The structural basis of modularity in ECF-type ABC transporters

Abstract

Energy coupling factor (ECF) transporters are used for the uptake of vitamins in Prokarya. They consist of an integral membrane protein that confers substrate specificity (the S-component) and an energizing module that is related to ATP-binding cassette (ABC) transporters. S-components for different substrates often do not share detectable sequence similarity but interact with the same energizing module. Here we present the crystal structure of the thiamine-specific S-component ThiT from Lactococcus lactis at 2.0 Å. Extensive protein-substrate interactions explain its high binding affinity for thiamine (K(d) ~10(-10) M). ThiT has a fold similar to that of the riboflavin-specific S-component RibU, with which it shares only 14% sequence identity. Two alanines in a conserved motif (AxxxA) located on the membrane-embedded surface of the S-components mediate the interaction with the energizing module. Based on these findings, we propose a general transport mechanism for ECF transporters.

Full text

Articles
nAture structurAl & moleculAr biology VOLUME 18 NUMBER 7 JULY 2011 755
ABC transporters catalyze the translocation of diverse compounds
across membranes and constitute one of the largest superfamilies of
proteins1. They consist of two integral membrane domains that form
a translocation pore and two nucleotide-binding domains (NBDs)
that drive transport by hydrolyzing ATP. The conserved NBDs are the
hallmark of ABC transporters, whereas the transmembrane regions
show a large variation in sequence and folds (Fig. 1a). Classical
ABC importers require additional extracellular or periplasmic
substrate-binding domains or substrate-binding proteins (SBPs) to
capture substrates, but the recently discovered ECF transporters use
S-components for substrate recognition. S-components associate with
an energizing module consisting of the membrane protein EcfT and
two identical or homologous NBDs (EcfA and EcfA′)2.
ECF-type transporters are found in Prokarya only and mediate
the uptake of vitamins and other nutrients needed in trace amounts
(such as Ni2+ or Co2+ ions)2. ECF-type ABC transporters fall into
two groups3. In group I the energizing module is used by a single
S-component (‘dedicated’ energizing modules). The biotin trans-
porter BioMNY from Rhodobacter capsulatus is the best-characterized
member of this group4. In ECF transporters of group II, the same
energizing module is shared by several different S-components with
different substrate specificities. These S-components are 20–25 kDa
in size and predicted to have 4–6 hydrophobic membrane-spanning
segments, but they are unrelated at the sequence level, and it is not
known whether they are evolutionary related. A crystal structure
(at 3.6-Å resolution) is available only for the riboflavin-specific
S-component RibU from Staphylococcus aureus (PDB 3P5N)5.
The group II proteins are particularly abundant in Gram-positive
organisms. For example, in Lactococcus lactis eight different
S-components interact with the same energizing module6, allowing
the ECF complexes to transport a wide variety of chemically dif-
ferent substrates. These complexes have a 1:1:1:1 subunit stoichio-
metry (S-component:EcfT:EcfA:EcfA′; Fig. 1a), indicating that the
S-components are integral parts of the translocating complex, rather
than peripherally associated substrate binding proteins6. Early in vivo
transport experiments in Lactobacillus casei showed that different
S-components dynamically compete for association with the energiz-
ing module7. The molecular basis for the dynamic interaction between
different S-components and the energizing module is unknown.
ThiT is the S-component involved in thiamine (vitamin B1)
transport8. It can bind thiamine with very high affinity (Kd = 120 pM)
and related compounds with nanomolar affinity. High-affinity
binding has been observed for other S-components9–11 and is most
likely important for their biological function. To understand the
molecular basis for high-affinity binding and the mechanism of trans-
port by ECF transporters, we determined the high-resolution crystal
structure of ThiT from L. lactis and present the work in the light of
previous and new biochemical data.
RESULTS
Structure determination of ThiT
We produced ThiT in L. lactis NZ9000, which has proven to be a suit-
able host for membrane protein production with properties comple-
mentary to those of Escherichia coli12. The crystal structure of ThiT
is the first of a polytopic membrane protein produced in L. lactis. We
purified ThiT bound to thiamine using the detergent n-nonyl-β--
glucopyranoside. Crystals of the native protein were formed in
space group C2 and diffracted to 2.0-Å resolution. We solved the
1University of Groningen, Groningen Biomolecular Science and Biotechnology Institute, Groningen, The Netherlands. 2University of Groningen, Zernike Institute for
Advanced Materials, Groningen, The Netherlands. Correspondence should be addressed to D.J.S. (d.j.slotboom@rug.nl).
Received 12 January; accepted 21 April; published online 26 June 2011; doi:10.1038/nsmb.2073
The structural basis of modularity in ECF-type
ABC transporters
Guus B Erkens1,2, Ronnie P-A Berntsson1,2, Faizah Fulyani1,2, Maria Majsnerowska1,2, Andreja Vujičić-Žagar1,2,
Josy ter Beek1,2, Bert Poolman1,2 & Dirk Jan Slotboom1,2
Energy coupling factor (ECF) transporters are used for the uptake of vitamins in Prokarya. They consist of an integral membrane
protein that confers substrate specificity (the S-component) and an energizing module that is related to ATP-binding cassette
(ABC) transporters. S-components for different substrates often do not share detectable sequence similarity but interact with
the same energizing module. Here we present the crystal structure of the thiamine-specific S-component ThiT from Lactococcus
lactis at 2.0 Å. Extensive protein-substrate interactions explain its high binding affinity for thiamine (Kd ~10−10 M). ThiT has a
fold similar to that of the riboflavin-specific S-component RibU, with which it shares only 14% sequence identity. Two alanines
in a conserved motif (AxxxA) located on the membrane-embedded surface of the S-components mediate the interaction with the
energizing module. Based on these findings, we propose a general transport mechanism for ECF transporters.
© 2011 Nature America, Inc. All rights reserved.
© 2011 Nature America, Inc. All rights reserved.

