Article

A toolkit and benchmark study for FRET-restrained high-precision structural modeling

Lehrstuhl für Molekulare Physikalische Chemie, Heinrich-Heine-Universität (HHU), Düsseldorf, Germany.
Nature Methods (Impact Factor: 32.07). 11/2012; 9(12). DOI: 10.1038/nmeth.2222
Source: PubMed
ABSTRACT
We present a comprehensive toolkit for Förster resonance energy transfer (FRET)-restrained modeling of biomolecules and their complexes for quantitative applications in structural biology. A dramatic improvement in the precision of FRET-derived structures is achieved by explicitly considering spatial distributions of dye positions, which greatly reduces uncertainties due to flexible dye linkers. The precision and confidence levels of the models are calculated by rigorous error estimation. The accuracy of this approach is demonstrated by docking a DNA primer-template to HIV-1 reverse transcriptase. The derived model agrees with the known X-ray structure with an r.m.s. deviation of 0.5 Å. Furthermore, we introduce FRET-guided 'screening' of a large structural ensemble created by molecular dynamics simulations. We used this hybrid approach to determine the formerly unknown configuration of the flexible single-strand template overhang.

Full-text

Available from: Paul J Rothwell, Jan 29, 2014
Understanding the Inhibitory Eect of Highly Potent and Selective
Archazolides Binding to the Vacuolar ATPase
Sandra Dreisigacker,
,
Dorota Latek,
,§
Svenja Bockelmann,
Markus Huss,
Helmut Wieczorek,
Slawomir Filipek,
Holger Gohlke,
#
Dirk Menche,
,
and Teresa Carlomagno*
,
Structural and Computational Biology Unit, EMBL, Mayerhofstrasse 1, D-69117 Heidelberg, Germany
Institute of Organic Chemistry, Ruprecht-Karls University Heidelberg, Im Neuenheimer Feld 270, D-69120 Heidelberg, Germany
Department of Animal Physiology, Faculty of Biology and Chemistry, University of Osnabru
̈
ck, Barbarastrasse 11,
D-49069 Osnabru
̈
ck, Germany
Faculty of Chemistry, University of Warsaw, Pasteura 1, 02-093 Warsaw, Poland
#
Heinrich-Heine-University Du
̈
sseldorf, Institute of Pharmaceutical and Medicinal Chemistry, Universita
̈
tsstrasse 1,
D-40225 Du
̈
sseldorf, Germany
*
S
Supporting Information
ABSTRACT: Vacuolar ATPases are a potential therapeutic
target because of their involvement in a variety of severe
diseases such as osteoporosis or cancer. Archazolide A (1) and
related analogs have been previously identied as selective
inhibitors of V-ATPases with potency down to the sub-
nanomolar range. Herein we report on the determination of
the ligand binding mode by a combination of molecular dock-
ing, molecular dynamics simulations, and biochemical experi-
ments, resulting in a sound model for the inhibitory mech-
anism of this class of putative anticancer agents. The binding
site of archazolides was conrmed to be located in the equatorial region of the membrane-embedded V
O
-rotor, as recently
proposed on the basis of site-directed mutagenesis. Quantication of the bioactivity of a series of archazolide derivatives, together
with the docking-derived binding mode of archazolides to the V-ATPase, revealed favorable ligand proles, which can guide the
development of a simplied archazolide analog with potential therapeutic relevance.
INTRODUCTION
The pol yk e ti d e family of archazolides isolated from
myxobacteria Archangium gephyra and cystobacter violaceus
14
are well-established natural products. In the past years, we
reported on the structure elucidation and the rst total synthesis
of archazolide A (1) (Figure 1), and, more recently, on the
syntheses of analogs and their bioactivity.
511
Archazolides are
interesting molecules both from the structural point of view and
in terms of their potential utility as pharmaceutical drugs. They
display impressive biological properties, including potent
cytotoxicity against several tumor cell lines in the subnanomolar
range.
24,7
The cytotoxicity is attributed to the binding of
archazolides to the vacuolar ATPase (V-ATPase), which leads to
an inhibition of this vital enzyme. V-ATPases use the energy
from ATP hydrolysis to translocate protons across membranes,
thereby energizing secondary active transport processes or
acidifying either the extracellular medium or the lumen of
intracellular organells.
12
Cleavage of ATP occurs at the cytosolic
V
1
-complex, while proton transport is mediated by the
membrane-integral V
O
-complex of the enzyme. The V
O
-complex
comprises a ring of c subunits. Each subunit contains a conserved
glutamate residue that is essential for proton translocation.
13
V-ATPase mediated extracellular acidication contributes for
example to bone resorption by osteoclasts or to metastasis of
tumor cells. Accordingly, V-ATPases represent a potential target
for the treatment of severe diseases such as osteoporosis or
cancer.
14,15
The function of V-ATPase is eectively inhibited by
archazolides, which bind selectively to the c subunits of the
membrane-integral V
O
-rotor.
7
In contrast, other ion translocating
enzymes, e.g., F-type and P-type ATPases, are not aected by
these inhibitors at concentrations up to 10 μM,
7
which makes
archazolides highly interesting as prospective potent and selec-
tive therapeutic agents. Furthermore, archazolide F (3), with
inhibitory proles in the picomolar range, is the most active
inhibitor of V-ATPase-controlled cell growth discovered so
far (Figure 1) and also displays a 10-fold higher eciency for
human cell lines compared to its derivatives archazolide A (1)
and B (2).
4
The development of archazolides into new pharmaceuticals
requires both easy accessibility of the substance and highly
selective biological activity. Unfortunately, neither isolation
nor total synthesis of archazolides accomplishes the amount
of natural product that would be needed in a pharmaceutical
Received: May 23, 2012
Published: July 2, 2012
Article
pubs.acs.org/jcim
© 2012 American Chemical Society 2265 dx.doi.org/10.1021/ci300242d | J. Chem. Inf. Model. 2012, 52, 22652272
Page 1
context. This limitation calls for a simpler, easier-to-synthesize
analog that retains all essential interactions of archazolides with
the target protein. To this end, in this work we study the
binding mode of archazolides to the V-ATPase with the goal of
revealing the key structural features for the high activity of the
natural product.
