds 1c, 3d and 3f restore LDL uptake by HepG2 liver cells in
the presence of PCSK9 D374Y
3b
3c
3d
3e
26.3
> 10 > 10 > 10
Next, selected compounds were subjected to a cell-based assay to
evaluate their impairment of the PCSK9-LDLR interaction in a func-
tional context. Human liver HepG2 cells expressing LDLR at their
plasma membrane were incubated with the gain-of-function mutant
PCSK9 D374Y (Fig. 1A and1,3) either alone or in combination with
compounds 1c, 3c, 3d or 3f. AMB-657286, a small molecule inhibitor of
the PCSK9-LDLR interaction developed by Shifa Biomedical,36 was used
as a reference control. As anticipated, PCSK9 D374Y significantly im-
paired the uptake of fluorescently labelled LDL particles by HepG2 liver
cells relative to an untreated control (Fig. 5C). Fig. 5C shows that co-
incubation with 0.1 or 1 µM of the reference compound AMB-657286
relieved the inhibition of uptake by PCSK9 D374Y, with LDL uptake
returning to baseline levels at 1 µM. On the addition of 0.1 or 1 µM 1c,
3d or 3f, there was a dose-dependent restoration of LDL internalization
by liver cells to baseline levels (Fig. 5C), suggesting that these com-
pounds disrupted the target protein–protein interaction. Compound 3c
was the exception of this group, with the compound displaying no
significant increase in LDL uptake by liver cells (Fig. 5C). It is note-
worthy that compounds 1c, 3d and 3f maintain activity through the
restoration of LDL uptake even though the gain-of function mutant
PCSK9 D374Y applied in this assay binds 6–25-fold more potently with
LDLR relative to its wild-type counterpart.37,38
0.759*
0.078*
2.13
> 10 > 10 7.66
0.227 > 10 3.29
> 10 > 10 > 10
> 10
2.62
> 10
3f
0.537
> 10 > 10 > 10
> 10
In an effort to rationalise why 3c had achieved sub-µM potency in
the PCSK9-LDLR binding assay (Table 3) yet failed to increase LDL
cellular uptake (Fig. 5C), we re-examined its potency in the binding
assay. As mentioned earlier, the dynamic range of the binding assay was
relatively narrow for some compounds, for reasons not yet determined.
In particular, 3c showed a maximum inhibition of 20% in this assay,
whereas compounds 1c, 3d, and 3f all showed at least 35% inhibition,
with 3f showing 57% inhibition at the highest concentration tested
(data not shown). It was therefore concluded that potency in the bio-
chemical binding assay served as a useful surrogate for cell-based re-
storation of LDL uptake presuming the range of inhibition was at least
35%. To date we have not established upper and lower limits of the
binding assay’s utility, other than to conclude that 20% is insufficient to
translate to cellular efficacy.
(Tables 1 and 2) combined with linked biaryl ethers (Table 3). The
phenoxyphenyl ethers (compounds 3c and 3f, Table 3) provide an ideal
balance between potency and selectivity. The isoquinoline group′s
added potency (compound 3d, Table 3) is consistent with a further
potential interaction between the ring nitrogen and the side-chain of
Lys 222 within the binding pocket of PCSK9 (Supplementary Figs. S1
and S2). Supplementary Fig. S2 shows that the angular methyl group on
the interior ring places the distal aromatic in a favourable tilt, resulting
in a sub-3 Å distance between side chain and inhibitor groups.
It is also noteworthy that compound 3f restored LDL uptake by
HepG2 cells at a dose (0.1 µM, Fig. 5C) that was lower than its IC50 as
evaluated by the in vitro PCSK9-LDLR binding assay (0.537 µM,
Table 3). While the in vitro binding assay data is consistent within the
chemical series detailed in Tables 1–3, this biochemical assay cannot
fully account for any indirect mechanisms that may confer activity in
cellular assays such as the HepG2 assay (Fig. 5C). A variety of factors
operating beyond the immediate control of the experimental conditions
in more complex cellular systems may explain the discrepancy between
effects observed at lower concentrations than those predicted by the
biochemical assay. Whether these be additional molecular targets and/
or physicochemical characteristics with effects on permeability etc, we
are working to clarify these as we develop this series further.
