Pathogenic siderophore ABC importer YbtPQ adopts a surprising fold of exporter

See allHide authors and affiliations

Science Advances  05 Feb 2020:
Vol. 6, no. 6, eaay7997
DOI: 10.1126/sciadv.aay7997


To fight for essential metal ions, human pathogens secrete virulence-associated siderophores and retake the metal-chelated siderophores through a subfamily of adenosine triphosphate (ATP)–binding cassette (ABC) importer, whose molecular mechanisms are completely unknown. We have determined multiple structures of the yersiniabactin importer YbtPQ from uropathogenic Escherichia coli (UPEC) at inward-open conformation in both apo and substrate-bound states by cryo–electron microscopy. YbtPQ does not adopt any known fold of ABC importers but surprisingly adopts the fold of type IV ABC exporters. To our knowledge, it is the first time an exporter fold of ABC importer has been reported. We have also observed two unique features in YbtPQ: unwinding of a transmembrane helix in YbtP upon substrate release and tightly associated nucleotide-binding domains without bound nucleotides. Together, our study suggests that siderophore ABC importers have a distinct transport mechanism and should be classified as a separate subfamily of ABC importers.


Iron (Fe), like other metal ions, is essential for all life forms and plays vital roles in many biological processes, primarily serving as cofactors for numerous proteins and enzymes (1). Thus, it is in the forefront of the battlefields between microorganisms and hosts. In general, infected hosts control the amount of available Fe ions by Fe-binding molecules and proteins, including intracellular heme (2) and ferritin (3), as well as extracellular transferrin (4), lactoferrin (5), and calprotectin (6). As a countermeasure, microbes have evolved an Fe uptake system using siderophores, a group of small molecules with high affinity to chelate Fe ions. For example, during urinary tract infections (UTIs), uropathogenic Escherichia coli (UPEC) secrete not only its genetically conserved siderophore enterobactin but also another virulence-associated siderophore yersiniabactin (Ybt). In the extracellular space, Ybt is able to chelate diverse metal ions, including, but not limited to, Fe3+, Cu2+, and Zn2+ (7). The metal-chelated Ybt complex is then transported back into the microbes first through the outer membrane protein ferric Ybt uptake receptor FyuA (8) in a TonB-dependent manner and then through inner membrane proteins such as adenosine triphosphate (ATP)–binding cassette (ABC) transporters YbtP and YbtQ (9, 10). After metal ions are released from the chelated complex in the cytosol, Ybt molecules are secreted out again to acquire more metal ions. Thus, the import of metal-chelated Ybt is an essential pathway for the microbes to acquire necessary metal ions. It has been established that deletion of YbtP or YbtQ gene in UPEC has no effect on the extracellular release of metal-free Ybt, but has detrimental effect on the intracellular metal uptake and later leads to profound decreased UTIs in a cystitis mouse model (11). Thus, it is suggested that the heterodimer YbtPQ is an ABC importer responsible for specific uptake of metal-chelated Ybt molecules. The importance of the uptake process renders YbtPQ a potential pharmaceutical target to treat related bacterial infections in human beings. Previously, studies have been focused on the identification of genes encoding YbtPQ (9) and their effect on the metal homeostasis at the cellular level (7, 10, 11). However, it remains largely unknown in the field how YbtPQ specifically selects and transports its substrates at the molecular level, which could be answered by high-resolution structural studies.

In this study, we used cryo–electron microscopy (cryo-EM) single-particle reconstruction to determine several structures of wild-type YbtPQ importer from the UPEC strain: YbtPQ in detergent lauryl maltose neopentyl glycol (YbtPQ-LMNG) at 3.7 Å resolution, YbtPQ in LMNG with iron-chelated siderophore Ybt (YbtPQ-LMNG-Ybt-Fe) at 3.4 Å resolution, and YbtPQ embedded in lipid nanodiscs (YbtPQ-nanodisc) at 4.9 Å resolution. To our surprise, we found that the overall structure of YbtPQ transporter does not resemble any known ABC importers, but rather adopts the fold of type IV ABC exporters, such as the multidrug resistance protein Sav1866 (12) and the lipid flippase MsbA (13). By structural comparison and mutagenesis studies, we identified a specific substrate-binding site around transmembrane helix 4 (TM4) in YbtP, which appears to undergo a unique structural rearrangement upon substrate interaction that may be substantial during the transport cycle. In addition, we noticed that in all YbtPQ structures, their nucleotide-binding domains (NBDs) are tightly associated even without any nucleotide bound to those domains. Furthermore, the association of NBDs is mediated by helix-helix interactions that are unique to YbtPQ.


Functional characterization of YbtPQ

First, we confirmed that YbtPQ is an ABC importer, not an exporter, by performing uptake experiments using the Ybt-Fe substrate in E. coli cells (Fig. 1A). In this experiment, E. coli BL21 C43 strain with empty vector defines the baseline for the Ybt-Fe uptake, serving as a negative control. It is clear that E. coli cells with overexpressed YbtPQ are able to import higher level of Ybt-Fe, suggesting that YbtPQ is a Ybt-Fe importer.

