Templated Construction of a Zn-Selective Protein Dimerization Motif

Abstract

Here we report that the approach of metal-templated ligand synthesis can be applied to construct a dimeric protein assembly (^BMOERIDC1₂), which is stabilized by non-covalent interactions and flexible covalent crosslinkers around the Zn templates. Despite its flexibility, ^BMOERIDC1₂ selectively binds Zn^II over other divalent metals and undergoes dimerization upon metal binding. Such simultaneous fulfillment of plasticity and selectivity is a hallmark of cellular signaling events that involve ligand/metalinduced protein dimerization.

Graphical abstract

Described here is the metal-templated construction of a dimeric protein assembly (^BMOERIDC1₂), which is stabilized by non-covalent interactions and flexible covalent crosslinkers around the Zn templates. ^BMOERIDC1₂ selectively binds Zn^II over other divalent metals and undergoes dimerization upon metal binding.

INTRODUCTION

We describe the metal-templated construction of a protein complex that undergoes dimerization in selective response to Zn binding. In previous examples of designed molecular recognition, chemists have devised small organic building blocks that self-assemble into discrete supramolecular structures upon selective binding of organic or inorganic substrates.¹ Such systems include Rebek² and Gibb’s³ cavitands that dimerize into nanocapsules via hydrocarbon sequestration, and Lehn⁴ and Raymond’s⁵ helicates that selectively self-organize into higher-order assemblies upon metal complexation. Our work represents an extension of such synthetic work into the realm of protein building blocks, while simultaneously demonstrating that a traditional approach in coordination chemistry (metal-templated ligand synthesis) can readily lead to the emergence of a selective, chemically-induced protein dimerization event, which is a cornerstone of cellular signaling pathways.⁶

Proteins are nature’s premier ligand scaffolds for constructing functional metal centers, as they can afford an exquisite tuning of metal selectivity and reactivity through a combination of covalent and noncovalent interactions. An absolute requirement for the generation of any metallic (or non-metallic) active site with a specialized function is that it is internalized within a network of such interactions.⁷ We have proposed that a straightforward route for engineering functional metal sites may proceed through a metalmediated nucleation of multiprotein complexes and a subsequent stabilization of the resulting protein interfaces around the now-internalized metal centers.^8,9 This approach was termed Metal-Templated Interface Redesign (MeTIR). The possibility of an evolutionary pathway similar to MeTIR has previously been suggested by Bolin¹⁰ and Armstrong¹¹ for the vicinal oxygen chelate (VOC) superfamily of metalloenzymes and by Hol¹² for hemocyanin and other dioxygen binding proteins. In further support, the inspection of many sophisticated metal active sites or binding pockets for non-metallic substrates reveal that they are formed in the interfaces between protein domains or subunits,^13,14 which likely were independently folded units.¹⁵

A synthetic analogy for MeTIR is provided by the time-proven method of metal-templated ligand synthesis, first reported by Busch¹⁶ and adopted by others,^17-19 in which a macrocyclic chelate is first organized and then covalently assembled around a metal ion. Inspired by this approach, we recently reported that a Zn^II-mediated, D₂-symmetrical, tetrameric superstructure of a monomeric protein, cytochrome cb562 (Zn4:MBPC14, Fig. 1a and b), can be converted into a stable, unimolecular assembly (C96RIDC14, Fig. 1c) through the installment of stable covalent and non-covalent bonds into the interfaces surrounding the four Zn centers.20 C96RIDC14 is not only one of the most stable multiprotein complexes ever engineered, but also displays impressive Zn^II-selectivity over other divalent ions including CuII and couples Zn^II binding to a large conformational change akin to a metal-dependent signaling protein.20 The seemingly contradictory traits of stability–both in terms of oligomerization and metal binding–and flexibility are afforded by the combination of fluid hydrophobic interactions incorporated into one set of interfaces (interface 1, i1) and short disulfide linkages (Cys96-Cys96’) built into a second set (i2) (Fig. 1c). In this study, we envisioned that the lengthening of the interfacial covalent linkages may further increase the flexibility of the assembly and result in the formation of alternative oligomeric forms while still retaining metal selectivity, which is the most basic but essential metal-based function. We report that such long covalent linkages across i3 result in the formation of a stable, folded structure consisting of two protomers (^BMOERIDC1₂) that dimerizes into a tetrameric form (Zn4:BMOERIDC14) upon selective Zn binding (Fig. 1d).

