Pyrrolidine derivatives as plasmepsin inhibitors: Binding mode analysis assisted by molecular dynamics simulations of a highly flexible protein

Article abstract of DOI:10.1002/cmdc.200900452

Plasmepsins II (EC number: 3.4.23.39) and IV (EC number: 3.4.23.B14) are aspartic proteases present in the food vacuole of the malaria parasite Plasmodium falciparum and are involved in host hemoglobin degradation. A series of pyrrolidine derivatives, originally synthesized as HIV-1 protease inhibitors, were tested for activity against plasmepsin (Plm). Inhibitors in the nanomolar range were discovered for the Plm II and IV isoforms. Detailed studies were carried out to identify putative binding modes that help to explain the underlying structure-activity relationships. Reasonable binding modes were generated for pyrrolidine-3,4-diester derivatives and a substituted 3,4-diaminopyrrolidine inhibitor by using a crystal structure of inhibitor-bound Plm II (PDB ID: 1LEE). Modeling studies indicated that the flap of available Plm crystal structures is not sufficiently opened to accommodate the 3,4-bis(aminomethylene)pyrrolidines. Molecular dynamics simulations were performed to analyze the flexibility of the protein in greater detail, leading to a binding mode hypothesis for the 3,4-bis(aminomethylene)pyrrolidines and providing further insight and general implications for the design of Plm II inhibitors.

Full text of DOI:10.1002/cmdc.200900452

DOI: 10.1002/cmdc.200900452
Pyrrolidine Derivatives as Plasmepsin Inhibitors: Binding
Mode Analysis Assisted by Molecular Dynamics
Simulations of a Highly Flexible Protein
Torsten Luksch,[a] Andreas Blum,[a] Nina Klee,[a] Wibke E. Diederich,[a] Christoph A. Sotriffer,[a, b]
and Gerhard Klebe*[a]
Plasmepsins II (EC number: 3.4.23.39) and IV (EC number:
3.4.23.B14) are aspartic proteases present in the food vacuole
of the malaria parasite Plasmodium falciparum and are involved
in host hemoglobin degradation. A series of pyrrolidine deriva-
tives, originally synthesized as HIV-1 protease inhibitors, were
tested for activity against plasmepsin (Plm). Inhibitors in the
nanomolar range were discovered for the Plm II and IV iso-
forms. Detailed studies were carried out to identify putative
binding modes that help to explain the underlying structure–
activity relationships. Reasonable binding modes were generat-
ed for pyrrolidine-3,4-diester derivatives and a substituted 3,4-
diaminopyrrolidine inhibitor by using a crystal structure of in-
hibitor-bound Plm II (PDB ID: 1LEE). Modeling studies indicated
that the flap of available Plm crystal structures is not sufficient-
ly opened to accommodate the 3,4-bis(aminomethylene)pyrro-
lidines. Molecular dynamics simulations were performed to an-
alyze the flexibility of the protein in greater detail, leading to a
binding mode hypothesis for the 3,4-bis(aminomethylene)pyr-
rolidines and providing further insight and general implications
for the design of Plm II inhibitors.
Introduction
Malaria is one of the main infectious diseases in the world,
with about one million fatalities per year.[1] It is caused by para-
sites of the genus Plasmodium. The species that infect humans
are P. falciparum, P. malariae, P. ovalae, and P. vivax. Of these
species, P. falciparum causes the most life-threatening form of
malaria. The organism is transmitted by the female Anopheles
mosquito.[1] General symptoms of the disease are fever, chills,
nausea, and flu-like illness, which in severe cases lead to coma
and death. Although there are several antimalarial drugs on
the market, the increasing number of multi-drug-resistant par-
asite strains requires the development of novel therapeutic
agents.[2,3]
with side chains from other ligands known to be well suited to
address the specificity pockets of the enzyme.[17] Ki values in
the low micromolar range were observed toward HIV-1 pro-
tease as well as cathepsin D. The crystal structure of this inhibi-
tor in complex with HIV-1 protease was subsequently deter-
mined (PDB ID: 1XL2). In a second design cycle, 10 additional
compounds were synthesized.[18]
Although HIV-1 protease belongs to the retroviral proteases
whereas Plms are aspartic proteases with a pepsin-like fold,
various examples of inhibitors designed for HIV-1 protease also
show strong inhibition of Plm II.[19–21] The two proteins share
sequence similarities that could be responsible for the similar
inhibition profiles; in particular, they exhibit identical catalytic
dyad and peptide recognition motifs.[16] Furthermore, both as-
partic proteases feature a common flap (two in HIV-1 protease)
that is involved in ligand binding. The flaps close upon sub-
strate accommodation, thereby excluding solvent from the
inner catalytic cavity of the enzyme.[22] Because our pyrrolidine
derivatives were designed to address the family-wide features
of aspartic proteases, we anticipated inhibition of the Plms as
well.
Plasmepsins have been described as promising targets for
the discovery of antimalarials. They belong to the family of as-
partic proteases and are involved in the initial steps of hemo-
globin degradation, which is the main source of amino acids
for the parasite.[4,5] The catabolism of human hemoglobin
takes place during the erythrocytic stage within the acidic
food vacuole of the parasite. Although recent evidence has
shown that P. falciparum can survive without vacuolar plas-
mepsins under culture,[6] many plasmepsin (Plm) inhibitors are
fatal to the parasite.[7–10] There are four plasmepsins present in
this lysosome-like organelle: Plm I,[8,11] Plm II,[4,12] Plm IV,[13] and
HAP (histo-aspartic protease).[13] The redundant functional role
of these enzymes in hemoglobin digestion has been demon-
strated by Plm gene deletion. This suggests that more effective
drugs may be obtained by simultaneously blocking multiple
Plm isoforms.[14,15]
[a] Dr. T. Luksch, Dr. A. Blum, N. Klee, Prof. Dr. W. E. Diederich,
Prof. Dr. C. A. Sotriffer, Prof. Dr. G. Klebe
Philipps-Universitꢀt Marburg
Institut fꢁr Pharmazeutische Chemie
Marbacher Weg 6, 35032 Marburg (Germany)
Fax: (+49)6421-28-28994
E-mail: Klebe@staff.uni-marburg.de
[b] Prof. Dr. C. A. Sotriffer
Recently, a novel 3,4-bis(aminomethylene)pyrrolidine scaffold
was designed and synthesized as a general aspartic protease
inhibitor lead structure.[16] The scaffold core was decorated
Current address: Universitꢀt Wꢁrzburg
Institut fꢁr Pharmazie und Lebensmittelchemie
Am Hubland, 97074 Wꢁrzburg (Germany)
ChemMedChem 2010, 5, 443 – 454
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
443

