How a Fragment Draws Attention to Selectivity Discriminating Features between the Related Proteases Trypsin and Thrombin

Article abstract of DOI:10.1021/acs.jmedchem.0c01809

In the S1 pocket, the serine proteases thrombin and trypsin commonly feature Asp189 and a Ala190Ser and Glu192Gln exchange. Nevertheless, thrombin cleaves peptide chains solely after Arg, and trypsin after Lys and Arg. Thrombin exhibits a Na+-binding site next to Asp189, which is missing in trypsin. The fragment benzylamine shows direct H-bonding to Asp189 in trypsin, while in thrombin, it forms an H-bond to Glu192. A series of fragments and expanded ligands were studied against both enzymes and mutated variants by crystallography and ITC. The selectivity-determining features of both S1 pockets are difficult to assign to one dominating factor. The Ala190Ser and Glu192Gln replacements may be regarded as highly conserved as no structural and affinity changes are observed between both proteases. With respect to charge distribution, Glu192, together with the thrombin-specific sodium ion, helps in creating an electrostatic gradient across the S1 pocket. This feature is definitely absent in trypsin but important for selectivity along with solvation-pattern differences in the S1 pocket.

Full text of DOI:10.1021/acs.jmedchem.0c01809

pubs.acs.org/jmc
Article
How a Fragment Draws Attention to Selectivity Discriminating
Features between the Related Proteases Trypsin and Thrombin
Anna Sandner, Khang Ngo, Johannes Schiebel, Angela Ilse Marca Pizarroso, Linda Schmidt,
Benjamin Wenzel, Torsten Steinmetzer, Andreas Ostermann, Andreas Heine, and Gerhard Klebe*
Cite This: J. Med. Chem. 2021, 64, 1611−1625
Read Online
sı
*
Metrics & More
Article Recommendations
Supporting Information
ACCESS
ABSTRACT: In the S1 pocket, the serine proteases thrombin and
trypsin commonly feature Asp189 and a Ala190Ser and Glu192Gln
exchange. Nevertheless, thrombin cleaves peptide chains solely
after Arg, and trypsin after Lys and Arg. Thrombin exhibits a
Na+-binding site next to Asp189, which is missing in trypsin. The
fragment benzylamine shows direct H-bonding to Asp189 in
trypsin, while in thrombin, it forms an H-bond to Glu192. A series
of fragments and expanded ligands were studied against both
enzymes and mutated variants by crystallography and ITC. The selectivity-determining features of both S1 pockets are difficult to
assign to one dominating factor. The Ala190Ser and Glu192Gln replacements may be regarded as highly conserved as no structural
and affinity changes are observed between both proteases. With respect to charge distribution, Glu192, together with the thrombin-
specific sodium ion, helps in creating an electrostatic gradient across the S1 pocket. This feature is definitely absent in trypsin but
important for selectivity along with solvation-pattern differences in the S1 pocket.
trypsin-like serine proteases, of which trypsin is one of the
most prominent and best studied enzymes, does also contain
many blood clotting factors such as thrombin and factor Xa,
two enzymes that cleave substrates exclusively after an arginine
residue. For both, approved anticoagulants have been
developed that enable the treatment of life-threatening
thromboembolic diseases.2,3
Trypsin-like serine proteases are structurally closely related.
In order to avoid undesired therapeutic side effects, it is
therefore essential to consider selectivity as a determining
parameter during the development of new anticoagulant drugs.
For instance, orally active thrombin and factor Xa inhibitors
must be selective against trypsin to avoid being trapped in the
intestine where trypsin is present in high concentrations of up
to 6 μM.4 A high selectivity is also required against their
physiological opponents, the fibrinolytic proteases plasmin and
tPA, which are responsible for the dissolution of blood clots.5
The ability of an inhibitor to discriminate the intended target,
e.g., thrombin, from trypsin can be used as a first indicator to
tailor selectivity.
INTRODUCTION
■
The success of a drug molecule in therapy strongly depends on
its potency, efficacy, and selectivity toward a given target. At
first, high potency is a prime prerequisite for its suitability
along with uncritical toxicity. Efficacy is correlated with
binding kinetics, which depends on the individual association
(kon) and dissociation (koff) rate constants and influences the
residence time of a drug molecule on its target. In addition,
high selectivity for a target protein may be prerequisite for
effective therapy without unwanted side effects; however, a
mixed profile toward several closely related isoforms of a target
family can also be desired. The question of which profile of a
drug molecule is optimal for the action toward a given
pharmacological target is difficult to answer in advance and
makes it problematic to assess the therapeutic value of drug-
development candidates prior to clinical trials. However, as
long as the molecular determinants of already known
selectivity profiles are hardly understood, it is difficult to
come up with novel compounds mastering unseen selectivity
challenges. An obvious criterion for high potency and often
also for sufficient selectivity is the optimal shape complemen-
tarity of the drug and binding site; however, steric fit is perhaps
the most intuitive but, by far, not the only criterion to achieve
selectivity.
Substantial selectivity discrimination results already from the
portion of an inhibitor binding to the S1 site, which, in both
Received: October 16, 2020
Published: January 20, 2021
An important class of target structures are serine proteases as
they fulfill numerous key functions within the human body.
Trypsin, for instance, plays a decisive role during digestion due
to its ability to catalyze the cleavage of proteins’ peptide bonds
C-terminally of most arginine or lysine residues.1 The family of
© 2021 The Authors. Published by
American Chemical Society
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625
1611

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 1. Inhibitory potency expressed as pKi (negative decadic logarithm of the binding constant) of three fragment-like inhibitors benzamidine
(1F), N-amidinopiperidine (2F), and benzylamine (3F) against trypsin (red), thrombin (blue), Ala190Ser thrombin (light blue), and Glu192Gln
thrombin (pale blue). For the latter mutated variant, no binding of benzylamine could be detected.
proteases, hosts an aspartate residue at the bottom of this
pocket.6 In light of this hypothesis, structural variations
between trypsin and thrombin within this pocket are made
responsible for the fact that trypsin cleaves substrates C-
terminally of Arg and Lys while thrombin specifically cleaves
substrates following an Arg residue only.7 A change from an
alanine to a serine residue in position 190 constitutes the most
obvious and prominent difference between the S1 pockets of
thrombin and trypsin and therefore was not only used to
explain the different substrate specificities but also for the
design of thrombin-selective inhibitors.6 Although both
proteases have a highly similar S1 pocket, our results suggest
a difference in the electrostatic properties of Asp189, which
may contribute to their selectivity differences.8,9 We
hypothesized that the observed differences may be at least
partially caused by a sodium ion binding close to Asp189 in
thrombin, which is absent in trypsin. This electrostatic
attenuation of Asp189 may be further pronounced by the
occurrence of a charged Glu192 residue at the rim of the S1
pocket in thrombin that is replaced by an uncharged Gln192 in
trypsin.
The current study tries to shed more light on the selectivity-
determining features between trypsin and thrombin. Our
investigations were initially stimulated by an intriguing
observation while studying the binding of benzylamine as a
probe fragment versus both proteases. In contrast to the
obvious assumption that such a simple fragment would bind
with its basic functional group to the carboxylate group of
Asp189 at the bottom of the S1 pocket of both proteases, this
binding mode is only found for trypsin. In thrombin instead,
the fragment avoids interacting with Asp189, and seeks contact
to Glu192 to which it likely forms a salt bridge. Such findings
could be important in a fragment-based lead optimization
project for thrombin and may suggest an alternative strategy to
follow as in the trypsin case. We believe that the result in
thrombin is even more surprising as Glu192 is rather solvent-
exposed at the rim of the thrombin-binding pocket and,
according to its pKa value in aqueous solution, Glu should be
less acidic than Asp and thus the less probable residue to get
involved in a salt bridge with a basic residue.10
important class of trypsin-like serine proteases of which
thrombin possesses a cation-binding site next to the S1
aspartate whereas other family members lack such a close-by
ion. In general, our data will improve the understanding of the
features in operation to determine selectivity in proteins.
RESULTS AND DISCUSSION
■
Binding Data of Fragments with Basic S1-Binding
Head Group. At first, we characterized the binding properties
of benzylamine (3F) in comparison to the related basic S1-
binding head groups benzamidine (1F) and N-amidinopiper-
idine (2F) against trypsin and thrombin using a fluorescence-
based kinetic enzyme inhibition assay.8,9 We included two
additional thrombin variants in this analysis where Ala190 was
replaced by Ser in the S1 pocket and Glu192 by Gln192 next to
the rim of this pocket. The idea of the two mutated thrombin
variants was to make the S1 pocket of thrombin more similar to
that of trypsin. In Figure 1, the obtained kinetic inhibition
constant (Ki) values are depicted.
The most potent fragment against trypsin is benzamidine
(1F) followed by N-amidinopiperidine (2F) and benzylamine
(3F) with the latter two being nearly equally potent. Against
thrombin, the two amidine fragments swap their order of
affinity,8 whereas benzylamine surprisingly loses more than two
orders of magnitude in potency. This trend is also reflected by
the two mutated forms. Moreover, the Ala190Ser thrombin
variant reduces more strongly the affinity of N-amidinopiper-
idine than of benzamidine. Benzylamine inhibits the Ala190Ser
variant similarly to the wild type, whereas the Glu192Gln
variant shows weak or even no affinity against this ligand in
that no binding could be recorded.
Crystal Structures with Benzylamine as S1-Binding
Head Group. Structurally, we could recently characterize 1F
and 2F in complex with trypsin and thrombin to a very high
resolution.8,11 In the case of trypsin, even a neutron diffraction
study could be accomplished, which allows structural
characterization of protons (and deuterons).12 Therefore, in
the trypsin case, the orientation and the rotational degrees of
freedom of water molecules are also observed.8 The residual
solvation patterns of the complexes formed in the S1 pocket of
both proteases, also in comparison to the uncomplexed apo
protein, deviate significantly and take impact on the binding
kinetics.8,13,14 This explains the affinity difference for both
ligands. However, the large affinity difference of benzylamine
(3F) against both enzymes stimulated us to investigate its
binding modes by crystal-structure analysis.
To rationalize this surprising difference, which might further
underscore our hypothesis about selectivity-discriminating
features, we applied high-resolution neutron and X-ray
crystallography, a fluorescence assay, isothermal titration
calorimetry (ITC), and site-directed mutagenesis to improve
our knowledge about the selectivity features of both proteases.
