Determinants of substrate specificity and biochemical properties of the sn-glycerol-3-phosphate ATP binding cassette transporter (UgpBAEC2) of Escherichia coli


  • Steven Wuttge,

    1. Division of Microbial Physiology, Humboldt-Universität zu Berlin, Berlin, Germany
    Search for more papers by this author
    • Contributed equally to this work.
  • Martin Bommer,

    1. Division of Structural Biology and Biochemistry, Department of Biology, Humboldt-Universität zu Berlin, Berlin, Germany
    2. Macromolecular Crystallography Group, Institute of Soft Matter and Functional Materials, Helmholtz-Zentrum Berlin für Materialien und Energie, Berlin, Germany
    Search for more papers by this author
    • Contributed equally to this work.
  • Franziska Jäger,

    1. Division of Microbial Physiology, Humboldt-Universität zu Berlin, Berlin, Germany
    Search for more papers by this author
  • Berta M. Martins,

    1. Division of Structural Biology and Biochemistry, Department of Biology, Humboldt-Universität zu Berlin, Berlin, Germany
    Search for more papers by this author
  • Sophie Jacob,

    1. Division of Microbial Physiology, Humboldt-Universität zu Berlin, Berlin, Germany
    Search for more papers by this author
  • Anke Licht,

    1. Division of Microbial Physiology, Humboldt-Universität zu Berlin, Berlin, Germany
    Search for more papers by this author
  • Frank Scheffel,

    1. Division of Microbial Physiology, Humboldt-Universität zu Berlin, Berlin, Germany
    Current affiliation:
    1. Milchwirtschaftliche Lehr- und Untersuchungsanstalt Oranienburg, Oranienburg, Germany
    Search for more papers by this author
  • Holger Dobbek,

    1. Division of Structural Biology and Biochemistry, Department of Biology, Humboldt-Universität zu Berlin, Berlin, Germany
    Search for more papers by this author
  • Erwin Schneider

    Corresponding author
    • Division of Microbial Physiology, Humboldt-Universität zu Berlin, Berlin, Germany
    Search for more papers by this author

For correspondence. E-mail; Tel. (+49) 30 2093 8121; Fax (+49) 30 2093 8126.


Under phosphate starvation conditions, Escherichia coli can utilize sn-glycerol-3-phosphate (G3P) and G3P diesters as phosphate source when transported by an ATP binding cassette importer composed of the periplasmic binding protein, UgpB, the transmembrane subunits, UgpA and UgpE, and a homodimer of the nucleotide binding subunit, UgpC. The current knowledge on the Ugp transporter is solely based on genetic evidence and transport assays using intact cells. Thus, we set out to characterize its properties at the level of purified protein components. UgpB was demonstrated to bind G3P and glycerophosphocholine with dissociation constants of 0.68 ± 0.02 μM and 5.1 ± 0.3 μM, respectively, while glycerol-2-phosphate (G2P) is not a substrate. The crystal structure of UgpB in complex with G3P was solved at 1.8 Å resolution and revealed the interaction with two tryptophan residues as key to the preferential binding of linear G3P in contrast to the branched G2P. Mutational analysis validated the crucial role of Trp-169 for G3P binding. The purified UgpAEC2 complex displayed UgpB/G3P-stimulated ATPase activity in proteoliposomes that was neither inhibited by phosphate nor by the signal transducing protein PhoU or the phosphodiesterase UgpQ. Furthermore, a hybrid transporter composed of MalFGUgpC could be functionally reconstituted while a UgpAEMalK complex was unstable.


Escherichia coli can utilize sn-glycerol-3-phosphate (G3P), a degradation product of phospholipids due to the action of phospholipases and phosphodiesterase, as sole source of carbon and phosphate (Wanner, 1996). Growth on G3P as carbon source depends on the G3P–Pi antiporter GlpT (Lemieux et al., 2004). In contrast, under phosphate starvation conditions, G3P is utilized as phosphate source when exclusively transported by an ATP binding cassette (ABC) importer composed of the periplasmic binding protein, UgpB, the transmembrane subunits, UgpA and UgpE, and a homodimer of the nucleotide binding and hydrolysing subunit, UgpC (Boos, 1998). ABC systems which are found in all three kingdoms of life comprise exporters, importers and non-transporting ATP hydrolysing proteins (Licht and Schneider, 2011). Binding protein-dependent ABC importers enable prokaryotes to take up a large variety of nutrients, osmoprotectants, various growth factors or trace elements but are also implicated in the maintenance of cell integrity, responses to environmental stresses, cell-to-cell communication, cell differentiation and in pathogenicity (Eitinger et al., 2011). ABC transporters function by an ‘alternating access’ mechanism, according to which ATP binding and hydrolysis triggers conformational changes that cause an alternation between an outward-facing (open to the exterior) and an inward-facing (open to the cytoplasm) structure of the translocation path (Licht and Schneider, 2011).

The Ugp transporter is a member of the CUT 1 subfamily of ABC importers encompassing transporters for the uptake of carbohydrates and polyols (Schneider, 2001) and thus being closely related to the maltose ABC transporter which is the best-understood ABC importer to date (Overduin et al., 1988; Bordignon et al., 2010). In particular, the nucleotide binding subunit UgpC can restore growth on maltose when overproduced in an E. coli strain lacking the homologous MalK subunit of the maltose transporter (Hekstra and Tommassen, 1993). Thus, UgpC could form a functional transport complex with the transmembrane domains of the maltose transporter MalF and MalG, albeit with low affinity. Likewise, excess MalK was also shown to complement a ugpC-negative strain (Hekstra and Tommassen, 1993).

Interestingly, the activity of the Ugp transporter is inhibited by phosphate in intact cells (Brzoska et al., 1994), but can be recovered by simultaneous overexpression of plasmid-borne ugpC (Boos, 1998). This observation led the authors to speculate that phosphate might bind to UgpC and thereby lower its affinity for UgpAE (Boos, 1998). Moreover, it provides a possible explanation for the lack of growth of E. coli on G3P as carbon source when transported by the Ugp system only (Brzoska et al., 1994).

