A solute-binding protein in the closed conformation induces ATP hydrolysis in a bacterial ATP-binding cassette transporter involved in the import of alginate

The Gram-negative bacterium Sphingomonas sp. A1 incorporates alginate into cells via the cell-surface pit without prior depolymerization by extracellular enzymes. Alginate import across cytoplasmic membranes thereby depends on the ATP-binding cassette transporter AlgM1M2SS (a heterotetramer of AlgM1, AlgM2, and AlgS), which cooperates with the periplasmic solute-binding protein AlgQ1 or AlgQ2; however, several details of AlgM1M2SS-mediated alginate import are not well-understood. Herein, we analyzed ATPase and transport activities of AlgM1M2SS after reconstitution into liposomes with AlgQ2 and alginate oligosaccharide substrates having different polymerization degrees (PDs). Longer alginate oligosaccharides (PD ≥ 5) stimulated the ATPase activity of AlgM1M2SS but were inert as substrates of AlgM1M2SS-mediated transport, indicating that AlgM1M2SS-mediated ATP hydrolysis can be stimulated independently of substrate transport. Using X-ray crystallography in the presence of AlgQ2 and long alginate oligosaccharides (PD 6–8) and with the humid air and glue-coating method, we determined the crystal structure of AlgM1M2SS in complex with oligosaccharide-bound AlgQ2 at 3.6 Å resolution. The structure of the ATP-binding cassette transporter in complex with non-transport ligand-bound periplasmic solute-binding protein revealed that AlgM1M2SS and AlgQ2 adopt inward-facing and closed conformations, respectively. These in vitro assays and structural analyses indicated that interactions between AlgM1M2SS in the inward-facing conformation and periplasmic ligand-bound AlgQ2 in the closed conformation induce ATP hydrolysis by the ATP-binding protein AlgS. We conclude that substrate-bound AlgQ2 in the closed conformation initially interacts with AlgM1M2SS, the AlgM1M2SS–AlgQ2 complex then forms, and this formation is followed by ATP hydrolysis.

Alginate is a linear acidic polysaccharide that comprises ␤-Dmannuronate (M) and its C-5-epimer ␣-L-guluronate (G). The Gram-negative and alginate-assimilating bacterium Sphingomonas sp. strain A1 incorporates polysaccharides into cells through cell-surface pits (10). Genes that are involved in alginate import and degradation form a cluster on the genome of strain A1 and encode the periplasmic alginate-binding proteins AlgQ1 and AlgQ2; the cytoplasmic membrane-bound ABC transporter (AlgM1M2SS); the cytoplasmic endotype alginate lyases A1-I, A1-II, and A1-III; the exotype alginate lyase (A1-IV); and the alginate-dependent transcription factor AlgO (11,12). Alginate concentrated in the cell-surface pit is incorporated into the periplasm and is delivered to the ABC transporter by alginate-binding proteins. The ABC transporter for alginate import is a heterotetramer comprising transmembrane domains (heterodimer of AlgM1 and AlgM2) and an ATPbinding domain (homodimer of AlgS). After import into the cytoplasm by the ABC transporter, alginate is depolymerized to monosaccharides by endo-and exotype alginate lyases. Moreover, the genetic cluster for import and degradation of alginate is expressed in the presence of the polysaccharide under the control of the transcription factor AlgO (13).
Characteristics of ATPase and transport activities of reconstituted AlgM1M2SS in liposomes and crystal structures of the transporter have mainly been determined using alginate oligosaccharides with polymerization degrees (PDs) of Ͻ4 (14 -16). In contrast with the E. coli maltose ABC transporter, interactions of AlgM1M2SS with AlgQ2 result in the formation of an alginate-binding tunnel that is accessible to solvent. In the present study, in vitro assays of the ATPase and transport activities of AlgM1M2SS were performed using long alginate oligosaccharides (PD Ն 5), and non-transport ligand-bound AlgQ2 in the closed conformation interacted with the ABC transporter in the inward-facing conformation. These X-ray crystallography structure analyses were performed using the humid air and glue-coating method and demonstrated that the solute-binding protein in the closed conformation induces ATP hydrolysis in this bacterial ABC transporter.