In previous work, we performed radioactive photoanity
labeling of archazolide A (1) and proximate cross-linking
studies with puried V-ATPase to identify the c-subunit of
the rotor as target for archazolides.
11
Furthermore, we used
ligand competition analyses and site-directed mutagenesis of
the S. cerevisiae subunit c to characterize the binding site of
archazolides in more detail.
11
We discovered that archazolide A
(1) prevents the covalent binding of the inhibitor N,N-
dicyclohexylcarbodiimide (DCCD) to glutamate E137 within
subunit c
11,16
and the binding of the plecomacrolide inhibitors
at the interface between two adjacent c subunits.
7,17
Last but
not least, we showed that in yeast the amino acid exchanges
Y142N and L144I in subunit c lead to V-ATPases that bind
ligand 1 10-fold more eciently than the wild-type protein.
11
These results indicated that the binding site of archazolides
comprises, or is close to, amino acids E137, Y142, and L144
(amino acids positions according to yeast subunit c), located in
the equ atorial region of the rotor structure (Figure 2).
However, the exact location and shape of the archazolide binding
pocket remained unclear.
Here, we aim at describing the binding mode of archazolides
to V-ATPase by a combination of biochemical experiments and
docking calculations. First, we dock ligand 1 to a model of the
V
O
-ring formed by ten c subunits of S. cerevisiae to identify
possible binding pockets and binding modes of archazolides to
the rotor structure. Second, we select the pose that is best in
agreement with the mutational analysis conducted in ref 11 and
with the activity of archazolide analogs reported in refs 24.
Third, we rene the selected binding pose in a membrane-like
environment by molecular dynamics (MD) simulations in
order to obtain a more realistic picture of the intermolecular
interactions. Last, we verify the validity of our pose by mea-
suring the bioactivity of archazolide derivatives with a yeast
mutant V-ATPase, which conrmed the predictions made on
the basis of the docked model. The binding pose of archa-
zolides to the V-ATPase presented here reveals the features of
the intermolecular interactions that are responsible for the high
anity of the ligand. These results are critical for further eorts
of our and other laboratories aimed at designing a potent ligand
based on the archazolide scaold with an easily accessible syn-
thetic route.
RESULTS AND DISCUSSION
The protein input structure for the docking calculations was
prepared by homology modeling of the yeast c subunit using
the crystallographic structure of the E. hirae rotor as template.
11
The two proteins display 35% (85%) sequence identity
(similarity) in the equatorial region, which speaks for the validity
of the homology model for docking studies (Figure S1). The
amino acid sequence in the equatorial region is conserved also in
other V-ATPase rotor subunits (Figure S2); in this work we
choose to model the yeast c subunit to provide a direct com-
parison with the mutational analysis, which is performed on the
yeast protein. To accelerate the docking calculations only three
units of the decameric rotor were used (Figure S3). The whole
rotor structure is stable during 10 ns of MD simulations con-
ducted in an explicit membrane (Figure S4). To sample a larger
conformational space of the binding pocket side chains, 22
protein structures were generated in a 1.1 ns MD run in aque-
ous solution (see Experimental Procedures for details). The
(modest) variability of the side chains in the equatorial region of
the rotor is shown in Figure S5. All 22 structures were used in
ensemble docking calculations. The electrostatic potential of the
archazolide binding site is shown in Figure S6.
18
To minimize the uncertainties of the modeling results, we
assumed that the bioactive conformation of the macrocycle of 1
is similar to that determined for free archazolide in solution by
NMR (Figure S7).
6
This choice follows the rationale that the
bioactive conformation of small molecules is usually highly
populated in solution to minimize unfavorable energetic contri-
butions upon binding.
19,20
In contrast, the carbamate side
chain, the methoxy-, and the hydroxyl groups were left free to
adapt to the protein binding region during docking (Figure S8).
The initial structure of archazolide derivatives 14 were
obtained by substitution of the corresponding functional groups
in the solution structure of 1 followed by energy minimization.
Figure 1. Structures of natural archazolides A, B, F (13) and
synthetic 15-dehydro-archazolide (4) and 1-descarbamoyl-archazolide
(5). Inhibition of the V
1
/V
O
holoenzyme activity and IC
50
values on
growth inhibition in mouse cell line L-929 are shown in the table.
Journal of Chemical Information and Modeling Article
dx.doi.org/10.1021/ci300242d | J. Chem. Inf. Model. 2012, 52, 226522722266
Page 2
This choice is justied by the high bioactivity of all derivatives
(Figure 1), which would not be compatible with a major change
of the interaction mode.
Next, we docked archazolide analogs 14 to the homology
model of the yeast c-ring using the program Autodock3.0
21
together with Drugscore
22
as objective function (Figure S9 and
S10). The resulting docking modes for each archazolide
derivative to the 22 models of subunit c were clustered before
analysis. Docking of archazolide A (1) to the homology model
yielded ve dierent binding poses B1B5 (Figure S11). B2
represents the most populated cluster, followed by B1 and B5.
Both B1 and B5 display a twin pose, with the same binding
mode for the macrocycle, but a dierent orientation of the side
chain containing the thiazole ring; in the twin poses, the
interaction partners of the methyl-carbamate and the i-butyl
side chain are eectively switched. Of the ve possible binding
modes of 1 to the yeast c-ring, B1, B3, and B4 can be safely
excluded because in these poses the archazolide does not
display any contacts to either Y142 or L144, which were
identied to be part of the binding pocket by mutational
analysis.
