Effects of copper occupancy on the conformational landscape of peptidylglycine α-hydroxylating monooxygenase

The structures of metalloproteins that use redox-active metals for catalysis are usually exquisitely folded in a way that they are prearranged to accept their metal cofactors. Peptidylglycine α-hydroxylating monooxygenase (PHM) is a dicopper enzyme that catalyzes hydroxylation of the α-carbon of glycine-extended peptides for the formation of des-glycine amidated peptides. Here, we present the structures of apo-PHM and of mutants of one of the copper sites (H107A, H108A, and H172A) determined in the presence and absence of citrate. Together, these structures show that the absence of one copper changes the conformational landscape of PHM. In one of these structures, a large interdomain rearrangement brings residues from both copper sites to coordinate a single copper (closed conformation) indicating that full copper occupancy is necessary for locking the catalytically competent conformation (open). These data suggest that in addition to their required participation in catalysis, the redox-active metals play an important structural role.

S ecreted peptides function as hormones, neurotransmitters, and growth factors. In the animal kingdom, many of these peptides-most of them produced by enzymatic cleavage of large precursors-must be amidated at their carboxy terminus to exhibit full biological activity. Surprisingly, these amides are not generated by an amination reaction. Instead, precursors or intermediates with a glycine at their C-terminus are transformed into active, terminally amidated des-glycine hormones by oxidative cleavage of the glycine N-Cα bond [1][2][3][4][5] . Two enzymes, peptidylglycine α-hydroxylating monooxygenase (PHM; EC 1.14.17.3) and peptidyl-α-hydroxyglycine α-amidating lyase (PAL; EC 4.3.2.5) working sequentially are the only proteins known to catalyze this reaction. A gene encoding active, bifunctional, integral membrane peptidylglycine α-amidating monooxygenase (PAM) was recently identified in Chlamydomonas reinhardtii, a unicellular green alga, raising the possibility that PAM existed in the last eukaryotic common ancestor 6,7 . In C. reinhardtii, as in placozoans and sponges-organisms that lack both nervous and endocrine systems-and in most animals, the PHM catalytic core follows an N-terminal signal sequence and is followed by the PAL catalytic core, a single transmembrane domain and an unstructured cytosolic domain 4,5,[7][8][9][10] . The PAM gene appears to have duplicated and in some organisms, including Drosophila melanogaster, PHM and multiple PAL proteins are encoded by separate genes 7 . It is now clear that a PAM protein that requires ascorbate, copper, and molecular oxygen appeared before the evolution of the nervous system. PHM, a two Cu enzyme, catalyzes the stereospecific hydroxylation of the glycine Cα of peptidylglycine substrates. PAL, a Zn-containing enzyme, completes the reaction, yielding the αamidated peptide product plus glyoxylate. Both domains have broad substrate specificity: peptides with all 20 amino acid amides have been isolated 3 . The cloning and successful expression of PHM, PAL, and bifunctional PAM have made detailed structural and functional studies possible [11][12][13][14][15] . The structures of the PHM and PAL domains of rat PAM, alone and in complexes with substrates, inhibitors, and other ligands, have been determined [11][12][13][14][15] . Other techniques used to study PHM have ranged from kinetics and kinetic isotope effect measurements [16][17][18][19][20] , to inhibitor design [21][22][23][24] , spectroscopy [25][26][27][28][29] including X-ray absorption spectroscopy (XAS) 17,[30][31][32][33] , and computational studies 11,[34][35][36] . These studies have provided significant insights into the mechanism of both domains, especially of PHM.
The reaction carried out by PHM-hydroxylation of an aliphatic carbon-is chemically highly sophisticated. The two copper atoms in PHM, Cu H , and Cu M (located in the N-and Csubdomains, respectively), each use a single reducing equivalent from ascorbate to catalyze the reduction of molecular oxygen for the hydroxylation of the Cα of glycine at the carboxy-terminus of the peptide substrate 37,38 . PHM is one of a limited number of enzymes that require copper for catalytic function; the residues that bind Cu H and Cu M are perfectly conserved from C. reinhardtii to human PHM. The transporters and chaperones needed to utilize copper in the secretory pathway are also highly conserved, suggesting an equally sophisticated metalation mechanism. The catalytic chemistry must therefore proceed within the confines of a complex cellular transport machinery, with the overall functioning of the system needed to balance the requirements of selective metalation with the structural determinants of catalytic function.