Figure 1.

Templated construction of Zn-selective protein complexes. The metal-chelating variant of cyt cb₅₆₂, MPBC-1 (a) undergoes Zn-mediated tetramerization into the D₂-symmetrical Zn₄:MBPC1₄ (b). (c) The combination of fluid, hydrophobic interactions installed into the i1 interface with a disulfide bond across i2 produces a tetramer ^96CMBPC1₄ that undergoes a Zn-selective conformational change,²⁰ whereas the combination of i1 interactions with a flexible linker across i3 (d) yields a Zn-selective dimerization motif (this study).

RESULTS AND DISCUSSION

The three C₂-symmetric interfaces of Zn₄:MBPC1₄ are highlighted in Fig. 1b and shown in detail in Fig. S1. Of these, i1 is the most extensive and closely packed, and readily lent itself to stabilization through the insertion of computationally prescribed and mostly hydrophilic-to-hydrophobic mutations (R34A/L38A/Q41W/K42S/D66W/V69I, blue patches in Fig. 1) to generate RIDC1.⁸ As the dihedral symmetry of the Zn-driven tetramer dictates, the combination of the stabilizing, non-covalent interactions in i1 with a short disulfide bond across the i2 interface (via the T96C mutation) should result in the formation of a stable tetramer (^C96RIDC1₄), as was borne out through hydrodynamic and crystallographic investigations.²⁰ i3, in contrast, is a minimal interface (490 Ų buried surface) formed entirely by the inner Zn coordination sphere (Fig. S1) and is not amenable to redesign or crosslinking through a disulfide bond. We surmised that a long, flexible chemical linker incorporated across i3 into RIDC1 could result in the formation of either a folded-over dimer (stabilized by the hydrophobic mutations in i1) (Fig. 1d) or a D₂symmetrical tetramer in the absence of Zn, in effect mimicking what nature accomplishes through genetically encoded peptide linkers for fusing proteins in order to realize the “chelate effect”.^11,21

Toward this end, we replaced the symmetrically related pairs of Gly82 residues (Fig. S1) near the vertex of i3 with cysteines for chemical coupling through bis-maleimide functionalized crosslinkers.The crosslinkers used, bis-maleimidoethane (BMOE), bis-maleimidobutane (BMB) and bismaleimidohexane (BMH) (Fig. 2), have spacer arm lengths of 8.0 Å, 10.9 Å and 13.0 Å, respectively, and are all expected to readily cover the anticipated C82S-C82’S distance of 10–11 Å. The coupling reactions proceed cleanly, whereby the crosslinking agents are added slowly to the protein solution to promote the formation of cross-coupled dimers over singly-modified (saturated) monomers. The dimers, ^BMOERIDC1₂, BMBRIDC12 and BMHRIDC12, were separated from monomers through size-exclusion chromatography, and their compositions were confirmed by mass spectroscopy (12,349 Da for the parent 82CRIDC1; 24,918 Da, 24,946 Da, and 24,974 Da for the BMOE, BMB, and BMH crosslinked dimers, respectively).

Figure 2.

Cys-specific, bis-maleimide functionalized crosslinkers.

