Pseudomonas aeruginosa pyoverdine maturation enzyme PvdP has a noncanonical domain architecture and affords insight into a new subclass of tyrosinases

Pyoverdines (PVDs) are important chromophore-containing siderophores of fluorescent pseudomonad bacteria such as the opportunistic human pathogen Pseudomonas aeruginosa in which they play an essential role in host infection. PVD biosynthesis encompasses a complex pathway comprising cytosolic nonribosomal peptide synthetases that produce a polypeptide precursor that periplasmic enzymes convert to the final product. The structures of most enzymes involved in PVD chromophore maturation have been elucidated, but the structure of the essential tyrosinase PvdP, a monooxygenase required for the penultimate step in PVD biosynthesis, is not known. Here, we closed this gap by determining the crystal structure of PvdP in an apo and tyrosine-complexed state at 2.1 and 2.7 Å, respectively. These structures revealed that PvdP is a homodimer, with each chain consisting of a C-terminal tyrosinase domain and an N-terminal eight-stranded β-barrel reminiscent of streptavidin that appears to have a structural role only. We observed that ligand binding leads to the displacement of a “placeholder” tyrosine that blocks the active site in the apo structure. This exposes a large, deep binding site that seems suitable for accommodating ferribactin, a substrate of PvdP in PVD biosynthesis. The binding site consists almost exclusively of residues from the tyrosinase domain. Of note, we also found that this domain is more closely related to tyrosinases from arthropods rather than to tyrosinases from other bacteria. In conclusion, our work unravels the structural basis of PvdP's activity in PVD biosynthesis, observations that may inform structure-guided development of PvdP-specific inhibitors to manage P. aeruginosa infections.

The mobilization of iron from environmental sources is difficult because iron is usually deposited in insoluble Fe 3ϩ compounds or otherwise tightly bound, e.g. to proteins. To overcome this growth-limiting factor, bacteria produce chelating agents (siderophores) that are capable of binding ferric iron tightly and transport it into their cells. A particularly well studied group of siderophores are the pyoverdines (PVDs), 2 which are pigments and important virulence factors of fluorescent pseudomonads. Almost 70 strain-specific PVDs have been described to date (1). PVDs consist of three parts: a short peptide backbone of 6 -12 amino acids is bound to a fluorescent dihydroxyquinoline chromophore, which is connected to an additional acyl side chain of variable length (2). Ferric iron is trapped with high affinity in a stable 1:1 octahedral complex between two hydroxamate groups (occasionally ␤-hydroxyamino acids) of the peptide backbone and the catecholate groups of the chromophore (3). Three strain-specific pyoverdines (PVDI-III) are known (4) from Pseudomonas aeruginosa with PVDI from strain PAO1 being the best investigated. At least 12 enzymes are involved in PVD biosynthesis of this strain (see Fig.  1A). The initial steps are catalyzed by cytoplasmic nonribosomal peptide synthetases. PvdL, PvdI, PvdJ, and PvdD not only synthesize the PVD peptide backbone but also moieties that will eventually become the fluorescent chromophore (5). Because the composition of the peptide is strain-specific, accessory proteins like PvdA, PvdF, and PvdH provide noncanonical amino acid building blocks (6 -8). It is believed that PvdE, which is an "export" ABC transporter in the inner membrane, then transports the nonfluorescent precursor to the periplasm for further maturation by the five enzymes PvdM, PvdN, PvdO, PvdP, and PvdQ (9 -12). The myristoyl membrane anchor of the pyoverdin precursor is removed by the hydrolase PvdQ (13), and the fluorescent chromophore of PVD is furnished by PvdP (14) and PvdO (15) before the pyridoxal phosphatecontaining PvdN modifies the acyl side chain at the 3-amino group of the chromophore (16 -18). In addition to PvdQ and PvdN, the structures of PvdM (Protein Data Bank (PDB) entry code 3B40) and PvdO (19) have been determined. PvdM possesses structural similarity to metal-dependent amidohydrolases, but its exact function in PVD biosynthesis is currently unknown.
The authors declare that they have no conflicts of interest with the contents of this article. This article was selected as one of our Editors' Picks. This article contains Figs. S1-S5. The atomic coordinates and structure factors ( The biosynthesis of the fluorescent dihydroxyquinoline moiety involves an oxidative cascade (20) in which the tyrosinase PvdP acts as a cresolase (monooxidase) to form a catechol from the D-Tyr moiety of the PVD precursor ferribactin first, followed by catecholase (oxidase) activity to create an o-quinone that undergoes intramolecular cyclization before PvdO performs a final oxidation to the fluorophore ( Fig. 1B; Refs. 14 and 15). Although the identity between PvdP and other tyrosinases is low, a tyrosinase-typical type-3 dicopper center involving six essential, highly conserved histidines has been revealed by homology modeling. Based on these findings, it has been proposed that PvdP is the first member of a new tyrosinase family (14). However, no homology can be detected for a significant part of the N terminus of PvdP, suggesting that the N terminus contains a second domain of unknown function. Because PvdP is essential for PVD biosynthesis (11) and was found to be up-regulated in an acute burned mouse P. aeruginosa infection model (21), it may be a suitable target for anti-infectives, but the lack of high-resolution structures hampers the development of such inhibitors. We therefore conducted crystallization and X-ray diffraction experiments, revealing that the homodimeric PvdP indeed is a two-domain protein with an unprecedented archi-tecture, consisting of an N-terminal streptavidin-like ␤-barrel and a C-terminal tyrosinase. The active site arranges into a new class of tyrosinases from a hitherto uncharacterized branch of type-3 copper proteins. Comparison of the apo structure with a ligand complex shows that the active site is blocked by a "placeholder" residue (Tyr 531 ) in the apo form, which is typical for tyrosinases. Ligand binding displaces this placeholder and exposes an extensive binding site to host the large ferribactin substrate. The binding site consists mainly of residues from the tyrosinase domain and is lined by a small proportion of the N terminus of the second monomer, suggesting that the streptavidin-like domain primarily possesses a structural rather than a functional role.