We believe that the collected insights may help in the
development of novel drug molecules targeting selectively the
In the case of trypsin, we succeeded to determine a
combined neutron and X-ray structure in complex with 3F
1612
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 2. XN structure of the benzylamine−trypsin complex. Selected protein residues are depicted as gray sticks. 3F is shown as green sticks
including all hydrogens. (A) Binding pose of 3F within the S1 pocket (PDB code: 5MOR), distances between hydrogen-bond donor deuterium and
acceptor atoms are given.15 (B) Dependency of the rotational state of waters W1 and W2 of the uncomplexed apo structure (PDB code: 5MOP8).
W1 and W2 were found to exist in two approximately equally populated rotational states, each represented by an individually colored hydrogen
(pink and cyan) with the indicated refined occupancies. In Figure S1 (Supporting Information), the scatter and electron density around
benzylamine and in the apo structure can be found.
Figure 3. (A) X-ray crystal structures of benzylamine (3F) (beige, water molecules red spheres, PDB code: 6T56) and 4-methyl benzylamine
(green, water molecules green spheres, PDB code: 6T55) in the S1 pocket of thrombin. Selected protein residues and the ligand 3F are depicted as
beige and green sticks, respectively. W2 is replaced from the complex by the 4-methyl group of the extended derivative. B: A second copy of 3F is
found in the S3/S4 pocket of thrombin forming an edge-to-face interaction with Trp215, distances in [Å].
(Figure 2 and Figure S1). To obtain the most reliable solvation
structure of the residual water molecules in the S1 pocket with
respect to their rotational and orientational degrees of
freedom, we applied combined XN refinement making
simultaneous use of the collected neutron and X-ray data.16
As anticipated, 3F binds in the cationic state with its
protonated ammonium group directed toward the carboxylate
group of Asp189. It forms a direct hydrogen bond with one of
the carboxylate oxygens (1.9 Å). The other two hydrogens of
upon 3F binding, W1 and W2 experience major changes in
their rotational degrees of freedom compared to the situation
in the uncomplexed protein. This observation is particularly
surprising, as in the complex with aniline, W2 remains
distributed over two states whereas W1 adopts an ordered
geometry.17 This contrasts to the trypsin complex with 2-
aminopyridine where, now, W1 remains with multiple
orientations and W2 transforms into an ordered state.17
Obviously, the residual solvation pattern and its intrinsic
dynamics depend on the bound ligand, which possibly also
influences the binding thermodynamics and the selectivity-
discriminating properties of accommodated ligands.
Surprisingly, benzylamine binds to thrombin with an entirely
different orientation as it recruits the carboxylate group of the
spatially rather flexible Glu192 for the interaction (Figure 3A).
Due to its pKa value of 9.3 in aqueous solution,18 it likely binds
also in a protonated state forming a salt bridge. In addition, the
ammonium group is hydrogen-bonded to the backbone
+
the NH3 head group are hydrogen-bonded to the backbone
carbonyl oxygens of Ser190 and Gly219 (dO····H: 1.8, 1.9 Å).
W1 adopts a fixed geometry with weak H-bonds (hydrogen
bonds) to the backbone carbonyl groups of Gly219 (2.4 Å)
and Lys224 (2.2 Å). In the apo structure (Figure 2B), W1 is
found with multiple orientations and mediates contacts
between the carboxylate group of Asp189, and the backbone
carbonyl oxygens of Lys224 and Gly219.8 Water molecule W2,
which is located on top of Tyr228 on the opposite site of the
pocket is also found in the complex with 3F with ordered
geometry and forms a weak contact to Ser190Oγ (2.8 Å).
Apparently, the residual solvation pattern seems to support the
fixation of the ligand in the S1 pocket. Remarkably, the latter
water molecule W2 is also disordered in the apo enzyme and
assumed to swap dynamically between two orientations. Thus,
carbonyl oxygens of Gly216 and 219 (dN···O: 3.1 Å; dN···O
:
3.2 Å) and to water molecule W6 (3.0 Å). To cross-validate
this rather unexpected result and see whether steric demand
may be important, we determined the crystal structure of the
highly similar 4-methyl benzylamine, which, however, adopts
the same binding pose in the S1 pocket. W2, found to occupy
1613
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 4. Inhibitory potency expressed as pKi of three D-Phe-Pro-CH2-R inhibitors 1L, 2L, and 3L against trypsin (red), thrombin (blue),
Ala190Ser thrombin (light blue), and Glu192Gln thrombin (pale blue).
the complex with 3F, is displaced by the attached methyl group
for steric reasons (Figure 3A). Likely, the loss of water W2
impacts the stabilization of the adjacent W1 in the complex
with the 4-methyl derivative as the difference density does not
disclose the presence of this water molecule in the complex as
well.
synthesis of the compounds is described in the Experimental
Section.
The affinities of 1L−3L deviate not only in total potency but
also in the individual trends from those of the pairwise
matching fragment-like P1 inhibitors. Overall, an affinity
improvement of roughly 1.5−2.5 orders of magnitude is
observed compared with the P1 fragments, which results from
additional interactions experienced by the extended ligand
scaffolds in the S2 and S3/S4 pockets of both enzymes. The
expanded inhibitors are more potent against thrombin than
trypsin, especially in the case of both amidine derivatives 1L
and 2L. This likely correlates with the fact that the
corresponding specificity pockets are structurally better
developed in thrombin and allow to form more favorable
interactions with this protein.9 Remarkable are the binding
data for the D-Phe-Pro-NH-CH2-benzylamine 3L as this ligand
seems to recover a major part of the affinity difference found
between trypsin and thrombin for the P1 fragment 3F.
Compound 3L is now nearly equally potent against both
enzymes and the two mutated thrombin variants.
The affinity data suggest that 3L must adopt a different
geometry in thrombin as 3F, its fragment-like head group
analog. The crystal structures confirm this assumption (Figure
5A). A superposition of the complexes with 3L in trypsin and
thrombin shows very similar binding modes (Figure 6).
Accordingly, all three inhibitors (1L, 2L, and 3L) now bind
with their basic P1 head groups into the S1 pocket toward
Asp189. As a mutual superposition shows, identical ligands
establish virtually the same binding modes in either trypsin or
thrombin. In thrombin, the basic P1 group of 3L is now
pointing toward Asp189 and it forms, similarly as the
corresponding P1 fragment in trypsin, H-bonds to Asp189,
Gly219, and the carbonyl oxygen of residue 190. Additionally,
Asp189 forms via W5 a water-mediated contact to Tyr228.
The side chain of Glu192 in the thrombin complex adopts two
alternative conformations. The lower populated one (II, 37%)
orients away from the bound ligand. Instead, the higher
occupied orientation (I, 63%) bends back toward the S1 pocket
and forms via W3 an H-bonded contact to Gly219NH. The
corresponding Gln192 residue adopts in trypsin a single
orientation, which aligns well with the back-folded conformer
II in thrombin. Nevertheless, 3L is still lower in affinity against
thrombin in comparison to 1L and 2L. Considering that the
extended tripeptide-like analogues collect a major part of their
affinity enhancement with the P2 and P3 portions, the
contribution of the benzylamine head group of 3L cannot
In the case of 3F, a second copy of the fragment is found in
the distal S3/S4 aryl binding site of thrombin with a slightly
inclined edge-to-face geometry on top of Trp215 (Figure 3B).
The amino group is hydrogen-bonded to the backbone
carbonyl oxygen of Glu97a, and a further contact is mediated
via a water molecule to the hydroxy group of Tyr60A.
Interestingly, the 4-methyl analog of 3F does not accom-
modate the S3/S4 pocket, likely because the attached methyl
group hampers the formation of an efficient edge-to-face
packing. Additionally, we tried to soak further 4-alkyl
benzylamine analogs with longer substituents than a 4-methyl
group into thrombin crystals; however, of these, none could be
observed as bound to the protease. Likely, the basic head
group preferentially selects the carboxylate group of Glu192
instead of Asp189. Supposedly, it forms more beneficial
interactions with the acid group of this residue (see discussion
below). For steric reasons, derivatives with larger 4-alkyl
substituents than methyl are no longer bound to thrombin’s S1
pocket, preventing a similar binding mode. Obviously, even
with the larger side chains, Asp189 is not selected as the
interaction partner. In comparison to the previously studied
basic head groups benzamidine and N-amidinopiperidine, both
showing good affinity toward both proteases, our structural
data for the 3F thrombin complex explain the strongly
deviating assay results obtained for trypsin and thrombin.
Supposedly, the second copy of 3F in the S3/S4 pocket of
thrombin is only weakly binding as no salt bridges are formed.
The refinement, however, indicates full population of 3F in the
S3/S4 pocket. Replacing Glu192 by Gln in thrombin results in
an even stronger or complete loss of binding affinity, which
falls below the detection limit of the applied assay (Figure 1).
Binding Data and Crystal Structures of Extended
Tripeptide-Like D-Phe-Pro-NH-CH2-R Ligands. The rather
surprising binding-mode difference of benzylamine in trypsin
and thrombin stimulated us to study also the properties of
elongated D-Phe-Pro-NH-CH2-R inhibitors with the three P1
basic R-groups benzamidine (1L), N-amidinopiperidine (2L),
and benzylamine (3L) (Figure 4) to assess whether the
extended ligand scaffold enforced a binding mode with the
benzylamine head group oriented toward Asp189. The
1614
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
keep up with that experienced by the P1 groups of 1L and 2L.
The latter amidine-substituted head groups are definitely more
appropriate to achieve strong binding to thrombin. Likely, the
situation is different in trypsin where 3L is even slightly more
potent than 2L.
Furthermore, the large affinity difference of 1L and 2L is
surprising against trypsin and thrombin. It matches for trypsin
with the data already found for the P1 fragment-like head
groups 1F and 2F where benzamidine is preferred.
Remarkably, it is reversed in the case of thrombin (Figures 1
and 4) where 2F is somewhat more potent than 1F. To shed
some light on the affinity difference between the benzamidine
and N-amidinopiperidine head groups toward both enzymes,
we performed in another contribution11 elaborate molecular
dynamics simulations along with GIST (grid inhomogeneous
solvation theory) evaluations19−21 of the trajectories to
estimate on the solvation thermodynamic parameters.19 This
study showed that the affinity difference between benzamidine
and N-amidinopiperidine fragments is explained, at least to
some part, by deviating solvation patterns in the S1 pocket.
