2.1. Cloning of PA4063

The sequence encoding the mature PA4063 protein (without the signal sequence for periplasmic export) was amplified using the chromosomal DNA from P. aeruginosa PAO1 as a template and the oligonucleotides PA4063_5 (CACCATGGCCCATGACGACCACGACCATGA) and PA4063_6 (GTGAATTCTACAGCTTCAGCTCGGCG), bearing NcoI and EcoRI cloning sites at the 5′-and 3′-termini, respectively. PCR was carried out using an Expand High Fidelity PCR System (Sigma–Aldrich). The amplified DNA fragment was then digested with NcoI and EcoRI, run on a preparative agarose gel (TAE) and purified using a Zymoclean Gel DNA Recovery Kit (Zymo Research). This fragment was subsequently inserted into the NcoI and EcoRI restriction sites of the expression vector pET-26b(+), generating plasmid pET-26b(+)PA4063. In this vector, the sequence encoding mature PA4063 is fused to the Erwinia carotovora pelB signal sequence, thus allowing periplasmic localization of the expressed protein. The sequence of the insert was subsequently confirmed by nucleotide sequencing. The vector was inserted into E. coli BL21 (DE3) cells for protein expression.

2.2. PA4063 expression and purification

E. coli BL21 (DE3) cells bearing pET-26b(+)PA4063 were inoculated into 4 l lysogeny broth at 37°C and expression of PA4063 was induced in mid-log phase cultures by the addition of 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). After 24 h of growth, the cells were harvested by centrifugation and resuspended in 20% sucrose, 30 mM Tris–HCl pH 8.0, 1 mM EDTA. Afterwards, lysozyme was added to a concentration of 1 mg ml⁻¹. After 10 min incubation on ice, the bacteria were centrifugated at 12 000g and the supernatant, containing the periplasmic proteins, was collected and extensively dialyzed against 50 mM NaH₂PO₄, 250 mM NaCl pH 7.8. Subsequently, the protein extract was loaded onto an Ni–NTA column (20 ml bed volume) pre-equilibrated with the same buffer. Proteins were fractionated according to their affinity for immobilized metal ions using a 0–250 mM linear gradient of imidazole. Most of the proteins eluted at imidazole concentrations below 60 mM, whereas PA4063 eluted as a pure protein at imidazole concentrations above 100 mM. Fractions containing PA4063 were collected, concentrated to a small volume and then extensively dialyzed against 20 mM HEPES, 100 mM NaCl pH 7.0. Protein concentration was determined by the method of Lowry.

Apo PA4063 was obtained by dialyzing the protein for 24 h against 50 mM sodium acetate, 2 mM EDTA pH 5.5. EDTA was then removed by dialysis against a buffer consisting of 50 mM sodium acetate, 100 mM NaCl pH 5 and the protein was finally dialyzed in 20 mM HEPES, 100 mM NaCl pH 7.0. To minimize metal contamination, all of these dialyses were carried out in buffers prepared with Chelex-treated water and in glassware pre-treated with nitric acid.

2.3. UV–Vis spectrophotometric analyses

Since zinc is spectrophotometrically silent, we analysed the ability of the protein to bind cobalt, a metal that has the ability to easily substitute for zinc in metal-binding proteins in vitro or in vivo (Ammendola et al., 2020). Experiments were carried out with apo PA4063 at a concentration of 670 µM. Spectra were recorded in the UV–Vis region following the addition of progressively increasing amounts of metal corresponding to 0.5 molar equivalents of the protein. Each spectrum was recorded immediately after the metal was added, subsequently evaluating any variations over time due to possible slow conformational changes induced by the metal. To gain insight into the ability of PA4063 to bind zinc, we analyzed the effect of zinc addition on the spectrum of the cobalt-bound protein. For this purpose, we carried out a gradual addition of zinc sulfate in aliquots of 0.5 zinc equivalents, as previously performed with cobalt.

2.4. Isothermal titration calorimetry (ITC)

ITC experiments were carried out using an iTC200 microcalorimeter (MicroCal, Malvern). 1.2 µl aliquots of metal solution (0.32 mM zinc acetate or cobalt nitrate) were injected into 11 µM apo PA4063 solution at 25°C. The experiments were carried out in the dialysis buffer used to prepare the apoprotein (20 mM HEPES pH 7.0, 100 mM NaCl) diluted with the percentage of water used to prepare the metal solutions. 0.32 mM metal solutions were prepared from water-dissolved 2 mM stock solutions. Data were fitted using the one-binding-site model of the MicroCal PEAQ-ITC Analysis Software (Malvern) version of Origin. The heat of binding (ΔH), the stoichiometry (n) and the dissociation constant (K_d) were then calculated from plots of the heat evolved per mole of ligand injected versus the molar ratio of ligand to protein using the software provided by the vendor. The data are the mean ± standard deviation of three or two experiments for zinc or cobalt, respectively. Optimization of the proper protein:ligand concentration ratio used to obtain the thermodynamic signature reported in this manuscript was carried out by assaying different protein concentrations (5.9, 7, 9.6 and 11 µM with zinc ligand; data not shown).