Structure determination
PvdP was produced without the N-terminal periplasmic localization signal by heterologous overexpression in Escherichia coli and purified via Ni 2ϩ -affinity chromatography followed by a size exclusion step. Because the removal of the N-terminal His 6 tag by tobacco etch virus protease (TEV) was Crystal structure of pyoverdine maturation enzyme PvdP not successful, all subsequent experiments were performed with His 6 -TEV-PvdP(26 -544) (where TEV indicates the TEV cleavage site). The protein was active as demonstrated by dopaquinone formation from L-or D-tyrosine in the presence of Cu(II)SO 4 , and the following enzyme kinetic parameters were determined: K m ϭ 1.077 Ϯ 0.103 mM, k cat ϭ 228 Ϯ 8 s Ϫ1 for L-Tyr and K m ϭ 1.074 Ϯ 0.209 mM, k cat ϭ 197 Ϯ 13 s Ϫ1 for D-Tyr (Fig. S1), indicative of nearly identical turnover of these surrogate substrates. Of note, although K m values are similar to previously reported numbers, k cat values were approximately 200-fold higher, which may be because the N-terminal periplasmatic localization signal was not omitted in the previous report (14).
Crystallization provided only thin plates that were very sensitive to handling and usually gave only strongly anisotropic diffraction patterns if any. Initial phases were derived from single anomalous diffraction data merged from three seleno-Lmethionine-containing His 6 -PvdP crystals. Because soaking destroyed the apo crystals, cocrystallization was used to obtain protein-ligand complex structures. Ellipsoidal resolution cutoffs were applied to compensate for the anisotropic diffraction, allowing to refine the apo structure at 2.09 Å and the complex with L-tyrosine at 2.7 Å. Although the electron density was unambiguous in general, residues at the N and C termini, a loop between strands ␤6 and ␤7, and in parts of the flexible C-terminal region beyond amino acid 485 were not visible in both structures. Data collection and refinement statistics are summarized in Table 1.