Fig. 1 Functional characterization of YbtPQ.

(A) Ybt-Fe uptake experiments. The relative amount of Ybt-Fe remaining in the solution after uptake by E. coli cells with either empty vector (dotted line, negative control) or overexpressed YbtPQ (solid line) is plotted against time. (B) ATP hydrolysis activity of YbtPQ-LMNG (top) and YbtPQ-nanodisc (bottom) plotted against ATP concentration. For both samples, Vmax and Km are not distinguishable at ~10 nmol mg−1 min−1 and ~0.5 mM, respectively. The error bars and SD are from the means of three independent experiments. (C) MST analysis of Ybt-Fe binding to the YbtPQ samples. Both YbtPQ-LMNG and YbtPQ-nanodisc have similar Kd values when binding the substrate.

Next, we characterized YbtPQ importer in terms of ATPase activity and substrate affinity. We overexpressed YbtPQ and purified it in detergent LMNG. The purification shows a homogeneous single peak for YbtPQ on gel filtration chromatography and two clean bands on SDS-PAGE (polyacrylamide gel electrophoresis) for YbtP and YbtQ, respectively (fig. S1, A and B). It is well known that the functionality of membrane proteins can be substantially impaired in various detergents on one hand and could be well maintained in lipid environment on the other hand. Thus, we made another sample, YbtPQ-nanodisc, by removing all detergents and incorporating the purified protein into a nanodisc system made of soybean polar lipid extract and membrane scaffold protein 1D1 (MSP1D1) (14). We then characterized the functions of both YbtPQ-LMNG and YbtPQ-nanodisc samples by measuring their ATPase activities and substrate Ybt-Fe–binding affinities. Both YbtPQ-LMNG and YbtPQ-nanodisc have a maximum activity of 10 to 12 nmol mg−1 min−1 with a Km (Michaelis constant) of around 0.5 mM (Fig. 1B). The substrate-binding affinities were measured using microscale thermophoresis (MST) (15). Again, the results show that both YbtPQ-LMNG and YbtPQ-nanodisc have a similar affinity to Ybt-Fe with a Kd (dissociation constant) value of around 0.08 to 0.13 μM (Fig. 1C). Together, the results confirm that the detergent LMNG did not affect the normal functions of YbtPQ. Thus, the final structures determined in LMNG are likely to be physiologically relevant.

Structure determination of YbtPQ

Because the functional characterization shows no substantial differences between YbtPQ-LMNG and YbtPQ-nanodisc, we decided to pursue structure determination of YbtPQ with both samples. We first analyzed both samples by negative-stain EM (fig. S2), which yielded two-dimensional (2D) averages with distinct features. We then analyzed both samples by cryo-EM (table S1). As shown in fig. S3, protein particles are evenly distributed in the cryo-images, and secondary features are visible in corresponding 2D averages. After data processing in both Relion (16) and cryoSPARC (17), we obtained final reconstructions of YbtPQ-LMNG at 3.7 Å overall resolution and those of YbtPQ-nanodisc at 4.9 Å resolution (fig. S4A). To improve the quality of the maps, we tried to subtract the signals of detergents and lipids around the protein particles. Unfortunately, this approach did not yield maps with better resolution. However, local resolution estimation of the YbtPQ-LMNG map shows well-resolved transmembrane domains (TMDs) approaching better than 3.5 Å resolution and NBDs with ~3.9 Å resolution (fig. S4B). The quality of the map in the TMDs allows de novo model building, with most side chains modeled in Coot (18). For the NBDs, it is known that the structural features of these domains in ABC transporters are highly conserved. We were able to use NBDs from several recent crystal structures of ABC transporters (12, 19, 20) as templates to dock into the YbtPQ-LMNG map. After several rounds of manual adjustment and real-space refinement in PHENIX (21), we obtained a near-atomic model of YbtPQ-LMNG. For the map of YbtPQ-nanodisc (fig. S5), we docked in the final YbtPQ-LMNG model and found that their overall conformations are identical.

To understand its substrate selectivity, we also determined the structure of YbtPQ, with Ybt-Fe bound. Here, we were convinced that YbtPQ-LMNG, rather than YbtPQ-nanodisc, would be a better sample moving forward because of the similar functional characterizations but better quality of cryo-EM reconstructions. We obtained a final reconstruction of YbtPQ-LMNG-Ybt-Fe with the overall resolution at 3.4 Å (fig. S6). Almost all helices in the TMDs are with a local resolution approaching better than 3.2 Å, allowing us to model previously ambiguous side chains in the YbtPQ-LMNG map.

