Improving the Selectivity of Engineered Protease Inhibitors: Optimizing the P2 Prime Residue Using a Versatile Cyclic Peptide Library

Article abstract of DOI:10.1021/acs.jmedchem.5b01148

Standard mechanism inhibitors are attractive design templates for engineering reversible serine protease inhibitors. When optimizing interactions between the inhibitor and target protease, many studies focus on the nonprimed segment of the inhibitor's binding loop (encompassing the contact β-strand). However, there are currently few methods for screening residues on the primed segment. Here, we designed a synthetic inhibitor library (based on sunflower trypsin inhibitor-1) for characterizing the P2′ specificity of various serine proteases. Screening the library against 13 different proteases revealed unique P2′ preferences for trypsin, chymotrypsin, matriptase, plasmin, thrombin, four kallikrein-related peptidases, and several clotting factors. Using this information to modify existing engineered inhibitors yielded new variants that showed considerably improved selectivity, reaching up to 7000-fold selectivity over certain off-target proteases. Our study demonstrates the importance of the P2′ residue in standard mechanism inhibition and unveils a new approach for screening P2′ substitutions that will benefit future inhibitor engineering studies.

Full text of DOI:10.1021/acs.jmedchem.5b01148

Subscriber access provided by UNIV OF NEBRASKA - LINCOLN
Article
IMPROVING THE SELECTIVITY OF ENGINEERED
PROTEASE INHIBITORS: OPTIMIZING THE P2 PRIME
RESIDUE USING A VERSATILE CYCLIC PEPTIDE LIBRARY
Simon John de Veer, Conan K Wang, Jonathan Malcolm Harris, David J Craik, and Joakim Erik Swedberg
J. Med. Chem., Just Accepted Manuscript • Publication Date (Web): 22 Sep 2015
Downloaded from http://pubs.acs.org on September 28, 2015
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Medicinal Chemistry is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
IMPROVING THE SELECTIVITY OF ENGINEERED PROTEASE INHIBITORS:
OPTIMIZING THE P2 PRIME RESIDUE USING A VERSATILE CYCLIC PEPTIDE LIBRARY
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
Simon J. de Veer†, Conan K. Wang‡, Jonathan M. Harris†, David J. Craik‡, Joakim E. Swedberg‡,*
†Institute of Health and Biomedical Innovation, Queensland University of Technology, Brisbane QLD
4059, Australia
‡Institute for Molecular Bioscience, The University of Queensland, Brisbane QLD 4072, Australia
*Corresponding author: Dr Joakim Swedberg, Institute for Molecular Bioscience, The University of
Queensland, Brisbane QLD 4072, Australia. Email: j.swedberg@imb.uq.edu.au
Page 1 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 42
1
2
3
4
5
6
ABSTRACT
7
8
9
Standard mechanism inhibitors are attractive design templates for engineering reversible serine protease
inhibitors. When optimizing interactions between the inhibitor and target protease, many studies focus on
the nonꢀprimed segment of the inhibitor’s binding loop (encompassing the contact βꢀstrand). However,
there are currently few methods for screening residues on the primed segment. Here, we designed a
synthetic inhibitor library (based on sunflower trypsin inhibitorꢀ1) for characterizing the P2ʹ specificity of
various serine proteases. Screening the library against thirteen different proteases revealed unique P2ʹ
preferences for trypsin, chymotrypsin, matriptase, plasmin, thrombin, four kallikreinꢀrelated peptidases,
and several clotting factors. Using this information to modify existing engineered inhibitors yielded new
variants that showed considerably improved selectivity, reaching up to 7,000ꢀfold selectivity over certain
offꢀtarget proteases. Our study demonstrates the importance of the P2ʹ residue in standard mechanism
inhibition, and unveils a new approach for screening P2ʹ substitutions that will benefit future inhibitor
engineering studies.
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 2 of 41
ACS Paragon Plus Environment

Page 3 of 42
Journal of Medicinal Chemistry
1
2
3
4
INTRODUCTION
5
6
7
8
9
Serine proteases are established therapeutic targets in a spectrum of medical conditions, including
coagulation and cardiovascular diseases, metabolic disorders, inflammation, and infectious diseases.1
Accordingly, the pharmaceutical industry has had a longstanding interest in developing therapeutic
molecules that can modulate the activity of one or more proteases, with most attention directed towards
protease inhibitors.2 These drug discovery programs have provided a number of major successes, and an
array of inhibitors targeting different proteases are currently used in the clinic,2 including several new
drugs on the market.3 Traditionally, proteaseꢀbased drug development has favored classical small
molecule inhibitors as this approach often yields suitable lead compounds that have excellent
pharmacokinetic properties. However, small molecule inhibitors typically have relatively few binding
determinants, which can lead to challenges when translating drug leads from the laboratory to the clinic.
First, the risk of limited selectivity is often higher for inhibitors that have fewer binding determinants.
Offꢀtarget inhibition can cause negative sideꢀeffects in patients, and was regarded to be a contributing
factor in the failure of first generation matrixꢀmetalloproteinase inhibitors in clinical trials for preventing
cancer metastasis.4 Second, the emergence of drug resistance is more likely for inhibitors that have fewer
binding determinants as only a small number of mutations are required for the target protein to evade drug
binding. This is particularly relevant to therapies for rapidly evolving diseases, and is an ongoing concern
for treating HIV5 and hepatitis C6 infections. To overcome these challenges, there is now growing interest
in engineering naturally occurring peptide and small protein inhibitors that interact via larger binding
surfaces and are capable of greater selectivity.7
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
Standard mechanism inhibitors8, 9 (or Laskowski inhibitors, as recommended by the MEROPS database)10
are a major class of serine protease inhibitors that fall between small molecule drugs and large protein
biologics (in terms of molecular weight). These inhibitors are found in all forms of life, and operate by an
intriguing mechanism that involves the protease cleaving then reꢀsynthesizing the inhibitor’s reactive site
bond.11, 12 Despite sharing a common mode of action, Laskowski inhibitors show immense variety in
Page 3 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 42
1
2
3
4
5
6
7
8
9
sequence and structure, and segregate into at least 19 different families that display unique
characteristics.9 This is an advantage for drug development, as it provides a series of engineering
templates that function in the same way but are structurally diverse, ranging in size from circular peptides
(as small as 1.5 kDa) to larger protein domains (6ꢀ8 kDa). Indeed, only one structural feature is strictly
conserved across all inhibitor families: the conformation of the inhibitor’s binding loop (termed the
canonical loop).8, 13 This segment is preꢀconfigured to match the geometry of the serine protease active
site, and displays a variable sequence that is complementary to the substrate specificity profile of intended
protease targets (spanning both primed (Pʹ) and nonꢀprimed (P) binding subsites, according to Schechterꢀ
Berger nomenclature,14 Figure 1A).
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
As with many proteinꢀprotein interactions, the sequence of the inhibitor’s binding loop has a major role in
determining which proteases the inhibitor will be effective against. Therefore, engineering the sequence
of the canonical loop is one of the main strategies for directing Laskowski inhibitors to new protease
targets. New sequence combinations can be identified by several different methods, with those capable of
screening both the nonꢀprimed and primed residues generally providing greater success. To date, the most
widely used approach has been phage display, which is one of only a few methods where all residues of
the canonical loop are accessible for optimization. Phage display was first applied to select potent
inhibitors of neutrophil elastase,15 then later used to develop the first engineered Laskowski inhibitor to
reach the clinic, Ecallantide. This drug is a selective plasma kallikrein inhibitor based on the first Kunitz
domain of tissue factor pathway inhibitor16 and is used to relieve inflammation in hereditary
angioedema.17
Alternatively, the template inhibitor can be engineered by rational approaches that focus on substituting a
preꢀdetermined variant sequence into the binding loop. This strategy has been successfully used to
engineer novel inhibitors for furin,18 HCV NS3 protease19 and coagulation proteases20 from diverse
Laskowski inhibitor templates. More recently, synthetic approaches have been developed to identify new
binding sequences by screening combinatorial21 or nonꢀcombinatorial22 libraries of tetrapeptide substrates
Page 4 of 41
ACS Paragon Plus Environment

