Coordination promiscuity guarantees metal substrate selection in transmembrane primary-active Zn²⁺ pumps

Abstract

Metal selectivity in P_1B-type ATPase pumps appears to be determined by amino acid motifs on their transmembrane helices. We reveal the principles governing substrate promiscuity towards first-, second- and third- row transition metals in a transmembrane Zn²⁺/Cd²⁺/Hg²⁺/Pb²⁺ P-type ATPase (ZntA), by dissecting its coordination chemistry. Atomic resolution characterization in detergent micelles and lipid bilayers reveals a “plastic” transmembrane metal-binding site that selects substrates by unique and diverse, yet defined, coordination geometries and ligand-metal distances.

Graphical Abstract:

P_1B-type ATPases are transmembrane primary-active pumps, conserved throughout all kingdoms of life, which utilize the energy generated by ATP hydrolysis to drive metal transport across biological membranes¹. These transporters constitute an essential system for the selective translocation of transition metal ions to control the cellular concentrations of essential-but-toxic (e.g.: Cu⁺, Zn²⁺) and toxic (e.g.: Ag⁺, Cd²⁺, Pb²⁺) metals².

The P_1B-type ATPase transport cycle follows the Post-Albers scheme. P-type pumps alternate between two states, E1 and E2, allowing ion accessibility on opposite sides of the lipid bilayer¹, coupled to conformational changes in cytoplasmic domains induced by ATP binding, hydrolysis, phosphorylation and dephosphorylation. Ion(s) bind to transmembrane site(s) (E1 state), are occluded within the membrane upon ATP hydrolysis and phosphorylation (E1P), and are then released from E2P to the opposite site, resulting in dephosphorylation to regenerate E1.

P_1B-type ATPases possess 8 transmembrane helices (MA, -MB, and -M1–6) containing signature sequences for ion recognition^{2a, 3}. Metal selectivity appears to be determined by conserved amino acid motifs in transmembrane helices TM4-5-6 providing ligands for coordination into transmembrane metal binding site(s) (TM-MBS). Based on these sequences, P_1B-type ATPases are classified into subtypes: P_1B-1 (Cu⁺/Ag⁺ exporters), P_1B-2 (Zn²⁺/Cd²⁺/Pb²⁺), P_1B-3 (Cu⁺/Cu²⁺), P_1B-4 (Zn²⁺/Co²⁺), P_1B-5 (probable Ni²⁺/Fe²⁺) and P_1B-6/−7-types (unknown selectivity)^{2a, 3a}.

The bioinorganic chemistry of pumps metal transport is unique. Selectivity and translocation are achieved through metal recognition with high thermodynamic stability and kinetic lability to guarantee selective binding and release through the transport cycle with a high turnover number. The P_1B-2-type subclass evolved to control the concentrations of essential-but-toxic Zn²⁺ and provide resistance towards toxic Cd²⁺, Hg²⁺ and Pb²⁺ exposure. Thus, they represent an ideal system to investigate the structural coordination chemistry underlying a defined but promiscuous selectivity. The structural framework underlying Zn-pumps function was established through biochemical studies on P_1B-2-type ZntA from E. coli and S. sonnei (99% identity) and from crystal structures in metal-free states^{3c, 45}. Four key conserved residues in transmembrane signature motifs on the TM 4, 5 and 6 are critical for Zn²⁺ pumping activity. Cys392, Cys394 and Asp714 guarantee high-affinity substrate binding while Lys693 acts as a built-in counterion preventing proton countertransport^3c. Atomic-level understanding of the coordination chemistry underlying metal binding and promiscuous selection at the transmembrane site remain poorly understood due to: i) difficulties in obtaining structural information on functional purified samples of metal-bound pumps; ii) the lack of comparative studies with transporters in their native lipid bilayer environment.

To reveal the coordination chemistry defining metal selection and examine the structural features underlying promiscuity we characterized biochemically and by X-ray absorption spectroscopy (XAS) the coordination properties of a Zn-pump from P. aeruginosa (PaZntA) bound to all its metal substrates.

