Crystal structure of PotD, the primary receptor of the polyamine transport system in Escherichia coli.

PotD protein is a periplasmic binding protein and the primary receptor of the polyamine transport system, which regulates the polyamine content in Escherichia coli. The crystal structure of PotD in complex with spermidine has been solved at 2.5-A resolution. The PotD protein consists of two domains with an alternating beta-alpha-beta topology. The polyamine binding site is in a central cleft lying in the interface between the domains. In the cleft, four acidic residues recognize the three positively charged nitrogen atoms of spermidine, while five aromatic side chains anchor the methylene backbone by van der Waals interactions. The overall fold of PotD is similar to that of other periplasmic binding proteins, and in particular to the maltodextrin-binding protein from E. coli, despite the fact that sequence identity is as low as 20%. The comparison of the PotD structure with the two maltodextrin-binding protein structures, determined in the presence and absence of the substrate, suggests that spermidine binding rearranges the relative orientation of the PotD domains to create a more compact structure.

The polyamine transport genes in Escherichia coli have been cloned and characterized (3)(4)(5)(6)(7). The proteins encoded by pPT104 constitute the spermidine-preferential uptake system, which belongs to a periplasmic transport system (8,9). This spermidine transport machinery consists of four protein subunits, PotA, -B, -C, and -D. The PotA (M r 43,000) protein, which is bound to the inner surface of the cytoplasmic membrane, is a strong candidate for an ATP-hydrolyzing, energy-generating factor. In fact, the PotA protein contains a consensus nucleotide-binding sequence, and exhibits ATPase activity (10). Both the PotB (M r 31,000) and PotC (M r 29,000) proteins have six transmembrane spanning segments linked by hydrophilic peptides with variable lengths, and hence they are assumed to jointly form a channel for spermidine and putrescine. The PotD protein is a periplasmic binding protein and consists of 348 amino acids, corresponding to a molecular mass of 39 kDa. Although it binds both spermidine and putrescine, spermidine is preferred (6).
The polyamines are unique substrates for a periplasmic binding protein, and their specific interactions with cognate binding proteins have never been studied in terms of threedimensional structure. Therefore, a crystallographic study of the PotD protein at an atomic resolution was performed to elucidate the detailed mechanism of its specific substrate recognition and the characteristics of the main chain folding. In this paper, we report the molecular structure of the PotDspermidine complex determined at 2.5-Å resolution by x-ray analysis.

MATERIALS AND METHODS
Structure Determination-Crystals, which belong to the monoclinic system space group P2 1 , with unit cell parameters a ϭ 145.3 Å, b ϭ 69.1 Å, c ϭ 72.5 Å, and ␤ ϭ 107.6°, were grown according to the procedure described previously (24). They contain four molecules in the asymmetric unit. The procedure for data collection was already reported (24).
The major heavy atom sites of K 2 PtCl 4 and Pb(NO 3 ) 4 derivatives prepared by soaking were determined from their difference Patterson maps. The initial analysis of the x-ray data showed that the structure factors for the reflections with the odd h indices were much smaller than those with the even h indices (ϽF(2n ϩ 1 , k, l)Ͼ ϭ 0.5*ϽF (2n, k, l)Ͼ) in a 6-Å resolution shell. This fact, along with the analysis of the heavy atom sites in the derivatives, confirmed that there are two dimers of the protein in the asymmetric unit, connected by an almost precise translational symmetry with 1 ⁄2 of the a-axis of the crystal (24). The heavy atom parameters were refined with the programs PROTEIN (25) and MLPHARE (26) against the 3.0-Å resolution data, including anomalous data from all derivatives. The latter program provided the mean figure of merit of 0.63. Solvent flattening (27) and noncrystallographic * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 (28) were applied to improve the phases. The 2-fold molecular averaging, using only reflections with h ϭ 2n, was successful in substantially improving the map. However, this map was still insufficient to achieve a complete chain tracing. The structure determination statistics for the MIR phasing are summarized in Table I.