The studies enumerated above, especially X-ray diffraction, revealed numerous ancillary properties of PHM that may or may not be necessary for catalysis: Cu H and Cu M are 11 Å apart in both the reduced and the oxidized states; the peptide substrate provides the path for the electron transfer; Cu H has an empty coordination site that remains empty even in the presence of high concentrations of small-molecule copper ligands; O 2 binds only to Cu M ; O 2 binds with end-on geometry and only in the presence of substrate; H 2 O 2 can bypass the requirement for O 2 and for an additional source of reducing equivalents but binds Cu M with side-on geometry. It is unknown whether these highly specific features are absolute mechanistic requirements or instead only evolutionary events that accompanied organismic specialization. A mechanism as sophisticated as that of PHM provides a unique platform for uncovering the differences among the evolution of the chemical requirements for product production, the constraints on metal ion coordination imposed by the metalation machinery, and those resulting from subsequent evolutionary events. The highly conserved Cu H and Cu M sites in PHM suggest that many of these features may be essential for function.
Information necessary to address these questions includes the effect of missing copper in one or both sites and the effect of modifying the coordination of Cu H on the structure, activity and other properties of PHM. Cu H is coordinated by three highly conserved histidine residues: H107, H108, and H172. Mutation of any one of these residues to alanine results in an inactive enzyme 1,39,40 . This loss of activity may be due to different consequences of the mutations. For example, do the mutations weaken or prevent Cu binding to the modified Cu H site, with loss of one of the two required electrons for O 2 reduction? Or, do the mutant proteins, even though they have an intact Cu M site-the site of substrate binding-fail to bind substrate? In this work, we describe structural studies on rat PHM aimed at addressing some of these important questions. Crystallographic data on apo wildtype PHM and on several crystal forms of three Cu H mutants (H107A, H108A, and H172A) presented here show that the copper ions are required not only for the catalytic steps but also for substrate binding and for locking the overall conformation of PHM in a configuration that is catalytically active. Absence of copper, or modified copper coordination, has a major effect on the overall flexibility of PHM, allowing the molecule to adopt conformations that have not been observed with wild-type PHM.

Results
Structure of rat apo-PHM. All of the PHM structures published to date contain two copper ions. To assess the possible structural role of copper, the structure of apo-PHM (PDB ID: 5WKW), which lacks bound copper, was determined to a resolution of 1.8 Å ( Table 1). The final R-values are R work = 20.9% and R free = 27.3%. As expected, there was no density for either of the two coppers-Cu H or Cu M (Fig. 1). Other than this, the structure is strikingly similar to the wild-type holoenzyme (PDB ID: 1PHM) (RMSD 1.6 Å for 1214 main chain atoms), with the largest differences restricted to the loops connecting the β-strands (Supplementary Figure 1A).
Structures of the PHM mutants (H107A, H108A, and H172A). The structures of each of the Cu H site PHM mutants were determined by X-ray diffraction in the absence and in the presence of 1-3 mM citrate, used as an additive to improve resolution. None of the crystallization conditions included Cu 2+ . Only the structures of crystals that diffracted to the highest resolution were fully refined and are reported here. All structures were determined without the addition of ascorbate and it is therefore expected that all observed coppers are in the oxidized state (Cu 2+ ).
Structure of the H107A-PHM mutant. The crystals of the Cu H mutant, H107A-PHM (PDB ID: 6ALV), grown in the conditions used for wild-type PHM, diffract to only 3.5 Å resolution. Nevertheless, the structure was determined by molecular replacement and refined to an R-value of 27.6% (R-free = 29.7%) with excellent geometry ( Table 1). The structure, except for the lack of Cu 2+ in the Cu H site, is very similar to that of wild-type PHM (RMSD = 0.42 Å for 283 Cα carbons) (Fig. 1c). Despite lacking its Cu 2+ , the rest of the Cu H site shows little change except for the absence of the side chain of residue 107. The mutant does have Cu 2+ in the Cu M site and the geometry of the coordination is almost identical to that of wild-type PHM (Fig. 1c). The main difference is a small (<0.5 Å) narrowing of the space between the two domains ( Supplementary Fig. 1B [41][42][43][44][45] .) The geometry is tetrahedral and is composed of the remaining two histidines (H108 and H172) plus two water molecules (Fig. 2c). In contrast, in molecule D, no density is found for Cu H (Fig. 2b). Instead, the two histidine residues from the former Cu H site interact with a bound citrate which, in turn, has a weak interaction with Cu M mediated by H 2 O (Fig. 2d). Citrate binds close to the empty site of Cu H and interacts directly with H108 and H172; the carboxylate of the citrate partially occupies the space that in wild-type PHM forms the peptide binding site (Fig. 2d). In addition, there is a movement of the loop spanning residues 45-53.