They result from differing solvation barriers for the rehydration
of both proteins.13 Likely, such effects, found for the P1 head
groups, cannot be neglected in discussing the affinity and
selectivity differences of 1L and 2L.
We studied the geometries of 2L with both proteins in more
detail. Again, virtually, the same binding poses are observed.
This is in contrast to the geometries found for the fragment-
like 2F, as we reported previously.8,9 The fragment-like ligand
2F binds to trypsin with two alternative configurations at the
ipso nitrogen of the guanidinium group (Figure 5B). It adopts
either a planar (63%) or pyramidal geometry (37%). The
second arrangement breaks up the conjugation of the
delocalized electron system and concentrates the positive
charge on the terminal nitrogens of the guanidinium group
directly involved in the salt bridge with the carboxylate group
of Asp189. Remarkably, in thrombin, 2F adopts full planar
geometry (Figure 5C).11 In our previous study,8 we performed
a search in the Cambridge Structural Database (CSD), which
showed that pyramidalization at N occurs predominantly in
cases where the guanidinum group is involved in strongly
polarizing bidentate salt bridges to a carboxylate ion in the
crystal structures. In contrast, examples lacking this contacting
salt bridge remain much closer to a planar arrangement. These
results suggest that the Asp189 carboxylate group has a
stronger polarizing effect in trypsin than in thrombin.7
Interestingly, the extended D-Phe-Pro analogue 2L is
exclusively found in a pyramidal geometry at the designated
ipso nitrogen in both proteins. This is explained by the spatially
restricting wall of the binding pockets, whereas a planar
geometry would force the attached substituent to occupy the
unfavorable axial position, creating sterically inconvenient 1,3-
diaxial interactions across the six-membered ring (Figure
5D,E). To avoid such unfavorable geometry, the restrictions
force 2L to adopt a pyramidal configuration at its ipso nitrogen.
This leads to a stronger localization of the electron density on
the exocyclic nitrogens and thus enhances the electrostatic
interactions with the carboxylate group of Asp189. This
observation correlates with a stronger relative affinity increase
of nearly 2 orders of magnitude in the case of thrombin
compared with trypsin while expanding the P1 head group N-
amidinopiperidine to the tripeptidic analogue 2L with the
simultaneous transition from a planar (2F) to pyramidal
geometry (2L).
Figure 5. (A) X-ray crystal structure of 3L with thrombin (blue, PDB
code: 6T53). The corresponding complex with 3F (beige, PDB code:
6T56) is superimposed. Glu192 adopts two alternative conformations,
I (63% populated) and II (37%), in the complex with 3L. The former
corresponds to the geometry found with 3F. (B) In the complex with
trypsin, N-amidinopiperidine (2F) adopts two different configurations
at the ipso nitrogen, showing planar (turquoise, PDB code: 5MNN8)
and pyramidal geometry (olive). (C) In the corresponding complex
with thrombin, only the planar geometry at the ipso nitrogen is found
(turquoise, PDB code: 4UE711). (D) The extended ligand 2L occurs
only with pyramidal geometry and adopts in both proteins (trypsin,
light pink (PDB code: 5MNQ8) and thrombin, magenta (PDB code:
6T57)) a virtually identical geometry. (E) The planar geometry found
for 2F (turquoise, PDB code: 4UE711) cannot be established by 2L
(magenta, PDB code: 6T57) in the binding pocket (shown for the
thrombin complex) for steric reasons. In the bound structure of 2L,
only an equatorial substitution at the C4 atom of the piperidine ring is
accepted, whereas an axial arrangement is avoided due to unfavorable
1,3-diaxial interactions (see insert).
Figure 6. Superposition of the complexes with 3L in trypsin (gray,
waters red spheres, PDB code: 6YDY) and thrombin (blue, waters
dark blue spheres, PDB code: 6T53). The side chain of Glu192
adopts two alternative conformations in thrombin (I: 63% toward the
ligand and II: 37%, away from the bound ligand). The corresponding
Gln192 in trypsin is found in only one single orientation toward the
ligand.
1615
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
a
b
Table 1. Thermodynamic data of ΔG° , ΔH°, and −TΔS° for binding of 1L−3L to trypsin and thrombin at pH = 7.8
thrombin
trypsin
a
b
c
d
a
b
c
d
+
ΔG° kJ mol−1
ΔH° kJ mol−1
−TΔS° kJ mol−1
ΔnH mol
ΔG° kJ mol−1
ΔH° kJ mol−1
−TΔS ° kJ mol−1
ΔnH mol
+
1L
2L
3L
−45.2
−42.9
−34.8
−38.8
−39.5
−19.2
−6.4
−3.4
−15.6
0.04
−0.15
−0.04
−42.6
−31.1
−33.4
−38.2
−21.8
−28.9
−4.4
−9.3
−4.6
0.17
0.12
0.20
a
b
ΔG° is given as the mean of titrations in three different buffers (HEPES, tricine, and tris). All ΔH° results are presented as buffer-corrected
values. −TΔS° has been calculated as the difference between ΔG° and the buffer-corrected ΔH° value. Molar amount of transferred protons
results from the slope regression line obtained for the buffer correction.
c
d
Figure 7. Crystallographically determined binding modes of 1L, 2L, and 3L with thrombin (upper row, A−C) and trypsin (lower row, D−F) in the
S1 pocket next to water molecule W2. (A) 1L (cyan) with thrombin (PDB code: 2ZDA22,25). (B) 2L (magenta) with thrombin (PDB code: 6T57).
(C) 3L (light blue) with thrombin (PDB code: 6T53). (D) 1L (light green) with trypsin (PDB code: 6ZQ2). (E) 2L (pale) with trypsin (PDB
code: 5MNQ8). (F) 3L (dark blue) with trypsin (PDB code: 6YDY). Of the bound ligands, only the polar head group is shown. In all complexes,
W2 forms interactions with the surrounding amino acids, mostly as hydrogen bonds or expanded polar contacts. With Tyr228, W2 interacts with
the aromatic moiety of this residue; all distances are in Å. The interaction inventory in thrombin is lower due to the Ser190Ala replacement.
Particularly, in the thrombin-3L complex (C), the interaction pattern is weakened as the bound ligand does not interact with W2.
ITC Titrations of 3L. We recorded ITC experiments
(isothermal titration calorimetry) with 3L and both enzymes
from three different buffers at a pH of 7.8 (Table 1) along with
those for 1L and 2L.8,22 All experiments show hardly any
buffer dependence. We can therefore conclude that 3L, having
a pKa value of 9.3 in aqueous solution, binds to the protein
without changing its charged protonation state. The same is
valid for the more basic head groups of 1L and 2L. As we could
show in previous studies, the terminal P3 amino group
becomes fully protonated upon binding. With respect to the
overall protonation inventory, this is compensated by the
similar amount of protons released from His57 upon ligand
binding.22−24
The formation of the five complexes of 1L, 2L, and 3L with
trypsin and 1L and 2L with thrombin is enthalpically favored,
which agrees with the observation that a salt bridge is formed
to Asp189. The binding of ligand 3L to thrombin falls
somewhat out of this series as its binding exhibits a remarkably
stronger entropic signature compared to the other complexes.
This observation is striking, although the latter complex
involves, similarly as with trypsin (Figure 6), the ligand’s basic
head group in a salt bridge with Asp189. All the more, the
entropically strongly favored profile is surprising. However,
there is an important difference considering the interaction
inventory of W2 in the six complexes (Figure 7). In the case of
1L and 2L, the charged amidino function establishes an
H-bond to W2 in both enzymes. In contrast, the ligand’s basic
head group of 3L orients away from W2; thus, any H-bonding
network between this water to the ligand is missing (Figure
7C,F). However, since trypsin exhibits a serine at position 190
and thrombin an alanine, W2 can only form an H-bond to
Ser190Oγ in the trypsin complex, whereas in the thrombin
complex, this water molecule remains without a similar
hydrogen-bonding interaction. It solely contacts
Phe227CO via a proper hydrogen bond, and a similar
contact is found to Val227 in the trypsin complex. The
expanded contact to Trp215 and the interaction with the
aromatic ring of Tyr228 help in keeping this water molecule in
position. Thus, the significantly reduced inventory of polar and
directional H-bonds around W2 in the complex of 3L with
thrombin suggests enhanced residual mobility of W2 in the
latter complex compared to all other ones, which embed W2 in
a network of several hydrogen bonds. Supposedly, this fact
contributes to the enhanced entropically favored thermody-
namic profile of the 3L-thrombin complex.
Affinity and Structural Data of the Mutated
Ala190Ser Variant of Thrombin. As mentioned, the only
obvious difference within the S1 pockets of thrombin and
trypsin is the replacement of alanine by serine in position 190.
This exchange has been made responsible for the selectivity
1616
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 8. Crystallographically determined binding modes of 1L and 4L with wild-type (A, C) and Ala190Ser (B, D) mutated variant of thrombin
((A) PDB code: 2ZDA, 1.73 Å22,25 and (B) PDB code: 5MM6, 1.29 Å) and 4L ((C) PDB code: 2ZC9, 1.59 Å22,25 and (D) PDB code: 5MLS, 1.62
Å). The structural pairs of the wild-type and mutated variants with the two ligands are highly similar. The only apparent difference concerns
Ser190Oγ. Its hydroxyl group stabilized W2 in the complex with 1L through an additional H-bond. In the complex with 4L, the Ser190 hydroxyl
group orients toward Tyr228 and forms an additional intramolecular H-bond.
discrimination of both proteases.6,7,26 To make thrombin more
trypsin-like, we exchanged Ala by Ser via site-directed
mutagenesis. In order to assess whether the Ala190Ser
replacement can be made responsible for the selectivity
difference, we determined binding affinities with our
fluorometric assay and solved the crystal structures of two
ligands in the complex with wild-type thrombin and the
mutated variant. The affinity difference with respect to the wild
type of the studied ligands (Figures 1 and 4) appears minor
apart from N-amidinopiperidine 2F where a difference of
approximately 2 orders of magnitude is surprisingly experi-
enced.
