Rational Design of Thermodynamic and Kinetic Binding Profiles by Optimizing Surface Water Networks Coating Protein-Bound Ligands

Article abstract of DOI:10.1021/acs.jmedchem.6b00998

A previously studied congeneric series of thermolysin inhibitors addressing the solvent-accessible S₂′ pocket with different hydrophobic substituents showed modulations of the surface water layers coating the protein-bound inhibitors. Increasing stabilization of water molecules resulted in an enthalpically more favorable binding signature, overall enhancing affinity. Based on this observation, we optimized the series by designing tailored P₂′ substituents to improve and further stabilize the surface water network. MD simulations were applied to predict the putative water pattern around the bound ligands. Subsequently, the inhibitors were synthesized and characterized by high-resolution crystallography, microcalorimetry, and surface plasmon resonance. One of the designed inhibitors established the most pronounced water network of all inhibitors tested so far, composed of several fused water polygons, and showed 50-fold affinity enhancement with respect to the original methylated parent ligand. Notably, the inhibitor forming the most perfect water network also showed significantly prolonged residence time compared to the other tested inhibitors.

Full text of DOI:10.1021/acs.jmedchem.6b00998

Subscriber access provided by United Arab Emirates University | Libraries Deanship
Article
Rational Design of Thermodynamic and Kinetic Binding Profiles by
Optimizing Surface Water Networks Coating Protein Bound Ligands
Stefan G Krimmer, Jonathan Cramer, Michael Betz, Veronica
Fridh, Robert Karlsson, Andreas Heine, and Gerhard Klebe
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.6b00998 • Publication Date (Web): 10 Nov 2016
Downloaded from http://pubs.acs.org on November 10, 2016
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 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Rational Design of Thermodynamic and Kinetic
Binding Profiles by Optimizing Surface Water
Networks Coating Protein Bound Ligands
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
†
,
†,
†
‡
‡
∥
∥
Stefan G. Krimmer, Jonathan Cramer, Michael Betz, Veronica Fridh, Robert Karlsson,
†
†,
Andreas Heine, and Gerhard Klebe *
†
Institute of Pharmaceutical Chemistry, University of Marburg, Marbacher Weg 6, 35032 Marburg,
Germany
‡
GE Healthcare BioꢀSciences AB, SEꢀ751 84 Uppsala, Sweden
KEYWORDS: protein–ligand interactions, thermodynamic optimization, hydrophobic effect,
solvent reorganization, water network prediction, molecular dynamics simulations, Xꢀray
crystallography, isothermal titration calorimetry, surface plasmon resonance, binding kinetics
ABSTRACT: A previously studied congeneric series of thermolysin inhibitors addressing the
solventꢀaccessible S ’ pocket with different hydrophobic substituents showed modulations of the
2
surface water layers coating the proteinꢀbound inhibitors. Increasing stabilization of water
molecules resulted in enthalpically more favorable binding signature, overall enhancing affinity.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Based on this observation, we optimized the series by designing tailored P ’ substituents to improve
2
and further stabilize the surface water network. MD simulations were applied to predict the putative
water pattern around the bound ligands. Subsequently, the inhibitors were synthesized and
characterized by highꢀresolution crystallography, microcalorimetry and surface plasmon resonance.
One of the designed inhibitors established the most pronounced water network of all inhibitors
tested so far, composed of several fused water polygons, and showed 50ꢀfold affinity enhancement
with respect to the original methylated parent ligand. Notably, the inhibitor forming the most
perfect water network also showed significantly prolonged residence time compared to the other
tested inhibitors.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
INTRODUCTION
Structureꢀbased drug design (SBDD) seeks to optimize ligand binding with respect to a given
target protein. Thermodynamic parameters such as ∆G, ∆H, –T∆S, ∆C and of binding kinetic
p
properties, such as kinetic association (k ) and dissociation (k ) rate constants, are considered to
on
off
rationalize and accelerate affinity optimization by a better characterization of the protein–ligand
1–7
binding process. Thermodynamic profiling, at its best based on isothermal titration calorimetry
8
–10
(
ITC),
is supposed to provide insights into the molecular interactions determining the affinity of
a ligand to its target. However, past experience has shown that optimization and prioritization of
compounds guided by thermodynamics is difficult, since the enthalpy–entropy profile reflects the
binding event as a whole. Hence, the high complexity of this event largely impedes factorization
into individual contributions to binding and often only succeeds for congeneric ligand series with
11–13
minor structural variations.
The correlation of structural properties with binding kinetic data is
presently poorly understood and reasonable correlations have only been established for a limited
ACS Paragon Plus Environment

Page 3 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
4
number of cases. Furthermore, we still face an incomplete comprehension of the fundamental
14–16
relationships between thermodynamics, kinetics and molecular interactions,
which can lead to
false predictions made under overly simplified assumptions. Especially, the impact of the versatile
and ubiquitously present water molecules is hardly understood. The involvement of water
molecules is a major cause for the inherent complexity, particularly arising from the ability of water
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
1
7
to establish Hꢀbonds and the resultant tendency to arrange in differently ordered structures. Water
molecules can actively mediate Hꢀbonds across the binding interface between protein and ligand
1
8
and thereby improve affinity. The displacement of wellꢀordered water molecules from apolar
surfaces is discussed as the driving force of the hydrophobic effect (and not the formation of
hydrophobic interactions), a process of utmost importance for molecular recognition and drug
1
1,19–23
action. According to the soꢀcalled “classical” hydrophobic effect,
the binding of an apolar
ligand portion to a hydrophobic protein cavity correlates with an entropic advantage due to the
displacement of wellꢀordered water molecules into the bulk water phase. In contrast, a “nonꢀ
classical” hydrophobic effect has been defined, which is enthalpy driven. It has been related to a
2
4–26
suboptimal hydration of a protein cavity prior to ligand binding.
In this case, the enthalpy gain
upon binding results from the displacement of orientationally mobile and thus entropically favored
water molecules into the bulk phase where they can establish better Hꢀbonds than previously
observed in the protein cavity.
Despite the popular binary classification of “classical” and “nonꢀclassical”, the hydrophobic
effect can range from entirely entropyꢀdriven to entirely enthalpyꢀdriven. This is determined by the
thermodynamic properties of the water molecules that are involved before and after binding and by
the ligand and the binding site (especially by their molecular shape and polarity), which affect the
2
7
formation of adjacent water networks. Of utmost importance for the thermodynamic signature is
the overall inventory of water molecules with respect to their release into or recruitment from the
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
bulk water phase, and the embedding of water molecules into Hꢀbonding networks of varying
completeness and perfection in the proteinꢀbound state. In particular, the way water molecules are
able to rearrange and establish Hꢀbonding networks around the newly formed solventꢀexposed
surface of the protein–ligand complex seems to have a significant impact on the thermodynamic
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
28–32
binding signature.
32
In a previous study, we structurally and thermodynamically analyzed a series of congeneric
thermolysin (TLN) inhibitors with a peptidomimetic parent scaffold (Figure 1A) decorated with
different hydrophobic P ’ substituents (Figure 1B). We selected the zinc metalloprotease TLN from
2
33–35
Bacillus thermoproteolyticus for our studies,
as this enzyme has been frequently used as class
3
6,37
representative and exhibits excellent crystallographic properties.
In addition, TLN is quite rigid
because of its high thermal stability, thus reducing structural adaptions and facilitating comparative
analyses. The active site of TLN is composed of three subpockets (Figure 1A). Firstly, the rather
unspecific S pocket, a hydrophobic cavity that recognizes aromatic portions such as Phe. Secondly,
1
the S ’ specificity pocket, a predominantly hydrophobic, deep and narrow pocket, which
1
preferentially accommodates hydrophobic amino acids such as Val, Leu, Ile and Phe, and thirdly
the S ’ subsite, a hydrophobic, flat, bowlꢀshaped pocket, which is wellꢀaccessible to bulk water
2
34
32
molecules. We selected the S ’ pocket for our studies. Within a previously investigated series,
2
we increased the size of the P ’ substituents addressing the wellꢀsolvated S ’ pocket systematically
2
2
from a sole methyl to a phenylethyl substituent as displayed in Figure 1B. By detailed ITC analyses
Figure 1B), we revealed a difference in binding affinity (K : dissociation constant of the
(
d
3
2
thermodynamic equilibrium) of more than one order of magnitude, or expressed as standard Gibbs
–1
free energy, a maximum ∆∆G° of 7.0±0.4 kJ mol . Remarkably, ∆∆G° factored in a huge
enthalpy–entropy variation, and indicated pronounced enthalpy–entropy compensation, a
2
0,38–40
phenomenon frequently observed in drug optimization.
We could correlate the observed
ACS Paragon Plus Environment