Structure of the H108A-PHM mutant. PHM H108A (PDB ID: 6AO6) crystallizes with the same unit cell and space group as wild-type PHM. In the structure, determined to 3.0 Å resolution, H108A-PHM shows no major conformational changes compared to wild-type PHM (Supplementary Fig. 1C) even though there is no electron density for copper at the Cu H site (Fig. 1d, Table 1). Crystals of H108A-PHM (PDB ID: 6ALA) obtained in a buffer containing citrate (3 mM) diffracted to 2.5 Å resolution and have a different space group and cell dimensions from those of wildtype PHM and H107A-PHM-cit (Fig. 3, Table 2). In these crystals there are two identical molecules with a bound citrate in the asymmetric unit present in a conformation significantly different from that of wild-type PHM. When the N-subdomain of the H108A-PHM-cit structure is aligned with that of wild-type PHM, the C-subdomain displays a rotation of~18°toward the N-subdomain, compared to its position in the wild-type PHM structure. As a consequence of this hinge movement, several secondary elements of the C-subdomain move as much as 15 Å closer to the N-subdomain (Fig. 3b). These changes result in a complete reorganization of the coordination of the copper in Cu M . The two histidines, His 242 and His 244, remain coordinated to the copper, but Met 314 and the fourth ligand in wild-type PHM (a water molecule) do not bind to the ion. Of the two vacated Cu 2+ coordination positions, one is occupied by the 2-carboxylate of citrate, and the other, surprisingly, by the N ε of His 107, originally a copper ligand of the Cu H site (Fig. 3c). In the wild-type PHM structure, the distance between His242 N ε and His107 N ε is 11.7 Å, while it is 3.9 Å in the H108A-PHM-cit structure (Fig. 3). The participation of His 107 in the coordination of Cu 2+ at the Cu M site brings the two domains of PHM closer together by approximately 7.8 Å.
Although citrate was added to the crystallization medium just to promote diffraction to higher resolution 46 does not contain copper in the Cu H site (Fig. 1e), copper at the Cu M site has the same coordination and geometry as the wildtype. The Cu H site displays only the minimum changes compatible with the mutation and the absence of its copper (RMSD 0.61 Å) (Fig. 1e, Table 1). The N-and C-subdomains come closer by a small displacement (<0.5 Å) ( Supplementary Fig. 1D).
Binding of peptide substrate. The structure of oxidized wildtype PHM with the substrate N-acetyl-di-iodotyrosyl glycine (N-Ac-di-I-YG) bound has been determined previously 12 . To assess whether the copper ions are required for substrate binding, crystals of apo-PHM (PDB ID: 5WM0) were soaked in solutions containing 1 mg/mL of N-Ac-di-I-YG for up to 12 h. Data were collected from these crystals which diffracted to 2.4 Å resolution and the structure, refined to an R factor /R free = 0.20/0.28, shows no density for the bound peptide (Fig. 4a, Table 2). Similarly, the refined structures of crystals of the H108A (PDB ID: 6AY0) and H172A (PDB ID: 6AN3) mutants grown without citrate and soaked in solutions of mother liquor containing 1 mg/ mL of N-Ac-di-I-YG for 2-12 h showed no density corresponding to the bound peptide (Fig. 4b, c, Table 2), suggesting that the presence of both copper ions is required for substrate binding. Co-crystallization with solutions containing 1 mg/mL peptide resulted in crystals that again did not show density for the bound peptide ( Supplementary Fig. 2).