Besides G3P, the transporter also accepts G3P diesters but sn-glycerol-2-phosphate (G2P) was not presumed to be a substrate. Utilization of G2P as source of phosphate was considered to require the activity of periplasmic alkaline phosphatase, the product of the phoA gene (Wanner, 1996). However, this notion was recently challenged by Yang et al. (2009) who reported growth of an E. coli ΔphoA strain on G2P as sole source of phosphate, provided the presence of an intact Ugp transporter. This conclusion was drawn from the observation that mutants carrying a deletion of ugpB, ugpA, ugpE or ugpC failed to grow on G2P.

The current knowledge on the Ugp transporter is solely based on genetic evidence and transport assays using intact cells. Thus, in order to prove or disprove these hypotheses, we set out to study the transporter's properties at the level of purified protein components.

We show both by binding assays and by solving the crystal structure of UgpB in complex with G3P that G2P is not a substrate of the transport system. Furthermore, UgpB/G3P-dependent ATPase activity of purified UgpAEC2 incorporated into proteoliposomes was not inhibited by inorganic phosphate up to 10 mM indicating that the observed in vivo effect is not due to direct binding of phosphate to the transporter. Likewise, neither the regulatory protein PhoU, previously proposed to inhibit the phosphate ABC transporter PstS–CAB2 (Rice et al., 2009), nor the cytoplasmic phosphodiesterase UgpQ, suggested to be associated with the Ugp transporter (Brzoska and Boos, 1988), did affect ATPase activity of reconstituted UgpAEC2. Finally, a hybrid complex composed of MalFG–UgpC2 was stable and displayed MalE/maltose-stimulated ATPase activity while a functional UgpAE–MalK2 complex could not be obtained.


UgpB binds G3P and glycerophosphocholine (GPC) but not G2P

To determine the substrate specificity of UgpB, the protein was purified from the overproducing E. coli host strain BL21 T1 DE3 (pFSA131) (Fig. 1), subjected to a de- and renaturation cycle to remove co-purified ligand (see Experimental procedures for details), and assayed for binding of G3P and G2P by isothermal titration calorimetry (ITC). The data reveal that G3P is bound with a dissociation constant (Kd) of 0.68 ± 0.02 μM (Fig. 2A). The Kd value is in the range of those reported for most carbohydrate binding protein components of ABC importers (Berntsson et al., 2010) including UgpB purified from wild-type cells (0.12 μM; Boos, 1998). UgpB also bound GPC, a G3P diester, with a Kd value of 5.1 ± 0.3 μM (Fig. 2B). In marked contrast, no binding was observed with G2P (Fig. 2C). Consequently, since complex formation of binding protein and substrate is a prerequisite for transport via the cognate ABC importer, we conclude that G2P is not transported by the Ugp system.

Figure 1.

SDS-gel of purified proteins used in this study. Pst (protein standard), UgpB (2 μg), UgpAEC2 (6 μg), MalFGUgpC2 (3 μg), PhoU (3 μg), UgpQ (3 μg) and MalFGK2 (2 μg).

Figure 2.

Substrate binding affinities of UgpB variants determined by isothermal titration calorimetry. UgpB wild type and variants were titrated with G3P, GPC or G2P. Integrated and normalized heats of reaction Q were plotted against the molar ratio R and fitted using a one-site binding model and corrected to remove unspecific ligand/buffer dilutions. (A) UgpB + G3P; (B) UgpB + GPC; (C) UgpB + G2P; (D) UgpB(W169H) + G3P; (E) UgpB(W169S) + G3P; (F) UgpB(W172H) + G3P; (G) UgpB(W172S) + G3P. Structures of the substrates used are indicated.

Crystal structure of UgpB reveals determinants of G3P binding

The crystal structure of UgpB in complex with G3P was determined at 1.8 Å resolution. The substrate is bound at the interface of an N- and C-terminal domain (Fig. 3A), thus following the general architecture of solute binding proteins (Berntsson et al., 2010). The N-terminal domain is composed of a five-stranded β sheet sandwiched in an α-helical structure. Two of the central strands lead up to an interdomain hinge, forming a part of the ligand binding pocket, where residues Ser-121 and Gly-284 are in hydrogen bonding distance to G3P (Fig. 3B). The C-terminal domain comprises a four-stranded β sheet enclosed by α helices. This domain contains the most prominent feature of the binding pocket, two tryptophan side chains (Trp-169 and Trp-172), which are in close contact with the backbone of G3P, and together with the abovementioned interdomain hinge enclose the ligand. The plane formed by G3P carbon atoms C1–C3 is virtually parallel to the Trp-169 indole ring with C1,2,3-ring distances of 3.6 Å, 4.0 Å and 3.8 Å respectively. Likewise, the C1, C2, C2-OH plane is close to parallel with the Trp-172 ring system, showing distances of 3.8 Å and 3.9 Å for the carbon atoms. The hydroxyl group oxygen is within hydroxyl-π hydrogen bonding (Steiner and Koellner, 2001) distance (3.5 Å) of the ring (see Fig. S2 for further details). Further stabilization occurs through hydrogen bonds from both domains: Arg-374 and Glu-66 bind to the 1- and 2-hydroxyl group of G3P, and Ser-247 as well as Tyr-42 and Tyr-323 are in hydrogen bonding distance to the phosphate group. Arg-374 is the only residue involved in both substrate binding and interdomain bridging. It is within hydrogen bonding distance of two residues from either domain, Glu-66 (N) and Glu-176 (C). Glu-176 in turn lies on the same helix as Trp-169 and Trp-172. A central arginine residue has been implicated in stabilizing the closed conformation of the glycine betaine and proline betaine binding protein ProX of Archaeoglobus fulgidus (Tschapek et al., 2011) via a similar network of interdomain hydrogens bonds.

Figure 3.

Topology of E. coli UgpB and substrate binding mode.