3.4. Compound 3f binds directly to PCSK9
While compound 3f (and the entire chemical series in general) was
designed to bind within the putative binding site of PCSK9 (Fig. 3A), it
remained conceivable that disruption of the target protein–protein in-
teraction was mediated through direct antagonism of LDLR as opposed
to PCSK9. In an effort to confirm direct binding of compound 3f to
PCSK9, biolayer interferometry analysis of the interaction was de-
ployed. The anti-PCSK9 monoclonal antibody evolocumab34 was in-
itially captured by an Anti-hIgG Fc Capture (AHC) biosensor and then
exposed to varying concentrations of PCSK9. Fig. 5A demonstrates that
10

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
B
A
20000
20000
17500
15000
12500
C
**
**
*
***
***
**
***
*
17500
15000
12500
10000
10000
10000
10000
0
0
0.1
1c
1
0.1
3c
1
0.1
1
0.1
1
0.1
1
µM
d
Y
e
t
74
a
3
e
r
D
t
9
n
3f
3d
PCSK9 D374Y
AMB
K
U
S
C
P
Fig. 5. Evolocumab and compound 3f bind directly to PCSK9. Biolayer interferometry analysis of evolocumab or compound 3f binding with PCSK9 using a ForteBio
Octet® RED96 system. (A) Evolocumab was immobilized to an Anti-hIgG Fc Capture (AHC) biosensor and then exposed to various concentrations of PCSK9
(1.56–50 nM as indicated) for 5 min. The interaction was subsequently allowed to dissociate for a period of 10 min. (B) Biotinylated PCSK9 was captured by a Super
Streptavidin (SSA) biosensor and then exposed to compound 3f at a variety of concentrations (0.2–1.6 µM as denoted) for 2 min. Compound 3f was then permitted to
dissociate for a further 10 min. Each sensogram exhibits the biolayer interferometry signal (nm) expressed as a function of time (s). (C) Compound 3f promotes the
uptake of fluorescently labelled LDL by human liver HepG2 cells in the presence of PCSK9 D374Y. HepG2 cells were co-incubated with the gain-of-function PCSK9
D374Y mutant (2 μg/mL) and compounds 1c, 3c, 3d, 3f or AMB (0.1 or 1 µM) as indicated for 16 h. Fluorescently labelled LDL (10 µg/mL) was subsequently added
for 4 h prior to the measurement of LDL uptake by HepG2 cells as detailed in the Materials and Methods. Each column represents the average cellular uptake of LDL as
indicated by the fluorescence in arbitrary units. Error bars represent the S.D. of triplicates. An unpaired student t test (two-tailed) was applied to each sample using
the PCSK9 D374Y only control as a comparator. P values < 0.05 are designated with asterisks: *** < 0.001, ** < 0.01, * < 0.05. “AMB” denotes the compound
AMB657286.
11

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
Fig. 6. Compound 3f restores the expression of
LDLR on the cellular surface of human lympho-
cytes in the presence of mutant PCSK9-D374Y.
Human PBMCs were cultured in the presence or
absence of 600 ng/mL recombinant PCSK9-
D374Y along with increasing doses of compound
3f or alirocumab for 4 h. Cell samples were wa-
shed with ice-cold PBS/1% BSA and then in-
cubated with an allophycocyanin-conjugated α-
LDLR antibody at room temperature for 20 min.
Following 2× ice-cold PBS/1% BSA washes and
a final wash with ice-cold PBS, samples were
analysed by flow cytometry as detailed in the
Materials and Methods. Each column represents
the average level of LDLR on the surface of
lymphocytes relative to the untreated control
which was normalized to 100%. Error bars re-
present the S.D. of triplicates. An unpaired stu-
ns
ns
ns
ns
**
**
**
125
100
75
50
25
0
125
100
75
50
25
0
dent
t test (two-tailed) was applied to each
treatment pair. P values < 0.05 are designated
with asterisks: ** < 0.01, ns denotes not sig-
nificant.
PCSK9 D374Y
- +
-
+
-
0.1
+
-
+
0.4
-
+
-
+
- +
10
1
4
Vehicle
[ 3f ] µM
Alirocumab
Table 4
A summary of the in vitro IC50 values (µM) of select compounds from Tables 1–3 presented alongside their pharmacokinetic properties in male CD1 mice. Mice were
dosed once (0.4 mg/kg) IV with a 5-in-1 cassette of either 1b, 1c, 1d, 1g and 3a or 2a, 3c, 3d, 3e and 3f. Blood samples were collected at designated timepoints
(0.033, 0.083, 0.25, 0.5, 1, 2, 4, 7 or 8 and 24 h) and then processed for analysis by LC/MS/MS to evaluate their compound levels. Pharmacokinetic parameters are
designated as follows: CL, clearance of compound; Vss, volume of distribution at equilibrium; T1/2, terminal half-life; AUClast, area under the blood con-
centration–time curve from t = 0 to the last measurable timepoint; AUC∞, area under the blood concentration–time curve from t = 0 to infinity, MRT∞, mean
residence time extrapolated to infinity.