To trace possible differences in structural terms, we
determined the binding modes of two inhibitors 1L and 4L
by crystallography in thrombin and its Ala190Ser variant
(Figure 8). By intention, we selected a ligand with a basic
benzamidino and non-basic m-chloroaromatic benzyl head
group for the S1 pocket. Both ligands had already been
characterized previously with wild-type thrombin (protein data
bank-codes (PDB codes): 2ZDA-1L and 2ZC9-4L)22,25 and
showed, according to the enzyme kinetic assay, a negligible
affinity difference (1Lwild‑type: 0.55 0.04 nM, 1LAla190Ser: 0.55
0.06 nM, 4Lwild‑type: 590 10 nM, and 4LAla190Ser: 320 50
nM).
All structures are highly similar. In the complex of the
mutated variant with 1L, comprising the basic head group, the
water molecule W2 connected to the ligand experiences an
additional hydrogen bond to the side-chain hydroxyl group of
Ser190 (Figure 8B). This contact is lacking in the wild type,
which possesses instead an alanine at this position. Obviously,
the Ala190Ser replacement has no effects on the binding mode.
We have observed similar differences in other structurally
highly related complexes of wild-type thrombin and trypsin9
and in the binding mode difference of 3L, with both enzymes
(Figure 6), which makes us confident that the mutated variant
reflects correctly the hardly given differences between both
proteases. This is at least valid for the affinity data, the
enthalpy/entropy partitioning might be different as suggested
by the 3L example above. We refrained from collecting ITC
data for the mutated variants as ITC requires large amounts of
protein, which are difficult to produce with the recombinant
technique. Furthermore, since production in Escherichia coli
does not provide the enzyme in a glycosylated form, the overall
solubility of the recombinantly produced thrombin is
significantly lower, an aspect that further complicates ITC
experiments. In the complex with 4L, which repels W2 for
steric reasons from the binding site, Ser190Oγ in the
Ala190Ser variant orients away from the hydrophobic P1
ligand portion and forms an H-bond to Tyr228OH (2.9 Å).
As hardly any differences in the affinity and binding pose are
recognized that could be linked to the Ala190Ser replacement,
we wanted to confirm this observation by selecting a larger set
of 54 diverse inhibitors (Table S1, Supporting Information).
The entries of this data set showed distinct binding affinities
toward the wild types of thrombin and trypsin (Figure 9A, R2 =
0.51). We therefore tested them in our kinetic enzyme assay
for inhibitory potency differences between the wild-type and
Ala190Ser variant of thrombin. A nearly perfect linear
correlation was found (R2 = 0.98), indicating that no
significant difference between both thrombin forms exists
(Figure 9B). This assumption is further supported by the fact
that the regression line exhibits virtually the same slope as the
main diagonal drawn in the diagram (Figure 9B).
Affinity and Structural Data for the Mutated Variant
Glu192Gln in Thrombin. A further difference between
thrombin and trypsin relates to residue Glu192, located at the
rim of the S1 pocket. This residue is replaced by an uncharged
Gln192 in trypsin. Glu192 shows substantial conformational
flexibility, and in the many studied thrombin−ligand
complexes,11 it is frequently involved in the binding of
particularly polar P1 head groups.9 The fragment complex of
3F with thrombin has impressively demonstrated this
surprising behavior of Glu192 resulting in a back-folded
conformation of this residue. On first glance, the latter complex
suggests that, as the fragment prefers the carboxylate group of
the less acidic Glu192 over that of Asp189, the concentration
1617
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 9. Correlation of assay data for wild-type trypsin and thrombin and four mutated variants of thrombin; the equation for the regression line
and mean R2 are listed. (A) Trypsin vs thrombin, (B) thrombin vs Ala190Ser thrombin, (C) thrombin vs Glu192Gln thrombin, (D) thrombin vs
Asp221Ala-Asp222Lys thrombin, (E) thrombin vs Tyr225Pro thrombin, (F) trypsin vs Tyr225Pro thrombin. (A−D) 54 data points; (E , F) 15
data points.
of the negative charge on Asp189 is attenuated in thrombin.
This could result from the nearby sodium ion (5.3 Å, PDB
code: 6TDT9). Obviously, an electrostatic gradient is formed
across the S1 pocket spanning from Glu192 to the Na+ ion,
which might assist in relocating the charges. A similar gradient
is not given at the trypsin site. To validate whether this
difference in charge distribution has any impact on binding and
possibly on selectivity, we generated the mutated Glu192Gln
variant of thrombin, which removes the negative charge at the
rim of the pocket.
The Glu192Gln variant was tested by the kinetic enzyme
assay across the abovementioned reference set of diverse
protease inhibitors (Figure 9C). Similar to the Ala190Ser
exchange, the replacement of charged Glu192 by uncharged
Gln in thrombin has hardly any impact on the affinity data
across the 54 tested compounds (R2 = 0.98). A minor
difference is perhaps indicated with respect to the Ala190Ser
variant as the slopes of the regression line deviates slightly from
the diagonal.
A crystal structure of the Glu192Gln variant could be
determined with D-Phe-Pro-NH-CH2-4-(2-amino)pyridine
(5L), a ligand that was already characterized crystallo-
graphically with the wild type of trypsin and thrombin in a
previous study (Figure 10B).9
Interestingly, all three complexes with 5L are highly similar.
However, one important difference is visible. Whereas the
binding mode of the ligand is ordered in trypsin, in wild-type
thrombin, and in the Glu192Gln variant, the ligand scatters
over two different orientations in the S1 pocket. While the
population of the two orientations in the variant Glu192Gln
refines to (I) 47% and (II) 53%, the population of the two
orientations in wild-type thrombin converged to (I) 76% and
(II) 24%. In our previous study, we could show that the scatter
over two orientations has a slight impact on the partitioning of
the thermodynamic properties. Disorder over multiple states
1618
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 10. Crystal structures of 5L (light green) with (A) thrombin (PDB code: 6T0P9), (B) the Glu192Gln thrombin variant (not deposited as
virtually identical with the wild-type structure), and (C) trypsin (PDB code: 6T3Q9). The binding mode in the S1 pocket next to Asp189 is shown.
The P1 residue in both thrombin complexes adopts two orientations. Selected protein residues are depicted as gray sticks and water molecules as
red spheres. (D) Structural formula of the ligand 5L.
Figure 11. (A) Close-up view of the sodium binding site in thrombin (PDB code: 2UUF14) next to the S1 pocket and Asp189. (B) Corresponding
region in trypsin (PDB code: 5MN117), which does not host a similar ion-binding site.
always increases the entropic contribution on binding and
reduces the enthalpic portion. Quite surprisingly, the charge on
residue 192 has only minor determining influence in thrombin
on the ligand’s binding mode as we observe the same geometry
in the wild-type and mutated variant of Glu192Gln and only
the ratio of the two orientations is altered. In trypsin, where
only one orientation of 5L is observed, Gln192 is directed
away from the S1 pocket. In addition, as seen with the
Ser190Ala variants, the hydroxyl group of the serine residue
fixes W2, which mediates the contact to the ligand.
Supposedly, these structural changes have minor influence on
the affinity distribution across the series of studied ligands
(Figure 10C). Nevertheless, the impact of the structural
changes is likely more pronounced considering the partitioning
of the free energy of binding into its enthalpic and entropic
contribution. This is indicated by the values found in Table 1
for inhibitors 1L−3L.
Affinity data for the mutated variants D221A-D222K
and Y225P in thrombin. We speculated already above that
the sodium ion, exclusively present in thrombin only, has an
influence on the charge distribution of Asp189. We therefore
extended our mutational studies on the sodium-binding site in
thrombin. However, modifying this site is more challenging as
four water molecules and two backbone carbonyl oxygen
atoms establish an octahedral coordination sphere around this
ion. This geometry is difficult to remove or replace without
introducing major changes of the protein architecture. Figure
11A shows a section of the thrombin structure hosting the
sodium ion-binding site. In Figure 11B, the related area in
trypsin is depicted.
1619
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Adjacent to the sodium-ion site, Asp221 and Asp222 are
found, which likely stabilize the binding of Na+ in this area.
Trypsin, which lacks the sodium-ion site in this region, holds at
the corresponding positions Ala and Lys residues. We therefore
planned the double mutant D221A-D222K as a promising
variant to alter the sodium-ion binding site.27 The larger and
positively charged Lys residue was assumed to interfere
unfavorably with the accommodation of Na+ for steric and
electrostatic reasons. As an alternative, we expected that the
replacement of tyrosine in thrombin by proline as the
corresponding residue in trypsin at position 225 would also
make thrombin more like the digestive protease and destabilize
the adjacent sodium-binding site.
CONCLUSIONS
■
In fact, the selectivity-determining features in the S1 pockets of
trypsin and thrombin are difficult to impose to one single
dominating factor, such as a single exchange of an amino acid
in the S1 pocket. Hence, selectivity results as a complex
interplay of several aspects. The local geometry of both pockets
is highly conserved and the sole Ala190Ser replacement may
be regarded as perfectly conserved, as indicated by the crystal
structures determined in parallel for thrombin, trypsin or the
Ala190Ser and Glu192Gln variants of thrombin. Thus, simple
steric features cannot explain the selectivity difference. The
Glu192Gln replacement at the rim of the S1 pocket has little
impact with regard to steric and dynamic properties, both
residues require similar space and multiple crystal structures
prove the flexibility of the residue in position 192. The affinity
data demonstrate the observed high similarity, remarkably a
deviating partitioning in enthalpy and entropy contributions
points more strongly to a given difference. Likely, this results
from the stronger scattering of residue 192 in thrombin
compared to trypsin, which also takes impact on the adopted
binding pose of the bound ligand. In terms of charge
distribution, Glu192, together with the thrombin-specific
sodium ion-binding site adjacent to Asp189, helps in creating
an electrostatic gradient across the S1 pocket, a feature
definitely differing in trypsin. The observed induced proto-
nation effects, reported in our previous paper for the basic
pyridine head groups (pKa = 5.0), along with the deviating
binding poses of 3F in both proteases, were a first indication
for a significant charge attenuation on the carboxylate group of
Asp189 in thrombin compared to trypsin.8,9 We advocate that
this attribute is one of the prime selectivity-determining
features between both proteases. In consequence, it influences
and controls the other, on a first glance, less obvious
selectivity-discriminating features, such as differences in the
solvation pattern and ordering of water molecules in both
enzymes. The difference in the partitioning of enthalpy and
entropy for the trypsin and thrombin complexes with 3L
further underscores the importance of the solvation pattern for
the properties of the formed complexes (Table 1).