A. Overall structure of E. coli UgpB showing the substrate as ball and stick model. The protein ribbon shows the N- (green helices, residues 1–120, 284–354) and C- (blue helices, residues 125–264, 369–413) terminal domains. Central to each domain is a five-stranded and four-stranded β sheet respectively (yellow).

B. sn-Glycerol-3-phosphate binding site as viewed from the N- towards the C-terminal domain. The binding pocket is formed by Trp-169 and Trp-172 and the numbered residues with superscript ‘N’, ‘C’ belong to said domains and superscript ‘Nβ’ belongs to the central β sheet. Interactions at hydrogen bond distance to protein side chains, the amide of Gly-284, a solvent water molecule and a glycerol molecule from the crystal cryoprotectant (yellow) are shown as dashed lines. Glu-176 does not interact with the substrate but together with Arg-374 is part of the Trp-169/Trp-172 α helix. Electron density maps are shown for Trp-169 and Trp-172 (blue mesh, 2mFobs-DFcalc, 1.2σ) and sn-glycerol-3-phosphate (green mesh, mFobs-DFcalc, 4σ, simulated-annealing omit map).

C. Slice through the surface representation of UgpB showing Trp-169 and 172, the substrate G3P (ball and stick), solvent glycerol (yellow sticks) and water molecules (red balls) at the solvent channel leading up to the substrate binding site.

D. Close-up of the solvent channel with modelled glycerophosphocholine (GPC, ball and stick) replacing both G3P and the glycerol molecule. GPC was placed by hand in Coot (Emsley et al., 2010), hydrogen atoms were added to the model and geometry optimization was carried out in Phenix (Adams et al., 2010). No clashes between GPC and UgpB were found with MolProbity (Chen et al., 2010).

In the ligand bound form, the phosphate group remains partially accessible to solvent (Fig. 3C) and makes contact to a water and a glycerol molecule. The latter is further stabilized in the solvent channel by potential hydrogen bonds to Asp-230, Trp-169 and backbone atoms from Ser-9 and Glu-11. In order to understand the significance (or insignificance) of the binding of this second glycerol, we compared its average crystallographic temperature (B factor), as an approximate measure of mobility and occupancy. Average B factors for the additional glycerol (33.8 Å2) are far higher than for G3P (11.4 Å2), or the protein (23.2 Å2), but similar to a solvent glycerol bound to the outside of the protein (33.4 Å2) or the solvent (33.5 Å2). This is consistent with the glycerol molecule not being tightly bound but originating from the cryoprotectant added after crystal growth. Furthermore, ITC measurements revealed no binding of glycerol to UgpB. Thus, while we would not consider the glycerol molecule a substrate, its hydrogen bonding network with UgpB might be utilized by the second glycerol moiety of GPC, a reported substrate of UgpB (Brzoska and Boos, 1988).

We constructed a theoretical model of GPC bound in place of G3P and the solvent glycerol. While neither appropriately positioned hydrogen bonding partners nor a choline binding motif such as the aromatic box in the choline/acetylcholine binding protein ChoX of Sinorhizobium meliloti (Oswald et al., 2008) was identified in UgpB, no clashes between the ligand and the protein were reported by MolProbity (Chen et al., 2010). The choline moiety sits in the solvent channel leading to the binding site (Fig. 3D). Absence of a specific interaction might explain both the lower affinity for GPC and substrate promiscuity with respect to diesters.

Despite their diverse sequences, the solute binding proteins are often very similar in structure and comparison of their active sites helps to highlight the importance of certain residues. UgpB fits the topology of type I solute binding proteins, which comprises most of the sugar binding proteins, as classified by Fukami-Kobayashi et al. (1999). According to structural alignments carried out using the Dali server (Holm and Rosenström, 2010) the most similar structure to UgpB is that of the maltose binding protein from Thermococcus litoralis (Diez et al., 2001) (PDB Code 1EU8) with both proteins sharing 17% identical amino acids. The most similar structure in the structural alignment-based classification of Berntsson et al. (2010) is the cluster B E. coli maltodextrin binding protein (Quiocho et al., 1997; PDB Code 4mbp) with 12% sequence identity. A superimposition of UgpB with seven similar structures identified by the Dali server, as well as pairwise structural alignments against cluster B proteins, is shown in Fig. S1. The highest conservation in both sequence and structure is seen in the β sheet of the C-terminal domain and the active site hinge. Matching of secondary structure elements allowed the superimposition of the UgpB active site with that of five of the homologues. The glycerol moiety of G3P superimposes well with one of the sugar rings from maltotetraose and maltotriose (Fig. S1B). The phosphate group protrudes into the space occupied by a tryptophan and a glutamate residue in the sugar binding proteins, and these residues are not conserved in UgpB. In turn, Arg-374 and Glu-66 binding the C1 hydroxyl group of G3P are not conserved in the sugar binding proteins. Most revealing however, Trp-169, one of the residues forming the substrate clamp, is exclusive to UgpB and occupies the space of the otherwise branched/circular sugar substrate. The second tryptophan (Trp-172) replaces a conserved tyrosine.

While superimposition of the UgpB binding site with that of more distantly related type II solute binding proteins is not possible, elements are shared. In a variation of the theme of Trp-169/172 in UgpB, polar-π hydrogen bonding has been observed in amine binding proteins: the glycine/proline betaine binding protein ProX of A. fulgidus has an aromatic box consisting of four tyrosine residues (Tschapek et al., 2011), the compatible solute binding protein OpuAC of Bacillus subtilis (Smits et al., 2008) features an aromatic prism consisting of three tryptophan residues and the choline/acetylcholine binding protein ChoX of S. meliloti (Oswald et al., 2008) has an aromatic box containing three conserved tryptophan and a tyrosine residue.