Page 5 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
to characterize the nonꢀprimed specificity of the target protease. We have previously used this strategy to
guide the design of inhibitor variants for several different members of the kallikreinꢀrelated peptidase
(KLK) family, including KLK422, 23 and KLK14.24, 25 The success of the substrateꢀguided approach
derives from observations that for all Laskowski inhibitors, the nonꢀprimed residues of the canonical loop
interact with the protease to form a βꢀsheet (spanning from P1 to at least P3) that replicates the binding of
a conventional substrate (Figure 1AꢀB).9, 26 As different proteases usually display a unique nonꢀprimed
specificity profile,21 modifying the sequence of the inhibitor’s contact βꢀstrand often leads to marked
changes in its potency and selectivity. Serine proteases also display diversity across their primed binding
sites that can be targeted in inhibitor engineering studies. Moreover, for Laskowski inhibitors, the primed
segment is involved in interactions that are specific to the inhibition mechanism, including the religation
reaction, that are not taken into account when screening substrates. Consequently, there is a need to
develop new, broadlyꢀapplicable strategies for screening the primed residues of the canonical loop to
increase the coverage of chemistryꢀbased methods for optimizing the binding sequence.
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 this study, we developed a versatile, inhibitorꢀfocused approach for screening preferences at a critical
residue in the primed segment, P2ʹ. This strategy was based on a novel, broadꢀrange protease inhibitor
that we recently produced by engineering the fourteen amino acid cyclic peptide, sunflower trypsin
inhibitorꢀ1 (SFTIꢀ1, Figure 1CꢀD).25 Using the engineered broadꢀrange inhibitor as a starting point, we
designed and synthesized a library of twenty P2ʹ variants to screen against diverse serine proteases. The
library screen revealed unique P2ʹ preferences for thirteen different proteases as well as an overall
preference for Ile, which is present at P2ʹ in many naturally occurring inhibitors, including SFTIꢀ1.
Additionally, we used the newly identified P2ʹ preferences to modify known matriptase, KLK14 and
thrombin inhibitors, generating new variants that displayed improved selectivity. Overall, our study
presents a versatile cyclic peptide library that makes it possible to rapidly screen P2ʹ preferences for
diverse serine proteases, and demonstrates that substituting the P2ʹ residue in Laskowski inhibitors, such
as SFTIꢀ1, is effective for modulating the selectivity of engineered inhibitor variants.
Page 5 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 42
1
2
3
4
RESULTS
5
6
7
8
9
Design, synthesis and characterization of a P2ʹ diverse inhibitor library. To devise a new screening
method for optimizing interactions between the primed segment of Laskowski inhibitors and various
protease targets, we first studied structures of diverse protease/inhibitor complexes (see Figure S3,
Supporting Information). Examining residues on the primed side of the scissile bond (P1ʹꢀP4ʹ) revealed
that interactions formed by the P1ʹ, P3ʹ and P4ʹ residues varied between intramolecular and intermolecular
interactions across different inhibitor families. However, the P2ʹ residue consistently extended out from
the inhibitor core and engaged a wellꢀdefined pocket (S2ʹ) within the protease active site (Figure 1EꢀF).
This suggested that the P2ʹ residue was an excellent target for substitution with the aim of changing the
inhibitor’s potency and selectivity. Therefore, we set out to characterize the P2ʹ specificity of different
proteases by designing a synthetic inhibitor library based on the naturally occurring cyclic peptide, SFTIꢀ
1 (compound 1, Table 1). SFTIꢀ1 is a potent inhibitor of trypsin (Ki = 0.1 nM)27 and with only fourteen
amino acids, SFTIꢀ1 (and engineered variants) can be readily produced by solid phase peptide synthesis,
including inhibitors containing nonꢀnatural amino acids.28, 29 Additionally, SFTIꢀ1 has relatively simple
architecture, comprising a canonical loop with minimal scaffolding, and thus is an excellent model system
for examining how various P2ʹ substitutions affect the inhibitor’s activity against different protease
targets.
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 SFTI variant that we selected for generating the inhibitor library was a broadꢀrange variant (2I,
compound 2 with Ile at P2ʹ) that we recently engineered,25 where residues 2, 12 and 14 of SFTIꢀ1 were
substituted to reꢀconfigure the inhibitor’s intramolecular hydrogen bond network, and P1 Lys was
replaced with Arg (Figure 2A). These replacements gave rise to an inhibitor that forms minimal side
chain interactions with the protease’s nonꢀprimed binding sites (except for the S1 subsite) as the P2 and
P4 residues participate in the inhibitor’s hydrogen bond network, and the P3 residue forms a disulfide
bond with Cys11. As 2I has relatively few selectivity determinants but a prominent hydrogen bond
network, it effectively inhibits a number of proteases, including trypsin and several members of the
Page 6 of 41
ACS Paragon Plus Environment

Page 7 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
kallikreinꢀrelated peptidase family.25 Using 2I as a template, we produced a library of twenty individually
synthesized inhibitors where the P2ʹ residue (Ile7) was substituted (Figure 2B) with each of the naturally
occurring amino acids (excluding Cys to prevent formation of disulfide bond isomers), or biphenylalanine
(BiP, single letter code: B) as an example of a nonꢀnatural amino acid.
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
Peptides were synthesized by standard Fmoc synthesis on acid labile 2ꢀchlorotrityl chloride resin.
Protected peptides were liberated from the solid support, then cyclized in solution using 1ꢀ
[Bis(dimethylamino)methylene]ꢀ1Hꢀ1,2,3ꢀtriazolo[4,5ꢀb]pyridinium 3ꢀoxid hexafluorophosphate (HATU)
and N,Nꢀdiisopropylethylamine (DIPEA) as the activator and base, respectively. After protecting group
removal and an initial purification by reverse phase HPLC, disulfide bonds were formed before further
reverse phase HPLC purification to achieve purities above 95% as determined by analytical UPLC (Table
1 and Figure S1, Supporting Information).
The purified peptides were subsequently analyzed by NMR. For SFTIꢀ1 and 2I, we previously showed
1
that the inhibitor’s 1D HꢀNMR spectrum exhibits sharp lines and wellꢀdispersed amide resonances that
are distributed across the 7–9 ppm range with downꢀfield shifted αH signals that are indicative of a βꢀ
1
sheet structure.25, 30 Similar to that of 2I, the 1D HꢀNMR spectra of the P2ʹ variants showed sharp lines
and wellꢀdispersed signals (Figure S2, Supporting Information), indicating that they were wellꢀfolded. In
addition, the number of signals (from backbone amides and side chain protons) present in the 7–9 ppm
region of the spectra suggested that the peptides adopt one main conformation in solution. Of all the
variants, 2P displayed the greatest conformational flexibility, with the presence of a set of additional low
intensity signals in its 1D 1HꢀNMR spectrum that suggest the presence of at least one other conformation.
This is expected as the replacement of an Ile with the conformationally restricted Pro would likely have
some effect on the backbone structure. In general though, the NMR data indicate that the SFTIꢀ1 variants
were wellꢀfolded and rigid.
Exploring the versatility of the inhibitor library against diverse serine proteases. The first proteases
screened against the P2ʹ inhibitor library were the classical serine proteases, trypsin and chymotrypsin. In
Page 7 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 42
1
2
3
4
5
6
7
8
9
a previous study, we showed that 2I was a potent inhibitor of trypsin (Ki = 0.7 nM),25 and thus each of the
twenty inhibitor variants was assayed against trypsin at a low nanomolar range (25 nM). Examining
trends across the library revealed that substituting the P2ʹ residue led to marked changes in the inhibitor’s
activity against trypsin (Figure 3). The strongest inhibition was provided by branched hydrophobic
residues, Ile (80% inhibition) and Val (75%), followed by Arg (65%) and Leu (52%). By contrast,
variants containing Gly or Pro showed essentially no activity whereas acidic residues, Asp and Glu, also
generated poor inhibitors. These findings were largely consistent with a previous study that used a
minimized BowmanꢀBirk loop peptide (similar to residues 3ꢀ11 of the peptides in this study) to examine
the effect of P2ʹ substitutions on trypsin inhibition,31 although the overall level of inhibition in our study
was several orders of magnitude higher.
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 inhibitor library was also able to provide insight into the P2ʹ specificity of chymotrypsin, even though
all library peptides contained Arg as the P1 residue which is not favored by chymotrypsin.21 It is
important to note that assays with chymotrypsin required much higher inhibitor concentrations (25 µM),
and this may impact on the level of discrimination between favored and nonꢀfavored P2ʹ residues.
However, the data generated is still useful for suggesting P2ʹ substitutions that can be explored further.
Overall, the P2ʹ residues that provided the strongest inhibition were similar to trypsin, including
hydrophobic residues, Val (76% inhibition) and Ile (60%), and Arg (57%) (Figure 3). However, aromatic
residues were also strongly preferred, including Trp (63%) and Phe (52%). To explore the structural basis
for these differences in the P2ʹ specificity profiles for chymotrypsin and trypsin, we used molecular
dynamics (MD) simulations to examine binding interactions in various protease/inhibitor complexes.
These analyses revealed that the S2ʹ pocket of chymotrypsin is larger than trypsin, allowing bulky P2ʹ
residues (Trp) to be accommodated in the space left by Thr151 (Figure S4A), Supporting Information).
By contrast, trypsin contains Tyr at residue 151, which is able to form cationꢀπ interactions with P2ʹ Arg
(Figure S4B), Supporting Information), but leaves less space for larger resides, including Trp.
Page 8 of 41
ACS Paragon Plus Environment

