Design, Synthesis, and Testing of Potent, Selective Hepsin Inhibitors via Application of an Automated Closed-Loop Optimization Platform

Article abstract of DOI:10.1021/acs.jmedchem.7b01698

Hepsin is a membrane-anchored serine protease whose role in hepatocyte growth factor (HGF) signaling and epithelial integrity makes it a target of therapeutic interest in carcinogenesis and metastasis. Using an integrated design, synthesis, and screening platform, we were able to rapidly develop potent and selective inhibitors of hepsin. In progressing from the initial hit 7 to compound 53, the IC₅₀ value against hepsin was improved from ～1 μM to 22 nM, and the selectivity over urokinase-type plasminogen activator (uPA) was increased from 30-fold to >6000-fold. Subsequent in vitro ADMET profiling and cellular studies confirmed that the leading compounds are useful tools for interrogating the role of hepsin in breast tumorigenesis.

Full text of DOI:10.1021/acs.jmedchem.7b01698

Subscriber access provided by Kaohsiung Medical University
Article
Design, synthesis and testing of potent, selective hepsin inhibitors
via application of an automated closed-loop optimization platform
Shishir M Pant, Amanda Mukonoweshuro, Bimbisar Desai, Manoj K Ramjee,
Christopher Nicholas Selway, Gary John Tarver, Adrian G Wright, Kristian
Birchall, Timothy M Chapman, Topi A Tervonen, and Juha Klefström
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.7b01698 • Publication Date (Web): 27 Apr 2018
Downloaded from http://pubs.acs.org on April 28, 2018
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Design, synthesis and testing of potent, selective hepsin
inhibitors via application of an automated closedꢀloop
optimization platform
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Shishir M. Pant§#, Amanda Mukonoweshuro†a, Bimbisar Desai†b, Manoj K. Ramjee†*, Christopher
N. Selway†, Gary J. Tarver†c, Adrian G. Wright†, Kristian Birchall‡, Timothy M. Chapman‡*, Topi
A. Tervonen§#, Juha Klefström§#*
† Cyclofluidic Ltd, Biopark, Broadwater Road, Welwyn Garden City, AL7 3AX, UK, +44 (0)1707
358673
a current address: Charles River Laboratories International Inc., Chesterford Research Park, Little
Chesterford, Cambridge, CB10 1XL, UK
b current address: Virginia Commonwealth University ꢀ School Of Engineering, 601 W Main St
#331, Richmond, VA 23220, USA
c current address: New Path Molecular Research, Babraham Research Campus, Cambridge, CB22
3AT, UK
‡
LifeArc, Accelerator Building, Open Innovation Campus, Stevenage, SG1 2FX, UK, +44 (0)20
7391 2700
§
Cancer Cell Circuitry Laboratory, Research Programs Unit / Translational Cancer Biology &
Medicum, University of Helsinki, P.O. Box 63 (Street address: Haartmaninkatu 8),
00014 University of Helsinki, Finland
Page 1 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 43
1
2
3
4
5
6
7
8
9
ABSTRACT: Hepsin is a membraneꢀanchored serine protease whose role in hepatocyte growth
factor (HGF) signalling and epithelial integrity makes it a target of therapeutic interest in
carcinogenesis and metastasis. Using an integrated design, synthesis and screening platform we
were able to rapidly develop potent and selective inhibitors of hepsin. In progressing from the
initial hit 7 to compound 53, the IC50 value against hepsin was improved from ~1 µM to 22 nM
and the selectivity over urokinaseꢀtype plasminogen activator (uPA) was increased from 30ꢀfold to
>6000ꢀfold. Subsequent in vitro ADMET profiling and cellular studies confirmed that the leading
compounds are useful tools for interrogating the role of hepsin in breast tumorigenesis.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
INTRODUCTION
Hepsin is a type II transmembrane serine protease with a chymotrypsinꢀlike serine protease
domain containing the catalytic triad of His, Asp and Ser residues.1 It is frequently overexpressed
in prostate, ovarian, renal and breast cancers2ꢀ8 and in vitro and in vivo studies have revealed the
functional role of hepsin in tumour progression and metastasis.3,8ꢀ10 The prevalence and
tumorigenic potential of hepsin in epithelial cancers and its location on the cell surface makes it an
accessible cancer therapy target.
Several classes of inhibitors ranging from small molecules, peptides and antibodies have been
developed against hepsin.11ꢀ16 However, poor understanding of the actions of hepsin in the
transformation process has made it difficult to establish a causality between pharmacological
inhibition of hepsin activity and reversal of cancerous phenotype in vitro. Recent findings suggest
that deregulated hepsin has two major oncogenic activities: firstly, hepsin induces processing of
several cancer critical growth factors including proꢀhepatocyte growth factor (proꢀHGF) and proꢀ
macrophage stimulating protein (proꢀMSP) into their active forms, leading to activation of tyrosine
Page 2 of 43
ACS Paragon Plus Environment