PaZntA possess a ferredoxin-like N-terminal MBD carrying at least a metal–binding site centered on a MDCxxE motif corresponding to the CxxC motif characterized in other ZntAs, flanked by a His/Cys-rich sequence potentially involved in additional metal binding⁵. However, regulatory cytoplasmic metal binding domains present in P_1B-₂-type ATPases (MBDs) are not essential to confer metal selectivity and transport.^{3c, 6} Thus, the N-terminal MBD was truncated to exclusively preserve the transmembrane metal binding site(s). PaZntA shares high homology with Cu(I) ATPases and Zn(II) ATPases, possessing 40% identical and 30% similar amino acids outside the N-MBD with S. sonnei ZntA and ~70% identical positions in TM4-5-6, allowing the generation of reliable structural homology models (Fig. 1A–B).

Figure 1:

(A) PaZntA_51–740 3D homology model based on the A. Fulgidus CopA cryo-EM model (PDB 3J09, which include the N-terminal MBD) and (B) close-up view of the expected high-affinity TM-MBS. Cys391, Cys393 and Asp721 (conserved in ZntAs) are responsible for transmembrane substrate binding, while Lys700 on M6 acts as a built-in counterion. A similar model was also obtained based on the structure of S. Sonnei ZntA (PDB 3J09). (C) Relative ATPase activity of PaΔZntA_121–740 in proteoliposomes as a function of different metals (40 µM, see Material and Methods; Zn(II)-ATPase activity ∼ 0.9 nmol/(mg·min) ) and (D) the influence of P-type ATPases inhibitors (VO₄^3-, AlF₄^-, ouabain) and non-hydrolysable nucleotide analogues (AMPPCP).

We established protocols for wtPaZntA and PaΔZntA_121–740 expression, purification in detergent micelles and reconstitution in artificial lipid bilayers (Fig. S1). Detergent-solubilized wtPaZntA showed a Zn²⁺-dependent ATPase activity with K_M of 32 ± 11 μM, and V_max of 1.7 ± 0.4 nmol/(mg·min). For PaΔZntA_121–740 the V_max was reduced to approx. 30 % (0.6 ± 0.1 nmol/(mg·min)) with K_M of 13.8 ± 7 μM, in agreement with the ATPase regulatory role by N-MBD (Fig. S2). Whether the N-MBD acts as a metal sensor and/or directly regulate phosphorylation/dephosphorylation rates as a function of metal bound to its MDCxxE motif remains to be established. To verify the PaΔZntA_121–740 functional reconstitution in proteoliposomes, the ATP hydrolysis and metal selectivity were determined by metal-dependent ATPase activity. Full-length wtPaZntA and PaΔZntA_121–740 proteoliposomes showed a Michaelis-Menten Zn(II)-dependent ATPase activity with K_M of 26 ± 4 and 26 ± 7 μM, and V_max of 1.9 ± 0.2 and 1.4 ± 0.2 nmol/(mg·min), respectively (Fig. S3). Substrate selectivity indicates that activity in PaΔZntA_121–740 proteoliposomes depends on group 12 transition metals (Zn²⁺/Cd²⁺/Hg²⁺) and Pb²⁺ (Fig. 1C). The selectivity profile follows the same order as detergent-solubilized wtPaZntA, confirming that selectivity stems from metal binding to the high-affinity transmembrane site and is not imparted by N-MBD (Fig. S4). Inhibitors of P_1B-type ATPases (AlF₄⁻ and VO₄⁻) abolished the metal-dependent activation of ATPase activity, but not inhibitors of other P-type ATPases (ouabain, Na/K ATPase). Furthermore, the absence of activity with the non-hydrolysable ATP analogue AMPPCP confirmed functional reconstitution of PaZntA in lipid bilayers (Fig. 1D), despite PaZntA appears to possess a lower specific ATPase activity than other Zn²⁺ P-type ATPases^{4b, 7}.