ID
PCSK9-LDLR IC50 (µM)
Kinase assay IC50 (µM)
IV PK data from cassette studies
c-Kit
Abl
PDGFR-β
Src
CL (L/h/kg)
Vss (L/kg)
T1/2 (h)
AUClast (h*ng/mL)
AUC∞ (h*ng/mL)
MRT∞ (h)
1b
1c
1d
1g
3a
2a
3c
3d
3e
3f
0.491
0.758
0.714
13.4
1.49
2.88
2.57
6
7.95
1.51
> 10
5.42
> 10
> 10
2.04
7.66
3.29
> 10
> 10
4.36
3.15
1.97
2.14
6.58
0.748
1.41
1.05
3.05
4.86
0.3
3.74
4.3
1.64
2.21
3.29
0.216
2.15
10.8
7.36
2.77
1.09
25.7
116
189
166
58.5
508
243
338
128
78
127
204
187
63.7
541
297
384
133
83
1.19
2.16
2.66
0.239
2.27
10.4
8.98
2.3
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
> 10
2.61
5.65
1.72
1.69
13.3
9.26
6.7
13.4
0.618
1.96
> 10
0.227
> 10
> 10
0.428*
0.759*
0.078*
2.13
> 10
> 10
4.7
0.96
34.9
0.537
9.13
722
1472
3.6. Compound 3f restores the expression of LDLR on the cellular surface of
alongside PCSK9 D374Y, there was no evidence of restoration of LDLR
expression on the surface of lymphocytes relative to controls that did
not receive the mutant PCSK9 D374Y (Fig. 6). However, higher doses of
3f (1–10 µM) did indeed rescue LDLR levels (Fig. 6), to a point where
3f- and PCSK9 D374Y-treated samples were not significantly different
from their PCSK9-untreated counterparts. Collectively, these results
indicate that compound 3f at 1–10 µM rescues LDLR from PCSK9-
mediated destruction, thereby enabling a restoration of receptor levels
at the cell surface.
human lymphocytes in the presence of mutant PCSK9 D374Y
While compounds 1c, 3d and 3f promoted the uptake of LDL by
HepG2 cells despite the presence of mutant PCSK9-D374Y, further
evidence was sought to establish the compounds’ mechanism of action
as inhibitors of the PCSK9-LDLR interaction. It is well established that
PCSK9 functions to direct the LDLR for lysosomal degradation following
endocytosis by hepatocytes, thereby preventing LDLR recycling to the
cell surface.10,11 Consistent with this notion, the addition of exogenous
PCSK9 D374Y to primary human lymphocytes in an ex vivo system39
significantly decreased the surface expression of LDLR by ~40% re-
lative to untreated controls (Fig. 6). The co-addition of alirocumab, a
humanized monoclonal antibody against PCSK940, completely relieved
the suppression of LDLR expression induced by PCSK9 D374Y (Fig. 6), a
result that validates the assay as appropriate for the detection of PCSK9
inhibition. On the introduction of 0.1 and 0.4 µM compound 3f
3.7. A pharmacokinetic evaluation of compound 3f and analogues
As a prelude to in vivo efficacy studies, the most potent compounds
were subsequently screened using a murine cassette-based pharmaco-
kinetic system to evaluate their clearance, half-life and volume of dis-
tribution parameters. Following IV administration with selected com-
pounds, blood was sampled from male CD1 mice (0–24 h) and then
12

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
Fig. 7. (A) The average plasma concentra-
tions of compound 3f in male C57BL/6 mice
following a single dose via the following
routes: IV at 5 mg/kg (circles), SC at 20 mg/
kg (squares) or PO at 20 mg/kg (triangles).
Mice were dosed and blood samples col-
lected at t = 0.083, 0.25, 0.5, 1, 2, 4, 7 and
24 h. Samples were processed and analysed
for their levels of 3f as described in the
Materials and Methods. Plasma concentra-
tions are shown as a function of time (n = 3
per administration arm). Error bars re-
present the S.D. of three individual re-
plicates. The dotted line running parallel
with the x-axis indicates the average IC50 of
compound 3f as evaluated in the PCSK9-
LDLR binding assay and is taken from
Table 1. (B) The pharmacokinetic properties
of 3f as calculated using the data presented
in panel (A). The data obtained from dosing
3f as part of an IV cassette was taken from
Table 4 and is presented here for the pur-
poses of comparison. Pharmacokinetic
parameters are designated as described in
the legend of Table 4 and as follows: CMax,
the maximum blood concentration achieved
by the compound; TMax, the time at which
CMax occurs; F, the fraction of compound
that is bioavailable.
100000
10000
1000
100
A
IV
SC
PO
10
1
0
10
20
Time (h)
B
Route of administration
PK parameter
Dose (mg/kg)
SC
PO
20
IV (single)
IV (cassette)
20
5
0.4
T
Max (h)
1.00
2207
2.00
52.6
CMax (ng/mL)
CL
(L/h/kg)
1.09
3.87
9.86
4381
4605
3.56
0.300
9.13
25.7
722
Vss (L/kg)
T1/2 (h)
5.47
N/A
92.3
N/A
AUClast
(h*ng/mL)
AUC∞
16076
16811
1472
34.9
(h*ng/mL)
MRT∞ (h)
F (%)
91.3
0.527
evaluated for levels of parent compound. Table 4 summarises the
pharmacokinetic values of each compound screened.