Page 3 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
kinase receptors mesenchymal epithelial transition factor (MET) and recepteur d'origine nantais
(RON).17ꢀ20 Secondly, our results with animal tumour models and an inducible form of hepsin
suggest an important role for hepsin in disruption of epithelial cohesion junctions, which include
desmosomes and hemidesmosomes.3,4 These defects will lead to loss of epithelial integrity, which
promotes tumour invasion and metastasis.21 In a recent study, we identified a urokinaseꢀtype
plasminogen activator (uPA) inhibitor WXꢀUK122,23 as a small molecule able to inhibit hepsin
activity at low micromolar concentration, prevent hepsin inductionꢀdependent desmosome and
hemidesmosome defects and restrain hepsinꢀinduced activation of HGF.3 However, WXꢀUK1 is a
relatively weak hepsin inhibitor (IC50 ~3 ꢁM) and exhibits poor selectivity which limits its utility
as a molecular tool to study the oncogenic role of hepsin.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Here we set out to identify novel subꢀmicromolar potency hepsin inhibitors with improved
selectivity profiles that would be capable of suppressing both axes of oncogenic hepsin function:
growth factor activation and damage to cell junctions. In order to accelerate the generation of new
lead molecules, we utilised an integrated design, synthesis and screening platform developed by
Cyclofluidic. The small molecule lead generation cycle involves iterative molecular design,
chemical synthesis, biological assay and data analysis steps, which feeds into the next cycle. Using
conventional processes each cycle may take one or more weeks leading to extended timelines for
lead generation, whereas the Cyclofluidic platform dramatically reduces the cycle time to a period
of one to two hours and enables rapid and automated lead generation. A comprehensive
description of the platform has been published,24 but for this work it was configured as shown
(Fig. 1) with integrated reactant handler, flow chemistry system, LC/MS/ELSD analytics and onꢀ
line biological assays for hepsin and uPA, and an onꢀline Chromatographic Hydrophobicity Index
(CHI) LogD assay.25 We report here the application of the platform to the identification of potent
Page 3 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 43
1
2
3
4
5
6
7
8
9
hepsin inhibitors, which are able to prevent deregulated hepsinꢀdependent induction of growth
factor signalling and disruption of cellꢀcell junctions and therefore serve as valuable tools in
further probing the role of hepsin.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1. Schematic of the integrated synthesis and screening platform.
RESULTS AND DISCUSSION
Hit Identification. Hepsin shows a strong preference for an arginine residue at the P1 position of
its substrates,17,26 therefore a series of fourteen commercially available amidine and guanidineꢀ
based inhibitors were assayed against hepsin (Table 1). uPA was selected as our initial
counterscreen as we sought to identify a starting point which did not display the uPA inhibition
exhibited by WXꢀUK1 (compound 13).
Page 4 of 43
ACS Paragon Plus Environment

Page 5 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
Table 1. Hepsin screening results for a set of commercially available amidine and guanidine
derivatives with uPA counterscreen
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Compound
Number
Hepsin
IC50 (µM)
uPA IC50
(µM)
Structure
NH
NH2
H
N
N
H
OH
O
>10
>10
ND
ND
1
2
O
N
H2N
O
NH2
NH
H
N
NH
O
S
N
H
O
O
N
HO
O
0.029
0.383
4.02
ND
ND
ND
3
4
5
0.005
1.13
ND
>10
6
7
NH
H2N
0.676
2.43
8
O
S
O
O
N
N
H
S
N
H
O
Page 5 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 43
1
2
3
4
5
6
Compound
Number
Hepsin
IC50 (µM)
uPA IC50
(µM)
Structure
7
8
9
4.46
ND
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
NH2
NH
O
S
O
N
N
O
H
N
N
6.55
>10a
10
11
O
0.272
0.484
6.19
ND
12
3.35
3.14
0.285
>10
13
14
a Literature value27
Two compounds, Camostat28 3 and Nafamostat29 6 exhibited apparent potent inhibition, however it
has been reported30 that this class of compound act as slow tightꢀbinding substrates and exhibit
apparent inhibition by stabilising the acylꢀenzyme intermediate. This was not the desired mode of
action for a selective hepsin inhibitor and these compounds were not considered further.
Page 6 of 43
ACS Paragon Plus Environment