11
Therefore, in the following analysis, we consider B2
and B5 only (Figure 3).
Binding Pose B2. In B2 the C17-OMe and the methyl
groups at C2, C12, and C16 are in close proximity to I134,
V138, Y142, L141, and L144, whereas the methyl groups at C5,
C10, C22, and the hydroxyl group at C7 face the solvent and
do not show any interactions with the protein surface. The
hydroxyl group at C15 interacts with the polar side chains of
E137 and Y66. In this position the ligand partially occupies the
DCCD binding pocket, which would explain why archazolide
binding prevents binding of DCCD (Figure S12A). The
thiazole side chain penetrates a hydrophobic cleft formed by
two adjacent subunits including residues T32, G61, I65, and
L139. This tight interaction is likely responsible for the
favorable binding energy of the complex. Notably, activity data
for archazolides with a G61S mutant protein indicated no eect
of this mutation on binding.
11
These data are thus contra-
dictory to B2, from which a longer amino acid side chain at
position 61 is predicted to partially obstruct the hydrophobic
cavity and therefore impact ligand binding. Docking calcula-
tions performed with the G61S mutant protein supported this
prediction, as the populations of the poses B1B3 drop
considerably upon substitution of G61 with a serine (Figure S13).
In addition, this pose does not easily recapitulate the poor bio-
activity of archazolide C,
2
with a glucose moiety at the C7, as the
hydroxyl group at this position is directed toward the solvent. All
in all, this pose can explain only a limited set of bioactivity data for
archazolide derivatives.
Binding Pose B5. In B5 the methyl groups at C2, C5, C10,
C12, and C16 are in close proximity to residues Y66, I134,
F135, E137, V138, and L141, providing favorable hydrophobic
interactions, whereas C22-Me is the only aliphatic group that is
not directed toward the protein. The C1C6 region of the
ligand entertains hydrophobic interaction with Y66, well in
agreement with the similar anity of the Y66F mutant protein
for archazolides as compared to the wild-type protein.
11
The
C7-OH forms a hydrogen bond to the backbone of I134, again
occupying the binding pocket of DCCD and thereby
preventing its binding (Figure S12B). In addition, this binding
pose is in good agreement with the low bioactivity of archa-
zolide C,
2
as a glucose moiety at C7 would clash with the
protein surface. The thiazole-bearing side chain contacts resi-
dues L141, L144, and I145, and, in contrast to pose B2, there is
no interaction with G61. The C15-OH is close to residue Y142.
If this residue is mutated to asparagine, the C15-OH could
form a hydrogen bond with the carbonyl group, thus explain-
ing the increased ligand activity with the Y142N mutant pro-
tein. In conclusion, this pose agrees well with all activity data of
archazolides with both the wild type and the mutant proteins.
Thus, we propose that binding mode B5 (Figure 3) is the
one representing best the interactions of archazolide with
V-ATPase.
Docking Results for Derivatives 24. The docking of the
archazolides derivatives 24 to the yeast c-ring model yielded
similar results as for 1. The ve binding modes B1B5 are
displayed also by derivatives 24 (Figure S14S16), although
with dierent populations. In agreement with the high activity
of derivatives 2 and 3, cluster B5 is highly populated (Figure
S17), probably due to a favorable interaction of the C1C6
region with Y66. Thus, the relevance of the binding pose B5 is
conrmed by docking across the whole set of archazolide
derivatives.
Renement of Pose B5 by MD Simulations. To conrm
the reliability of pose B5, the complexes of archazolide A (1)
Figure 2. Left: Sphere representation of a decameric V
O
-ring formed of S. cerevisiae c subunits. The c subunits, consisting of 4 transmembrane helices
each, are shown in dierent colors. Residues involved in the binding of the V
O
-ring to archazolides, identied in previous studies, are colored in red
(E137), blue (Y142), and orange (L144), respectively. Right: Zoom of the protein region containing the critical residues, with E137, Y142, and L144
represented as spheres. Amino acids are labeled in one letter code.
Journal of Chemical Information and Modeling Article
dx.doi.org/10.1021/ci300242d | J. Chem. Inf. Model. 2012, 52, 226522722267
Page 3
with the wild-type and the Y142N mutant of V-ATPase,
corresponding to the binding pose B5, were rened during a
short 3 ns MD simulation in an explicit phosphatidylethanol-
amine (PEA) membrane using the AMBER03
23
force eld in
the YASARA
24
molecular modeling program. Eight dierent
sets of weak restraints were imposed in eight MD runs
representing dierent combinations of contacts between the
ligand and protein residues 134, 137, 142, and 144, as seen in
binding pose B5. These residues were chosen because either
biochemical data or molecular docking shows their interaction
with the ligand (Table S1). To compare all 3 ns MD
simulations, we monitored hydrogen bonds involving the
C7-OH, the C15-OH, and the NH of the methyl carbamate
side chain of archazolide. The conformation of the archazolide
macrocycle remained stable during the MD run (Figure S18).
As far as the interaction between archazolide and the binding
site of DCCD around residues 134 and 137 is concerned, stable
hydrogen bonds were formed between the C7-OH of the ligand
and either the carbonyl backbone of I134 or the side chain of
E137 in most of the runs. A hydrogen bond with the backbone
of V138 was less frequently observed. If the C7-OH of
archazolide forms a hydrogen bond to the carbonyl backbone
of I134, the side chain of E137 interacts with the side chain of
Y66. This hydrogen bond between the essential glutamate and
the highly conserved adjacent tyrosine (Figure S2) is also
present in the proton or sodium binding sites of F-
ATPases.
2527
The hydrogen bond between the C15-OH of
archazolide and the side chain of residue 142 was observed
considerably more often for the Y142N mutant than for the
wild type protein (Table S2), in very good agreement with the
higher anity of archazolide for the mutant protein.