G. Klebe et al.
MED
To suggest reasonable binding modes, previously solved
Plm II crystal structures were analyzed for their suitability to ac-
commodate the pyrrolidine-type scaffold described herein. Be-
cause only limited structural information is available for the
highly flexible Plms, molecular dynamics (MD) simulations
were also performed and analyzed in detail.
Table 1. Experimental inhibitory activities of the compounds studied.
1
Initial MD simulations for Plm II were performed in 2001.[23]
Within a short simulation time of 100 ps, residues that play an
important role in hydrogen bonding to the inhibitor were
identified. In the research group of ꢁqvist, several MD simula-
tions of protein–ligand complexes were carried out in order to
determine binding modes and to predict Ki values for com-
pounds with various scaffolds such as dihydroxyethylenes,
macrocyclic inhibitors, and allophenylnorstatine inhibitors.[24–28]
Recently, MD simulations were also used to obtain insight into
the maturation process from pro-Plm II to Plm II[29] and to
derive guidelines for the design of selective Plm inhibitors.[30]
MD studies have further been used to assess the relative stabil-
ity of putative ligand binding modes, as recently studied for
the binding of halofantrine to Plm.[31] All these studies indicate
that Plm is a difficult and highly flexible target and that MD
simulations are helpful and necessary to overcome the limited
structural information available.
Compd
R
Enzyme Ki [mm]
Plm II
51.2
Plm IV
Cat D
1.1
1.2
168
>1000
47.2
11.4
19.0
41.2
>1000
53.4
1.3
1.4
1.5
7.4
53.8
134
>1000
163
1.0
>1000
1.6
1.7
0.8
0.5
0.10
0.09
0.37
Results and Discussion
2
Enzyme inhibition
A total of 10 compounds were tested for activity against Plm II,
Plm IV, and the human aspartic protease cathepsin D (Cat D).
They can be divided into three molecule classes with different
core structures (1, 2, and 3; Table 1). Compounds 1.1–1.7
belong to the first series of molecules containing a 3S,4S-pyrro-
lidinediol scaffold 1, which is symmetrically decorated, via
esterification, with substituents of various size and polarity.
The second compound class is composed of a 3S,4S-diamino-
pyrrolidine core 2 and is substituted with two benzyl and two
benzoyl moieties. Compounds 3.1 and 3.2 are racemic mix-
tures of trans-3,4-bis(aminomethylene)pyrrolidines 3, both dec-
orated with four side chains that presumably interact with the
sub-pockets of different aspartic proteases (Table 1).
Compd
R
Enzyme Ki [mm]
Plm II
0.43
Plm IV
Cat D
11.7
2
1.5
3
R2:
R3:
R1:
Well-diffracting crystals for HIV-1 protease in complex with a
3,4-bis(aminomethylene)pyrrolidine derivative were ob-
tained.[17] The crystals were grown in the presence of a racemic
inhibitor mixture.[17] Because the 3R,4R-bis(aminomethylene)-
pyrrolidine stereoisomer was found in the crystal structure
(PDB ID: 1XL2), it was hypothesized that the R,R enantiomer
binds more strongly than the S,S enantiomer. Encouraged by
these results, a second compound series to inhibit HIV-1 pro-
tease was synthesized (compounds 1.1–1.7) by following an
enantioselective synthetic route.[18] To optimize key interactions
toward the flap residues of HIV-1 protease, the scaffold was al-
tered by decreasing the linker length. In addition, the number
of appended side chains was limited to two. Compounds 1.1–
1.7, originally synthesized for HIV-1 protease, were also tested
against Plm II, Plm IV, and Cat D.
Compd
R4
Enzyme Ki [mm]
Plm II
Plm IV Cat D
3.1
3.2
0.63
0.35 0.67
0.50
0.18 0.37
While the 3,4-bis(aminomethylene)pyrrolidines 3.1 and 3.2
exhibited Ki values in the sub-micromolar range, the pyrroli-
dine-3,4-diester derivatives functionalized with two monocyclic
444
www.chemmedchem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2010, 5, 443 – 454

Pyrrolidine Derivatives as Plasmepsin Inhibitors
aromatic rings resulted only in single- and double-digit micro-
molar activities. For example, 1.3 showed a Ki value of 11 mm
for Plm II, 7 mm for Plm IV, and 53 mm for Cat D.
tor 2, with Ki values of 430 nm for Plm II and 1.5 mm for Plm IV,
was obtained. In comparison to the most potent compounds
of series 1, inhibitor 2 shows a decreased ligand efficiency
(0.21 for Plm II and 0.19 for Plm IV), but reasonable ligand “fit
quality” values of 0.84 for Plm II and 0.76 for Plm IV were still
achieved. Fortunately, slight improvements in selectivity
toward human cathepsin D were observed.
Within the series of derivatives bearing 1 as the core ele-
ment, compounds with isobutyl, neopentyl, and cyclohexyl
moieties were the weakest binders to the Plms, with Ki values
in the high double- and triple-digit micromolar range (1.1, 1.2,
and 1.5). For Cat D, no inhibition was detected up to 1 mm. A
gain in affinity was achieved by replacing the alkyl chains with
monocyclic aromatic ring systems. With a Ki value of 11 mm
toward Plm II, 1.3 is about fourfold more potent than 1.2 (Ki =
47 mm). Similar results were obtained for Plm IV (sixfold in-
crease in affinity). The removal of one methylene unit (com-
pound 1.4) relative to 1.3 has almost no effect on the binding
affinity to Plm II, whereas a sevenfold decrease was detected
for Plm IV. The pyrrolidine-3,4-diester equipped with two 2-
naphthyl moieties further increased the potency (compound
1.6). The Ki value of 1.6 for Plm II is 1 mm, with slightly better Ki
values for Plm IV and Cat D (0.8 and 0.5 mm, respectively). The
most potent inhibitor in this series is 1.7, with a Ki value of
100 nm for Plm II and 90 nm for Plm IV. It shows a 1600-fold
greater binding affinity for Plm II than 1.5, the weakest-binding
inhibitor of all compounds measured.
Docking studies
Currently, 15 P. falciparum Plm II–inhibitor complex crystal
structures are available; these can be subdivided into three
groups exhibiting different ligand binding modes.[36] Pepstati-
n A and related ligands are tightly embraced by the protein
(PDB IDs: 1SME, 1ME6, 1XDH, 1XE5, 1XE6, 1W6I, 1W6H, 1M43,
and 2R9B). The hydroxy group of the statine moiety forms hy-
drogen bonds to the catalytic dyad (Figure 1a). A difference is
only observed for the peptide-based ligand in a recently
solved crystal structure (PDB ID: 2R9B), in which the hydrogen
bonds to the catalytic dyad are mediated by a water molecule.
Further key interactions are formed by two inhibitor amide
oxygen atoms to the backbone nitrogen of V78 and to the
side chain of S79. Both amino acids belong to a b-hairpin
structure, known as a “flap”, that closes with a virtually perpen-
dicular orientation to the binding cleft. The flap is a flexible
The calculated ligand efficiency of 1.1 is 0.31 (in kcal
molꢀ1 · [no. of heavy atoms]ꢀ1
) for Plm II and 0.27 for
Plm IV.[32,33] These values are promising, as inhibitors with
ligand efficiencies ~0.3 are con-
sidered efficient binders.[34] The
ligand efficiency for the largest
compound 1.7 is 0.29 for both
Plm II and IV, and is thus similar
to 1.1. This is remarkable, as it
has been observed that ligand
efficiency depends on ligand
size, with smaller ligands having
greater efficiencies, on average,
than larger ligands.[35] Based on
the observation that the maxi-
mum observed ligand efficiency
decreases for molecules with
higher numbers of heavy atoms,
Reynolds et al. introduced a new
scoring scheme termed ligand
“fit quality” that normalizes the
ligand efficiency values: Most ef-
ficient binders in the data set
were scaled to have a top score
of 1.0 across a wide range of
molecular sizes.[35] The calculated
ligand “fit quality” values of 1.7
for both Plms are 0.95, which in-
dicates near-optimal ligand bind-
ing.
Replacing the ester functional-
ities with amide groups, and fol-
Figure 1. Hydrogen bonds of a) pepstatin A and b) rs370 in the active site of Plm II. c) Putative hydrogen bonds of
3S,4S-diesterpyrrolidine 1 and 3S,4S-diaminopyrrolidine 2 to Plm II. d) Putative hydrogen bonds of 3,4-bis(amino-
lowing subsequent decoration
with four benzyl moieties, inhibi- methylene)pyrrolidine 3 to Plm II.
ChemMedChem 2010, 5, 443 – 454
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
445