75 6 VOLUME 18 NUMBER 7 JULY 2011 nAture structurAl & moleculAr biology
Articles
structure (Fig. 1b and Supplementary Fig. 1) by multi-wavelength anom-
alous dispersion (MAD) phasing, using crystals of selenomethionine
(SeMet)-substituted protein (Table 1). The asymmetric unit contained
two copies of ThiT that were virtually identical (r.m.s. deviation = 0.2 Å,
Fig. 1c). The entire ThiT sequence could be fitted in the electron
density, with the exception of the N-terminal His tag and the subse-
quent five or six residues (difference between the two copies of ThiT
in the asymmetric unit), which apparently were disordered in the
crystals. Almost all residues (98%) were in the preferred regions of
the Ramachandran plot, and the remaining 2% were in the addition-
ally allowed regions. The Rwork and Rfree val-
ues after refinement were 20.4% and 23.2%,
respectively. The orientation of the two ThiT
molecules in the asymmetric unit is incom-
patible with the formation of a continuous
lipid bilayer, because the membrane plane
would have to be rotated ~145° at the dimer
interface (Fig. 1c), which is highly improb-
able. Therefore, we believe that the functional
unit of ThiT is a monomer in the absence of
the energizing module, consistent with previ-
ous light-scattering experiments8.
Well-defined non-protein electron density
became visible during refinement that could
be assigned unambiguously to thiamine
(discussed below). In addition, we found elec-
tron density that fitted acyl chains surround-
ing the hydrophobic parts of the protein. In
six cases, there was connected electron density
that could fit the headgroup of the detergent
n-nonyl-β--glucopyranoside, and in these
cases, we modeled the entire detergent mol-
ecule. In the remaining six cases, it was not
clear whether the acyl chains belonged to
the detergent or to copurified lipids. In these
cases, we modeled the acyl chains from the
detergent in the electron density.
Overall fold of ThiT and S-components
The overall fold of ThiT from L. lactis
is similar to that of RibU from S. aureus
(r.m.s. deviation = 3.5 Å for 145 Cα atoms),
α
α
π
3
10
ABC importers
TMD
SBD
NBD NBD
NBD NBD
TMD TMD TMD RibU ThiT EcfT
EcfA EcfA
Other
S
ABC exporters ECF transporters
(importers)
∼145°
L1
H1
H2
L2 L4 C terminus
Cytoplasm
N terminus
H3
L3 Outside L5
H4 H5
H6
Outside
Cytoplasm N
90°
C
ab c
d
Figure 1 The structure of ThiT. (a) Architecture of ABC and ECF
transporters. The three types of ABC transporters are SBD-dependent
ABC-type importers (left), ABC exporters (center) and the recently
discovered ECF transporters (right). The NBDs (blue) are conserved
among all ABC transporters; the TMDs (various colors) are different.
Substrates are indicated as black dots. (b) Surface (left) and
secondary structure ribbon representation (right) of ThiT. In the
surface model, hydrophobic residues are gray and hydrophilic residues are
green; positively charged residues are blue and negatively charged residues
are red (the latter are not visible in this orientation). The ribbon is colored from N-terminal blue to
C-terminal red. The bar indicates the approximate position of the lipid membrane (35 Å). The membrane
topology for ThiT is depicted below and colored as in the ribbon model. H1–H6, helices 1–6; L1–L5, loops 1–5.
(c) The dimer of ThiT in the asymmetric unit. One monomer is colored purple, the second gray and bound
thiamine in black. The dashed lines indicate the position of the membrane for both monomers. (d) The unusual structure of helix 4, colored green;
the rest of ThiT is depicted in gray. The lines indicate the vertical position of the secondary structure elements.
Table 1 Data collection, phasing and refinement statistics
Native SeMet-labeled ThiT (Se-MAD)
Data collection
Space group C2C2
Cell dimensions
a, b, c (Å) 61.4, 84.3, 127.0 65.2, 83.7, 128.5
α
,
β
,
γ
(°) 90.0, 95.7, 90.0 90.0, 95.9, 90.0
Peak Inflection Remote
Resolution (Å) 47.5–2.0 48.9–2.9 48.9–2.9 48.9–2.9
Rsym 6.6 (48.6) 6.9 (27.3) 5.1 (12.7) 4.4 (7.2)
I / σI8.4 (2.1) 15.7 (6.0) 21.2 (10.2) 26.7 (15.5)
Completeness (%) 99.0 (96.6) 99.9 (99.9) 99.9 (100) 99.9 (100)
Redundancy 3.7 6.9 6.9 6.9
Refinement
Resolution (Å) 47.5–2.0
No. reflections 41,123
Rwork / Rfree 20.4 / 23.2
No. atoms
Protein 2,748
Thiamine 36
Water 73
B-factors
Protein 41
Thiamine 36
Water 51
R.m.s. deviations
Bond lengths (Å) 0.018
Bond angles (°) 1.68
Values in parentheses are for highest-resolution shell.
© 2011 Nature America, Inc. All rights reserved.
© 2011 Nature America, Inc. All rights reserved.

nAture structurAl & moleculAr biology VOLUME 18 NUMBER 7 JULY 2011 757
Articles
although at the sequence level the two proteins are unrelated (14%
sequence identity). In the past few years, a growing number of mem-
brane protein structures have been determined that share a related
fold without being related in sequence (for example, the LeuT and
the aquaporin folds13,14). These observations raise noteworthy
questions about the evolution of membrane proteins and the rela-
tion between primary and tertiary structures. The lack of sequence
conservation between S-components of the ECF transporters is all
the more unexpected, because these proteins interact with a common
partner, the shared ECF energizing module.
ThiT contains six hydrophobic helical segments that cross the
membrane (Fig. 1b). A part of the L1 loop is also embedded in the
lipid bilayer, which is necessary because helix 2 is too short to span
the entire thickness of the membrane. The position of L1 may play an
important mechanistic role in the translocation of thiamine across the
membrane (see Discussion). The structure of helix 4 is highly irregu-
lar. It starts with the backbone hydrogen-bond pattern of a regular
α-helix, then turns into a π-helix15, returns to an α-helical confor-
mation, and finally continues as a long (seven residues) 310-helix16
(Fig. 1d, helix in green). This unusual combination of structural fea-
tures has not been observed in any other membrane protein structure.