Interestingly, the smaller asparagine side chain at position
142 favors the formation of the hydrogen bond between the
methyl carbamate side chain of archazolide and the backbone of
L141, which was observed more frequently for the Y142N
mutant than for the wild type protein. In both MD runs a close
interaction of the i-butyl side chain of archazolide and L144 was
observed. Even when starting from another conformer of
archazolide A, with a rotation of 180° around the CC
Ar
bond
in the archazolide side chain, the contact of the i-butyl group
with L144 was restored during the MD run, in very good
agreement with the sensitivity of the archazolide anity for
V-ATPase to mutations of L144.
The MD run with restraints set no. 3 for the wild type
protein (Table S1) and the Y142N mutant (Table S2) was
extended to 5 ns, and the snapshot of the trajectory with the
Figure 3. Stereoview of binding modes B2 (A) and B5 (B). Important residues belonging to one c subunit of the V
O
-rotor protein are labeled in
yellow and cyan, respectively. The ligand is shown in sticks (carbon: gray; oxygen: red; nitrogen: blue; sulfur: yellow; hydrogen: white). Key
interactions are shown as dashed lines.
Journal of Chemical Information and Modeling Article
dx.doi.org/10.1021/ci300242d | J. Chem. Inf. Model. 2012, 52, 226522722268
Page 4
lowest energy is shown in Figure 4. In essence, binding mode
B5 could be successfully rened by an MD simulation in an
explicit membrane; the rened pose retains the essential protein
ligand interactions seen in docking and explains well the bio-
chemical data available for the wild-type and mutant V-ATPases.
Experimental Validation of Binding Mode B5. The
validity of binding mode B5 was proved experimentally by
testing the activity of archazolide analogs with mutant yeast V-
ATPase. The distinctive landmark of B5 is the positioning of
the C15-OH next to Y142, in contrast to all other binding
modes B1B4, where the C15-OH is located close to I134,
E137, and Y66. For archazolide A (1) previous studies showed
that the Y142N mutant is ten times more sensitive to the ligand
than the wild-type protein. If binding mode B5 is correct,
modications at position C15 should promote a dierent
behavior of the archazolide derivative with the Y142N mutant
protein compared to the parental ligand 1. Vice versa, if one
of the other binding modes is correct, a ligand with modied
C15-OH should behave like 1 with the Y142N mutant pro-
tein. Following this reasoning, we synthesized ligand 4, which
is dehydrogenated at C15 and tested its biological activity with
boththewildtypeandtheY142Nmutantprotein.
Interestingly, 4 shows similar inhibitory proles for both the
wild-type and mutant protein. This change of behavior with
respect to the parental compound 1, which is ten times more
active toward the Y142N mutant V-ATPase, supports the
vicinity of the C15-OH to Y142, thereby endorsing B5 as a
plausible binding mode (Figure 5). The increased activity of 1
with the Y142N mutant protein can be explained by an
additional hydrogen bond formed between the hydroxyl group
at C15 and the carbonyl group of the asparagine side chain in
the Y142N mutant. This carbonyl group is a much better
hydrogen bond acceptor than the hydroxyl group of the wild
type tyrosine. In 4, the oxidation of C15 leads to a sp
2
-
hybridization state, which, in conjunction to the triene system
at C9C14, stiens the macrocycle of the ligand and likely
results in an increased distance between the asparagine side
chain of the Y142N mutant protein and the C15 carbonyl func-
tion. The absence of hydrogen bonds between the C15
functional group of ligand 4 and the residue 142 of the protein
would explain the similar sensitivity of the wild type and the
Y142N mutant protein to this ligand. Following the same
reasoning, we expected that 1-descarbamoyl-archazolide (5),
which lacks the carbamate side chain, would show the same
eciency as 1 to the Y142N mutant proteins, since the
hydroxyl function at C15 remains unmodied. This prediction
is fully conrmed in bioactivity tests (Figure 5), showing an
increased activity of ligand 5 toward mutant Y142N compared
Figure 4. Lowest energy MD-re ned binding mode B5 to the wild type (A) and mutant Y142N (B) V-ATPase. Important residues belonging to one
c subunit of the V
O
-rotor protein are colored in yellow and cyan, respectively. The ligand is shown as sticks (carbon: gray; oxygen: red; nitrogen:
blue; sulfur: yellow; hydrogen: white). Key interactions are shown as dashed lines.
Figure 5. Inuence of mutation Y142N in the yeast c subunit on the binding of ligand 4 (left) and ligand 5 (right). V-ATPase activities were
measured on isolated yeast vacuoles of the wild type strain BMA64-1B and mutant strain Y142N.
Journal of Chemical Information and Modeling Article
dx.doi.org/10.1021/ci300242d | J. Chem. Inf. Model. 2012, 52, 226522722269
Page 5
to the wild type protein. Additionally, the lack of the side chain
does not aect binding, since, in contrast to B1B4,inB5 the
carbamate does not interact extensively with the protein. These
data further support B5 as the binding mode of archazolides to
V-ATPases.
CONCLUSION
In this study, we describe the binding mode of archazolide A to
the V-ATPAse using an interdisciplinary experimental and
computational approach. We demonstrate that the functionality
at C15 of the ligand is close to residue 142 of the protein, as
shown by docking and conrmed by mutant cycle experiments.
The conjugated region between C1 and C6 of the ligand is
in proximity to the amino acids Y66 and E137, mainly
entertaining hydrophobic interactions. This is consistent with
the fact that the Y66F mutant binds archazolide A as strong as
the wild-type protein. We observe a hydrogen bond between
the hydroxyl group at C7 of archazolides and the backbone of
residue I134, in full agreement with the loss of activity for
archazolide C, with a bulky glucose moiety at C7, which would
clash with the protein surface. The DCCD binding side is
partially occupied by the archazolide, thereby preventing the
binding of this V-ATPase inhibitor. Additionally, the binding
pocket of the plecomacrolide inhibitors
17
is occupied too,
which explains why archazolides compete with the binding of
plecomacrolides to the V-ATPase.