G. Klebe et al.
MED
structural unit, present in all aspartic proteases, that interacts
with substrates and inhibitors.[37,38]
conformers of Plm II found either in the complex with pepstati-
n A (PDB ID: 1SME) or in the more open complex structure
(PDB ID: 1LEE). The pyrrolidine nitrogen atom was considered
as protonated, whereas the catalytic dyad was considered to
be fully deprotonated as suggested by pKa model calculations
on HIV-1 protease.[43] The ligand functionalities assigned to in-
teract with the flap region (Figure 1c,d) were rotated toward
the backbone nitrogen atom of V78 and to the side chain of
S79.
The second group of ligands found in co-crystal structures
(PDB IDs: 1LEE, 1LF2, and 1LF3) comprises a hydroxypropyl-
amine scaffold that forms a similar pattern of conserved hydro-
gen bonds to the catalytic aspartates and the flap residues V78
and S79[39] as the pepstatin-A-like ligands (Figure 1b). In these
complexes, the binding cavity remains more open than in the
pepstatin A complexes.[40]
The third group of ligands present in co-crystal structures
(PDB IDs: 2BJU, 2IGY, and 2IGX) induces significant differences
in the protein conformation.[41] The N-substituted piperidine
scaffold interacts with the catalytic aspartates via an interstitial
water molecule. Moreover, no interactions to the flap residues
are observed. Instead, major changes in the flap region occur,
and a new tunnel-shaped hydrophobic cavity opens up.
In the design of derivatives with 1, 2, and 3 as core struc-
tures, it was anticipated to directly address both the aspartate
residues of the catalytic dyad and the flap residues (Fig-
ure 1c,d). Therefore, the choice of Plm II in the open confor-
mation and in the conformation found in the complex struc-
ture with pepstatin A was considered as a promising starting
point for our considerations to generate appropriate binding
modes.
For all three ligand series, hydrogen bonds could not be
formed to the flap residues as found in the orientation ob-
served in the closed Plm II conformer with bound pepstatin A
(PDB ID: 1SME). However, manual placement of 1 and 2 into
the binding pocket of the more open complex structure (PDB
ID: 1LEE) could be achieved. In order to optimize these initial
ligand geometries, an energy minimization was performed.
During this minimization using the MOLOC MAB force field[44]
the protein was kept rigid, allowing full flexibility of the inhibi-
tor. In all minimized complexes the pyrrolidine core position
remains between the two catalytic aspartates, making short-
distance hydrogen bonds to both carboxylate groups. Addi-
tional hydrogen bonds are formed to the flap residues V78
and S79 (Figure 2). The side chains of all energy-minimized in-
hibitors point toward the S2 and S2’ pockets.
A co-crystal structure of HIV-1 protease with a scaffold relat-
ed to 3 is available (PDB ID: 1XL2). The central pyrrolidine
moiety is found at the pivotal position between the two cata-
lytic aspartate residues.[17] The ring adopts an envelope confor-
mation, and both ring substituents are found in an axial posi-
tion. Experimental pKa values for unsubstituted pyrrolidines
suggest the central nitrogen atom of the ligand is protonat-
ed.[42] Furthermore, pKa calculations of a pyrrolidine derivative
in complex with HIV-1 protease
As described above, the activity of compounds 1.1–1.7 in-
creases in the order: isobutyl < benzyl < 2-naphthyl < 1-
naphthyl (Table 1). The modeled binding modes support these
findings (Figure 2). S2 and S2’ are spacious pockets that are
not fully occupied by small side chains such as a 1-isobutyl
substituent (Figure 2a). The larger benzyl moiety can form
better hydrophobic interactions (Figure 2b). A further gain in
affinity is observed for two 2-naphthyl substituents. The inhibi-
suggest that both catalytic as-
partates of the HIV-1 protease
should be deprotonated upon
ligand binding.[43] Recently, the
crystal structure of 1.7 in com-
plex with HIV-1 protease was de-
termined (PDB ID: 3BC4).[18] Al-
though binding of the pyrroli-
dine nitrogen atom to the cata-
lytic dyad was observed, no
polar interactions with the flap
region were detected due to an
alternate protein conformation
revealing an open flap. Based on
these differences, only the struc-
ture in the closed-flap conforma-
tion (PDB ID: 1XL2) was consid-
ered for the generation of a pu-
tative binding mode for Plm II.
Taking the aforementioned
findings as a prerequisite for our
model generation, we tried to
manually place the ligand core
Figure 2. Modeled binding modes of compounds a) 1.1, b) 1.3, c) 1.7, and d) 2 in the binding pocket of Plm II
structures of 1–3 into different (PDB ID: 1LEE).
446
www.chemmedchem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2010, 5, 443 – 454

Pyrrolidine Derivatives as Plasmepsin Inhibitors
tor with two 1-naphthyl groups shows a tenfold improved af-
finity relative to the 2-naphthyl-substituted diester, with a Ki
value of 100 nm for Plm II. This can be attributed to the more
favorable burial of the ligand side chains in the S2 and S2’
pockets (Figure 2c).
1 ns to 10 ns and closely monitored further conformational
transitions along the trajectory.
In the remainder, we focus exclusively on the binding site
region as the area of highest interest in the context of struc-
ture-based drug design, particularly with respect to the issue
of flexibility. A total of 32 residues were selected as part of the
binding pocket. Mobility in this region is quantitatively ana-
lyzed by a mutual comparison of individual frames along the
trajectory in terms of a 2D RMSD plot. For this purpose, 500
snapshots were extracted at regular 20 ps intervals along the
entire 10 ns trajectory. It is immediately apparent that three
main conformational sub-states can be distinguished
(Figure 3).
A similar binding mode can be proposed for 2 bearing the
3S,4S-diaminopyrrolidine scaffold: Both aspartates as well as
the flap residues V78 and S79 are addressed by 2 (Figure 2d).
The four side chains of this inhibitor are placed in the four
sub-pockets (S1, S2, S1’, and S2’) forming hydrophobic interac-
tions.
The compounds of series 1 and 2 were synthesized as pure
enantiomers, exhibiting an S,S configuration at both stereo-
genic centers.[18] The synthesis of 2 is outlined below in the Ex-
perimental Section. The derivatives of series 3 have been pre-
pared and tested as racemic mixtures.[17] Considering the crys-
tal structure with HIV-1 protease, only the 3R,4R enantiomer
was detected,[17] which leads to the assumption that this enan-
tiomer represents the more potent binder.
In analogy to the results observed for HIV-1 protease, im-
proved binding of the R,R enantiomer of 3 is also anticipated
for Plm II. Docking of these stereoisomers of 3.1 and 3.2 to the
Plm II conformer (PDB ID: 1LEE) was attempted. However, no
reasonable binding mode could be generated in which all in-
teractions with the flap region and the catalytic dyad are pres-
ent. Clearly, the lid of the flap would have to open further in
order to allow hydrogen bond formation with the two aspar-
tates and residues V78 and S79 simultaneously. Nevertheless,
because potent binding is experimentally observed, the ques-
tion arises whether Plm II can adapt to host 3.1 to 3.2 in an ex-
tended binding pocket. To gather insight into the dynamic
properties of the enzyme, MD simulations were performed.
Figure 3. 2D RMSD plot using snapshots extracted every 20 ps. The mutual
RMSD between two snapshots is represented by the color code displayed at
right. All values are given in ꢁ. For the RMSD measurements, all side chain
atoms were included.
Molecular dynamics simulations
The available structural data indicate that Plm II is a highly flex-
ible protein. For example, in the asymmetric unit of the Plm II–
pepstatin A co-crystal structure (PDB ID: 1SME) two molecules
are found that do not exhibit the same conformation, indicat-
ing intrinsic inter-domain flexibility.[38] Furthermore, a compari-
son of uncomplexed Plm II (PDB ID: 1LF4) with Plm II in com-
plex with EH58 (PDB ID: 1LF3) reveals movements of residues
in the binding pocket to accommodate bulky groups of the in-
hibitor or to form hydrogen bonds.[40]
The first conformational sub-state is sampled for 4.3 ns at
the beginning of the trajectory (“conformationall 1”, Call1). The
second conformational family Call2 is then maintained for
about 4.4 ns before a transition to Call3 occurs, which remains
for the following 1.2 ns. Conformations within the same sub-
state show RMSD values mostly <1.4 ꢁ, while conformations
of different sub-states show RMSD values of up to 2.3 ꢁ.
To detect the molecular changes giving rise to this confor-
mational flip, we further divided the binding pocket into indi-
vidual sub-pockets (S1, S2, S1’, and S2’) and generated 2D
RMSD plots for each of them (Figure 4).
Until now, only three distinct binding modes for the differ-
ent inhibitor classes have been discovered. Structural studies
indicate, however, that Plm II allows significant conformational
flexibility and could possibly adapt in a way to accommodate
derivatives 3.1 and 3.2 with the extended pyrrolidine core
structure.
The S2 pocket (T35, S215, T217, S218, A219, T221, and I290)
remains nearly unchanged and represents essentially only one
conformation along the entire trajectory (Figure 4a). The
RMSDs of these residues are comparatively low, with move-
ments of 1.3 ꢁ on average.
To further sample the conformational space of the protein
and to explore its adaptive properties, a multi-nanosecond MD
simulation was performed. We recently reported MD simula-
tions of uncomplexed Plm II. The obtained results helped to ra-
tionalize unexpected binding in two examples.[36,45] Encour-
aged by these results, we extended the simulation time from
Similar behavior is assumed for the S1’ pocket (Q194, N210,
I212, L292, F294, and I300), but here the overall RMSDs among
the snapshots are significantly higher (Figure 4b) and reach
ChemMedChem 2010, 5, 443 – 454
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
447