Page 9 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
To further test the versatility of the inhibitor library, we screened the catalytic domain of matriptase, a
type II transmembrane serine protease. Matriptase attracts significant interest as a therapeutic target in
many epithelial cancers,32 and has been the focus of previous inhibitor design studies, including several
based on SFTIꢀ1.28, 29, 33 Across the P2ʹ inhibitor library, the three most effective variants were Val (54%
inhibition), Ile (41%) and Asn (37%) (Figure 3). Additionally, Asp (29%) appeared to be better tolerated
by matriptase than seen previously for trypsin and chymotrypsin (comparing Asp to the optimal residue
for each protease). Examining interactions between matriptase and 2N (Figure S4C, Supporting
Information) indicated that Asn formed favorable interactions within the S2ʹ pocket by engaging His143
(His752) – chymotrypsin numbering is used throughout and where relevant, specific protease numbering
is given in brackets. In 2D, Asp formed similar interactions with His143 (His752). Collectively, these
findings demonstrate that the P2ʹ inhibitor library is effective against diverse serine proteases and has the
capacity to discern a unique specificity profile for individual enzymes.
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
Closelyꢀrelated proteases from the kallikreinꢀrelated peptidase family show distinct P2ʹ
preferences. Next, we used the P2ʹ inhibitor library to examine specificity divergence among four
members of the largest known multiꢀgene serine protease family, the kallikreinꢀrelated peptidases
(KLKs). KLK proteases are involved in a wide range of physiological functions34 and are under
investigation as therapeutic targets for several indications, including hormoneꢀdependent cancers (prostate
and ovarian cancers)35 and skin diseases.36, 37 The nonꢀprimed (P1ꢀP4) specificity of several KLKs has
been studied using combinatorial peptide libraries,38 and later by nonꢀcombinatorial library screens22, 24 in
order to guide engineering of the SFTIꢀ1 βꢀstrand. Optimizing the P2ʹ residue provides a further
opportunity to improve the potency and selectivity of engineered Laskowski inhibitors targeting different
KLKs.
Screening the inhibitor library against KLK4, KLK5, KLK7 and KLK14 revealed that each protease had a
unique P2ʹ specificity profile (Figure 3). KLK4 and KLK14 displayed similar preferences, with the most
effective variant being Val (92% and 86% inhibition, respectively), followed by Ile, Asn and Arg.
Page 9 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 42
1
2
3
4
5
6
7
8
9
However, KLK4 also favored aromatic residues (BiP and Phe) but not acidic residues (Glu and Asp),
whereas the converse was observed for KLK14. KLK5 and KLK7 also displayed similar preferences,
including Asn which provided 93% and 97% inhibition for each protease, respectively. Polar residues
were also favored by both proteases, although residues such as His, Gln and Ser provided higher
inhibition for KLK7 (compared to Val and Ile), but not KLK5. By contrast, the preference for aromatic
residues appeared more pronounced for KLK5, such that the BiP, Tyr, Trp and Phe variants were among
the most effective inhibitors for this protease.
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 further investigated the distinct P2ʹ preferences observed for each kallikrein protease by performing
MD simulations. These analyses revealed that the structural basis for KLK4’s preference for Arg at P2ʹ
was different to that previously seen for trypsin. Rather than interacting with an aromatic residue, Arg7
formed a salt bridge with Glu39 in the KLK4 S2ʹ pocket (Figure S5A, Supporting Information).
Additionally, KLK5’s strong preference for aromatic residues could be explained by Tꢀoriented π
interactions involving Phe150 (Figure S5B, Supporting Information). For KLK7, the most favored residue
(Asn) appeared to form a hydrogen bond with the backbone carbonyl of Leu40 (Figure S5C, Supporting
Information). Finally, the preference for Lys at P2ʹ for KLK14 seemed to involve cationꢀπ interactions
with Tyr150 (Figure S5D, Supporting Information). These findings illustrate that even closelyꢀrelated
proteases (from the same multiꢀgene family) show differences in their P2ʹ specificity profile that can be
traced to specific points of sequence diversity in the S2ʹ pocket.
Profiling the P2ʹ specificity of proteases from the coagulation and fibrinolysis cascade. The
coagulation and fibrinolysis cascade was among the first proteolytic pathways to be targeted in major
drug discovery programs.2 Dysregulation of this proteolytic cascade is a common theme in many diseases,
including inflammation, cancer and cardiovascular diseases. Overactive blood coagulation (thrombosis) is
the most frequent cause of myocardial infarction and stroke.39 Additionally, both coagulation and
fibrinolysis need to be controlled during surgery to reduce bleeding and prevent activation of a systemic
inflammatory response.40 The broadꢀspectrum Laskowski inhibitor, aprotinin, has been used to inhibit
Page 10 of 41
ACS Paragon Plus Environment

Page 11 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
proteases that drive both coagulation and fibrinolysis, but it was discontinued in 2008 after 15 years of
clinical use following a controversial safety trial.41 Consequently, there has recently been a renewed effort
to develop inhibitors that selectively target key serine proteases in the coagulation and fibrinolysis
pathways.
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 major direct contributor to coagulation is thrombin, which activates platelets and cleaves fibrinogen
to generate fibrin.42 Screening the P2ʹ inhibitor library against thrombin (Figure 3) revealed a clear
preference for Tyr (62% inhibition) followed by Asn (49%) and Phe (41%). Molecular modeling
indicated that the Tyr aromatic ring was accommodated by the wide and flat hydrophobic portion of the
S2ʹ pocket, while the hydroxyl group formed hydrogen bonds with Arg73 (Arg431) and Asn143 (Asn506)
(Figure S6B, Supporting Information). The actions of thrombin and the coagulation cascade are countered
by the fibrinolysis pathway, where plasmin acts to disperse blood clot by degrading fibrin. We have
previously studied the nonꢀprimed (P1ꢀP4) specificity of plasmin using a nonꢀcombinatorial peptide
library,43 and screening preferences at P2ʹ using the inhibitor library revealed that Lys (60%) was the most
favored residue, followed by Ile (50%), Val (25%) and Arg (23%). Molecular modeling suggested that
P2ʹ Lys can form a salt bridge with Glu143 (Glu685) deep within the plasmin S2ʹ pocket (Figure S6A,
Supporting Information) while the shallow part of the pocket is hydrophobic, which suits Ile. These
findings were consistent with a previous study where plasmin’s preference for basic residues at P2ʹ was
used to engineer a selective plasmin inhibitor based on tissue factor pathway inhibitorꢀ2.44
We were also able to use the P2ʹ inhibitor library to screen proteases in the coagulation cascade that lie
upstream of thrombin and regulate its activation. Interestingly, factor IXa tolerated both basic and acidic
residues at P2ʹ, with the strongest inhibition provided by Glu (54%) followed by Arg and Lys (43%). This
anomaly was resolved by modeling analyses, which revealed that Glu was able to form a salt bridge with
Arg142 (Arg358) (Figure S6C, Supporting Information), whereas basic residues formed a different salt
bridge with Asp39 (Asp249). Factor Xa has a narrow S2ʹ pocket and was the only protease that did not
tolerate Ile (Leu and Trp were also disfavored). Rather, most acidic, aromatic or small residues were
Page 11 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 42
1
2
3
4
5
6
7
8
9
accepted, including Asp (31%) and Tyr (26%). Molecular modeling suggested P2ʹ Asp formed several
hydrogen bonds, including a salt bridge with Arg142 (Arg366) (Figure S6D, Supporting Information),
while Tyr formed a hydrogen bond with Glu39 (Glu257). The inhibitor library also revealed several
differences in the P2ʹ specificity for factor XIa and factor XIIa. The most preferred P2ʹ residue for factor
XIa was Val (65%), followed by Ile (47%) and Asp (34%), whereas factor XIIa favored aromatic and
branched hydrophobic residues, with the strongest inhibition provided by Trp (87%), closely followed by
Ile (85%), Val (80%), BiP (80%) and Tyr (75%). Factor XIa has a narrow and hydrophobic S2ʹ pocket,
which ideally fits Val (Figure S6E, Supporting Information), while Arg39 (Arg413) located at the far end
of the pocket enables hydrogen bonding with acidic residues. By contrast, factor XIIa has a much larger
S2ʹ pocket with Tyr150 (Tyr377) at its base, allowing for aromatic ring interactions with Trp (Figure S6F,
Supporting Information) or Tyr.
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
Exploiting P2ʹ specificity differences to engineer new inhibitor variants that show improved
selectivity. Previous studies have used SFTIꢀ1 to design new inhibitors for a diverse array of protease
targets.22, 25, 28, 29, 33, 45ꢀ47 These engineered variants were developed using a range of different strategies,
but in several cases, the P2ʹ residue (Ile) was not targeted for refinement. Therefore, to explore whether
the P2ʹ specificity data from our study can be used in the design process to improve the selectivity of
engineered inhibitors, we studied the effect of substituting Ile7 in existing SFTI variants that were
developed to inhibit matriptase, KLK14 and thrombin.
1
As with members of the inhibitor library, we acquired 1D HꢀNMR spectra to confirm that the new
engineered inhibitors (Table 1) had folded correctly. As shown in Figure S2 (Supporting Information), the
spectra showed sharp lines and wellꢀdispersed signals, indicating that the engineered inhibitors were wellꢀ
folded. The intensity and number of amide resonances in the spectra suggests that most of the inhibitors
adopt one main conformation in solution, with the exception of 4, which adopts multiple conformations.
To investigate this further, we acquired 2D homonuclear TOCSY and NOESY spectra of 4. Based on the
presence of NOEs between the Hα/Hδ of a Pro and the Hα of the preceding residue, we found that the
Page 12 of 41
ACS Paragon Plus Environment