For example, in the RibU structure, helix 4 is a regular transmembrane
α-helix. The π-bulge irregularity is important for ligand binding and
will be discussed in more detail below. The 310-helical segment allows
a very tight packing of transmembrane segment 4 with helices 2 and 3,
and—to a lesser extent—helix 5. This close packing is further facili-
tated by the presence of numerous conserved glycines in helices 3, 4
and 5 and in loop L2 (Supplementar y Fig. 2).
Structural basis of high-affinity thiamine binding
The thiamine-binding site is located in a pocket near the extracellular
side of the membrane and lined by helices 4, 5 and 6 and the loops L1
and L5. The thiamine molecule was modeled in clear electron density
(Fig. 2a) and has a conformation that is different from the catalytic
V-shaped conformation observed in enzymes that use thiamine pyrophos-
phate as a cofactor17. A large number of interactions shape the binding site
and account for the high binding affinity (Fig. 2b and Supplementary
Fig. 3). Glu84 in helix 4 stabilizes the positively charged N2 of the
thiazole ring, and the adjacent residue Tyr85 forms a hydrogen bond
with the hydroxyl group of thiamine. Both residues are located in helix 4,
and their favorable orientation is dependent on the π-bulge irregular-
ity in this helix. Hydrogen bonds are formed by Tyr146 (through an
ordered water molecule) and Asn151 in helix 6 with the N1 and N3 of
the pyrimidine ring, respectively. Trp34 in L1 and His125 (helix 5) are
involved in aromatic π-stacking on opposite sides of the thiazole ring,
and Trp133 in L5 stacks with the pyrimidine ring. Gly129 allows the
pyrimidine ring to pack closely against the C-terminal end of helix 5.
The side chains that interact directly with the substrate are held in place
by an intricate network of hydrogen bonds and aromatic interactions
with other binding residues. In addition, more distant residues that are
not directly involved in substrate coordination contribute to this net-
work (Supplementary Table 1). The structure of the thiamine-binding
site is in excellent agreement with previous mutagenesis studies8:
mutations of Trp133, Gly129, Asn151 and Tyr146 to alanine reduced
the binding affinity by a factor of between 20 and 1,000. In contrast,
the W34A mutant still bound thiamine with wild-type affinity. Trp34
acts as a lid on the binding site, and its removal apparently has little
effect on the rest of the binding site.
The hydroxyl group of thiamine is accessible from the extracellu-
lar environment through a narrow opening expanding into a cavity
(Fig. 2c). The cavity does not allow the substrate to enter but provides
sufficient space to accommodate the phosphate moieties of thiamine
monophosphate and thiamine pyrophosphate (TMP and TPP), which
also bind with high affinity to ThiT8. Entrance of thiamine into the
binding site from the external side of the membrane would require
conformational changes of loops L1, L3 and L5, which form a cage of
aromatic side chains on top of the substrate.
Substrate transport by ThiT requires the energizing module
The residues involved in thiamine binding are highly conserved among
ThiT orthologs (Supplementary Fig. 2). In Figure 2c, the degree of
conservation is projected on the ThiT structure. In addition to the bind-
ing site residues, a few other amino acids are also strongly conserved
(for example, Pro43 at the beginning of helix 2 and several glycines in
L2, helix 3 and the 310 part of helix 4). These residues are likely to have
a structural role (causing close packing of helices 3 and 4 and capping
the short helix 2). We do not observe an obvious translocation path
lined with conserved amino acids within ThiT, in contrast to what has
been suggested for RibU5 (see Discussion). The absence of a transloca-
tion path is consistent with thiamine-transport assays (Fig. 3). When
expressed in E. coli, ThiT alone did not support thiamine transport,
but coexpression of the energizing module (EcfAA′T) and ThiT led to
robust thiamine uptake. Overexpression of ThiT in L. lactis resulted in
a
b
c
90°
Degree of conservation
Low High
NC
Tyr85
Tyr146
Trp133
Glu84
His125 Asn151
Trp34
*
Figure 2 The high-affinity thiamine-binding site. (a) Electron density
for thiamine shown in gray mesh (2Fo – Fc map contoured at 1.5σ),
with the modeled thiamine molecule. (b) Residues forming hydrogen
bonds and aromatic interactions with thiamine. Carbon atoms of the
thiamine molecule and side chains of the binding residues are shown in
green and blue, respectively. Hydrogen bonds are indicated by the red
dashes. Tyr146 interacts with thiamine through a water molecule (black
asterisk). (c) Ribbon and sliced-surface models of the ThiT structure with
the conserved amino acids in ThiT homologs colored according to their
conservation score. The arrow indicates the access to the cavity that can
accommodate phosphate moieties of TMP and TPP.
© 2011 Nature America, Inc. All rights reserved.
© 2011 Nature America, Inc. All rights reserved.

75 8 VOLUME 18 NUMBER 7 JULY 2011 nAture structurAl & moleculAr biology
Articles
increased levels of thiamine binding to the cells but not in increased
transport rates. In L. lactis the energizing module is constitutively
expressed from the chromosomal copy of the ecfAA
′
T genes, and the
results show that thiamine transport in L. lactis is limited by the amount
of EcfAA′T rather than by the amount of ThiT. These experiments,
together with previous data8, show that ThiT binds thiamine but does
not translocate the vitamin in the absence of the energizing module.
Interaction with the EcfT subunit
The lack of sequence similarity between ThiT and RibU is unusual,
because both proteins interact with ECF energizing modules, but as
the two proteins are from different organisms, they do not interact
with the same energizing module. However, ThiT and RibU proteins
from the same organism are also unrelated in sequence (for example,
in L. lactis there is only 16% identity between them). Because of the
absence of detectable sequence conservation between the two different
S-components, we searched for structural motifs that could be the dock-
ing sites for the energizing module. Interaction with the hydrophobic
EcfT subunit is expected to take place within the hydrophobic core of the
lipid bilayer. A superposition of the ThiT and RibU structures (Fig. 4a)
revealed that helices 1, 2, 3—and to a lesser extent helix 6—align well,
but that helices 4 and 5 adopt very different conformations. The struc-
tural variability in the latter region is required for correct positioning of
the side chains that create the binding sites for either thiamine or ribofla-
vin, and this makes it unlikely that the EcfT component interacts here.