2.5. Crystallization

Automated crystallization screening (Crystal Phoenix, Art Robbins Instruments) was performed at 294 K in a sitting-drop vapour-diffusion setup using a 1:1 ratio of protein and reservoir solutions (0.6 µl drop volume, 70 µl reservoir). The protein solution consisted of 10 g l⁻¹ PA4063, 20 mM HEPES pH 7.5, 100 mM NaCl. Screening was performed with and without 5 mM zinc acetate added to the protein solution (zinc:protein ratio of about 10:1). Two crystallization conditions (A and B) were identified in the absence of zinc. Condition A, consisting of 2 M ammonium sulfate and 2 M NaCl, produced a crystal with tetragonal symmetry (space group P4₁2₁2) that diffracted to 1.95 Å resolution but could not be easily reproduced. Condition B, consisting of 2.7–3 M ammonium sulfate and 1% 2-methyl-2,4-pentanediol (MPD), produced poorly diffracting rod-shaped crystals with hexagonal symmetry (space group P6₅) that were further optimized by hand and by microseeding, growing up to 0.5 mm in length, which reached a maximum resolution of 2.7 Å.

Since crystallization did not occur in the presence of zinc, crystal soaking was performed to identify zinc-binding sites. Crystals from both conditions A and B were soaked in 5 mM zinc acetate (zinc:protein ratio of around 10:1) for 1 h.

2.6. Data collection and structure solution

PA4063 has no sequence similarity to any protein present in the PDB; therefore, phasing was accomplished by single-wavelength anomalous dispersion (SAD) experiments. The presence of a single methionine in the sequence led us to prefer heavy-atom derivatization over selenomethionine substitution. A native diffraction data set was collected from a tetragonal crystal to a resolution of 1.95 Å using a wavelength of 1 Å. However, the better-diffracting tetragonal crystals were not easily reproducible; thus, heavy-atom derivatization was only attempted on the hexagonal crystals. Crystals were soaked in 5 mM KAu(CN)₂ for 1 h followed by back-soaking in mother liquor (three quick steps in 2 µl drops, the last with an additional 20% glycerol). The best SAD data set was collected at the gold absorption peak (λ = 1.000 Å) to a resolution of 2.7 Å.

A zinc-soaked hexagonal crystal was used for X-ray diffraction data collection at the wavelength corresponding to the zinc absorption peak (λ = 1.271 Å) to 3.3 Å resolution.

Three diffraction data sets were collected on the XRD2 beamline at the Elettra synchrotron-radiation source, Trieste, Italy using a Dectris PILATUS 6M detector at 100 K (Lausi et al., 2015).

An additional data set was collected to 2.82 Å resolution from a zinc-soaked tetragonal crystal at a wavelength of 0.967 Å. This last data set was collected using a Dectris EIGER2 X 4M detector on beamline ID30B at the ESRF synchrotron-radiation source, Grenoble, France.

The data sets were processed and scaled with XDS (Kabsch, 2010) and AIMLESS (Evans & Murshudov, 2013).

SAD phasing was accomplished by exploiting the gold anomalous signal with SHELXC/D/E (Sheldrick, 2010). A significant anomalous signal was detected to 3.4 Å resolution (i.e. where CC_anom falls below a 0.15 threshold). SHELXD located two gold sites with occupancies of 1 and 0.33, and subsequent phasing and density modification with SHELXE allowed 147 residues to be autotraced in space group P6₅ (the final model has an estimated mean FOM of 0.521 and a pseudo-free CC of 58.80%). Model building of one monomer in the asymmetric unit was performed either automatically using Buccaneer (Cowtan, 2006) or manually with Coot (Emsley et al., 2010). The traced monomer was used to perform molecular replacement with MOLREP (Vagin & Teplyakov, 2010) to find the locations of the other two monomers in the trimeric asymmetric unit. The structure was refined with REFMAC5 (Murshudov et al., 2011). Only the gold ion corresponding to the stronger peak was included in the final model (the final occupancy refined to 0.72) since the putative second site, located in a completely hydrophobic small pocket, appeared to be unconvincing. The PA4063 structure at 2.7 Å resolution obtained from the hexagonal crystal was then used to solve the higher resolution structure of PA4063 from the tetragonal crystal (resolution of 1.95 Å) by molecular replacement.

Structural coordinates have been deposited in the Protein Data Bank with accession codes 7ahw (apo tetragonal crystal), 7aly (gold derivative, hexagonal crystal), 7bgo (zinc-bound, hexagonal crystal) and 7amx (zinc-bound, tetragonal crystal). Data-collection and refinement statistics are summarized in Table 1.