Overall architecture
PvdP crystallized in space group P2 1 with four or two monomers in the asymmetric unit of the apo or complex structure, respectively. These monomers possess an unprecedented twodomain architecture consisting of an N-terminal ␤-barrel domain (BBD) and a C-terminal tyrosinase domain (TYD) connected by a short linker (residues 189 -192; Fig. 2, A and B). The interface between both domains consists of a continuous sequence on the TYD side (residues 292-339; helices ␣9, ␣10, and N terminus of ␣11), which assumes an L-shaped structure that matches the ␤-barrel on its nonsolvent-exposed face. Both domains share a 713-Å 2 interface that is filled with water molecules and has an average gap width of 3.7-4.2 Å. Only a few interactions aside of van der Waals forces stabilize the interaction between both domains, namely five H-bonds: the backbone amide of Leu 297 interacts with the carbonyl oxygen atom of Asp 86 ; the side chain of Arg 301 establishes three H-bonds with the backbone of a loop containing Leu 104 , Ala 106 , and Glu 108 ; and His 333 bridges to Asp 56 .
The monomers contained in the asymmetric unit associate to homodimers in which the BBD of one monomer tightly interacts with the TYD of the other ( Fig. 2A). The overall surface area of the dimer is 37,345 Å 2 . Analysis with PISA (22) indicates a dissociation energy, ⌬G diss , of 34.4 kcal/mol. Both molecules share an interface area of 3225 Å 2 of which the N termini of the BBD (residues 36 -47) contribute 1163 Å 2 . Removal of the N termini reduces ⌬G diss to 17.2 kcal/mol. All monomers are virtually identical and superimpose with a maximum C␣ r.m.s.d. of 0.176 Å between the four chains of the apo structure and of 0.385 Å between the apo and the ligand-bound structure.

N-terminal domain (BBD)
The N-terminal domain (residues 36 -188) comprises an eight-stranded antiparallel ␤-barrel that is closed off with helix ␣1 at one end and by the solvent-exposed loop L1 at the other. With the exception of one histidine (His 142 ), the inside of the barrel is filled by the side chains of hydrophobic residues. Of note, Trp 128 reaches especially deep into the barrel and blocks the passage (Fig. 3A). Without this side chain, a cavity with a diameter of 9.6 Å would form. Interestingly, searches with PDBeFold (http://www.ebi.ac.uk/msd-srv/ssm) 3 reveal structural similarity to streptavidin (Fig. 3B). Superposition with a monomer of WT streptavidin from Streptomyces avidinii (WTStrep) resulted in an r.m.s.d. of 2.7 Å (89 residues). Strands ␤1-␤4 of PvdP and WTStrep are similar in size (5-7 residues), but the following strands, ␤5-␤8, of PvdP are shorter (6 -9 residues compared with 10 -13 amino acids in WTStrep), giving PvdP a more symmetrical appearance. However, because the BBD of PvdP lacks the biotin-binding pocket of streptavidin, it is not surprising that soaking or cocrystallization trials with biotin did not result in incorporation of the ligand. In fact, the position of both BBDs of the PvdP dimer relative to the active sites implies a structural role, similar to structurally related BBDs found in quinohemoprotein amine dehydrogenase (PDB entry code 1JMX; Ref. 24) and in erythrocruorin (PDB entry code 2GTL; Ref. 25), two heterooligomeric proteins with functions unrelated to PvdP (Fig. S2). Although the sequence identity to the BBD of PvdP is less than 20% for both of these proteins, their potential ligand-binding sites are also blocked with bulky hydrophobic amino acids, emphasizing their structural role.
Because the similarity between BBD and streptavidin or the other mentioned proteins was not detected at the sequence level or by sophisticated structure prediction methods such as Phyre 2 (26), we used different bioinformatics approaches to identify homologues and to learn about the potential function of this domain. Although searches with HHPred (27) in various databases as well as a search in CATH (28) returned only insignificant hits, a BLAST search (29) in the NCBI reference sequence database (30) excluding pseudomonads returned five proteins from ␥-proteobacteria, namely from Enterobacter cloacae (SAJ31658.1), Cellvibrio japonicus (WP_012487482.1), Azotobacter vinelandii (WP_061289382.1), Stenotrophomonas rhizophila (KWW15420.1), and the unclassified ␥-proteobacterium L18 (WP_027977676.1). Alignment with the complete sequence of PvdP (Fig. S3) shows that these proteins share large patches of conserved amino acids also beyond the BBD; i.e. they also contain a tyrosinase domain, implying that they may be involved in a similar biosynthetic pathway. Indeed, searches with PvdO, the other enzyme involved in PVD fluorophore maturation in P. aeruginosa (15), identified similar proteins in the genomic vicinity of two of these species (C. japonicus and A. vinelandii), suggesting that these strains could synthesize related compounds. However, with respect to PvdP, three of the proteins listed above lack an N-terminal periplasmic localization signal (Fig. S3), indicating that a potential siderophore biosynthesis must be organized differently in these organisms. Interestingly, the search also returned an uncharacterized fully identical but N-terminally truncated version of PvdP designated as coming from E. cloacae e403. Surprisingly, the genome of this bacterium seems to contain most PVD biosynthesis genes, and all of them share almost 100% identity to the respective proteins from P. aeruginosa PAO1. The N-terminal truncation of the PvdP-like gene comprises the first 83 amino acids and would also affect helix ␣1 and strands ␤1 and ␤2, which will likely impede the stability of the resulting protein. This may indicate that PVD biosynthesis is not functional in the respective isolate. In summary, although theses searches do not provide indications toward the function of the BBD, they show that there may be other ␥-proteobacteria whose capacity to biosynthesize PVD-like siderophores has not been recognized.