We first probed whether the crosslinked dimers assemble into the same Zn-mediated D₂-symmetric tetrameric architecture that formed the basis of templated redesign. Sedimentation velocity (SV) experiments show that even at a concentration of 2.5 μM (lowest dimeric concentration measurable) ^BMOERIDC1₂, ^BMBRIDC1₂ and ^BMHRIDC1₂ all are entirely tetrameric in the presence of Zn^II, with a peak sedimentation coefficient of approx. 4.5 S (Fig. 3), which matches that of the parent complex, Zn₄:RIDC1₄.⁸ In comparison, the non-crosslinked RIDC-1 shows a distribution of 60% dimer and 40% tetramer under comparable solution conditions (5 μM RIDC1 and Zn^II). Although SV is a non-equilbrium technique, this comparison nevertheless indicates the stabilization of Zn-induced tetramers due to crosslinking, likely due to entropic reasons. The crystal structures of the Zn-adducts of ^BMOERIDC1₂ (PDB ID: 3QVY), ^BMBRIDC1₂ (PDB ID: 3QW0) and ^BMHRIDC1₂ (PDB ID: 3QW1) were determined at 2.3 Å, 1.8 Å and 2.6 Å resolution, respectively. As shown in a structural superposition in Fig. 4a, the supramolecular architectures of the resulting complexes are identical to one another, and to Zn₄:MBPC1₄²² and Zn₄:RIDC1₄.⁸ Importantly, the Zn-coordination environments, which lie in close proximity to the crosslinking sites, are unchanged as intended by the use of flexible linkers (Fig. 4b). The electron density for the entire length of the BMOE linker is evident, while for the longer BMB and BMH linkers only the Cys82-thioether-succinamide moiety can be discerned (Fig. 4c).

Figure 3.

Hydrodynamic analysis of maleimide crosslinked constructs in the presence of Zn^II. Sedimentation velocity (SV) profiles for various crosslinked (2.5 μM dimer) and a non-crosslinked construct (5 μM monomer) in the presence of 5 μM Zn^II demonstrate almost full conversion to a Zn^II induced tetramer of the three crosslinked constructs at relatively low protein and metal concentrations.

Figure 4.

Comparison of the supramolecular structures of the Zn^II mediated tetramers of maleimide crosslinked constructs. a) Structural superposition of Zn₄:RIDC1₄ (orange), Zn₄:^BMOERIDC1₄ (magenta), Zn₄:^BMBRIDC1₄ (cyan) and Zn₄:^BMHRIDC1₄ (marine). b) Superposition of the Zn coordination environments (same coloring scheme). c) F_o-F_c omit electron density maps (2.5 s) of the BMOE, BMB and BMH crosslinkers. For BMH, the hexane linker region was not modeled due to the lack of electron density.

Having verified their Zn-driven assembly into the desired “macrocyclic” architecture, we next investigated the structures of ^BMOERIDC1₂, ^BMBRIDC1₂ and ^BMHRIDC1₂ in the absence of Zn. SV measurements indicate that dimers are the predominant oligomeric form under the conditions tested for all crosslinkers. At low proteins concentrations (2.5 μM), a small population of possibly tetrameric species (4.5–4.8 S) is observed (Fig. 5a), while at higher concentrations (>25 μM), the SV distributions for such high-order aggregates become considerably broader and tail into larger sedimentation coefficients that are suggestive of linear polymeric chains (Fig, 5b). These results suggest that the crosslinked dimers prefer to adopt the folded-over conformation rather than a pre-organized tetramer reminiscent of the Zn-induced structure.

Figure 5.

Hydrodynamic analysis of metal free maleimide crosslinked constructs. a) Sedimentation velocity (SV) profiles for various crosslinked (2.5 μM dimer) and a non-crosslinked construct (5 μM monomer) in the absence of metals. b) Same as in (a), but at 25 μM and 50 μM concentrations, respectively, highlighting the presence of higher ordered species formed by the crosslinked constructs even in the absence of any metal ions. c) Hypothetical conformations for ^BMOERIDC1₂ and their calculated sedimentation coefficients. The rightmost structure is that of crystallographically characterized RIDC1₂ (non-crosslinked) with its calculated and experimentally determined sedimentation coefficient.