To further evaluate the pharmacokinetic properties of 3f, cassette
studies were followed by a dedicated bioavailability study of 3f via oral,
intravenous and subcutaneous routes. The pharmacokinetic profile and
parameters of male C57BL/6 mice dosed with 3f via these routes is
presented in Fig. 7. While 3f was not well absorbed orally (Fig. 7) the
compound demonstrated excellent bioavailability via subcutaneous
administration (91%, Fig. 7) and was judged suitable for in vivo proof-
of-concept studies.
The first round of studies (represented in the top half of Table 4)
established that compounds bearing a truncated phenoxyphenyl core
(e.g. compound 3a, Table 4) were likely to be cleared less rapidly than
aminopyrimidines (1b–1d and 1g, Table 4). While 3a did not show
sufficient potency against the PCSK9-LDLR target (IC50 = 13.4 µM,
Table 4) its low clearance and mean residence time (0.748 L/h/kg and
2.27 h, respectively, Table 4) suggested that the more potent and se-
lective analogue 3f might behave accordingly. Indeed, when dosed in
the second round (indicated by compounds in the bottom half of
Table 4), 3f proved to be superior in all aspects, with potential for once
per day dosing as suggested by an extended half-life (T1/2 = 25.7 h,
Table 4) and low clearance (CL = 0.3 L/h/kg, Table 4).
In addition to its high subcutaneous bioavailability, it was noted
that the plasma concentrations of 3f in mice treated SC with 20 mg/kg
3f exceeded the PCSK9-LDLR binding inhibition IC50 value of 537 nM
(Table 3) over at least 16 h (horizontal dotted line, Fig. 7), with 24 h
concentrations at approximately 93 nM. Notionally, this may translate
to potential once-a-day dosing at 20 mg/kg/day for proof-of-concept
13

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
Fig. 8. (A) The average total cholesterol
plasma levels (g/L) of female C57BL/6
wild-type mice treated SC daily for
1.00
800
600
400
200
0
Day 0
Day 5
A
B
****
Day 10 14 days with either vehicle (black circles)
or 3f (3.28 or 16.4 mg/kg/day, inverted
Day 15
**
yellow triangles and blue triangles, re-
spectively) shown as a function of time.
In parallel, a separate cohort was dosed
with evolocumab (10 mg/kg in 100%
saline) once every 5 days (red squares).
Female PCSK9−/− knockout mice were
also treated in parallel with vehicle
(green circles) or 16.4 mg/kg/day 3f
(grey squares) as indicated (n = 5 per
treatment group). Error bars represent
the S.D. of 5 replicates. An unpaired stu-
dent t test (two-tailed) was applied to
each sample within the evolocumab and
3f (16.4 mg/kg/day) treated test groups
using the equivalent vehicle only control
0.75
**
*
**
****
****
0.50
0.25
0.00
0
1
2
3
4
5
10
15
as the comparator. P values < 0.05 are
designated with asterisks: **** < 0.0001,
** < 0.01, * < 0.05. (B) A subset of the
Vehicle
Evoloc
3f(16.4mg) 3f(3.28mg)
Time (days)
plasma samples described in (A) were concurrently assayed for their total PCSK9 levels. Each column represents average PCSK9 plasma levels (ng/mL) as a function
of time (0, 5, 10 and 15 days as indicated) and are grouped according to their treatment as detailed in (A). “Evoloc” indicates the cohort treated with 10 mg/kg
evolocumab every 5 days while “3f (16.4 mg)” and “3f (3.28 mg)” designate groups that were treated with 16.4 or 3.28 mg/kg/day 3f, respectively. Error bars
represent the SEM of 5 replicates. An unpaired student t test (two-tailed) was applied to each sample using the equivalent vehicle only control as a comparator. P
values < 0.05 are designated with asterisks: **** < 0.0001, ** < 0.01.
studies, although there is also the possibility of drug accumulation over
several days. With this in mind, a lower dose regimen (4 mg/kg/day)
was incorporated into proof-of-concept studies alongside 20 mg/kg/
day2.
suggesting that the decrease of ~10% observed in the wild-type cohort
treated with 16.4 mg/kg/day 3f was mediated via a PCSK9-dependent
mechanism. The levels of circulating triglycerides were largely un-
affected by any treatment (Supplementary Fig. S3). It should be noted
that no change in mouse body weight in any treatment cohort was
observed nor was there any change in behaviour for each experimental
group (data not shown).