The presence of a single transmembrane metal binding site was investigated by PaΔZntA_121–740 titration with all its metal substrates followed by metal quantification by Inductively-coupled plasma mass spectrometry (ICP-MS) revealing metal-to-protein ratios of: 0.99 ± 0.04 mol/mol for Zn²⁺_, 1.36 ± 0.23 mol/mol for Cd²⁺, 0.95 ± 0.24 mol/mol for Hg²⁺, and 0.82 ± 0.18 mol/mol for Pb²⁺. Cd²⁺ binding was confirmed by absorption spectroscopy (Fig. S5). The metal-induced absorption intensity at 252nm plotted as function of Cd-to-protein ratios showed a breakpoint at ~1 Cd²⁺ equiv./mol, confirming the presence of a single high-affinity transmembrane binding site⁸.

To obtain details of the structure and coordination of the metal centers, we generated metal-bound forms of PaΔZntA_121–740 (1 eq. M²⁺/PaΔZntA_121–740, M²⁺= Zn²⁺, Cd²⁺, Hg²⁺, Pb²⁺) in Cymal-7 micelles. Zn and Cd K-edge, and Hg and Pb L₃-edge extended X-ray absorption fine structure (EXAFS) spectra were collected. The data are presented in Fig. 2A–H, with the corresponding best fits and Fourier transforms. To prevent interference from potential binding to the His₆ tag, thrombin cleavage was optimized to obtain a tag-free protein (Fig S6).

Figure 2:

XAS analysis of PaΔZntA_121–740–M²⁺. K-edge experimental EXAFS data (black line) and best fits (red line) with the corresponding Fourier transforms for ΔZntA_121–740–Zn²⁺ (A,B), ΔZntA_121–740–Cd²⁺ (C,D), and L3-edge EXAFS data for ΔZntA_121–740–Pb²⁺ (E,F) and ΔZntA_121–740–Hg²⁺ (G,H; see Materials and Methods). The parameters for the best fits are listed in Table 1. (I) Models of the metal binding sites in PaΔZntA_121–740–Zn²⁺, PaΔZntA_121–740–Cd²⁺, PaΔZntA_121–740–Hg²⁺ and PaΔZntA_121–740–Pb²⁺ with the coordination geometries and metal-ligand bond distances.

In PaΔZntA-Zn²⁺, the best EXAFS fit was obtained with two N/O ligands at 1.97 Å, and two S ligands at 2.27 Å (Fig. 2A–B and Table 1). Fitting with 1S3N/O coordination and penta-coordinated geometry including an additional independent ligand (either S or N/O) resulted in worse fits. The bond distances are consistent with those of protein sites where Zn is bound in a tetrahedral or distorted tetrahedral coordination⁹ by mixed S and N/O ligands, and with that of E. coli ZntA¹⁰ (Fig. 2I). Based on functional data for ZntA homologues and the key residues conservation among Zn²⁺ P-type pumps, the proposed ligands are two thiolate sulfurs (Cys391 and Cys393) and two oxygen ligands from Asp721 in bidentate fashion, likely resulting in distortion from ideal tetrahedral geometry due to constraints imposed by the Asp O-C-O angle.^{3c, 4d, 4f}. We generated mutants in all the proposed coordinating residues (C391A, C393A, C391A, C393A, D721A and D721N) and determined the effect on the Zn²⁺-dependent ATPase activity (Fig.3). In agreement with the proposed model the ATPase activity was dramatically affected in all mutants. In addition, single mutation in either coordinating Cys (C391 or C393) abolished the activity to the same extent as the double Cys mutant (C391 or C393) confirming the nature of the 2 S scatters in EXAFS analysis and excluding possible Cl⁻ coordination (similar scattering properties in EXAFS). Moreover, mutation of D721 to either Asn or Ala resulted in similar abolishment of pump activity (Fig. 3), supporting bidentate coordination by the Asp side chain rather than the possible presence of a H₂O molecule in the coordination shell.

Table 1.

EXAFS best curve-fitting parameters for ΔZntA_121–740–M²⁺.