Overall architecture of YbtPQ

YbtPQ is a heterodimeric ABC importer with a molecular size of ~130 kDa, occupying a space of 120 Å × 90 Å × 60 Å (Fig. 2A). Both YbtP and YbtQ are single polypeptide chains containing one N-terminal TMD and one C-terminal NBD. The sequence identity between YbtP and YbtQ is ~33%, and the similarity between is ~51%. The YbtPQ structure has a pseudo-twofold symmetry along the transport pathway that is perpendicular to the membrane plane. We used CAVER (22) to calculate possible tunnels inside the YbtPQ structures and found that the transport pathway is only solvent accessible to the intracellular side, suggesting that YbtPQ adopts an inward-open conformation (Fig. 2B). Similar to some known ABC exporters, such as Sav1866 and MsbA, YbtP and YbtQ each has six TMs, two of which “swapped” to its counterpart (Fig. 2C). Specifically, TM4 and TM5 in YbtP are associated with TM1, TM2, TM3, and TM6 in YbtQ to form a TMD, while TM4 and TM5 in YbtQ are associated with TM1, TM2, TM3, and TM6 in YbtP to form another TMD. The overall folding of the TMDs is similar, but they are not superimposed very well [root mean square (RMS) deviation of 5.04 Å for ~330 residues]. The major misalignment appears to be in the regions of TM4-TM5 and TM2-TM3. The regions are close to the interface between NBDs and TMDs formed by two sets of intracellular loops (ICLs): ICL1 in between TM2 and TM3 and ICL2 in between TM4 and TM5. When TMDs are aligned, NBDs are rotated away from each other ~10° around the axis perpendicular to the membrane plane (Fig. 2C). However, the NBDs of YbtP and YbtQ alone are very similar to each other, with an RMS deviation of 2.57 Å (~250 residues) when superimposed, because they adopt the conserved folds as NBDs in other ABC transporters.

Fig. 2 Cryo-EM analysis of YbtPQ and its overall architecture.

(A) The atomic model of YbtPQ-LMNG fits in its cryo-EM map. YbtP is in blue, YbtQ is in pink, while the cryo-EM density is in gray. The dotted black circle indicates a protruded density, probably consisting of copurified lipids. (B) A tunnel open to the cytoplasm with a minimum diameter of 1.5 Å is shown in green mesh, with a substrate molecule, Ybt-Fe, in dark gray sitting in the substrate-binding pocket. Zoom-in of this same view of the substrate-binding pocket is shown in Fig. 3. (C) Superposition of YbtP and YbtQ. Six TMs and two ICLs are labeled. When the two proteins are aligned with TMDs, their NBDs appear to be 10° rotated away.

ABC transporters are generally classified into seven distinct types (23) based on their functional and structural differences: types I to III are importers, types IV and V are exporters, and type VI functions as extractors, whereas type VII appears to be part of the efflux pumps. Because YbtPQ uptakes metal-chelated siderophores, it is previously thought to be a member of type II importers, similar to the vitamin B12 transporter BtuCD (24) from E. coli and the heme importer HmuUV (25) from Yersinia pestis. However, to our surprise, the overall structure of YbtPQ drastically deviates from not only type II importers but also all known ABC importers. Instead, YbtPQ closely resembles structures of type IV exporters. This observation can be explained in twofold: First, bacterial ABC importers normally have four polypeptide chains, with each folding as one domain of either TMD or NBD (26), while YbtP and YbtQ are both single polypeptide chains containing one TMD and one NBD; second, bacterial ABC importers generally have a periplasmic substrate-binding protein (SBP) to deliver the substrates, such as maltose-binding protein (MBP) for the maltose importer MalFGK2 (27) and BtuF for the vitamin B12 importer BtuCD (28) from E. coli. For YbtPQ, extensive searches for a corresponding SBP have been conducted and no appropriate candidate has yet been identified (7, 9, 29). It remains unclear whether such SBP is required for YbtPQ. Thus, YbtPQ may represent a new type of ABC importers.

An intracellular protruded density close to YbtQ

From the three cryo-EM structures present in this study, we observed a common protruded density unsymmetrically localized on one side of the importer, right below the membrane belt region (Fig. 2A and figs. S4 to S6). This feature might be a helpful factor for the accurate particle alignment during cryo-EM data processing, which is vital to obtain near-atomic resolution. After the model building, we noticed that the protruded density was on the side of YbtP and located in the space nearby the connection between the N-terminal elbow helix and the proceeding TM1. We first considered the possibility that the motif is similar to the “lasso motif” described in recent ABC transporter structures, such as cystic fibrosis transmembrane conductance regulator (CTFR) (30) and multidrug resistance protein 1 (MRP1) (31). However, in the CTFR and MRP1 structures, the lasso motif is ~60 residues long with two defined small helices. In YbtPQ, the protrusion does not have any features of secondary structure and it appears to be much smaller in size. Thus, we reasoned that it could be either posttranslational modifications on YbtQ or small molecules, especially lipids, copurified with the protein. Thus, we analyzed the YbtPQ-LMNG sample using matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS). We detected two protein peaks with the deconvoluted masses at 66,429.4 and 66,394.9 Da, which correspond well with the theoretical average mass of YbtP at 66,428.7 Da and that of YbtQ at 66394.8 Da (fig. S7A), suggesting no posttranslational modifications for YbtPQ. However, in the same sample, we also detected many small molecules that correspond well with different lipids when searched against the LIPID MAPS Proteome Database (fig. S7B). Thus, it is likely that the protruded density is a mixture of copurified lipids. Because of its unstructured nature and many hits of possible copurified lipids, we did not fit any specific molecules into the density. It remains interesting to see whether the copurified lipids affect the normal function of YbtPQ.