Mutational analysis of the G3P binding site reveals crucial role of Trp-169

To further explore the role of Trp-169 and Trp-172 in ligand binding, we performed mutational analysis. Both residues were individually replaced by either histidine or serine and, in addition, the double mutant Trp-169His, Trp-172His was constructed. The variants were purified and binding of G3P was monitored by ITC. As seen in Fig. 2F and G, replacing Trp-172 by histidine only slightly increased (twofold) the dissociation constant of G3P (Kd = 1.59 ± 0.04 μM), while substitution by serine caused a strong (130-fold) increase in Kd (90 ± 3 μM). These results indicate that a tryptophan at position 172 is not essential for G3P binding, provided π interactions are retained as in case of the imidazole side chain of histidine. In marked contrast, Trp-169 could neither be replaced by histidine nor serine without complete loss of binding activity (Fig. 2D and E), underscoring a crucial role in G3P binding as suggested from the crystal structure. Consistently, the double mutant did not exhibit binding activity either (not shown). Notably, none of the mutants had gained binding activity for G2P (not shown). Finally, and in line with these results, all mutants, except UgpB(W172H), failed to stimulate ATPase activity of the purified UgpAEC2 complex incorporated into liposomes (Table 1) (see also below).

Table 1. Properties of UgpB variants
UgpB variantKd (G3P) (μM)Stimulation of ATPase activity of UgpAEC2a (n-fold)
  1. aStimulated ATPase activity corresponds to 2.2 ± 0.2 μmol Pi min−1 mg−1.
Wild type0.68 ± 0.027.3
W172H1.59 ± 0.046.2
W172S90 ± 32.1

Likewise, sequential mutation of the tryptophan residues in the binding pocket of B. subtilis OpuAC to phenylalanine or tyrosine reduced, and to alanine virtually abolished affinity for glycine betaine (Smits et al., 2008). Mutation to alanine of aromatic box tyrosine residues in A. fulgidus ProX led to a 60- to 2400-fold reduction in affinity for glycine betaine (Tschapek et al., 2011).

UgpB/G3P-stimulated ATPase activity of reconstituted UgpAEC2 is insensitive to phosphate, PhoU or UgpQ

In a previous study, uptake of G3P into E. coli cells by the Ugp transporter was found to be inhibited by internal phosphate concentrations greater than 5 mM which was proposed to be mediated by direct phosphate binding to the transporter, most likely to UgpC (Brzoska et al., 1994). To reexamine this finding in vitro, the UgpAEC2 complex was purified (Fig. 1) and analysed by monitoring substrate-dependent ATPase activity in proteoliposomes. To this end, UgpAEC2 was mixed with E. coli total phospholipids in dodecylmaltoside and proteoliposomes were formed by removal of detergent via adsorption to hydrophobic beads (see Experimental procedures for details). The transporter displayed a low intrinsic activity of 0.2–0.3 μmol Pi min−1 mg−1 which was significantly stimulated by UgpB/G3P, usually in the range of five to sevenfold, depending on the preparation (Fig. 4B). UgpB/G3P can either be enclosed in the lumen of the proteoliposomes during their formation or added to the medium. In the latter case, Mg2+ ions that are required for activity permealize the proteoliposomes thereby allowing access of ATP to their lumen (Liu et al., 1997) (a cartoon illustrating the assay conditions is also shown in Fig. 5A). The observed specific activity of 1.7–2.2 μmol Pi min−1 mg−1 is in good agreement with data reported for its close relative, the maltose/maltodextrin transporter of E. coli/Salmonella (Mannering et al., 2001; Landmesser et al., 2002). [Please note that the apparent excess of UgpC over UgpEA as seen in Fig. 1 is no longer observed after reconstitution (see Fig. 4B, insert). Thus, the intrinsic ATPase activity measured in the absence of UgpB/G3P cannot be explained by the presence of soluble UgpC.] Furthermore, UgpB/G3P-stimulated ATPase activity was inhibited by orthovanadate (Fig. 4B), which blocks release of phosphate from the nucleotide binding site after one step of ATP hydrolysis and thus trapping the transporter in the outward-facing conformation (Chen et al., 2001). Sensitivity to vanadate indicates coupling of ATP hydrolysis to substrate translocation (Chen et al., 2001) while basal (uncoupled) ATPase activity is resistant to the inhibitor (see Fig. 4B). Together, we conclude that the purified UgpAEC2 complex has the earmarks of a functional transporter.

Figure 4.

ATPase activities of UgpAEC2 and complex variant MalFGUgpC2.

A. Cartoon illustrating subunit composition of native (UgpBAEC2, MalEFGK2) and hybrid (MalE–FG–UgpC2; UgpB–AE–MalK2) complexes. Arrows show the exchange of NBDs which in case of MalK was unsuccessful (indicated by ‘×’).

B. ATPase activities of UgpAEC2 and hybrid complex MalFGUgpC2 in proteoliposomes stimulated by wild-type UgpB/G3P and MalE/maltose respectively. Values represent the mean of three independent measurements. Inserts: SDS gel of reconstituted UgpAEC2 (2 μg) and MalFGUgpC2 (1 μg).

Figure 5.

ATPase activity of UgpBAEC2 in the presence of phosphate.

A. Cartoon illustrating the experimental setup of the ATPase assay. UgpAEC2 is incorporated into proteoliposomes in two possible orientations. With the nucleotide binding subunits, UgpC2, facing the medium, ATPase activity is initiated by UgpB/G3P that was enclosed during proteoliposome formation. Putative inhibitors that are thought to bind to UgpC2 (phosphate ions, PhoU, UgpQ; question marks indicate hypothetical interaction) are added to the medium side (left). Addressing UgpAEC2 complexes that expose the UgpC dimer to the lumen of the proteoliposomes is achieved by enclosing putative inhibitors during vesicles formation and supplying UgpB/G3P externally. Mg2+ ions that are required for enzymatic activity will permealize the proteoliposomes (Liu et al., 1997), thus allowing access of ATP and other small molecules (e.g. phosphate ions) to the lumen (right).