3.8. Compound 3f lowers total cholesterol levels in the plasma of wild-type
mice
While 5 days in-life is sufficient for evolocumab to achieve a decrease
in total cholesterol (Fig. 8A), a longer period was required for compound
3f at 16.4 mg/kg/day to significantly reduce total cholesterol in mice
plasma (day 10, Fig. 8A). This can readily be explained by the antibody’s
improved pharmacological profile (e.g. likely a greater affinity for
PCSK9, [compare Fig. 5A and B]) relative to that of the small molecule
inhibitor 3f. That compound 3f required 10 days of in-life treatment at
16.4 mg/kg/day to induce a statistically significant decrease in total
cholesterol (Fig. 8A) may reflect a requirement by the small molecule
inhibitor to accumulate at sufficient levels in a suitable form (e.g. the
fraction of 3f unbound by other non-PCSK9 plasma proteins) to effect a
decrease in total cholesterol. This point is a subject of ongoing studies.
In parallel with the analysis of samples for total cholesterol, plasma
samples were also assayed for their levels of PCSK9 protein. Fig. 8B
shows that two conditions, namely exposure to evolocumab (10 mg/kg/
5 days on day 15) and compound 3f (16.4 mg/kg/day on day 10),
significantly increased the levels of circulating PCSK9 beyond those
observed in vehicle-treated controls. An increase in circulating PCSK9
levels may be attributed to drug-induced impairment of PCSK9 uptake
by the LDLR of liver cells. More broadly, the results suggest that both
evolocumab and compound 3f are engaging PCSK9 and suppressing its
function in vivo as intended. It is notable that a similar effect was ob-
served by Zhang et al.45 who reported a large increase in circulating
PCSK9 in the plasma of mice treated with their monoclonal antibody
(1B20) against PCSK9. Mice exposed to 10 mg/kg/day 1B20 demon-
strated an increase of ~750% in total plasma PCSK9 levels, a result that
was ascribed to the sequestration of PCSK9 within plasma by the an-
tibody.45 It is highly likely that a similar response was recapitulated in
the present study by evolocumab and compound 3f. Collectively, these
results have established a proof-of-concept that the impairment of
PCSK9 by a small organic molecule can yield a significant, albeit
modest decrease in total plasma cholesterol in vivo.
Wild-type mice are often considered to be sub-optimal as models for
LDL-c-lowering therapies as the majority of their total plasma choles-
terol is carried as HDL-c (80–90%) with a low fraction incorporated as
LDL-c41,42, a profile that is typically inverted relative to the human
state. Despite this limitation, the use of wild-type mice is appropriate in
this context as HDL-c levels (and by proxy, total cholesterol levels) are
modulated by the LDLR as they contain apolipoprotein E, a ligand for
the LDLR43. Moreover, the putative binding site of compound 3f is fully
conserved between human, mouse and other mammalian forms of
PCSK9 (Supplementary Fig. S1), making cross-species testing feasible.
In an effort to validate wild-type mice as an appropriate in vivo
model, female C57BL/6 wild-type mice were treated SC with evolo-
cumab (10 mg/kg every 5 days). Fig. 8A demonstrates that evolocumab
decreased total cholesterol plasma levels by ~30% relative to vehicle-
only controls on the final day of the experiment. Since evolocumab is an
FDA-approved monoclonal antibody raised against PCSK9 and used to
lower LDL-c in humans with hypercholesterolemia,34 it can be con-
cluded that the murine model was appropriate for the detection of
agents that lower LDL-c levels. In parallel, mice treated with 3.28 mg/
kg/day 3f exhibited no decrease in total cholesterol (Fig. 8A), however
a cohort exposed to 16.4 mg/kg/day 3f displayed a modest yet sig-
nificant decrease in total cholesterol plasma levels of ~10% (Fig. 8A).
In a result consistent with literature values,44 PCSK9 knockout mice
demonstrated typically low cholesterol levels throughout the time
course of the experiment (Fig. 8A). Notably, PCSK9−/− mice treated
with 16.4 mg/kg/day 3f did not exhibit any further decrease in total
cholesterol relative to vehicle-only treated PCSK9-/- controls (Fig. 8A),
2 While it was intended that mice were to be dosed with 4 and 20 mg/kg/day,
human error led to the use of 3.28 and 16.4 mg/kg/day instead.