Substrate-binding pocket for Ybt-Fe

By comparing the maps of YbtPQ-LMNG and YbtPQ-LMNG-Ybt-Fe, we observed a clear density located within the pocket formed by four adjacent helices: TM4 and TM5 in YbtP and TM2 and TM3 in YbtQ (Fig. 3 and fig. S8A). The shape of the density correlates well with the shape of the substrate based on the previously determined crystal structure of Ybt-Fe (32). We docked one Ybt-Fe molecule into the density and analyzed its interactions with surrounding residues. The substrate pocket is mostly hydrophobic, surrounded with at least five aromatic residues all with clear side-chain densities in the cryo-EM map: W243, Y244, and F255 from YbtP, and Y92 and Y141 from YbtQ (Fig. 3A and fig. S8B). Together with other noncharged residues with small side chains, the pocket hydrophobically accommodates at least three of the four aromatic ring structures in Ybt-Fe. The four rings (fig. S8C) are phenol, thiazoline, thiazolidine, and another thiazoline. The first phenol ring in Ybt interacts with YbtP-F255 and YbtQ-Y92 in a face-to-edge style within the distance of 4 to 6 Å. The second and third rings of thiazoline and thiazolidine roughly align in a plane between the stacked YbtP-W243 and YbtP-W244 side chains, in a parallel displaced fashion. The last thiazoline ring does not seem to have strong pi-pi interactions with YbtQ-Y141 but may experience a hydrophobic push from the bulky side chain because they are in close distance. In addition, there is also a hydrogen-bonding network within the pocket that helps to maintain the positions of side chains and Ybt-Fe (Fig. 3B). The hydroxyl group of YbtQ-Y92 interacts with two hydroxyl groups in Ybt. The last thiazoline ring in Ybt seems to be stabilized by hydrogen bonding with surrounding residues YbtP-R185 and YbtQ-Y141. To confirm that the residues mentioned above are important for substrate binding, we performed mutagenesis studies and measured their substrate-binding affinities by MST (table S2). Because YbtP-N307 is around the substrate-binding pocket without apparent interaction with the substrate, we used it as a control. As expected, YbtP-N307A has the same substrate-binding affinity compared with the wild type. Among the six important residues mentioned above, YbtP-F255A seems to have the same substrate-binding affinity and YbtP-W243A and YbtQ-Y141A have several fold lower affinity, while YbtP-R185A, YbtP-Y244A, and YbtQ-Y92A exhibit much lower affinity compared with the wild type. Thus, our mutagenesis study results confirm the critical roles of YbtP-W243A, YbtQ-Y141A, YbtP-R185A, YbtP-Y244A, and YbtQ-Y92A in substrate binding.

Fig. 3 Network of interactions in the substrate-binding pocket.

(A) Important hydrophobic interactions between Ybt-Fe and YbtPQ in front (left panel) and side (right panel) views. Four helices forming the pocket are shown as ribbon, and five aromatic residues are shown as stick. (B) Important hydrogen bonds in the substrate-binding pocket in front (left panel) and top (right panel) views. Hydrogen bonds formed between residues and Ybt-Fe are shown as dotted lines. YbtP residues are in blue, and YbtQ residues are in pink. Ybt-Fe is colored by heteroatom, with the Fe molecule in orange.

Helical unwinding of YbtP-TM4 to release Ybt-Fe

Both models of YbtPQ with and without the substrate are in the same inward-open conformation, and they can be superimposed well with an RMS deviation of 1.69 Å over all residues. However, we noticed a marked difference along the YbtP-TM4 region, exactly where the substrate-binding pocket is (Fig. 4A). YbtP-TM4 starts with residue L166 at the top and ends with residue T216 at the bottom. Both residues are at about the same position in the aligned structures. In the map of YbtPQ-LMNG, without substrate Ybt-Fe, the densities in the top and bottom helical regions of YbtP-TM4 are well resolved, but the density in the middle region is poorly resolved (Fig. 4B). Thus, we modeled residues R185 to H196 with the side-chain occupancy set to 0. In the map of YbtPQ-LMNG-Ybt-Fe, with substrate Ybt-Fe, the whole YbtP-TM4 region is well resolved (Fig. 4B), and we were able to accurately model residues R185 to H196.

Fig. 4 YbtP-TM4 unwinding upon substrate release.

(A) Superposition of YbtP from YbtPQ structures with and without Ybt-Fe. YbtP with Ybt-Fe is in blue, while YbtP without Ybt-Fe is in gray. (B) Model (ribbon) and map (green mesh) of the isolated YbtP-TM4 region from both YbtPQ structures.