G. Klebe et al.
MED
the S2’ pocket does not directly
affect transitions of the other
sub-pockets.
Therefore,
the
movements of the particular
sub-pockets appear to be rather
independent of each other.
The amino acids N76–V78 are
known to interact with P2’ sub-
stituents. Their mobility is pro-
nounced, as they are involved in
closing and opening of the flap.
These movements are of particu-
lar interest with respect to the
search of a putative binding-
competent conformer to accom-
modate ligands with scaffold 3.
This is discussed in greater detail
in the next section. Residues in
this pocket close to the catalytic
dyad (G36–A38) show little struc-
tural mobility, whereas large de-
viations are observed for N39,
M75, I133, and Y192 (Table 2).
Side chain movements for N39
(Nc1 and Nc2) are especially pro-
nounced, which could be impor-
tant for ligand binding. Upon a
1808 rotation of the terminal
amide function, hydrogen bond
acceptor and donor functionali-
Figure 4. 2D RMSD plot using snapshots extracted every 20 ps. The mutual RMSD between two snapshots is rep-
resented by a relative color code shown at right. All values are given in ꢁ. Only residues of the S2 pocket are con-
sidered in a), of the S1’ pocket in b), of the S2’ pocket in c), and of the S1 pocket in d). For the RMSD measure-
ments, all side chain atoms were included.
ties
are
mutually
inter-
changed.[36] Similarly, no particu-
lar state is predominantly popu-
values of up to 4.0 ꢁ (red areas). The residues responsible for
Table 2. RMSD for residues of the S2’ pocket.[a]
this fluctuation are F294 (average RMSD with respect to the
reference crystal structure: 2.8ꢁ0.8 ꢁ) and L292 (2.4ꢁ0.6 ꢁ),
whereas all other residues in this pocket show less pronounced
changes. A maximal RMSD value of 5.8 ꢁ (with respect to the
reference crystal structure) was detected for F294. These re-
sults are consistent with the observations by Ersmark et al.[20,24]
describing a slight expansion of the S1’ pocket by side chain
rotations of F294 and L292 necessary for the accommodation
of bulky P1’ groups.[20,24]
Residue
Average
RMSD [ꢁ]
Standard
Deviation [ꢁ]
Maximum
RMSD [ꢁ]
Minimum
RMSD [ꢁ]
G36
S37
1.5
1.5
1.3
2.0
2.2
3.6
2.2
0.3
0.4
0.3
0.3
0.3
1.3
0.6
3.1
2.7
2.5
3.4
3.5
8.1
5.2
0.5
0.3
0.4
0.3
1.1
0.4
0.3
A38
N39
M75
I133
Y192
Two major distinct states of the S2’ pocket (G36–N39, M75–
V78, L131–I133, and Y192) can be discerned from its 2D RMSD
plot (Figure 4c). The distribution of clusters highly resembles
the findings for the entire binding pocket, indicating that the
overall parent conformations (Call1, Call2, and Call3) are mainly
determined by changes in the S2’ pocket. Interestingly, a
sudden and clear-cut transition from the first (CS2’1) to the
second state (CS2’2) is observed after 4.3 ns. The second confor-
mational family is maintained for about 4.4 ns (CS2’2), before a
new one-step transition back to geometries closely related to
CS2’1 occurs. A comparison of the 2D RMSD plots (Figure 4a–d)
indicates that a transition of the conformational states within
[a] RMSD for residues of the S2’ pocket with respect to the reference crys-
tal structure, averaged over the entire trajectory, with standard deviation
as well as maximum and minimum values; measurements were carried
out including all side chain atoms.
lated for Y192. This residue fluctuates between different states
with a maximal deviation of 5.2 ꢁ. Therefore, the hydroxy
group of Y192 could possibly be addressed by ligands to form
hydrogen bonds as observed for all pepstatin-A-like ligands.
An extensive flexibility of the S2’ pocket has also been de-
scribed by Ersmark et al. mainly based on side chain move-
448
www.chemmedchem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2010, 5, 443 – 454

Pyrrolidine Derivatives as Plasmepsin Inhibitors
ments of M75.[24] Analysis of the Mc2 angle along the trajectory
resulted in similar observations (Figure 5). The c2 angle exhibits
a high standard deviation of 458 from the mean value of 1188.
the conformational states sampled during the simulations
(Figure 7). Based on the 2D RMSD plot for the S2’ pocket (Fig-
ure 4c), four MD-generated snapshots are compared, showing
all residues of this pocket (G36–N39, M75–V78, L131–I133, and
Y192). The snapshots are representatives from three different
clusters: The first frame (1S2’) belongs to the cluster in the
lower left of Figure 4c (CS2’1), the second and third frames (2S2’
and 3S2’) originate from the center cluster (CS2’2), and the
fourth frame (4S2’) is extracted from the cluster in the upper
right (CS2’1).
The superposition of the first (1S2’) and second frame (2S2’)
shows that, with the exception of I133, no significant confor-
mational variations are given (Figure 7a). The mobility of I133
is not restricted to the side chain level, but includes backbone
movement as well. The Ca atom is shifted by 4.2 ꢁ, and the
side chain rotates outward from the S2’ pocket. As shown by a
comparison of the surface representations of the two frames,
I133 swings out and opens up a wider pocket in 2S2’ (Fig-
ure 7b,c), enabling the accommodation of ligands with more
bulky groups. The potent inhibitor 1.7 with a bulky 1-naphthyl
substituent predicted to point into the S2’ pocket supports
this finding. The transition from frame 2S2’ to frame 4S2’ is illus-
trated in Figure 7d. As in Figure 7a, the S2’ flanking residues
roughly remain at their positions, except for I133. A superposi-
tion of 1S2’, a representative of cluster 1 (CS2’1, Figure 4c lower
left) with 4S2’ from cluster 3 (CS2’1, Figure 4c upper right) clearly
shows that I133 adopts the same orientation in both cases,
and indeed both frames represent the same conformational
state (Figure 7e). Figure 7 f shows that residues found in two
snapshots extracted from the same cluster occur with closely
related geometry (both frames belong to cluster CS2’2, Fig-
ure 4c).
Figure 5. Fluctuation of the Mc2 angle in M75 in the MD simulation of Plm II,
with the corresponding value from the crystal structure shown as reference.
In addition to the conformation found in the crystal structure,
one particular conformation is highly populated and seems to
be energetically favorable. Similar observations are made for
the c3 angle: a high standard deviation of 658 from the aver-
age value of 548 caused by two highly populated alternative
conformations. In the crystal structure, the M75 side chain ex-
tends into the S2’ pocket. Upon rotation it oscillates toward
the interior of the enzyme, thereby opening and closing the
pocket.
However, despite this dynamic behavior of M75, residue
I133 plays the key role. The RMSD plot (Figure 6) shows an ele-
As Ersmark et al. collected data only up to 2 ns, whereas in
our simulation the first major conformational change occurs
after 4.3 ns, the results are in agreement that as long as the
sub-pockets remain in one conformational state, no significant
displacement of the backbone portion is observed during the
MD simulations.[24]
The 2D RMSD plot for the S1 pocket (M15, I32, S79, F111,
T114, F120, and I123) is shown in Figure 4d. In contrast to the
2D RMSD plots for the S2’ pocket and the entire binding site,
the borderlines between two conformational states are less
pronounced. The first occurring conformational family is main-
tained for 1.4 ns (CS11/2). After the appearance of the cluster
CS11 for 2.8 ns, a conformational flip back to CS11/2 is observed
for 1 ns. This is followed by a transition to CS12. This conforma-
tional state is retained for 4 ns. While the clusters CS11 and CS12
are distinct from each other, CS11/2 exhibits similarities to both
CS11 and CS12 and is therefore considered as an intermediate
conformational state. While F120, I32, and I123 appear rather
rigid during the MD simulation, residues M15, S79, F111, and
T114 show elevated flexibility (Table 3).
Figure 6. All-atom RMSD of I133 from the reference crystal structure as a
function of simulation time.
vated average RMSD value of 3.6 ꢁ with a high standard devia-
tion (1.3 ꢁ) and a maximal deviation of 8.1 ꢁ. During the first
4.3 ns the average RMSD is <3 ꢁ with low standard deviations.
For the next 4.4 ns the RMSD increases to 5 ꢁ with high stan-
dard deviations, and returns back to 3 ꢁ during the last 1.3 ns.
These time steps are identical to those observed in the overall
2D RMSD plot (Figure 3), indicating that this residue is respon-
sible for the distinct conformational states in the S2’ pocket.
RMSD fluctuations should not be over-interpreted and must
be complemented by qualitative visual inspections. A compari-
son of frames extracted from each cluster provides insight into
In most of the available crystal structures the hydroxy group
of S79 in the flap region is involved in the formation of crucial
hydrogen bonds to a bound ligand. Residue S79 experiences
an average RMSD of 2.4 ꢁ with a standard deviation of 0.7 ꢁ,
resulting in maximal values of ~5 ꢁ. Similarly, the orientation
ChemMedChem 2010, 5, 443 – 454
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
449