Despite extensive efforts, we could not crystallize the crosslinked dimers in the absence of metal ions, which likely is a result of their flexibility as well as their tendency to form large aggregates at high concentrations. In order to gain some insight into the structures of the crosslinked dimers, we compared their experimentally determined sedimentation coefficients–which are not only dependent on molecular weight but also shape–to those of various theoretical dimeric conformations of ^BMOERIDC1₂ (Fig. 5c) using Hydropro.²³ These simulations show that as the BMOE-dimer transitions from a fully extended configuration to a compact form, its sedimentation coefficient increases from 2.2 to 2.6 S, which is the lower end of the observed range (2.6–2.8 S) for the crosslinked dimers. Validating this analysis, the maximal sedimentation coefficient of the structurally characterized, antiparallel dimer RIDC1₂ was determined to be ~2.75 S,⁸ which is the most compact structure shown in the series (Fig. 5c). Based on these findings, we conclude that the chemically-crosslinked dimers likely assume a folded structure in solution, driven by the interactions between hydrophobic residues installed for i1 redesign. Given that the crosslinked dimers in their folded form should have many metal-coordinating residues (His, Glu and Asp) in close proximity and be relatively flexible, one would expect them to promiscuously coordinate different transition metal ions. Indeed, Co^II, Ni^II or Cu^II binding to ^BMOERIDC1₂ all induce the formation of a tetrameric species, with a maximal sedimentation coefficient of 4.5 identical to that of Zn₄:^BMOERIDC1₄ (Fig. 6a). However, since the redesign of i1 and the placement of crosslinkers across i3 were based on the four tetrahedral Zn templates, we anticipated that there should be increased affinity for Zn^II binding relative to the other divalent metal ions. To test this possibility, metal binding competition assays were performed with the metal chelating fluorophore, MagFura-2,²⁴ whereby increasing concentrations of Co^II, Ni^II, Cu^II or Zn^II were titrated into a mixture of ^BMOERIDC1₂ and MagFura-2, and the changes in MagFura2 fluorescence were monitored (Fig. 6b–d, Fig. S2). In parallel, titrations were performed to determine the affinity of Mag-Fura-2 for these metal ions (Fig. S3), which yielded similar values to those previously reported.^25,26 Because ^BMOERIDC1₂ does not self-associate into a pre-formed tetramer in the absence of metal binding, and the titration solutions contain a mixture of ^BMOERIDC1₂ and ^BMOERIDC1₄ in their various metal-bound forms, it is not possible to extract accurate metal binding equilibrium constants. Nevertheless, the titrations show that ^BMOERIDC1₄ can accommodate four equivalents each of Co^II, Ni^II, or Zn^II with affinities that rival that of MagFura-2 for those ions. The competition isotherms can be reasonably well described by a 4-equivalent-binding-site model, yielding half-saturation concentrations of 9 μM for Co^II, 0.5 μM for Ni^II and 0.04 μM for Zn^II (Table 1). It should be noted that these are not true dissociation constants, but rather approximate metal concentrations where half of the starting ^BMOERIDC1₂ population is converted into the fully metal-bound, M₄:^BMOERIDC1₄ species. Given that i, i+4 bis-histidine motifs on a-helices–which constitute the metal binding motifs on the parent MBPC1 molecule–bind Co^II, Ni^II, or Zn^II with micromolar affinities that are within an order of magnitude of each other (K_d-Zn ≈ K_d-Ni < K_d-Co),^27,28 our results clearly show an increased thermodynamic selectivity for Zn^II coordination over the other two ions.

Figure 6.

^BMOERIDC1₂ tetramerization mediated by various divalent metal ions with a preference for Zn^II binding. a) SV profiles for 2.5 μM ^BMOERIDC1₂ dimer in the presence of 5 μM Co^II, Ni^II. Cu^II, and Zn^II, reflecting the formation of a tetrameric species on binding all four metal ions. (b-d) Binding isotherms for Mag-Fura-2 / ^BMOERIDC1₂ competition experiments with (b) Co²⁺, (c) Ni²⁺, and (d) Zn²⁺, with tick marks shown on the top x-axis that correspond to theoretical titration end points for the indicated number of binding sites per tetramer. Solid lines represent fits obtained using DynaFit to a four-equivalent-bindingsites model, with K_d values listed in Table 1. Cu²⁺ competition experiments (e) demonstrate 12 metal binding events.