Model Building and Crystallographic Refinement-The initial model was constructed on the basis of the averaged electron density map at 3.0-Å resolution, using the program O (29). Rigid body refinement was then carried out using the program X-PLOR, version 3.0 (30), with all the data to 3.0-Å resolution. The initial model was used to define the molecular envelope, and then the 4-fold averaging technique was repeated using all the data, for further improvement of the phases, by the program DM in the CCP4 package (26). The averaging was reiterated until convergence was achieved, while the phases were expanded from 3.0-to 2.7-Å resolution. The correlation coefficients increased from 0.33 to 0.85. Consequently, this map allowed us to achieve a complete chain tracing. The model was manually modified by the program FRODO (31) on an Evans & Sutherland PS390 graphics system, and it was refined against the 2.5-Å resolution data set, using the program X-PLOR. During the refinement, the (͉F o ͉ Ϫ ͉F c ͉) and (2͉F o ͉ Ϫ ͉F c ͉) maps were used for manual adjustments of the model, and for locating the four spermidine and the solvent molecules. Only the water molecules that form geometrically reasonable hydrogen bonds with the protein atoms were included in the refinement calculation. The Ramachandran plots for the main chain torsion angles have been analyzed with the PROCHECK program (32), and the and torsion angles for the nonglycine residues lie within the allowed regions. The overall geometry of the model is satisfactory, as shown in Table II.

RESULTS
Overall Structure of PotD-The crystal contains two dimeric molecules of the PotD protein in the asymmetric unit. The final structure refined at 2.5-Å resolution includes four identical protein molecules, each of which contains 325 amino acids, and one ordered spermidine molecule, in addition to 236 ordered water molecules in the asymmetric unit. The first two residues (aspartate residues 24 and 25) at the N terminus (since the signal sequence is eliminated in the crystallized PotD protein, Asp 24 is defined as the N-terminal residue) are not well defined in the electron density map, and their conformations appear to be disordered in the crystal. The primary sequence with the secondary structure elements and the ribbon representation are shown in Fig. 1.
The PotD molecule has an ellipsoidal shape with dimensions of 30 ϫ 40 ϫ 55 Å. It consists of two distinct domains divided by a deep cleft. Each domain is formed by two noncontiguous polypeptide segments. Nevertheless, the two domains are very similar in the arrangements of their secondary structure ele-ments. The first domain (N domain; residues 26 -131 and 257-302) consists of five ␤-strands and six ␣-helices. The other domain (C domain; residues 132-256 and 303-348), with a larger size, contains five ␤-strands and seven ␣-helices. The ␤-sheet within each domain is flanked by several ␣-helices on both sides. The polypeptide chain crosses over three times between the two domains, which noncovalently interact with each other by an extensive interface. The three crossing segments and the interface form a deep cleft with approximate dimensions of 20 Å long, 5 Å wide, and 14 Å deep (Fig. 1B). The PotD protein, with its many ␤-␣-␤ repeats, is classified as an ␣/␤ type. Of the amino acids, 40 and 18% are located in the ␣-helices and the ␤-sheets, respectively. The remaining amino acids (42%) belong to loops and coils. There is no substantial difference among the backbone structures of the four independent molecules in the asymmetric unit. Their root mean square (r.m.s.) 1 deviations for the superimposed C␣ atoms are as low as 0.40 Å. However, when the C␣ atoms of the N and C domains are superimposed separately among the corresponding domains in the asymmetric unit, the C domain shows a larger r.m.s. deviation value (0.42 Å) than the N domain (0.33 Å).
Subunit Contacts-The crystal contains two dimeric molecules in the asymmetric unit. The dimensions of the dimer are approximately 70 ϫ 70 ϫ 55 Å. Each monomer in a dimer is related by a noncrystallographic 2-fold axis (Fig. 2). The r.m.s. deviation values for the C␣ atoms between these two related dimers was calculated to be 0.52 Å. The dimerization mainly involves the interactions between the N domain (␤A, ␣1, and 1 The abbreviation used is: r.m.s., root mean square.  centric reflections), where I hj ϭ measured diffraction intensity, ͗I h ͘ ϭ mean value of all intensity measurements of (h, k, l) reflections, F PH ϭ structure amplitude of a derivative, F P ϭ structure amplitude of the native crystal, and F H(CALC) ϭ the calculated contribution of the heavy atoms.

