Structure-based design of potent small-molecule binders to the S-component of the ECF transporter for thiamine

Article abstract of DOI:10.1002/cbic.201402673

Energy-coupling factor (ECF) transporters are membrane-protein complexes that mediate vitamin uptake in prokaryotes. They bind the substrate through the action of a specific integral membrane subunit (S-component) and power transport by hydrolysis of ATP

Full text of DOI:10.1002/cbic.201402673

DOI: 10.1002/cbic.201402673
Full Papers
Structure-Based Design of Potent Small-Molecule Binders
to the S-Component of the ECF Transporter for Thiamine
Lotteke J. Y. M. Swier,[a] Leticia Monjas,[b] Albert Guskov,[a] Alrik R. de Voogd,[b]
Guus B. Erkens,[a] Dirk J. Slotboom,*[a] and Anna K. H. Hirsch*[b]
Energy-coupling factor (ECF) transporters are membrane-pro-
tein complexes that mediate vitamin uptake in prokaryotes.
They bind the substrate through the action of a specific inte-
gral membrane subunit (S-component) and power transport
by hydrolysis of ATP in the three-subunit ECF module. Here,
we have studied the binding of thiamine derivatives to ThiT, a
thiamine-specific S-component. We designed and synthesized
derivatives of thiamine that bind to ThiT with high affinity; this
allowed us to evaluate the contribution of the functional
groups to the binding affinity. We determined six crystal struc-
tures of ThiT in complex with our derivatives. The structure of
the substrate-binding site in ThiT remains almost unchanged
despite substantial differences in affinity. This work indicates
that the structural organization of the binding site is robust
and suggests that substrate release, which is required for
transport, requires additional changes in conformation in ThiT
that might be imposed by the ECF module.
Introduction
ATP-binding cassette (ABC) transporters are found in all king-
doms of life and are involved in many crucial processes.[1] The
recently discovered energy-coupling factor (ECF) transporters
form a structurally distinct subtype of ABC transporters.[2,3]
These transporters mediate the uptake of a wide variety of
substrates, among which are water-soluble vitamins, cofactors,
One of the transmembrane domains is responsible for sub-
strate binding and is termed the S-component. The mode of
substrate binding is one of the defining characteristics of ECF
transporters, because other ABC transporters that import sub-
strates in prokaryotes invariably make use of soluble extracyto-
plasmic binding proteins or domains. The identical or similar
NBDs of ECF transporters (EcfA and EcfA’) bind and hydrolyze
ATP. It is hypothesized that the energy released in this process
induces changes in conformation within and between the dif-
ferent domains of the transporter, and that this allows trans-
port of the substrate across the membrane. These changes in
conformation are thought to be transduced from the NBDs to
the S-component and vice versa through the action of the
second transmembrane domain, called EcfT. EcfA, EcfA’, and
EcfT together form the eponymous ECF module.
their precursors, and the transition-metal ions Co2+ and Ni2+ [4]
.
ECF transporters are found in prokaryotes, but have not been
identified in eukaryotes. In the kingdom of bacteria, the trans-
porters are found in particular in the phylum of the Firmicutes.
This phylum includes many human pathogens such as Staphy-
lococci, Streptococci, Enterococci, and Listeriae.[3,5] These patho-
gens lack enzymes for the de novo synthesis of many vitamins
and cofactors and are dependent on membrane proteins such
as the ECF transporters for the uptake of these molecules. Be-
cause of their absence in humans, ECF transporters are poten-
tial targets for the development of new antibiotics.
Although the architectures of three ECF transporters have
been revealed recently,[6–8] the transport mechanism is not fully
elucidated. To investigate this mechanism, we focused on sub-
strate binding to the S-component ThiT from Lactococcus
lactis. ThiT is a 21-kDa protein that binds its natural substrate
thiamine (vitamin B1) with sub-nanomolar affinity.[9] Thiamine is
a precursor of thiamine diphosphate (TDP), which serves as
a cofactor in many enzymes that perform decarboxylations
and are involved in essential cellular processes.[10] TDP and its
precursor thiamine monophosphate (TMP) can bind to ThiT as
well, but with lower affinities.[9] The crystal structure of ThiT
from L. lactis has been solved previously at 2.0 Š resolution
(PDB ID: 3RLB).[11]
ECF transporters feature the basic architecture of ABC trans-
porters, consisting of two cytoplasmic nucleotide-binding do-
mains (NBDs) and two transmembrane domains (Figure 1A).
[a] L. J. Y. M. Swier,+ Dr. A. Guskov, Dr. G. B. Erkens, Prof. D. J. Slotboom
Groningen Biomolecular Sciences and Biotechnology Institute
University of Groningen
Nijenborgh 4, 9747 AG Groningen (The Netherlands)
E-mail: d.j.slotboom@rug.nl
Homepage: http://www.rug.nl/research/membrane-enzymology/
[b] L. Monjas,+ A. R. de de Voogd, Dr. A. K. H. Hirsch
Stratingh Institute for Chemistry, University of Groningen
Nijenborgh 7, 9747 AG Groningen (The Netherlands)
E-mail: a.k.h.hirsch@rug.nl
Homepage: http://www.rug.nl/research/bio-organic-chemistry/hirsch/
[+] These authors contributed equally to this work.
Here, we rationally designed derivatives of thiamine such
that the various moieties of the original thiamine molecule
were altered one-by-one, in order to explore the space within
the substrate-binding pocket. The aim of this study was to
trap the protein in previously undetected conformational
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cbic.201402673.