Page 5 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
variations in the thermodynamic profiles with crystallographically observed structural changes of
the P ’ substituents and, triggered thereby, modulations of the adjacent surface water solvation
2
3
2
layer. The ∆H° contributions appeared to be more favorable (and –T∆S° less favorable) where a
betterꢀordered water network was established next to the surface of the newly formed protein–
ligand complex. In contrast, –T∆S° apparently increased (and ∆H° decreased) where the first
solvation layer next to the bound ligand was unfavorably disrupted. Furthermore, the increasing
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
contribution to desolvate the gradually growing P ’ substituents seemed to enhance binding entropy,
2
1
9
as expected for the “classical” hydrophobic effect of a wellꢀhydrated, apolar cavity. Strikingly, the
inhibitor with the highest affinity (ligand 1, Figure 1B) showed both, a pronounced burial of its
relatively large P ’ group along with a wellꢀestablished surface water network wrapping around this
2
3
2
substituent in its proteinꢀbound state. This resulted in a wellꢀbalanced thermodynamic profile
driven by favorable enthalpic and entropic contributions, overall resulting in an increase in binding
affinity. Consequently, optimization of the surface water network wrapping around the partly
solventꢀexposed P ’ substituent appears an useful approach to enhance ligand binding, since the
2
enthalpic gains seem to overall improve affinity.
In the present study, we want to validate this working hypothesis by systematically improving of
the surface water network around a newly formed protein–ligand complex to modulate its
thermodynamic binding profile and thus increase affinity of a bound ligand. Starting with the best
and already fairly wellꢀoptimized ligand 1 of our previous series (Figure 1B), we designed five
additional ligands (2–6, Figure 1C) based on the carboxybenzylꢀGlyꢀ(PO )ꢀLꢀLeuꢀNH ꢀP ’ parent
2 2
2
scaffold (Figure 1A), and attached distinct apolar P ’ substituents to generate differently shaped
2
solventꢀexposed surfaces in complex with TLN. Prior to ligand synthesis, we used molecular
dynamics (MD) simulations to predict the quality and completeness of the surface water network
4
1
established around the newly formed complex. The MD simulations suggested the highest
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
completeness and quality for the complex with 3 and lowest for its epimer 6. As the designed
ligands seem to be promising candidates to validate our hypothesis, we synthesized all five to study
the established water networks around the formed complexes by Xꢀray crystallography and
thermodynamically by ITC. Furthermore, as we also expected an impact on the binding kinetic
properties, we studied the association and dissociation rate constants k and k by surface plasmon
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
on
off
resonance (SPR).
RESULTS
Ligand design and solvation pattern prediction using MD simulations. Our design hypothesis
to maximize the desolvation of increasingly bulky P ’ substituents along with an energetically
2
favorable (“optimal”) surface water network to enhance binding affinity started with the binding
3
2
pose of 1, which was already characterized in a previous study (PDB code 4MZN). This ligand
showed a wellꢀestablished surface water network toward the left rim of the binding pocket
including an energetically favorable fiveꢀmembered water polygon, interconnected by Hꢀbonds
(Figure 2A, left panel). Deficiencies in the network are suggested on the lower and rightꢀhand side
of the S ’ pocket (direction relative to the view of the figure). Here, the tertꢀbutyl portion of the
2
ligand with the 2,2ꢀdimethylbutanyl P ’ group (Figure 2A, right panel), exhibiting two additional
2
terminal methyl groups relative to 1, stabilized a more complete network in this region, and also
established the favorable pentagonal polygon to the left. However, its water network is unfavorably
disrupted on top of the tertꢀbutyl portion, resulting in an entropically highly favored system with an
overall lowered affinity (Figure 1B). We therefore envisioned merging the features of both P2’
substituents into sizeꢀincreased 2 (comprising only one of the two additional methyl groups of the
tertꢀbutyl portion) and 3 (exhibiting both additional methyl groups of the tertꢀbutyl portion). As our
ACS Paragon Plus Environment

Page 7 of 88
Journal of Medicinal Chemistry
1
2
3
4
design involved the creation of stereogenic centers, we also considered the epimers 5 and 6 of 2 and
3
in our subsequent MD evaluation.
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
To predict and analyze the pattern of water solvation sites around the designed P ’ substituents,
2
4
1
we applied our recently introduced MD approach to simulate TLNꢀ1, TLNꢀ2, TLNꢀ3, TLNꢀ5 and
TLNꢀ6. Ligand 4 (the epimer of 1, Figure 1C) was not considered for the MD simulations, as this
ligand was only synthesized at a late stage of the study with the purpose to complete the congeneric
series and to further validate the influence of the rearrangement of water molecules on the
thermodynamics of protein–ligand binding. This strategy provided the opportunity to validate our
MD protocol on TLNꢀ1, as we had already determined a highꢀresolution crystal structure (1.17 Å)
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
2
for this complex. MD simulations were run for 20 ns and the spatial positions of water molecules
were recorded every 2 ps along the trajectory, from which the propensity of water molecules to
occupy the indicated solvation sites was calculated. For the novel complexes, the crystal structure
of TLNꢀ1 was used as a template. During the MD simulations, nonꢀhydrogen atoms and nonꢀwater
molecules were constrained to their starting coordinates. Similar protocols were applied to model
the designed complexes with 2, 3, 5 and 6 (for further details, see the Experimental Section). For
TLNꢀ1, the computed results are superimposed with the difference electron densities of the water
molecules found in the crystal structure (Figure 2B). The displayed solvation sites encompass a
probability greater than 48% to record a water molecule along the trajectory at this site. This
contour level was adjusted by visual inspection of the computed map to qualitatively match with the
contouring of the crystallographically determined F –F difference electron density at the
o
c
commonly applied 3σ level. The results matched convincingly well. Only the site W8 is predicted as
less populated compared to the crystal structure and W9 was suggested as being slightly displaced
by the MD approach. Mutually facing the distributions of the computed solvation sites indicates that
TLNꢀ3 displays the most densely packed and complete surface water network in the series (Figure
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
2
C). The simulation of TLNꢀ5 also assigned a rather elaborate surface water network to this
complex, whereas TLNꢀ2 and in particular TLNꢀ6 show a gap in the water network capping their
P2’ substituents.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Stereoselective synthesis of the congeneric phosphonamidate inhibitors 2–6. Stimulated by
the simulation results, we decided to synthesize 2–6. Compound 4, the epimer of 1, was included
for reasons of comparison. Scheme 1 illustrates representatively the synthesis route to prepare 2–6.
The stereogenic center in the P ’ portion of the ligands was synthetically accessible by a strategy
2
employing a chiral oxazolidinone auxiliary. 4ꢀBenzyl oxazolidinone (7 or 8) was treated with nꢀ
BuLi followed by the respective acid chlorides. In a diastereoselective enolate alkylation of the
resulting Nꢀacyloxazolidinones (9–13), intermediates 14–18 were synthesized in a diastereomeric
ratio of 13:1–16:1. With the exception of 16, the concentration of the desired diastereomers could
be improved to a ratio of >20:1 by recrystallization from cyclohexane. The auxiliary was removed
by hydrolysis with LiOH/H O . The chiral carboxylic acids 19–23 were subsequently reduced to the
2
2
corresponding alcohols (24–28) with LiAlH . The conversion to the peptidic intermediates 34–38
4
was carried out by a multistep procedure involving a Mitsunobu reaction with DPPA followed by
an in situ Staudinger reduction of the resulting azides. Due to their highly volatile nature, the
intermediate amines 29–33 were not isolated. After an aqueous workup procedure, they were
reacted with BocꢀLeuꢀOH under standard EDC coupling conditions.
4
2
Phosphonic acid monoester 39 was synthesized following a modified literature procedure. For
the phosphonamidate coupling reaction, 39 was activated with SOCl . The peptidic intermediates
2
3
4–38 were deprotected using a 4 M solution of HCl in dioxane and reacted with the activated
ACS Paragon Plus Environment