Refinement statistics
The R-factor for all data (36,198 reflections) with F Ͼ 1.0 s (F) between 6-and 2.5-Å resolution is 0.207. The free R-factor (45)  Spermidine Binding-Crystals of PotD have been grown in the presence of spermidine. Indeed, the omit map at 2.5-Å resolution shows an elongated electron density (Fig. 3), which is assigned to a spermidine molecule bound to the central cleft between the two domains. The same densities have been found within all four molecules in the asymmetric unit, indicating that the bound spermidine molecules adopt the identical conformation. Interestingly, the spermidine molecule is bent within the PotD molecule, whereas all kinds of the crystal structures of spermidine in the Cambridge structural data base exhibit a linear shape (33,34).
The substrate binding site is located at the middle of the cleft between the two domains. This site forms a hydrophobic box, which is composed of four aromatic side chains, Trp 34 , Tyr 37 , Trp 255 , and Tyr 293 in the N domain and Trp 229 in the C domain. These aromatic side chains anchor the methylene backbone of the spermidine molecule through van der Waals interactions. The methylene bonds of spermidine are sandwiched between the aromatic side chains of Trp 34 and Trp 255 , which are arranged in parallel (Fig. 4A). The side chain of Trp 229 , oriented perpendicular to the previous side chains, covers the spermidine like a lid over the cleft. Another important feature in the binding site is that four acidic residues, Glu 36 , Asp 168 , Glu 171 , and Asp 257 , recognize the charged nitrogen atoms of spermidine through numerous ionic interactions (Fig. 4B). The conformations of these aromatic and acidic residues are conserved well among the four molecules in an asymmetric unit. One terminal amino group of propyl amine moiety in the spermidine forms the salt bridges with the carboxyl side chains of Asp 168 and Glu 171 , and the hydrogen bonds with the side chain of Gln 327 and Tyr 85 . The secondary amino group in the middle is recognized through the side chain of Asp 257 , and the other terminal amino group forms a salt bridge with the side chain of Glu 36 and a hydrogen bond with the side chain of Thr 35 . These aromatic and acidic side chain atoms embed the spermidine molecule in the cleft so as to prevent no solvent access.

Comparison with Other Periplasmic Binding Proteins-
Among the periplasmic binding proteins, the PotD backbone is most similar to that of MBP from E. coli, although their two sequences exhibit an identity less than 20%. In particular, the similarity between the two N domains is remarkable. When the two domains of PotD are optimally superimposed on the corresponding domains of MBP, the r.m.s. deviation values are evaluated to be 1.64 Å for the 100 C␣ atoms between the N domains and 2.63 Å for the 100 C␣ positions between the C domains. Furthermore, the active residues of the PotD and MBP proteins can be observed in similar regions of the two topologies (Fig. 5).
Another notable similarity between the two structures is that a conserved sequence motif is found in both PotD and MBP (35). This conserved sequence motif spans residues 46 -54 (FT-KETGIKV) of PotD, which corresponds to the loop between ␣1 and ␤B, and residues 53-61 (FEKDTGIKV) of MBP (Fig. 1). These sequences exhibit a remarkably similar conformation between the two molecules, as proved by the very small r.m.s. deviation value of 0.40 Å for the nine C␣ positions.