Unfortunately, our current understanding of the influence of
water solvation features on ligand and substrate binding,
particularly with respect to solvation binding kinetics, is quite
rudimentary. Nevertheless, even though both proteases
comprise very similar recognition pockets, their properties
strongly differ in the internal architectures and charge
distribution, enabling the formation of deviating water
inventories important for ligand binding and unbinding.
Already, in the uncomplexed proteins, the carboxylate group
of Asp189 in thrombin (PDB code: 2UUF) is solvated by a
network of three water molecules, whereas in trypsin, the same
group is solvated by only two water molecules. Particularly, the
neutron diffraction study of apo trypsin demonstrates that the
immediate solvation pattern of the carboxylate group of
Asp189 is far from ideal.8 The imperfect and highly perturbed
geometry of the first solvation shell around Asp189 is an
important prerequisite for a potent substrate and, in
consequence, ligand binding. Otherwise, the desolvation of
the charged carboxylate in the deep S1 pocket would be
energetically very costly. Furthermore, in trypsin, a water
reservoir is found below Asp189, which is the source of water
molecules needed for the association and dissociation process
of ligands. Remarkably in thrombin, a water channel, which
Both mutated variants of thrombin were generated and
tested by our kinetic enzyme assay across the abovementioned
reference set of diverse protease inhibitors (Figure 9D−F).
The D221A-D222K variant has, at least with respect to the
enzyme kinetic analysis, no strong impact on the binding data
as suggested by an R2 = 0.935. However, the slope of the
regression line deviates here more strongly from the main
diagonal. This variant has already been described by Pineda et
al.27 and a crystal structure could be determined (PDB code:
1TWX27). Overall, the geometrical variations seem to be
minor (root-mean-square deviation (rmsd) = 0.5 Å). Small
translocations of carbonyl groups coordinating the sodium ion
in the wild type and the lack of a density peak at the former
sodium-binding site suggest the absence of the ion. However, it
remains to be considered that the diffraction power of a
sodium ion is nearly identical to that of a water molecule and
at a resolution of 2.4 Å, the assignment of water becomes
problematic. Nevertheless, it might well be in agreement with
our assay results that the D221A-D222K variant is still rather
close to the structural properties of wild-type thrombin.
Finally, the kinetic data of the Y225P variant indicate
stronger deviations from the wild-type assay results. This is
indicated by an R2 = 0.74 with respect to wild-type thrombin
or 0.66 to wild-type trypsin, respectively. Here, the slope of the
regression lines deviate significantly from the main diagonal.
Due to the strongly reduced activity of the Y225P variant, a
large amount of protein material was required to perform the
measurements. Since the expression of this variant provided
only a very low yield, we had to limit our testing of the
inhibitory power to a set of a few selected ligands.
Unfortunately, we also did not succeed to crystallize the
Y225P variant to determine its structure. Nevertheless, if we
assume that the mutated Y225P variant alters the stability and
in consequence the occupancy of the sodium-binding site in
thrombin more strongly than in the other studied variants, the
assay data support our hypothesis of the sodium-ion influence.
The presence of the positively charged sodium ion seems to
have a deviating impact on the binding features of both
proteases, supposedly via a significant attenuation of the
negative charge concentrated on the carboxylate group of
Asp189 in thrombin. Functionally, the occupancy or absence
of a sodium ion in thrombin has been discussed28 to explain
why a kinetically slow and fast form of the enzyme exists. The
present comparative study of trypsin and thrombin, inves-
tigated in structural and thermodynamic terms under
equilibrium conditions, suggests affinity differences but does
not allow conclusions on differences in the enzyme kinetic
properties.
1620
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
2.50 ppm (1H NMR), 39.5 ppm (13C NMR), DMSO-d6 (deuterated
dimethyl sulfoxide). The multiplicity of the signals is described with
the following abbreviations: s = singlet, d = doublet, t = triplet, q =
quartet, dd = doublet of doublet, m = multiplet, br = broad signal.
The coupling constants J are given in hertz. NMR spectra of ligand 3L
represent the signals of the main conformer.
facilities water exchange with a bulk water phase, replaces the
water reservoir in trypsin. Elaborate MD simulations have
shown that this difference has a decisive influence on the
solvation kinetics and, in consequence, on the selectivity of
ligands binding to both enzymes.13 As a result, the dissociation
of ligands from trypsin affords a larger barrier because they
must dissociate before the site becomes re-hydrated, while in
thrombin, re-hydration and ligand dissociation proceed
simultaneously. Therefore, the ligand-binding mechanism is
not only determined by the established protein−ligand
interactions but also by the differing solvation barriers of
both proteins, which adds to the deviating desolvation
properties of the different ligand molecules, yet another factor
in selectivity discrimination.
1H NMR (500 MHz, DMSO-d6) δ = 8.38−8.35 (m, 4H), 8.24 (s,
br), 7.38−7.22 (m, 9H), 4.36−4.21 (m, 4H), 4.02−3.98 (m, 2H),
3.53−3.49 (m, 1H), 3.12−3.08 (dd, J = 13.2, 5.9 Hz, 1H), 3.01−2.95
(dd, J = 13.3, 8.6 Hz, 1H), 2.69−2.63 (m, 1H), 1.80−1.67 (m, 3H),
1.47−1.41 (m, 1H). 13C NMR (126 MHz, DMSO-d6) δ = 170.9,
167.0, 158.4 (q, J = 33.7 Hz), 139.9, 134.5, 132.3, 129.5, 128.8, 128.6,
127.4, 127.0, 116.5 (q, J = 295.6 Hz), 60.0, 52.0, 46.7, 42.0, 41.6, 36.7,
29.4, 23.7.
MS spectra was measured on a Q-Trap 2000 system with an
electrospray interface (ESI). MS (ESI+) m/z calculated for
C22H29N4O2 [M+H]+: 381.23; found: 381.30.
The series of ligands studied with trypsin by our neutron
diffraction investigations unravel unexpected differences in the
residual dynamics of water molecules in the S1 pocket.
Surprisingly, the rotational and translational behaviors of the
water molecules differ not only between uncomplexed and
complexed states but also between the complexes hosting
different ligands. In the trypsin−benzylamine complex studied
here, water W1 alters its dynamic properties from a disordered
to an ordered state (Figure 2A,B). Water molecule W2 is also
disordered in apo trypsin and becomes ordered in the complex
with 3F by experiencing a weak H-bond with Ser190Oγ. This
residue is replaced by alanine in the S1 pocket of thrombin and
suggests on a first glance that the lacking Oγ and thus the loss
of a hydrogen bond have consequences for selectivity. The
thermodynamic data suggest no difference in ΔG but in the
enthalpy/entropy partitioning. Our panel of 54 tested ligands
also speaks against an impact on affinity; however, the
partitioning of enthalpy and entropy cannot be excluded and
remains to be shown for a larger number of cases. As our
crystallographic analysis shows, the Ala190Ser thrombin
variant is structurally closer to trypsin at position 190, but
with respect to the remaining solvation features of the pocket,
it is still a chimera closer to thrombin than trypsin. This makes
the interpretation of mutational differences rather inscrutable.
Unfortunately, the X-ray structures collected with thrombin do
not disclose the required details about the dynamics of the
water molecules in the S1 pocket. Accordingly, more
experimental data in terms of neutron diffraction studies
complemented by MD simulations are required to further trace
the water influence of selectivity between these closely related
serine proteases.
Flash column chromatography was performed on silica gel (0.04−
0.063 mm, Macherey-Nagel, Duren, Germany) and monitored by
̈
thin-layer chromatography (TLC) on aluminum sheets (Silicagel 60
F254, Merck, Darmstadt, Germany) with UV light at 254 or 366 nm.
Preparative HPLC was performed with a Varian HPLC system
gradient system (reversed-phase column: Nucleodur C18, 5 μm, 100 Å
, 32 × 250 mm, Macherey-Nagel, Duren, Germany). All solvents were
̈
of HPLC grade, and in a gradient run (solvent A: 0.1 % TFA in water,
solvent B: 0.1 % TFA in acetonitrile), the percentage of solvent B was
increased by 0.5% solvent min−1 at a flow rate of 20 mL min−1. The
purity of 3L used for this study was at least 95% as determined by
analytical HPLC with a Shimadzu LC-10A system using a gradient
with the same solvents as described above for the preparative HPLC
(increase of 1 % sovent B/min, reversed-phase column: Nucleodur
C18, 5 μm, 100 Å, 4.6 × 250 mm, Macherey-Nagel, Duren, Germany).
̈
All solvents were of HPLC grade and in a gradient run, the percentage
of acetonitrile was increased by 1% solvent min−1 at a flow rate of 1
mL min−1. The detection was recorded at a wavelength of 220 nm. 3L
was obtained as TFA salt after lyophilization.
Quantitative 1H NMR Experiments. In order to assess the
accuracy of our enzyme kinetic and thermodynamic results, purities of
the ligands 1F−3F was determined by quantitative 1H NMR
(qNMR) spectroscopy, as similarly described in ref 17, on a JEOL
ECA-500 MHz spectrometer. Purity was in all cases >95%.
Enzyme Inhibition Kinetics. The kinetic inhibition constants, Ki,
of a series of ligands and fragments (see Figure S3), relative to human
thrombin, trypsin, and thrombin variants A190S, E192Q, D221A-
D222K, and Y225P, were determined photometrically by a kinetic
fluorescence assay according to Stu
̈
rzebecher et al.30 For human
thrombin and all thrombin variants, the tripeptide Tos-Gly-Pro-Arg-
AMC·TFA (S1 = 10 μM, S2 = 5 μM, S3 = 2.5 μM) was used as a
substrate. For trypsin, we used Mes-D-Arg-Gly-Arg-AMC·2TFA (S1 =
25 μM, S2 = 12.5 μM, S3 = 6.25 μM). Both peptidic substrates
contained the fluorophoric 7-amido-4-methylcoumarins (AMC) (see
Figure S2). The amount of the released AMC after substrate cleavage
was determined at λex = 355 and λem = 460 nm. The Km values of
purified human thrombin, recombinantly expressed thrombin, and the
recombinant Ala190Ser variant in E. coli for the chromogenic
substrate were determined and agree reasonably well and allow for
a direct comparison of the assay data (Km(purified human thrombin):
3.1 μM, Km(recomb. thrombin, E. coli): 4.7 μM, and Km(A190S
variant, E. coli): 5.0 μM).