14

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
Fig. 9. (A–D) The location of the putative binding site of nilotinib in the context of pre-existing PCSK9 X-ray crystal structures represented as cartoons. The colour
scheme from panel (A) has been maintained with the residue side-chains lining the putative nilotinib binding site highlighted in copper orange and emphasised as
dotted atomic spheres. (B, C) The locus of the nilotinib binding site as shown in the X-ray crystal structure (PDB code) 4NMX.10 The green PCSK9 catalytic domain is
in complex with a peptide inhibitor designated Pep2-8 and shown in dark teal as dotted atomic spheres.10 The complex is presented in the absence (B) and presence
(C) of the prodomain displayed in red. Note that the crystal structure does not include the C-terminal domain of PCSK9. (D) The binding site of nilotinib shown in the
context of the PCSK9-LDLR X-ray crystal structure complex as detailed by Lo Surdo et al. (28 and PDB code 3P5C). The LDLR is depicted in purple and was crystallised
with mature PCSK9 as a truncation of the full-length receptor.28
3.9. The interaction of compound 3f with PCSK9
4NMX and10). Pep(2–8) disrupts LDLR binding through a direct inter-
action with the EGF-A binding region of PCSK9 (Figs. 1A and 3A), a
locale that is clearly distinct from the putative binding site of nilotinib
(compare Fig. 9A and B). Moreover, a comparison of the binding site in
the absence (Fig. 9B) and presence (Fig. 9C) of the prodomain de-
monstrates a clear overlap in the three-dimensional space occupied by
niltonib and the prodomain, suggesting a source of potential competi-
tion for binding to PCSK9. It is currently unclear how compound 3f and
analogues engage the catalytic domain when it is pre-complexed with
Having established proof-of-concept with 3f in vivo (Fig. 8), it is
reasonable to briefly speculate on the interaction of the compound with
PCSK9 at a molecular level. Fig. 9 provides a direct comparison of the
modelled putative binding site (using nilotinib as an exemplar ligand)
in the catalytic domain of PCSK9 (Fig. 9A) with existing X-ray crystal
structures of the protein including PCSK9 in complex with a peptide
inhibitor (Pep2-8) developed by Genentech (Fig. 9B and C, PDB code
15

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
its prodomain, as was the case in the binding assay and as circulating
PCSK9 in vivo.1,3 In any case, it is clear that compound 3f engages
PCSK9 to effect disruption of its binding with LDLR (Table 3 and
Fig. 9D). X-ray crystallography studies of the 3f-PCSK9 complex are
underway and results will be reported in due course.
PCSK9 protein binding to the low density lipoprotein receptor. J Biol Chem.
2014;289:942–955.
11. Schroeder CI, Swedberg JE, Withka JM, et al. Design and synthesis of truncated EGF-
A peptides that restore LDL-R recycling in the presence of PCSK9 in vitro. Chem Biol.
2014;21:284–294.
12. Zhang Y, Ultsch M, Skelton NJ, et al. Discovery of a cryptic peptide-binding site on
PCSK9 and design of antagonists. Nat Struct Mol Biol. 2017;24:848–856.
13. Londregan AT, Wei L, Xiao J, et al. Small molecule proprotein convertase subtilisin/
kexin type 9 (PCSK9) inhibitors: Hit to lead optimization of systemic agents. J Med
Chem. 2018;61:5704–5718.
3.10. Conclusions
14. Dadu RT, Ballantyne CM. Lipid lowering with PCSK9 inhibitors. Nat Rev Cardiol.
2014;11:563–575.
In summary, a small molecule inhibitor of the PCSK9-LDLR inter-
action has been identified following the discovery that nilotinib mod-
estly disrupted the interaction. A subsequent medicinal chemistry
campaign focused on minimising the off-target kinase inhibitory ac-
tivity of nilotinib, promoting activity in disrupting the PCSK9-LDLR
interaction and introducing structural features intended as favourable
in the context of pharmacokinetic properties. Compound 3f emerged
from the campaign by demonstrating disruption of the PCSK9-LDLR
interaction at nanomolar levels in vitro and little inhibitory activity
against a small panel of tyrosine kinases. The compound restored LDL
uptake by liver cells cultured in the presence of PCSK9 and demon-
strated excellent bioavailability when delivered subcutaneously in
mice. Most significantly, compound 3f lowered total cholesterol levels
in the plasma of wild-type mice, thereby providing proof-of-concept
that the notion of a small molecule inhibitor against PCSK9 is ther-
apeutically viable.
15. Sabatine MS, Giugliano RP, Wiviott SD, et al. Open-label study of long-term eva-
luation against, efficacy and safety of evolocumab in reducing lipids and cardio-
vascular events. N Engl J Med. 2015;372:1500–1509.
16. Lintner NG, McClure KF, Petersen D, et al. Selective stalling of human translation
through small-molecule engagement of the ribosome nascent chain. PLoS Biol.
2017;15:e2001882.
17. Taechalertpaisarn J, Zhao B, Liang X, Burgess K. Small molecule inhibitors of the
PCSK9.LDLR interaction. J Am Chem Soc. 2018;140:3242–3249.
18. Fitzgerald K, Frank-Kamenetsky M, Shulga-Morskaya S, et al. Effect of an RNA in-
terference drug on the synthesis of proprotein convertase subtilisin/kexin type 9
(PCSK9) and the concentration of serum LDL cholesterol in healthy volunteers: a
randomised, single-blind, placebo-controlled, phase 1 trial. Lancet. 2014;383:60–68.