C-terminal domain (TYD)
The C-terminal domain of PvdP is mostly ␣-helical with the exception of a small, solvent-exposed ␤-sheet (␤9a and ␤9b). Its a Statistics refer to data truncated by STARANISO to remove weak reflections affected by anisotropy (46). b The resolution limits for three directions in reciprocal space (a*, b*, c*) are indicated here. To accomplish this, STARANISO computed an ellipsoid postfitted by least squares to the cutoff surface, removing points where the fit was poor. Note that the cutoff surface is unlikely to be perfectly ellipsoidal, so this is only an estimate. c n.a., not applicable. d Values in parentheses are for the highest-resolution shell. e The anisotropic completeness was obtained by least squares fitting an ellipsoid to the reciprocal lattice points at the cutoff surface defined by a local mean I/(I) threshold of 1.5, rejecting outliers in the fit due to spurious deviations (including any cusp) and calculating the fraction of observed data lying inside the ellipsoid so defined. Note that the cutoff surface is unlikely to be perfectly ellipsoidal, so this is only an estimate.
. k R free is the same as R work with 5% of reflections chosen at random and omitted from refinement. l B-factors calculated with Moleman2 (55). m Statistics calculated with the MolProbity web server (56) (http://molprobity.biochem.duke.edu/) (Please note that the JBC is not responsible for the long-term archiving and maintenance of this site or any other third party-hosted site.).

Crystal structure of pyoverdine maturation enzyme PvdP
most prominent feature is a four-helix bundle (␣5, ␣8, ␣13, and ␣16) that provides six histidine residues to form the active site (His 216 , His 220 , His 271 , His 375 , His 379 , and His 432 ; Figs. 2, 4, and 5A). This arrangement is typical for type-3 copper proteins, and because PvdP displays tyrosinase activity (14), we refer to this domain as the TYD. Type-3 copper proteins contain two copper-binding sites termed CuA and CuB that are established by three histidines each, and it has recently been suggested that they have evolved into three subclasses that can be distinguished by the length of the sequences that separate the histidines in both copper-binding sites (31). Notably, with a His-X 3 -His-X n -His-CuA motif (His 216 , His 220 , and His 271 ) and a very long insertion between the second and third histidines of the CuB motif (His 375 , His 379 , and His 432 ), PvdP seems to fall into the ␤-subclass of type-3 copper proteins (Fig. 4). This ␤-subclass has mainly been associated with arthropods, whereas bacterial enzymes in general belong to the ␣-subclass (␣-subclass motifs are His-X n -His-X 8 -His for the CuA site and His-X 3 -His-X n -His for the CuB site, respectively). However, many other conserved ␤-subclass residues described previously (31) are not found in PvdP, suggesting that PvdP establishes a new, previously unrecognized subclass of type-3 copper proteins.
In addition to the six histidines, the TYD of PvdP contains several strictly to highly conserved type-3 copper protein residues such as phenylalanines four positions upstream of the third histidine in both copper-binding sites (CuA, Phe 267 / His 271 ; CuB, Phe 428 /His 432 ; Fig. 5A), which both point toward the active site. Other conserved residues are Pro 426 , which sits at the N-terminal end of helix ␣16 from the four-helix bundle and is probably required to provide a kink that leads to an almost 90°bend after helix ␣15 near the active site, and Asp 436 in the middle of helix ␣16 where it interacts with the highly conserved Arg 272 of helix ␣8 of the four-helix bundle.
A feature that sets PvdP truly apart from related proteins is the long insertions between the second and third histidines of both copper-binding sites. In PvdP, these sequences contain 50 (CuA) and 52 (CuB) amino acids, whereas they are much shorter in other tyrosinases (Fig. 4). For example, in the ␤-subclass tyrosinase from Drosophila melanogaster, they consist of

Crystal structure of pyoverdine maturation enzyme PvdP
only 24 (CuA) and 35 (CuB) amino acids. The extensions in PvdP lead to the formation of additional secondary structural elements such as the short helix ␣6 and the long helix ␣7 in the CuA site and to a long loop containing two short helices (␣14 and ␣15) between helices ␣13 and ␣16 in the CuB site, respectively. Another unusually long sequence is located between the last histidine of the CuA (His 271 ) and the first histidine of the CuB site (His 375 ). This stretch includes helices ␣9 -␣12, and, with the exception of ␣12, all of these helices are involved in the interface with the BBD (Fig. 2B).