EXPERIMENTAL SECTION
■
Site-directed Mutagenesis, Expression, and Purification.
Mutation and overexpression of the thrombin Glu192Gln gene and
mutated variants thereof as well as protein purification were done as
described in Supporting Information.
Synthesis of D-Phe-Pro-NH-CH2-3F (3L). The amide formation
was afforded under coupling conditions similar to a procedure of Ngo
et al. with tert-butyl 4-(aminomethyl)benzylcarbamate and Boc-D-
Phe-Pro-OH.9 The resulting compound (0.175 g, 0.301 mmol) was
used for Boc deprotection with trifluoroacetic acid (TFA) (3.0 mL)
for 2 h at room temperature. The mixture was concentrated in vacuo,
and the crude product was purified by preparative high-performance
liquid chromatography (HPLC). 3L (0.186 mg, 0.273 mmol, 91%)
was obtained as a colorless solid after lyophilization.
The inhibitors were dissolved in DMSO at specific concentrations
depending on their expected inhibition strength. With respect to their
activity, the proteases were diluted in the assay buffer (50 mM Tris
(Tris(hydroxymethyl)aminomethane), 154 mM NaCl, 0.1% (w/v)
polyethylene glycol (PEG) 8000, pH 7.8) to obtain a concentration
suitable for measurement. The measurement was performed with a
Fluoroskan Ascent from Thermo Fisher Scientific, Waltham (U.S.A.)
at intervals of 20 × 15 s, preceded by an initial mixing time of 40 s.
The Ki values were determined as described by Dixon.31
1H and 13C nuclear magnetic resonance spectroscopy (NMR)
spectra were measured on a JEOL ECX-400 or JEOL ECA-500
instrument. Chemical shifts are reported in parts per million using
residual peaks for the deuterated solvents as an internal standard:29
Isothermal Titration Calorimetry (ITC). The ITC experiments
for 1L in thrombin were previously published by Baum et al.22 All
1621
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
ITC experiments were performed on a Microcal ITC200 (Malvern
Panalytical, Malvern, U.K.) device at 25 °C with a reference power of
5 μcal/s. Ligand solutions of 0.5 mM 2L and 0.6 mM 3L were titrated
into 50 and 60 μM thrombin, respectively. Thrombin samples were
prepared by a solution of human α-thrombin (Beriplast, CSL Behring,
Marburg, Germany)32 in a buffer of 100 mM NaCl, 0.1% (w/v) PEG
8000, HEPES (4-(2-hydroxyethyl)-1-piperazine ethane sulfonic acid),
Tricine (N-(tri(hydroxymethyl)methyl)glycine), or Tris at pH 7.8,
which was dialyzed at 4 °C overnight. Ligands were dissolved in the
same buffer with 3% (v/v) DMSO to assure a complete solubility.
The titration protocol consisted of an initial volume of 0.3 μL
followed by 15 injections of 1.4 μL for 2L and 12 injections of 1.9 μL
for 3L separated by an interval of 180 s.
ITC measurements for 1L and 2L on trypsin were previously
described by Schiebel et al.8 3L was titrated analogous to the protocol
given in ref 8 using a direct titration approach. Protein samples were
prepared by dialyzing a solution of bovine β-trypsin (Sigma Aldrich,
product number T8003) against a buffer composed of 80 mM NaCl,
10 mM CaCl2, 0.1% (w/v) PEG 8000, and 100 mM, HEPES, Tricine,
or Tris at a pH of 7.6 and a temperature of 4 °C, while 3L was
dissolved in buffer contained 3% (v/v) DMSO. The titrations in the
three different buffer systems enabled correction for any proton
transfer occurring between the buffer and protein upon complex
formation. Data analysis followed the same scheme as described in an
earlier study.15 The ITC protocol consisted of an initial 0.3 μL
volume injection and 20−25 injections with a volume between 1.5
and 1.8 μL separated by a 220 s time interval. The reported
thermodynamic values are given as the average over three measure-
ments.
Growth and Preparation of Crystals. Crystals for Neutron
Diffraction Experiments. Bovine β-trypsin crystals were generated as
we reported previously with the following ligand-specific adapta-
tions.8,17 Benzylamine was used at a concentration of 50 mM as a co-
crystallization ligand. Using the sitting-drop vapor-diffusion method, a
large trypsin-benzylamine crystal (1.0 mm × 1.0 mm × 2.3 mm = 2.3
mm3) was grown at 4 °C in a center-well dish purchased from VWR
(product no. NUNC150260). The crystallization drop was composed
of 100 μL of trypsin-benzylamine solution (at 60 mg/mL) and 100 μL
of mother liquor, which in turn consisted of 0.1 M HEPES pH 7.5, 0.2
M (NH4)2SO4, and 15% (w/v) PEG8000. The remaining part of the
crystal preparation was identical to those described in detail in our
previous contributions.8,17 For neutron diffraction experiments, the
crystal was placed in a quartz glass capillary and exposed to the
monochromatic neutron beam on the BIODIFF instrument (λ = 2.67
Å) operated by the FRM II and JCNS at the Heinz Maier-Leibnitz
Zentrum (MLZ, Garching, Germany) in two different orientations;
169 and 143 neutron diffraction images were recorded in the first and
second pass, respectively (run 1: φ range of 67.6° with Δφ = 0.4°,
exposure = 50 min; run 2: φ range of 57.2° with Δφ = 0.4°, exposure
= 50 min). After finalization of the data collection, all diffraction
images have been processed using HKL2000.33
Crystals for X-ray Diffraction of Trypsin, Thrombin, and
Mutated Variants. Crystal Preparation, Soaking and Freezing of
the Thrombin Wild Type. Whereas the preparation of the hanging
drop crystallization plates with silicon grease was done at room
temperature, all other preparation steps as well as the pipetting of the
crystallization samples and the storage of the crystallization plates
were performed at 4 °C. Thrombin obtained from CSL Behring in
Marburg, Germany32 was dissolved in crystallization buffer (20 mM
NaH2PO4, 350 mM NaCl, 2 mM benzamidine, 0.5 mM sulfated
hirudin fragment (54-65, also named “hirugen” acquired from Bachem
in Bubendorf, Switzerland34), pH 7.5) at a concentration of 10 mg/
mL. Each crystallization drop consisted of 1 μL protein solution (8
mg/mL thrombin) in crystallization buffer and 1 μL of reservoir
solution (20 mM NaH2PO4, 27% (w/v) PEG8000, pH 7.5). In each
reservoir, 500 μL of reservoir solution was added. The drop was
seeded with former thrombin wild-type crystals using the streak
seeding method. The crystals finished growing after 3−4 weeks.
For soaking, a 50 mM stock solution for 1L−4L and 100 mM for
1F−3F was prepared in DMSO. The soaking process was performed
identically for all ligands and thrombin variants. It will be described
exemplarily for the E192Q thrombin variant with 5L.
Crystal Preparation, Soaking and Freezing of E192Q Crystals
with 5L. Each crystallization drop consisted of 1 μL of protein
solution in crystallization buffer (4 mg/mL active E192Q thrombin
variant, 2 mM benzamidine, 0.5 mg/mL sulfated hirudin) and 1 μL of
reservoir solution (20 mM NaH2PO4, 27% (w/v) PEG8000, pH 7.5).
The further procedure is identical to that of wild-type thrombin.
For soaking, a 50 mM stock solution of 5L was diluted in dimethyl
sulfoxide (DMSO) with a 1:10 ratio in the soaking solution (20 mM
NaH2PO4, 175 mM NaCl, 13.5% (w/v) PEG8000, pH 7.5). This led
to the final soaking concentration, which contained 5 mM of 5L and
10% DMSO. A crystal, which had no visible imperfections, was
transferred to a cover slip in a drop of the given solution. Similar to
crystallization, the wells containing 500 μL reservoir buffers were
closed with the cover slip and sealed with silicone gel. After 24 h, the
soaking was stopped and the crystal was transferred to a
cryoprotectant buffer (20% glycerol, 16 mM NaH2PO4, 21.6% (w/
v) PEG8000), containing 5 mM 5L for approximately 1 min. For data
collection, the crystal was frozen in liquid nitrogen.
Crystallization of Trypsin. Trypsin crystals were grown at 18 °C on
sitting-drop plates in 2 μL crystallization drops, respectively. A
trypsin−ligand solution of 40 m/mL trypsin from bovine pancreas
(Merck, Darmstadt, Germany) in protein buffer (1 mM HCl, 10 mM
CaCL2, 3 mM inhibitor in 100% DMSO) was prepared. Each
crystallization drop consisted of 1 μL of trypsin−ligand solution and
1 μL of reservoir buffer, which contained 0.1 M imidazole, 0.1 M
(NH4)2SO4, 0.1% (w/v) NaN3, and 25% (w/v) PEG8000 at pH 7−8.
The crystals finished growing after 2 weeks and were flash-frozen in
liquid nitrogen in a cryoprotectant solution composed of reservoir
buffer, 20% glycerol, and 3 mM ligand 1L or 3L, respectively. The
crystallization of 2L and 1F−3F in trypsin was done as previously
described by Schiebel et al.8
Structure Determination and Refinement. Joint Neutron and
X-ray Refinement of Trypsin·3F. The refinements against only
neutron or X-ray diffraction data as well as the joint refinement
against neutron and X-ray diffraction data have been performed as
reported previously.8 Please note that the deuterated trypsin−
benzylamine crystal used for room temperature X-ray diffraction
data collection (resulting in the structure with the PDB code 5MNL)
was found to be non-isomorphous to the crystal used for the neutron
diffraction experiment (see Tables S1 and S2). In contrast, a non-
deuterated trypsin−benzylamine crystal also used for a room-
temperature X-ray diffraction experiment (resulting in the structure
with the PDB code 5MNM) was found to be isomorphous to the
crystal used for the neutron diffraction experiment. The X-ray data
from this non-deuterated crystal was thus used for the joint X-ray/
neutron refinement.