ChemBioChem 2015, 16, 819 – 826
819
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Results
Design of small-molecule modulators
The crystal structure of ThiT in complex with thiamine (1; PDB
ID: 3RLB) shows that the protein consists of six transmembrane
helices (Figure 1B).[11] The substrate-binding pocket is located
on the extracytoplasmic side of the membrane and is formed
by residues in helices 4–6 and residues in loops L1 and L3,
which are located between helices 1 and 2 and helices 3 and
4, respectively. The flexible L1 loop shields the substrate-bind-
ing pocket from the surrounding solvent in a lid-like manner,
with Trp34 playing a major role in the substrate occlusion (Fig-
ure 1C).[12] The substrate-binding site is connected to the exte-
rior of the protein through an opening of 7–9 Š in diameter,
which provides multiple interaction sites for extended small
molecules. Thiamine interacts with several residues in the bind-
ing pocket (Figure 1D and Figure S1 in the Supporting Infor-
mation): the aminopyrimidine moiety acts as an anchor in-
volved in hydrogen bonds with residues Glu84, His125, Tyr146,
and Asn151, as well as in a p–p-stacking interaction with
Trp133. The thiazolium ring is engaged in p–p-stacking and
cation–p interactions with Trp34 and His125, and the positively
charged nitrogen atom is stabilized by an ionic interaction
with Glu84. The hydroxy group forms a hydrogen bond with
Tyr85. We designed small molecules and tested their ability to
interact with ThiT in order to determine which functional
groups are crucial for high-affinity binding (Scheme 1). MOLOC
software[13] was used for molecular modeling, and the pre-
ferred orientations of the molecules in the binding pocket and
the estimated free energies of binding were obtained by use
of the FlexX docking module[14] and the scoring function
HYDE[15,16] in the LeadIT suite, respectively.
Figure 1. Architecture of an ECF transporter and the structure of thiamine-
bound ThiT of L. lactis. A) Schematic representation of an ECF transporter in
which the nucleotide-binding domains EcfA and EcfA’ are shown in red, the
transmembrane domain EcfT in blue, the S-component in green, and the
membrane in orange. The cytoplasmic and extracytoplasmic space are indi-
cated by the words “In” and “Out”, respectively. B) Secondary-structure rep-
resentation of ThiT, colored from the N terminus in blue to the C terminus in
red. Helices 1–6 are indicated by H1–H6. Thiamine is shown in stick repre-
sentation and colored according to the following color code: C: dark green,
O: red, N: blue, S: yellow. This color code for the oxygen, nitrogen, and
sulfur atoms is maintained throughout the article. C) Close-up of the sub-
strate-binding pocket of ThiT with surface and secondary-structure represen-
tation. The surface of the L1 loop between helices 1 and 2 is colored light
blue, whereas the remainder of the surface is shown in gray. Residue Trp34
is shown in dark blue and stick representation, and thiamine is represented
as in (B). D) The residues involved in the binding of thiamine are shown in
stick representation with their carbon atoms colored gray, whereas the
other atoms and thiamine are colored as in (B). The names of the residues
are given by the three-letter codes for amino acids. The dashed lines indi-
cate electrostatic interactions and hydrogen bonds, and the gray spheres
represent water molecules. Figure 1B–D was generated with Pymol and use
of chain A of structure PDB ID: 3RLB.[11]
It has previously been reported that when the thiazolium
ring of thiamine (1) is substituted by a pyridinium ring, as in 2,
states, in order to provide new insights into the mechanism of
binding and transport. Furthermore, because the ECF trans-
porter for thiamine is essential for the uptake of this vitamin in
some pathogenic bacteria, this study is a first step on the way
to new antibiotics capable of blocking thiamine transport and
thereby interfering with essential cellular processes within
these bacteria.
Scheme 1. Small-molecule binders for ThiT. A) Structure of thiamine (1), with
the names of the rings and the hydroxyethyl group indicated. B) Structures
of pyrithiamine (2), thiamine monophosphate (3), thiamine diphosphate (4),
and the designed and synthesized small molecules 5–9.
ChemBioChem 2015, 16, 819 – 826
www.chembiochem.org
820
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Table 1. Binding affinities of ThiT for various small molecules as deter-
mined in the ligand-binding assay, with the errors indicated as standard
deviations, together with the experimentally measured and estimated
Gibbs free energies of binding (DG). The estimated values are based on
the scoring function HYDE.
Compound
KD ÆS.D. [nm]
DGexp [kJmolÀ1
]
DGest [kJmolÀ1
]
1
2
3
4
5
6
7
8
9
0.122Æ0.013[9]
0.180Æ0.070[9]
1.01Æ0.14[9]
1.60Æ0.00[9]
4.23Æ1.69[a]
266Æ131[b]
À57
À56
À51
À50
À48
À38
À47
À36
À46
À53
À54
À50
À52
À52
À48
À46
À43
À40
5.21Æ3.19[c]
528Æ135[c]
7.44Æ1.67[c]
[a] The error represents the standard deviation, obtained from four ex-
periments. [b] The error represents the standard deviation, obtained from
three experiments. [c] The errors each represent the range of the data
from two experiments.