Comparison of DBH structure with PHM. DBH contains a catalytic domain (DBH-cat; residues 209 to 507) that is related in sequence (human DBH to rat PHM-29% identity and 44% similarity), copper content, structure 49 , and mechanism [50][51][52] to PHM (Fig. 5a). Similar to PHM, DBH-cat has two subdomains -N-terminal (residues 209 to 356) and C-terminal (residues 357 to 507)-each containing a copper in the active form of the enzyme (also called Cu H and Cu M ). The reported crystal structure of DBH (PDB ID 4ZEL; 2.9 Å resolution) contains two molecules in the asymmetric unit, referred to as molecule A and molecule B. Only one of the four possible copper ions (Cu M in molecule A) is reported in the PDB, albeit with low occupancy (reflected by the high B = 151.88). The structures of the individual DBH-cat subdomains are highly similar between the two molecules in the asymmetric unit (RMSD: N-terminal 0.73 Å for 141 Ca, C-terminal 0.21 Å for 150 Ca) even though the conformations of the DBH-cat domains in the two molecules are quite different (RMSD 4.10 Å for 260 Ca). In DBH molecule A, the two subdomains come closer together (closed arrangement), shortening significantly the distance between the putative Cu H and the Cu M sites (distance between DBH Cu H and Cu M sites is about 4-5 Å while in wild-type PHM, this distance is 11.7 Å). In DBH molecule B, the arrangement of the two subdomains is similar to that of wild-type PHM, making the distance between the putative Cu H -Cu M sites~11 Å ( Supplementary  Fig. 3).  (Fig. 5c), respectively. Nevertheless, the overall conformation of both DBHcat molecules in the crystal structure is different from that observed in the structure of wild-type PHM and the highly similar apo-PHM. Molecule A is the most different (RMSD 3.21 Å for 219 Ca). Although, the DBH-cat of molecule B is more similar to wild-type PHM, it still shows significant differences (RMSD 2.49 Å for 213 Ca).
The arrangement of subdomains in the DBH-cat of molecule A (closed) is highly similar to that of the H108A-PHM-cit structure (RMSD 1.19 Å for 195 Cα), including the position of the single copper ion of each structure (difference in position of the copper atom in the two aligned structures <1.6 Å). The similarity of these two structures, which contain a single copper, suggests that they may represent one-copper intermediates in the assembly of the two-copper competent PHM.

Discussion
The structures of metalloenzyme active sites are exquisitely tailored to their catalytic function. PHM is an example of a copper enzyme that utilizes highly evolved reaction chemistry to catalyze a difficult hydroxylation believed to proceed via cupric-superoxomediated radical chemistry. However, this catalytic reactivity must proceed within the confines of equally sophisticated cellular transport mechanisms designed to ensure that the enzymes are metalated selectively and in response to cellular signals. In mammalian cells, the copper homeostatic machinery utilizes a pathway comprised of importers, chaperones, and energy-driven membrane pumps that eventually results in metalation of the catalytic metal centers via transporter-enzyme complexes 53 . Thus, the overall functioning of the system requires a fine balance between the requirements for selective metalation and the structural determinants of catalytic function. It has been suggested that ATP7A, a P-type ATPase, is a major required component of the system that metalates the PHM catalytic center 53 . It is likely that a component of this copper transfer mechanism may involve shared ligand complexes at the H-center where one or more of its histidine copper-ligands do not coordinate the copper. To address these issues, it is necessary to determine (a) the structural elements at the active site that facilitate the catalytic chemistry and (b) the conformational landscape which enables the enzyme to mature from its apo to the fully metalated catalytically competent form.
Since the Cu H site mutants exhibit dramatically reduced activity, their structures may provide clues to both of these objectives. As a step toward this goal, we have determined the structures of the apo-enzyme and three His variants at the Hcenter-H107A, H108A, and H172A, all without the addition to Cu 2+ in the crystallization media. All the mutants have previously been characterized in solution for kinetic parameters, metal content, and structural integrity using steady-state kinetics, inductively coupled plasma-optical emission spectrometry (ICP-OES), emission paramagnetic resonance (EPR), and XAS 19,33,54-56 . All of the mutants appeared to bind substrate in their di-copper forms with K m values between 3 and 18 μM (wildtype K m which is 8.3 μM) using dansyl-tyr-val-gly as substrate.