Co-immunoprecipitation of AlgQ2 by AlgM1M2SS
Co-immunoprecipitation assays were conducted to analyze interactions between AlgM1M2SS and AlgQ2. In these analyses, recombinant AlgM1M2SS and AlgQ2 were expressed in E. coli cells and were purified to homogeneity (Fig. 1B). After interactions of anti-His tag antibody conjugated Dynabeads Protein G (Veritas) and AlgM1M2SS containing a histidine tag at the C terminus of AlgM2, AlgQ2 was added in the presence or absence of ATP and ⌬3M. Dynabeads Protein G was then collected, and bound proteins were subjected to SDS-PAGE, followed by Western blotting using anti-AlgQ2 antiserum (18) (Fig. 2). AlgQ2 was observed only in the presence of both ATP and ⌬3M (Fig. 2, lanes 2 and 3), and no interactions between AlgM1M2SS and AlgQ2 occurred in the absence of ⌬3M (Fig.  2, lane 1). No AlgQ2 was detected in the immunoprecipitation assay using other His tag protein as a negative control in place of AlgM1M2SS (data not shown). Because AlgQ2 comprises Nand C-terminal domains and adopts substrate-free open and substrate-bound closed conformations through domain dynamics (16,19), these co-immunoprecipitation assays suggested that ⌬3M-bound AlgQ2 interacts with AlgM1M2SS in the closed conformation. However, the substrate-free AlgQ2 in the open conformation did not interact with AlgM1M2SS. Moreover, following the addition of vanadate, the intensity of the AlgQ2 band increased slightly in the presence of ATP and ⌬3M (Fig. 2, lane 3). The phosphate analogue vanadate bound the ATP-binding site with ADP and inhibited ATP hydrolysis. Hence, the increase in AlgQ2 band intensity probably reflects inhibited dissociation of AlgQ2 from AlgM1M2SS, suggesting that this dissociation event is coupled with ATP hydrolysis.

ATPase and transport activities of AlgM1M2SS
ATPase and transport activities of reconstituted AlgM1M2SS in liposomes were measured in the presence of various alginate oligosaccharides and AlgQ2 (Fig. 3). Specific ATPase activity of AlgM1M2SS in the absence of saccharide was determined to around 30 nmol/min/mg. In these experiments, all alginate oligosaccharides enhanced ATPase activity, and saccharide chain lengths, structures of non-reducing ends (saturated or unsaturated), and modifications at reducing saccharide residues with PA in alginate oligosaccharides had limited effects on ATPase activities. In contrast, transport activities varied with saccharide chain lengths, with the highest transport activity in the presence of alginate tetrasaccharide as a substrate, and almost no transport of alginate oligosaccharides with PD of Ͼ5. In addition, transport of these long oligosaccharides was comparable with that of the non-alginate oligosaccharide chitotetraose, indicating that long alginate oligosaccharides with PD of Ͼ5 were not transport substrates of AlgM1M2SS in this assay system.

X-ray diffraction experiments using the humid air and glue-coating (HAG) method
Previously, the crystal structure of AlgM1M2SS was determined in complex with ⌬3M-bound AlgQ2 at a resolution of 3.2 Å (14), and similar crystals were frozen using the HAG method (20) to improve X-ray diffraction images. The HAG method is a novel technique for crystal mounting, in which crystals are coated with a water-soluble polymer and are then exposed to controlled humid air using humidifier HUM-1F (RIGAKU). AlgM1M2SS crystals in complex with ⌬3M-bound AlgQ2 were obtained in drop solutions comprising 22% PEG 3000, 0.1 M ADA-NaOH (pH 6.6), 0.15 M NaCl, and 16 mM CHAPSO. Crystals were then picked up using a loop covered with 10% polyvinyl alcohol (PVA) 4500 and were subjected to X-ray diffraction experiments with air controlled at 86 -92% relative humidity. In these experiments, the crystals had optimal mosaicity (0.607) at 90% relative humidity, and after freezing at 90% relative humidity, crystals were exposed to X-rays and provided diffraction data at 3.4 Å resolution, allowing determination of crystal structure by molecular replacement. The overall structure of AlgM1M2SS adopted the inward-facing conformation (Fig. 4A), and the electron density map that was obtained using the HAG method included an additional electron density at C-terminal regions of AlgM1 and AlgM2. In contrast with the previous structure, three residues (from Thr-321 to Val-323) in AlgM1 and seven residues (from Gly-285 to Val-291) in AlgM2 were accurately traced in the electron density map, indicating improved X-ray diffraction data after crystal freezing of AlgM1M2SS in complex with AlgQ2 using the HAG method.
To obtain a snapshot of the structure of the ABC transporter in complex with the non-transport ligand-bound solute-binding protein, AlgM1M2SS and AlgQ2 were crystallized in the presence of long alginate oligosaccharides. In these experiments, crystals of AlgM1M2SS, AlgQ2, and 6 -8M were