In contrast, the surface formed by helices 1, 2, 3 and 6 is very simi-
lar in RibU and ThiT. Prompted by this observation, we searched for
patterns of sequence conservation in this region that might have been
overlooked previously. Sequence comparison of all eight S-components
from L. lactis revealed that there is a shared alanine motif (AxxxA,
where x can be any amino acid) on the exposed face of helix 1 (Fig. 4b,
Supplementary Fig. 4). Such tetrad repeats of small amino acids are well
known to promote helix-helix interactions between membrane proteins
in the lipid bilayer18–20. To test whether the alanine motif was involved
in the interaction with EcfT, we separately mutated the two alanines
into tryptophan residues and coexpressed the ThiT mutants with the
energizing module in E. coli. Both mutations (A15W and A19W) lead
to a complete loss of transport activity (Fig. 4c), even though the ThiT
mutants were expressed to at least the same level as the wild type (Fig. 4c,
inset). Notably, the integrity and thiamine-binding capacity of the ThiT
a b c d
Lipid bilayer interface
Interaction with
energizing module
A19W
A15W
WT
Alanine
motif
L1 loop
A19W
Membranes
Membranes
Membranes
Elution
Elution
Elution
A15WWT
ThiT-Strep
[
3
H]Thiamine uptake (pmol (D
600 nm
.ml)
–1
)
ThiT-
Strep
1.5
5
5
4
4
6
2
3
1
1.0
0.5
0
0 10 20 30 40
Time (min)
EcfT-His
Figure 4 Interaction of ThiT with the energizing module. (a) Superposition of the RibU structure in gray (PDB 3P5N)5 on the ThiT structure (colored as
in Fig. 1b). The dashed line indicates the proposed interface with the energizing module. The helices are numbered from N terminus to C terminus as
in Figure 1b. (b) The ThiT structure as seen from the interface with the energizing module. The surface of the L1 loop region is highlighted in blue and
indicated by the dashed circle. Rearrangement of the L1 loop would expose the bound thiamine (black sticks) to the lateral EcfT interface. The alanine
motif in helix 1 that is shared by all S-components in L. lactis is colored red. (c) Thiamine uptake by recombinant E. coli cells coexpressing EcfAA′T with wild-
type ThiT (), A15W ThiT (n) or A19W ThiT (). Thiamine uptake by a control strain harboring an empty plasmid is indicated by the black circles (•).
The error bars indicate the upper and lower measured values. The inset shows western blot analyses of the expression levels of His-tagged EcfT (using
antibodies directed against the His tag) and Strep-tagged ThiT (using antibodies against the Strep tag). (d) Pull-out experiment. EcfAA′T with a His
tag on EcfT was coexpressed with Strep-tagged wild-type ThiT, or variants A15W or A19W in E. coli as in c. Membranes were solubilized and the
complexes were purified using nickel-affinity chromatography. The elution fractions were analyzed by SDS-PAGE followed by western-blot analysis
using antibodies against ThiT-strep.
a b
1.0 3.5
0.8
0.6
[3H]Thiamine uptake (pmol (D600 nm.ml)–1)
[3H]Thiamine uptake (pmol (D600 nm.ml)–1)
0.4
0.2
0
0 10 20 30 40
Time (min) Time (min)
0
3.0
2.5
2.0
1.5
1.0
0.5
0
12345
Figure 3 Transport of [3H]thiamine in E. coli and L. lactis cells.
(a) Thiamine uptake by recombinant E. coli cells. E. coli cells
expressing ThiT and EcfAA′T from L. lactis (•) or ThiT alone (),
and control cells containing an empty expression plasmid (), were
assayed for thiamine uptake. All cells were energized with glucose.
(b) Thiamine uptake by recombinant L. lactis cells. De-energized
control cells (harboring an empty plasmid but containing chromosomal
copies of the genes thiT (also known as llmg_0334) and ecfAA
′
T (cbiO,
cbiO, cbiQ2) (), de-energized cells expressing ThiT from a plasmid (∇),
energized control cells (•) and energized cells expressing ThiT () were
assayed for thiamine uptake. Thiamine binding, rather than transport,
was observed in the de-energized cells. The levels of binding depended
on the expression levels of ThiT. In the energized cells harboring the
empty plasmid, rapid thiamine uptake was observed. In energized cells
overexpressing ThiT, the offset on the y axis—indicative of binding—
increased, rather than the uptake rate. All experiments were conducted
at least in duplicate. The error bars indicate the range (a) or s.d. (b).
© 2011 Nature America, Inc. All rights reserved.
© 2011 Nature America, Inc. All rights reserved.

nAture structurAl & moleculAr biology VOLUME 18 NUMBER 7 JULY 2011 759
Articles
mutants were unaffected by the mutations: the solitary ThiT mutants
(in the absence of the energizing module) could be overexpressed and
purified in similar amounts as the wild-type ThiT and bound thiamine
with high affinity (Supplementary Fig. 5). We conclude that the muta-
tions in the alanine motif disrupt the functional interaction between
ThiT and the energizing module.
To show that the mutations not only affected the functional inter-
action between the energizing module and ThiT but also affected
complex formation, we conducted a pull-out experiment. We pre-
viously showed that the entire complex (EcfAA′T–ThiT) could be
pulled out by nickel-affinity chromatography using a His tag on EcfT6.
Here we show that in contrast to wild-type ThiT, the mutants A15W
and A19W were not enriched together with EcfT (Fig. 4d), indicating
that the interaction between ThiT and the energizing module had
indeed been disrupted by the single amino acid substitutions.