Table 1.

Apparent Dissociation Constants for Divalent Metal Ion Adducts of Mag-Fura2 and ^BMOERIDC1₂.

	Kd (μM)
metal	Mag-Fura-2	BMOERIDCl2
Zn	0.047 (0.005)	0.041 (0.010)
Ni	0.175(0.014)	0.54 (0.055)
Co	1.41 (0.22)	9.29(1.02)

The ultimate test for Zn^II selectivity is provided by Cu^II, which lies atop the Irving-Williams series²⁹² for the relative stabilities of divalent metal complexes (Mn^II < Fe^II < Co^II < Ni^II < Cu^II > Zn^II), and readily outcompetes Zn^II for most natural and synthetic Zn-specific binding sites.^30–32 This is particularly the case for flexible peptides or proteins that cannot impose steric constraints to offer selectivity between metalions.³³ In accord with the high affinity of Cu^II for protein-based donor atoms, our titrations show that ^BMOERIDC1₄ has at least twelve binding sites for Cu^II that can compete with MagFura-2 (i.e., their K_d’s are within an order of magnitude of 8.5 nM)³⁴ (Fig. 6e), some likely on the surface and some in the core. The high plurality of Cu^II binding to ^BMOERIDC1₄ and the availability of surface binding sites make it challenging to assess the Zn^II vs. Cu^II selectivity of the internal coordination sites through titrations. We therefore cocrystallized ^BMOERIDC1₂ in the presence of a mixture of Zn^II and Cu^II. The structure of the resulting complex (PDB ID: 3QVZ) was determined at 2.6 Å resolution, which revealed that it possesses the same supramolecular architecture as Zn₄:^BMOERIDC1₄, with four metals associated with the core and seven with the surface of the tetramer (Fig. 7). To identify these metal ions, anomalous difference maps were calculated using X-ray diffraction data collected at the Zn and Cu K-edges (1.265 and 1.377 Å). These maps indicate the core ions to be Zn, which display anomalous scattering only at the Zn K-edge and not at the Cu K-edge, and the majority (6 out of 7) of surface ions to be Cu. X-ray fluorescence scans further confirm the presence of both Cu and Zn in the thoroughly washed crystals (Fig. S4). Subsequently, we investigated through size exclusion chromatography (SEC) whether Zn₄:^BMOERIDC1₄ is a true thermodynamic product. The SEC elution profile of Zn₄:^BMOERIDC1₄ ran in the presence of Zn indicates the presence of both the tetrameric and the dimeric (as well as higher-order) species (Fig. S5).³⁵ Although the observed dimeric species is a minor component, as would be predicted from the SV experiments (Fig. 3a), the SEC findings suggest that Zn₄:^BMOERIDC1₄ is in exchange with its dimeric form in solution, and that Zn vs. Cu selectivity observed under crystallization conditions likely is thermodynamically driven.³⁶ We attribute the thermodynamic basis for Zn selectivity not only to the formation of a covalently and non-covalently stabilized macrocycle built around the tetrahedral Zn coordination sites, but also to the fact that this macrocyle has been programmed to house four Zn ions in its interior, all of which contribute to the overall stability of Zn₄:^BMOERIDC1₄ formation.

Figure 7.

Structure of Zn₄:^BMOERIDC1₄ grown in the presence of Cu^II. A ribbon representation of Zn₄:^BMOERIDC1₄ crystallized in the presence of CuSO₄ is shown along with the anomalous difference maps generated from data collected (a) at the Cu K-edge (green, 5 s) and (b) at the Zn K-edge (blue, 5 s). The surface-associated Cu ions are shown with arrows in (a). The heme Fe centers, as well as well-ordered surface Cu ions have anomalous contributions at both energies, whereas the core Zn ions only contribute at the higher energy Zn K-edge.