ATP hydrolysis by bacterial ABC importer
obtained in a drop solution comprising 22% PEG 3000, 0.1 M ADA-NaOH (pH 6.6), 0.15 M sodium formate, 8 mM CHAPSO, and 4 mM ATP and were picked up using a loop covered with 10% PVA 4500 and 5% ethylene glycol for X-ray diffraction experiments with controlled humid air. Dehydration of the crystal at 85% relative humidity for 10 min significantly improved mosaicity (1.714 -0.732), which was reduced to 0.846 at 83% relative humidity. The crystal was finally frozen and subjected to X-ray diffraction experiments at 85% relative humidity. Subsequently, the crystal structure of AlgM1M2SS in complex with long alginate oligosaccharide-bound AlgQ2 was determined at 3.6 Å using molecular replacement. In the resulting complex, AlgM1M2SS and AlgQ2 adopted the inward-facing and closed conformations, respectively (Fig. 4C). However, although crystallization was conducted in the presence of ATP, no density map of ATP was observed. Diffraction data and refinement statistics are shown in Table 1.

Alginate recognition by AlgQ2 in the complex
In addition to AlgM1M2SS in complex with long alginate oligosaccharide-bound AlgQ2, crystal structures of AlgQ2 complexed with ⌬3M and 7-10M were determined to analyze configurations of oligosaccharides bound to AlgQ2. The crystal of ⌬3M-bound AlgQ2 was then generated in drop solution comprising 30% PEG 4000, 0.1 M Tris-HCl (pH 8.5), and 0.2 M LiSO 4 . AlgQ2 was also co-crystallized with 7-10M in drop solution comprising 25% PEG 4000, 0.1 M Tris-HCl (pH 8.5), and 0.2 M calcium chloride. Finally, crystal structures of ⌬3Mand 7-10M-bound AlgQ2 were determined by molecular replacement at 1.55 and 2.0 Å resolution, respectively, and overall structures of both were shown to adopt the closed conformation (Fig. 4, B and D).
The present structures of AlgM1M2SS in complex with ⌬3M and 6 -8M-bound AlgQ2 were essentially the same as the previously determined structure (Protein Data Bank entry 4TQU), with root mean square deviations (RMSDs) of Ͻ0.6 Å for 1641 C␣ atoms. Moreover, no significant differences between structures of AlgM1M2SS in complex with ⌬3M and 6 -8M-bound AlgQ2 were observed at the AlgM1M2SS-AlgQ2 interface, suggesting that saccharide chain lengths had no effects on structures and conformations of AlgM1M2SS and AlgQ2.
The electron density map corresponded to four saccharide residues of AlgQ2 in the complex of AlgM1M2SS and ⌬3Mbound AlgQ2, whereas three saccharide residues in ⌬3M were bound to AlgQ2 in a monomer form (Fig. 4, A and B). These observations suggest two binding modes of ⌬3M in the AlgQ2 interaction with AlgM1M2SS. However, the present data were insufficient for analyses of occupancy and conformations of the two trisaccharides in the complex of AlgM1M2SS and ⌬3Mbound AlgQ2, and the coordinates corresponding to a tetrasaccharide were included for convenience. AlgQ2 recognizes the first to fourth saccharide residues from the non-reducing end of the substrate at subsites 1-4 (15,16,19). In the structure of AlgM1M2SS complexed with ⌬3M-bound AlgQ2, ⌬3M was accommodated at subsites 1-3 or subsites 2-4 in AlgQ2, and in the latter scenario, ⌬3M was considered a translocation product of ⌬3M at subsites 1-3 with the slide of one residue.

ATP hydrolysis by bacterial ABC importer
This translocation of ⌬3M was probably carried out through docking of ⌬3M-bound AlgQ2 and the ABC transporter AlgM1M2SS.
In structures of the AlgM1M2SS and AlgQ2 complex and AlgQ2 alone, the electron density map corresponded to only five saccharide residues of AlgQ2, despite the addition of 6 -8M to the crystallization drop solution (Fig. 4, C and D). Hence, AlgQ2 may strictly recognize five saccharide residues of alginate, and the sixth or later saccharide residue is flexible due to limited interactions with AlgQ2.
Structures of the substrate-binding site of AlgQ2 in four kinds of crystals are presented in Fig. 5. Tables 2 and 3 show amino acid residues of AlgQ2 interacting with saccharide residues via hydrogen bonds (Ͻ3.5 Å) and C-C interactions (Ͻ4.4 Å), respectively. The interaction of AlgQ2 with AlgM1M2SS produced little effect on the structure of the subsite in AlgQ2 (Fig. 5) or on the substrate-binding mode (Tables 2 and 3). Saccharide residues are represented as M1, M2, M3, M4, and M5 in that order from the non-reducing end. As reported previously (16), the number of hydrogen bonds between residues of amino acids and saccharides was greatest in M1, confirming that AlgQ2 strongly recognizes the saccharide of the non-reducing end. M5 also formed C-C interactions with multiple amino acid residues, suggesting that AlgQ2 recognizes alginate at subsites 1-5.