Page 9 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
phosphonic acid. The final deprotection of the coupling products 40–44 with aqueous LiOH
solution followed by semiꢀpreparative HPLC purification afforded inhibitors 2–6 in high purity.
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
Structure determination of TLN–2 to TLN–6 by Xꢀray crystallography. Crystal structures of
2
–6 (chemical structures in Figure 1C) in complex with TLN were collected at high resolutions
between 1.16–1.38 Å (Table 1). As shown in Figure 3, the conformations of the amino acids of their
TLN binding sites are highly conserved and superimpose perfectly well, likewise the binding mode
of the parent scaffold of all six ligands is virtually identical. The binding mode of this conserved
3
0,43,44
part was already described extensively.
In brief, the scaffold coordinates monodentally the
zinc ion with its negatively charged phosphonamidate oxygen. The carbamate group is disordered
over two conformations with approximately equal occupancy. The isoꢀbutyl portion of the ligand’s
leucine component is buried in the hydrophobic S ’ specificity cavity, a binding motif crucial for
1
44
achieving high ligand affinity. The S pocket is occupied by a glycerol molecule from the
1
cryobuffer, on top of which the carboxybenzyl moiety of the ligand is positioned. Also picked up
from the buffer, a DMSO molecule is binding adjacent to the carbamate group of the ligand. Thus,
1
–6 differ solely in their P ’ substituents and in the water networks adjacent to them. In the
2
following, the positions of fifteen water molecules W1–W15 (referring to water molecules found at
distinct positions in the first solvation layer around the P ’ groups) are described according to the
2
nomenclature and relative to the view angle chosen in Figure 4. At the right upper rim of the S2’
pocket, a second glycerol molecule is found (Figure 4A–F), which establishes weak Hꢀbonds
(
distance >3.2 Å) to water molecules W10 and W11 in some of the crystal structures. We observed
the two glycerol molecules in all 19 crystal structures that we determined in the previous
29,30,32,45
studies.
A glycerol molecule is well known to replace three water molecules as a kind of
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
rigidified surrogate in a crystal structure. Nonetheless, to validate and exclude whether these
glycerol molecules take any artificial influence on the ligand pose and adjacent water structure, we
succeeded to establish alternative crystallization conditions also yielding wellꢀdiffracting crystals
using PEG400 and methylꢀ2,4ꢀpentanediol (MPD) instead of glycerol as cryoprotectant. We also
collected diffraction data of glycerolꢀfree crystals of TLN in complex with 3, 5 and 6 in order to
validate whether the glycerol molecule exerts any artificial influence on the structural arrangement
of the observed water molecules (Figure S2, Supporting Information). In summary, in the crystal
structures with PEG400 and MPD as cryoprotectant, the positions of the glycerol OH groups are
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
occupied by water molecules, and the established water networks in the S ’ pocket are very similar
2
to the below described complexes with glycerol as cryoprotectant. The minor differences concern
only highly mobile water molecules (borderline cases with respect to the placement of water
molecules in the refinement model) and the differences observed between the crystals exposed to
different cryoprotectants fall into the same range as deviations recognized if diffraction data
collected for different crystals of similar protein structures is compared. In the following, we
compare the crystal structures in terms of observed electron densities in the S ’ pockets (F –F omit
2
o
c
electron densities in Figure 4A–F) and refined B factors of the water molecules (B factors of all
water molecules from the first solvation layer of TLNꢀ1 to TLNꢀ6 displayed as a heatmap in Figure
5
) and we avoid to discuss only the presence or absence of a water molecule in the refinement
model.
Arrangement of water molecules in the S ’ pocket of TLNꢀ1, TLNꢀ2 and TLNꢀ3. The
2
3
2
structure of TLNꢀ1 was published previously and served as starting point for the design of our
new congeneric ligand series. Hence, the water networks observed in TLNꢀ2 to TLNꢀ6 are
ACS Paragon Plus Environment

Page 11 of 88
Journal of Medicinal Chemistry
1
2
modifications of that in TLNꢀ1. In TLNꢀ1, the ligand places its 2(S)ꢀmethylbutyl P ’ group into the
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
S2’ pocket (Figure 4A). The right cleft of the pocket is addressed by the terminal methyl group of
the P ’ portion, whereas the terminal ethyl group is oriented toward the left. In TLNꢀ1, W5–W9
2
form a fiveꢀmembered polygon with Hꢀbond distances between 2.8–3.1 Å exhibiting low B factors
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(Figure 5). In total, twelve water molecules are detected in the first solvation layer around the P2’
substituent covering its large, solventꢀexposed hydrophobic surface patch (Figure 6). Several of the
water molecules are anchored to the protein surface via polar interactions: W1 (Arg203 primary
nitrogen, 3.0 Å; His231 carbonyl oxygen, 2.9 Å), W2 (ligand 1 carboxybenzyl carbonyl O, 2.9 Å;
Asn112 amide nitrogen, 3.0 Å), W6 (Asp226 carboxy oxygen, 2.8 Å), W11 (Asn111 backbone
carbonyl oxygen, 2.8 Å) and W13 (Tyr193, 3.1 Å). In contrast, W3, W5, W7, W8, W9, W10 and W12
are only stabilized by Hꢀbonds to other water molecules or by van der Waals interactions with the
apolar surface patch of the P ’ substituent of 1.
2
In TLNꢀ2, the 2(S),3ꢀdimethylbutyl group of 2 orients its P ’ group similarly to 1, and the
2
additional methyl group is disordered over two positions (Figure 4B): conformation A (56%
occupancy) is sticking out into the solvent (no contacts within hydrophobic interaction distance of
≤
4.6 Å), whereas conformation B (44% occupancy) is oriented downward, alongside the protein
surface. The observed disorder of the P ’ substituent is a result of the shallow, widely open S ’
2
2
pocket. Due to the steric requirement of the methyl group in conformation A, two distinct, mutually
excluding sites are observed for W3 (occupancies constrained to 50/50 in the refinement model). As
a consequence of the reduced occupancy and owing to a strong correlation of B factors with
4
6
occupancy, the refined B factor for this water molecule (Figure 5) has to be regarded with care
and will hardly reflect its actual mobility. Furthermore, the distance between W9 and W10 increases
from 3.4 Å in TLNꢀ1 to 4.5 Å in TLNꢀ2, clearly exceeding the maximum distance for an
4
7,48
energetically favorable Hꢀbond.
The constraint which modifies the water structure between
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
TLNꢀ1 and TLNꢀ2 is best visualized by use of the solvent accessible surface areas (SASAs; solvent
excluded surface area plus radius of a water molecule) as displayed for the P ’ portions in Figure 7:
2
W3, W9, W10, and W12 in TLNꢀ1 would penetrate inside the SASA of TLNꢀ2, thus these water
molecules in TLNꢀ2 must be shifted. Because of the expanded water structure, W12 and W13
become increasingly destabilized (Figure 5).
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
In TLNꢀ3, the ligand exhibits a 2(S)ꢀ3,3ꢀtrimethylbutyl substituent (Figure 4C), representing the
bulkiest P ’ portion in the (S)-configurated series. The binding mode of 3 closely resembles that of
2
2
if both conformations A and B of TLNꢀ2 would be merged. Two additional sites for water
molecules (W14 and W15) are refined, resulting in the formation of a sixꢀmembered polygon and
wellꢀdefined electron density (Figure 4C) with strongly decreased B factor (Figure 5) of the
incorporated W13. Furthermore, water molecules W8 and W9, both participating in the five and sixꢀ
membered polygonal water networks in TLNꢀ3, are here better stabilized than in TLNꢀ2 (Figure 5).
In TLNꢀ3, W10 is shifted distal from the P ’ substituent and is indicated by less electron density
2
along with a higher B factor compared to TLNꢀ2.
Arrangement of water molecules in the S ’ pocket of TLNꢀ4, TLNꢀ5, and TLNꢀ6. Ligands 4,
2
5
and 6 are the epimers of 1, 2 and 3. The overall quality of the crystal structure TLNꢀ4 is slightly
lower compared to that of the other five complexes (resolution, Rwork/Rfree values, Wilson B factor;
see Table 1). Nevertheless, only about ten water molecules less are observed in TLNꢀ4 at a total
amount of >400. In TLNꢀ4 (Figure 4D), the P ’ portion is flipped over by 180° compared to TLNꢀ1:
2
The terminal P ’ ethyl group is oriented toward the right rim, whereas the P ’ methyl group is
2
2
directed to the left of the S ’ pocket. Only six water molecules are detected in the crystal structure
2
adjacent to the P ’ group mainly stabilized via Hꢀbonds to protein residues.
2
ACS Paragon Plus Environment