3.1. Metal-binding properties of PA4063

Preliminary UV–Vis spectrophotometric analysis revealed that PA4063 can bind either zinc or cobalt, showing a higher affinity for zinc (Supplementary Fig. S1). To obtain quantitative information on the number of metal-binding sites and on the stability of the protein–metal interaction, ITC experiments were run on the apoprotein. The titration profiles with zinc or cobalt ions are shown in Fig. 1. The dissociation constant (K_d) and all of the additional thermodynamic parameters obtained by fitting the enthalpy curves are reported in Table 2 and Supplementary Fig. S2, respectively.

Figure 1 Binding of selected metals to apo PA4043. Analysis by ITC measurements was performed by titrating 11 µM PA4063 with 320 µM zinc acetate (a) or cobalt nitrate (b) solution in 84% dialysis buffer (150 mM NaCl, 20 mM Tris pH 7.2). The upper panels show the raw ITC data, while the lower panels show the integrated energy values normalized for injected protein. Binding isotherms were fitted using a single-binding-site model.

As expected for specific binding, integration of the titration peaks of apo PA4063 produced a sigmoidal enthalpy curve for each metal (Fig. 1, black markers) and a binding stoichiometry of ∼2. Under these experimental conditions, data fitted with the single-site-binding equation indicate that the two binding sites on the protein are equivalent, displaying the same affinity and the same mechanism of binding for each metal. In fact, the thermodynamic signature (i.e. profiling of the thermodynamic parameters; Supplementary Fig. S2) includes favourable enthalpic and entropic factors in both cases, thus suggesting that both binding events are mainly guided by novel inter­actions and dehydration processes that take place upon metal coordination, rather than by conformational change(s) of the protein. On the other hand, comparison of the K_d for the two metals indicates that the affinity of apo PA4063 for zinc is twofold higher than that for cobalt.

3.2. Apo PA4063 structure

PA4063 crystallized in two different forms corresponding to space groups P4₁2₁2 and P6₅. The asymmetric unit of the hexagonal crystal comprises three protein molecules, but the interactions are so limited that they can be considered as simple packing contacts between monomers as in the case of the tetragonal crystal form. This observation is in line with size-exclusion chromatography, which shows that the protein is monomeric in solution regardless of the presence of zinc (Supplementary Fig. S3).

Despite the differences in crystal packing with regard to both lattice and asymmetric unit composition, the conformation of the protein in the two crystal forms is almost identical (C^α r.m.s.d. in the range 0.49–0.74 Å). Thus, since the tetragonal crystal diffracted to higher resolution (1.95 Å versus 2.7 Å), the structural analysis presented here focuses on the latter model.

The structure is monomeric, with an α/β-sandwich architecture resembling the ferredoxin fold, although relevant differences can be detected. Indeed, PA4063 comprises a six-stranded antiparallel β-sheet made up of three β-hairpins (β3β4–β2β1–β6β7) interrupted in sequence by two main α-helices, four short helical turns and a small two-stranded β-sheet (β5–β8), collectively constituting the second layer of the `sandwich' (Fig. 2). The secondary structure of PA4063 can then be represented as ββ(α)α(αα)ββ(αβ)αββ(βα), where minor elements are shown in parentheses. A disulfide bridge connects Cys81 on β3 and Cys131 on β4.

Figure 2 Structure of apo PA4063. Two views of the structure showing the layers of the α/β sandwich. (a) The α-helical side; secondary-structure elements are indicated. (b) The β-sheet side: the unstructured His-rich stretches, underlined in the sequence, are represented as orange dashes and histidines located in the folded region are represented as orange sticks. (c) Residues involved in zinc coordination are marked with asterisks, histidines are in bold and unstructured regions are underlined.

Comparison with the canonical ferredoxin-like fold, with a signature βαββαβ secondary structure, reveals that PA4063 differs in both composition and arrangement of β elements. In fact, even when neglecting minor elements, PA4063 has two more β-strands at the ends of the sequence that are inserted in the sheet to form β-hairpins, implying that the strands are coupled both in sequence and structure, while in ferredoxin the sheet is organized such that only two of the strands that are consecutive in sequence are contiguous in the 3D structure (see Supplementary Fig. S4 for a topology diagram).

His-rich sequences of 19 and 24 amino acids, located at the N-terminus and between β3 and β4, respectively, are disordered and are not visible in the crystal structure, as usually observed in other proteins containing this motif (Blindauer, 2015; Ilari et al., 2011). It is worth noting that both His-rich fragments face the same side of the structure, corresponding to the region where the two zinc ions bind (see below). Only three histidines belong to the structured portion, two of which (His124 and His126) are located on β4 and the third of which (His48) is on the loop between β2 and the following helix (Fig. 2).