DISCUSSION
Interaction with the energizing module
GxxxG and AxxxA motifs are frequently found in membrane proteins
and are often involved in helix-helix interactions18–20. The results
presented here show that the conserved alanine motif (AxxxA) on the
exposed, membrane-embedded face of helix 1 in ThiT mediates the
interaction with the EcfT protein. Our observation that a single amino
acid substitution (either A15W or A19W) is sufficient to disrupt the
functional and physical interaction between ThiT and the energizing
module provides a first clue about the mechanism of interaction.
It will require mapping the complete interface to gain a full under-
standing of the forces driving the association in the complex.
Interaction with the nucleotide binding domains
The free energy required for the transport of thiamine comes from
the hydrolysis of ATP in the EcfA subunits. In all ABC transporters
for which crystal structures are available, each NBD contains a groove
that binds to a cytoplasmic segment of the membrane domain. The
structural elements in the membrane domains that fit in these grooves
are short helical segments named ‘coupling helices’. The interaction by
means of coupling helices allows ATP hydrolysis to be linked to trans-
port21,22. Because EcfA and EcfA′ possess all the sequence motifs that
are mechanistically important in NBDs associated with ABC trans-
porters22–28 (Supplementary Fig. 6), it is reasonable to assume that the
two proteins also communicate with the membrane subunits by means
of coupling helices, but neither ThiT nor RibU has an obvious coupling
helix. To explain this paradox, we propose that the EcfT subunit of the
energizing module may contain two coupling helices. EcfT proteins
are predicted to have a long and conserved cytoplasmic loop with two
moderately hydrophobic helical segments3,29 (Supplementary Fig. 7).
The size of this cytoplasmic domain (109 amino acids) allows the pres-
ence of two rather than one coupling helix for interaction with both
EcfA and EcfA′ (Fig. 5a). The free energy released by ATP hydrolysis
is then transferred through the EcfT subunit to the S-components.
This hypothesis is supported by mutagenesis studies showing that two
conserved motifs (30–40 amino acids apart) in the cytoplasmic loops
of different EcfT proteins are important for stabilizing the energizing
module–S-component complexes29. The absence of a coupling helix in
the S-components might facilitate their exchange from complexes with
the shared energizing module. Such a dynamic interaction was already
suggested in the 1970s, as a result of in vivo transport experiments7.
Based on the RibU structure5, it has been suggested that clusters of
positively charged amino acids in the cytoplasmic loops of ThiT and
RibU might mediate the interaction with the EcfA subunit. However,
we suspect that the clustering of arginines and lysines in cytoplas-
mic loops is more likely a manifestation of the ‘positive inside rule’:
membrane proteins are usually enriched for these residues in their
cytoplasmic loops, and this bias predicts the membrane topology30,31.
The substrate translocation path
We do not observe an obvious translocation path for thiamine within the
ThiT molecule. The substrate binding site is located near the extracel-
lular side of the membrane, and the tight helix packing does not allow a
substrate to pass through the interior of the protein. Although the same
is true for the RibU structure, some researchers speculate on a trans-
port path for riboflavin through the core of the cylinder-shaped RibU
molecule5 that would be opened by a conformational change and that is
lined with moderately conserved amino acids. In the ThiT structure, we
do not see a similar arrangement of conserved residues. Furthermore, a
number of amino acids in RibU that were classified as conserved on the
basis of a multiple sequence alignment with 12 orthologs are not well
conserved when a much larger set of RibU sequences is used, whereas
the binding site residues are still conserved.
The absence of an intramolecular translocation path in ThiT is
consistent with the lack of thiamine transport activity in the absence
of the energizing module (Fig. 3). Similarly, for riboflavin trans-
port by RibU from S. aureus, the energizing module was required5
as well. We hypothesize that instead of using an intramolecular
translocation path within the S-components, the substrates are
translocated at the interface between the S-component and EcfT,
in line with the mechanism that classical ABC transporters use to
import substrates1. Contrary to ThiT and RibU, the biotin-specific
S-component BioY from R. capsulatus has been proposed to have a
homo-oligomeric quaternary structure in vivo32 and may facilitate
biotin transport in the absence of the energizing module5.
Interaction of ThiT with the EcfT subunit on the surface formed by heli-
ces 1, 2, 3 and 6 (Fig. 4a) immediately suggests a mechanism for substrate
transport to the cytoplasm. Rearrangement of the membrane-embedded
L1 loop, instigated by ATP binding and hydrolysis in the NBDs, could open
a lateral gate for thiamine, facing the EcfT subunit (Fig. 4b). Furthermore,
the L1 loop interacts intimately with three regions (helices 5 and 6 and L3)
that contain substrate-binding residues (Fig. 5b). Repositioning of L1 will
disrupt these interactions and perturb the binding site, thereby reducing
the binding affinity and allowing thiamine to leave.
Symmetry of the membrane domains
All available crystal structures of ABC transporters show structural sym-
metry between the two transmembrane subunits22–28. The symmetry is
Coupling helices
H6
H5
H2
Glu38
Thr158
L1 loop
Ser154
Tyr122
Lys121
a b
Figure 5 Working model for substrate translocation and interaction with
EcfA. (a) Model for coupling helix interaction in ECF transporters. The
S-component is colored red, EcfT is orange and the EcfA subunits are
blue. (b) Specific interactions between the L1 loop and helices 5 and 6.
Glu38 in L1 forms a salt bridge with Lys121 in helix 5 and a hydrogen
bond with Tyr122; the side chains of Ser154 and Thr158 in helix 6 form
hydrogen bonds with backbone NH groups in loop L1. In addition, Trp34
in L1 makes an aromatic interaction with Tyr74 in L3 (not shown).
© 2011 Nature America, Inc. All rights reserved.
© 2011 Nature America, Inc. All rights reserved.

76 0 VOLUME 18 NUMBER 7 JULY 2011 nAture structurAl & moleculAr biology
most obvious when the two subunits are identical but is also present in
the case of heterodimers, where the folds are related . We were not able to
detect sequence similarity between the S-components and EcfT, but we
cannot exclude the possibility that these proteins are structurally similar
but their sequences have diverged beyond recognition . However, the
predicted topology and structural organization of EcfT proteins are dif-
ferent from those of the S-components29. Together with our hypothesis
that the EcfT component may contain two coupling helices for interac-
tion with the EcfA subunits, these data suggest that ECF transporters are
less likely to be symmetrical than classical ABC transporters .