Page 7 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Compounds 4 and 11 were also not considered structurally attractive starting points; the next most
potent compounds 731,32 and 833 were of more interest and their potency was confirmed in the
assay. From these two, compound 7 provided a moderate level of potency, good selectivity for
hepsin over uPA and considerable scope for structural modification; it was therefore selected as a
suitable starting point for SAR exploration with the goal of optimizing hepsin potency alongside
selectivity against uPA.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
General Chemistry Methods. General methods for the batch and flow synthesis of compound 7
analogs are described below, and full details are provided in the supporting information.
Batch Method: Initial batch synthesis of compound 7 analogs34 of general structure 19 was
undertaken as illustrated in Scheme 1.
Scheme 1. Batchꢀbased synthetic route to compound 7 analogs
Reaction of a sulfonyl chloride 16 with an amino acid 15 in aqueous sodium hydroxide, followed
by acidification and extractive work up gave an intermediate sulfonamide 17. The crude
intermediate was reacted with an aminoamidine 18 using 2ꢀ(3Hꢀ[1,2,3]triazolo[4,5ꢀb]pyridinꢀ3ꢀ
yl)ꢀ1,1,3,3ꢀtetramethylisouronium hexafluorophosphate (HATU) as the coupling agent to give the
product 19. Regioselective reaction at the aniline moiety rather than the amidine of 18 was
ensured through the protection of the very basic amidine as a hydrochloride salt, readily achieved
Page 7 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 43
1
2
3
4
5
6
through the use of NꢀethylꢀNꢀisopropylpropanꢀ2ꢀamine (DIPEA) as reaction base. The desired
product was purified by reverse phase HPLC.
7
8
9
Flow Method: In order to successfully prepare a range of analogues utilising the CyclOps platform
a flow chemistry method was developed based on the batch synthesis (Scheme 2). Amino acids
are inherently insoluble in solvents more suited to flow synthesis; therefore the use of preformed
trimethyl silyl esters of the acids addressed both the insolubility of the amino acids and offered the
advantage of transient protection of the carboxylic acid functionality. No formal deꢀprotection
step was required as the ester proved labile under the reaction conditions.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Scheme 2. Flowꢀbased synthetic route to compound 7 analogs
Stock solutions A of silylated amino acid trimethylsilyl esters 20 were preꢀprepared in batch by
reaction of an amino acid 15 with chlorotrimethylsilane and DIPEA in dichloromethane under
microwave conditions for 10 to 20 min at 100°C. They were used without further purification by
reaction in a flow reactor with stock solution B of a sulfonyl chloride 16 in dichloromethane at
Page 8 of 43
ACS Paragon Plus Environment

Page 9 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
60°C and a total flow rate of 100 µL/min. After 20 min the intermediate sulfonamide product 21
eluted from reactor 1. It was immediately mixed under flow conditions with a stock solution C of
amidine dihydrochloride 18 in NMP containing 1 equivalent of DIPEA and stock solution D of
HATU and DIPEA in NMP. The reaction mixture was heated at 100°C in reactor 2 for 10 min at a
total flow rate of 200 µL/min to provide a solution of the final product 19, which was purified by
HPLC prior to screening.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The regioselective formation of the desired amide in the presence of the amidine was again
achieved by ensuring the more basic amidine was protected as the hydrochloride salt throughout
the coupling reaction and the two steps could be seamlessly telescoped together to enable the flow
synthesis of a range of structures. All flow based synthetic methods were undertaken using the
Vapourtec35 R4 chemistry apparatus.
The scope of the reaction described above was extended to cover synthesis of amide analogs 23
and 24 and also urea analogs 26 and 27, using acid chlorides 22 and isocyanates 25, respectively,
instead of the sulfonyl chloride in step 1, Scheme 3. In general, the relevant acid chloride or urea
building blocks were used as direct substitutes of the sulfonyl chlorides.
Scheme 3. Twoꢀstep amide and urea formation followed by amidoamidine formation from
amino acid trimethylsilyl esters in flow
Page 9 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 43
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Platform SAR Strategy. The integrated platform offers an efficient and effective means to rapidly
explore the SAR of a defined area of virtual chemistry space. Compound 7 can be readily
disconnected to three reaction components, each readily accessible through commercial suppliers.
A subset of available reactants was selected which when enumerated represent the virtual
chemistry space for the platform SAR generation (Scheme 4).
Scheme 4. Representation of the virtual chemistry space for platform SAR generation
O
Y
R1
R1
O
H
N
NH2
Y
R2
N
R3
OH
R3
R2 Cl
Y = C, SO
+
H2N
+
H
O
O
5 R2COCl
31 R2SO2Cl
8 amino amidines
5,472 enumerated products
19 amino acids
This set of 63 reactants (see Supporting Information for structures) represented a chemistry space
of 5,472 virtual molecules for initial exploration on the integrated platform. This initial set of
reactants was chosen based on commercial availability, synthetic compatibility and diversity.
Exploration of the SAR represented by this virtual chemical space was undertaken using a
multiparameter optimization method to select the predicted most potent compounds against the
Page 10 of 43
ACS Paragon Plus Environment