We propose that both YbtPQ structures (with and without substrate) represent the protein status after the substrate import, because they are both in the inward-open conformation. However, YbtPQ-LMNG-Ybt-Fe represents the status before the substrate release, while YbtPQ-LMNG is after the substrate release. On the basis of the structural differences in YbtP-TM4, we argue that the substrate release process may be facilitated by purposely unwinding part of this particular helix from residue Q186 to N190. Upon YbtP-TM4 unwinding, residues in the middle region move away from the substrate, completely destroying the substrate-binding pocket to release Ybt-Fe. For example, YbtP-R185 loses its strong hydrogen bonds with Ybt-Fe. Furthermore, the hydrophobic environment in the pocket can no longer be maintained. In the literature, similar helix conformational change is observed in the structure of human multidrug exporter P-glycoprotein (33). However, the helical unwinding in P-glycoprotein is associated with the conformational changes in the alternating-access mechanism, between inward-open and outward-open conformations. It is different from the YbtP-TM4 helical unwinding that is associated with the substrate release. Because the helical status of YbtP-TM4 in the outward-open conformation is unknown, we could not rule out the possibility that it might also be related to the conformational switches during the transport cycle.

Interface between TMD and NBD

ATP binding and hydrolysis trigger conformational changes in NBDs, which is transmitted into the conformational changes in TMDs to either open or close the translocation pathway in ABC transporters. The transmission is mediated through the interfaces between TMDs and NBDs. In YbtPQ, the transmission interfaces consist of ICL1, ICL2 from TMDs, and surrounding residues from NBDs (Fig. 5A and fig. S9). YbtP-ICL1 and YbtQ-ICL2 form short helices lying roughly parallel to the membrane plane, while YbtP-ICL2 and YbtQ-ICL1 adopt more relaxed conformation. The ICLs form bundles, as there are close interactions between YbtP-ICL1 and YbtQ-ICL2 and between YbtP-ICL2 and YbtQ-ICL1. The bundled ICLs may move together when the alternative access conformational switch happens during the transport cycle.

The residues in NBDs interacting with the ICLs are around the following two conserved motifs: A-loop and Q-loop. A-loop (xxYxx) contains a conserved aromatic residue in the middle that helps to bind to the adenine ring of ATP, while Q-loop (xxxxxxQx) contains a conserved Gln residue at the end that helps to form the reactive site for ATP hydrolysis. Residues in the A-loop interact with YbtP-ICL1 and YbtQ-ICL2, while residues in and around the Q-loop interact with YbtP-ICL2 and YbtQ-ICL1 (fig. S9). It has been reported in the structure of Sav1866 (12) that a short sequence (TEVGERG, conserved in many ABC exporters, also named X-loop) right before the ABC signature motif (LSGGQ, a unique feature of ABC transporters, not present in other P-loop nucleotide triphosphatases) contributes to the interactions between ICLs and NBDs. Unexpectedly, YbtPQ, as an importer, also has the short sequence but seems to be far away from the ICLs and does not directly interact with them.

In general, when ATP binds to NBDs, its adenine ring forms π-π interaction with the conserve aromatic residue in A-loop, and its phosphate groups interact with Walker A motif. Walker A is another conserved motif with a sequence of GxxGxGK(S/T). ATP binding also brings the two NBDs closer (34). Specifically, ABC signature motif in one subunit will come closer to Walker A motif in the other subunit for proper positioning of the γ-phosphate in ATP. Meanwhile, the conserved D-loops (xxxD) of NBDs need to come along each other, bringing conserved Walker B motifs closer to ATP, to form the correct geometry of the catalytic sites. Thus, the distances between conserved motifs are good indicators to elucidate whether NBDs are in closed or open conformation. The distance between Walker A and ABC signature motifs can be determined by the Cα atoms of conserved Gly and Ser, respectively, because both residues are at the end of a helix, while the distance between D-loops is determined by the Cα atoms of two spatially closest residues. In our structures, there are no ATP bound in the ATP-binding sites, as the aromatic rings of conserved Tyr residues (YbtP-Y351 and YbtQ-Y353) adopt a geometry parallel to Walker A motif, which is not ready for a π-π stacking with the adenine ring in ATP (Fig. 5B). In addition, Walker A motifs are 12.5 to 13 Å away from ABC signature motifs when measured between YbtP-G382 and YbtQ-S483 and between YbtQ-G383 and YbtP-S483, while D-loops are 7.2 Å away from each other when measured between YbtP-A511 and YbtQ-A510. These key distances between conserved motifs are longer when compared with known structures of ABC transporters, with ATP or ATP analogs bound (table S3), suggesting that NBDs are in the open conformation. We propose that when ATP binds, the open NBDs will switch to a closed conformation by physically pulling the A-loop, Q-loop, ABC signature, Walker A, and D-loop motifs closer. This will, in turn, pull the ICLs toward the center of YbtPQ because of their strong interactions with the A-loop and Q-loop. Considering that TMDs are rigid bodies, the intracellular movement of the ICLs will be transmitted to the extracellular side of TMDs to open the translocation channel.

Fig. 5 NBDs in YbtPQ.