Archazolides bind on the surface of the c ring rotor. While it
is commonly assumed that ligands binding on the surface of the
rotor exert their inhibitory activity by blocking the rotation, we
notice that the interaction of archazolides with the essential
glutamate could additionally prevent the E137-mediated proton
exchange between the c ring rotor and subunit a. Both facts can
rationalize the potent inhibitory eect of the V-ATPase function
by archazolides.
In conclusion, here we present the rst binding mode for the
most potent V-ATPase inhibitor discovered so far. This binding
mode, derived in silico and validated experimentally by mutant
cycle experiments, explains well all previously existing SAR
(Structure-Activity-Relationship) data. The next step along the
way to promote archazolide-based ligands as therapeutic agents
is the design of a simplied molecule, with an easy synthetic
route that is capable of forming the same critical intermolecular
interactions as the archazolide scaold. Eorts in this direction
are ongoing in our laboratories.
EXPERIMENTAL METHODS
Preparation of Protein Input Structures. The homology
model of the yeast rotor was generated by aligning the amino
acid sequence of the S. cerevisiae c subunit onto one of the 10-
fold symmetric crystal structure of E. hirae (PDB-ID: 2BL2)
and generating the homology model as described in ref 11. We
then minimized the homology model using MacroModel9.7
28
and the OPLS2005
29
force eld in aqueous solution and
simplied the whole yeast rotor to three subunits for the
docking (Figure S3). Protein side chain conformers were
generated during a 1 ns MD simulation at 300 K at a constant
pressure of 1 bar in the CHARMM22
30
force eld with the
CHARMM-XXL
31
program, using the TIP3P
32
model for water
and restrained backbone (force constant of 1 kcal mol
1
Å
2
).
Twenty snapshots were taken at equal time intervals of 50 ps
each. To overcome clash barriers, we used a temperature gra-
dient from 300 to 500 K for another 0.1 ns and extracted two
additional snapshots at equal time intervals of 50 ps. For
practical purposes, the resulting 22 conformers were then
minimized in aqueous solution in MacroModel9.7
28
using
OPLS2005
29
following the common procedure and served as
protein input structures for the docking. While using dierent
force elds may inuence the outcome of the rotamer search,
we do not expect a strong bias here given that modern protein
force elds behave comparably with respect to structural and
dynamical properties, as it has been shown by Price and
Brooks.
33
The mutant G61S was generated by substitution of residue
61 in the yeast homology model with serine, followed by min-
imization in the OPLS2005
29
force eld in aqueous solution
using Macromodel9.7.
28
Molecular Docking. Archazolide A (1) and its derivatives
(24) were docked to three subunits of the yeast homology
model using Autodock3.0
21
with the Drugscore
22
scoring
function. This combination has proven reliable for binding
mode prediction in a re-docking evaluation.
34
Default para-
meters were used, except for the number of GA runs, which was
set to 100, the population size, which was set to 200, and the
number of generations, which was set to 50000. The search
area and the electrostatic map size was chosen to be a cube with
edge length of 22.7 Å (60 grid points, 0.375 grid point
distance), centered on the triangle dened by amino acids
E137, Y142, and L144 (Figure S9 and S10). The ligand
macrocycle was kept xed according to the experimentally
determined conformation of archazolide A (1),
6
whereas the
thiazole-bearing side chain and the methoxy- and the hydroxyl
groups were fully rotatable throughout the docking. The struc-
tures of all archazolide derivatives (24) were generated in
Maestro9.2 from the solution conformation for archazolide A
(1). Subsequently, the modied regions were subjected to 100
steps of minimization with MacroModel9.7
28
using the
OPLS2005
29
force eld in aqueous solution, whereas the
unmodied ones were frozen to the solution conformation of 1.
2200 solutions were obtained for each archazolide analog
(22 protein conformers × 100 docking poses). The 2200
structures were clustered at a ligand root mean squared deviation
(rmsd) value of 2.0 Å. Clusters with population >5% (110 out of
2200 structures) were analyzed further. The structure with the
lowest docking energy was regarded as cluster representative.
The solution conformation of archazolide A (1) was also
docked to three subunits of the yeast mutant G61S using the
same conditions, except that we did not create side chain con-
formers. 100 solutions were obtained (1 protein conformer ×
100 docking poses), which were clustered at a ligand rmsd
value of 2.0 Å. As shown in Figure S13 the population of
binding modes B1B3 drops considerably upon substitution of
G61, whereas B5 comprises 70% of all solutions. The remaining
18 solutions are distributed among 7 other binding modes, with
populations between 1 to 3 structures (data not shown).
Molecular Dynamics Simulations. The complexes of
archazolide A and two variants of the V-ATPase protein (the
wild type and the Y142N mutant) were rened in the YASARA
program (YASARA Biosciences) using the AMBER03
23
force
eld with the TIP3P
32
model for water. The membrane used in
the molecular dynamics simulations consisted of phosphatidy-
lethanolamine lipids (PEA). The complexes were embedded
into the membrane based on the location of hydrophobic resi-
dues.
24
The width and the position of the membrane were
tted to the data for the V-ATPase homologue of E. hirae
(PDB-ID: 2BL2) deposited in the OPM database.
35
Force eld
Journal of Chemical Information and Modeling Article
dx.doi.org/10.1021/ci300242d | J. Chem. Inf. Model. 2012, 52, 226522722270
Page 6
parameters for lipids and the ligand were derived using
AutoSMILES.
36,37
The periodic box contained about 135000
atoms. The s imulations were conducted at a constant
temperature of 298 K using the Berendsen thermostat and a
constant pressure of 1 bar. The simulation time step was set to
2.5 fs. During the rst 0.25 ns of the MD simulation each of the
complexes (wild-type and mutant) was kept frozen to allow the
membrane to relax and to adapt to the embedded protein
complex without disrupting its structure. During the next 3 ns
of the simulation we imposed weak restraints on the
experimentally predicted interactions between the ligand and
residues 134, 137, 142, and 144 in the adjacent protein chain.