Page 11 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
primary target hepsin and with the best predicted selectivity over uPA. This multiparameter
objective was defined as maximising the predicted potency against hepsin (i.e. lower IC50 values)
and minimising the predicted potency for uPA (i.e. higher IC50 values) and is referred to as the
platform objective.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In order to maximise the exploration of the selected chemical reactants whilst seeking selective
and potent structures, the initial strategy chosen was Best Objective Under Sampled (BOUS). At
the start of a platform experiment the algorithm predicts IC50 values based on the prior knowledge
data set (the results of the commercial compound screen were initially used for this). The
desirability of each compound is then determined using the platform objective.
To determine the order of synthesis for the platform for each eligible virtual product, the total is
calculated of the number of times the constituting reactants have been used previously. The
compounds with the lowest total previous usage are considered to be under sampled, and
synthesizing compounds from this pool ensures all reactants are sampled as equally as possible.
At the end of each complete synthesis and screen cycle the prior knowledge data set is updated to
include the new inhibition values, the selection algorithm is repeated and the next compound for
synthesis and screening is selected for the platform. Each closed loop cycle took approximately 90
minutes on the platform and the process is fully automated (i.e. subsequent cycles continue
without any intervention). Following a set of automated experiments the developing SAR is
evaluated and a new set of experiments devised that may include changes in the reactant sets used
on the platform as well as the optimization strategy. In this work a series of five distinct sets of
automated experiments were carried out, and the results from these are described below as
Platform Experiments 1 to 5.
Page 11 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 43
1
2
3
4
5
6
7
8
9
Platform Experiments 1 – Best Objective UnderꢀSampled Experiment. An initial set of 24
closedꢀloop synthesis and screening experiments, using the BOUS strategy to select from within
the 5,472 virtual set, quickly established that compounds derived from 4ꢀaminobenzamidine
showed the best activity against hepsin and the other seven aminoamidines in the reactants set
were only weakly active in the examples made. A number of chemical reactivity trends were also
established, the main ones being that aromatic sulfonyl chlorides worked well, but in general
aliphatic sulfonyl chlorides failed to react, and hindered amino acids usually failed to react too. In
addition a number of reactants were found to be insoluble in the required solvents. Based on
reactivity and this initial SAR, a number of reactants were removed from the set and the
experiment was continued with a subset of thirteen amino acids, thirtyꢀtwo sulfonyl/acid chlorides
and five aminoamidines. A final total of sixtyꢀthree synthesis to screen cycles using BOUS were
attempted and fortyꢀfour gave reliable SAR data (70%) (see Supporting Information for
structures). Three compounds 28, 29 and 30 gave hepsin IC50 values below 1 µM, with uPA
activity generally around tenꢀfold weaker (Table 2). The CHI LogD in this series was also
improved compared to 7 (CHI LogD = 4.2). It was apparent that the highest hepsin inhibition was
observed in 4ꢀmethylbenzene sulfonamide derivatives of small aliphatic amino acids such as
leucine and norleucine, combined with paraꢀsubstituted aminobenzamidines. The matched pair 28
and 31 confirmed the importance of the amidine to binding; replacing it with a bromine led to a
large loss in activity.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 2. Selected results from Best Objective UnderꢀSampled experiment and aminoamidine
scan
Page 12 of 43
ACS Paragon Plus Environment