Active site
No metal ions were observed in the CuA or CuB sites of the apo structure of PvdP. Instead, the CuA site was occupied by a water molecule tetrahedrally coordinated by His 216 , His 220 , His 271 , and another water molecule. The entrance to the active site is blocked by Tyr 531 , which belongs to the short helix ␣19 close to the C terminus (Figs. 2B, 5A, and S5). Superimposition with the tyrosine complex of the Bacillus megaterium tyrosinase (PDB entry code 4P6R; Ref. 32) shows that Tyr 531 occupies the substrate-binding site and hence acts as a placeholder, reminiscent of similar residues but from different regions in related polyphenoloxidases and hemocyanins (33) and again implicating that PvdP belongs to a new subclass of type-3 copper proteins. The hydroxyl group of Tyr 531 points toward the CuA site and is located within van-der-Waals contact distance to N⑀2-His 220 . The phenol ring of Tyr 531 stacks with the imidazole group of His 379 of the CuB site. The distance of the hydroxyl group toward the position of the metal ions at the CuA or CuB sites would be 3.6 and 3.9 Å as extrapolated from superimposition of PvdP apo on PvdP Tyr .
The PVD precursor ferribactin (Fig. 1B) was not available to us. We therefore attempted soaking and cocrystallization with the surrogate substrates D-and L-tyrosine in the presence of CuSO 4 but failed to obtain crystals. CuSO 4 was therefore replaced by ZnCl 2 because substitution of the cofactor Cu 2ϩ by Zn 2ϩ is known to lead to reduced activity of tyrosinases (32,34), and we hypothesized that impeded turnover would support crystallization in the presence of substrates. Indeed, crystals with L-tyrosine could be obtained, but they were of lower quality than for the apoenzyme and belonged to a different crystal form (P2 1 with one PvdP dimer in the asymmetric unit), but single anomalous diffraction data collected at the zinc absorption edge (1.28 Å) clearly revealed the presence of Zn 2ϩ ions in both the CuA and CuB sites (Fig. S4). The distance between the cations is 3.6 Å, and they are bridged by a water molecule at a distance of 2.0 Å toward each Zn 2ϩ . The most striking difference with respect to the apo structure is that not only the placeholder residue Tyr 531 is displaced from the active site but also the structure surrounding it has become disordered such that residues 514 -541 could not be traced in the electron density. This opens a large solvent-exposed and negatively charged pocket whose dimensions seem to reflect the size of the substrate molecule ferribactin (Fig. 5, B and C).
The placeholder position is occupied by poorly defined electron density that can be interpreted as MES from the crystallization buffer or as the surrogate substrate L-tyrosine. However, because superimposition with ligand complexes of related enzymes shows that tyrosine and derivatives bind in a similar fashion in these proteins, we interpreted the additional electron density as L-tyrosine (Figs. 5, A and C, and S4). The inferior quality of the electron density probably is a consequence of low solubility and weak binding of L-tyrosine, which is an order of magnitude smaller, has the opposite chirality, and is a zwitterion compared with the neutral D-tyrosyl moiety of the natural substrate ferribactin. This is also reflected in the fact that L-tyrosine, similar to D-tyrosine, is a relatively poor substrate of

Crystal structure of pyoverdine maturation enzyme PvdP
PvdP (K m values approximately 1 mM for L-and D-Tyr; see above).
Interestingly, the ligand-binding site consists mainly of residues from the tyrosinase domain, and only a small extended stretch of the first 10 amino acids from the N terminus of the second monomer lines parts of its perimeter (Fig. 5B). This again emphasizes the primarily structural function of the BBD but may also point toward a critical role of dimerization in PvdP.