G. Klebe et al.
MED
Figure 7. Fluctuation of the S2’ pocket. a) Superposition of the frames 1S2’ (blue) and 2S2’ (salmon) extracted from cluster CS2’1 and CS2’2 (Figure 4c). b) Surface
representation of 1S2’ in blue. c) Surface representation of 2S2’ in salmon. d) Superposition of the frames 3S2’ (green) and 4S2’ (gray). e) Superposition of the
frames 1S2’ (blue) and 4S2’ (gray). f) Superposition of the frames 2S2’ (salmon) and 3S2’ (green).
2D RMSD plots with different distinct states exhibiting clear
Table 3. RMSD for residues of the S1 pocket.[a]
borderlines (Figure 8b,c).
The flexibility of M15 was recently addressed by us.[45] Fluc-
tuations in the Mc1 and Mc2 angles result in an expansion of
the S1 pocket. As a consequence, the pocket is capable of ac-
commodating norstatine derivatives with large hydrophobic
side chains.
Residue
Average
RMSD [ꢁ]
Standard
Deviation [ꢁ]
Maximum
RMSD [ꢁ]
Minimum
RMSD [ꢁ]
M15
I32
S79
F111
T114
F120
I123
2.1
1.1
2.4
1.7
1.9
1.0
0.8
0.5
0.2
0.7
0.6
0.5
0.3
0.2
3.8
2.2
5.1
3.8
4.3
2.4
2.1
0.7
0.7
0.6
0.3
0.4
0.2
0.2
To analyze the conformations of F111 in greater detail, repre-
sentative frames extracted from the CS1_M1 and CS1_M2 cluster
(Figure 8b) were selected and visually inspected. The superpo-
sition of both frames shows that movements of S79, M15, and
F111 cause the main structural differences in the S1 pocket
(Figure 8d). Primarily responsible for the occurrence of the dis-
tinct sub-states in Figure 8b is the rotation around the Fc2
angle of F111.
[a] RMSD for residues of the S1 pocket with respect to the reference crys-
tal structure, averaged over the entire trajectory, with standard deviation
as well as maximum and minimum values; measurements were carried
out including all side chain atoms.
Putative binding mode generation for 3,4-bis(amino-
methylene)pyrrolidines
of the hydroxy group, described by the Sc1 angle, is involved
in the pronounced flexibility. This observation might be impor-
tant for finding a putative binding mode for the 3,4-bis(amino-
methylene)pyrrolidine derivatives. No highly populated pre-
ferred state is observed for T114, indicating that this residue
fluctuates over multiple states.
As discussed above, a conformational opening of the flap is re-
quired to establish key interactions to the 3,4-bis(aminomethy-
lene)pyrrolidines. The involved residues N76–S79 at the tip of
the flap region indeed show pronounced flexibility during the
simulations (Table 4).
M15 and F111 are responsible for the occurrence of distinct
states of the S1 pocket, as illustrated in Figure 8a–c. Discarding
these two crucial residues for the generation of the 2D RMSD
plot results in one preferred conformer of the S1 pocket de-
scribed by the remaining residues (Figure 8a). Discarding
either one of the flexible residues from the analysis produces
The backbone atoms of the flap region move ~5 ꢁ in total
relative to the crystal structure. Snapshots along the trajectory
that show shifts of ~2 ꢁ should be suited to reasonably ac-
commodate scaffold 3. Because this is well within the range of
the observed average RMSDs, a MD-generated snapshot could
450
www.chemmedchem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2010, 5, 443 – 454

Pyrrolidine Derivatives as Plasmepsin Inhibitors
optimized using MOLOC, assign-
ing full flexibility of the ligand
and keeping the protein rigid.
The phthalimide substituent was
placed into the S2’ pocket, sug-
gesting hydrogen bond forma-
tion to S37. The benzyl moiety is
placed into the S1’ pocket, the
2-bromo-4,5-dimethoxyphenyl
substituent into the S2 pocket,
and the 2,4-dichlorophenyl ring
occupies the S1 pocket. For 3.1
an equivalent binding mode
could be generated, pointing its
unsubstituted phenyl substitu-
ent into the S2’ pocket.
Relevance and implications for
inhibitor design
Plms are difficult targets for
structure-based ligand design
due to their high structural flexi-
bility. As a consequence, routine
drug-design protocols based on
a single rigid protein conformer
of the binding site are insuffi-
cient to successfully predict rea-
sonable binding modes. It is
likely that the originally selected
binding pocket would only be
suitable to accommodate a limit-
ed fraction of the ligands tested
by virtual screening. As long as
Figure 8. 2D RMSD plot considering residues of the S1 pocket discarding: a) M15 and F111, b) M15 only (two
main conformational sub-states can be distinguished that originate from fluctuations of F111), and c) F111 only
(the observed conformational sub-states originate from fluctuations of M15). d) A frame taken from the cluster
CS1_M1 (red) is superimposed with a representative member from cluster CS1_M2. The Fc2 angle of F111 remains
nearly unchanged for the first 5 ns, but then shows a rotation of about 908. This conformation is retained for the
last 4 ns. S79 also shows elevated flexibility, but no preferentially populated
state is observed, indicating that this residue fluctuates over multiple states.
For the RMSD measurements, all side chain atoms were included.
be easily identified that fulfills the requirements in order to es-
tablish key interactions to 3. Ligand 3.2 was manually placed
between the catalytic aspartates of the selected snapshot,
keeping the ring geometry as observed in the HIV-1 protease
crystal structure (PDB ID: 1XL2; Figure 9). The amide oxygen
atoms were rotated toward V78 and S79. The geometry was
Table 4. RMSD for residues comprising the flap.[a]
Residue
Average
RMSD [ꢁ]
Standard
Deviation [ꢁ]
Maximum
RMSD [ꢁ]
Minimum
RMSD [ꢁ]
N76
Y77
V78
S79
3.0
1.7
2.7
2.4
0.6
0.4
0.8
0.7
5.7
3.3
5.6
5.1
1.2
0.4
0.5
0.6
[a] RMSD with respect to the reference crystal structure for residues com-
prising the flap region, averaged over the entire trajectory, with standard
deviation as well as maximum and minimum values; RMSD measure-
ments were carried out including all side chain atoms.
Figure 9. Manual placement of the pyrrolidine scaffold 3 (inhibitor 3.2) in a
MD-generated snapshot exhibiting an open conformation of the flap. After
geometry minimization with MOLOC, hydrogen bonds to D34, D214, V78,
S79, and S37 were formed (dashed yellow line).
ChemMedChem 2010, 5, 443 – 454
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
451