Conclusion
The high-resolution structural data presented here, together with the
RibU structure, provide the first glimpse of the transport mechanism
of ECF transporters. Further structural and biochemical studies of the
complete complexes will now be necessary for full understanding. ECF
transporters are exclusively prokaryotic and numerous human patho-
gens are dependent on the uptake of ECF substrates for survival33,34.
For example, ThiT from the human pathogen Listeria monocytogenes
has proven to be essential for intracellular replication35. Structural
and mechanistic understanding of ECF transporters may enable the
development of new antibiotics that target these proteins.
METHODS
Methods and any associated references are available in the online
version of the paper at http://www.nature.com/nsmb/.
Accession codes. The coordinates of the ThiT structure have been
deposited in the Protein Data Bank under accession code 3RLB.
Note: Supplementary information is available on the Nature Structural
&
Molecular
Biology website.
ACKNOWLEDGMENTS
We thank the European Synchrotron Radiation Facility and Swiss Light Source
for providing excellent beamline facilities. We thank R. Duurkens for conducting
transport experiments, D. Colpa for assistance with the pull-out experiments and
A.-M. Thunnissen for critically reading the manuscript. This research was supported
by the Netherlands Organization for Scientific Research (NWO) (Vidi and ALW Open
Programma grants to D.J.S., TOP subsidy grant 700.56.302 to B.P. and Top Talent
grant to J.t.B.) and by the European Union European Drug Initiative on Channels and
Transporters (EDICT) program.
AUTHOR CONTRIBUTIONS
G.B.E., R.P.-A.B. and D.J.S. designed the experiments. G.B.E., R.P.-A.B., F.F., M.M.,
A.V.-Z. and J.t.B. conducted the experiments. G.B.E., R.P.-A.B., B.P. and D.J.S.
analyzed the data. G.B.E., B.P. and D.J.S. wrote the manuscript.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
Published online at http://www.nature.com/nsmb/.
Reprints and permissions information is available online at http://www.nature.com/
reprints/index.html.
1. Davidson, A.L., Dassa, E., Orelle, C. & Chen, J. Structure, function, and evolution
of bacterial ATP-binding cassette systems. Microbiol. Mol. Biol. Rev. 72, 317–364
(2008).
2. Rodionov, D.A. et al. A novel class of modular transporters for vitamins in prokaryotes.
J. Bacteriol. 191, 42–51 (2009).
3. Eitinger, T., Rodionov, D.A., Grote, M. & Schneider, E. Canonical and ECF-type
ATP-binding cassette importers in prokaryotes: diversity in modular organization and
cellular functions. FEMS Microbiol. Rev. 35, 3–76 (2011).
4. Hebbeln, P., Rodionov, D.A., Alfandega, A. & Eitinger, T. Biotin uptake in prokaryotes
by solute transporters with an optional ATP-binding cassette-containing module.
Proc. Natl. Acad. Sci. USA 104, 2909–2914 (2007).
5. Zhang, P., Wang, J. & Shi, Y. Structure and mechanism of the S component of a
bacterial ECF transporter. Nature 468, 717–720 (2010).
6. ter Beek, J., Duurkens, R.H., Erkens, G.B. & Slotboom, D.J. Quaternary structure and
functional unit of energy coupling factor (ECF)-type transporters. J. Biol. Chem. 286,
5471–5475 (2011).
7. Henderson, G.B., Zevely, E.M. & Huennekens, F.M. Mechanism of folate transport
in Lactobacillus casei: evidence for a component shared with the thiamine and
biotin transport systems. J. Bacteriol. 137, 1308–1314 (1979).
8. Erkens, G.B. & Slotboom, D.J. Biochemical characterization of ThiT from
Lactococcus lactis: a thiamin transporter with picomolar substrate binding affinity.
Biochemistry 49, 3203–3212 (2010).
9. Duurkens, R.H., Tol, M.B., Geertsma, E.R., Permentier, H.P. & Slotboom, D.J. Flavin
binding to the high affinity riboflavin transporter RibU. J. Biol. Chem. 282,
10380–10386 (2007).
10. Eudes, A. et al. Identification of genes encoding the folate- and thiamine-binding
membrane proteins in Firmicutes. J. Bacteriol. 190, 7591–7594 (2008).
11. Henderson, G.B., Kojima, J.M. & Kumar, H.P. Differential interaction of cations
with the thiamine and biotin transport proteins of Lactobacillus casei. Biochim.
Biophys. Acta 813, 201–206 (1985).
12. Kunji, E.R., Slotboom, D.J. & Poolman, B. Lactococcus lactis as host for
overproduction of functional membrane proteins. Biochim. Biophys. Acta 1610,
97–108 (2003).
13. Waight, A.B., Love, J. & Wang, D.N. Structure and mechanism of a pentameric
formate channel. Nat. Struct. Mol. Biol. 17, 31–37 (2010).
14. Theobald, D.L. & Miller, C. Membrane transport proteins: surprises in structural
sameness. Nat. Struct. Mol. Biol. 17, 2–3 (2010).
15. Cartailler, J.P. & Luecke, H. Structural and functional characterization of π bulges
and other short intrahelical deformations. Structure 12, 133–144 (2004).
16. Vieira-Pires, R.S. & Morais-Cabral, J.H. 310 helices in channels and other membrane
proteins. J. Gen. Physiol. 136, 585–592 (2010).
17. Shin, W., Pletcher, J., Blank, G. & Sax, M. Ring stacking interactions between
thiamin and planar molecules as seen in the crystal structure of a thiamin picrolonate
dihydrate complex. J. Am. Chem. Soc. 99, 3491–3499 (1977).
18. Lemmon, M.A., Flanagan, J.M., Treutlein, H.R., Zhang, J. & Engelman, D.M.
Sequence specificity in the dimerization of transmembrane alpha-helices.