Discussion
The structure analysis presented here reveals that PvdP possesses a novel two-domain architecture exclusively found in a small number of ␥-proteobacterial species. Although the C-terminal domain has a core architecture commonly found in tyrosinases, the N-terminal domain is unique in primary sequence but resembles streptavidin in tertiary structure. Interestingly, sequence database searches identified one protein (GenBank TM accession number SAJ31658.1) that, with the Crystal structure of pyoverdine maturation enzyme PvdP exception of a deletion of the first 60 amino acids, is 100% identical to PvdP. The sequence of this protein was derived by whole-genome shotgun sequencing from a biological sample and was assigned as coming from E. cloacae strain e403. However, the high sequence identity makes the assignment to E. cloacae questionable, and indeed, other proteins from the same strain deposition are identical to P. aeruginosa proteins as well, corroborating this assumption. Conversely, the finding that searches with the PvdP sequence identified similar proteins in species not previously investigated for siderophore production such as C. japonicus, A. vinelandii, S. rhizophila, or ␥-proteobacterium L18 shows that PVD production may be more widespread than anticipated.
PvdP is only the third bacterial tyrosinase whose structure has been determined. The other two representatives are tyrosinases from Streptomyces castaneoglobisporus (TyrSC; PDB entry code 1WXC; Ref. 35) and B. megaterium (TyrBM; PDB entry code 3NM8; Ref. 36), but they are only relatively distantly related to PvdP. In fact, sequence analysis places PvdP closer to arthropod rather than bacterial tyrosinases (31), albeit still with large dissimilarities at the sequence level (Fig. 4). The evolutionary distance of PvdP may also be responsible for the failure to discover related structures of the C-terminal domain with default parameters in the structure similarity search program PDBeFold (http://www.ebi.ac.uk/msd-srv/ssm). 3 Such relatives could be identified with DALI (37), clearly revealing similarities to other type-3 copper proteins such as hemocyanins, arylphorins, and phenoloxidases in addition to tyrosinases. According to this analysis, the closest homologue to the tyrosinase domain of PvdP is the type-3 copper protein domain of a prophenolxidase from Manduca sexta (PDB entry code 3HHS; Ref. 38), which aligns with an r.m.s.d. of 3.9 Å over 177 residues (Z-score ϭ 11.3). However, the overall structure of this protein is grossly different from PvdP, and the sequence identity within the aligned structural elements is only 14%.
The large evolutionary distance of PvdP to related proteins offers an opportunity to re-evaluate the importance of several sequence motifs that have been identified as having key roles in other tyrosinases (for a recent review, see Ref. 39). These enzymes execute a complicated reaction cycle that involves different redox states of the two copper atoms and can eventually lead to loss of the metal due to side reactions that generate Cu 0 atoms. This may hint at generally rather weak metal affinity to enable enzyme reactivation by copper reloading and could explain why we observed the apoprotein in a metal-free state. At the same time, low metal affinity would require evolving means for recharging the enzyme with Cu 2ϩ , and indeed, several such mechanisms are discussed for different types of type-3 copper proteins. For example, the role of two structurally neighboring methionines in supplying the protein with Cu 2ϩ has been demonstrated in TyrBM (Met 61 and Met 184 in TyrBM; Ref. 40). These residues are not conserved in PvdP, but interestingly, a similar solvent-exposed methionine motif (Met 270 and Met 274 in PvdP) that shields the active site and may hence serve as a copper load port can be found at another location in PvdP, suggesting a similar role that deserves attention in future studies (Fig. 5A). Another highly conserved methionine that may be involved in H 2 O 2 scavenging via sulfur oxidation (39), Met 215 in TyrBM, is replaced by a leucine in PvdP (Leu 416 ), which may reflect the fact that P. aeruginosa has other very effective detoxification systems for reactive oxygen species (41).
Mechanistic studies indicate that the phenol group of the substrate has to be deprotonated to initiate binding to the CuA site of tyrosinases. Experimental and in silico evidence suggests that this is achieved by a conserved glutamate/asparagine dyad (Glu 195 and Asn 205 in TyrBM) that binds a water molecule to perform the deprotonation (32). In PvdP, the glutamic acid is conserved (Glu 371 ), but we found an aspartic acid (Asp 376 ) instead of asparagine. Nevertheless, a water molecule bound in a similar place as in other tyrosinases can be observed in the better resolved apo structure, indicating that PvdP applies a similar mechanism for substrate activation (Fig. 5A).
An interesting feature of most type-3 copper proteins is the presence of a placeholder residue that occupies the substratebinding site in the resting state of the enzyme and needs to be displaced via structural rearrangements or by proteolytic cleavage to activate the enzyme. PvdP provides a new placeholder motif to the type-3 copper protein family by bearing the placeholder residue Tyr 531 in an ␣-helix within its flexible C terminus as indicated by high B-factors and the partial absence of traceable electron density. The fact that the enzyme was active in in vitro assays both toward the model substrate tyrosine and toward ferribactin (14) indicates that PvdP does not undergo proteolytic activation but rather uses displacement of the last 30 amino acids (beyond Phe 512 at the C terminus of helix ␣18) to provide a binding interface for the large PVD precursor molecule. In the apo structure, Tyr 531 binds the active center at a position in which the hydroxo group of the side chain is 1.4 Å more distant to the CuA site than in the complex with the tyrosine ligand. This probably avoids autoxidation of Tyr 531 and is achieved by locating Tyr 531 in a short -helix (Pro 529 -Arg 533 ) that interacts with Gly 417 at its N terminus, thus avoiding further slipping of Tyr 531 into the active center (Fig. 5A).
The finding that the N-terminal domain adopts a streptavidin-like fold that could not be predicted from its sequence was a surprise to us. The current analysis of the structure suggests that it takes a structural rather than a functional role by establishing contacts with the tyrosinase domain of the second chain of the PvdP homodimer, leading to an unprecedented overall structure within the type-3 copper protein family. This family is known to contain largely different quaternary structural arrangements, reaching from monomers such as TyrBM to large complexes consisting of up to 48 chains as in the case of hemocyanin from the horseshoe crab Limulus polyphemus (42). The diversity of these structures probably reflects a long evolutionary history of this protein family, explaining why PvdP deviates in so many details from previously studied tyrosinases and establishes a previously unrecognized subclass of type-3 copper enzymes.
In summary, the data presented here unravel the structural basis of the activity of PvdP in pyoverdin biosynthesis. The finding that PvdP, although keeping essential residues involved in the chemistry catalyzed by tyrosinases, replaces several sequence motifs involved in mechanisms not directly associated with catalysis provides deeper insight into this protein family and may also serve as a starting point for the structure-Crystal structure of pyoverdine maturation enzyme PvdP guided development of PvdP-specific inhibitors against disease inflicted by P. aeruginosa. Toward this, the large body of known natural and synthetic tyrosinase inhibitors (43) should provide leads into the chemical nature of such compounds.