Scheme 2. Synthesis of compounds 5, 7, and 9. a) sBuLi, Et2O, À788C, 5 min,
93%; b) i: sBuLi, Et2O, À788C, 1 h, ii: ethylene oxide, BF3·Et2O, À788C to
room temperature, 2 h, 68%; c) i: sBuLi, Et2O, À78 to À458C, 1 h, ii: DMF,
À788C to RT, 2 h, iii: NaBH4, MeOH, RT, 1 h, 79% over two steps; d) i:
iPrMgCl, THF, 08C, 10 min, ii: tBuLi, À78 to À408C, 1 h, iii: DMF, À788C to
RT, 16 h, 65% (13) and 62% (16); e) 3-anilinopropionitrile, NaOEt, DMSO,
microwave, 708C, 30–90 min, 71% (14 and 17); f) acetamidine·HCl, NaOEt,
EtOH, reflux, 1–3 d, 68% (5) and 61% (7); g) MnO2, DMF, RT, 2 h, 70%.
the affinity of the new molecule for ThiT is essentially the same
(Table 1).[9] When the hydroxy group from the hydroxyethyl
side chain is mono- or diphosphorylated (compounds 3 and 4),
the binding affinity decreases by one order of magnitude
(Table 1).[9] Firstly, we performed docking studies for com-
pounds 1–4, for which the KD values had been determined pre-
viously, to benchmark how well the docking program performs
on our target protein and class of compounds. According to
the estimated Gibbs free energies of binding (DGest), we could
predict that compounds 1 and 2 should have higher binding
affinities for ThiT than compounds 3 and 4, and this matched
with the experimental data obtained from intrinsic-fluores-
cence titration assays (Table 1).
quenching with methanol, intermediate 11 was obtained in
93% yield. In the second step, the hydroxyethyl side chain was
introduced by treatment with sBuLi followed by reaction of
the resulting anion with ethylene oxide, affording the desired
product 12 in 68% yield. In the third step, treatment with
iPrMgCl (for the in situ protection of the hydroxy group), fol-
lowed by tBuLi, afforded an anion, which was treated with
DMF to yield the aldehyde 13 (65%).[18] This aldehyde was con-
densed with 3-anilinopropionitrile, affording 14 in 71% yield.
The aminopyrimidine ring was formed by cyclization of the
enamine 14 with acetamidine under basic conditions, affording
the desired product in 68% yield. Compound 5 was obtained
in 21% overall yield, an improvement on the 11 % overall yield
reported previously.[17]
In our design, we maintained the aminopyrimidine moiety
of thiamine in order to preserve the interactions between this
part of the molecule and ThiT. We explored different modifica-
tions in the rest of the molecule: the thiazolium ring and the
hydroxyethyl side chain. We substituted the thiazolium ring
with different aromatic moieties, such as a thiophenyl (com-
pound 5) or a phenyl ring (compound 6), to evaluate the
effect on the p–p-stacking interaction of this moiety with resi-
dues Trp34 and His125. In order to analyze the importance of
the hydroxyethyl side chain of thiamine (1), we designed deriv-
atives of 5 and 6 featuring either a hydroxymethyl side chain
(compounds 7 and 8) or an aldehyde group (compound 9) to
evaluate the importance of the hydrogen bond with Tyr85.
The predicted binding poses of the new derivatives (Figure S2
in the Supporting Information) are almost identical to that of
thiamine, with differences only in the hydrogen bonds formed
by the new groups introduced in place of the hydroxyethyl
substituent of thiamine.
Compound 7 was synthesized by the same route. In this
case, in the second step 11 was treated with sBuLi, and the re-
sulting anion was allowed to react with DMF, affording an alde-
hyde that was reduced to the corresponding alcohol 15 with
NaBH4 in 79% yield over two steps. Compound 9 was obtained
by oxidation of 7 with MnO2, which afforded the desired prod-
uct in 70% yield.
Compounds 6 and 8 were synthesized by an analogous
route (Scheme 3), and by modification of a previously reported
synthesis of 8.[19] Commercially available bromides 18 and 19
were converted into the corresponding aldehydes by treat-
ment with iPrMgCl, tBuLi, and DMF, in 74 and 70% yield for 20
and 21, respectively. These aldehydes were each condensed
with 3-anilinopropionitrile under basic conditions to afford
22 and 23, in 45 and 24% yield, respectively. The final step,
Synthesis of small-molecule modulators
Compound 5 was synthesized by optimizing a previously re-
ported synthetic route (Scheme 2).[17] On starting from 2,3,5-tri-
bromo-4-methylthiophene (10), after treatment with sBuLi and
ChemBioChem 2015, 16, 819 – 826
www.chembiochem.org
821
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Table 2. Conditions for co-crystallization of ThiT with small-molecule
binders and diffraction results.
Compound Concentration[a] Resolution of
RMSD alignment [Š]
crystal structure [Š] chain A chain B
[mm]
2
5
6
7
8
9
100
10
40
10
10
10
2.1
2.5
2.0
2.2
2.4
2.2
0.282
0.341
0.310
0.305
0.332
0.287
0.276
0.322
0.379
0.274
0.361
0.317
[a] For purification and crystallization.
Scheme 3. Synthesis of compounds 6 and 8. a) i: iPrMgCl, THF, 08C, 10 min,
ii: tBuLi, À78 to À408C, 1 h, iii: DMF, À788C to RT, 16 h, 74% (20) and 70%
(21); b) 3-anilinopropionitrile, NaOEt, DMSO, 408C, 16 h, 45% (22) and 24%
(23); c) acetamidine·HCl, NaOEt, EtOH, reflux, 1–2 d, 66% (6) and 61% (8).