Restraints were included in dierent combinations (Table S1
and S2): 1. hydrogen bond between the C7-OH of archazolide
and the backbone or side chain of I134 and E137, respectively; 2.
hydrogen bond between the C15-OH of archazolide and the
side chain of Y142; 3. hydrophobic contact between the i-butyl
side chain of archazolide and the side chain of I144. The
restraints were imposed in the form of the SoftSquare poten-
tials
38
(a hydrogenacceptor distance equal to 2.0 ± 1 Å and
for the van der Waals interaction with L144 a distance of: 3.0 ±
1.0 Å). We assessed the MD runs by monitoring hydrogen
bond formation (Table S1 and S2). The hydrogen bonds
statistics was computed using the VMD software
39
(donor
acceptor distance: 3.5 Å, angle cuto: 30 degrees). The MD run
with the restraints set no. 3 (Table S1 and S2) was extended to
5 ns, and the representative snapshot with the lowest energy
was provided as a nal result.
To check the stability of the V-ATPase model we further
extended the MD simulation with restraints set no. 3 of the
wild type protein to 10 ns and computed the rmsd with respect
to the starting homology model (Figure S4).
Synthesis of Archazolide Derivatives. The structures of
15-dehydro-archazolide (4) and 1-descarbamoyl-archazolide
(5) were synthesized according to the published procedure.
10
Biological Assays. Purication of yeast vacuoles and
following activity assays were carried out as described in ref
11. The concentration of inorganic phosphate in the samples
was determined according to ref 40.
ASSOCIATED CONTENT
*
S
Supporting Information
Protein sequence alignments, yeast homology model, electro-
statics of yeast V-ATPase and archazolide A, computational
details for the molecular docking as well as the molecular
dynamics simulations, additional material of all docking-derived
binding modes, and hydrogen bond analyses of the MD
trajectories. This material is available free of charge via the
Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
*Phone: +49 6221 3878552. Fax: +49 6221 3878519. E-mail:
teresa.carlomagno@embl.de.
Present Addresses
§
International Institute of Molecular and Cell Biology, 4 Ks.
Trojdena, 02-109 Warsaw, Poland.
Kekule
́
-Institute of Organic Chemistry and Biochemistry,
University Bonn.
Notes
The authors declare no competing nancial interest.
ACKNOWLEDGMENTS
We thank the German Science Foundation (DFG, Graduate
College 850:Modeling of Molecular Properties ) for generous
nancial support (stipend to S.D.) and cluster usage as well as
Dennis Kru
̈
ger, Alexander Metz, and Doris Klein (Heinrich-
Heine-University Du
̈
sseldorf) for assistance in using Drugscore
and fruitful discussions. T.C. acknowledges nancial support
from EMBL and Grant I/81 637 from the Volkswagen Stiftung.
D.L. acknowledges the computational grant G35-6 from ICM
Warsaw. S.B., M.H., and H.W. acknowledge nancial support
from the Volkswagen Stiftung (Grant I/82 801).
ABBREVIATIONS
DCCD = N,N-dicyclohexylcarbodiimide; RDC = residual
dipolar coupling; GA = genetic algorithm; MD = molecular
dynamics; PEA = phosphatidylethanolamine
REFERENCES
(1) Sasse, F.; Steinmetz, H.; Ho
̈
fle, G.; Reichenbach, H. Archazolids,
new cytotoxic macrolactones from Archangium gephyra (Myxobac-
teria). Production, isolation, physico-chemical and biological proper-
ties. J. Antibiot. (Tokyo) 2003, 56 , 520525.
(2) Menche, D.; Hassfeld, J.; Steinmetz, H.; Huss, M.; Wieczorek, H.;
Sasse, F. Archazolid-7-O-β-D-glucopyranoside Isolation, Structural
Elucidation and Solution Conformation of a Novel V-ATPase
Inhibitor from the Myxobacterium Cystobacter violaceus. Eur. J. Org.
Chem. 2007, 2007, 11961202.
(3) Menche, D.; Hassfeld, J.; Steinmetz, H.; Huss, M.; Wieczorek, H.;
Sasse, F. The first hydroxylated archazolid from the myxobacterium
Cystobacter violaceus: isolation, structural elucidation and V-ATPase
inhibition. J. Antibiot. (Tokyo) 2007, 60, 328331.
(4) Horstmann, N.; Essig, S.; Bockelmann, S.; Wieczorek, H.; Huss,
M.; Sasse, F.; Menche, D. Archazolid A-15-O-beta-D-glucopyranoside
and iso-archazolid B: potent V-ATPase inhibitory polyketides from the
myxobacteria Cystobacter violaceus and Archangium gephyra. J. Nat.
Prod. 2011, 74, 11001105.
(5) Hassfeld, J.; Fares, C.; Steinmetz, H.; Carlomagno, T.; Menche,
D. Stereochemical determination of Archazolid A and B, highly potent
vacuolar-type ATPase inhibitors from the Myxobacterium Archangium
gephyra. Org. Lett. 2006, 8, 47514754.
(6) Fares, C.; Hassfeld, J.; Menche, D.; Carlomagno, T. Simultaneous
determination of the conformation and relative configuration of
archazolide a by using nuclear overhauser effects, J couplings, and
residual dipolar couplings. Angew. Chem., Int. Ed. Engl. 2008, 47,
37223726.
(7) Huss, M.; Sasse, F.; Kunze, B.; Jansen, R.; Steinmetz, H.;
Ingenhorst, G.; Zeeck, A.; Wieczorek, H. Archazolid and apicularen:
novel specific V-ATPase inhibitors. BMC Biochem. 2005, 6, 13.