3.3. Zinc-bound PA4063 structure

The structure of PA4063 in complex with zinc was solved by soaking crystals in zinc acetate solution since crystallization did not occur in presence of zinc. Soaking experiments were first performed with hexagonal crystals, despite their lower X-ray diffraction power, due to the poor reproducibility of the tetragonal crystals, leading to data collection to 3.3 Å resolution.

Diffraction data were collected close to the absorption edge of zinc (λ = 1.27 Å) in order to unambiguously locate the metal. Two strong anomalous peaks were observed for each monomer (>9σ), revealing two binding sites, according to the ITC experiments, indicated as site a and site b (Fig. 3 and Supplementary Fig. S5).

Figure 3 Zinc-bound PA4063 structure. (a) Superimposition of apo (yellow) and zinc-bound (blue) PA4063 from the hexagonal crystal form. The anomalous difference density map (orange mesh), contoured at 5σ, unequivocally locates two zinc-binding sites. (b) Superimposition of apo (green) and zinc-bound (magenta) PA4063 from the better-diffracting tetragonal crystal form. The difference map (mFo − DFc, green mesh), set to 3σ, confirms the position of the zinc ions. (c) and (d) show the details of zinc coordination in sites a and b, respectively.

Later, a higher resolution (2.85 Å) data set was collected from a zinc-soaked tetragonal crystal, confirming the positions of the two zinc-binding sites and allowing a better determination of the complex geometry and coordination distances.

Site a is located on the flat surface of the β-sheet in correspondence with the histidine residues. The same mode of binding is detected in all three monomers of the hexagonal crystal: besides His124 and His126, the zinc ion binds to Glu87, likely as a monodentate ligand, and a water molecule, adopting a tetrahedral coordination geometry (Fig. 3c). The water molecule is not visible in the tetragonal crystal, where zinc appears to be coordinated by only three ligands.

All distances between zinc and ligands are in the range 2–2.5 Å. Comparison with the apo structure shows that coordination requires rearrangement of the donor side chains, which move closer to the site, while Glu35 is displaced. The second shell of coordination comprises many acidic amino acids (Glu33, Glu35, Asp37 and Asp122) that facilitate zinc binding through a double mechanism. Indeed, the global negative charge could have the effect of attracting cations to the site and increasing the basicity of the histidines, thus favouring metal coordination.

Site b is about 20 Å from site a, on the edge of the β-sheet where the His-rich sequences are inserted. Notably, a few residues belonging to these sequences, which are disordered in apo PA4063, become structured and take part in zinc coordination, although some variability is present. As for site a, the coordination is tetrahedral and three donor groups are provided by one glutamate, Glu47, and two histidines, His16 and His18 (Fig. 3d). Indeed, 5–6 additional residues (14/15–19) are visible at the N-terminus with respect to the apo structure. The fourth donor was modelled as a water molecule in monomer A of the hexagonal crystal, while in monomers B and C, as well as in the tetragonal crystal, the electron-density map indicates His120 to be the donor. Even in this case, the residue belongs to the His-rich stretch and only becomes structured upon zinc binding.

When evaluating the contribution of the His-rich sequences to the binding and the structural variation thus induced, it must be considered that the zinc-bound structure was obtained by soaking, thus the rearrangement could be influenced and limited by crystal packing. On the other hand, the fact that crystallization does not take place in the presence of zinc supports the hypothesis that the His-rich stretches participate in the binding but in a flexible and inhomogeneous way, thus interfering with the crystallization process.

3.4. Surface analysis

To delve into the structural characterization, we analysed the shape and charge distribution of the surface of PA4063 (Fig. 4). The protein has no relevant cavities and this feature, together with the exposure of both zinc-binding sites to solvent and the relatively low metal affinity, indicates that a catalytic function is improbable. The electrostatic potential surface reveals an extended negative patch on the flat surface of the β-sheet, which is likely to contribute to metal binding to site b, while the α-helical face has a rather positive character.

Figure 4 Surface analysis of PA4063. The picture shows two different views of the electrostatic surface potential of zinc-bound PA4063 generated with PyMOL. The protein orientation is the same as shown in Fig. 2: (a) α-­helical side, (b) β-sheet side. Disordered His-rich stretches are represented as orange dashed lines. In (b) the zinc ion in site a is represented as a grey sphere, while that in site b is buried and not visible.

A similar charge distribution is observed in Atx1, a ferredoxin-like copper chaperone present in eukaryotes that is responsible for metal transfer to the ATPase Ccc2. In the case of Atx1, the positive surface was indicated as critical for protein–protein interaction (Banci et al., 2006), while no role was proposed for the negative surface, which does not correspond to a metal-binding site in this protein.