The decrease in activity was thus almost entirely due to the large (>150 fold) decrease in k cat for this substrate 56 . In the structures with one or both coppers missing, no substrate binding was observed, either by soaking or co-crystallization. It is clear that in solution, H-site mutants can exist in forms that retain copper and substrate binding capacity, and the finding that crystal forms exist for H108A and H172A that are isostructural with wild-type broadly confirms these findings. However, the crystallographic determination of other conformers that either lack copper at the H-site, or that exhibit different conformations and/or binding modes, suggests that the mutations stabilize alternative conformers that may be intermediates in catalysis or metal transfer chemistry. The crystal forms that lack copper at the H-site may be stabilized by crystal-packing constraints or have lost copper to the copper-free crystallization media because of the weaker affinity of Cu(II) for the histidinedepleted site. Notwithstanding the fact that this result was unexpected, and recognizing that the mutants in solution may retain their di-copper structures, the crystal structures reported herein provide an unprecedented window into the role of metal occupancy in modulating the conformational landscape of PHM.
While loss of both coppers in the structure of apo PHM has no significant effect on the conformation of the molecule, mutations that cause loss of copper from a single site appear to have significant effects. We have previously shown that PHM mutations at the Cu M (M314I) site result in conformational changes and a reduction in thermal stability 14 . The modifications of the Cu H site reported in the present paper also result in significant structural effects. In wild-type PHM, this site has an unusual geometry: Cu 2 + is only coordinated by three histidines (His 107, His 108 and His 172) with a T-shaped geometry. The fourth coordination position has been shown to remain unoccupied even in the presence of high concentrations of strong copper liganding small molecules (nitrite, azide, CO) 15 , which is unusual, given the strong preference of cupric centers to adopt square or tetragonal geometry. Here, insights into the Cu H site were gained by studying the effects of substituting each of the three histidine ligands by alanine, one at a time.
The structure of the H107A mutant crystallized in the same conditions as the wild-type is highly similar to that of wild-type PHM. The coordination of the copper at the Cu M site remains unchanged and even the two remaining histidines at Cu H site show only minor changes (Fig. 1c). Inclusion of 1-3 mM citrate, which produced crystals (H107A-PHM-cit) that diffracted to higher resolution, had unexpected effects on the structure of H107A-PHM and provided information about alternative conformations that can be adopted by this mutant (Fig. 2). H107A-PHM-cit has two molecules in the asymmetric unit. Interestingly, the copper sites of these two molecules (A and D) present in the same unit cell have very different configurations (Fig. 2), probably because citrate is only present in one of the molecules (molecule D). In molecule A, copper binds at the Cu H site despite the absence of His 107 by retaining the coordination with His 108 and His 172 and adding two water molecules as ligands. Even though molecule A does not have a bound citrate, it shows a change in the conformation of the loop containing residues 126 to 130 that brings Glu 128 into the proximity of the Cu M site (Fig. 2a). Molecule D, in contrast, has no copper at the Cu H site and contains a bound citrate that bridges the two remaining histidines of the copperless Cu H site to the Cu M site (Fig. 2b). It is clear that these changes are a consequence of the combination of the two modifications: the substitution of His 107 and the presence of citrate. Loss of coordination by H107 allows the Cu H site to adopt a tetrahedral coordination by the inclusion of two water molecules, implying that square planar geometry at the H site is destabilized relative to the tetrahedral alternative. This observation strengthens the case for a functional requirement for a nonreactive open coordination site in Cu H that allows electron transfer but prevents binding of small molecules that could modify the electrochemical potential of the copper. How the particular coordination of wild-type PHM accomplishes this feat remains unexplained.