Refinement of thrombin complexes. The data sets of the complex
of 1L and 4L in thrombin were previously collected, refined, and
published by Baum et al.25 The data of the thrombin complex with 5L
has already been described by Ngo et al.9 and Ru
̈
hmann et al.
previously described the complexes of the fragments 1F and 2F in
thrombin.35
Thrombin complex structures with 4-methylbenzylamine, 3L, and
2L were collected at 100 K at the synchrotron BESSY II in Berlin at
beamline 14.1, and 3F at beamline 14.2. All structures were scaled
with XDS.36 The PDB entry 2ZGB9 was used as a starting model for
the conduction of the molecular replacement with Phaser37 from the
CCP4 suite. Complexes with 3F and 3L were refined anisotropically,
whereas the thrombin complex with 4-methylbenzylamine and 2L
were refined isotropically using the program Phenix.refine.38 Five
percent of the data were randomly selected and used for the
calculation of Rfree. TLS parameters were used subsequently to these
isotropic refinements, whereupon the definition of the TLS groups
was performed with Phenix.refine.38 The restraints and coordinates of
the ligands were obtained from Grade Webserver.39 The model
building was done with Coot.40 The final structures were labeled
according to the work of Bode et al.41,42
1622
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Refinement of Trypsin Complexes. The data sets of the complex of
1F, 2L, and 2F in trypsin were previously collected, refined, and
published by Schiebel et al.8 Ngo et al. described the complexes of 5L
in trypsin.9 The data set of the trypsin complex with 1L was collected
at 100 K at the synchrotron Bessy II in Berlin. For the trypsin
complex structure with ligand 3L, the data set was collected in-house
at a MAR345 device at 100 K. Data were processed with XDS.36 The
PDB entry 5MNK was used as a starting model for the conduction of
the molecular replacement with Phaser37 from the CCP4 suite. For
3L, the B factors (Debye−Waller factor) were refined isotropically
with TLS groups using the program Phenix.refine.38 The B-factors of
the 1L complex were refined anisotropically except water. Five
percent of the data were randomly selected and used for the
calculation of Rfree. The restraints and coordinates of the ligands were
obtained from Grade Webserver.39 The model building was done with
Coot.40 The final structures were labeled according to the work of
Bode et al.41,42
Refinement of E192Q·5L. The data set of the E192Q complex with
5L was collected at 110 K at the Synchrotron Elettra in Trieste at the
beamline XRD2 and scaled using XDS.36 The PDB entry 1H8D was
used as a starting model for the conduction of the molecular
replacement with Phaser37 from the CCP4 suite. The structure was
then refined isotropically with group B factors using the program
Phenix.refine.38 Five percent of the data were randomly selected and
used for the calculation of Rfree. The restraints and coordinates of 5L
were obtained from Grade Webserver.39 The model building was
done with Coot.40 The final structure was labeled according to the
work of Bode et al.41,42 The structure of complex 5L in E192Q is
virtually identical to complex 5L in wild-type thrombin. Only a
significantly poorer resolution differentiates the structures; thus, the
complex 5L in E192Q was not additionally deposited.
Khang Ngo − Institut für Pharmazeutische Chemie, Philipps-
Universität Marburg, 35032 Marburg, Germany
Johannes Schiebel − Institut für Pharmazeutische Chemie,
Philipps-Universität Marburg, 35032 Marburg, Germany
Angela Ilse Marca Pizarroso − Institut für Pharmazeutische
Chemie, Philipps-Universität Marburg, 35032 Marburg,
Germany
Linda Schmidt − Institut für Pharmazeutische Chemie,
Philipps-Universität Marburg, 35032 Marburg, Germany
Benjamin Wenzel − Institut für Pharmazeutische Chemie,
Philipps-Universität Marburg, 35032 Marburg, Germany
Torsten Steinmetzer − Institut für Pharmazeutische Chemie,
Philipps-Universität Marburg, 35032 Marburg, Germany;
orcid.org/0000-0001-6523-4754
Andreas Ostermann − Heinz Maier-Leibnitz Zentrum,
Technische Universität München, 85748 Garching, Germany
Andreas Heine − Institut für Pharmazeutische Chemie,
Philipps-Universität Marburg, 35032 Marburg, Germany
Complete contact information is available at:
https://pubs.acs.org/10.1021/acs.jmedchem.0c01809
Author Contributions
A.S., K.N., J.S., and A.I.M.P. performed expression, crystal-
lization, crystal structure analyses, and ITC. L.S., A.S., and
A.I.M.P. expressed protein and mutated variants. B.W., A.S.,
A.I.M.P., L.S., and T.S. performed enzyme kinetic measure-
ments. A.O. helped with the neutron diffraction experiments.
A.H. and G.K. helped interpret the data. A.S., K.N., J.S., and
G.K. wrote the manuscript. All authors have given approval to
the final version of the manuscript.
ASSOCIATED CONTENT
■
sı
* Supporting Information
Funding
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.jmedchem.0c01809.
Short-term fellowship granted by EMBO to J.S. European
Research Council ERC Advance grant to G.K. (no. 268145-
DrugProfilBind).
Images illustrating X-ray and neutron 2mFo-DFc maps,
enzymatic cleavage reaction of 7-amino-4-methylcou-
marin as the assay substrate, chemical formulas and
strings of the ligands used in the comparative kinetic
enzyme assays, exemplary thermograms, crystallographic
tables, thermodynamic data of ligand 3L binding to
thrombin and trypsin, and site-directed mutagenesis,
expression, and purification of thrombin variants (PDF)
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank the Heinz Maier-Leibnitz Zentrum for their
generous allocation of beamtime at the FRM II and for travel
support. In addition, we are grateful to the staff of the EMBL
beamlines at the DESY for their support during the collection
of ultrahigh-resolution and room-temperature X-ray data. We
would like to thank the Bessy II synchrotron for generous
allocation of beamtime and support by the beamline scientists
and the Helmholtz Zentrum Berlin for support of travel costs.
We also thank Pavel Afonine (Lawrence Berkeley National
Labs Berkeley, CA, U.S.A.), Chad Brautigam (University of
Texas Southwestern, TX, U.S.A.), and Andre Mitschler
(CNRS, Illkirch, France), who supported our work with
advice and technical help concerning refinement, ITC
evaluation, and the mounting of large protein crystals in
capillaries. The support of Hans-Dieter Gerber (Univ.
Marburg, Germany) in helping with the qNMR measurements
is gratefully acknowledged. Tobias Schrader (Heinz Maier-
Molecular formula of thrombin variants (CSV)
Accession Codes
Atomic coordinates and experimental details for all new crystal
structures will be released from the PDB upon article
publication: Thrombin complexes: 2L (6T57), 3L (6T53),
3F (6T56), 4-methyl benzylamine (6T55). Thrombin A190S
complexes: 1L (5MM6), 4L (5MLS). Trypsin complexes: 1L
(6ZQ2), 3L (6YDY), 3F (5MNK), 3F (5MNL), 3F (5MNM),
3F (5MO1), 3F (5MOR).
AUTHOR INFORMATION
■
Corresponding Author
Gerhard Klebe − Institut für Pharmazeutische Chemie,
Philipps-Universität Marburg, 35032 Marburg, Germany;
orcid.org/0000-0002-4913-390X; Phone: +49 6421 28
21313; Email: klebe@staff.uni-marburg.de
Leibnitz Zentrum, JCNS, Munchen, Germany) and Christian
̈
Sohn (Univ. Marburg, Germany) helped in data collection and
sample preparation. Moreover, we are thankful for the financial
support of J.S. through a short-term fellowship granted by
EMBO. The presented work was supported by the European
Research Council (ERC) grant to GK (no. 268145-
DrugProfilBind). We kindly acknowledge CSL Behring,
Authors
Anna Sandner − Institut für Pharmazeutische Chemie,
Philipps-Universität Marburg, 35032 Marburg, Germany
1623
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
(12) O’Dell, W. B.; Bodenheimer, A. M.; Meilleur, F. Neutron
protein crystallography: A complementary tool for locating hydrogens
in proteins. Arch. Biochem. Biophys. 2016, 15, 48−60.
Marburg (Germany) for supplying us with generous amounts
of human thrombin from the production of Beriplast.
(13) Hufner-Wulsdorf, T.; Klebe, G. Role of water molecules in
̈
ABBREVIATIONS
■
protein−ligand dissociation and selectivity discrimination: Analysis of
the mechanisms and kinetics of biomolecular solvation using
molecular dynamics. J. Chem. Inf. Model. 2020, 60, 1818−1832.
(14) Ahmed, H. U.; Blakeley, M. P.; Cianci, M.; Cruickshank, D. W.
J.; Hubbard, J. A.; Helliwell, J. R. The determination of protonation
states in proteins. Acta Crystallogr. Sect. D Struct. Biol. 2007, 63, 906−
922.
(15) Hubbard, R. E.; Haider, M. K. Hydrogen bonds in proteins:
R o l e a n d s t r e n g t h . E L S 2 0 1 0 , D O I : 1 0 . 1 0 0 2 /
9780470015902.a0003011.pub2.
(16) Adams, P. D.; Mustyakimov, M.; Afonine, P. V.; Langan, P.
Generalized X-ray and neutron crystallographic analysis: More
accurate and complete structures for biological macromolecules.
Acta Crystallogr. Sect. D Biol. Crystallogr. 2009, 65, 567−573.
(17) Schiebel, J.; Gaspari, R.; Sandner, A.; Ngo, K.; Gerber, H. D.;
Cavalli, A.; Ostermann, A.; Heine, A.; Klebe, G. Charges shift
protonation: Neutron diffraction reveals that aniline and 2-amino-
pyridine become protonated upon binding to trypsin. Angew. Chem. -
Int. Ed. 2017, 56, 4887−4890.
(18) Hall, H. K. Correlation of the base strengths of amines. J. Am.
Chem. Soc. 1957, 79, 5441−5444.
(19) Nguyen, C. N.; Kurtzman Young, T.; Gilson, M. K. Grid
inhomogeneous solvation theory: Hydration structure and thermody-
namics of the miniature receptor cucurbit[7]uril. J. Chem. Phys. 2012,
137, 044101.
(20) Nguyen, C.; Gilson, M. K.; Young, T. Structure and
thermodynamics of molecular hydration via grid inhomogeneous
solvation theory. arXiv:1108.4876.