Page 13 of 88
Journal of Medicinal Chemistry
1
2
3
4
In TLNꢀ5 (Figure 4E), in contrast to 4, the P ’ substituent of 5 adopts a conformation analogous
2
to that of the (S)-configurated P ’ substituents of 1 to 3. One methyl group from the terminal isoꢀ
2
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
propyl portion of 5 is not defined in the electron density most likely owing to enhanced mobility. It
was therefore not modelled in the structure. The absence of this methyl group is also observed in
TLNꢀ5 with MPD as cryoprotectant (Figure S3B, Supporting Information). The missing of water
molecules W5, W10 and W12 results in an incomplete water network in TLNꢀ5 (compared to that of
the epimeric TLNꢀbound 2). In the crystal structure TLNꢀ5 with MPD as cryoprotectant, W5 is
highly mobile but sufficient electron density is detected to allow placement of a water molecule in
the refinement model (Figure S2B, Supporting Information). Furthermore, in the structure of TLNꢀ5
with glycerol as cryoprotectant, some F –F difference electron density is observed at the positions
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
o
c
of W5 and W10 (Figure 4E), which, however, is too weak to justify placement of a fully populated
water molecule in the refinement model of TLNꢀ5. In TLNꢀ6 (Figure 4F), W5 and W10 were added
to the refinement model, but they refined as highly mobile (Figure 5). Furthermore, W10 is missing
in the crystal structure of TLNꢀ6 with MPD as cryoprotectant (Figure S2C, Supporting
Information). Consequently, the water molecules W5 and W10 are highly mobile in TLNꢀ5 and
TLNꢀ6, and the local concentration of their electron density is at the borderline for water placement
in the refinement model. Thus, the water networks of TLNꢀ5 and TLNꢀ6 are highly similar. Overall,
in the (R)ꢀseries a lower amount of water molecules with increased residual mobility (especially of
W5–W10, Figure 5) is observed compared to the (S)ꢀseries.
Thermodynamic signatures of TLNꢀligand complex formation measured by ITC. As we
recently documented, ITC measurements comparable on the same scale and with minimal error
margins can only be obtained if all ligands are studied with the same optimized measurement
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
4
5
protocol using the same protein batch. We therefore reevaluated 1 along with 2–6 in the present
study (for further details see Chapter 4, Supporting Information). Across the (S)ꢀconfigurated series
(
1→2→3), the binding affinity ∆G° improved with growing number of methyl groups (Figure 8).
Remarkably, this effect is determined by an increasingly favorable –T∆S°, which is only partly
compensated by a loss in ∆H° (slope of –T∆S° is steeper than of ∆H°), leaving overall a gain in
∆G°. Interestingly, for the (R)ꢀconfigurated series (4→5→6), no affinity enhancement is detected.
The mutual compensation of ∆H° and –T∆S° fully nullifies any affinity improvement as the ∆H°
compensation is stronger (slopes of –T∆S° and ∆H° are equal with opposite sign) compared to the
(S)ꢀseries. Accordingly, in the (S)ꢀseries, a small but significant advantage in ∆H° is experienced
parallel to the growing of the P ’ substituent into the S ’ pocket relative to the (R)ꢀseries.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
2
To further validate whether the glycerol molecules found in our cryoprotected crystal structures
have any distorting effect on the thermodynamic signature, we performed ITC titrations with the
addition of different concentrations of glycerol (up to 10%, Figure S6 in the Supporting
Information). These titrations revealed a systematical increase of ∆H° with increasing glycerol
concentration paralleled by a compensating decrease of –T∆S°. Most importantly, the relative
difference between ∆H° and –T∆S° remained unchanged, thus no specific effect and only an overall
systematic influence of the added glycerol was observed. Comparable systematic influences, for
example by the type of salt (NaCl or NaSCN) and its concentration used in the measurement buffer,
4
5
have been described previously. Similarly, systematic influences on ∆H° and –T∆S° were also
observed for the measurement with the addition of different concentrations of DMSO (Figure S7,
Supporting Information). These findings underscore that ITC data should only be recorded applying
highly comparable measurement conditions and evaluated relative to each other in congeneric
4
5
compound series.
ACS Paragon Plus Environment