(A) Interface between TMDs and NBDs in front and top views. NBDs are in gray, with YbtP shown as ribbon and YbtQ shown as surface. The interacting loops in NBDs are in orange (A-loop) and black (Q-loop). ICLs are shown as ribbon. (B) ATP binding sites and conserved motifs in top view. Walker A motif is in red, ABC signature motif is in green, and D-loop is in purple. The distance between Cα atoms of conserved residues from Walker A, ABC, and D-loop is indicated by dotted lines. (C) Association between NBDs in bottom view. Important residues for hydrophobic interactions between helices are highlighted in blue.

Associated NBDs without ATP binding

Although NBDs in YbtPQ are in open conformation without ATP bound, they remain tightly associated in the map. To our knowledge, most wild-type ABC transporters exhibit dissociated NBDs when there is no ATP bound, with a few exceptions. For example, without ATP, the crystal structure of TM287/288 (19) shows that NBDs are associated through hydrogen bonding between D-loop and Walker A motif in the same subunit and the switch Gln residue from the other subunit. In the structure of yeast mitochondrial ABC transporter Atm1 (35), NBDs are associated through the dimerization of two long C-terminal helices, with each coming from one subunit. In YbtPQ, NBDs are associated through two sets of helix-helix interactions, which is completely different from reported examples (Fig. 5C): L571-A582 near the C termini of YbtP interacts with helix I514-M527 in YbtQ right after its D-loop, and pseudo-symmetrically, G571-A579 near the C termini of YbtQ interacts with helix E514-L527 in YbtP right after its D-loop. The helix-helix contacts are mainly made of hydrophobic interactions, because no obvious hydrogen bonds between the helices could be found. The sequences of the helices contain repeated Leu, Ile, or Ala residues, mimicking the classic “leucine zipper coiled coil.” It is understandable that during the transport cycle, the helices might slide along each other in response to ATP binding. In doing so, the D-loops could be brought closer to each other and help to form the correct geometry for the following ATP hydrolysis.


The transport mechanism of YbtPQ can be explained by the classical model of alternating access. First, in the absence of the substrate and ATP, YbtPQ rests in an inward-open conformation, as presented in the structure of YbtPQ-LMNG. To import substrate Ybt-Fe, YbtPQ has to switch to outward-open conformation from the resting state. At this point, the detailed mechanism of the inward-to-outward switch is still unknown. In other ABC importers, such switch is facilitated by SBP and ATP binding. For instance, in bacterial importer MalFGK2, the transporter rests in an inward-open conformation and switches to outward-open upon MBP and ATP binding (27). For YbtPQ, although no specific SBP has been identified, one possibility is that SBPs from other siderophore uptake systems may be used here, for example, FepB of the enterobactin uptake system (36) and FhuD of the hydroxamate siderophore uptake system (37). Once SBP delivers the substrate to the transporter, YbtPQ’s conformational transition happens, which will effectively stimulate ATP binding and hydrolysis. This explains the basal level of ATPase activity detected in purified YbtPQ samples, because they are in the inward-open conformation without SBP binding. After ATP hydrolysis, YbtPQ goes back to inward-open conformation and exposes substrate Ybt-Fe to the cytosolic side, as shown in the structure of YbtPQ-LMNG-Ybt-Fe. The substrate is then released by unwinding the helix YbtP-TM4 to deform the substrate-binding pocket. Now, a transport cycle is completed and YbtPQ is back in the resting state.

In summary, our structures of YbtPQ represent a new type of bacterial ABC importers and provide novel information regarding substrate release after transport. Future studies to determine YbtPQ structures in other major conformations are important for our understanding of its complete transport cycle. Furthermore, as structures of other siderophore ABC transporters are determined, it remains interesting to see whether they follow the example of YbtPQ.


YbtPQ expression and purification

The genes encoding full-length YbtP (UniProt: A0A1D7Q186) and YbtQ (UniProt: A0A0F3WF96) from UPEC strain UTI89 were cloned into the pET15b (Novagen) vector with a His tag on the N terminus of YbtP and then expressed in the E. coli strain of BL21(DE3) C43 (Sigma-Aldrich) at 18°C overnight with 0.2 mM isopropyl β-d-1-thiogalactopyranoside (IPTG; UBPBio). Six liters of harvested cells was resuspended in 20 mM tris buffer (pH 7.5), 150 mM NaCl, and 1 mM phenylmethylsulfonyl fluoride (PMSF) and lysed by passing through a microfluidizer (Microfluidics) at 15,000 psi. The cell lysis was first centrifuged at 15,000g to get rid of unbroken cells and large cell debris, and then pelleted by ultracentrifugation at 150,000g for 2 hours. The pellets containing cellular membrane fraction were resuspended in 20 mM tris buffer (pH 7.5), 150 mM NaCl, and 1 mM PMSF. To solubilize proteins within the membrane, 1% LMNG was added and the mixture was incubated for 2 hours at 4°C before ultracentrifugation at 150,000g. The supernatant was collected and incubated with previously buffer-balanced Ni–nitrilotriacetic acid (NTA; Qiagen) resin. YbtPQ, as a complex, was eluted with 20 mM tris buffer (pH 7.5), 150 mM NaCl, 300 mM imidazole, and 0.01% LMNG and then incubated overnight in a cold room with thrombin (Enzyme Research Laboratories) at a molar ratio of 1:100 to cut off the His tag. YbtPQ without His tag was reapplied to fresh Ni-NTA resin again, and the flow-through was further purified by gel filtration on a Superose 6 column (GE Healthcare Life Sciences). The peak fractions with purified YbtPQ were concentrated to ~6 mg/ml and stored for further experiments.