	M2+ eq.	Scattering paths [b]	N [a]	R (Å)	σ2 (Å2)	F
ΔZnta-Zn2+	1	Zn-S	2	2.268(5)	0.0069(4)	0.412
		Zn-N/O	2	1.969(4)	0.0030(3)
ΔZnta-Cd2+	1	Cd-S	2	2.529(8)	0.0025(4)	0.472
		Cd-N/O	2	2.29(1)	0.002(1)
ΔZnta-Pb2+	1	Pb-S	2	2.642(6)	0.0061(3)	0.584
		Pb-N/O	1	2.27(1)	0.006(1)
ΔZnta-Hg2+	1	Hg-S	2	2.312(7)	0.007(1)	0.392
		Hg-N/O	(2)	2.433(5)	0.0017(7)

^[a]Coordination numbers (N), interatomic distances (R), Debye-Waller factors σ² (mean-square deviations in interatomic distance). The fit-error function F is: $F=\sqrt{\sum k^{\text{6}}{\text{(}\chi {\text{(}k\text{)}}_{\text{calcd }}\text{– }\chi \text{ (}k{\text{)}}_{\text{exp}}\text{)}}^{\text{2}}\text{ /}\sum k^{\text{6}}{\left(\chi \text{ (}k{\text{)}}_{\text{exp}}\right)}^2}$, X(k): EXAFS oscillation; k: photo-electron wave number). In parentheses are the standard deviations for best-fit parameters (± values on last digit)

Figure 3:

ATPase activity of PaΔZntA_121–740 mutants (C391A, C393A, C391A/C393A, D721N and D721A) relative to wtPaΔZntA_121–740 in detergent micelles, in the presence of the corresponding metal substrates (40 µM).

In PaΔZntA-Cd²⁺, the best fit was obtained with two N/O ligands at 2.29 Å, and two S ligands at 2.52 Å (Fig. 2C–D and Table 1, extended k-range fitting up to 14 Å⁻¹ is reported in Fig. S7). These bond-lengths are consistent with other protein Cd²⁺ sites in tetrahedral/distorted tetrahedral coordination (Fig. 2I)⁹. Because of the larger Cd²⁺ ionic radius, the bond lengths significantly increased by 0.3 Å indicating plasticity in the selection site to accommodate first- and second- row transition metal substrates. The capability of uptake and binding of Zn²⁺ and Cd²⁺ at the plastic TM-MBS is compatible with the presence of a conserved electronegative funnel connecting the cytoplasmic membrane interface to the intramembranous high-affinity site^3c. The funnel and the absence of a selectivity filter allow specific uptake of cellular Zn²⁺ or Cd²⁺. In agreement with the proposed Zn²⁺ model, mutation of coordinating C391, C393 and D721 residues dramatically affected the Cd²⁺ dependent ATPase activation.

To date, the understanding of Pb²⁺ transporters chemistry underlying detoxification is limited. By generating PaΔZntA_121–740-Pb²⁺ samples, we have characterized for the first time the binding site and bonding geometry in a Pb²⁺ transporter. The EXAFS data revealed the existence of two ligand shells. The best EXAFS fit in PaΔZntA-Pb²⁺ was obtained with one N/O ligand at 2.27 Å, and two S ligands at 2.64 Å, (Fig. 2E–F and Table 1). The bond distances are consistent with trigonal pyramidal Pb²⁺ sites in which one tetrahedron position accommodates the Pb²⁺ lone pair (hybridization) (Fig. 2I). Based on the abolished Pb²⁺-dependent ATPase activity in all our PaΔZntA_121–740 mutants (Fig. 3) and corresponding mutagenesis studies in E. coli^{4d, 4f} and S. sonnei^3c ZntA, the Pb²⁺ site is formed by Cys391, Cys393 and Asp721^{3c, 4d, 4f}.

This novel PbS₂O/N site indicates that the binding environment in ZntAs differs from the frequently observed favorable [Pb(II)(SR)₃]⁻ in Pb-substituted zinc fingers¹¹ and de novo designed three helical bundles¹². Pb²⁺ can be coordinated by S/O/N/P-containing ligands with coordination numbers from 2 to 9. Nevertheless, the favorable geometry in a thiol-rich environment is trigonal pyramidal with a lone pair occupying the apical position (hemidirected)¹³. Analysis of PDB for Pb²⁺-bound proteins reveals that thiolate ligands are indeed typically present in trigonal pyramidal [Pb(II)(SR)₃]⁻ geometries⁹. Because of the soft nature of Pb²⁺ expected from Pearson’s theory and the high enthalpy of Pb-S bond formation¹⁴, Pb(II) is thermodynamically more tightly bound by thiolates than carboxylate ligands. Thus, the observed 2S1N/O binding indicates a reduced Pb²⁺ affinity compared to 3S sites. This and the intrinsic kinetic lability of CysS-Pb(II) bonds could favor dissociation from the site in the Post-Albers cycle. Indeed, Pb(II) activation results in the highest ATPase turnover rate among all substrates (Fig. 1C).

Finally, we analyzed the EXAFS data of PaΔZntA_121–740-Hg²⁺ (Fig. 2G–H and Table 1). Frequently, in metalloproteins Hg²⁺ is bound in digonal or trigonal 2S/3S coordination. We initially obtained a best fit with 2S ligands at 2.32 Å. However, the introduction of additional 1 or 2 N/O ligands resulted in comparable F values, preventing definitive assignment of the coordination environment from EXAFS. However, analysis of the X-ray absorption near edge structure (XANES) features in PaΔZntA_121–740-Hg²⁺ (Fig. S8), revealed the absence of a pronounced 2p_3/2 → 6s/5d transition at 12280 eV, the absence of a shoulder at 12295 eV and an absorption maximum below 12330 eV (in linear bis-L-cysteinate Hg(Cys)₂ is above 12330 eV)¹⁵, consistent with a non-linear Hg²⁺ complex. The conservation and proximity of the Asp721 and the high frequency of Hg²⁺ structures in the PDB adopting irregular 3-/4- coordinated geometries, suggest linear coordination distortion and Hg²⁺ binding in irregular trigonal/tetrahedral geometries with possible weak N/O interactions (Fig. 2I). To address this, the effect of D721A or D721N mutations on the Hg²⁺-dependent ATPase activation profiles was investigated. D721A and D721N PaΔZntA_121–740 mutants failed to be activated by Hg²⁺ (Fig. 3), strongly supporting the involvement of Asp721 side chain in Hg²⁺ coordination. Indeed, linear 2S Hg²⁺ coordination resulting in high affinity and reduced kinetic lability could prevent metal release or dramatically reduce turnover numbers.

To address whether the transmembrane site plasticity is preserved in the lipid bilayer, where structural constrains are imposed by the anisotropic nature of the protein-membrane interaction, we characterized the metal-bound forms of the pump reconstituted in lipid bilayers. PaΔZntA_121–740ZntA was successfully reconstituted in proteoliposomes and its metal-bound forms generated for comparative XANES in PaΔZntA_121–740-Zn²⁺/Cd²⁺/Hg²⁺/Pb²⁺ (Fig. S8) and for comparative EXAFS in the Zn²⁺/Cd²⁺/Hg²⁺ bound forms (Fig. S9). The close correspondence of all XANES features for PaΔZntA_121–740-M²⁺ in detergent-micelles with those in lipid bilayers (with minor differences only in PaΔZntA_121–740-Zn²⁺) and the similar ligand-to-metal distances from EXAFS fits, revealed sites with almost identical coordination (Table S1).

The results reveal “coordination plasticity” in metal recognition to guarantee substrate promiscuity in metal pumps. In contrast to ion channels, where size-selectivity filters select substrates in gated or non-gated conducting pores¹⁶, selectivity in zinc pumps is governed primarily by coordination chemistry and geometry resulting in metal binding with different coordination numbers and bond distances, while preserving the apparent coordination properties. Moreover, coordination divergence form ideal ligands sets and geometries typically present in high affinity binding sites could result in inefficient metal release required by the catalytic cycle. Thus, divergence from these ideal templates appears required for substrate promiscuity and efficient metal translocation.