G. Klebe et al.
MED
full flexibility of the protein structure cannot be handled simul-
taneously or in an adaptive manner during docking in the vir-
tual screening campaign, the currently favored strategy we
propose is to use a limited set of representative protein con-
formers to capture all possible alternatives in a comprehensive
way. This set of parent conformers may be derived from either
multiple experimental crystal structures or individual frames
from MD simulations.
uncomplexed state was carried out, and transitions along the
trajectory were monitored. Supported by the MD simulation, a
putative binding mode for the inhibitors with a 3,4-bis(amino-
methylene)pyrrolidine scaffold 3 could be suggested. This pro-
vides further insight into binding pocket adaptations that pos-
sibly have implications for Plm II inhibitor design. In particular,
the observed opening of the S1 and S2’ pockets suggests the
design of inhibitors with extended P1 and P2’ residues. Fur-
thermore, the search for new core scaffolds with different
linker lengths can be of interest based on the observed flap
flexibility.
With respect to crystal structures, the available data un-
doubtedly display relevant conformations. In our design they
served as a set of reference points generating reasonable bind-
ing modes for derivatives with 1 and 2 as core structures.
However, because the presently determined crystal structures
represent only three major distinct conformational states, it
seems unlikely that they sufficiently cover the broad spectrum
of potentially binding-competent protein conformers. This fact
became evident for inhibitors with 3 as the scaffold.
In this study, highly flexible residues could be identified by
MD simulation through analysis of residue mobility, and dis-
tinct conformers were identified. Based on the observed con-
formational opening of the flap during the MD simulation, a
putative binding mode for the 3,4-bis(aminomethylene)pyrroli-
dine derivatives (3.1 and 3.2) was suggested. In general, the
flexibility of the flap residues indicates that different linker
lengths are well tolerated for inhibitors that are designed to di-
rectly address both the aspartates of the catalytic dyad and
the flap residues. Furthermore, simulations of the uncom-
plexed state helped to understand surprising findings in two
cases.[36,45] The flexibility of M15 is responsible for the confor-
mational opening of the S1 pocket. Therefore, elongated P1
substituents might be suitable to increase ligand affinity. The
opening of the S2’ pocket could be observed, allowing to ac-
commodate bulky ligand side chains. This hypothesis is sup-
ported by the potent inhibitor 1.7 equipped with a bulky 1-
naphthyl substituent directed into the S2’ pocket. Hence, new
ligands designed to explore the available space and properties
of the S2’ pocket in greater detail could possibly lead to more
potent inhibitors.
Experimental Section
Assays and Ki determinations
The substrate used for the Plm II assay (Bachem) is a synthetic pep-
tide [Abz-Thr-Ile-Nle-(p-nitro-Phe)-Gln-Arg-NH2]. The Michaelis–
Menten constant (KM) for Plm II is 63 mm. The assays were per-
formed with a Tecan Spectra Fluor spectrometer at excitation
wavelength 337 nm and emission wavelength 414 nm. A volume
of 180 mL assay buffer (0.1m acetic acid/sodium acetate pH 4.5,
20 mm substrate, 2% DMSO) was added to 96-well plates. For the
assay 18 mL of enzyme solution (0.1m acetic acid/sodium acetate
buffer, pH 4.5) were mixed with 2 mL inhibitor solution (dissolved in
DMSO). After incubation for 5 min the enzyme–inhibitor solution
was added to the 180 mL substrate solution (final enzyme concen-
tration 1 nm). Substrate hydrolysis was recorded as an increase in
fluorescence intensity over a period of 3 min.
IC50 values were converted into Ki values by applying the following
equation:
Ki ¼ ½IC50ꢀðEt=2Þꢂ½1 þ ðS=KMÞꢂꢀ1
in which Et is the total enzyme concentration (1 nm), KM is the Mi-
chaelis–Menten constant (63 mm), and S is the substrate concentra-
tion (18 mm). Plm IV activity assays were performed similarly as de-
scribed for Plm II. The enzyme concentration was 10 nm, KM =
28 mm, and S=18 mm. Cathepsin D activity assays were essentially
performed as described for Plm II. The enzyme concentration was
1 nm, KM =16 mm, and S=9 mm.
Conclusions
MD simulations
In summary, we have discovered novel nonpeptidic inhibitors
of Plm II and IV, featuring a pyrrolidine scaffold as core ele-
ment. These compounds, originally designed for HIV-1 pro-
tease, show activity in the nanomolar range. Plausible binding
modes could be suggested for the pyrrolidine-3,4-diester deriv-
atives 1.1–1.7 and the 3,4-diaminopyrrolidine 2, based on the
Plm crystal structure (PDB ID: 1LEE) as a starting point. Putative
binding modes for Plm II are in agreement with structure–ac-
tivity relationships. Unfortunately, with the available crystallo-
graphic information it was not possible to generate a reasona-
ble binding mode for the 3,4-bis(aminomethylene)pyrrolidine
derivatives 3.1 and 3.2. None of the three major binding-site
conformers observed in the presently available crystal struc-
tures allow accommodation of the ligands. However, the struc-
tural studies indicate that Plm II is a highly flexible protein. To
tackle this problem in greater detail, a MD simulation of the
The MD simulation and all setup steps were performed as previ-
ously described[36,45] using the Amber 8.0 suite of programs[46] and
the Amber 1999 force field. The plasmepsin structure (PDB ID:
1LEE) was used as starting point. After removal of the ligand and
all crystallographic water molecules, hydrogens were added with
PROTONATE. Protonation states were estimated with Poisson–
Boltzmann calculations for all histidines and the aspartates of the
catalytic dyad (D34 and D214) at pH 5: Both catalytic aspartates
were assumed to be deprotonated, and all histidines were set to
the protonated form. Subsequently, 200 steps of minimization
were applied to the protein using a generalized Born solvation
model. After addition of sodium counterions to ensure neutrality,
the system was solvated in a box containing ~10300 TIP3P water
molecules.[47] The solvated system was subjected to 200 steps of
minimization.
The MD simulation was then started by heating the solvent to
300 K over a period of 20 ps and cooling to 100 K over a period of
452
www.chemmedchem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2010, 5, 443 – 454