Crystals of H108A-PHM, despite crystallizing in the same space group as wild-type PHM with similar cell dimensions, diffract only to 3.0 Å resolution and the structure shows no density for the copper in the Cu H site. In every other respect, the structure of this mutant is highly similar to that of the wild-type (RMSD 0.57 Å for 1220 atoms of the main chain). The situation is different for the crystals obtained in the presence of 1-3 mM citrate: they diffract to 2.5 Å, the space group and cell dimensions are different from those of the wild-type and contain two identical molecules in the asymmetric unit with bound citrate. Neither molecule contains copper in the Cu H site. Compared to wild-type PHM, the C-subdomain shows an~18°rotation around an axis that goes through residue 201. The 2-carboxylate of the bound citrate coordinates the copper at the Cu M and leads to a rearrangement of the copper coordination sphere (Fig. 3). The two histidines, His 242 and His 244, remain liganded to the copper but Met 314 does not coordinate the copper any longer. The water that occupies the fourth ligand position in wild-type PHM is replaced, surprisingly, by the N ε of His 107-which in wildtype PHM coordinates the copper at Cu H . This interaction requires that the N-and the C-domains come closer together. The changes in conformation that result in this approach are extensive. The most salient feature is the flattening of the β-sheet of the N-terminal domain closest to the central cavity, which involves straightening of strand 5, the strand that starts with residue 107. It is difficult to quantitate these changes but one possible measure of the effect of this straightening may be the change in the distance between the α-carbons of residues Met 314 and His 107. In a b c  Fig. 4 Binding of a peptide substrate to apo-PHM and to two PHM mutants-H108A and H172A. Superposition of wild-type PHM + peptide structure (olive) with apo-PHM in light blue (a), H108A-PHM in dark green (b), and H172A-PHM in pink (c). The peptide N-Acetyl-di-iodotyrosyl glycine (Ac-Di-I-YG) in wild-type PHM is shown in olive. No peptide bound to apo-PHM nor to either of the two mutants wild-type PHM, the distance is 19.6 Å and it is reduced to 12.5 Å in the H108A-PHM citrate complex. A measure of the effects that these changes in coordination have in other sections of the βsheet involved in the Cu H site-i.e., strand 9-is reflected in the distance between Met 314 and His 172; this distance is 17.9 Å in wild-type PHM and 13.8 Å in H108A-PHM-cit. Other notable changes are those in the loop spanning residues 129 to 136. This loop connects strands 5 and 7, and although its local conformation remains unchanged from that in wild-type PHM, its overall position is shifted away from the C-terminal domain. Again, these changes seem to be due as much to the mutation as to binding of citrate. In any case, they reflect the range of possible conformations that the molecule can adopt. The H172A-PHM mutant is highly similar in structure to the oxidized wild-type PHM (PDB ID: 1YIP). More surprising, although it has no copper in the Cu H site, the conformation of the two remaining histidines in this site, His 107 and His 108, is almost unmodified from that of wild-type PHM and the Cu M site retains the same coordination of the copper observed in wild-type PHM.
To determine whether copper at both sites is required to bind peptide substrate, crystals of apo-PHM were both soaked and cocrystallized in mother liquor containing N-Ac-di-I-YG. This peptide binds to wild-type holo-PHM in both the oxidized and the reduced forms 41 . In contrast, the structure of apo-PHM determined with data collected from the N-Ac-di-I-YG soaked or co-crystallized crystals showed no additional density corresponding to the bound peptide, indicating that copper at both sites is required not only for the later steps of the catalysis but also for binding substrate. Furthermore, two water molecules that are a part of a network formed by Q170, Q272, H108, and the peptide substrate are present in wild-type PHM + peptide structure (PDB ID: 1OPM) but absent in H108A-PHM and H172A-PHM (Fig. 6a, b). These water molecules are possibly required for the peptide to bind to the enzyme. In the H108A and H172A mutants, lack of copper in the Cu H site could prevent the two water molecules from forming the network required for peptide binding. In the wild-type enzyme, substrate binding has been shown to induce a new mode of CO binding at the M-center which lowers the C≡O infrared stretching frequency and suggests electronic activation of the diatomic ligand. This process does not occur in the mutants, consistent with a lack of substrate binding although clearly other factors could also be responsible 57 .