Ala, alanine; AMC, fluorophoric 7-amido-4-methylcoumarins;
Arg, arginine; Asp, aspartate; B factors, Debye−Waller factor;
DMSO, dimethyl sulfoxide; ERC, European Research Council;
GIST, grid inhomogeneous solvation theory; Gln, glutamine;
Glu, glutamate; Gly, glycine; H bonds, hydrogen bonds;
HEPES, 4-(2-hydroxyethyl)-1-piperazine ethane sulfonic acid;
His, histidine; HPLC, high-performance liquid chromatog-
raphy; Ile, isoleucine; ITC, isothermal titration calorimetry; Ki,
kinetic inhibition constant; koff, dissociation rate; kon,
association rate; Lys, lysine; NMR, nuclear magnetic resonance
spectroscopy; PDB, Protein Data Bank; PEG, polyethylene
glycol; Phe, phenylalanine; Pro, proline; rmsd, root-mean-
square deviation; Ser, serine; TFA, trifluoroacetic acid; Thr,
threonine; TLC, thin-layer chromatography; tPA, tissue
plasminogen activator; Tricine, N-(tri(hydroxymethyl)-
methyl)glycine; Tris, Tris(hydroxymethyl)aminomethane;
Trp, tryptophane; Tyr, tyrosine; Val, valine; W, water
REFERENCES
■
(1) Schmidt, A.; Jelsch, C.; Østergaard, P.; Rypniewski, W.; Lamzin,
V. S. Trypsin Revisited: Crystallography at (SUB) atomic resolution
and quantum chemistry revealing details of catalysis. J. Biol. Chem.
2003, 278, 43357−43362.
(2) Sanderson, P. E. J. Small, noncovalent serine protease inhibitors.
Med. Res. Rev. 1999, 19, 179−197.
(3) Straub, A.; Roehrig, S.; Hillisch, A. Oral, direct thrombin and
factor Xa inhibitors: the replacement for warfarin, leeches, and pig
intestines? Angew Chem Int Ed Engl. 2011, 50, 4574−4590.
(4) Metheny, N. A.; Stewart, B. J.; Smith, L.; Yan, H.; Diebold, M.;
Clouse, R. E. pH and concentrations of pepsin and trypsin in feeding
tube aspirates as predictors of tube placement. J. Parenter. Enter. Nutr.
2016, 21, 279−285.
(21) Ramsey, S.; Nguyen, C.; Salomon-Ferrer, R.; Walker, R. C.;
Gilson, M. K.; Kurtzman, T. Solvation thermodynamic mapping of
molecular surfaces in AmberTools: GIST. J. Comput. Chem. 2016, 37,
2029−2037.
(22) Baum, B.; Muley, L.; Heine, A.; Smolinski, M.; Hangauer, D.;
Klebe, G. Think twice: Understanding the high potency of
bis(phenyl)methane inhibitors of thrombin. J. Mol. Biol. 2009, 391,
552−564.
(5) Gustafsson, D.; Antonsson, T.; Bylund, R.; Eriksson, U.;
Gyzander, E.; Nilsson, I.; Elg, M.; Mattsson, C.; Deinum, J.; Pehrsson,
(23) Czodrowski, P.; Sotriffer, C. A.; Klebe, G. Protonation changes
upon ligand binding to trypsin and thrombin: Structural interpreta-
tion based on pKa calculations and ITC experiments. J. Mol. Biol.
2007, 367, 1347−1356.
̈
S.; Karlsson, O.; Nilsson, A.; Sorensen, H. Effects of melagatran, a
new low-molecular-weight thrombin inhibitor, on thrombin and
fibrinolytic enzymes. Thromb. Haemost. 1998, 79, 110−118.
(6) Katz, B. A.; Sprengeler, P. A.; Luong, C.; Verner, E.; Elrod, K.;
Kirtley, M.; Janc, J.; Spencer, J. R.; Breitenbucher, J. G.; Hui, H.;
McGee, D.; Allen, D.; Martelli, A.; Mackman, R. L. Engineering
inhibitors highly selective for the S1 sites of Ser190 trypsin-like serine
protease drug targets. Chem. Biol. 2001, 8, 1107−1121.
(24) Dullweber, F.; Stubbs, M. T.; Musil, D.; Sturzebecher, J.; Klebe,
̈
G. Factorising ligand affinity: A Combined thermodynamic and
crystallographic study of trypsin and thrombin inhibition. J. Mol. Biol.
2001, 313, 593−614.
(25) Baum, B.; Mohamed, M.; Zayed, M.; Gerlach, C.; Heine, A.;
Hangauer, D.; Klebe, G. More than a simple lipophilic contact: A
detailed thermodynamic analysis of nonbasic residues in the S1
pocket of thrombin. J. Mol. Biol. 2009, 390, 56−69.
(26) Kim, S.; Hwang, S. Y.; Kim, Y. K.; Yun, M.; Oh, Y. S. Rational
design of selective thrombin inhibitors. Bioorg. Med. Chem. Lett. 1997,
7, 769−774.
(7) Hilpert, K.; Ackermann, J.; Banner, D. W.; Gast, A.; Gubernator,
K.; Hadvary, P.; Labler, L.; Mueller, K.; Schmid, G.; Tschopp, T. B.;
van de Waterbeemd, H. Design and synthesis of potent and highly
selective thrombin inhibitors. J. Med. Chem. 1994, 37, 3889−3901.
(8) Schiebel, J.; Gaspari, R.; Wulsdorf, T.; Ngo, K.; Sohn, C.;
Schrader, T. E.; Cavalli, A.; Ostermann, A.; Heine, A.; Klebe, G.
Intriguing role of water in protein-ligand binding studied by neutron
crystallography on trypsin complexes. Nat. Commun. 2018, 9, 3559.
(9) Ngo, K.; Collins-Kautz, C.; Gerstenecker, S.; Wagner, B.; Heine,
A.; Klebe, G. Protein-induced change in ligand protonation during
trypsin and thrombin binding: Hint on differences in selectivity
determinants of both proteins? J. Med. Chem. 2020, 63, 3274−3289.
(27) Pineda, A. O.; Zhang, E.; Guinto, E. R.; Savvides, S. N.;
Tulinsky, A.; Di Cera, E. Crystal structure of the thrombin mutant
D221A/D222K: the Asp222:Arg187 ion-pair stabilizes the fast form.
Biophys. Chem. 2004, 112, 253−256.
(28) Wells, C. M.; Di Cera, E. Thrombin is a sodium ion activated
enzyme. Biochemistry 1992, 31, 11721−11730.
(10) Castaneda, C. A.; Fitch, C. A.; Majumdar, A.; Khangulov, V.;
̃
(29) Gottlieb, H. E.; Kotlyar, V.; Nudelman, A. NMR chemical shifts
of common laboratory solvents as trace impurities. J. Org. Chem. 1997,
62, 7512−7515.
Schlessman, J. L.; García-Moreno, B. E. Molecular determinants of the
pKa values of Asp and Glu residues in staphylococcal nuclease.
Proteins Struct. Funct. Bioinforma. 2009, 77, 570−588.
(30) Sturzebecher, J.; Sturzebecher, U.; Vieweg, H.; Wagner, G.;
̈ ̈
(11) Ruhmann, E.; Betz, M.; Heine, A.; Klebe, G. Fragment binding
̈
Hauptmann, J.; Markwardt, F. Synthetic inhibitors of bovine factor Xa
and thrombin comparison of their anticoagulant efficiency. Thromb.
Res. 1989, 54, 245−252.
can be either More enthalpy-driven or entropy-driven: Crystal
structures and residual hydration patterns suggest why. J. Med.
Chem. 2015, 58, 6960−6971.
1624
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
(31) Dixon, M. The graphical determination of Km and Ki. Biochem.
J. 1972, 129, 197−202.
̈
(32) Bonacker, L.; Schroder, M. CSL Behring, https://www.
beriplast.de/ , (accessed Dec 3, 2020).
(33) Otwinowski, Z., Minor, W. Processing of X-ray Diffraction Data
Collected in Oscillation Mode, Methods in Enzymology, Volume 276:
Macromolecular Crystallography, part A, pp 307−326, 1997. Carter,
C. W., Jr., Sweet, R. M., Eds., Academic Press: New York.
(34) AG Bachem Holding. Bachem. https://shop.bachem.com/
4014553.html , (accessed Dec 3, 2020).
̈
(35) Ruhmann, E. Structural and thermodynamic characterization of
fragment and ligand binding to thrombin. PhD thesis univ. of Marburg,
Germany, 2015.
(36) Kabsch, W. Integration, scaling, space-group assignment and
post-refinement. Acta Crystallogr. Sect. D 2010, 66, 133−144.
(37) McCoy, A. J.; Grosse-Kunstleve, R. W.; Adams, P. D.; Winn, M.
D.; Storoni, L. C.; Read, R. J. Phaser crystallographic software. J. Appl.
Crystallogr. 2007, 40, 658−674.
́
(38) Adams, P. D.; Afonine, P. V.; Bunkoczi, G.; Chen, V. B.; Davis,
I. W.; Echols, N.; Headd, J. J.; Hung, L. W.; Kapral, G. J.; Grosse-
Kunstleve, R. W.; McCoy, A. J.; Moriarty, N. W.; Oeffner, R.; Read, R.
J.; Richardson, D. C.; Richardson, J. S.; Terwilliger, T. C.; Zwart, P.
H. PHENIX: a comprehensive Python-based system for macro-
molecular structure solution. Acta Crystallogr. Sect. D 2010, 66, 213−
221.
(39) Smart, O.; Womack, T. Grade Web Server. Global Phasing
http://grade.globalphasing.org, (accessed Dec 3, 2020).
(40) Emsley, P.; Cowtan, K. Coot: model-building tools for
molecular graphics. Acta Crystallogr. Sect. D 2004, 60, 2126−2132.
(41) Bode, W.; Mayr, I.; Huber, R.; Stone, S. R.; Hofsteenge, J. A-
Thrombin: Interaction with D-Phe-Pro-Arg chloromethylketone and
significance of the Tyr-Pro-Pro-Trp insertion segment. EMBO J.
1989, 8, 3467−3475.
(42) Cera, D.; Gruber, A.; Us, O. R.; Gandhi, P.; Louis, S.; Us, M.
O.; Pelc, L.; Us, I. L.; Louis, S.; Us, M. O.; Wood, C.; Us, M. O.;
Saint, A.; Louis, S. Expression of thrombin variants, United States
Patent, US 0,089,4029,7 B2, Jan 27, 2015.
1625
https://dx.doi.org/10.1021/acs.jmedchem.0c01809
J. Med. Chem. 2021, 64, 1611−1625