Discussion
ABC importers are classified as type I or II based on their structures. Type I ABC transporters include the maltose transporter MalFGK 2 and AlgM1M2SS, whereas the vitamin B 12 transporter BtuCD is categorized as a type II ABC transporter. MalFGK 2 exhibits significant ATPase activity in the presence of both maltose-binding protein (MalE) and maltose, whereas the MalE mutant (sMBP) accommodates maltose and sucrose, which is a non-transport substrate. MalFGK 2 reportedly exhibited similar ATPase activity in the presence of sMBP and either maltose or sucrose (8), suggesting that ATP hydrolysis is induced by interactions between the ABC transporter and the solute-binding protein with any substrate, but not by translo-

ATP hydrolysis by bacterial ABC importer
cation of the substrate by the ABC transporter. However, the structure of the MalFGK 2 and sucrose-bound sMBP complex has not been determined, and the initial interaction mode between the ABC transporter and solute-binding protein remains unknown. In contrast, when present in excess of ATP (ATP/ABC transporter ϭ 1:20), the ATPase activity of MalFGK 2 in the presence of MalE was higher than that in the presence of both MalE and maltose (9). These observations suggest that interactions between MalE in the open conformation and MalFGK 2 induce ATP hydrolysis. Accordingly, in the presence of excess ATP (ATP/ABC transporter ϭ 1000:1), ATPase activity of MalFGK 2 in the presence of MalE and maltose was higher than that in the presence of MalE alone, suggesting that MalE interacts with MalFGK 2 in closed and open conformations (6). Because the complex structure of MalFGK 2 and MalE has not yet been characterized at the moment of formation, the ensuing interaction scheme remains controversial. Similarly, the solute-binding protein BtuF reportedly induces ATP hydrolysis activity of the type II transporter BtuCD (21), although the details of the interaction scheme between BtuCD and BtuF are not fully understood. The present in vitro assays indicated that long alginate oligosaccharides with PD of Ͼ5 were not transported by liposome-reconstituted AlgM1M2SS. Moreover, in alginate-importing A1 cells, AlgM1M2SS is expressed in conjunction with cell-surface pits (11), suggesting that the ABC transporter requires additional elements, such as the cell-surface pit, for import of alginate polymers. However, long alginate oligosaccharides with PD of Ͼ5 stimulated ATPase activity of AlgM1M2SS, suggesting that substrate binding by the solutebinding protein triggers ATP hydrolysis in AlgM1M2SS. Accordingly, X-ray crystallography demonstrated that long alginate oligosaccharide-bound AlgQ2 in the closed conformation interacts with AlgM1M2SS. To the best of our knowledge, this is the first structural characterization of the complex of the ABC transporter with the non-transport ligand-bound solutebinding protein.
Although the structure of the AlgM1M2SS and ⌬3M-bound AlgQ2 complex was determined previously, it remains unclear which step of the transport cycle was structurally characterized, because ⌬3M was used as a transport substrate of AlgM1M2SS. In contrast, the complex structure of AlgM1M2SS with non-transport 6 -8M-bound AlgQ2 is indicative of the initial moment of the interaction between AlgQ2 and AlgM1M2SS. This structure indicates that ATP binding and hydrolysis occur after the formation of the complex, because no ATP is included in the complex structure (Fig. 6). In conclusion, substratebound AlgQ2 in the closed conformation initially interacts with AlgM1M2SS, and the AlgM1M2SS-AlgQ2 complex forms subsequently, followed by ATP hydrolysis.

Preparation and fluorescence labeling of alginate oligosaccharides
M oligosaccharides were prepared from alginate as reported previously (15). Briefly, alginate was hydrolyzed using HCl and was divided into M-block-and G-rich saccharides (G-block)

ATP hydrolysis by bacterial ABC importer
based on differences in the solubility at pH 2.85. Treatment of neutralized M-block with alginate lyase provided unsaturated M oligosaccharides, and acid hydrolysis of M-block with HCl provided saturated M oligosaccharides. M oligosaccharides were then labeled with 2-aminopyridine as reported previously (14) and were then eluted through a HiTrap Q HP column (GE healthcare). Remaining reagents and unreacted oligosaccharides were excluded using anion exchange chromatography, and purified PA-saccharides were obtained.