B. Autoradiograph showing hydrolysis of [γ-32P]ATP catalysed by UgpBAEC2 in the absence and presence of phosphate. Proteoliposomes containing UgpAEC2 were incubated in 50 mM Tris-HCl, pH 7.5, 3 mM MgCl2, in the absence or presence of UgpB/G3P or in the presence of UgpB/G3P and 10 mM phosphate for 5 min at 37°C prior to the addition of [γ-32P]ATP (final concentration: 1 mM). At the indicated times, aliquots were removed and subjected to thin-layer chromatography followed by detection of 32Pi with a PhosphorImager (see Experimental procedures for details).

C. Quantification of 32Pi from B as a function of time. A representative experiment is shown.

Then, we studied the effect of phosphate on ATPase activity of the reconstituted UgpAEC2 complex in the presence of UgpB/G3P (Fig. 5A). Since the routinely used colorimetric assay based on determining the release of phosphate from ATP was not applicable, we monitored the formation of 32Pi from [γ-32]ATP (Fig. 5B). As shown in Fig. 5C, no inhibition was observed at 10 mM sodium phosphate, thereby excluding a direct inhibitory interaction of phosphate with the transporter.

Since these results suggested that the observed inhibition of the Ugp system at high internal phosphate concentration might be mediated by a yet unknown cytosolic factor, we examined a possible role of the signal-transducing protein PhoU. In the context of the phosphate signalling system in E. coli, PhoU is involved in shifting the biochemical activity of the sensor kinase PhoR from an autokinase activity to a phosphatase activity leading to dephosphorylation and, thus, inactivation of the transcriptional regulator phospho-PhoB under high phosphate conditions (Hsieh and Wanner, 2010). In a recent study, based on genetic evidence, PhoU was also speculated to directly inhibit the activity of the phosphate ABC transporter PstS–CAB2 (Rice et al., 2009). Hence, we reasoned that PhoU might have a similar effect on the activity of the Ugp system. Consequently, we purified His6–PhoU from the cytosolic fraction of an overproducing strain (see Experimental procedures for details) by Ni-NTA chromatography, cleaved the His6 tag by protease treatment (Fig. 1) and subsequently analysed release of 32Pi from [γ-32]ATP by reconstituted UgpB–AEC2 in phosphate buffer in the presence and absence of PhoU. As in case of phosphate alone, no inhibition of ATPase activity was observed (data not shown).

Another possible candidate for mediating inhibition of the Ugp transporter by internal phosphate is the phosphodiesterase UgpQ. The protein, which is encoded by the last gene in the ugp operon, hydrolyses phosphodiesters to G3P and the corresponding alcohols. UgpQ was suggested to be associated with UgpAEC2 based on the finding that its activity could be observed only in intact cells but not in cellular extracts (Brzoska and Boos, 1988). Although this notion was recently challenged by Ohshima et al. (2008), who reported enzymatic activity of UgpQ in the cytosolic fraction in the presence of Mg2+ ions, we purified UgpQ (Fig. 1) similar to PhoU from the soluble fraction of an overproducing E. coli strain (see Experimental procedures for details). However, like in the case of PhoU, ATP hydrolysis catalysed by reconstituted UgpB–EAC2 was unaffected by UgpQ in the presence or absence of phosphate.

Hybrid complex MalFGUgpC exhibits MalE/maltose-dependent ATPase activity

Previous complementation studies revealed that overproduced UgpC could restore maltose transport in E. coli cells carrying a deletion of the malK gene (Hekstra and Tommassen, 1993). This finding prompted us to investigate whether a hybrid MalFG–UgpC complex would be stable in vitro. To this end, malFG and ugpC were co-expressed from plasmids pHL16 and pFSA132, respectively, in E. coli strain BL21 T1 DE3, and a hybrid complex was purified from solubilized membrane vesicles by taking advantage of a His6 tag fused to the N terminus of UgpC. As shown in Fig. 1, all proteins could be identified on a SDS gel, indicating the formation of a stable complex. Next, we analysed MalE/maltose-dependent ATPase activity of the hybrid complex incorporated into liposomes (Fig. 4A) and found a vanadate-sensitive, specific activity comparable to that of the native maltose transporter (Chen et al., 2001; Landmesser et al., 2002) (Fig. 4B).

In contrast, a hybrid complex composed of UgpAE and MalK could not be obtained. SDS-PAGE of the eluates from the Ni-NTA matrix clearly showed His-tagged MalK but only substoichiometric amounts of UgpAE (data not shown). Since the His-tagged version of MalK used here was shown previously to form a functional complex with its native partners MalF and MalG (Wuttge, Landmesser, Thaben and Schneider, unpubl. data), the result cannot be explained by assuming a defect of the protein in pulling down interacting proteins.


We have investigated the Ugp transporter at the level of purified proteins to address open questions from previous in vivo studies concerning substrate specificity, inhibition by phosphate and hybrid formation with subunits of the maltose transporter.

The structure of UgpB closely resembles that of sugar binding proteins in type I cluster B (Fukami-Kobayashi et al., 1999; Berntsson et al., 2010). Differences in the active site account for binding of the smaller substrate. A clamp formed by two tryptophan residues holds the glycerol moiety of the linear G3P substrate. The phosphate group is within hydrogen bonding distance to four protein side chains (Fig. 3B). One of the tryptophan residues (Trp-169) is unique to UgpB and explains its specificity. A branched G2P substrate would clash with either of the two tryptophan residues seen in the G3P bound structure. Together with results from binding experiments, these data clearly demonstrate that G2P, in contrast to G3P and GPC, is not a substrate of UgpB and hence is not transported by the Ugp system.

Our findings challenge the recent in vivo study which indicated G2P is transported by the Ugp system. The reason for this discrepancy is unclear. However, since each ugp gene was required for the observed growth of the mutant strain on G2P, the possibility that a solute binding protein other than UgpB could deliver G2P to the transporter is unlikely (Yang et al., 2009). Thus, as already taken into account by the authors, G2P might be converted in the periplasm to G3P by a yet unknown enzyme.