19. van Poelgeest EP, Hodges MR, Moerland M, Tessier Y, Levin AA, Persson R, Lindholm
MW, Dumong Erichsen K, Orum H, Cohen AF, Burggraaf J. Antisense-mediated re-
duction of proprotein convertase subtilisin/kexin type 9 (PCSK9): a first-in-human
randomized, placebo-controlled trial. Br J Clin Pharmacol. 2015;80:1350–1361.
20. Ding Q, Strong A, Patel KM, et al. Permanent alteration of PCSK9 with in vivo
CRISPR-Cas9 genome editing. Circ Res. 2014;115:488–492.
21. Wang X, Raghavan A, Chen T, et al. CRISPR-Cas9 targeting of PCSK9 in human
hepatocytes in vivo-brief report. Arterioscler Thromb Vasc Biol. 2016;36:783–786.
22. Petersen DN, Hawkins J, Ruangsiriluk W, et al. A small-molecule anti-secretagogue of
PCSK9 targets the 80S ribosome to inhibit PCSK9 protein translation. Cell Chem Biol.
2016;23:1362–1371.
Declaration of Competing Interest
23. McClure KF, Piotrowski DW, Petersen D, et al. Liver-targeted small-molecule in-
hibitors of proprotein convertase subtilisin/kexin type 9 synthesis. Angew Chem Int Ed
Engl. 2017;56:16218–16222.
B.J.E., A.K.S., J.P., G.E.K. and J.B. are employees of Nyrada Inc., a
subsidiary of Noxopharm Ltd. G.E.K. is the founder and an Executive
Chairperson of Noxopharm Ltd., the founder and a Non-Executive
Director of Nyrada Inc., and a shareholder of both Noxopharm Ltd. and
Nyrada Inc. I.D. is a Non-Executive Director and shareholder of
Noxopharm Ltd. and is the owner and Director of Altnia Group, a
shareholder of Nyrada Inc. J.T.P. and G.L. are members of the Scientific
Advisory Board of Nyrada Inc. B.J.E, J.T.P., G.L., I.D., G.E.K and J.B.
have share/stock options in Nyrada Inc. J.T.P., H.T., J.Z. and I.D. are
listed as inventors on the patent that discloses a subset of the com-
pounds described in this manuscript. J.T.P., G.L., B.N., K.C., H.T., J.Z.,
Q.Z., J.W., Y.T., W.T., Y.X., A.K.R and S.K. are employees or members of
their respective companies or laboratories who were contracted and/or
consulted by Nyrada Inc. to perform specific research activities.
24. Xu S, Luo S, Zhu Z, Xu J. Small molecules as inhibitors of PCSK9: Current status and
future challenges. Eur J Med Chem. 2019;162:212–233.
25. Min DK, Lee HS, Lee N, et al. In silico screening of chemical libraries to develop
inhibitors that hamper the interaction of PCSK9 with the LDL receptor. Yonsei Med J.
2015;56:1251–1257.
26. Gilson MK, Liu T, Baitaluk M, et al. A public database for medicinal chemistry,
computational chemistry and systems pharmacology. Nucl Acids Res.
2015;44(2016):D1045–D1053.
27. MedChemSoft, Solutions. Modeling kinase inhibition: Our predictions vs crystal structure.
2016 (accessed July 09, 2019).
28. Lo Surdo P, Bottomley MJ, Calzetta A, et al. Mechanistic implications for LDL re-
ceptor degradation from the PCSK9/LDLR structure at neutral pH. EMBO Rep.
2011;12:1300–1305.
29. Berman HM, Westbrook J, Feng Z, et al. The Protein Data Bank. Nucl Acids Res.
2000;28:235–242.
30. Kee AJ, Chagan J, Chan JY, et al. On-target action of anti-tropomyosin drugs reg-
ulates glucose metabolism. Sci Rep. 2018;8:4604.
31. Manley PW, Drueckes P, Fendrich G, et al. Extended kinase profile and properties of
the protein kinase inhibitor nilotinib. Biochim Biophys Acta. 1804;2010:445–453.
32. Cai H, Smith DA, Memarzadeh S, Lowell CA, Cooper JA, Witte ON. Differential
transformation capacity of Src family kinases during the initiation of prostate cancer.
Proc Natl Acad Sci U S A. 2011;108:6579–6584.
Appendix A. Supplementary material
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.bmc.2020.115344.
33. Xing L, Rai B, Lunney EA. Scaffold mining of kinase hinge binders in crystal structure
database. J Comput Aided Mol Des. 2014;28:13–23.