The recently published 2.9 Å resolution structure of apo human dopamine β-hydroxylase (hDBH, PDB ID: 4ZEL) 49 has two monomers in the asymmetric unit (molecules A and B) with different conformations. In chain A, the two subdomains display a closed conformation in which the C-subdomain is closer to Nsubdomain (hinge rotation of~18°) resulting in a structure very similar to that observed in our H108A-PHM structure in complex with citrate (Fig. 5d, e). Furthermore, the individual active site residues of chain A of hDBH align closely with those of H108A-PHM in complex with citrate (Fig. 5f). The single copper ions present in both structures have similar coordination and are in approximately the same positions in the aligned structures (distance ≤1.6 Å). The other chain of the hDBH crystals (molecule B), however, shows an open active site similar to that of wild-type PHM (Fig. 5a). Christensen and co-workers 49 modeled a copper in the Cu M site of chain A and used the positions of the Culiganding residues in chains A and B to infer the position of the three other coppers. Based on this, the authors suggested the possibility that Cu H and Cu M could come as close as 4-5 Å to each other during the catalytic cycle, close enough to form a binuclear copper center 49 . However, the fact that this conformation was only seen in structures with a single copper ion (H108A-PHM-cit and hDBH chain A) is more compatible with the conclusion that having two bound coppers has an important structural role: locking PHM (and probably DBH) in the catalytically competent conformation. This role is unusual for a catalytic redox active metal 58 . The closed structures observed in hDBH and H108A-PHM-cit may represent intermediates in the loading of copper ions during the final assembly. This is an attractive hypothesis since the molecule with only one copper could have access to a number of alternative conformations which may assist in loading the second copper, and/or may be required for recognition of its cognate transfer partner 32,33,59 . All PHM structures determined previously (reduced, oxidized, with and without ligands, and in a precatalytic complex) have had copper ions in both sites including the Cu M mutant M314I 14 . In all these structures, the distance between the two coppers is approximately~11 Å 12,14,15,41 . This observation, together with the fact that PHM crystals can carry out sustained catalysis, supports a mechanism that does not require a large conformational change to bring the copper ions into close enough proximity for a direct electron transfer between the ions.
The structures of the Cu H site mutants (H107A, H108A, and H172A), even those with unoccupied copper sites, are all found in the open conformation. The incorporation of citrate in the crystallization media as an additive had a positive effect on the resolution but, surprisingly, resulted in structures that show significant changes with respect to wild-type, providing a glimpse into the changed conformational landscape that results from the absence of one copper.

Methods
Protein expression, mutagenesis, and purification. Stable cell lines secreting PHM and its mutants, H107A, H108A, and H172A, were created using our established methods 12,13,33,37,60 . Briefly, CHO cell lines were transfected with a vector encoding the protein of interest and dhfr; the cells were selected by growth in αMEM plus 10% dialyzed fetal bovine serum, subcloned and tested for enzyme expression and clonality until stable, clonal lines were obtained. Typically, it took several months to obtain a cell line in which at least 1-2% of total protein synthesis was devoted to the protein of interest, and in which the cell doubling time remained less than a day 37,60 . Wild-type and mutant cell-lines were thawed from freezer stock into T75 flasks with 20 mL of DMEM/F12 medium containing 10% FCII serum (Fisher). At 80% confluence, the cells were passed into five NUNC triple flasks (500 cm 2 area per flask) which were then grown until the cells were 80% confluent. Cells were trypsinized and resuspended in 50 mL medium containing 10% FCII serum and then inoculated into the extra-capillary space (ECS) of a Hollow Fiber Bioreactor (Fibercell Systems 4300-C2008, MWCO 5 kD, 3000 cm 2 surface area) precultured with 2 L of 50 mM PBS pH 7.35 and 2 L of DMEM/F12 10% FCII serum [61][62][63] . Individual bioreactors containing each of the mutants were fed with DMEM/F12/10% FCII serum for 2-3 weeks, after which the serum level was lowered to 0.5% FCII serum. Thereafter, the bioreactors were fed with 0.5% serum-containing medium every other day and spent medium (20 mL) from the ECS was harvested and frozen at −20°C for subsequent purification 63 .
Preparation of samples for apo and mutant proteins. Purified enzymes (wildtype and mutants) were dialyzed against 20 mM sodium phosphate buffer, pH 8.0. The apo-protein isolated from recombinant CHO cells contained no copper, and was used without further addition of copper. Mutant proteins, H107A, H108A, and H172A-PHM, were also devoid of copper as isolated and were subsequently reconstituted with cupric sulfate by slow addition of 2.5 molar equivalents Cu(II) per protein followed by two cycles of dialysis to remove unbound cupric ions. Protein Concentrations were determined using OD280(1%) = 0.980 on a Cary 50 spectrophotometer. Copper concentrations were determined using a Perkin-Elmer Optima 2000 DV inductively coupled plasma optical emission spectrometer.