Chemicals and reagents
All chemicals were from Sigma-Aldrich unless otherwise indicated. Molecular biology reagents were purchased from Fermentas.

Expression and protein purification
PvdP expression was achieved in E. coli Rosetta2(DE3) (Novagen) using 1 liter of SOC medium supplemented with 30 mg/liter kanamycin at 37°C at 130 rpm. Induction with 0.1 mM isopropyl 1-thio-␤-D-galactopyranoside was started when cell density reached A 600 nm 0.6 -0.8 at which point the temperature was decreased to 20°C. Cells were harvested after 20 h of incubation. Selenomethionine (SeMet)-containing PvdP was obtained as follows. Precultures were grown in 100 ml of LB including 30 mg/liter kanamycin and incubated overnight at 37°C. 20 ml of the culture were harvested, washed twice with M9 minimal medium without antibiotic, and used for starting 1-liter cultures of M9 medium including kanamycin. After reaching an A 600 nm of 0.5, an amino acid mixture (100 mg/liter Lys, Phe, and Thr and 50 mg/liter Ile, Leu, and Val) was supplied. After a 0.5-h incubation at 20°C, 60 mg of SeMet powder and 0.1 mM isopropyl 1-thio-␤-D-galactopyranoside were added. Cultures were shaken at 130 rpm for 24 h before harvesting. For purification of protein for crystallization experiments, 2 (native) or 4 liters (SeMet) of cell culture were used. Purification started with a 5-ml HisTrap FF column (GE Healthcare) and buffer A (50 mM Tris/HCl, pH 8.0, 0.1 M NaCl) versus buffer B (50 mM Tris/HCl, pH 8.0, 0.1 M NaCl, 0.5 M imidazole) on an ÄKTApurifier system (GE Healthcare). Because cleavage of the tag was not successful, His 6 -tagged PvdP was directly run on an Superdex S75 26/600 gel filtration column (GE Healthcare) using buffer A. Using this buffer system, PvdP reversibly aggregated when concentrated to more than 12 mg/ml.