Ybt-Fe uptake

E. coli cells with empty vector (control) and expressed YbtPQ were normalized and resuspended in phosphate buffer (pH 7.5) at room temperature. Ybt-Fe was added into the suspension at a final concentration of 0.2 mM to start the uptake experiments. At the following time points, 1, 2, 5, 10, and 20 min, cells were pelleted and the amount of Ybt-Fe in the solution was detected by absorbance at 310 nm as previously described (38). The experiments were repeated three times, and the average readings were used to plot the curve of the Ybt-Fe uptake.

YbtPQ-nanodisc preparation

An established protocol was followed to incorporate YbtPQ into nanodiscs with lipids of choice (14). Briefly, membrane scaffold protein MSP1D1 (Addgene) was expressed at 37°C in the E. coli BL21(DE3) strain and purified by one-step affinity chromatography using Ni-NTA resin. The purified MSP1D1 was concentrated to ~5 mg/ml and stored at −80°C as aliquots. As for the lipids, a stock of soybean polar lipid extract (Avanti Polar Lipids) was made in sodium chelate buffer. To assemble the nanodiscs, purified YbtPQ, MSP1D1, and soybean lipids were mixed at a molar ratio of 1:2:200 on ice for 30 min. To remove the detergents, Bio-Beads were added and the sample was incubated overnight at 4°C. Homogeneously assembled YbtPQ-nanodiscs were purified by gel filtration on a Superose 6 column in a buffer containing 20 mM tris (pH 7.5) and 150 mM NaCl.

Microscale thermophoresis

Substrate-binding affinity was measured using a Monolith NT.115Pico system (NanoTemper). A standard protocol was followed. Briefly, YbtPQ-LMNG with His tag was labeled with red-tris-NTA second-generation dye. For each reaction, 10 nM YbtPQ sample was mixed with Ybt-Fe at various concentrations from 0 to 50 μM. The sample was loaded into Monolith NT.115 capillaries, and microthermophoresis was performed using 20% light-emitting diode power and medium MST power. Kd values were calculated using the NanoTemper software. The reported Kd values were averages of three independent experiments.

ATPase activity assay

The ATPase activity was determined using the ATPase/GTPase Activity Assay Kit (Sigma-Aldrich) featuring a detectable colorimetric product from malachite green reacting with released phosphate group. For each reaction, a 15-μl mixture was prepared with 0.05 to 0.1 mg of YbtPQ-LMNG or YbtPQ-nanodisc by adding different concentrations of ATP from 0 to 2 mM. The mixtures were incubated at 37°C for 15 min, and then 100 μl of the dye reagent was added. Absorbance of each sample, including standard curve points at 620 nm, was measured 30 min later. The ATPase activity curve was fitted to the Michaelis-Menten equation using Excel. The experiments were repeated at least three times independently.

Cryo-EM sample preparation and data acquisition

The following three protein samples were used for cryo-EM analysis: YbtPQ-LMNG (~3 mg/ml), YbtPQ-nanodisc (2 mg/ml), and YbtPQ-LMNG (~3 mg/ml) incubated with 200 μM Ybt (EMC Microcollections), 200 μM FeCl3, 5 mM ATP, and 5 mM MgCl2 for 10 min. Three microliters of these samples was applied to plasma-cleaned C-flat holy carbon grids (1.2/1.3, 400 mesh) and prepared using a Vitrobot Mark IV system (Thermo Fisher Scientific) with the environmental chamber set at 100% humidity and 4°C. The grids were blotted for 2.5 to 3 s and then flash-frozen in liquid ethane cooled by liquid nitrogen. Cryo-EM data were collected on a Talos Arctica microscope (Thermo Fisher Scientific) operated at 200 keV and equipped with a K3 direct detector (Gatan). All movies were recorded at ×45,000 magnification with a calibrated pixel size of 0.864 Å using Leginon software (39). Defocus range was set from −1.2 to −2.3 μm. The parallel illumination setup on the Arctica gave a dose rate of ~24 electrons per pixel per second. Each movie was dose-fractionated to 50 frames, with a total exposure of 2 s, leading to a total dose of ~64 electrons.