Page 15 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
Binding kinetics of TLNꢀligand complex formation measured by SPR. The binding kinetic
parameters of 1–3 and 5–6 (Figure 9) were determined by singleꢀcycle SPR measurements
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
preformed in triplicate for each ligand. Kinetic analysis of the SPR sensorgrams was performed by
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
9,50
global analysis of the triplicate data applying a 1:1 binding model,
individual analysis of the sensorgrams. The relative standard deviations of the individual analysis
results were about 30% for k and 10% for k (Table S8, Supporting Information). Ligand 4 was
which agreed well with
on
off
not tested, as this ligand was synthesized at a later stage of the study. As depicted on the kinetic
map (Figure 9), 1 showed the fastest k within the series, whereas the dissociation rates for 2, 5 and
off
6 are slower and fall within a narrow window. Ligand 3 instead shows a significantly prolonged
dissociation rate compared to all other members of the series.
Buried SASAs of the TLNꢀligand complexes. Figure 10 shows the computed buried solvent
accessible surface areas (SASAs) of 1–6 in complex with TLN. One methyl group of the P ’ portion
2
of 5 remained undetected in the electron density (Figure 4E). Therefore, the missing methyl group
was modeled in two conformations A and B based on the crystal structure (Figure S9, Supporting
m
m
Information) and the buried SASAs of these two conformations were calculated. The SASA buried
2
within the (S)ꢀseries increases monotonously from 1→2→3 by approximately 15 Å per added
methyl group. TLNꢀ4 exhibits the largest buried SASA of all six ligands and the buried SASAs of 5
and 6 are slightly larger than those of their respective epimers 2 and 3.
DISCUSSION
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
In the current study, we wanted to validate our hypothesis that ligand binding to an open, rather
flat and solventꢀexposed binding pocket can be enhanced by optimizing the surface water network
wrapping around exposed parts of the bound ligand. We started with the previously characterized
3
2
peptidomimetic TLN inhibitor 1, and modified its hydrophobic solventꢀexposed P ’ substituent
2
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
attached to the parent scaffold. We improved binding affinity by maximizing the desolvation of the
increasingly bulky P ’ substituents along with an enhancement of the water network wetting the
2
surface of the formed complex. Moreover, we expected that the residence time of the complex
could be expanded with increasing quality and perfection of the formed water network. Prior to
synthesis, we predicted the putative water pattern around the designed ligands by MD simulations.
In our previous purely descriptive studies, we observed that small changes of the solventꢀexposed
ligand surface can strongly modulate the stability and complexity of the formed water network
2
9,30,32
adjacent to the bound ligands.
Since the spatial arrangement of water molecules across apolar
surfaces is governed by a complex architecture and the addition of a sole methyl group can already
lead to the unfavorable disruption of the adjacent water network, optimal hydration of a partly
solventꢀexposed apolar P ’ substituent is a challenging design task.
2
As starting point we chose 1, which already displayed rather potent inhibition properties but
showed local deficiencies in the solvation pattern next to its P ’ substituent. This was indicated by a
2
comparison with structures of closely related complexes. By merging features of their P2’
substituents with those of 1, we designed a small series of ligands comprising chiral aliphatic P2’
substituents. To predict their impact on the quality of the wetting surface water network, we
41
followed our recently introduced MD approach. To further validate our hypothesis, we considered
both chiral orientations, as the simulations suggested significant differences between the
stereoisomers. Since the designed ligands exhibited a second stereogenic center at P ’, epimeric
1
ACS Paragon Plus Environment

Page 17 of 88
Journal of Medicinal Chemistry
1
2
pairs of ligands resulted. This leads to the disadvantage that differences in the desolvation cost to
transfer the corresponding ligands of the epimeric pairs from the bulk solvent phase to the protein
pocket cannot be entirely excluded. However, as the two stereogenic centers are separated by
several bonds, we assume quite similar physicochemical properties for the matching epimeric pairs.
Across the series, the MD analysis suggested small but significant differences in the completeness
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
of the surface water network formed next to the different P ’ substituents rendering the (S)ꢀ
2
configurated 3 as most promising candidate of the series. Subsequently, we synthesized the ligands
stereoselectively and characterized the complexes formed via crystallography, microcalorimetry
and surface plasmon resonance. Concerning the crystal structure analysis, in all cases we obtained
diffraction data with very high resolution, also falling into a narrow window (mean: 1.22±0.10 Å,
Table 1). This is important to reliably compare details of the water structures between the different
crystal structures, as deviating resolution can affect such details and will complicate the mutual
51
matching of B factors.
The parent scaffold of all six ligands adopts virtually the same binding pose (Figure 3). Thus, the
observed differences between the studied ligands 1–6 most likely originate predominantly from the
desolvation differences of the gradually increasing and partly buried P ’ substituents and from
2
deviations of the formed surface water networks “wetting” the newly formed complexes. They
show varying degrees of completeness and perfection, which in turn, influence the thermodynamic
and binding kinetic signature of complex formation. Whereas the fixation of water molecules on the
surface of the bound ligand increases binding enthalpy and reduces entropy, in contrast binding
entropy is favored and enthalpy lowered by enhanced mobility up to the displacement of water
molecules into the bulk water phase. The mobility and occupancy of individual water molecules is
indicated by the spatial concentration of the electron densities (Figure 4) and the assigned B factors
(Figure 5). A strong fixation of a water molecule results from the embedding into a geometrically
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
rather constrained Hꢀbonding network, also involving the formation of Hꢀbonds to adjacent
functional groups of the protein’s amino acids.
Although essential in the current series, the arrangement of water molecules is possibly only one
component determining the thermodynamic signature of the binding process. The total
thermodynamic signature is in any case the sum of many contributions and might be composed of
partly compensating or mutually enhancing contributions. Hence, it is even more important that we
attempt to only evaluate relative differences of the studied complexes and not their absolute values.
Furthermore, ligand binding can be accompanied by global conformational adjustments of the
protein partly masking the thermodynamic signature of the local binding event. However, from our
experience with the system, global adjustments of TLN are unlikely. The enzyme has proven to be
highly rigid and the sole differences introduced between the congeneric ligands are their solventꢀ
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
exposed P ’ groups. Fenley et al. recently evaluated a largeꢀscale MD trajectory of BPTI (bovine
2
pancreatic trypsin inhibitor) and observed remarkable transitions between states of unchanged
52
overall Gibbs free energy but significantly altered enthalpy/entropy inventory. This entropy–
enthalpy transduction might suggest a physical mechanism underlying entropy–enthalpy
compensation in such systems. However, we propose that in our congeneric ligand series, where
binding occurs to a rigid protein, the ligands always address the same or very similar configurations
of TLN.
Prediction of solvation sites by MD simulations and their agreement with
crystallographically determined solvation sites. Based on our earlier study on four TLN
29
ligands, we developed an MD simulation protocol to predict water networks adjacent to solventꢀ
4
1
exposed ligand groups. This protocol correctly reproduced the rupture of a water network between
ACS Paragon Plus Environment