Page 13 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
Hepsin
Compound
IC50
uPA
IC50
(µM)
Selectivity CHI
Structure
Number
ratio
LogD
(µM)
7
8
9
0.5
0.9a
5
3a
10
3
1.2
28
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
0.9
0.6a
11
20a
12
33
1.2
29
0.9
58
8
9
1
1.4
30
31
58
NAb
73
19
0.26
1.1
32
a Repeat synthesis and screening data; b Data not available.
Platform Experiments 2 – Aminoamidine Scan. To further examine the effect of changing the
aminoamidine component an additional five compounds were synthesized and screened using the
platform, keeping 4ꢀmethylbenzene sulfonyl chloride and norleucine constant while varying the
aminoamidine moiety. The results confirmed the SAR in this region is very tight: for example
changing from the paraꢀ to the metaꢀaminobenzamidine derivative 32 (Table 2) resulted in a
significant loss of activity, as did replacing the amidine with a nitrile or Nꢀmethylation at the
amide (data not shown).
Page 13 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 43
1
2
3
4
5
6
7
8
9
Platform Experiments 3 – Chase Objective Strategy. Based on the SAR obtained up to this
point, the original set of 5,472 virtual compounds was refined down to a much smaller set of 297
virtual compounds derived from 27 sulfonylating/acylating agents, 11 amino acids and 1
aminoamidine. The new chemical space was efficiently explored using a Chase Objective strategy
to run multiple closedꢀloop synthesis and screening experiments, with 80% success rate overall.
The highest hepsin activity was seen in products derived from aliphatic amino acids; a summary of
these are shown Table 3. In particular, the product 33 derived from (R)ꢀcyclohexylglycine (Table
3) gave a hepsin IC50 of 33 nM and greater than one hundredꢀfold selectivity when initially run on
the platform, and similar values on repeat offꢀline screening of the collected fraction. CHI LogD
values are also within a generally acceptable range, although the properties are strongly influenced
by the highly basic amidine moiety.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 3. Selected results following Chase Objective experiments
Compound Hepsin uPA IC50 Selectivity CHI
Structure
Number IC50 (µM) (µM)
ratio
LogD
0.033
0.070a
0.038b
4.4
133
>142
>1000
>10a
>41b
1.3
33
O
O
H
N
R
S
0.3
1.1
12
22
40
20
1.3
1.2
34
35
N
H
O
O
NH2
NH
Page 14 of 43
ACS Paragon Plus Environment

Page 15 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
Compound Hepsin uPA IC50 Selectivity CHI
Structure
Number IC50 (µM) (µM)
ratio
LogD
6
7
8
9
0.7
16
23
1.5
36
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
37b
0.25
60
240
0.9
a Manual screen of platform synthesized sample; b Offꢀline synthesis and manual screen.
To confirm this result, 33 was synthesized offꢀline and assayed manually (Figure 2). The hepsin
IC50 was 38 nM and uPA IC50 was >41 µM, indicating >1000ꢀfold selectivity.
Figure 2. Hepsin and uPA IC50 curves for compound 33.
The synthesis of the racemic phenylglycine derivative 37 (Table 3) failed on the platform, so it
was synthesized offꢀline and assayed manually to compare activity of the phenyl and cyclohexyl
matched pair. The hepsin IC50 was 250 nM and uPA IC50 was 60 µM, indicating ~240ꢀfold
Page 15 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 43
1
2
3
4
5
6
7
8
selectivity, but overall about 7ꢀfold weaker against hepsin than the (R)ꢀcyclohexylglycine
compound 33. However, later evaluation of the (R)ꢀphenylglycine derivative 53 showed it to have
similar potency to 33.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Platform Experiments 4 – Amino Acid Scan. To further examine the effect of changing the
amino acid component, the synthesis and screening of a series of nine compounds was attempted
on the platform with 4ꢀmethylbenzene sulfonyl chloride and 4ꢀaminobenzamidine kept constant
while the amino acid was varied. Seven of these were successfully completed and the results,
shown in Table 4, indicated that the cyclopentyl derivative 39 had good selectivity and reasonable
hepsin activity although it was weaker than the cyclohexyl analogue 33 and phenyl analogue 37. It
was also clear that βꢀamino acid derivatives such as 38 and 40 were generally weaker than αꢀ
amino acids. Interestingly the tetrahydropyranylglycine derivative 45 was ~30ꢀfold weaker than
compound 33 indicating that polarity was not well tolerated at this position. Unfortunately the
syntheses of 42 from glutamine and 43 from dimethyl lysine failed, so the tolerance for other polar
side chains could not be tested.
Table 4. Hepsin and uPA activity data from variation of the amino acid
Hepsin uPA
Compound
Number
Selectivity
ratio
CHI
LogD
Structure
IC50
IC50
(µM) (µM)
3
1
0.3
1.6
38
Page 16 of 43
ACS Paragon Plus Environment