ThiT in complex with thiamine revealed very subtle differences
only in the spatial orientations of the a-helices. As in the case
of ThiT complexed with thiamine, two molecules of ThiT
(chains A and B) were present in the unit cell, binding one
small molecule each. Comparison of the binding sites in the
different chains showed only slight differences in the orienta-
tions of water molecules, except for the structures with com-
pounds 6 and 8. Superposition of the six structures with the
corresponding chain of ThiT in complex with thiamine gave
RSMD values between 0.274 and 0.379 Š, indicating that they
are almost identical (Table 2). When zooming into the residues
of ThiT involved in substrate binding, we noticed significant
changes in the orientations of the side chains only for Trp34 in
the complexes with compounds 6 and 8. Figure 2A–F shows
the binding site residues in chain A of ThiT co-crystallized with
compounds 2 and 5–9, respectively. The binding sites in
chain B of ThiT are shown in Figure S3. The p–p-stacking inter-
actions of the thiazolium ring sandwiched between Trp34 and
His125 are conserved in most of the co-crystallized structures,
as we predicted (Figure S2). In the co-crystallized structures
with compounds 6 and 8, Trp34 has moved in the direction of
the pyrimidine ring, making a CH–p interaction with this ring.
In the binding site of compound 6 in chain B of the structure
(Figure S3D), Trp34 is located even further away from the pyri-
midine ring, making it possible for a PEG molecule to bind
nonspecifically in the binding pocket. In the case of com-
pound 8, Trp34 is in a position similar to that in the structure
of ThiT in complex with thiamine in the binding site in chain A
(Figure 2E). However, the electron density for this residue ex-
tends to the side of the pyrrole moiety of the indole, showing
that it could also adopt the position shown in the binding site
in chain B (Figure S3F).
the cyclization of the enamines 22 and 23 with acetamidine,
afforded the desired products 6 and 8 in 66 and 61% yield,
respectively.
Ligand-binding measurements by an intrinsic-fluorescence
titration assay
We determined the KD values for compounds 5–9 by monitor-
ing the intrinsic protein fluorescence upon addition of ligand
(Table 1). The KD values for compounds 1–4 had been deter-
mined by the same method in a previous study.[9] Replacement
of the positively charged nitrogen atom in the thiazolium ring
of thiamine (1) by a carbon atom in 5 resulted in an increase
in KD to 4.23 nm. Replacement of the thiazolium ring of 1 by
a phenyl ring in compound 6 led to a decrease in binding af-
finity, resulting in a KD of 266 nm. Variations of the hydroxyeth-
yl side chain of 5 did not have any significant effect, as shown
by the KD values of 5.21 and 7.44 nm for compounds 7 and 9,
respectively. Shortening the hydroxyethyl side chain in com-
pound 6 by one carbon atom in compound 8 led to an in-
crease in KD up to 528 nm.
Co-crystallization of ThiT with small-molecule modulators
and structure determination
To visualize how the small molecules bind to ThiT, we per-
formed co-crystallization of substrate-free ThiT (purified in
0.35% (w/v) n-nonyl b-d-glucopyranoside (NG)) in the presence
of various concentrations of compounds 2 and 5–9 (Table 2),
at 58C by the vapor diffusion hanging drop method. All crys-
tals were obtained under the same conditions as for ThiT with
thiamine, with use of a reservoir solution of 0.15m NH4NO3
and 20% (w/v) PEG 3350. The crystals grew to full size in one
to two weeks and had the symmetry of space group C2. All
structures were solved to 2.0–2.5 Š resolution (Table 2). The
data processing and refinement statistics are shown in
Table S1 in the Supporting Information.
Despite the movement of the side chain of Trp34 and the
broadening of the entrance to the binding pocket in the cases
of these two compounds, the backbone of the L1 loop main-
tains its position. The electrostatic interaction of Glu84 with
the positively charged nitrogen atom in the thiazolium ring is
only conserved in the complex with compound 2, this being
the only derivative in which the charged nitrogen atom is
maintained. The hydroxyethyl group of compound 5 forms
water-mediated hydrogen bonds with Asn29 and Tyr85. The
same group of compound 6 forms hydrogen bonds through
a single molecule of water with both Tyr85 and His72. The hy-
droxymethyl or aldehyde groups of compounds 7 and 9 form
A comparison of the overall crystal structures of ThiT com-
plexed to the synthesized compounds with the structure of
ChemBioChem 2015, 16, 819 – 826
www.chembiochem.org
822
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Figure 2. Co-crystal structures of ThiT with small-molecule binders. The residues involved in the binding of the small-molecule binders are shown with use
of the same format and color code as in Figure 1D. A)–F) The carbon atoms of compounds 2 and 5–9 are shown in magenta (PDB ID: 4MUU), cyan (4MHW),
violet purple (4MES), lime (4POV), orange (4N4D), and deep salmon (4POP), respectively. The figure was created by using chain A of the corresponding PDB
file.
water-mediated hydrogen bonds with the backbone carbonyl
group of Glu84, the amino group of Lys121, and the hydroxy
group of Tyr85. In the case of compound 8, the phenyl ring
has flipped over, relative to its position for compound 6, so as
to allow the hydroxymethyl group to form a direct hydrogen
bond with the backbone carbonyl group of Asn151. Besides
the ordered water molecule through which Tyr146 interacts
with the pyrimidine ring of the ligands, there are a few other
water molecules involved in connecting binding site residues
with each other. The side chain of Lys121 interacts with the
backbone of Glu84 through an ordered water molecule in all
structures, which likely assures correct alignment of His125
with the ligand. In addition, the backbone of Trp34 makes in-
teractions with the side chains of Tyr146 and Asn151 through
ordered water molecules; this could help the backbone of
loop L1 to keep its position.
are in agreement with those predicted with computational
methods, thus highlighting the power of docking studies
when probing protein–ligand interactions.
From the KD values obtained from the intrinsic-fluorescence
titration measurements, we conclude that the presence of the
positively charged nitrogen atom in the central ring contrib-
utes to the picomolar binding affinity. Compounds 1 and 2, in
which the thiazolium ring is replaced by a pyridinium ring, dis-
play comparable KD values of 122 pm and 180 pm, respectively.