Biochemistry 31, 12719–12725 (1992).
19. Russ, W.P. & Engelman, D.M. The GxxxG motif: a framework for transmembrane
helix-helix association. J. Mol. Biol. 296, 911–919 (2000).
20. Schneider, D., Finger, C., Prodohl, A. & Volkmer, T. From interactions of single
transmembrane helices to folding of alpha-helical membrane proteins: analyzing
transmembrane helix-helix interactions in bacteria. Curr. Protein Pept. Sci. 8,
45–61 (2007).
21. Hollenstein, K., Dawson, R.J. & Locher, K.P. Structure and mechanism of ABC
transporter proteins. Curr. Opin. Struct. Biol. 17, 412–418 (2007).
22. Locher, K.P., Lee, A.T. & Rees, D.C. The E. coli BtuCD structure: a framework
for ABC transporter architecture and mechanism. Science 296, 1091–1098
(2002).
23. Aller, S.G. et al. Structure of P-glycoprotein reveals a molecular basis for poly-
specific drug binding. Science 323, 1718–1722 (2009).
24. Hollenstein, K., Frei, D.C. & Locher, K.P. Structure of an ABC transporter in complex
with its binding protein. Nature 446, 213–216 (2007).
25. Gerber, S., Comellas-Bigler, M., Goetz, B.A. & Locher, K.P. Structural basis of
trans-inhibition in a molybdate/tungstate ABC transporter. Science 321, 246–250
(2008).
26. Kadaba, N.S., Kaiser, J.T., Johnson, E., Lee, A. & Rees, D.C. The high-affinity
E. coli methionine ABC transporter: structure and allosteric regulation. Science 321,
250–253 (2008).
27. Oldham, M.L., Khare, D., Quiocho, F.A., Davidson, A.L. & Chen, J. Crystal structure
of a catalytic intermediate of the maltose transporter. Nature 450, 515–521
(2007).
28. Dawson, R.J. & Locher, K.P. Structure of a bacterial multidrug ABC transporter.
Nature 443, 180–185 (2006).
29. Neubauer, O. et al. Two essential arginine residues in the T components
of energy-coupling factor transporters. J. Bacteriol. 191, 6482–6488
(2009).
30. von Heijne, G. & Gavel, Y. Topogenic signals in integral membrane proteins.
Eur. J. Biochem. 174, 671–678 (1988).
31. von Heijne, G. Control of topology and mode of assembly of a polytopic membrane
protein by positively charged residues. Nature 341, 456–458 (1989).
32. Finkenwirth, F. et al. Subunit composition of an energy-coupling-factor-type biotin
transporter analysed in living bacteria. Biochem. J. 431, 373–380 (2010).
33. Glass, J.I. et al. Essential genes of a minimal bacterium. Proc. Natl. Acad.
Sci. USA 103, 425–430 (2006).
34. Thanassi, J.A., Hartman-Neumann, S.L., Dougherty, T.J., Dougherty, B.A. & Pucci, M.J.
Identification of 113 conserved essential genes using a high-throughput gene disruption
system in Streptococcus pneumoniae. Nucleic Acids Res. 30, 3152–3162 (2002).
35. Schauer, K., Stolz, J., Scherer, S. & Fuchs, T.M. Both thiamine uptake and
biosynthesis of thiamine precursors are required for intracellular replication of
Listeria monocytogenes. J. Bacteriol. 191, 2218–2227 (2009).
Articles
© 2011 Nature America, Inc. All rights reserved.
© 2011 Nature America, Inc. All rights reserved.

nAture structurAl & moleculAr biology
doi:10.1038/nsmb.2073
ONLINE METHODS
Protein overexpression. Native ThiT containing an N-terminal His8 tag
(ThiT-nHis) was overexpressed in L. lactis strain NZ9000 (ref. 36). The cells were
grown semi-anaerobically in GLS medium (2% (w/v) Gistex (Brenntag), 2.5% (w/v)
glucose, 100 mM KH2PO4, 110 mM K2HPO4 and 5 µg ml−1 chloramphenicol).
The initial pH was 6.8 and decreased during cell growth. At D600 nm of 1.5 (at this
point the pH was 6.5), expression was induced by the addition of 0.1% (v/v) of
culture supernatant from the nisin A–producing strain NZ9700 (ref. 36). The cells
were induced for 2 h and reached a final D600 nm of 4–5. The preparation of mem-
brane vesicles was carried out as previously described8. Expression of SeMet (Acros
Organics)-substituted ThiT was done in L. lactis as previously described37.
Purification of thiamine-bound ThiT-nHis. ThiT-nHis was purified as pre-
viously described8 with the following modifications: during solubilization,
100 µM thiamine-HCl (Sigma) was added to ensure saturation of all binding sites
with thiamine. Membrane vesicles from L. lactis NZ9000, expressing ThiT, were
solubilized in 1.0% (w/v) n-dodecyl-β--maltopyranoside (DDM, Anatrace),
but in all subsequent steps, this detergent was replaced by 0.35% n-nonyl-β-
-glucopyranoside (NG, Anatrace). Size-exclusion chromatography (SEC) was
done on a Superdex-200 column (GE Healthcare) in 20 mM HEPES buffer,
150 mM NaCl and 0.35% NG (pH 7.0, adjusted with NaOH). The peak frac-
tions after SEC were concentrated on a Vivaspin 30-kDa molecular weight cutoff
(MWCO) concentrator (VWR International) to 6–8 mg ml−1. Concentrated ThiT
was used directly to set up crystallization trials.
Crystallization. Initial crystals of ThiT were obtained under several conditions
by screening commercially available crystallization conditions with ThiT puri-
fied in n-octyl-β--glucopyranoside (OG, Anatrace), using a dispensing robot
(mosquito, TTP Labtech). These crystals were small and diffracted only to ~50 Å.