Previous complementation experiments have shown that the nucleotide binding subunits MalK and UgpC, when overproduced, can be exchanged between both transport systems. These results indicated that the respective binding affinities are less for the foreign than for the cognate transmembrane subunits. Our in vitro data clearly demonstrate that there are distinct differences in affinity between MalFG/UgpC and UgpEA/MalK. While the former could be stably purified and display normal ATPase activity, the latter was apparently not formed. Interestingly, C-terminal amino acid residues of MalG (Gly-293, Lys-295) were found in the crystal structure of the MalE–FGK2 catalytic intermediate to form hydrogen bonds with one copy of MalK (Oldham et al., 2007). While the interacting residues of MalK are conserved in UgpC, the C terminus of UgpE (275GLVDSEK281) differs in sequence and length from that of its paralogue MalG (288GLTAGGVKG296; conserved residues are underlined, MalK-interacting residues are shown in boldface).

In contrast, all residues of MalF that are in contact with the other copy of MalK are conserved in UgpA. Thus, it is tempting to speculate that the lack in sequence identity at the C terminus of UgpE might account for the instability of a complex between UgpEA and MalK in vitro.

Exchangeability of nucleotide binding domains (NBDs) is not confined to transporters operating in the same organism, like UgpC and MalK, but was also demonstrated for LacK, energizing the lactose ABC transporter of Agrobacterium radiobacter and MalK of Salmonella enterica subspecies enterica serovar Typhimurium (Wilken et al., 1996). Moreover, in Gram-positive bacteria, operons encoding carbohydrate ABC transporters belonging to the CUT 1 subfamily often lack a gene for an NBD subunit. Consequently, NBD genes located elsewhere on the chromosome might be shared by these systems. This notion was first proven right experimentally by demonstrating that the MsiK protein of Streptomyces is required for uptake of maltose and cellobiose, respectively, by distinct transport systems (Hurtubise et al., 1995; Schlösser et al., 1997). Subsequently, it was shown that also trehalose transport is energized by MsiK in Streptomyces reticuli (Schlösser, 2000). Most recently, the MsmK protein of Streptococcus pneumoniae was found to be shared by ABC transporter core complexes, including those with specificity for sialic acid, trehalose and maltooligosaccharides (Marion et al., 2011). Intriguingly, sharing NBDs was identified as a general concept of ABC transporters belonging to the subfamily of ECF (energy coupling factor) import systems. ECF transporters, which mediate the uptake of micronutrients and are widespread among prokaryotes, consist of pairs of NBDs (A components), a conserved transmembrane protein (T component) and a transmembrane substrate capture protein (S component). While in subclass 1 porters the T component and the NBDs are dedicated to one S component, members of subclass 2 share the TAA module between multiple S components (Eitinger et al., 2011).

It was previously proposed by Boos and co-workers that the activity of the Ugp transporter is inhibited by external Pi concentrations greater than 0.5 mM (Brzoska et al., 1994). Our failure to demonstrate direct inhibition of the reconstituted Ugp system by phosphate (Fig. 5) might be explained by assuming that an unknown cytosolic protein of E. coli is altered by phosphate (directly or indirectly) in such a way that it binds to UgpC and thus arresting the transporter. This would be reminiscent of the inhibition of the maltose transporter by EIIAGlc, a component of the phosphoenolpyruvate glucose phosphotransferase system (Postma et al., 1996). Recently, the PhoU protein was suggested to inhibit the activity of the phosphate ABC transporter, PstS–CAB2 (Rice et al., 2009). Thus, it was tempting to speculate that PhoU might also be involved in phosphate-dependent inhibition of the Ugp transporter. However, our results clearly show that this is not the case. Likewise, our data also exclude that UgpQ, the product of the last gene of the ugp operon, which displays phosphodiesterase activity, mediates inhibition of the transporter by phosphate. Thus, the mechanism by which internal phosphate affects transporter activity in vivo remains to be elucidated.

Together, the data presented here add the Ugp transporter to the list of model systems suited for studying molecular details of the mechanism by which canonical ABC importers exert their functions. In particular, as a close relative of the maltose transporter, UgpAEC2 will be of further use for addressing the question how interaction with its cognate substrate binding protein, UgpB, is achieved in the absence of a long periplasmic loop which is crucial for function of the former (Bordignon et al., 2010). Such studies are underway in our laboratory.

Experimental procedures


β-Glycerophosphate (G2P) disodium salt hydrate, sn-glycerol 3-phosphate bis(cyclohexylammonium) salt and sn-glycero-3-phosphocholine (1:1 cadmium chloride adduct) were purchased from Sigma-Aldrich Chemistry GmbH (Hamburg, Germany). [γ-32P]Adenosine 5′-triphosphate (3000 Ci mmol−1) was purchased from Hartmann Analytik (Braunschweig, Germany). Guanidine hydrochloride was obtained from Roth (Karlsruhe, Germany).

Gene cloning and mutagenesis

The genes ugpB (lacking its signal sequence), ugpC, ugpQ and phoU were amplified from E. coli K12 by colony PCR and subcloned into expression vector pET15 providing an N-terminal His tag (Novagen, Germany). The resulting plasmids were named pFSA131 (ugpB), pFSA132 (ugpC), pWS28 (phoU) and pWS30 (ugpQ). A PCR fragment containing ugpAE was obtained likewise and cloned into pAC11a (pACYC184 containing multiple cloning site of pET11a; kind gift of U. Wehmeier, University of Wuppertal) yielding plasmid pFSA136. Single substitutions in ugpB were introduced by Stratagene's QuikChange lightning kit, resulting in plasmids pWS23(W169S), pWS24(W169H), pWS25(W172S), pWS26(W172H) and pWS27(W169H, W172H).