Pyrrolidine Derivatives as Plasmepsin Inhibitors
5 ps, keeping the protein fixed. After this, the entire system was
gradually brought to 300 K over a period of 25 ps. The simulation
was carried out for 10 ns under constant temperature and pressure
(NPT) applying periodic boundary conditions. A time step of 2 fs
and PME (Particle–Mesh–Ewald)[48] for evaluating the electrostatic
interactions were used. Energy data were saved every 20 fs, and
protein coordinates every 0.5 ps. CARNAL was used for further
analysis of the trajectory, and VMD 1.8.2 for visualization.
EtOAc 3:7) yielded 0.58 g (59%) of the bis-benzylated pyrrolidine 5
as a pale-yellow oil: 1H NMR (400 MHz, CDCl3, 21.08C), rotamers:
d=7.36–7.22 (m, 10H), 3.87–3.58 (m, 6H), 3.14–3.00 (m, 4H), 1.51
(s, 2H), 1.46 ppm (s, 9H); 13C NMR (100 MHz, CDCl3, 21.08C), rotam-
ers: d=28.6, 50.4, 51.0, 52.4, 61.2, 62.1, 79.4, 127.2, 128.2, 128.6,
140.1, 154.7 ppm; MS (ESI) m/z (%): 785 (24) [2M+Na]+, 763 (45)
[2M+H]+, 382 (100) [M+H]+; HRMS-ESI m/z [M+H]+ calcd for
C23H32N3O2: 382.2495, found: 382.2507; Anal. calcd for
C23H31N3O2·0.5H2O: C 70.74, H 8.26, N 10.76, found: C 70.71, H 8.04,
N 10.60.
The Ca atom RMSD with respect to the reference crystal structure
was 2.1ꢁ0.3 ꢁ averaged over the trajectory, indicating a high sta-
bility of the overall fold (Figure 10). Considering all atoms, the tra-
jectory-averaged RMSD value is 2.7ꢁ0.3 ꢁ, indicating more pro-
nounced motions once the side chains are included in the analysis.
Bis-amide 6: Bis-benzylated pyrrolidine 5 (0.33 g, 0.86 mmol), 2-
phenylacetic acid (0.29 g, 2.15 mmol), and 1-hydroxybenzotriazole
(0.29 g, 2.15 mmol) were dissolved in DMF (5 mL) and cooled to
08C. Et3N (0.45 mL, 3.23 mmol) was added, and the solution was
stirred at 08C for 5 min. Subsequently, 1-ethyl-3-(3’-dimethylamino-
propyl)carbodiimide hydrochloride (0.41 g, 2.15 mmol) was added,
and the mixture was stirred at 08C for 10 min. After allowing the
reaction mixture to reach room temperature, stirring was contin-
ued for 13 h. The reaction was quenched by the addition of H2O
(10 mL), followed by EtOAc (10 mL). The aqueous layer was extract-
ed with EtOAc (3ꢂ10 mL). The combined organic layers were
washed with brine (2ꢂ10 mL), dried over MgSO4, filtered, and the
solvent was removed under reduced pressure. Purification via FC
(hexanes/EtOAc 7:3) yielded 0.47 g (89%) of the bis-amide 6 as col-
1
orless crystals: mp: 788C; H NMR (500 MHz, [D6]DMSO, 1008C), ro-
tamers: d=7.39–7.11 (m, 20H), 4.92 (brs 2H), 4.48 (psd, 4H), 3.74–
3.53 (m, 4H), 3.30 (s, 2H), 3.16 (brs, 2H), 1.32 ppm (s, 9H); 13C NMR
(125.7 MHz, [D6]DMSO, 1008C), rotamers: d=27.4, 27.5, 27.6, 27.7,
45.0, 49.9, 56.6, 78.1, 125.9, 126.5, 127.6, 127.7, 127.9, 128.6, 134.9,
137.6, 152.8, 171.1 ppm; MS (ESI) m/z (%): 1257 (100) [2M+Na]+,
640 (63) [M+Na]+, 618 (12) [M+H]+; HRMS-ESI m/z [M+Na]+ calcd
for C39H43N3O4Na: 640.3151, found: 640.3157; Anal. calcd for
C39H43N3O4·0.5H2O: C 74.73, H 7.08, N 6.70, found: C 74.95, H 7.08,
N 6.81.
Figure 10. Ca atom RMSD in the MD simulation of Plm II.
Synthesis
General: Reported yields refer to the analytically pure product ob-
tained by column chromatography. All proton and carbon NMR
spectra were recorded on a Jeol Eclipse+ spectrometer (1H and
1
13C NMR: 400 or 500 MHz as indicated). H NMR spectra were refer-
enced to CDCl3 (d=7.26 ppm) or [D6]DMSO (d=2.50 ppm).
13C NMR spectra were referenced to CDCl3 (d=77.16 ppm) or
[D6]DMSO (d=39.52 ppm). Chemical shift values (d) are given in
ppm and coupling constants (J) are given in Hz. Abbreviations:
br=broad, ps=pseudo, s=singlet, d=doublet, t=triplet, q=
quartet, smul=symmetric multiplet, m=multiplet. Mass spectra
were obtained from a double focusing sector field Micromass VG-
Autospec spectrometer. Combustion analyses were determined on
a Vario Micro Cube by Elementar Analysen GmbH. Melting points
were determined using a Leitz HM-Lux apparatus and are uncor-
rected. Flash chromatography (FC) was performed using silica
gel 60 (0.04–0.063 mm) purchased from Macherey & Nagel. Sol-
vents and reagents that are commercially available were used with-
out further purification. All moisture-sensitive reactions were car-
ried out using oven-dried glassware under a positive stream of
argon.
Scheme 1. Synthesis of inhibitor 2. Reagents and conditions: a) MeOH, pow-
dered molecular sieves (4 ꢁ), benzaldehyde, NaBH4, 2 h, 59%; b) 2-phenyl-
acetic acid, 1-hydroxybenzotriazole, DMF, EDCI, 13 h, 89%; c) Et2O, HCl in
Et2O, 24 h, 91%.
Bis-benzylated pyrrolidine 5: Powdered molecular sieves (4 ꢁ,
0.65 g) and benzaldehyde (0.79 mL, 7.8 mmol) were added to a so-
lution of tert-butyl-(3S,4S)-3,4-diaminopyrrolidine-1-carboxylate[49]
(0.52 g, 2.6 mmol) in dry MeOH at room temperature. The suspen-
sion was stirred for 1 h under Ar. Subsequently, NaBH4 (0.40 g,
10.4 mmol) was added portionwise at 08C. After stirring for 1 h at
08C, EtOAc (15 mL) and a saturated solution of NaHCO3 (20 mL)
were added. The suspension was filtered, and the filtrate was ex-
tracted with EtOAc (3ꢂ20 mL), dried over MgSO4, filtered, and con-
centrated under reduced pressure. Purification via FC (hexanes/
Pyrrolidine 2 (Scheme 1): HCl in Et2O (5 mL, 2m) was added to bis-
amide 6 (0.11 g, 0.18 mmol) in dry Et2O (5 mL) at room tempera-
ture, and the solution was stirred for 24 h under Ar. The solvent
was subsequently removed under reduced pressure. Purification
was carried out by FC (CH2Cl2/MeOH 9:1) to yield 0.17 g (91%) of
pyrrolidine
2 as a
pale-yellow powder: mp: 438C; 1H NMR
(500 MHz, [D6]DMSO, 1008C), rotamers: d=7.39–7.10 (m, 20H),
5.67 (s, 1H), 4.52 (psd 6H), 3.70–3.48 (m, 4H), 2.89 (brs, 2H),
2.74 ppm (brs, 2H); 13C NMR (125.7 MHz, [D6]DMSO, 1008C), rotam-
ChemMedChem 2010, 5, 443 – 454
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
453