Open and Closed Forms-The crystal structures of MBP have been reported in both states of the open and closed forms (Brookhaven Protein Databank entry 1MBP, closed liganded form; 1OMP, open unliganded form), which correspond to the substrate-free form and the complex with the substrate, respectively. Substrate binding induces few conformational changes within each domain. However, it generates a substantial alteration in the relative orientation of the two domains. Substrate binding to MBP yields a hinge bending angle of 35°about an axis through the central hinge residues (residues 111 and 261) (36). The interdomain orientation of PotD is much closer to the closed form of MBP, indicating that both the PotD and MBP molecules adopt a similar domain arrangement upon binding the substrates (Fig. 6). These findings suggest that the spermidine-bound PotD molecule assumes the closed form, which presumably was converted from the open, ligand-free form.
Dimer Formation-The PotD molecule forms a dimer in this crystal structure, although PotD exists as monomeric form in the presence of the substrate (data not shown). The ionic interactions between the two subunits are so extensive that they are unlikely to have accidentally taken place during crystallization. All of the crystal structures of the periplasmic binding proteins exhibit monomeric molecules except for a MBP mutant crystal produced in the presence of maltose. This mutant crystal structure revealed a dimeric molecule (36). Furthermore, the MBP protein is purified as a dimer from an E. coli strain that is constitutive for the expression of the maltose system and that has been grown in the absence of maltose (37). Therefore, we assume that the switch from the dimer to the monomer of the PotD protein may have physiological significance in polyamine transport.
Spermidine Recognition-The residues that participate in the recognition of spermidine spread over the two domains. The N domain comes into more extensive contact with the spermidine through the walls of the hydrophobic box, while the C domain provides the lid of the box. The ligand serves as a pin that links the two domains, and it is completely embedded in the protein atoms that lie between them (Fig. 4B).
The PotD protein can bind putrescine as well as spermidine, although the affinity of putrescine is much lower than that of spermidine. The dissociation constants (K d ) for spermidine and putrescine are 3.2 M and 100 M, respectively (23). These K d values reflect the spermidine-preferential recognition for the primary receptor of the polyamine transport system. The shorter putrescine molecule could possibly make ionic interac- tions with the acidic residues of Glu 36 and Asp 257 , and form van der Waals interactions with the aromatic residues Trp 34 , Tyr 37 , Trp 229 , Trp 255 , and Tyr 293 . These interactions may stabilize the closed conformation. However, the smaller number of interactions, as compared with those with spermidine, would decrease the stability of the closed form.
Consensus Motif-The highly conserved sequence motif observed in MBP and PotD suggests a common functional role in the transport system, although mutations introduced in this region of MBP do not clearly cause a transport defect (21,38). When the PotD structure is optimally superimposed onto the MBP structure, the motif is located in the surface loop of the N domain ( Fig. 1), which is distant from the substrate binding site. On the other hand, it is directly connected to the loop ␤A to ␣1, which participates in the substrate binding and is opposite the hinge between the domains. This motif, which also lies on the molecular surface, does not participate in the dimerization. It may be possible that the physiological role of the consensus sequence is to interact with the membrane components, PotB and/or PotC, which could be an initial switch to release the substrate from the protein.
Interactions with the Membrane-bound Components-In the course of periplasmic receptor-dependent transport, the substrate is initially recognized by a specific binding protein. The subsequent translocation across the cytoplasmic membrane requires a set of membrane protein components. The membranebound components in the polyamine transport machinery are three nonidentical proteins (PotA, -B, and -C). Polyamine uptake appears to be initiated by the formation of a complex between the two membrane-bound components (PotB and PotC) and the periplasmic receptor (PotD). The substrate-free PotD protein slightly inhibits spermidine uptake to the cytoplasm (23). This result implies that the closed form of PotD is preferentially recognized by the membrane components.
In spite of many relevant reports (38 -43), mutational analyses have not clearly identified the interface of periplasmic binding proteins with their membrane protein components yet. However, it should be noted that the consensus sequence lies in the N domain, which shows a higher similarity of PotD and MBP in terms of the three-dimensional structure. Furthermore, in most of the periplasmic binding proteins including PotD and MBP (Fig. 5), the folding topology of the N domain is more conserved than the C domain (22). Therefore, it is likely that the interface of PotD with the membrane components is located in the N domain rather than the C domain.