Protein purifications

UgpAEC2, MalFGUgpC2, UgpAEMalK2

The UgpAEC2 complex was overproduced in E. coli strain BL21 T1 DE3 (Sigma-Aldrich) harbouring plasmids pFSA132 and pFSA136 and purified according to a published procedure (Landmesser et al., 2002). Briefly, the cells were grown in tryptone–phosphate medium (Moore et al., 1993) at 30°C to an OD650 of 0.8. Expression of ugpAE and ugpC was induced by the addition of 0.5 mM isopropyl-β-thiogalactoside (IPTG). At the end of log phase (usually 3 h after induction) the cells were harvested, resuspended in buffer A [50 mM Tris-HCl, pH 7.5, 0.1 mM EDTA, 20% glycerol, 0.1 mM phenylmethylsulphonyl fluoride (PMSF), 2 mM dithiothreitol (DTT)] and disrupted by passage through a Constant Cell Disruption System (Constant Systems Limited) at 1.7 kbar. After a low-speed centrifugation for 10 min at 10 000 g followed by high-speed centrifugation for 90 min at 200 000 g, the membrane fraction was resuspended in buffer B (50 mM Tris-HCl pH 7.5, 20% glycerol, 0.1 mM PMSF). Membrane vesicles (at 5 mg ml−1 protein) were solubilized with n-dodecyl-β-d-maltoside (DDM) at a final concentration of 1.1%. After incubation for 1 h at 4°C under gentle stirring, solubilized proteins were separated from the insoluble fraction by ultracentrifugation at 200 000 g for 30 min. The supernatant was subsequently incubated with Ni-nitrilotriacetic acid (NTA) agarose (Qiagen) previously equilibrated with buffer C (50 mM Tris-HCl, pH 7.5, 10% glycerol, 0.1 mM PMSF, 0.01% DDM) for 1 h at 4°C under gentle shaking. After washing the resin with 15 bed volumes of buffer C and buffer C supplemented with 20 mM imidazole, respectively, UgpAEC2 was eluted with buffer C supplemented with 250 mM imidazole. The elution fractions were pooled and concentrated with Vivaspin 4 (100 kDa cut-off) (Sartorius) to a final volume of 2.5 ml, followed by passage through a PD10 column (GE Healthcare) which was previously equilibrated with buffer C. Finally, the protein was shock-frozen in liquid nitrogen and stored at −80°C.

The hybrid complex MalFG–UgpC2 was purified similarly using plasmids pHL16 (malFG) (lab collection) and pFSA132 (ugpC).

Attempts to purify a UgpAE–MalK2 hybrid were carried out with cells of E. coli strain BL21 T1 DE3 carrying plasmids pFSA136 (ugpAE) and pHL08 (malK) (lab collection).

UgpB, MalE

UgpB (wild type and variants) was overproduced in E. coli strain BL21 T1 DE3 harbouring the respective plasmids. Cells were grown in LB medium (Miller, 1972) at 30°C to an OD650 of 0.5 and gene expression was subsequently induced by addition of 0.5 mM IPTG. After growth continued for 3 h cells were harvested, disintegrated and UgpB variants were purified from the cytosolic fraction. Pooled fractions were concentrated by Amicon® Ultra-4 (cut-off 10 kDa) (Millipore), passed through a PD10 column (GE Healthcare) to remove imidazole, subsequently shock-frozen in liquid nitrogen and stored at −80°C. For crystallization, UgpB was concentrated up to 90 mg ml−1.

For ITC measurements and ATPase assays, purified UgpB was subjected to a denaturation/renaturation procedure (Hall et al., 1998) to remove tightly bound substrate. To this end, UgpB was dissolved in buffer D (50 mM Tris-HCl pH 7.5, 10 mM DTT, 6 M guanidine hydrochloride) to a final protein concentration of 1 mg ml−1 and dialysed against 1 l of buffer D overnight at 4°C. Subsequently, dialysis was continued with two changes of buffer D. UgpB was renatured by dialysis against 1 l of 50 mM Tris-HCl, pH 7.5, over 4 days, with two buffer changes per day, and finally concentrated with Amicon® Ultra-4 (cut-off 10 kDa) columns.

MalE was prepared as described previously (Daus et al., 2007).


Purification of PhoU was carried out essentially as described for UgpB except that cells of E. coli strain BL21 T1 DE3 (pWS28) were grown in tryptone–phosphate medium (Moore et al., 1993). After purification, the His6 tag was cleaved by treatment with thrombin (Thrombin CleanCleave™ Kit, Sigma-Aldrich) according to the manufacturer's instructions. Samples were then passed through a PD10 column (GE Healthcare) followed by passage through a Ni-NTA resin. Tag-less PhoU was collected in the flow through, shock-frozen in liquid nitrogen and stored at −80°C.


UgpQ was purified from the cytosolic fraction of E. coli strain BL21 T1 DE3 (pWS30) exactly as described for PhoU.

Preparation of proteoliposomes

Proteoliposomes were prepared essentially as described by Scheffel et al. (2004). Briefly, E. coli total lipid extract (Avanti) (20 mg) was dried extensively under vacuum in a glass tube and redissolved in 1 ml of 50 mM Tris-HCl, pH 7.5, containing 1% octyl-β-d-glucopyranoside (OG). The mixture was flushed under a stream of nitrogen, sealed, incubated for 1 h at 4°C under gentle shaking and sonicated for 5 min in an ultrasonic bath (Bandelin Sonorex Super RK 106) at 4°C. Subsequently, UgpAEC2 (50 μg) and 50 mM Tris-HCl, pH 7.5, were added to the lipid–detergent mixture (25 mg) to give a final volume of 300 μl. Proteoliposomes were formed by adsorption of detergent to 100 mg of Biobeads (Bio-Rad) at 4°C overnight. After incubation for 1 h with a new batch of Biobeads the mixture was separated from beads and centrifuged for 1 min at 10 000 g to pellet remaining beads and non-reconstituted protein. Subsequently, the proteoliposomes were centrifuged for 1 h at 200 000 g at 4°C and resuspended gently in 200 μl of 50 mM Tris-HCl, pH 7.5.