G. Klebe et al.
MED
[20] K. Ersmark, I. Feierberg, S. Bjelic, J. Hultꢄn, B. Samuelsson, J. ꢁqvist, A.
Hallberg, Bioorg. Med. Chem. 2003, 11, 3723.
[21] K. T. Andrews, D. P. Fairlie, P. K. Madala, J. Ray, D. M. Wyatt, P. M. Hilton,
L. A. Melville, L. Beattie, D. L. Gardiner, R. C. Reid, M. J. Stoermer, T. Skin-
ner-Adams, C. Berry, J. S. McCarthy, Antimicrob. Agents Chemother. 2006,
50, 639.
ers: d=39.9, 48.0, 50.3, 60.4, 125.8, 126.0, 126.5, 127.6, 127.9,
128.5, 135.0, 137.9, 170.8 ppm; MS (ESI) m/z (%): 518 (100) [M+H]+;
HRMS-ESI m/z [M+H]+ calcd for C34H36N3O2: 518.2808, found:
518.2838; Anal. calcd for C34H35N3O2·0.5H2O: C 77.54, H 6.89, N
7.98, found: C 76.99, H 6.94, N 7.96.
[22] S. Bjelic, M. Nervall, H. Gutiꢄrrez-de-Terꢅn, K. Ersmark, A. Hallberg, J.
ꢁqvist, Cell. Mol. Life. Sci. 2007, 17, 2285.
[23] S. K. Pranav Kumar, V. M. Kulkarni, Drug Des. Discovery 2001, 17, 293.
[24] K. Ersmark, M. Nervall, E. Hamelink, L. K. Janka, J. C. Clemente, B. M.
Dunn, M. J. Blackman, B. Samuelsson, J. ꢁqvist, A. Hallberg, J. Med.
Chem. 2005, 48, 6090.
[25] H. Gutiꢄrrez-de-Terꢅn, M. Nervall, B. M. Dunn, J. C. Clemente, J. ꢁqvist,
FEBS Lett. 2006, 580, 5910.
[26] H. Gutiꢄrrez-de-Terꢅn, M. Nervall, K. Ersmark, P. Liu, L. K. Janka, B. Dunn,
A. Hallberg, J. ꢁqvist, Biochemistry 2006, 45, 10529.
[27] K. Ersmark, M. Nervall, H. Gutiꢄrrez-de-Terꢅn, E. Hamelink, L. K. Janka,
J. C. Clemente, B. M. Dunn, A. Gogoll, B. Samuelsson, J. ꢁqvist, A. Hall-
berg, Bioorg. Med. Chem. 2006, 14, 2197.
[28] K. Ersmark, I. Feierberg, S. Bjelic, E. Hamelink, F. Hackett, M. J. Blackman,
J. Hultꢄn, B. Samuelsson, J. ꢁqvist, A. Hallberg, J. Med. Chem. 2004, 47,
110.
Acknowledgements
We gratefully acknowledge financial support of our research by
the DFG-Graduiertenkolleg GRK 541 “Proteinfunktion auf atomar-
er Ebene” (T.L.), the joint DFG research project KL1204/5-3 and
DI827/3-2, and the Studienstiftung des deutschen Volkes (N.K.).
The authors are grateful to Serghei Glinca for technical help. The
clone of the protease was kindly provided by Prof. Dr. Daniel E.
Goldberg, Department of Molecular Microbiology, Washington
University School of Medicine, St. Louis, MO (USA).
Keywords: induced-fit adaptations · ligand design · molecular
dynamics · plasmepsins · pyrrolidine derivatives
[29] R. Friedman, A. Caflisch, Proteins Struct. Funct. Bioinf. 2008, 73, 814.
[30] P. A. Valiente, P. R. Batista, A. Pupo, T. Pons, A. Valencia, P. G. Pascutti,
Proteins Struct. Funct. Bioinf. 2008, 73, 440.
[31] R. Friedman, A. Caflisch, ChemMedChem 2009, 4, 1317.
[32] I. D. Kuntz, K. Chen, K. A. Sharp, P. A. Kollman, Proc. Natl. Acad. Sci. USA
1999, 96, 9997.
[33] A. L. Hopkins, C. R. Groom, A. Alex, Drug Discovery Today 2004, 9, 430.
[34] C. H. Reynolds, S. D. Bembenek, B. A. Tounge, Bioorg. Med. Chem. Lett.
2007, 17, 4258.
[1] World Health Organization (WHO), The World Health Report 2004,
http://www.who.int/whr/2004/en/ (accessed January 4, 2010).
[2] G. E. Linares, J. B. Rodriguez, Curr. Med. Chem. 2007, 14, 289.
[3] Antimalarial Chemotherapy: Mechanisms of Action, Resistance, and New
Directions in Drug Discovery (Ed.: P. J. Rosenthal), Humana Press, Totowa,
2001.
[35] C. H. Reynolds, B. A. Tounge, S. D. Bembenek, J. Med. Chem. 2008, 51,
2432.
[4] I. Y. Gluzman, S. E. Francis, A. Oksman, C. E. Smith, K. L. Duffin, D. E.
Goldberg, J. Clin. Invest. 1994, 93, 1602.
[36] T. Luksch, N.-S. Chan, S. Brass, C. A. Sotriffer, G. Klebe, W. E. Diederich,
ChemMedChem 2008, 3, 1323.
[5] K. K. Eggleson, K. L. Duffin, D. E. Goldberg, J. Biol. Chem. 1999, 274,
32411.
[37] A. Gustchina, I. T. Weber, FEBS Lett. 1990, 269, 269.
[38] A. M. Silva, A. Y. Lee, S. V. Gulnik, P. Majer, J. Collins, T. N. Bhat, P. J. Col-
lins, R. E. Cachau, K. E. Luker, I. Y. Gluzman, S. E. Francis, A. Oksman, D. E.
Goldberg, J. W. Erickson, Proc. Natl. Acad. Sci. USA 1996, 93, 10034.
[39] O. A. Asojo, E. Afonina, S. V. Gulnik, B. Yu, J. W. Erickson, R. Randad, D.
Medjahed, A. M. Silva, Acta Crystallogr. Sect. A 2002, 58, 2001.
[40] O. A. Asojo, S. V. Gulnik, E. Afonina, B. Yu, J. A. Ellman, T. S. Haque, A. M.
Silva, J. Mol. Biol. 2003, 327, 173.
[6] J. A. Bonilla, T. D. Bonilla, C. A. Yowell, H. Fujioka, J. B. Dame, Mol. Micro-
biol. 2007, 65, 64.
[7] G. H. Coombs, D. E. Goldberg, M. Klemba, C. Berry, J. Kay, J. C. Mottram,
Trends Parasitol. 2001, 17, 532.
[8] S. E. Francis, I. Y. Gluzman, A. Oksman, A. Knickerbocker, R. Mueller, M. L.
Bryant, D. R. Sherman, D. G. Russell, D. E. Goldberg, EMBO J. 1994, 13,
306.
[9] R. P. Moon, L. Tyas, U. Certa, K. Rupp, D. Bur, C. Jacquet, H. Matile, H.
Loetscher, F. Grueninger-Leitch, J. Kay, B. M. Dunn, C. Berry, R. G. Ridley,
Eur. J. Biochem. 1997, 244, 552.
[41] L. Prade, A. F. Jones, C. Boss, S. Richard-Bildstein, S. Meyer, C. Binkert, D.
Bur, J. Biol. Chem. 2005, 280, 23837.
[42] S. Cabani, G. Conti, L. Lepori, Trans. Faraday Soc. 1971, 67, 1933.
[43] P. Czodrowski, C. A. Sotriffer, G. Klebe, J. Chem. Inf. Comput. Sci. 2007,
47, 1590.
[10] K. Ersmark, B. Samuelsson, A. Hallberg, Med. Res. Rev. 2006, 26, 626.
[11] D. E. Goldberg, A. F. G. Slater, R. Beavis, B. Chait, A. Cerami, G. B. Hender-
son, J. Exp. Med. 1991, 173, 961.
[44] P. R. Gerber, K. Mꢃller, J. Comput. Aided Mol. Des. 1995, 9, 251.
[45] S. Weik, T. Luksch, A. Evers, J. Bçttcher, C. A. Sotriffer, A. Hasilik, H.-G.
Lçffler, G. Klebe, J. Rademann, ChemMedChem 2006, 1, 445.
[46] D. A. Case, T. E. Cheatham, T. Darden, H. Gohlke, R. Luo, K. M. Merz, A.
Onufriev, C. Simmerling, B. Wang, R. J. Woods, J. Comput. Chem. 2005,
26, 1668.
[12] J. B. Dame, G. R. Reddy, C. A. Yowell, B. M. Dunn, J. Kay, C. Berry, Mol.
Biochem. Parasitol. 1994, 64, 177.
[13] M. J. Humphreys, R. P. Moon, A. Klinder, S. D. Fowler, K. Rupp, D. Bur,
R. G. Ridley, C. Berry, FEBS Lett. 1999, 463, 43.
[14] A. L. Omara-Opyene, P. A. Moura, C. R. Sulsona, J. A. Bonilla, C. A. Yowell,
H. Fujioka, D. A. Fidock, J. B. Dame, J. Biol. Chem. 2004, 279, 54088.
[15] J. Liu, I. Y. Gluzman, M. E. Drew, D. E. Goldberg, J. Biol. Chem. 2004, 280,
1432.
[47] W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, M. L. Klein,
J. Chem. Phys. 1983, 79, 926.
[48] T. Darden, D. York, L. Pedersen, J. Chem. Phys. 1993, 98, 10089.
[49] A. Blum, J. Bçttcher, A. Heine, G. Klebe, W. E. Diederich, J. Med. Chem.
2008, 51, 2078.
[16] E. Specker, J. Bçttcher, H. Lilie, A. Heine, A. Schoop, G. Mꢃller, N. Griebe-
now, G. Klebe, Angew. Chem. 2005, 117, 3200; Angew. Chem. Int. Ed.
2005, 44, 3140.
[17] E. Specker, J. Bçttcher, S. Brass, A. Heine, H. Lilie, A. Schoop, G. Mꢃller,
N. Griebenow, G. Klebe, ChemMedChem 2006, 1, 106.
[18] J. Bçttcher, A. Blum, S. Dçrr, A. Heine, W. E. Diederich, G. Klebe, Chem-
MedChem 2008, 3, 1337.
[19] A. Nezami, I. Luque, T. Kimura, Y. Kiso, E. Freire, Biochemistry 2002, 41,
2273.
Received: November 1, 2009
Revised: December 18, 2009
Published online on January 28, 2010
454
www.chemmedchem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2010, 5, 443 – 454