Page 13 of 42
Journal of Medicinal Chemistry
1
2
3
4
conformational heterogeneity was due to cisꢀtrans isomerization of the Asp7ꢀPro8 and Pro8ꢀPro9 peptide
5
6
bonds.
7
8
9
Matriptase. To date, the most common target in SFTIꢀbased engineering studies has been the cancerꢀ
related protease, matriptase.28, 29, 33 In one study, a high affinity SFTI variant for matriptase was developed
using a truncated inhibitor scaffold (spanning residues 1ꢀ12) with a nonꢀcyclic peptide backbone and an
amidated Cꢀterminus29 (Compound 3, Table 2). However, an ongoing challenge in designing SFTIꢀbased
inhibitors for matriptase has been achieving a suitable degree of selectivity. Indeed, 3 was reported to
show only 6ꢀfold selectivity for matriptase over trypsin.29 Our inhibitor library screen indicated that
matriptase and trypsin had divergent P2ʹ specificity profiles (Figure 3). In particular, matriptase seemed to
tolerate Asp at P2ʹ reasonably well (compared to Ile), whereas Asp was poorly favored by trypsin. With
the aim of improving the selectivity of 3 for matriptase over trypsin, we synthesized 3 and the
corresponding Asp7 variant, 4. Screening both compounds against matriptase revealed that Asp7 was
tolerated almost as well as Ile7 with only a slight decrease in activity observed for 4 (Table 2). Modeling
4 in complex with matriptase indicated that Asp7 formed a favorable hydrogen bond with His143
(His752) (Figure 4A), that was similar to that seen for P2ʹ Asn. However, equivalent assays with trypsin
revealed that the Asp7 variant showed a 350ꢀfold decrease in activity compared to 3 (P2ʹ Ile), which
improved the inhibitor’s selectivity to 33ꢀfold (under the assay conditions in our study, we found that 3
was a more potent inhibitor of trypsin than matriptase). Further, 4 showed 196ꢀfold selectivity over
plasmin due to a decrease in activity compared to 3, while neither variant 3 nor 4 showed any inhibition
of thrombin up to 50 µM (Table 2).
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
KLK14. A further therapeutic application where SFTI variants are currently under development is for
inhibiting kallikrein proteases implicated in skin diseases.25, 47 In a recent study, we designed a potent and
selective inhibitor for KLK14 (5, Table 1) by engineering the SFTIꢀ1 contact βꢀstrand (P4ꢀP1 residues) as
well as residues 12 and 14.25 This variant is a potent KLK14 inhibitor (Ki = 2.0 nM) and shows
approximately 180ꢀfold selectivity over KLK5 and trypsin. Our data from the P2ʹ library screen indicated
Page 13 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 42
1
2
3
4
5
6
7
8
9
that Lys7 was reasonably well tolerated by KLK14, but was strongly disfavored by related skin proteases,
KLK5 and KLK7 (Figure 3). This suggested that replacing Ile7 with Lys could further increase the
selectivity of 5. The P2ʹ Lys variant (6) showed a slight decrease in activity against KLK14 (Ki = 7.0 nM,
Table 2), but had dramatically lower activity against KLK5 and KLK7 (Ki > 50,000 nM), which
represents more than 7,000ꢀfold selectivity over these proteases. Molecular modeling of 6 in complex
with KLK14 indicated that Lys7 was involved in cationꢀπ interactions with Tyr150 (Tyr158) (Figure 4B)
as seen with 2K in the initial library screen. Additionally, 6 showed 8.4ꢀfold weaker inhibition of trypsin
compared to 5 (457ꢀfold selectivity), but maintained reasonable activity against KLK4 (Table 2), which
was consistent with data from the inhibitor library screen (Figure 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
Broad-range kallikrein inhibitors. In addition to engineering selective inhibitors for skin kallikrein
proteases, there is growing interest in developing multiꢀtarget inhibitors where a single variant shows
potent activity against KLK5, KLK7 and KLK14.48 We found that the starting compound in the P2ʹ
inhibitor library (2I) was a potent inhibitor of all three kallikrein proteases (Table 2). A further
consideration is whether the multiꢀtarget inhibitor should also block matriptase as this protease has been
implicated in activating KLK5 and KLK7 in skin diseases.49 However, matriptase influences several
functions in the skin50 and deficient matriptase activity is implicated in causing a skin disease (ichthyosis
hypotrichosis syndrome),51 thus a broadꢀrange inhibitor showing activity against matriptase may carry a
risk of negative sideꢀeffects. In assays with matriptase, 2I showed weak activity (Ki = 13,600 nM),
corresponding to at least 800ꢀfold selectivity for KLK5, KLK7 and KLK14. Yet, when the remaining P2ʹ
variants in the library screen were considered, it appeared that replacing Ile7 with Asn (2N) could provide
a more effective KLK5, KLK7 and KLK14 inhibitor without compromising the variant’s low activity
against matriptase. Whereas 2N showed a slight decrease in activity against KLK5 and KLK14 compared
to 2I, inhibition of KLK7 was improved by over 20ꢀfold (Table 2). Additionally, 2N showed weaker
activity against matriptase (Ki = 19,300 nM), which equates to over 3,700ꢀfold selectivity.
Page 14 of 41
ACS Paragon Plus Environment