Isothermal titration calorimetry measurements

Binding affinities of UgpB (wild type and variants) for G3P, GPC and G2P were determined with a VP-ITC device (MicroCal) as described by Bulut et al. (2012). Briefly, UgpB was filled in the reaction cell at a concentration of 40 μM in buffer containing 50 mM Tris-HCl, pH 7.5, 100 mM NaCl. At 4 min intervals (30 steps) 10 μl of the indicated substrate (480 μM) were titrated to UgpB. In case of Ugp(W172S) the substrate concentration was raised 10 times. GPC was prepared as described in Ohshima et al. (2008).

Crystallization and structure determination of UgpB

Crystals were grown by the sitting drop vapour diffusion method. Initial crystals were obtained using the JCSG+ screen (Newman et al., 2005). Optimization yielded the following crystallization condition: 5 mM cobalt(II) chloride, 5 mM nickel(II) chloride, 5 mM cadmium chloride, 5 mM magnesium chloride, 0.1 M HEPES, pH 7.5, 12% (w/v) polyethylene glycol 3350, 0.3% OG, which was mixed in a ratio of 1:1 with a protein solution containing 90 mg ml−1 UgpB, 20 mM Tris-HCl, pH 7.5, 5 mM G3P and 100 mM NaCl. Resulting plate-like crystals grew to 50 × 50 μm size within 1 week at 18°C. The reservoir solution mixed 3:1 with 87% glycerol served as cryoprotectant. A high-resolution data set (0.918 Å X-ray wavelength, 1.8 Å resolution) was recorded at Beamline 14.1 at the Bessy II Synchrotron Facility at the Helmholtz Zentrum Berlin, Germany (Mueller et al., 2012). An additional long-wavelength data set (1.9 Å wavelength, 2.32 Å resolution) was measured on Beamline 14.2. Data were processed using XDS (Kabsch, 2010) (Table S1).

Experimental phasing was performed by the single-wavelength anomalous dispersion (SAD) method using the Phenix AutoSol/Autobuild Wizard (Terwilliger et al., 2009; Adams et al., 2010). Anomalous differences from the long-wavelength data set revealed 16 heavy atoms (later identified as one cadmium, 14 sulphurs and one phosphorus). This substructure was used for initial phasing, followed by automated model building. A total of 389 residues were built automatically for the correct hand solution. This initial structure was then refined against the high-resolution (1.8 Å) data set. After manual adjustment of the model with Coot (Emsley et al., 2010) and refinement in Phenix refine, 418 residues and 476 water molecules could be located. The G3P ligand was built in eLBOW (Moriarty et al., 2009). The crystal structure was validated using MolProbity (Chen et al., 2010). A single Ramachandran outlier is well defined in the electron density map and co-ordinates a metal ion as part of a crystal contact. Cys-163 and Cys-241 form a disulfide bridge in a low-resolution structure obtained with a reduced X-ray dose (Fig. S2), while our final 1.8 Å resolution structure shows the reduced open configuration. We attribute this difference to photo reduction by the synchrotron beam, which frequently affects redox centres in high-resolution structures (Ravelli and McSweeney, 2000). Figures were prepared in PyMol (Schrödinger, LLC). Co-ordinates were deposited in the PDB under Accession Number 4AQ4.

ATPase assays and protein determination

Hydrolysis of ATP was routinely assayed with proteoliposomes containing 0.5 μM of complex protein in 50 mM Tris/HCl, pH 7.5, in the presence or absence of 100 μM substrate and 10 μM UgpB. The reaction was initiated by adding 2 mM ATP and 4 mM MgCl2 and ATP hydrolysis was measured by assaying the release of inorganic phosphate at 37°C using ammonium molybdate as described in Chifflet et al. (1988). Inhibition of ATPase activity of reconstituted transport complexes by orthovanadate was assayed accordingly after incubation with the inhibitor (0.5 mM) for 5 min at 37°C (Chen et al., 2001).

Where indicated, ATPase activity was alternatively measured by determining the release of 32Pi from [γ-32P]ATP according to Weinstock et al. (1981). Standard reaction mixtures (125 μl total) contained 0.5 μM reconstituted complex, 10 μM UgpB/G3P (either incorporated during vesicle formation or added externally) and 3 mM MgCl2 in 50 mM Tris-HCl, pH 7.5. To test their inhibitory effects, sodium phosphate (10 mM; pH 7) was added from both sides, whereas PhoU and UgpQ were provided at the opposite side from UgpB (Fig. 5A). After incubation at 37°C for 5 min, the reaction was started by the addition of 12.5 μl of [γ-32P]ATP (10 mM, 6.25 μCi). At each indicated time point, 25 μl was transferred to a reaction tube containing ice-cold 20% trichloroacetic acid. Subsequently, aliquots (1.7 μl) were spotted on polyethyleneimine–cellulose plates (Merck, Darmstadt) and developed in 1 M formic acid and 0.5 M LiCl. Plates were then dried under a hood, wrapped in plastic foil and release of 32Pi was detected by a PhosphorImager system (Molecular Imager FX, Bio-Rad, Munich, Germany).

Protein concentrations were determined by measuring absorption at 280 nm. The respective molar extinction coefficients were calculated from the amino acid sequences using the ProtParam tool (


This work was supported by the Deutsche Forschungsgemeinschaft (SCHN 274/15-1 to E. S. and Cluster of Excellence ‘UniCat’ EXC 314 to H. D.) and the Humboldt-Universität zu Berlin through the Joint Berlin MX program. B. M. M. is supported by the DFG (MA3348/2-1). We would like to thank all staff at the BESSY II beamlines, in particular Michael Krug for his help with XDS, Tobias Werther (Division of Structural Biology and Biochemistry, Department of Biology, Humboldt-Universität zu Berlin) for excellent assistance with isothermal titration calorimetry measurements, Haydar Bulut (Division of Physical and Theoretical Chemistry, Department of Chemistry and Biochemistry, Freie Universität Berlin) for his advice on the crystallization of solute binding proteins and N. Ohshima (Gunma University, Japan) for making his lab protocol to prepare cadmium-free glycerophosphocholine available to us.