Measuring KD values in the low nanomolar or picomolar range
with high accuracy is difficult, and with a ThiT concentration of
50 nm in the intrinsic-fluorescence titration assay, these values
cannot be considered significantly different. Replacement of
the thiazolium ring by a thiophene ring, and thereby removing
the positive charge, lowered the binding affinity about 30-fold.
Nonetheless, derivatives 5, 7, and 9 still bind with high affinity
to ThiT (KD values between 4.2 and 7.4 nm), showing that
these analogues are suitable thiamine mimics. Removal of the
positively charged nitrogen atom and the methyl group in
compounds 6 and 8, relative to 2, led to decreases in binding
affinity, resulting in KD values in the upper nanomolar range.
This indicates that the presence of the charged nitrogen atom
in the context of a six-membered ring provides a bigger contri-
bution to interactions involved in binding than in the case of
the five-membered ring. From the co-crystallization studies we
could see that Glu84 is engaged in an electrostatic interaction
with the positively charged nitrogen atom, and that this is lost
when this nitrogen is missing. In addition, the charged nitro-
gen atom also enables cation–p interactions with Trp34 and
His125. In the absence of the charged nitrogen, only p–p-
Discussion
We have studied the binding of thiamine derivatives to ThiT,
the S-component for thiamine in L. lactis, by modifying moiet-
ies of thiamine in a modular manner. Because the crystal struc-
ture of ThiT features promising pockets for expansion of thi-
amine adjacent to the thiazolium ring and the hydroxyethyl
side chain, we focused on modifications of these moieties
whilst keeping the pyrimidine moiety intact. The estimated DG
values (DGest) from the docking studies correlate very well with
the experimentally measured DG values derived from the KD
measurements by intrinsic-fluorescence titration (DGexp). In
addition, the binding poses observed by X-ray crystallography
ChemBioChem 2015, 16, 819 – 826
www.chembiochem.org
823
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
stacking interactions can be formed, resulting in a weaker in-
teraction of the ligand with Trp34 and His125. In the cases of
co-crystallization with compounds 6 and 8, Trp34 was dis-
placed in the direction of the pyrimidine ring, leaving only
His125 to form p–p-stacking interactions with the phenyl ring
in these compounds. Both changes decreased the binding
affinities, with KD values of 266 and 528 nm for compounds 6
and 8, respectively.
On comparing compounds 5 and 7, each featuring a thio-
phenyl ring, with compounds 6 and 8, each with a phenyl
ring, the binding affinity is about 50 to 70 times lower in the
case of the phenyl ring. The difference in affinity could be ex-
plained by the sulfur atom in compounds 5 and 7, which can
engage in S–p interactions with Trp34 and His125, and by the
methyl group at C-4, which could also contribute through CH–
p interactions with Tyr85. Furthermore, different alignments of
the dipole moments of the aromatic moieties could also ac-
count for the observed difference in affinity. Removing the
methyl group in compounds 6 and 8 leads to a loss of van der
Waals interactions. In the cases of these two compounds, the
side chain of Trp34 adopts conformations different from those
in the other co-crystal structures; this could also be a reason
for the drop in binding affinity.
Conclusions
At the outset of this study, we hypothesized that the designed
thiamine derivatives might trap ThiT in previously undetected
conformational states, and that this might allow us to follow
the changes in conformation that take place for substrate
binding and release. Surprisingly, in spite of differences in affin-
ity of over three orders of magnitude (compounds 1 and 8)
the crystal structures of ThiT with the bound thiamine deriva-
tives were almost identical. The only difference in the binding-
site residues that we observed is a rotation of the side chain of
Trp34 located in loop L1 when compounds 6 and 8 are bound.
The different orientation of Trp34 opens the entrance to the
binding site pocket slightly, but the backbone of loop L1 main-
tains its position, keeping the substrate-binding pocket intact.
Although this observation is consistent with the EPR studies
on ThiT that showed that the main structural difference be-
tween apo-ThiT and the thiamine-bound protein is the confor-
mation of loop L1,[12] it does not provide the anticipated mech-
anistic insight into the mechanism of binding. Apparently, the
structure of the binding site in the S-component is so robust
that it stays intact, even if lower-affinity substrates bind. This
robustness has direct consequences for the transport mecha-
nism employed by the ECF transporter. We hypothesize that
the S-component “topples” in the membrane only if the loop
L1 is closed. Interaction with the ECF module is required for
“toppling”, resulting in the exposure of the binding site to the
cytoplasmic side of the membrane, where the substrate can be
released. Possibly ATP hydrolysis is needed for opening of the
binding site and substrate release. Indeed, in the context of
the entire complex (S-component and ECF module), the sub-
strate-binding site of the S-component FolT (for folate) is dis-
rupted and the loop L1 is displaced outward.[6] We speculate
that a substantial input of free energy is required to accom-
plish the changes in conformation in the S-component’s bind-
ing site that are necessary for transport and release of the sub-
strate. This energy is probably provided by the association
with the ECF module and by binding and hydrolysis of ATP by
the NBDs.