Page 15 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Thrombin. Activation of thrombin is a key step in the coagulation pathway, and thus a considerable
amount of research has focused on the development of potent and selective thrombin inhibitors.42
Although SFTIꢀ1 is a poor inhibitor of thrombin (Ki = 5,000 nM), a previous study showed that
substituting the P1 Lys residue with Arg (7) could improve the inhibitor’s activity by almost 6ꢀfold (Ki =
860 nM).28 Our findings from the P2ʹ inhibitor library revealed that Tyr7 (62%) led to stronger inhibition
of thrombin than Ile7 (39%). Additionally, Tyr7 appeared to be less favored by plasmin and matriptase
compared to Ile. Screening 7 and the corresponding P2ʹ Tyr variant (8) against thrombin revealed that
replacing Ile7 with Tyr led to a further increase in activity (Table 2). Modeling 8 in complex with
thrombin revealed similar interactions involving P2ʹ Tyr (Figure 4C) as seen previously with 2Y. Further,
comparing inhibition of other proteases by 7 and 8 revealed that 8 showed 5ꢀfold lower activity against
trypsin and plasmin, and more than 90ꢀfold lower activity against matriptase. While 8 remains a very
potent trypsin inhibitor and may have limited potential for therapeutic development, these proofꢀofꢀ
concept findings indicate that P2ʹ Tyr can be incorporated into a future SFTIꢀbased inhibitor for thrombin
to improve its 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
DISCUSSION AND CONCLUSIONS
In this study, we developed a new library of cyclic peptide inhibitors that can be used to characterize the
P2ʹ specificity of different serine proteases in order to improve the selectivity of engineered inhibitor
variants. Screening the inhibitor library against thirteen different proteases revealed several common
preferences but ultimately, a unique P2ʹ specificity profile was identified for each enzyme. These effects
were explored further using MD simulations where we found that the distinct P2ʹ preferences for each
protease were often linked to regions of sequence diversity in the S2ʹ binding pocket. Finally, we used the
newly identified P2ʹ specificity profiles to further optimize existing engineered inhibitors based on SFTIꢀ
1, demonstrating that our data from the inhibitor library screen is transferrable to other SFTI variants.
Together with existing substrateꢀguided approaches, the strategy described in this study now makes it
Page 15 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 42
1
2
3
4
5
6
7
8
possible to use synthetic chemistryꢀbased methods to screen key residues on both the primed and nonꢀ
primed sides of the inhibitor’s binding loop, allowing engineered variants to achieve higher potency and
selectivity.
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
This study has first demonstrated that the P2ʹ residue has considerable potential for modulating the
potency and selectivity of engineered Laskowski inhibitors. In existing structures of serine proteases, the
S2ʹ pocket is typically wellꢀdefined and has a unique configuration in different proteases due to variation
in sequence or loop conformation, hinting that interactions involving the P2ʹ residue are important to the
enzyme’s binding specificity. Additionally, studies on mesotrypsin have shown that interactions involving
the S2ʹ subsite and P2ʹ residue have a major role in the protease’s notable resistance to Laskowski
inhibitors (including aprotinin).52, 53 However, until now, there has been a lack of systematic, broadlyꢀ
applicable methods for screening the P2ʹ specificity profile of target proteases (in the context of protease
inhibition) to identify suitable modifications that can be harnessed in engineering studies. Using our
library of P2ʹ inhibitor variants, we found that a range of residues with different biophysical properties
were favored by various proteases, to the extent that almost all of twenty P2ʹ residues included in the
study were highly preferred by at least one of the thirteen enzymes screened (Figure 3). Moreover, using
these data to guide substitution of the P2ʹ residue in a series of existing engineered inhibitors (for
matriptase, KLK14 and thrombin) yielded new variants that showed marked improvements in selectivity.
Interestingly, the P2ʹ substitutions were found to be effective regardless of whether the initial library
screen was performed at low (25 nM) or high concentration (10ꢀ25 µM), further demonstrating the
versatility of the method. Additionally, the slight decrease in potency for the new KLK14 and matripase
variants was fully consistent with data from the library screen, as Ile (present in the starting variant) was
more favored than new residue (Lys and Asp, respectively) and the substitution was made with the
intention of improving the inhibitor’s selectivity.
The findings presented here also provide new insights into the significance of Ile at P2ʹ in SFTIꢀ1 and its
prevalence in structurallyꢀrelated inhibitors from the BowmanꢀBirk family.8 SFTIꢀ1 has been found to
Page 16 of 41
ACS Paragon Plus Environment

Page 17 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
inhibit a number of proteases with reasonably high affinity, including trypsin, matriptase, cathepsin G,
KLK4, KLK5, KLK14.22, 25, 27, 28 Across all P2ʹ residues included in the inhibitor library, we found that Ile
was one of the most preferred residues for almost all proteases screened (Figure 3), together with Val,
which is chemically reminiscent to Ile. Additionally, Ile is found at P2ʹ in almost half of the 190
BowmanꢀBirk inhibitor sequences (from different species) listed in the MEROPS database10 (Figure 5A).
In an earlier study, Leatherbarrow and colleagues suggested that Ile may be prevalent at P2ʹ in Bowmanꢀ
Birk inhibitors either to favor broadꢀrange inhibition or to allow the binding loop to adopt an optimal
conformation for interaction with the protease.31 Our findings across thirteen diverse serine proteases
provide new evidence that supports the former hypothesis (broadꢀrange inhibition). Indeed, Ile appeared
to be dispensable at P2ʹ as although it was favored by most proteases, we identified chemically diverse
P2ʹ residues that performed as well or better than Ile for each protease screened, suggesting that Ile is not
a structural requirement for BowmanꢀBirk inhibitors. Interestingly, branched hydrophobic residues (Val,
Ile and Leu) are also found at P2ʹ in over 80% of serine protease proꢀsegments that undergo cleavage
during zymogen activation (Figure 5B). This suggests that broadꢀrange Laskowski inhibitors and their
protease targets have both capitalized on the widespread preference for branched hydrophobic residues at
P2ʹ for the purpose of generating binding motifs that are tolerated by many different serine proteases. By
extension, the broad tolerance of Ile and Val at P2ʹ is also significant in the context of inhibitor
engineering, as retaining these types of residues at P2ʹ is likely to limit the selectivity of new inhibitor
variants.
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
Considering the broader aspects of the Laskowski mechanism and serine protease catalysis, the primed
segment of the binding loop has an especially important role in the inhibitor’s activity, as seen in this
study. During serine protease catalytic processing, once the scissile bond has been cleaved, the primed
segment becomes the initial leaving group, whereas the nonꢀprimed segment remains covalently bound to
the protease via an ester bond (acylꢀenzyme complex).54 For Laskowski inhibitors, the positioning of the
neoꢀNꢀterminus (P1ʹ residue) appears to be critical for efficient religation as it is proposed to engage the
Page 17 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 42
1
2
3
4
5
6
7
8
9
catalytic His residue (in place of the hydrolytic water molecule) to initiate the resynthesis reaction.55, 56
Therefore, the degree of complementarity between the primed segment of the inhibitor and the protease
will probably have a strong influence on whether cleavage of the reactive site bond leads to hydrolysis or
religation (as seen previously with mesotrypsin).52, 53 Since these interactions are specific to the inhibitory
mechanism, it may not be as effective to use substrateꢀbased screens to identify primed segment
substitutions as the set of sequences that are cleaved rapidly will also include those that generate efficient
leaving groups. Indeed, although existing P2ʹ cleavage specificity data for trypsin, thrombin and factor
Xa57 shows a number of similarities with our inhibitor library data, the overall level of correlation is only
modest. However, by using a substrateꢀguided approach to optimize the nonꢀprimed residues, followed by
an inhibitorꢀbased screen to refine the primed segment, we were able to overcome this problem and
developed a potent KLK14 inhibitor, for example, that showed more than 7,000ꢀfold selectivity over
closelyꢀrelated skin proteases, KLK5 and KLK7.
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 potency and selectivity of the new inhibitor variants described in this study compare favorably with
those developed by existing methods, including phage display. While phage display has major advantages
relating to the number of variants that can be screened, its ability to scan both primed and nonꢀprimed
residues simultaneously, and it is entirely inhibitorꢀbased, our synthetic chemistry approach is versatile,
allows validation that each library member is present (in the same concentration), and focuses solely on
variants that bind within the active site. Thus, selection of variants that are not effective inhibitors of the
target protease, as seen in a previous phage display study involving thrombin,16 is avoided. However,
there are several factors that should be considered when transferring this study’s findings to other
engineered inhibitors. First, the activity of each P2ʹ variant was assessed by calculating percentage
inhibition at a single concentration (to enable high throughput screening), and these values do not reflect
the inhibition constant (Ki). Although preferred residues from the library screen generally led to more
selective inhibitors (assessed by Ki), there were examples were percentage inhibition values in the library
screen did not absolutely correlate with changes in Ki, particular when the library was screened at high
Page 18 of 41
ACS Paragon Plus Environment