Cryo-EM data processing

The dataset of YbtPQ-LMNG with Ybt-Fe and ATP was processed using software suite cryoSPARC v2.5.0 (17). Specifically, full-frame motion correction was done using default settings within cryoSPARC. Contrast transfer function (CTF) parameters of each micrograph were estimated using Gctf v1.06 (40). Micrographs with a calculated defocus range of −0.8 to 2.8 μm, fitted resolution of better than 6 Å, and global motion of less than 35 pixels were selected for further processing. Approximately 2000 particles were manually picked and classified into 20 classes. Five class averages with distinct feature and view were selected as templates for automatic picking. These particles were extracted with a box size of 336 × 336 pixels after local motion correction and subjected to 2D classification. Classes with clear visible secondary features were selected and subjected to ab initio reconstruction of four to six initial models. These initial models were heterogeneously refined, and the best group was selected for further homogeneous refinement and nonuniform refinement. The final resolution was estimated to be 3.4 Å based on the gold standard (41) Fourier shell correlation (FSC) with a cutoff of 0.143. Local resolution estimation was done with BlocRes (42).

The other two datasets, YbtPQ-LMNG and YbtPQ-nanodisc, were processed using RELION v3.0 (16). Specifically, movies were motion-corrected using UCSF (University of California, San Francisco) MotionCor2 (43), and then CTF was estimated using CTFFIND4 (44). Approximately 2000 particles were manually picked, and 20 class averages were generated by 2D classification. Six good class averages were selected as templates for the following autopicking in RELION. Following standard processing of particle extraction, 2D classification, initial model building, and 3D classification, best groups of particles were selected and refined to ~4 to 6 Å resolution. From there, multiple rounds of CTF refinement and Bayesian polishing were performed. The final resolution was estimated by gold standard FSC to be 3.7 Å for YbtPQ-LMNG and 4.9 Å for YbtPQ-nanodisc. Note that the default radiation damage weighting protocols in RELION and cryoSPARC are used for data processing. All details regarding the three datasets can be found in table S1.

Model building, refinement, and validation

There was no crystallographic structural information available for YbtPQ. However, the complex adopted similar fold as other ABC transporter such as sav1866 multidrug transporter from Staphylococcus aureus (12), TM287/TM288 from Thermotoga maritima (19), and P-glycoprotein from Caenorhabditis elegans (20). Thus, these homologous models were used as a guide for manual model building. The cryo-EM maps used here are final refinement maps from datasets YbtPQ-LMNG and YbtPQ-LMNG with Ybt-Fe and ATP. The maps were sharpened by different B-factor values for better visualization of both low and high local resolution information, which was very helpful during the process. The model building was carried out in Coot program (18). All transmembrane helices were built de novo, because their densities were of high quality. The NBDs were docked with homologous models, rigid-body refined, and then manually adjusted to fit the working maps. The substrate density within YbtPQ was fit with the known crystal structure of Ybt-Fe (32).

The final rounds of model refinement were carried out by real-space refinement in PHENIX (21), with secondary structure restraints imposed. The quality of the model was assessed by MolProbity (45). To validate the refinement, the model was refined against half-maps, and FSC curves were calculated. The final statistics was shown in table S1. All figures were prepared using the programs UCSF Chimera (46) and PyMOL (PyMOL Molecular Graphics System, DeLano Scientific). Superposition of the structures was carried out in Chimera using Needleman-Wunsch alignment algorithm and BLOSUM-62 matrix.


Supplementary material for this article is available at

Fig. S1. Purification of YbtPQ in LMNG.

Fig. S2. Negative-stain EM of YbtPQ.

Fig. S3. Cryo-EM data quality of YbtPQ.

Fig. S4. Cryo-EM reconstruction of YbtPQ.

Fig. S5. Cryo-EM reconstruction of YbtPQ-nanodisc.

Fig. S6. Cryo-EM reconstruction of YbtPQ-LMNG-Ybt-Fe.

Fig. S7. MALDI-MS analysis of YbtPQ-LMNG.

Fig. S8. Ybt-Fe substrate density in cryo-EM map.

Fig. S9. Sequence alignment of YbtPQ with homologs.

Table S1. Cryo-EM data collection, refinement, and validation statistics.

Table S2. Substrate-binding affinity of YbtPQ mutagenesis measured by MST.

Table S3. Distances between key conserved motifs in various ABC transporters.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank J. Henderson for providing the genomic material of YbtPQ and the staff in the shared resources of Biophysics, Mass Spectrometry, and cryo-EM at the University of Colorado Anschutz Medical Campus for their assistance. We also thank J. Kieft for critical reading and suggestions. Funding: H.Z. was supported by the Boettcher Foundation Webb-Waring Biomedical Research Award. Author contributions: Z.W., W.H., and H.Z. designed the experiments, collected and analyzed the data, and wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and material availability: Cryo-EM maps of YbtPQ were deposited in the Electron Microscopy Data Bank under accession codes EMD-20263 (YbtPQ-nanodisc), EMD-20262 (YbtPQ-LMNG), and EMD-20264 (YbtPQ-LMNG-Ybt-Fe). Coordinates of atomic models of YbtPQ-LMNG and YbtPQ-LMNG-Ybt-Fe were deposited in the Protein Data Bank under accession codes 6P6I and 6P6J, respectively. All other data are available from the corresponding author upon reasonable request.

Stay Connected to Science Advances

Navigate This Article