Having investigated the importance of the central ring, we
next examined the importance of the hydrogen-bonding inter-
action with Tyr85. To do this, we modified the hydroxyethyl
group in compounds 7–9. Compound 7 shows that shortening
the spacer by one carbon atom to a hydroxymethyl group
does not change the binding affinity significantly relative to
compound 5. In the cases of compounds 6 and 8, the same
shortening increases the KD values roughly twofold. In the case
of compound 8, the interaction with Tyr85 is lost and instead,
a new hydrogen bond is formed with the carbonyl group of
Asn151. This carbonyl group is in closer proximity than Tyr85,
but the interaction with Asn151 does not seem to contribute
to the binding affinity to the same degree as an interaction
with Tyr85. In addition, compound 8 also interacts with the
backbone carbonyl group of Trp34 through an ordered water
molecule. The lack of interactions between compound 8 and
residues Glu84, Tyr85, and Lys121 through an ordered water
network might weaken binding, thereby lowering the binding
affinity to a larger extent relative to the binding affinity of
compound 6. The interaction of the aldehyde group of com-
pound 9 with Tyr85 appears to contribute to the binding affini-
ty almost as well as that of the hydroxyethyl group in com-
pound 5, so there is no real preference for a hydrogen bond
donor or acceptor on this side of the substrate, given that
most interactions between the ligands and Tyr85 are water-
mediated. In addition, in the binding site in chain B of ThiT in
complex with thiamine (Figure S3A), the hydroxyethyl group
of thiamine has turned away from Tyr85 and is now engaged
in a hydrogen bond with the side chain of Glu38. This indicates
that the interaction between Tyr85 and the ligand is not partic-
ularly important and does not make a major contribution to
the high-affinity binding.
The observation that the thiamine derivatives bind to ThiT
with high affinities is very promising for another reason. It sug-
gests that the versatile scaffold of the derivatives that lack the
thiazolium ring (compounds 5, 7, and 9) could be further
modified, which is an important step in the design of new de-
rivatives. They can assist not only in the elucidation of the
transport mechanism but also in the development of new anti-
biotics targeted against pathogenic bacteria that depend on
this type of transporters.
Experimental Section
Materials: The detergents n-dodecyl b-d-maltopyranoside (DDM),
n-decyl b-d-maltopyranoside (DM), and NG were purchased from
Anatrace. All reagents used for the synthesis of the small-molecule
modulators were purchased from Sigma–Aldrich, Acros Organics,
or TCI Europe, and were used without further purification unless
ChemBioChem 2015, 16, 819 – 826
www.chembiochem.org
824
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
noted otherwise. All solvents were reagent-grade and, if necessary,
dried and distilled prior to use.
KCl (200 mm), glycerol (10%, w/v)] to a final concentration of 6–
10 mgmLÀ1. For the ThiT co-crystallization experiments, certain
concentrations of compounds 2 and 5–9 (Table 2) were added to
all buffers used during the purification. The membrane vesicles
were solubilized with DDM (1%, w/v) for 1 h at 48C with gentle
rocking. The undissolved material was removed by centrifugation
for 30 min at 80000 rpm (MLA-80 rotor) and 48C. The supernatant
was incubated with Ni2+-Sepharose resin (column volume=
0.5 mL), equilibrated with buffer B, for 1 h at 48C with gentle rock-
ing. Subsequently, the suspension was poured onto a 10 mL dis-
posable column (Bio-Rad) and, after collection of the flow-through,
the column material was washed with 20 column volumes of buf-
fer C [KPi (pH 7.0, 50 mm), KCl (200 mm), imidazole (50 mm), DM
(0.15%, w/v)]. In cases of purification of protein for crystallization,
DM was replaced with NG (0.35%, w/v). The protein was eluted in
three fractions (of 350, 650, and 500 mL) of buffer D (KPi (pH 7.0,
50 mm), KCl (200 mm), imidazole (500 mm), DM (0.15%, w/v)). In
cases of purification of protein for crystallization, DM was replaced
with NG (0.35%, w/v). EDTA (1 mm) was added to the second elu-
tion fraction, which contained most of the purified protein, and
this fraction was loaded on a Superdex 200 10/300 gel filtration
column (GE Healthcare) equilibrated with buffer E [KPi (pH 7.0,
50 mm), KCl (150 mm), DM (0.15%, w/v)]. In cases of purification of
protein for crystallization, buffer F [HEPES (pH 7.0, 20 mm), KCl
(150 mm), NG (0.35%, w/v)] was used for gel-filtration chromatog-
raphy. After gel-filtration chromatography, the fractions containing
ThiT were used directly for further analysis or concentrated by use
of a Vivaspin 2 concentrator device with a molecular weight cut-off
Modeling and docking of small-molecule modulators: The crystal
structure of ThiT in complex with thiamine (PDB ID: 3RLB) was
used for modeling.[11] Thiamine derivatives were designed by using
the program MOLOC,[13] and the energy of the system was mini-
mized by use of the MAB force field implemented in this software,
while the protein coordinates and the crystallographically localized
water molecule (HOH196) were kept fixed. Hydrogen bonds and
hydrophobic interactions were measured in MOLOC. The designed
thiamine derivatives were subsequently docked into the binding
pocket of ThiT with the aid of the FlexX docking module in the
LeadIT suite.[14] During docking, the binding site in the protein was
restricted to 8.0 Š around the co-crystallized thiamine, and the 30
top-scored solutions were retained and subsequently post-scored
with the scoring function HYDE.[15,16] After careful visualization to
exclude poses with significant inter- or intramolecular clash terms
or unfavorable conformations, the resulting solutions were subse-
quently ranked according to their binding energies.