Page 19 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
concentration (for example, matriptase and 2I versus 2N). Second, all variants in the inhibitor library
contain Arg as the P1 residue. Therefore, we cannot exclude the possibility that certain residues may
become more or less favored (due to coꢀoperativity effects) when the P1 (or an adjacent) residue is
changed. For example, in an engineered variant containing Phe at P1, we have noticed that P2ʹ Tyr
becomes highly favored by chymotrypsin. Third, although the geometry of the binding loop is conserved
across all Laskowski inhibitors, how well the specificity trends identified in this study transfer to
scaffolds from other inhibitor families remains to be investigated. Here, it is important to note that in
SFTIꢀ1 (and many other BowmanꢀBirk inhibitors), the primed segment has a unique conformation as the
P2ʹ residue is followed immediately by two Pro residues (or in some cases, one Pro and a small residue).
Thus, the inhibitor backbone bends away (forming an antiꢀparallel βꢀstrand with the P4ꢀP1 residues)
without engaging the conventional S3ʹ or S4ʹ binding sites.
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 conclusion, our study demonstrates that the P2ʹ residue has a strong influence on Laskowski inhibitor
potency and selectivity, and unveils a novel, inhibitorꢀbased library screen to guide optimization of the
P2ʹ residue in future engineering studies. We anticipate that the P2ʹ specificity profiles determined in this
study will help to improve the potency and/or selectivity of engineered SFTI variants targeting many of
the proteases we screened, many of which have pharmaceutical indications. Additionally, the inhibitor
library is itself a valuable resource that, once synthesized and validated, can be screened against further
proteases under investigation as targets for inhibitor design. Indeed, this strategy contributes to making
more of the canonical loop accessible to optimization by synthetic chemistryꢀbased methods, and when
combined with a systematic approach for refining the nonꢀprimed binding sequence, provides a promising
blueprint for engineering potent and selective serine protease inhibitors.
Page 19 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 42
1
2
3
4
EXPERIMENTAL SECTION
5
6
7
Protein Expression and Protein Sources
8
9
Recombinant KLK proteases used in this study were expressed in zymogen (pro) form in Pichia pastoris
strain Xꢀ33 (KLK5 and KLK7) or Spodoptera frugiperda Sf9 insect cells (KLK4 and KLK14), then
purified and activated as described recently.25 Active site titration was performed using either α2ꢀ
antiplasmin (Sino Biological) (KLK5), α1ꢀantitrypsin (SigmaꢀAldrich) (KLK7) or 4ꢀmethylumbelliferylꢀ
pꢀguanidinobenzoate (MUGB, SigmaꢀAldrich) (KLK4 and KLK14). Active KLKs were supplemented
with 20% (v/v) glycerol to allow storage at ‒80°C with minimal losses of activity. Remaining proteases
were obtained from commercial distributers: βꢀtrypsin, αꢀchymotrypsin, plasmin and factor XIa (Sigmaꢀ
Aldrich); matriptase (R&D Systems); αꢀthrombin, factor IXaꢀβ, factor Xa and factor αꢀXIIa (Molecular
Innovation).
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
Peptide Synthesis
Cyclic inhibitors (including the P2ʹ inhibitor library and proteaseꢀtargeted variants) were synthesized on
2ꢀchlorotrityl chloride resin (0.8 mmol/g) starting at Gly1, while linear inhibitors were synthesized on
Rink amide MBHA resin (0.8 mmol/g) using a Symphony automated peptide synthesizer (Protein
Technologies, Inc). For coupling reactions, Fmoc Nꢀprotected amino acids (4 equiv) were activated with
4 equiv Oꢀ(6ꢀchlorobenzotriazoleꢀ1ꢀyl)ꢀ1,1,3,3ꢀtetramethylaminium hexafluorophosphate (HCTU) and 8
equiv N,Nꢀdiisopropylethylamine (DIPEA) in DMF (2 × 5 min reactions per residue). The Fmoc
protecting group was subsequently removed using 30% piperidine in DMF (2 × 3 min). Assembled
peptides (retaining sideꢀchain protecting groups) were cleaved from the resin using 1% TFA in DCM, and
collected by precipitation in diethyl ether. Headꢀtoꢀtail cyclization was performed in DMF (50 mL per 0.1
mmol peptide) with 4 equiv 1ꢀ[Bis(dimethylamino)methylene]ꢀ1Hꢀ1,2,3ꢀtriazolo[4,5ꢀb]pyridinium 3ꢀoxid
hexafluorophosphate (HATU) and 8 equiv DIPEA at room temperature for 3 hours. To remove DMF and
activators, the postꢀcyclization solution was mixed with an equal volume of DCM, and then washed with
Page 20 of 41
ACS Paragon Plus Environment

Page 21 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
5 volumes of H2O (2ꢀ3 times). The organic phase was recovered, solvent was removed by rotary
evaporation, and sideꢀchain protecting groups were removed by cleavage using
TFA/triisopropylsilane/H2O (96:2:2) (20 mL per 0.1 mmol peptide). Fully deprotected peptides were
collected by precipitation in diethyl ether. Formation of the intramolecular disulfide bond was achieved
by vigorous stirring at room temperature (3 hours) in 0.1 M sodium bicarbonate buffer (pH 8.5)
containing 10 µM oxidized glutathione (total volume: 100 mL per 0.1 mmol peptide).
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
Peptide paraꢀnitroanilide (pNA) substrates were synthesized using 2ꢀchlorotrityl chloride resin that had
been derivatized with paraꢀphenylenediamine (Sigma Aldrich), as previously described.58 Assembly of
linear peptides and liberation of protected peptides from the solid support was performed as above.
Oxidation of the Cꢀterminal paraꢀaminoanilide group was carried out in solution using 6 equiv Oxone
(Sigma Aldrich), as previously described.58 Sideꢀchain protecting groups were removed by TFA cleavage
as described for SFTI variants.
Peptide Purification and Mass Spectrometry Analysis
SFTI variants and peptideꢀpNA substrates were purified by reverse phase HPLC (Shimadzu Prominence)
using a 5 µm ZORBAX ExtendꢀC18 PrepHT column (21.2 × 250 mm) and a linear gradient of 10%
acetonitrile/0.05% TFA to 50% acetonitrile/0.05% TFA. SFTI variants were purified twice, both before
and after formation of the disulfide bond (see above). Peptide purity (> 95%) was confirmed by UPLC
using a 5 µm Agilent 300 SB C18 column (2.1 × 50 mm) at 50°C with mobile phases as above (Figure
S1, Supporting Information). Peptide masses were determined using a quadrupole timeꢀofꢀflight (QꢀTOF)
LCꢀMS system (Agilent Technologies, Table 1). Typical yields for cyclic SFTI variants were between 15ꢀ
20 mg (9ꢀ12%), except compounds 2G and 2P where the yield was 5 mg (3%).
1D NMR
Unless otherwise stated, peptide TFA salts were dissolved in H2O:D2O (9:1) at a concentration of 1.2 mM
(pH 3ꢀ4) including 4,4ꢀdimethylꢀ4ꢀsilapentaneꢀ1ꢀsulfonic acid (DSS) as an internal standard (2 µg/mL,
Page 21 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 42
1
2
3
1
referenced at 0.00 ppm). H NMR spectra were collected on a Bruker Avanceꢀ600 MHz spectrometer at
298 K. For TOCSY and NOESY experiments, a mixing time of 80 ms and 200 ms was used, respectively.
Spectra were processed with TOPSPIN 2.1 and further analyzed with CCPNMR 2.2.2 if required.
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
Inhibition Assays
Inhibition assays were performed in lowꢀbinding 96 well plates using 250 µL assay buffer (0.1 M Tris pH
8.0, 0.1 M NaCl and 0.005% Triton Xꢀ100) and 100 µM peptideꢀpNA substrates. Enzyme concentrations,
proteaseꢀspecific buffer additives and substrates are listed in Table S1 (Supporting Information).
Proteolytic activity was measured by monitoring release of the pNA moiety using a plate reader
spectrophotometer (λ = 405 nm, reading interval: 10 s, assay time course: 300 s), and the level of
inhibition was calculated by comparing kinetic rates to control assays without inhibitor. Substrate
constants (Michaelis–Menten) and inhibition constants (Morrison Ki) were determined by nonꢀlinear
regression using GraphPad Prism 6.0 from three independent experiments performed in triplicate.
Molecular Modeling
For proteases where their structure in complex with SFTIꢀ1 has not been determined by xꢀray
crystallography, protease/SFTI complexes were generated by overlay of the trypsin/SFTIꢀ1 complex
(PDB ID: 1SFI) with existing structures for each protease: chymotrypsin (PDB ID: 1CHO), KLK5 (PDB
ID: 2PSX), KLK7 (PDB ID: 2QXI), plasmin (PDB ID: 3UIR), thrombin (PDB ID: 3VXE), factor IXa
(PDB ID: 3LC3), factor Xa (PDB ID: 2JKH), factor XIa (PDB ID: 3SOR). Additionally, homology
models were created for KLK14 and factor XIIa using SWISSꢀMODEL59 and the templates KLK5 (PDB
ID: 2PSX) and hepatocyte growth factor activator (PDB ID: 2R0L), respectively. For KLK4 and
matriptase, existing protease/SFTI complexes were used (PDB ID: 4KEL and 3PF8). Systems were
solvated with TIP3P water and neutralized by Na+/Clꢀ counter ions to a final concentration of 100 mM in
VMD 1.9.1.60 This generated systems of approximately 22,000ꢀ25,000 atoms including 6,000ꢀ7,000 water
molecules.
Page 22 of 41
ACS Paragon Plus Environment