Optimization of the crystallization conditions gradually improved the diffrac-
tion properties and finally yielded crystals diffracting to 7–8 Å. Rescreening
with the detergent n-octyl-β--thioglucopyranoside (OTG, Anatrace) resulted
in bigger crystals that diffracted up to 5–6 Å. A major improvement was obtained
with ThiT purified in NG. Crystals ranging in size from 50 to 300 µm could be
grown at 5 °C from a solution containing 15–20% (w/v) PEG 3350 (Hampton
Research) and 0.1–0.3 M NH4NO3. The crystals appeared within 1 week, grew
to full size in 3–4 weeks and diffracted to ~2 Å. For cryoprotection, a solution
of 40% (w/v) PEG 3350 was prepared with the same concentration of NH4NO3
as for the crystallization condition. Replacing PEG 3350 by PEGs with a higher
or lower molecular weight resulted in crystals in most cases, but the best
diffraction properties were obtained with PEG 3350. SeMet-substituted ThiT-nHis
could be purified and crystallized under conditions identical to those for the
native protein.
Structure determination. Diffraction data were collected at the ESRF and
SLS beamline facilities. Multi-wavelength Anomalous Dispersion (MAD) data
on SeMet-ThiT to 2.9 Å were collected at 100 K on ID29 at ESRF around the
K-absorption edge of selenium with wavelengths for remote of 0.9768 Å, for
inflection of 0.9793 Å and for peak of 0.9791 Å, in that order. Native data to 2.0 Å
were collected at 100 K and 1.0723 Å on ID23-1. Data processing and reduction
were carried out using XDS38 and programs from the CCP4 suite39. Relevant
statistics for the data collection, phasing and model refinement can be found in
Table 1. Initial phase information was found and the initial model built using
Phenix AutoSol40 and Resolve (within Phenix). Four selenium sites were found
within the asymmetric unit, corresponding to two SeMet substitutions per protein
molecule (Met17 and Met68). All SeMet peaks were above 25σ (Supplementary
Fig. 8a). The full model was built in ARP/wARP41, using the native data. A few
cycles of refinement in Refmac5 (ref. 42), including noncrystallographic symme-
try with loose restraints, interspersed with manual model building using Coot43,
were necessary to complete the model (Supplementary Fig. 8b). The final protein
model contains residues 7–182 for chain A and 6–182 for chain B; thus, only
the initial five or six residues and the His tag are missing. Water molecules were
automatically placed in Fo – Fc Fourier difference maps at a 3σ cutoff level and
validated to ensure correct coordination geometries, using Coot. All structure
figures were prepared with PyMOL (http://www.pymol.org/).
Transport of [3H]thiamine in L. lactis and E. coli. Transport experiments with
L. lactis cells were conducted as previously described8. To de-energize the cells,
we added 20 mM N-methyl-α--glucopyranoside instead of glucose. The cells
were grown in chemically defined medium without thiamine. The scarcity of
thiamine in the growth medium induces the expression of the chromosomal thiT
gene. The energizing module (EcfAA′T) was expressed constitutively from the
chromosomal copies of the genes.
E. coli MC1061 cells containing plasmids for expression of ThiT alone or both
EcfAA′T and ThiT6 were grown on LB medium with 100 µg ml−1 ampicillin. At
an D600 nm of ~0.5, expression was induced by adding 10−3% (w/v) -arabinose.
After 2 h of induction, the cells were harvested, washed and resuspended in ice-
cold buffer (50 mM potassium phosphate, pH 7.5) to a final D600 nm of 5 and kept
on ice. For the transport assays, the cells were energized with 10 mM glucose for
15 min at 30 °C. Subsequently, [3H]thiamine (American Radiolabeled Chemicals)
was added to a final concentration of 25 nM, and at the indicated time points,
200-µl samples were taken and mixed with 2 ml stop buffer (ice-cold 50 mM
potassium phosphate, pH 7.5). The suspension was rapidly filtered over a BA-85
nitrocellulose filter, which was subsequently washed once with 2 ml stop buffer.
Filters were dried for 1 h at 80 °C, and 2 ml of Emulsifier-Scintillator Plus liquid
(PerkinElmer) was added. The levels of radioactivity were determined with a
PerkinElmer Tri-Carb 2800 TR isotope counter. For time point zero, 200 µl of
cell suspension was added to 2 ml stop buffer containing radioactive thiamine,
and this mixture was directly filtered.
Purification of EcfAA′T–ThiT. The experiments were conducted as previously
described6 with wild-type ThiT, ThiT A15W and ThiT A19W coexpressed with
the energizing module. The ThiT mutants were prepared using standard clon-
ing techniques.
36. Kuipers, O.P., de Ruyter, P.G.G.A., Kleerebezem, M. & de Vos, W.M. Quorum sensing-
controlled gene expression in lactic acid bacteria. J. Biotechnol. 64, 15–21
(1998).
37. Berntsson, R.P. et al. Selenomethionine incorporation in proteins expressed in
Lactococcus lactis. Protein Sci. 18, 1121–1127 (2009).
38. Kabsch, W. Automatic processing of rotation diffraction data from crystals of initially
unknown symmetry and cell constants. J. Appl. Cryst. 26, 795–800 (1993).
39. Collaborative Computational Project Number 4. The CCP4 suite: programs for protein
crystallography. Acta Crystallogr. D Biol. Crystallogr. 50, 760–763 (1994).
40. Adams, P.D. et al. PHENIX: a comprehensive Python-based system for macromolecular
structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221 (2010).
41. Langer, G., Cohen, S.X., Lamzin, V.S. & Perrakis, A. Automated macromolecular
model building for X-ray crystallography using ARP/wARP version 7. Nat. Protoc. 3,
1171–1179 (2008).
42. Murshudov, G.N., Vagin, A.A. & Dodson, E.J. Refinement of macromolecular
structures by the maximum-likelihood method. Acta Crystallogr. 53, 240–255
(1997).
43. Emsley, P., Lohkamp, B., Scott, W.G. & Cowtan, K. Features and development of
Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501 (2010).
© 2011 Nature America, Inc. All rights reserved.
© 2011 Nature America, Inc. All rights reserved.