Page 17 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
Hepsin uPA
Compound
Number
Selectivity
ratio
CHI
LogD
Structure
IC50
IC50
(µM) (µM)
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
0.86
11
64
68
74
0.4
1.4
0.8
39
40
6
O
H
N
S
N
1.8
7.2
4
41
O
H
O
NH2
NH
42a
NR
NR
43a
44
45
NR
>19
1.4
NR
24
<1
1.8
0.4
>19
>13
Page 17 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 43
1
2
3
4
5
6
Hepsin uPA
Compound
Number
Selectivity
ratio
CHI
LogD
Structure
IC50
IC50
(µM) (µM)
7
8
9
3
28
9
3
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
0.5
1.3
46
1.5
5.3
0.22b
47
a Synthesis failed; b Manual synthesis and screening.
In addition, the (S)ꢀcyclohexylglycine derivative 47 (Table 4) was synthesized manually to provide
a comparison of the hepsin activity of the two enantiomers (Figure 3). The (R)ꢀcyclohexylglycine
derivative 33 was measured three times in sequential experiments and was slightly weaker in this
comparison than previously measured but the data suggests the (S)ꢀenantiomer is 2ꢀ to 3ꢀfold
weaker, but note the enantiomeric purity of the two samples has not been measured directly.
Page 18 of 43
ACS Paragon Plus Environment

Page 19 of 43
Journal of Medicinal Chemistry
1
2
3
4
Figure 3. Hepsin activity comparison of the (S) and (R) enantiomers of the cyclohexylglycine
5
6
b
derivatives. a Platform used to purify and collect sample with manual offꢀline screening; Repeat
7
8
of ‘a’; c Pure solid sample with manual screening.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Platform Experiments 5 – Extending the Chemical Space. To further explore the SAR a new
medicinal chemistry design was undertaken and a set of 186 virtual compounds, derived from 31
sulfonylating / acylating agents, 6 amino acids and a single aminoamidine (4ꢀaminobenzamidine),
was defined. The selection of the building blocks forming this virtual space was made based on an
evaluation of the existing SAR, commercial availability, synthetic tractability and also the desire to
extend into untapped areas, for example the acylating reagents included some isocyanates to yield
urea derivatives. The full set of reactants is shown in the Supporting Information. The platform
was run for 22 synthesisꢀtoꢀscreen cycles using BOUS. It was immediately apparent that only the
cyclohexylglycine and cyclopentylglycine derivatives showed good hepsin activity and selectivity.
Hence a Chase Objective experiment was set up with only cyclohexylglycine and
cyclopentylglycine amino acids along with the remaining sulfonylating / acylating agents. A
further 21 loops were run in this mode to extend the SAR. The analogs derived from pꢀ
fluorobenzenesulfonyl chloride 48 and pꢀchlorobenzenesulfonyl chloride 49 showed some of the
best activity from this set against hepsin and good selectivity over uPA (Table 5). The matched
cyclohexylglycine triplet of the sulfonamide 33, amide 50 and urea 51 show the activity sequence
sulfonamide >> urea > amide. The larger pꢀtrifluoromethylphenylsulfonamide 52 showed reduced
potency versus its methyl analogue 33.
Table 5. Examples of active compounds and matched pairs synthesized on the platform
Page 19 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 43
1
2
3
4
5
6
Hepsin uPA
Compound
Number
Selectivity
ratio
CHI
LogD
Structure
IC50
IC50
(µM) (µM)
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
0.09
0.10
>14
18
>155
180
NA
1.5
48
49
5.0
24.5
43
5
22
7
1.7
1.5
1.7
50
51
52
O
H
N
R
2
N
H
N
H
O
NH2
NH
0.68
4.7
Offꢀline synthesis of samples for further analysis and testing. Four inhibitors spanning a range
of potencies were synthesized offꢀline in batch mode and assayed manually against hepsin and
uPA (alongside 7 and 8 as standards) in order to allow comparison with the results obtained on the
Page 20 of 43
ACS Paragon Plus Environment

Page 21 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
platform and to generate sufficient solid sample for further testing. This group included the
product derived from (R)ꢀphenylglycine, compound 53 (Table 6). The IC50 values are shown in
Table 6 and the initial IC50 curves are shown in Figure 4. The most potent compound, the
phenylglycine derivative 53 initially showed a potency of 4 nM, but the curve was not a complete
sigmoidal shape at the lower concentration, so a second assay covering lower concentrations was
run for this compound (Figure 5). The curve showed a better shape and the potency was 22 nM
against hepsin with >6000 fold selectivity over uPA. The (R)ꢀcyclohexylglycine derivative 33 was
slightly weaker than 53 but still shows excellent potency and selectivity.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 6. Hepsin and uPA activity of four selected compounds synthesized offꢀplatform with
two original hit compounds included as standards
Hepsin
IC50
(µM)
uPA
IC50
(µM)
Compound
Number
Structure
Selectivity
0.056
0.19
1946
280
34000
33
48
1450
0.85
>1946
>2300
52
Page 21 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 43
1
2
3
4
5
6
Hepsin
IC50
(µM)
uPA
IC50
(µM)
Compound
Number
Structure
Selectivity
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
0.022
0.87
140
26
6363
53
7
30
18
8
0.33
6
Page 22 of 43
ACS Paragon Plus Environment