Page 23 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Each protease/inhibitor complex was equilibrated using a stepwise relaxation procedure. First, all heavy
atoms were harmonically restrained with a force constant of 2 kcal/(mol Å2) before a conjugate gradient
minimization was performed (500 steps) using NAMD 2.961 and CHARMM27 force fields parameters.
This was followed by heating to 298 K and simulating 500 ps under NPT conditions with periodic
boundary conditions. A Langevin thermostat with a damping coefficient of 0.5 psꢀ1 was used to maintain
the system temperature and the system pressure was maintained at 1 atm using a Langevin piston barostat.
The particle mesh Ewald algorithm was used to compute longꢀrange electrostatic interactions at every
time step and nonꢀbonded interactions were truncated smoothly between 10ꢀ12 Å. Hydrogen bonds were
constrained by the SHAKE algorithm (or the SETTLE algorithm for water). For the second stage, the
restraints were retained on the protease and inhibitor αꢀcarbons only (1 ns), while all constraints were
released in the third stage (1 ns). Production runs (10 ns) were performed under NVT conditions with
otherwise identical force field and simulation parameters as above. Coordinates were saved every 100
simulation steps producing 50,000 frames per trajectory.
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
Sequence Alignments
The BowmanꢀBirk inhibitor (family I12) sequence alignment was performed using all 190 sequences
listed in the MEROPS database v9.1210 (date accessed: 9 April 2015). For the human serine protease
alignment, all sequences (139) were extracted from the UniProt proteome database
(http://www.uniprot.org/proteomes/) (date accessed: 9 April 2015). Sequences were aligned using Clustal
Omega 1.2,62 with only serine protease sequences found to contain a canonical proꢀsequence included in
the alignment (112 out of 139). Graphics illustrating sequence diversity were generated using WebLogo
3.4 (http://weblogo.threeplusone.com/). The structureꢀbased sequence alignment of 13 serine proteases in
Figure S7 (Supporting Information) was generated using Chimera 1.8.1 and the PDB structures described
above.
Supporting Information Available: Contains compound characterization (analytical UPLC traces and
1D NMR spectra), molecular modeling, structureꢀbased protease sequence alignment, and proteaseꢀ
Page 23 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 42
1
2
3
4
specific kinetic assay conditions. This material is available free of charge via the Internet at
5
6
http://pubs.acs.org.
7
8
9
Corresponding Author Information: Dr Joakim Swedberg; Phone, +61 (0)7 3346 2021; Email,
j.swedberg@imb.uq.edu.au; Address, Institute for Molecular Bioscience, The University of Queensland,
Brisbane QLD 4072, Australia.
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
Acknowledgement: We thank Dr Peta Harvey (The University of Queensland) for assistance with NMR
experiments. This study was funded by the Australian National Health and Medical Research Council
(NHMRC) [Grant 1059410]. David Craik is a NHMRC Senior Principal Research Fellow [Grant
1026501] and Joakim Swedberg is a NHMRC Early Career Fellow [Grant 1069819].
Abbreviations Used: KLK, kallikreinꢀrelated peptidase; MD, molecular dynamics; pNA, paraꢀ
nitroanilide; SFTIꢀ1, sunflower trypsin inhibitorꢀ1
Page 24 of 41
ACS Paragon Plus Environment

Page 25 of 42
Journal of Medicinal Chemistry
1
2
3
4
REFFERENCES
5
6
7
8
9
(1) Bachovchin, D. A., and Cravatt, B. F. The pharmacological landscape and therapeutic potential of
serine hydrolases, Nat. Rev. Drug Discov. 2012, 11, 52ꢀ68.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(2) Drag, M., and Salvesen, G. S. Emerging principles in proteaseꢀbased drug discovery, Nat. Rev.
Drug Discov. 2010, 9, 690ꢀ701.
(3) Mullard, A. 2013 FDA drug approvals, Nat. Rev. Drug Discov. 2014, 13, 85ꢀ89.
(4) Coussens, L. M., Fingleton, B., and Matrisian, L. M. Matrix metalloproteinase inhibitors and
cancer: trials and tribulations, Science 2002, 295, 2387ꢀ2392.
(5) Rosenbloom, D. I., Hill, A. L., Rabi, S. A., Siliciano, R. F., and Nowak, M. A. Antiretroviral
dynamics determines HIV evolution and predicts therapy outcome, Nat. Med. 2012, 18, 1378ꢀ1385.
(6) Rong, L., Dahari, H., Ribeiro, R. M., and Perelson, A. S. Rapid emergence of protease inhibitor
resistance in hepatitis C virus, Sci. Transl. Med. 2010, 2, 30ra32.
(7) Scott, C. J., and Taggart, C. C. Biologic protease inhibitors as novel therapeutic agents, Biochimie
2010, 92, 1681ꢀ1688.
(8) Laskowski, M., Jr., and Kato, I. Protein inhibitors of proteinases, Annu. Rev. Biochem. 1980, 49,
593ꢀ626.
(9) Laskowski, M., Jr., and Qasim, M. A. What can the structures of enzymeꢀinhibitor complexes tell
us about the structures of enzyme substrate complexes?, Biochim. Biophys. Acta 2000, 1477, 324ꢀ
337.
(10) Rawlings, N. D., Waller, M., Barrett, A. J., and Bateman, A. MEROPS: the database of proteolytic
enzymes, their substrates and inhibitors, Nucleic Acids Res. 2014, 42, D503ꢀ509.
(11) Finkenstadt, W. R., and Laskowski, M., Jr. Peptide bond cleavage on trypsinꢀtrypsin inhibitor
complex formation, J. Biol. Chem. 1965, 240, 962ꢀ963.
(12) Finkenstadt, W. R., and Laskowski, M., Jr. Resynthesis by trypsin of the cleaved peptide bond in
modified soybean trypsin inhibitor, J. Biol. Chem. 1967, 242, 771ꢀ773.
Page 25 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 42
1
2
3
4
5
6
(13) Bode, W., and Huber, R. Natural protein proteinase inhibitors and their interaction with
proteinases, Eur. J. Biochem. 1992, 204, 433ꢀ451.
7
8
9
(14) Schechter, I., and Berger, A. On the size of the active site in proteases. I. Papain, Biochem.
Biophys. Res. Commun. 1967, 27, 157ꢀ162.
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
(15) Roberts, B. L., Markland, W., Ley, A. C., Kent, R. B., White, D. W., Guterman, S. K., and Ladner,
R. C. Directed evolution of a protein: selection of potent neutrophil elastase inhibitors displayed on
M13 fusion phage, Proc. Natl. Acad. Sci. U. S. A. 1992, 89, 2429ꢀ2433.
(16) Markland, W., Ley, A. C., and Ladner, R. C. Iterative optimization of highꢀaffinity protease
inhibitors using phage display. 2. Plasma kallikrein and thrombin, Biochemistry 1996, 35, 8058ꢀ
8067.
(17) Cicardi, M., Levy, R. J., McNeil, D. L., Li, H. H., Sheffer, A. L., Campion, M., Horn, P. T., and
Pullman, W. E. Ecallantide for the treatment of acute attacks in hereditary angioedema, N. Engl. J.
Med. 2010, 363, 523ꢀ531.
(18) Lu, W., Zhang, W., Molloy, S. S., Thomas, G., Ryan, K., Chiang, Y., Anderson, S., and Laskowski,
M., Jr. Arg15ꢀLys17ꢀArg18 turkey ovomucoid third domain inhibits human furin, J. Biol. Chem.
1993, 268, 14583ꢀ14585.
(19) Martin, F., Dimasi, N., Volpari, C., Perrera, C., Di Marco, S., Brunetti, M., Steinkuhler, C., De
Francesco, R., and Sollazzo, M. Design of selective eglin inhibitors of HCV NS3 proteinase,
Biochemistry 1998, 37, 11459ꢀ11468.
(20) Grzesiak, A., Krokoszynska, I., Krowarsch, D., Buczek, O., Dadlez, M., and Otlewski, J. Inhibition
of six serine proteinases of the human coagulation system by mutants of bovine pancreatic trypsin
inhibitor, J. Biol. Chem. 2000, 275, 33346ꢀ33352.
(21) Harris, J. L., Backes, B. J., Leonetti, F., Mahrus, S., Ellman, J. A., and Craik, C. S. Rapid and
general profiling of protease specificity by using combinatorial fluorogenic substrate libraries,
Proc. Natl. Acad. Sci. U. S. A. 2000, 97, 7754ꢀ7759.
Page 26 of 41
ACS Paragon Plus Environment

Page 27 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
(22) Swedberg, J. E., Nigon, L. V., Reid, J. C., de Veer, S. J., Walpole, C. M., Stephens, C. R., Walsh,
T. P., Takayama, T. K., Hooper, J. D., Clements, J. A., Buckle, A. M., and Harris, J. M. Substrateꢀ
guided design of a potent and selective kallikreinꢀrelated peptidase inhibitor for kallikrein 4, Chem.
Biol. 2009, 16, 633ꢀ643.
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
(23) Swedberg, J. E., de Veer, S. J., Sit, K. C., Reboul, C. F., Buckle, A. M., and Harris, J. M. Mastering
the canonical loop of serine protease inhibitors: enhancing potency by optimising the internal
hydrogen bond network, PLoS One 2011, 6, e19302.
(24) de Veer, S. J., Swedberg, J. E., Parker, E. A., and Harris, J. M. Nonꢀcombinatorial library screening
reveals subsite cooperativity and identifies new high efficiency substrates for kallikreinꢀrelated
peptidase 14, Biol. Chem. 2012, 393, 331ꢀ341.
(25) de Veer, S. J., Swedberg, J. E., Akcan, M., Rosengren, K. J., Brattsand, M., Craik, D. J., and Harris,
J. M. Engineered protease inhibitors based on sunflower trypsin inhibitorꢀ1 (SFTIꢀ1) provide
insights into the role of sequence and conformation in Laskowski mechanism inhibition, Biochem.
J. 2015, 469, 243ꢀ253.
(26) Madala, P. K., Tyndall, J. D., Nall, T., and Fairlie, D. P. Update 1 of: Proteases universally
recognize beta strands in their active sites, Chem. Rev. 2010, 110, PR1ꢀ31.
(27) Luckett, S., Garcia, R. S., Barker, J. J., Konarev, A. V., Shewry, P. R., Clarke, A. R., and Brady, R.
L. Highꢀresolution structure of a potent, cyclic proteinase inhibitor from sunflower seeds, J. Mol.
Biol. 1999, 290, 525ꢀ533.
(28) Li, P., Jiang, S., Lee, S. L., Lin, C. Y., Johnson, M. D., Dickson, R. B., Michejda, C. J., and Roller,
P. P. Design and synthesis of novel and potent inhibitors of the type II transmembrane serine
protease, matriptase, based upon the sunflower trypsin inhibitorꢀ1, J. Med. Chem. 2007, 50, 5976ꢀ
5983.
(29) Fittler, H., Avrutina, O., Glotzbach, B., Empting, M., and Kolmar, H. Combinatorial tuning of
peptidic drug candidates: highꢀaffinity matriptase inhibitors through incremental structureꢀguided
optimization, Org. Biomol. Chem. 2013, 11, 1848ꢀ1857.
Page 27 of 41
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 42
1
2
3
4
5
6
7
8
9
(30) Korsinczky, M. L., Schirra, H. J., Rosengren, K. J., West, J., Condie, B. A., Otvos, L., Anderson,
M. A., and Craik, D. J. Solution structures by 1H NMR of the novel cyclic trypsin inhibitor SFTIꢀ1
from sunflower seeds and an acyclic permutant, J. Mol. Biol. 2001, 311, 579ꢀ591.
(31) Gariani, T., McBride, J. D., and Leatherbarrow, R. J. The role of the P2' position of BowmanꢀBirk
proteinase inhibitor in the inhibition of trypsin. Studies on P2' variation in cyclic peptides
encompassing the reactive site loop, Biochim. Biophys. Acta 1999, 1431, 232ꢀ237.
(32) Antalis, T. M., Bugge, T. H., and Wu, Q. Membraneꢀanchored serine proteases in health and
disease, Prog. Mol. Biol. Transl. Sci. 2011, 99, 1ꢀ50.
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
(33) Quimbar, P., Malik, U., Sommerhoff, C. P., Kaas, Q., Chan, L. Y., Huang, Y. H., Grundhuber, M.,
Dunse, K., Craik, D. J., Anderson, M. A., and Daly, N. L. Highꢀaffinity cyclic peptide matriptase
inhibitors, J. Biol. Chem. 2013, 288, 13885ꢀ13896.
(34) Magdolen, V., Sommerhoff, C. P., Fritz, H., and Schmitt, M. Kallikrein-related peptidases.
Characterization, regulation, and interactions within the protease web, Walter de Gruyter GmbH:
Berlin, 2012; pp. 1ꢀ410.
(35) Lawrence, M. G., Lai, J., and Clements, J. A. Kallikreins on steroids: structure, function, and
hormonal regulation of prostateꢀspecific antigen and the extended kallikrein locus, Endocr. Rev.
2010, 31, 407ꢀ446.
(36) de Veer, S. J., Furio, L., Harris, J. M., and Hovnanian, A. Proteases: common culprits in human
skin disorders, Trends Mol. Med. 2014, 20, 166ꢀ178.
(37) Prassas, I., Eissa, A., Poda, G., and Diamandis, E. P. Unleashing the therapeutic potential of human
kallikreinꢀrelated serine proteases, Nat. Rev. Drug Discov. 2015, 14, 183ꢀ202.
(38) Debela, M., Magdolen, V., Schechter, N., Valachova, M., Lottspeich, F., Craik, C. S., Choe, Y.,
Bode, W., and Goettig, P. Specificity profiling of seven human tissue kallikreins reveals individual
subsite preferences, J. Biol. Chem. 2006, 281, 25678ꢀ25688.
(39) Mackman, N. Triggers, targets and treatments for thrombosis, Nature 2008, 451, 914ꢀ918.
Page 28 of 41
ACS Paragon Plus Environment

Page 29 of 42
Journal of Medicinal Chemistry
1
2
3
4
5
6
(40) Edmunds, L. H., Jr. Managing fibrinolysis without aprotinin, Ann. Thorac. Surg. 2010, 89, 324ꢀ
331.
7
8
9
(41) Deanda, A., Jr., and Spiess, B. D. Aprotinin revisited, J. Thorac. Cardiovasc. Surg. 2012, 144, 998ꢀ
1002.
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
(42) Di Nisio, M., Middeldorp, S., and Buller, H. R. Direct thrombin inhibitors, N. Engl. J. Med. 2005,
353, 1028ꢀ1040.
(43) Swedberg, J. E., and Harris, J. M. Plasmin substrate binding site cooperativity guides the design of
potent peptide aldehyde inhibitors, Biochemistry 2011, 50, 8454ꢀ8462.
(44) Bajaj, M. S., Ogueli, G. I., Kumar, Y., Vadivel, K., Lawson, G., Shanker, S., Schmidt, A. E., and
Bajaj, S. P. Engineering kunitz domain 1 (KD1) of human tissue factor pathway inhibitorꢀ2 to
selectively inhibit fibrinolysis: properties of KD1ꢀL17R variant, J. Biol. Chem. 2011, 286, 4329ꢀ
4340.
(45) Kocsis, A., Kekesi, K. A., Szasz, R., Vegh, B. M., Balczer, J., Dobo, J., Zavodszky, P., Gal, P., and
Pal, G. Selective inhibition of the lectin pathway of complement with phage display selected
peptides against mannoseꢀbinding lectinꢀassociated serine protease (MASP)ꢀ1 and ꢀ2: significant
contribution of MASPꢀ1 to lectin pathway activation, J. Immunol. 2010, 185, 4169ꢀ4178.
(46) Gitlin, A., Debowski, D., Karna, N., Legowska, A., Stirnberg, M., Gutschow, M., and Rolka, K.
Inhibitors of Matriptaseꢀ2 Based on the Trypsin Inhibitor SFTIꢀ1, ChemBioChem 2015, 16, 1601ꢀ
1607.
(47) de Veer, S. J., Ukolova, S. S., Munro, C. A., Swedberg, J. E., Buckle, A. M., and Harris, J. M.
Mechanismꢀbased selection of a potent kallikreinꢀrelated peptidase 7 inhibitor from a versatile
library based on the sunflower trypsin inhibitor SFTIꢀ1, Biopolymers 2013, 100, 510ꢀ518.
(48) Tan, X., Soualmia, F., Furio, L., Renard, J. F., Kempen, I., Qin, L., Pagano, M., Pirotte, B., El
Amri, C., Hovnanian, A., and ReboudꢀRavaux, M. Toward the first class of suicide inhibitors of
kallikreins involved in skin diseases, J. Med. Chem. 2015, 58, 598ꢀ612.
Page 29 of 41
ACS Paragon Plus Environment