In summary, we have shown here that a traditional approach in synthetic coordination chemistry (metaltemplated ligand synthesis) can be applied to the design of a large and dynamic protein assembly, which displays an extent of Zn selectivity that is challenging to achieve in small molecule systems. The advantage of proteins as ligand scaffolds lies in their chemically rich surfaces that enable the formation of an extensive network of stable, yet reversible non-covalent interactions around the metal centers. The ensuing, simultaneous fulfillment of plasticity and selectivity, such as that displayed in our system, is a hallmark of all cellular signaling systems, and particularly those that couple selective metal binding to conformational changes or oligomerization to activate downstream processes, such as the prototypical EF-hand signaling proteins (calmodulin,³⁷ S100³⁸) or any metalloregulatory gene transcription factor.³⁹

EXPERIMENTAL SECTION

Protein Mutagenesis. Purification, Modification with Crosslinkers.

The G82C-RIDC1 cyt cb₅₆₂ construct (^C82RIDC1) was generated, expressed, and purified according to published procedures.^22,40,41 Purified ^C82RIDC1 was buffer exchanged into 20 mM TRIS (pH 7) and 50 mM dithiothreitol (DTT) and concentrated using Amicon stirred cells with MW=10 kDa cutoff-filters. Concentrated protein was eluted through a 10 DG desalting column (BioRad) to eliminate excess DTT and diluted with an appropriate volume of 20 mM TRIS (pH 7) to yield a protein concentration of 100 μM. Concentrated stock solutions of BMOE, BMB, or BMH (Pierce) were added to the sample in ten equal aliquots every 30 sec over 5 min to give a final crosslinker concentration of 100 μM. The reaction was then allowed to proceed at room temperature under constant stirring for 30 min. Crosslinking reactions were quenched by the addition of 50 mM DTT and allowed to incubate at room temperature for 15 min. Crosslinking yields were approx. 50%. Crosslinked proteins were subsequently purified through multiple size exclusion chromatography runs using a Pall ACA54 resin (fractionation range = 5–70 kDa) using an elution buffer of 20 mM TRIS (pH 7) and 150 mM NaCl. Dimer purity was assessed by SDS-PAGE electrophoresis, and the compositions of the dimers were verified by MALDI mass spectrometry.

Analytical Ultracentrifugation and SV simulations.

Sedimentation velocity (SV) measurements were made on a Beckman XL-I Analytical Ultracentrifuge (Beckman-Coulter Instruments) using an An60 Ti rotor at 41,000 rpm for a total of 250 scans per sample. Data were collected at 415 nm for low concentration (2.5 μM crosslinked dimer, 5 μM RIDC1 monomer) samples, and 560 nm for higher concentrations (25 μM crosslinked dimer). All data were processed in SEDFIT⁴² with the following fixed parameters: buffer density (r) = 0.99764 g/ml; buffer viscosity = 0.0089485 poise; Vbar, which was calculated to be 0.7310 ml/g for ^BMOERIDC1₂ and ^BMBRIDC1₂; 0.7306 ml/g for ^BMHRIDC1₂. The buffer density (r) and viscosity for 20 mM TRIS (pH 7) were calculated through SEDNTERP (http://www.jphilo.mailway.com/default.htm). Theoretical sedimentation coefficients were calculated using Hydropro Version 7.C,²³ using the same parameters employed in SV data processing.

Crystallography.

Crystals of all crosslinked dimers were obtained by sitting drop vapor diffusion at room temperature. The crystallization conditions for the four different crystal forms described in this study are listed in Table S1 and Table S2, along with corresponding data collection and refinement statistics. All protein stocks solutions were in 20 mM TRIS (pH 7) buffer. Appropriate crystals were transferred to a solution of mother liquor containing 20% glycerol as cryoprotectant and frozen in liquid nitrogen prior to data collection at 100 K.