Page 19 of 88
Journal of Medicinal Chemistry
1
2
2
9
two ligands differing by only one single methyl group. The rupture was responded by a dramatic
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
–1
loss in binding enthalpy (∆∆H°methyl→H = +13.3±0.6 kJ mol ) and an increase in binding entropy (–
–1
T∆∆S°methyl→H = –7.7±0.4 kJ mol ), overall resulting in a lowered affinity (∆∆G°
=
methyl→H
–
1
+
5.7±0.3 kJ mol ). This example underlines the pronounced effect of the ligandꢀcapping water
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
network on the thermodynamic signature.
Importantly, our MD simulation approach does not require any a priori knowledge about water
positions but allows the prediction of hydration sites adjacent to protein–ligand complexes from
scratch in agreement with experiment. In the present study, the tool was applied to predict TLNꢀ2,
TLNꢀ3, TLNꢀ5 and TLNꢀ6. Its predictive power could be further assessed by simulating TLNꢀ1, as
3
2
its crystal structure had been determined prior to the present study. For TLNꢀ1, the computed
hydration sites match remarkably well with the difference electron density assigned to water
molecules by crystallography (Figure 2B). Only the population of W8 is underestimated, which is in
line with our previous observation that waterꢀtoꢀmethyl interactions are predicted as too weak by
41
the AMBER force field.
To estimate whether the designed P ’ substituents exhibit higher or lower hydration propensities
2
than those of 1, the predicted solvation sites were mutually compared. The simulations of TLNꢀ1
and TLNꢀ3 suggest some advantages of the latter adjacent to the P ’ substituents (Figure 2B,C).
2
This should render 3 superior to 1 with respect to affinity, since the enthalpic component of
stabilizing the water network is larger and can compensate for the increasing enthalpic cost to
(entropical beneficially) desolvate 3 over 1, as its substituent comprises two additional methyl
groups, resulting in a significantly larger buried surface area (Figure 10). For TLNꢀ6, a possible
rupture of the water network in the center of the apolar surface next to the P ’ substituent is
2
suggested. Facing the subsequently determined crystal structures with our predictions, TLNꢀ3
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
(
Figure 11A) exhibits the water molecules capping the apolar P ’ substituent which are likewise
2
predicted too weak (as W8 in TLNꢀ1). The in the crystal structure wellꢀstabilized W8 and W9 are
weakly indicated by the MD approach, and the relatively mobile W14 (high B factor and weak
electron density) is not predicted by the computer analysis, at least on the 48% threshold level. W15
is correctly predicted by a tubular solvation site, which also hosts W13. In TLNꢀ6, W8 and W9 are
not predicted by MD, which, however, agrees with the experimental observation that these water
molecules are significantly less stabilized in TLNꢀ6 than in TLNꢀ1 or TLNꢀ3 (cf. high B factors,
Figure 5). The shifted position of W10 was correctly predicted, resulting in a large gap toward W11
beyond Hꢀbonding distance. Water molecule W12 is not observed in the electron density, whereas
the MD simulation predicts a hydration site at this position. In summary, TLNꢀ1, TLNꢀ3 and TLNꢀ6
are convincingly predicted on qualitative level and the relative ranking of the epimer complexes
TLNꢀ3 (strong fixation) and TLNꢀ6 (weak fixation) was correctly assigned.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
The predicted solvation sites in TLNꢀ2 and TLNꢀ5 differ more strongly from the
crystallographically observed electron densities. This results from the disorder of the P2’
substituents of 2 and 5 indicated in the crystal structures, which was difficult to predict as the
disorder was not considered in the MD simulations (Figure S10, Supporting Information).
Correlation of structural data with thermodynamic signature of complex formation. Our
starting ligand 1 has been established as most potent binder from the previously studied series
3
2
(Figure 1B). We related its superior affinity to an entropically beneficial burial of its surface along
with the establishment of an extensive enthalpically favored surface water network. In TLNꢀ1, a
particularly favorable fiveꢀmembered water polygon is formed (Figure 4A, W5–W9), which is wellꢀ
5
3,54
known for its favorable energetic contribution.
ACS Paragon Plus Environment

Page 21 of 88
Journal of Medicinal Chemistry
1
2
3
4
Compared to TLNꢀ1, the additional methyl group in TLNꢀ2 creates disorder over two
conformational states, perturbing the neighboring water structure (Figure 7) and increasing the
mobility of the adjacent water molecules: W3 is observed in two orientations, and W9, W10, W12
and W13 show enhanced B factors (Figure 5). Consequently, enthalpy decreases from TLNꢀ1 to
TLNꢀ2 (Figure 8). This is overcompensated by a more favorable entropy resulting from the burial
of a larger apolar surface area (Figure 10) and the enhanced sidechain mobility, which is
entropically beneficial. Altogether, the more favorable entropy contribution is only partly
compensated by enthalpic losses, resulting overall in a slightly enhanced affinity of 2 over 1 by
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
–1
∆∆G°1→2 = –1.0±0.1 kJ mol .
The thermodynamic signature of 3 is comparable to that of 2, with a slightly elevated entropic
and virtually unchanged enthalpic term leading to an increased affinity of 3 over 2 (Figure 8). The
surface water networks of both complexes differ in several regards. In TLNꢀ3 (Figure 4C), a
complete sixꢀmembered water network polygon (W8, W9, W12–W15) is established adjacent to the
fiveꢀmembered one and is integrated in an eightꢀmembered ring structure (W3–W5, W9–W12, W14).
Such fused polygonal water arrangements can be considered as optimal solvation shell to coat the
surface of a formed protein–ligand complex in terms of Hꢀbonding and thus inherent enthalpy
contributions. Overall, TLNꢀ3 shows the most perfect water network along with an increased apolar
surface burial compared to TLNꢀ2, resulting in a slightly superior affinity. This leads to an
unchanged enthalpic contribution as desolvation costs are compensated by the formed enthalpically
favored water structure. For 3, however, an entropic benefit remains originating from the burial and
desolvation of the additional methyl group (“classical” hydrophobic effect).
The complexes TLNꢀ5 and TLNꢀ6 formed with (R)ꢀconfigurated ligands also show elaborate
surface water networks involving the formation of the stabilizing fiveꢀmembered water polygon
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
(
W5–W9). Detailed analysis suggests that the involved water molecules experience much higher B
factors than in the corresponding complexes with the (S)ꢀconfigurated epimers (Figure 5). This
supposedly less stable arrangement results from the inverted stereochemistry and increases the
steric demand of the (R)ꢀconfigurated P ’ substituents. Furthermore, one terminal methyl group of 5
2
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
is not detectable in the crystal structure suggesting enhanced mobility, likely increasing entropy and
reducing the stabilization of the adjacent water network. In contrast to the (S)ꢀseries, the (R)ꢀ
configurated substituents do not enhance affinity (Figure 8). The gradual enhancement in –T∆S°
with growing number of methyl groups is similar within the (S) and (R)ꢀseries. However, in the (R)ꢀ
series the loss in ∆H° completely nullifies the advantage in –T∆S°. Thus, overall the affinity
enhancement across the (S) series ligands results from an enthalpic advantage of the growing P2’
substituents. They achieve more elaborated and energetically improved surface water networks.
Ligand 4 shows an unexpected binding mode, as its P ’ group is flipped by 180° compared to the
2
other five ligands (Figure 4D). Remarkably, this ligand shows the largest surface burial across the
series, significantly higher than that of its epimer 1 (Figure 10). The flipped orientation takes
considerable impact on the established water structure, showing a much lower amount of recruited
water molecules compared to the other TLN complexes. This results in the striking observation that
TLNꢀ4, even though exhibiting the largest surface burial, shows the lowest affinity across all six
ligands (Figure 8). This underpins our observation that the sole burial of hydrophobic surface
portions of a ligand is clearly not sufficient to explain the binding features. Due to the considerable
conformational change of the P ’ substituent, a direct comparison of ∆H° and –T∆S° of TLNꢀ4 with
2
the thermodynamic signatures of the other five ligands is complicated.
ACS Paragon Plus Environment