Page 23 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
Figure 4. Hepsin and uPA IC50 curves for four compounds synthesized offꢀplatform with two
original hit compounds as standards
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 5. IC50 curve for compound 53 against hepsin.
Selectivity & in vitro ADMET profiling. The leading compounds were profiled against a panel of
10 serine proteases to assess their broader selectivity (Table 7). The three compounds tested
displayed similar selectivity profiles, with the most potent hepsin inhibitor 53 exhibiting 54% and
34% inhibition of trypsin and plasma kallikrein respectively at 1 ꢁM inhibitor concentration.
Table 7. Serine protease selectivity profiling of key compounds; values given are %
inhibition (mean of 2 experiments) at compound concentrations of 1 µM and 10 µM
Serine Protease, % inhibition
at given compound
concentration
33
48
53
1 µM 10 µM 1 µM 10 µM 1 µM 10 µM
Thrombin
11
NI
NI
NI
NI
52
12
NI
NI
NI
NI
98
NI
NI
NI
NI
NI
40
10
NI
NI
NI
NI
91
NI
NI
NI
NI
NI
27
44
NI
NI
NI
NI
80
Chymase
Elastase
Cathepsin G
Dipeptidyl peptidase
Tryptase
Page 23 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 43
1
2
3
4
5
6
7
8
9
Trypsin
83
NI
NI
NI
18
NI
96
29
NI
NI
22
NI
77
12
NI
NI
10
NI
97
34
NI
NI
24
NI
54
34
NI
NI
16
NI
93
85
NI
NI
24
NI
Plasma kallikrein
ELA2 (Neutrophil Elastase 2)
Matriptase
HGFA
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Factor Xa
NI = no significant inhibition observed (<10%)
Compounds 33, 48 and 53 were also profiled in in vitro ADMET assays (Table 8). Due to the
influence of the highly basic benzamidine group, all three compounds showed high solubility and
no PAMPA permeability; however, given the extracellular location of the protease domain of
hepsin, the lack of permeability observed does not limit their application in cellꢀbased assays.
Metabolic stability was acceptable in both mouse and human microsomes, with the exception of
the low stability of 48 in human microsomes. The compounds showed no cytotoxicity in HepꢀG2
cells up to 10 µM concentration.
Table 8. In vitro ADMET profiling of key compounds
Compound number
33
48
53
Kinetic solubility (µM)
MLM t1/2 (min)
>200
74
200
48
>200
108
53
HLM t1/2 (min)
89
16
PAMPA Papp (nm/s)
Cytotox HepꢀG2 viableꢀ50 (µM)
0
0
0
>10
>10
>10
MCF10AꢀIndu20ꢀhepsin cell based assay. We employed nonꢀtransformed mammary epithelial
MCF10A cells expressing doxycycline (Dox) inducible hepsin (MCF10AꢀIndu20ꢀhepsin) as a
Page 24 of 43
ACS Paragon Plus Environment