(8) Menche, D.; Hassfeld, J.; Li, J.; Rudolph, S. Total synthesis of
archazolid A. J. Am. Chem. Soc. 2007, 129, 61006101.
(9) Menche, D.; Hassfeld, J.; Li, J.; Mayer, K.; Rudolph, S. Modular
total synthesis of archazolid A and B. J. Org. Chem. 2009, 74, 7220
7229.
(10) Menche, D.; Hassfeld, J.; Sasse, F.; Huss, M.; Wieczorek, H.
Design, synthesis and biological evaluation of novel analogues of
archazolid: a highly potent simplified V-ATPase inhibitor. Bioorg. Med.
Chem. Lett. 2007, 17, 17321735.
(11) Bockelmann, S.; Menche, D.; Rudolph, S.; Bender, T.; Grond,
S.; von Zezschwitz, P.; Muench, S. P.; Wieczorek, H.; Huss, M.
Archazolid A binds to the equatorial region of the c-ring of the
vacuolar H+-ATPase. J. Biol. Chem. 2010, 285, 3830438314.
(12) Beyenbach, K. W.; Wieczorek, H. The V-type H+ ATPase:
molecular structure and function, physiological roles and regulation.
J. Exp. Biol. 2006, 209, 577589.
(13) Mandel, M.; Moriyama, Y.; Hulmes, J. D.; Pan, Y. C.; Nelson,
H.; Nelson, N. cDNA sequence encoding the 16-kDa proteolipid of
Journal of Chemical Information and Modeling Article
dx.doi.org/10.1021/ci300242d | J. Chem. Inf. Model. 2012, 52, 226522722271
Page 7
chromaffin granules implies gene duplication in the evolution of
H+-ATPases. Proc. Natl. Acad. Sci. U.S.A. 1988, 85, 55215524.
(14) Hinton, A.; Bond, S.; Forgac, M. V-ATPase functions in normal
and disease processes. Pflugers Arch. 2009, 457, 589598.
(15) Forgac, M. Vacuo lar ATPases: rotary proton pumps in
physiology and pathophysiology. Nat. Rev. Mol. Cell Biol. 2007, 8,
917929.
(16) Finbow, M. E.; Eliopoulos, E. E.; Jackson, P. J.; Keen, J. N.;
Meagher, L.; Thompson, P.; Jones, P.; Findlay, J. B. C. Structure of a
16 kDa integral membrane protein that has identity to the putative
proton channel of the vacuolar H+-ATPase. Protein Eng. 1992, 5,
715.
(17) Bowman, B. J.; McCall, M. E.; Baertsch, R.; Bowman, E. J. A
Model for the Proteolipid Ring and Bafilomycin/Concanamycin-
binding Site in the Vacuolar ATPase of Neurospora crassa. J. Biol.
Chem. 2006, 281, 3188531893.
(18) Baker, N. A.; Sept, D.; Joseph, S.; Holst, M. J.; McCammon, J.
A. Electrostatics of nanosystems: Application to microtubules and the
ribosome. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 1003710041.
(19) Erde
́
lyi, M.; Pfeiffer, B.; Hauenstein, K.; Fohrer, J.; Gertsch, J.;
Altmann, K.-H.; Carlomagno, T. Conformational Preferences of
Natural and C3-Modified Epothilones in Aqueous Solution. J. Med.
Chem. 2008, 51, 14691473.
(20) Marraud, M.; Aubry, A. Crystal structures of peptides and
modified peptides. Biopolymers 1996, 40,4583.
(21) Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.; Hart,
W. E.; Belew, R. K.; Olson, A. J. Automated docking using a
Lamarckian genetic algorithm and an empirical binding free energy
function. J. Comput. Chem. 1998, 19, 16391662.
(22) Gohlke, H.; Hendlich, M.; Klebe, G. Knowledge-based scoring
function to predict protein-ligand interactions. J. Mol. Biol. 2000, 295,
337356.
(23) Duan, Y.; Wu, C.; Chowdhury, S.; Lee, M. C.; Xiong, G.; Zhang,
W.; Yang, R.; Cieplak, P.; Luo, R.; Lee, T.; Caldwell, J.; Wang, J.;
Kollman, P. A point-charge force field for molecular mechanics
simulatio ns o f p rotein s ba sed on condensed-phase quantum
mechanical calculations. J. Comput. Chem. 2003, 24, 19992012.
(24) Krieger, E. 2003, YASARA (www.yasara.org/sd).
(25) Murata, T.; Yamato, I.; Kakinuma, Y.; Leslie, A. G. W.; Walker,
J. E. Structure of the Rotor of the V-Type Na+-ATPase from
Enterococcus hirae. Science 2005, 308, 654659.
(26) Pogoryelov, D.; Yildiz, O.; Faraldo-Gomez, J. D.; Meier, T.
High-resolution structure of the rotor ring of a proton-dependent ATP
synthase. Nat. Struct. Mol. Biol. 2009, 16, 10681073.
(27) Meier, T.; Ferguson, S. A.; Cook, G. M.; Dimroth, P.; Vonck, J.
Structural Investigations of the Membrane-Embedded Rotor Ring of
the F-ATPase from Clostridium paradoxum. J. Bacteriol. 2006, 188,
77597764.
(28) Macromodel, version 9.7; Schro
̈
dinger, LLC: New York, NY,
2009.
(29) Jorgensen, W. L.; Maxwell, D. S.; Tirado-Rives, J. Development
and Testing of the OPLS All-Atom Force Field on Conformational
Energetics and Properties of Organic Liquids. J. Am. Chem. Soc. 1996,
118, 1122511236.
(30) MacKerell, A. D.; et al. All-Atom Empirical Potential for
Molecular Modeling and Dynamics Studies of Proteins. J. Phys. Chem.
B 1998, 102, 35863616.