Page 23 of 88
Journal of Medicinal Chemistry
1
2
3
4
Kinetic analysis of the ligands. The overall rather slow association of ligands binding to TLN is
likely governed by a large conformational transition (induced fit) of the protein, rendering
influences of individual ligands on kon rather insignificant. As a matter of fact, the experimental
determination of association rate constants is dependent on the concentration of the studied samples
and thus prone to additional experimental uncertainties (e.g. weighting errors or repeated freezeꢀ
thaw cycles of the inhibitor solution affecting its concentration). For these two reasons, we refrain
from a detailed interpretation of k . This uncertainty also affects K values determined by SPR, and
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
on
d
will also afflict a direct comparison with K values from ITC measurements. In any case, there
d
might be inherent differences from a theoretical point of view between both techniques involved,
making a direct comparison of K values taken from both methods difficult. ITC observes a system
d
under thermodynamic equilibrium conditions based on a particular binding model, whereas SPR
records a steady state situation in a flow cell using immobilized protein, which does not necessarily
relate to the same binding model as different structural states might determine binding kinetics. This
5
5
may lead to differences in the determined K values.
d
Based on the thermodynamic equilibrium K values determined by ITC, 3 clearly shows the
d
highest affinity of the series (Figure 8), whereas in the SPR measurements the second largest Kd
value was determined for this inhibitor (Table S7, Supporting Information). Ligand 3 shows,
however, a significantly longer residence time compared to all other ligands (Figure 9). Since the
local interactions of a specific inhibitor conceivably have a higher influence on the dissociation
kinetics, we relate this decreased dissociation rate constant of 3 to the formation of the pronounced
surface water network caging the hydrophobic P ’ substituent to stabilize the complex and thus
2
prolong residence time of the ligand.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
CONCLUSION
Drug optimization aims for the tailored optimization of binding parameters to endow a ligand
with the required potency, selectivity and binding efficacy. We increasingly recognize that the sole
optimization to enhanced binding affinity is not sufficient to render a ligand as most promising
candidate for further development. Consequently, additional parameters such as thermodynamic
and binding kinetic signatures are consulted to obtain a more detailed view on the binding process.
Furthermore, increasing resolution of the crystal structures determined across narrow series of
protein–ligand complexes discloses tiny differences in the binding poses, and adaptations of the
target protein or modulations of the “wetting” surface water networks. For the medicinal chemist
who performs ligand optimization by means of chemical synthesis, it is essential that these
modulations, which finally improve the ligand’s profile, result from properties of the bound ligand
and its partly solventꢀexposed substituents. They allow fineꢀtuning of affinity, enthalpy, entropy and
binding kinetic properties, as they are accomplished by wellꢀestablished medicinal chemistry
optimization steps. In the present study, we show that by means of optimizing the composition of a
partly exposed, apolar ligand substituent bound to a flat, solventꢀexposed binding pocket of a
protein, the relevant binding parameters can be fineꢀtuned using rational design principles. We
therefore had to analyze, predict and characterize the substituent’s burial and in parallel the quality
and perfection of the adjacent formed surface water network, which coats the formed complex. The
advantage of this concept is that the chemical adjustments needed to drive the thermodynamic and
kinetic parameters into a desired range are performed using the normal toolbox available to
medicinal chemists. Our strategy requires the following steps along an iterative design cycle: (i)
molecular design of the exposed substituent to optimize the pocket burial and adjacent surface
water layer, (ii) molecular dynamics simulations to validate the proposed surface water network,
(iii) ligand synthesis, followed by (iv) structural, microcalorimetric and binding kinetic
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment

Page 25 of 88
Journal of Medicinal Chemistry
1
2
characterization of the formed complexes. The Table of Contents Graphic shows the stepwise
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
affinity enhancement across our studied series. The most potent complex TLNꢀ3 (with the 2(S)ꢀ3,3ꢀ
trimethylbutyl P ’ substituent, at the far right of the diagram in the Table of Contents graphic) is by
2
about 1.5 orders of magnitude more potent in terms of affinity than the initial purely methylated
complex (at the far left of the diagram). As the detailed thermodynamic characterization shows, this
only partly results from enhancements of the “classical” hydrophobic effect. Moreover, additionally
important, mainly more enthalpyꢀdriven effects result from the optimization of the surface water
network that coats TLNꢀ3 almost perfectly and establishes several fused polygonal water
arrangements, which are characterized by a particular stability. Apart from the enhancement of the
thermodynamic profile, TLNꢀ3 shows prolonged residence time that results from a more stable
protein–ligand complex. Obviously, the optimized surface water layer captures and holds the ligand
more tightly to the protein, thus increasing the barrier for its release. It is possible that this is a
rather general concept to modulate binding kinetics, namely by enhancing the interaction of a bound
ligand with the adjacent surface water network. Exploitation of this property might allow the
medicinal chemist to fineꢀtune binding kinetic parameters via ligand optimization for many drug
targets.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
EXPERIMENTAL SECTION
Water Network Prediction by Molecular Dynamics Simulations. The crystal structure of
TLNꢀ1 (PDB code 4MZN) was used for modeling of the ligands as well as for the MD simulations.
In order to provide a common environment for modeling and simulation, the Cartesian coordinates
of protein, zinc ion, and cryobuffer molecules (DMSO, glycerol) were used. The preparation was
4
1
performed similarly to our protocol described previously.
After protonation, all
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
crystallographically observed water molecules were removed. All ligands were modeled within the
binding site of TLNꢀ1. As template structures for the modelling of the ligands, 1 was used for the
3
2
(
S)ꢀconfigurated 2 and 3, and the ligand in its complex to TLN from the PDB entry 4MTW was
used as a template structure for the (R)ꢀconfigurated 5 and 6, as they provided suitable exit vectors.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Modeling and a subsequent minimization of the S ’ groups was performed using the molecule
2
5
6
builder function and the AMBER99 force field implemented in MOE. Atomic charges for the
5
7
ligands were calculated with the RESP methodology based on quantum mechanical calculations
5
8
obtained by Gaussian09 at HF/6ꢀ31G* level. The MD simulations were performed with the
5
9
AMBER14 package, using the ff99SB force field and periodic boundary conditions. During all
simulation steps, all atoms, except hydrogen atoms and water molecules, were restrained to their
coordinates of the crystal structure. In a 20 ns production phase, water molecule positions were
recorded every 2 ps. This trajectory was analyzed to calculate the solvation sites using the
60
41
VOLMAP plugin in VMD. The protocol is described in detail in our earlier contribution.
1
13
31
Ligand Synthesis and Purification. H, C and P NMR spectra were recorded on a JEOL
ECXꢀ400 or JEOL ECAꢀ500 instrument. All chemical shift values are reported in ppm relative to
31
the nonꢀdeuterated solvent signal. An external standard was used for P NMR spectra (referenced
13
to: 85% H PO ) and C NMR spectra in D O (referenced to: trimethylsilyl propanoic acid). ESIꢀ
3
4
2
MS spectra were recorded on a QꢀTrap 2000 system by Applied Biosystems. For the description of
multiplicity the following abbreviations were used: s = singlet, d = doublet, t = triplet, q = quartet,
quint = quintet, dd = doublet of doublet, m = multiplet, br = broad signal. For high resolution ESIꢀ
MS a LTQꢀFT Ultra mass spectrometer (Thermo Fischer Scientific) was used. EIꢀMS analysis was
carried out on a Micromass AutoSpec instrument. For HPLC chromatography a Shimadzu LCꢀ20
ACS Paragon Plus Environment