Page 25 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
cellꢀbased assay to test whether the key compounds inhibit oncogenic hepsin functions. In this cell
system, addition of Dox to the cells induces a level of hepsin protein expression and activity that is
comparable to breast tumour cell lines.3 The inducible system for hepsin expression is described in
Figure 6 (aꢀc). Hepsin induction leads to downꢀmodulation of its cognate inhibitor HAIꢀ1, which
serves as one readout of oncogenic hepsin activity (Figure 6b).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 6. Doxycyclineꢀinducible hepsin expression in nonꢀtransformed mammary epithelial
MCF10A cells. a) Schematic presentation of pINDUCER20ꢀHPN (Indu20ꢀHPN) vector construct.
b) Doxycycline induction of Indu20ꢀHPN in MCF10A (MCF10AꢀIndu20ꢀhepsin) cells leads to
increased expression of processed forms of hepsin and down modulation of HAIꢀ1. c) Doxꢀ
Page 25 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 43
1
2
3
4
5
6
induction leads to increased cleavage of hepsin selective substrate BOCꢀQRRꢀAMC. For hepsin
induction cells were treated with 100 ng/mL Dox for 24 h. *p< 0.05, **p< 0.01, ***p< 0.001.
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
We used the MCF10AꢀIndu20ꢀhepsin cells to determine the cellular IC50 values of the compounds
against hepsin using fluorogenic BOCꢀQRRꢀAMC peptide as a hepsin selective substrate. All four
compounds displayed IC50 values in the range of 0.10ꢀ3.15 µM. Compound 53 was the most potent
inhibitor of hepsin with an IC50 of 0.10 µM and is therefore significantly more potent than WXꢀ
UK1 (IC50 6.49 µM). Compound 33 and 48 inhibited hepsin with IC50 values of 0.99 µM and 1.29
µM respectively, while 52 was the least potent inhibitor with an IC50 of 3.15 µM (Figure 7a,b).
The proteolytic activity of hepsin is regulated by its cognate inhibitor HAIꢀ1 at the cell surface and
oncogenic levels of hepsin lead to enhanced shedding of HAIꢀ1 from the cell membrane, a
phenomenon not seen with the overexpression of catalytically inactive hepsin.3 To test whether the
compounds were able to prevent the loss of HAIꢀ1, MCF10AꢀIndu20ꢀhepsin cells were treated
with the compounds prior to induction of hepsin with doxycycline. In these experiments,
compound 53 prevented the hepsinꢀinduced loss of HAIꢀ1 at 10 µM concentration. We have
previously shown that WXꢀUK1 inhibits HAIꢀ1 down modulation at 20 µM concentration.
However, neither WXꢀUK1 nor any of the other compounds was able to efficiently prevent HAIꢀ1
loss at 10 µM concentration, which is consistent with the superior potency of 53 seen in
biochemical and cell based IC50 assays (Figure 7c).
Page 26 of 43
ACS Paragon Plus Environment

Page 27 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 7. Inhibition of the proteolytic activity of hepsin. a) Determining the IC50 from the cellular
doseꢀresponse curve. The inhibitory action of compounds 33, 48, 52, 53 and WXꢀUK1 was
measured in MCF10AꢀIndu20ꢀhepsin cells using BOCꢀQRRꢀAMC as a substrate. Due to the
higher potency of 53, the IC50 curve was generated with a broader concentration range. b) IC50
values of the tested compounds. c) Inhibition of hepsinꢀdependent downꢀmodulation of cellular
HAIꢀ1. MCF10AꢀIndu20ꢀhepsin cells were preꢀtreated with either DMSO or the compounds for 30
min followed by induction of hepsin with 100 ng/mL doxycycline for 6 h. HAIꢀ1 protein levels
were analyzed by immunoblot.
Inhibition of hepsinꢀdependent activation of HGF/MET signalling pathway. Hepatocyte
growth factor is produced in its latent form (ProꢀHGF) by mesenchymal cells and requires a
Page 27 of 43
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 43
1
2
3
4
5
6
7
8
9
proteolytic cleavage for conversion into active α and β chains.36 Several studies have shown that
hepsin activates ProꢀHGF, consequently activating MET receptor tyrosine kinase.3,17,37 The
HGF/MET pathway is associated with poor prognosis and acquired drug resistance in several
cancers.38,39 We examined whether the lead compounds inhibit hepsinꢀmediated processing of
recombinant human ProHGF (rProꢀHGF) by incubating rProꢀHGF with recombinant human
hepsin (rhHPN) in the presence of inhibitors. Compound 53 demonstrated the highest inhibition of
rProꢀHGF activation shown by a decrease in α and β chain processing (Figure 8a). To further
evaluate the capacity of our compounds to prevent HGF/MET signalling, we induced hepsin with
Dox in MCF10AꢀIndu20ꢀhepsin cells and treated them with the inhibitors and rProꢀHGF.
Compound 53 displayed the best potency against HGF/MET activation as demonstrated by the
decrease in MET phosphorylation (Figure 8b).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
We have recently shown that hepsin protein is overexpressed in all major subtypes of breast
cancer, with the highest frequency of overexpression in triple negative breast cancer (TNBC).3 We
tested the efficacy of the inhibitors on HGF/MET signalling in a TNBC cell line Hs578T, which
expresses high levels of proteolytically active hepsin and low levels of HAIꢀ1.3 Treatment of
Hs578T cells with rProꢀHGF led to an increase in MET phosphorylation, which was inhibited by
all four compounds (Figure 8c). Consistent with our previous results, 53 demonstrated superior
potency compared to the other compounds (Figure 8c). We confirmed the results with 53 in two
TNBC cell lines, Hs578T and HCC1937. In these assays, 1 µM 53 inhibited MET phosphorylation
similarly to the clinically tested MET inhibitor crizotinib at 100 nM (Figure 8d,e).
Page 28 of 43
ACS Paragon Plus Environment

Page 29 of 43
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Page 29 of 43
ACS Paragon Plus Environment