Enzyme kinetic measurements
Enzyme kinetic parameters for L-and D-tyrosine were determined with a colorimetric assay detecting the generation of dopachrome at 475 nm (⑀ ϭ 3600 M Ϫ1 cm Ϫ1 ; Ref. 14) in an Evolution 260 UV-visible spectrophotometer (Thermo Fisher Scientific) thermostated at 303.15 K. The substrate concentration was varied between 0 and 4 mM by mixing the required ratios of buffer (50 mM CHES, pH 9, 0.25 mM CuSO 4 ) containing no or 4 mM substrate to 1 ml in 1-cm plastic cuvettes. The reaction was initiated by adding 126 g of His 6 -tagged PvdP and then followed for 300 s in 10-s intervals. All measurements were performed in triplicates and evaluated with the Enzyme Kinetics Module in SigmaPlot (Systat Software, Inc.) using a simple Michaelis-Menten model.

Crystallization
PvdP was crystallized with the sitting drop vapor diffusion method in 96-well format using Intelli-Plates (Art Robbins Instruments) at 293.15 K. Native PvdP concentrations ranged from 4 to 12 mg/ml, and promising crystals were identified with a precipitant consisting of 20% PEG 3350 and 0.18 M ammonium citrate. These crystals could only be optimized by microseeding (1:10 to 1:1000 diluted seed stock). Final crystals were small thin plates and often contained defects. For SeMet-protein, conditions had to be rescreened, resulting in 25% PEG 3350, 0.2 M NaCl, 0.1 M Tris/HCl, pH 8.4, as the precipitant. Reliability of crystal growth was again enhanced using microseeding. For ligand-bound PvdP, the buffer was changed to 50 mM CHES, pH 9.0, prior to crystallization (14), which resulted in better solubility of PvdP. Crystals grew in 0.86 M (NH 4 ) 2 SO 4 , 0.1 M MES, pH 5.5. Cocrystallization was used to obtain ligand complexes. PvdP was incubated with 0.5 mM ZnCl 2 and 1 mM L-tyrosine for 0.5 h on ice before setting up crystallization experiments. Prior to flash cooling in liquid nitrogen, crystals were washed in precipitant supplemented with 10% (2R,3R)-2,3-butanediol and 0.5 mM L-tyrosine, 0.3 mM ZnCl 2 in the case of cocrystallized PvdP.

Data collection
X-ray diffraction data were collected at 100 K at the PETRAIII (Deutsches Elektronen-Synchrotron (DESY), Hamburg, Germany), BESSYII (Berlin, Germany), and SLS (Paul Scherrer Institute, Villigen, Switzerland) synchrotrons. For PvdP apo and PvdP Tyr , data from single crystals were collected on beamline P11 (PETRAIII) or beamline 14.2 (BESSYII) using 3600 nonoverlapping frames of 0.1°. Data were indexed and integrated with XDS (45) and then submitted to the STARANISO server (http://staraniso.globalphasing.org/cgi-bin/ staraniso.cgi) 3 to calculate the ellipsoidal resolution limit, setting 1.5 I/(I) as the lowest acceptable signal. Because no search model for molecular replacement was available, PvdP SeMet was used for phasing. To improve the anomalous signal, multiple SeMet-containing crystals were measured, and data sets were tested for scalability using XSCALE (45). Data of three crystals were sufficiently similar to be merged: of crystal I, four wedges (-rotation of 0°, 10°, 20°, and 30°) were measured and scaled to 10 wedges from crystal II (seven data sets at 0°, 5°, 10°, 15°, 20°, 25°, and 30°and three data with -/-rotation of 7°/10°, 17°/10°, Crystal structure of pyoverdine maturation enzyme PvdP and 27°/10°). These data were collected at the SLS on beamline PXIII at a wavelength of 0.978 Å. However, more data were required to obtain initial phases, which were contributed by crystal III measured at PETRAIII beamline P11 at a wavelength of 0.979 Å. Scaling all of these SeMet data with XSCALE increased redundancy to 49-fold and led to usable anomalous signal to 3.5-Å resolution. After data reduction, SeMet data were converted using XDSCONV (45).

Structure solution and refinement
Structure solution and refinement were carried out using programs from the Phenix software suite (47). Initial phases were obtained with the HySS subroutine of Autosol using the scaled and merged SeMet data and an apo-PvdP data set to 4.0 Å. The output was then used for Autobuild and Buccaneer from the CCP4 suite (48,49). Both programs calculated different parts of the structure and still contained missing or misplaced connections, which were then curated manually. The initial model was used for phasing the PvdP apo and PvdP Tyr data. Iterative refinement was done with phenix.refine after manual inspection using Coot (50). All figures were prepared with PyMOL (51). Coordinates and diffraction data have been deposited in the PDB (52) with entry codes 6EYS for PvdP apo and 6EYV for PvdP Tyr .