Data were integrated using MOSFLM and scaled in SCALA,⁴³ except for Zn₄:^BMBRIDC1₄, which was processed using SAINT and Bruker SADABS. All structures were determined through molecular replacement with MOLREP,⁴⁴ using the RIDC1 monomer structure (PDB ID:3HNI) as the search model, followed by rigid-body, positional, thermal and TLS refinement with REFMAC⁴⁵ using appropriate noncrystallographic symmetry restraints. The crystals for both forms of Zn₄: ^BMOERIDC1₄ (crystallized in the presence and the absence of CuSO₄) were found to have 48% and 44% twin fractions, respectively. Both data sets were refined using intensity-based twin refinement as implemented in REFMAC using the twin operator: −1.000h-1.000k, 1.000k, -l. All structures were manually built in COOT⁴⁶ to produce the final models. All figures were produced using PYMOL.⁴⁷

Metal Binding Titrations.

For Mag-Fura-2 competition titrations, 1 mg Mag-Fura-2 (Invitrogen) was resuspended in 1 ml Millipore purified water. Mag-Fura-2 concentration was determined using an extinction coefficient of 22,000 M⁻¹cm⁻¹ at 369 nm. The titrations for determining Mag-Fura-2 affinities for Ni^II and Co^II were performed in a 1-cm cuvette with a 2 ml sample volume containing 0.5 μM MagFura-2 and 1 mM CaCl₂ in Chelex-treated 20 mM 3-(N-morpholino)propanesulfonic acid (MOPS, pH 7) buffer with 150 mM NaCl. The sample was allowed to equilibrate with stirring for 3 min. after each addition of the competing metal stock before recording fluorescence excitation scans monitoring emission at 505 nm. Zn^II titrations were performed with 1 mM Ni^II (sulfate salt) as the competing ion in place of CaCl₂. Titration data (fluorescence excitation monitored at 330 nm) were fit to a 1-site binding model using Dynafit⁴⁸ assuming a Mag-Fura-2:Ca²⁺ K_d of 25 μM, which was found to be constant over a wide variety of pH and ionic strength conditions.²⁶ The resulting K_d values are listed in Table 1. The Mag-Fura2:Zn^II K_d was back-calculated using the Mag-Fura-2:Ni^II K_d determined here.

^BMOERIDC1₂/Mag-Fura-2 competition assays were performed in a 1-cm cuvette with a 1.5-ml sample volume containing 25 μM ^BMOERIDC1₂ and 10 μM Mag-Fura-2 in Chelex-treated MOPS (pH 7) buffer with 150 mM NaCl. As with the titrations described above, samples were allowed to equilibrate under stirring for 3 min. after the addition of each aliquot of the appropriate divalent metal stock before recording fluorescence excitation scans monitoring emission at 505 nm. Ni^II and Co^II binding to MagFura-2 results in quenching of the florescence signal, which was subsequently followed using the emission readout using 372 nm excitation. Zn^II binding results in an increased fluorescence excitation at 323 nm, which was used as the excitation wavelength. All titration curves could be adequately described with a minimal (4 ´ 1) binding model consisting of four binding sites with equivalent half-saturation values listed in Table 1. Cu^II titrations were performed similarly to Ni^II, Co^II and Zn^II, except Mag-Fura-2:Cu^II binding was monitored by the increase in absorbance at 300 nm, rather than by fluorescence.

Analytical Size Exclusion Chromatography (SEC).

Size exclusion chromatography was performed on a Biologic DuoFlow FPLC system (Bio-Rad) equipped with a UV detector set to monitor absorbance at 280 nm. For each run, 100 μl of 25 μM ^BMOERIDC1₂ in 20 mM TRIS (pH 7) buffer with 50mM NaCl and either 50 μM ZnCl₂ or 5 mM EDTA were injected on an analytical SEC column packed with Pall ACA54 resin (fractionation range = 5–70 kDa). The samples were eluted from the column at 0.5 ml/min in 20 mM TRIS (pH 7) with 50 mM NaCl, and additionally, 5 mM EDTA for the metal free sample. The dead volume in Fig. S5 is based on a ferritin (MW ~ 500 kDa) standard run under conditions identical to the metal free sample.