Page 27 of 88
Journal of Medicinal Chemistry
1
2
system equipped with a diode array detector was used. Analytic separations were carried out with a
MN Nucleodur 100ꢀ5 C18 ec 4.6×250 mm column using a waterꢀacetonitrile gradient. For semiꢀ
preparative separations a Water XSelect CSH C18 10×250 mm column using a waterꢀacetonitrile
gradient was used. The purity of all inhibitors used for biophysical experiments was greater than
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
9
5%, as determined by HPLC
General procedure for the synthesis of acyloxazolidinones 9–13: nꢀBuLi (1.2 eq) was slowly
added to a solution of 7 or 8 (1.0 eq) in THF at ꢀ78 °C under argon. The solution was allowed to
warm to room temperature over 30 minutes. The respective acid chloride (1.1 eq) was added to the
yellow solution and the reaction mixture was stirred for 2 h at ꢀ78 °C. The reaction was quenched
with saturated NH Cl solution and extracted with EtOAc (3×20 mL). The combined organic
4
extracts were washed with brine and dried over MgSO . The crude product was purified by silica
4
gel column chromatography (cyclohexane/EtOAc 5:1).
(S)-4-Benzyl-3-(3-methylbutanoyl)oxazolidin-2-one (9): Compound 9 was synthesized according
to the general procedure using nꢀBuLi (2 M in hexanes, 2.4 mL, 6.00 mmol), (S)-4ꢀ
benzyloxazolidinꢀ2ꢀone (7, 886 mg, 5.00 mmol) and 3ꢀmethylbutanoyl chloride (663 mg, 5.50
1
mmol). The product was obtained as a colorless oil (1243 mg, 4.76 mmol, 95%). H NMR (400
MHz, CDCl ) δ = 1.02 (t, J = 6.6, 6H), 2.17–2.28 (m, 1H), 2.71–2.76 (m, 1H), 2.79 (dd, J = 10.2,
3
6
.0, 1H), 2.89 (dd, J = 16.2, 6.7, 1H), 3.31 (dd, J = 13.3, 3.3, 1H), 4.12–4.23 (m, 2H), 4.63–4.74
1
3
(
m, 1H), 7.18–7.39 (m, 5H). C NMR (101 MHz, CDCl ) δ = 22.6, 22. 7, 25.1, 38.1, 44.1, 55.3,
3
+
66.2, 127.4, 129.1, 129.5, 135. 5, 153.6, 172.8. MS (ESI+) m/z calculated for C H NO [M+H] :
1
5
20
3
2
62.32; found: 262.09.
S)-4-Benzyl-3-(3,3-dimethylbutanoyl)oxazolidin-2-one (10): Compound 10 was synthesized
(
according to the general procedure using nꢀBuLi (2.5 M in hexanes, 3.4 mL, 8.40 mmol), (S)-4ꢀ
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
benzyloxazolidinꢀ2ꢀone (7, 1240 mg, 7.00 mmol) and 3,3ꢀdimethylbutanoyl chloride (1036 mg,
1
7
.70 mmol). The product was obtained as a colorless oil (1776 mg, 6.45 mmol, 92%). H NMR
(
400 MHz, CDCl ) δ = 1.09 (s, 9H), 2.71 (dd, J = 13.2, 10.1, 1H), 2.86 (d, J = 14.9, 1H), 2.99 (d, J
3
=
14.9, 1H), 3.35 (dd, J = 13.3, 3.2, 1H), 4.07–4.22 (m, 2H), 4.62–4.78 (m, 1H), 7.16–7.43 (m,
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
13
5
H). C NMR (101 MHz, CDCl ) δ = 29.7, 31.6, 38.2, 46.3, 55.5, 65.9, 127.4, 129.1, 129.5, 135.6,
3
+
153.6, 172.0. MS (ESI+) m/z calculated for C H NO [M+H] : 276.35; found: 276.08.
1
6
22
3
(R)-4-Benzyl-3-butyryloxazolidin-2-one (11): Compound 11 was synthesized according to the
general procedure using nꢀBuLi (2.5 M in hexanes, 6.6 mL, 16.50 mmol), (R)-4ꢀbenzyloxazolidinꢀ
2
ꢀone (8, 2660 mg, 15.00 mmol) and butyryl chloride (2238 mg, 21.00 mmol). The product was
1
obtained as a colorless oil (3483 mg, 14.08 mmol, 94%). H NMR (400 MHz, CDCl ) δ = 1.01 (t, J
3
=
7.4 Hz, 3H), 1.66–1.80 (m, 2H), 2.77 (dd, J = 13.3, 9.7 Hz, 1H), 2.83–3.01 (m, 2H), 3.30 (dd, J =
1
3
1
3.3, 3.2 Hz, 1H), 4.14–4.24 (m, 2H), 4.63–4.72 (m, 1H), 7.18–7.36 (m, 5H). C NMR (101 MHz,
CDCl ) δ = 13.8, 17.8, 37.5, 38.1, 55.3, 66.3, 127.5, 129.1, 129.6, 135.4, 153.6, 173.4. MS (ESI+)
3
+
m/z calculated for C H NO [M+NH ] : 265.33; found: 265.19.
1
4
18
3
4
(R)-4-Benzyl-3-(3-methylbutanoyl)oxazolidin-2-one (12): Compound 12 was synthesized
according to the general procedure using nꢀBuLi (2.5 M in hexanes, 2.8 mL, 7.00 mmol), (R)-4ꢀ
benzyloxazolidinꢀ2ꢀone (8, 1050 mg, 5.90 mmol) and 3ꢀmethylbutanoyl chloride (787 mg, 6.50
1
mmol). The product was obtained as a colorless oil (1368 mg, 5.23 mmol, 88%). H NMR (500
MHz, CDCl ) δ = 1.01 (dd, J = 8.3, 6.7, 2H), 2.15–2.29 (m, 1H), 2.71–2.81 (m, 2H), 2.89 (dd, J =
3
16.2, 6.7, 1H), 3.31 (dd, J = 13.3, 3.3, 1H), 4.09–4.23 (m, 2H), 4.63–4.73 (m, 1H), 7.18–7.37 (m,
13
5
1
H). C NMR (126 MHz, CDCl ) δ = 22.5, 22.7, 25.1, 38.1, 44.1, 55.3, 66.2, 127.4, 129.0, 129.5,
3
+
35.4, 153.5, 172.8. MS (EI) m/z calculated for C H NO [M] : 261.32; found: 261.
15 19
3
ACS Paragon Plus Environment

Page 29 of 88
Journal of Medicinal Chemistry
1
2
3
4
(
R)-4-Benzyl-3-(3,3-dimethylbutanoyl)oxazolidin-2-one (13): Compound 13 was synthesized
according to the general procedure using nꢀBuLi (2.5 M in hexanes, 3.4 mL, 8.40 mmol), (R)-4ꢀ
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
benzyloxazolidinꢀ2ꢀone (8, 1240 mg, 7.00 mmol) and 3,3ꢀdimethylbutanoyl chloride (1036 mg,
1
7
.70 mmol). The product was obtained as a colorless oil (1876 mg, 6.81 mmol, 97%). H NMR
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(
500 MHz, CDCl ) δ = 1.09 (s, 9H), 2.71 (dd, J = 13.3, 10.0 Hz, 1H), 2.86 (d, J = 14.9 Hz, 1H),
3
2.99 (d, J = 14.9, 1H), 3.34 (dd, J = 13.3, 3.3 Hz, 1H), 4.10–4.18 (m, 2H), 4.64–4.73 (m, 1H),
1
3
7
.21–7.36 (m, 5H). C NMR (126 MHz, CDCl ) δ = 29.7, 31.5, 38.1, 46.2, 55.5, 65.9, 127.4,
3
+
1
29.0, 129.5, 135.6, 153.6, 172.0. (MS ESI+) m/z calculated for C H NO [M+H] : 290.38;
1
7
24
3
found: 290.13.
General procedure for the synthesis of acyloxazolidinones 14–18: A solution of diisopropylamine
(1.3 eq) in dry THF under argon was cooled to ꢀ78 °C. nꢀBuLi (1.2 eq) was slowly added to the
solution and the reaction mixture was allowed to warm to room temperature over 60 minutes. After
cooling to ꢀ80 °C the respective oxazolidinone 9–13 (1.0 eq) was added dropwise to the mixture.
After 60 minutes MeI (4.0 eq) was added to the solution. The reaction mixture was stirred for 5 h
without further cooling. The reaction was quenched with saturated NH Clꢀsolution and extracted
4
with EtOAc (3×20 mL). The combined organic extracts were washed with brine and dried over
MgSO . The crude reaction product was purified by silica gel column chromatography
4
(cyclohexane/EtOAc 6:1).
(S)-4-Benzyl-3-((S)-2,3-dimethylbutanoyl)oxazolidin-2-one (14): Compound 14 was synthesized
according to the general procedure using diisopropylamine (591 mg, 5.84 mmol), nꢀBuLi (2.5 M in
hexanes, 2.2 mL, 5.39 mmol), oxazolidinone 9 (1173 mg, 4.49 mmol) and MeI (2550 mg, 17.96
mmol). Recrystallization of the chromatographically pure product from cyclohexane gave the
1
diastereomerically enriched product as a colorless solid (dr 20:1, 892 mg, 3.24 mmol, 68%). H
ACS Paragon Plus Environment