References
34. Fala L. Repatha (Evolocumab): second PCSK9 Inhibitor approved by the FDA for
patients with familial hypercholesterolemia. Am Health Drug Benefits.
2016;9:136–139.
1. Awan Z, Baass A, Genest J. Proprotein convertase subtilisin/kexin type 9 (PCSK9):
lessons learned from patients with hypercholesterolemia. Clin Chem.
2014;60:1380–1389.
35. Gibbs JP, Doshi S, Kuchimanchi M, et al. Impact of target-mediated elimination on
the dose and regimen of evolocumab, a human monoclonal antibody against pro-
protein convertase subtilisin/kexin type 9 (PCSK9). J Clin Pharmacol.
2017;57:616–626.
2. Lambert G, Sjouke B, Choque B, Kastelein JJ, Hovingh GK. The PCSK9 decade. J Lipid
Res. 2012;53:2515–2524.
36. Abdel-Meguid SS, Elshourbagy N, Meyers H, Mousa SA. (WO2014150326) Anti-
proprotein convertase subtilisin kexin type 9 (Anti-PSCK9) compounds and methods of
using the same in the treatment and/or prevention of cardiovascular diseases. 2014.
37. Cunningham D, Danley DE, Geoghegan KF, et al. Structural and biophysical studies
of PCSK9 and its mutants linked to familial hypercholesterolemia. Nat Struct Mol Biol.
2007;14:413–419.
3. Horton JD, Cohen JC, Hobbs HH. PCSK9: a convertase that coordinates LDL cata-
bolism. J Lipid Res. 2009;50(Suppl):S172–S177.
4. Hampton EN, Knuth MW, Li J, Harris JL, Lesley SA, Spraggon G. The self-inhibited
structure of full-length PCSK9 at 1.9 A reveals structural homology with resistin
within the C-terminal domain. Proc Natl Acad Sci U S A. 2007;104:14604–14609.
5. Kwon HJ, Lagace TA, McNutt MC, Horton JD, Deisenhofer J. Molecular basis for LDL
receptor recognition by PCSK9. Proc Natl Acad Sci U S A. 2008;105:1820–1825.
6. Abifadel M, Varret M, Rabes JP, et al. Mutations in PCSK9 cause autosomal dominant
hypercholesterolemia. Nat Genet. 2003;34:154–156.
38. Fisher TS, Lo Surdo P, Pandit S, et al. Effects of pH and low density lipoprotein (LDL)
on PCSK9-dependent LDL receptor regulation. J Biol Chem. 2007;282:20502–20512.
39. Thedrez A, Blom DJ, Ramin-Mangata S, et al. Homozygous familial hypercholester-
olemia patients with identical mutations variably express the LDLR (low-density li-
poprotein receptor): implications for the efficacy of evolocumab. Arterioscler Thromb
Vasc Biol. 2018;38:592–598.
7. Cohen J, Pertsemlidis A, Kotowski IK, Graham R, Garcia CK, Hobbs HH. Low LDL
cholesterol in individuals of African descent resulting from frequent nonsense mu-
tations in PCSK9. Nat Genet. 2005;37:161–165.
40. Schwartz GG, Steg PG, Szarek M, et al. Investigators, alirocumab and cardiovascular
outcomes after acute coronary syndrome. N Engl J Med. 2018;379:2097–2107.
41. Wang X, Paigen B. Genetics of variation in HDL cholesterol in humans and mice. Circ
Res. 2005;96:27–42.
8. Cohen JC, Boerwinkle E, Mosley Jr TH, Hobbs HH. Sequence variations in PCSK9,
low LDL, and protection against coronary heart disease. N Engl J Med.
2006;354:1264–1272.
9. Mullard A. Nine paths to PCSK9 inhibition. Nat Rev Drug Discov. 2017;16:299–301.
10. Zhang Y, Eigenbrot C, Zhou L, et al. Identification of a small peptide that inhibits
42. Yin W, Carballo-Jane E, McLaren DG, et al. Plasma lipid profiling across species for
16

B.J. Evison, et al.
Bioorganic&MedicinalChemistry28(2020)115344
the identification of optimal animal models of human dyslipidemia. J Lipid Res.
9 (PCSK9): hepatocyte-specific low-density lipoprotein receptor degradation and
critical role in mouse liver regeneration. Hepatology. 2008;48:646–654.
45. Zhang L, McCabe T, Condra JH, et al. An anti-PCSK9 antibody reduces LDL-choles-
terol on top of a statin and suppresses hepatocyte SREBP-regulated genes. Int J Biol
Sci. 2012;8:310–327.
2012;53:51–65.
43. Marais AD. Apolipoprotein E in lipoprotein metabolism, health and cardiovascular
disease. Pathology. 2019;51:165–176.
44. Zaid A, Roubtsova A, Essalmani R, et al. Proprotein convertase subtilisin/kexin type
17