(31) Chemistry at HARvard Macromolecular Mechanics
(CHARMM), version 35b2, 2008.
(32) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R.
W.; Klein, M. L. Comparison of simple potential functions for
simulating liquid water. J. Chem. Phys. 1983, 79, 926935.
(33) Price, D. J.; Brooks, C. L. Modern protein force fields behave
comparably in molecular dynamics simulations. J. Comput. Chem.
2002, 23, 10451057.
(34) Sotriffer, C. A.; Gohlke, H.; Klebe, G. Docking into Knowledge-
Based Potential Fields: A Comparative Evaluation of DrugScore.
J. Med. Chem. 2002, 45, 19671970.
(35) Lomize, M. A.; Pogozheva, I. D.; Joo, H.; Mosberg, H. I.;
Lomize, A. L. OPM database and PPM web server: resources for
positioning of proteins in membranes. Nucleic Acids Res. 2012, 40 (D1),
D370D376.
(36) Jakalian, A.; Jack, D. B.; Bayly, C. I. Fast, efficient generation of
high-quality atomic charges. AM1-BCC model: II. Parameterization
and validation. J. Comput. Chem. 2002, 23, 16231641.
(37) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D.
A. Development and testing of a general amber force field. J. Comput.
Chem. 2004, 25, 11571174.
(38) Brunger, A. T. X-PLOR version 3.1: A System for X-ray
Crystallography and NMR; Yale University Press: New Haven, CT,
1993.
(39) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual molecular
dynamics. J. Mol. Graphics Modell. 1996, 14,3338.
(40) Wieczorek, H.; Cio, M.; Klein, U.; Harvey, W. R.; Schweikl,
H.; Wolfersberger, M. G. Isolation of goblet cell apical membrane from
tobacco hornworm midgut and purication of its vacuolar-type
ATPase. In Methods Enzymology; Sidney Fleischer, B . F., Ed.;
Academic Press: 1990; Vol. 192, pp 608616.
Journal of Chemical Information and Modeling Article
dx.doi.org/10.1021/ci300242d | J. Chem. Inf. Model. 2012, 52, 226522722272
Page 8
  • Source
    • "Preliminary work has focused on the impact of sequence-specific contributions to internal friction [20,81828384 . Advances in nuclear magnetic reso- nance8586878889 and single molecule spectroscopies909192 combined with novel computational and theoretical methodologies939495 should pave the way for comprehensive characterization of IDP dynamics and assessing their impact on the dynamical regulation of cellular phenotypes [96,97]. Overall, it is clear that continued synergistic investigations must be brought to bear in order to build on the insights that have been forthcoming with regard to connecting information encoded in IDP sequences to their form and function. "
    [Show abstract] [Hide abstract] ABSTRACT: Intrinsically disordered proteins (IDPs) showcase the importance of conformational plasticity and heterogeneity in protein function. We summarize recent advances that connect information encoded in IDP sequences to their conformational properties and functions. We focus on insights obtained through a combination of atomistic simulations and biophysical measurements that are synthesized into a coherent framework using polymer physics theories. Copyright © 2015 Elsevier Ltd. All rights reserved.
    Full-text · Article · Jun 2015 · Current Opinion in Structural Biology
  • Source
    • "The first type of experiment uses FRET to define the ternary structures of dynamic protein complexes from subunits previously characterized at higher resolution by crystallography or NMR. Seidel has established sophisticated methods taking advantage of fluorescence intensity, lifetime, polarization and wavelength information obtained from smFRET measurements to build models using maximal information on the donor and acceptor chromophores [47]. For example, they recently used FRET (as well as double electron–electron resonance spectroscopy) to build a model of the dimer of the immune defense protein human guanylate binding protein 1 in the presence of GTP [48]. "
    [Show abstract] [Hide abstract] ABSTRACT: Fluorescence spectroscopy is a powerful method for monitoring protein folding in real-time with high resolution and sensitivity, but requires the site-specific introduction of labels into the protein. The ability to genetically incorporate unnatural amino acids (Uaas) allows for the efficient synthesis of fluorescently labeled proteins with minimally perturbing fluorophores. Here, we describe recent uses of labeled proteins in dynamic structure determination experiments and advances in unnatural amino acid incorporation for dual site-specific fluorescent labeling. The advent of increasingly sophisticated bioorthogonal chemistry reactions and the diversity of Uaas available for incorporation will greatly enable protein folding and stability studies.
    Full-text · Article · Apr 2015 · Current Opinion in Chemical Biology
  • Source
    • "Similarly to a global positioning system, iterative measurements of multiple distances can afford a high-resolution structural view of macromolecular complexes [20] . Because the resolution can increase with the number of probed distance constrains, sub-A ˚ ngströ m resolution has been achieved with advanced analysis of DNA complexes with HIV-1 reverse transcriptase [21]. One of the demands for smFRET-based techniques to serve as structural biology tools is the freedom to site-specifically, but noninvasively, attach at least two synthetic fluorophores anywhere in the sequence of protein. "
    [Show abstract] [Hide abstract] ABSTRACT: Contemporary structural biology research promises more than just static snap-shots of molecular machineries. This goal is not just facilitated by combining different structural biology techniques, but also by new tools from the field of protein and genetic engineering, as well as from chemistry. Genetic encoding of noncanonical amino acids (ncAAs) through codon-suppression technology provides an excellent opportunity to probe biomolecules using different structural biology methods. In this article, we review the applications of ncAA incorporation into proteins for determining structural information through various techniques with the main focus on crosslinking mass spectrometry and single-molecule FRET-based techniques. Furthermore, advances and limitations of the incorporation of multiple ncAAs are discussed, with respect to design of an ideal host organism for modern and integrative structural biology research. Copyright © 2015 Elsevier Ltd. All rights reserved.
    Full-text · Article · Mar 2015 · Current Opinion in Structural Biology
Show more