FACE
FACE (17) was conducted to determine the purity of alginate oligosaccharides, which were then mixed with labeling reagent containing 75 mM 8-amino-1,3,6-naphthalenetrisulfonic acid, 0.5 M sodium cyanoborohydride, 7.5% (v/v) acetic acid, and 50% (v/v) dimethyl sulfoxide and were incubated at 37°C for 16 h in the shade. Labeled alginate oligosaccharides were then freezedried and resolved in 10% glycerol and were then subjected to electrophoresis in 30% polyacrylamide gels with running buffer containing 25 mM Tris-HCl (pH 8.5) and 0.192 M glycine. Bands of labeled oligosaccharides in gels were detected using an ultraviolet lamp.

Assays of ATPase and transport activities
Expression and purification of AlgM1M2SS was conducted as reported previously (14). Briefly, E. coli BL21(DE3)Gold Proteoliposomes were prepared using L-␣-phosphatidylcholine (Sigma-Aldrich) and phospholipid derived from soybean as reported previously (14). ATPase assays were performed using type II-S, whereas transport assays were performed using type IV-S. AlgM1M2SS was added to liposomes that were destabilized in n-octyl-␤-D-glucoside and were diluted to less than critical micelle concentrations to produce proteoliposomes. Magnesium ions and ATP were confined in proteoliposomes for transport assays (14).
Assays of ATPase and transport activity were conducted as reported previously (14). Briefly, reactions were conducted in the presence of 0.1 M AlgM1M2SS, 1 M AlgQ2, 5 mM Mg-ATP, and 20 M oligosaccharides. ATPase activity was determined as the amount of inorganic phosphate released per min, and transport activity was measured in the presence of PA-saccharides. The reaction mixture comprised 0.1 M AlgM1M2SS, 1.3 M AlgQ2, 2 mM ATP, 10 mM MgCl 2 , 20 M PA-saccharide, and 20 mM Tris-HCl (pH 8.0). Liposomes were collected by ultracentrifugation after reactions, and transport activities were determined based on fluorescence intensities of imported PA-saccharides.

Co-immunoprecipitation assays
Interactions between AlgM1M2SS and AlgQ2 were determined using co-immunoprecipitation assays with binding/ washing buffer comprising 10% glycerol, 0.1 M NaCl, 0.045% 6-cyclohexyl-hexyl-␤-D-maltoside, 0.25% CHAPSO, and 20 mM Tris-HCl (pH 8.0). AlgM1M2SS exhibits significant ATPase activity in this surfactant solution. Precipitates were collected from 30 l of Dynabeads Protein G using DynaMag-2 (Life Technologies) and were then mixed with 200 l of binding buffer containing 0.5 g of rabbit anti-His-tag antibody (Cosmo Bio). After incubation at 4°C for 15 min, precipitates were washed once with wash buffer, and 200 l of binding buffer containing 20 g of AlgM1(d24)M2(H10)SS(WT) was added. After incubation at 4°C for 15 min, precipitates were washed once with wash buffer, and 200-l aliquots of binding buffers containing 20 g of AlgQ2, 0 or 2 mM Mg-ATP, 0 or 0.2 mM ⌬3M, and 0 or 5 mM sodium orthovanadate were added. After incubation at 4°C for 15 min, precipitates were washed four times in wash buffer and were then dissolved in 10-l aliquots of binding buffer and 10 l of 2ϫ SDS buffer comprising 4% SDS, 12% 2-mercaptoethanol, 20% glycerol, 0.1 M Tris-HCl (pH 6.8), and bromphenol blue. After boiling for 3 min, mixtures were centrifuged, and supernatants were subjected to SDS-PAGE analyses. AlgQ2 was then detected in Western blots using rabbit anti-AlgQ2 antiserum as a primary antibody and donkey anti-rabbit IgG as a secondary antibody (GE Healthcare).

Crystallization and X-ray diffraction experiments
Crystallization and crystal freezing of AlgQ2 (15) and crystallization of the AlgM1(d24)M2(H10)SS(E160Q) and AlgQ2 complex (14) were conducted as reported previously. Crystals of AlgM1(d24)M2(H10)SS(E160Q) and AlgQ2 complexes were frozen using the HAG method (20). Crystals were picked up using a loop covered with 10% PVA 4500 with or without ethylene glycol and were frozen in a cold nitrogen gas stream at optimized humidity.