Synthesis of small-molecule modulators—general methods: All
reactions were carried out under nitrogen (if not otherwise indicat-
ed), with use of dried glassware. Reactions were monitored either
by GC-MS (GCMS-QP2010 Shimadzu) with a HP-5 column (Agilent
Technologies) or by thin-layer chromatography (TLC) on silica-gel-
coated aluminium foils (silica gel 60/Kieselguhr F254, Merck). Flash-
column chromatography was performed on silica gel (SiliCycle 40–
63 mm). Melting points were determined with a Büchi B-545 appa-
ratus. NMR spectra were recorded with a Varian AMX 400 spec-
trometer at 258C. Chemical shifts (d) are reported in ppm relative
to the residual solvent peak. FTIR spectra were measured with a
PerkinElmer FTIR spectrometer. High-resolution mass spectra were
recorded with a Thermo Scientific LTQ Orbitrap-XL mass spectrom-
eter. Details of the synthesis and characterization are described in
the Supporting Information.
of 50 kDa (Sartorius stedim) to
a final concentration of 6–
10 mgmLÀ1 and used for crystallization.
Ligand-binding measurements by intrinsic-fluorescence titration
assay: The ligand-binding measurements by intrinsic-fluorescence
titration were performed as described previously,[9] by use of
a Spec Fluorlog 322 fluorescence spectrophotometer (Jobin Yvon),
with some modifications. Purified ThiT was diluted in buffer E to
a concentration of about 50 nm in a final volume of 800 mL. The
substrates were added in 1 mL steps from a Harvard apparatus
syringe pump with a 500 mL gas-tight glass syringe (Hamilton).
With use of an excitation wavelength of 280 nm, emission was
measured at 350 nm and 258C. After each addition of substrate,
10 s were allowed for equilibration and mixing, and the signals
were averaged over 15 s. The data analysis was performed as
described.[9]
Expression and purification of ThiT: Expression of native ThiT
with N-terminal His8-tag was performed as described previously,[9,11]
with some modifications. Briefly, L. lactis NZ9000 cells[20] with
pNZnHisThiT plasmid were grown semi-anaerobically in chemically
defined medium[21] without thiamine and supplemented with
glucose (2.0%, w/v) and chloramphenicol (5 mgmLÀ1) in a 10 L bio-
reactor at 308C and pH 6.5. At an OD600 of 1.5, expression of ThiT
was induced by addition of culture supernatant (0.1%, v/v) from
a Nisin-A-producing strain.[20] The cells were induced for 3 h and
harvested at a final OD600 of 7–8. After harvesting and washing by
centrifugation for 15 min at 6000 rpm (JLA9.1000 rotor) and 48C,
the cells were resuspended in buffer A [potassium phosphate
buffer (KPi, pH 7.0, 50 mm)], frozen in liquid nitrogen, and stored
at À808C.
Co-crystallization of ThiT with small-molecule modulators and
structure determination: The concentrated ThiT (purified in NG
(0.35%, w/v)) in the presence of compounds 2 or 5–9 (see Table 2
for the concentrations used) was used to set up 24-well hanging
drop crystallization plates with use of a reservoir solution (NH4NO3
(0.15m), PEG 3350 (20%, w/v)). Hanging drops of 2 mL were set up
in a 1:1 ratio of concentrated protein solution to reservoir solution.
The plates were incubated at 58C, and after one to two weeks,
crystals appeared. For cryoprotection, a cryoprotection solution
(NH4NO3 (15 mm), PEG 3350 (40% or 50%, w/v)) was used when
the crystals were fished. Diffraction data were collected at the
DESY, Hamburg (ThiT co-crystallized with compounds 6, 7, or 9) or
the ESRF, Grenoble (ThiT co-crystallized with compounds 2, 5, or 8).
Data processing was carried out by use of XDS,[22] and molecular
replacement was performed with the aid of Phaser MR in the CCP4
suite,[23] with use of the structure of thiamine-bound ThiT (PDB ID:
3RLB).[11] Refinements were performed by use of both Phenix re-
finement[24] and Refmac.[25] Manual model building and the place-
ment of substrate, detergent, and PEG molecules were done with
The preparation of membrane vesicles was performed as described
previously,[9] with some modifications. After the resuspended cells
had been thawed, MgSO4 (5 mm) and DNase (375 mgmLÀ1) were
added. The cells were lysed by high-pressure disruption (Constant
Cell Disruption Systems, Ltd, UK, twofold passage at 269 MPa and
48C), and the cell debris was separated from the membrane vesi-
cles by low-speed centrifugation for 30 min at 15000 rpm (JA25.50
rotor) and 48C. The membrane vesicles were harvested by high-
speed centrifugation for 2 h at 40000 rpm (45Ti rotor) and 48C,
resuspended in buffer A to a final concentration of 10 mgmLÀ1
,
frozen in liquid nitrogen, and stored at À808C.
For the purification of substrate-free ThiT, membrane vesicles were
rapidly thawed and resuspended in buffer B [KPi (pH 7.0, 50 mm),
ChemBioChem 2015, 16, 819 – 826
www.chembiochem.org
825
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Coot.[26] Refinement statistics are shown in Table S1. The crystal
structures were deposited in the PDB with the following PDB IDs.
ThiT+compound 2: 4MUU. ThiT+compound 5: 4MHW. ThiT+
compound 6: 4MES. ThiT+compound 7: 4POV. ThiT+compound
8: 4N4D. ThiT+compound 9: 4POP.
[9] G. B. Erkens, D. J. Slotboom, Biochemistry 2010, 49, 3203–3212.
[10] R. Kluger, K. Tittmann, Chem. Rev. 2008, 108, 1797–1833.
ˇ ´
[11] G. B. Erkens, R. P. Berntsson, F. Fulyani, M. Majsnerowska, A. Vujicic-
ˇ
Zagar, J. Ter Beek, B. Poolman, D. J. Slotboom, Nat. Struct. Mol. Biol.
2011, 18, 755–760.
[12] M. Majsnerowska, I. Hänelt, D. Wunnicke, L. V. Schäfer, H. J. Steinhoff,
D. J. Slotboom, Structure 2013, 21, 861–867.
[13] P. R. Gerber, K. Müller, J. Comput.-Aided Mol. Des. 1995, 9, 251–268.
[14] LeadIT (version 2.1.2), BioSolveIT, GmbH.
Acknowledgements
[15] I. Reulecke, G. Lange, J. Albrecht, R. Klein, M. Rarey, ChemMedChem
2008, 3, 885–897.
[16] N. Schneider, S. Hindle, G. Lange, R. Klein, J. Albrecht, H. Briem, K. Beyer,
H. Claußen, M. Gastreich, C. Lemmen, M. Rarey, J. Comput.-Aided Mol.
Des. 2012, 26, 701–723.
[17] H. Zhao, L. P. S. De Carvalho, C. Nathan, O. Ouerfelli, Bioorg. Med. Chem.
Lett. 2010, 20, 6472–6474.
[18] J. Ko, J. Ham, I. Yang, J. Chin, S. J. Nam, H. Kang, Tetrahedron Lett. 2006,
47, 7101–7106.
[19] S. Mann, C. Perez Melero, D. Hawksley, F. J. Leeper, Org. Biomol. Chem.
2004, 2, 1732–1741.
[20] O. P. Kuipers, P. G. G. de Ruyter, M. Kleerebezem, W. M. de Vos, J. Biotech-
nol. 1998, 64, 15–21.
The research leading to these results has received funding from
the European Community’s Seventh Framework Programme
(FP7/2007–2013) under BioStruct-X (grant agreement N8283570),
the Ministry of Education, Culture and Science (Gravitation pro-
gram 024.001.035), the Netherlands Organisation for Scientific
Research (NWO) (NWO ChemThem grant 728.011.104 and NWO
Vici grant 865.11.001), and the European Research Council (ERC)
(ERC Starting Grant 282083). The Deutsches Elektronen-Synchro-
tron (DESY) and the European Synchrotron Radiation Facility
(ESRF) are acknowledged for beam line facilities.
[21] R. P. Berntsson, N. Alia Oktaviani, F. Fusetti, A.-M. W. H. Thunnissen, B.
Poolman, D.-J. Slotboom, Protein Sci. 2009, 18, 1121–1127.
[22] W. Kabsch, Acta Crystallogr. Sect. D Biol. Crystallogr. 2010, 66, 125–132.
[23] M. D. Winn, C. C. Ballard, K. D. Cowtan, E. J. Dodson, P. Emsley, P. R.
Evans, R. M. Keegan, E. B. Krissinel, A. G. W. Leslie, A. McCoy, S. J. McNi-
cholas, G. N. Murshudov, N. S. Pannu, E. A. Potterton, H. R. Powell, R. J.
Read, A. Vagin, K. S. Wilson, Acta Crystallogr. Sect. D Biol. Crystallogr.
2011, 67, 235–242.
[24] P. D. Adams, P. V. Afonine, G. Bunkóczi, V. B. Chen, I. W. Davis, N. Echols,
J. J. Headd, L.-W. Hung, G. J. Kapral, R. W. Grosse-Kunstleve, A. J. McCoy,
N. W. Moriarty, R. Oeffner, R. J. Read, D. C. Richardson, J. S. Richardson,
T. C. Terwilliger, P. H. Zwart, Acta Crystallogr. Sect. D Biol. Crystallogr.
2010, 66, 213–221.
Keywords: energy-coupling factor transporter · membrane
proteins · molecular recognition · structure-based design · X-
ray diffraction
[1] J. ter Beek, A. Guskov, D. J. Slotboom, J. Gen. Physiol. 2014, 143, 419–
435.
[2] D. J. Slotboom, Nat. Rev. Microbiol. 2014, 12, 79–87.
[3] D. A. Rodionov, P. Hebbeln, A. Eudes, J. Ter Beek, I. A. Rodionova, G. B.
Erkens, D. J. Slotboom, M. S. Gelfand, A. L. Osterman, A. D. Hanson, T. Ei-
tinger, J. Bacteriol. 2009, 191, 42–51.
[4] T. Eitinger, D. A. Rodionov, M. Grote, E. Schneider, FEMS Microbiol. Rev.
2011, 35, 3–67.
[25] G. N. Murshudov, A. A. Vagin, E. J. Dodson, Acta Crystallogr. Sect. D Biol.
Crystallogr. 1997, 53, 240–255.
[5] C. T. Jurgenson, T. P. Begley, S. E. Ealick, Annu. Rev. Biochem. 2009, 78,
569–603.
[26] P. Emsley, B. Lohkamp, W. G. Scott, K. Cowtan, Acta Crystallogr. Sect. D
Biol. Crystallogr. 2010, 66, 486–501.
[6] K. Xu, M. Zhang, Q. Zhao, F. Yu, H. Guo, C. Wang, F. He, J. Ding, P.
Zhang, Nature 2013, 497, 268–271.
[7] T. Wang, G. Fu, X. Pan, J. Wu, X. Gong, J. Wang, Y. Shi, Nature 2013, 497,
272–276.
Received: November 24, 2014
[8] M. Zhang, Z. Bao, Q. Zhao, H. Guo, K. Xu, C. Wang, P. Zhang, Proc. Natl.
Acad. Sci. USA 2014, 111, 18560–18565.
Published online on February 12, 2015
ChemBioChem 2015, 16, 819 – 826
www.chembiochem.org
826
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim