Realization of a Designed Three-Helix Bundle Capable of Binding Heavy Metals in a Tris(Cysteine) Environment

An important objective of de novo protein design is the preparation of metalloproteins, as many natural systems contain metals which play crucial roles for the function and/or structural integrity of the biopolymer.[1, 2] Metalloproteins catalyze some of the most important processes in nature, ranging from energy generation and transduction to complex chemical transformations. At the same time, metals in excess can be deleterious to cells and some ions are purely toxic, having no known beneficial effects (e.g., Hg^II or Pb^II). One hopes to use a first principles approach for realizing both known metallocenters and also to prepare novel sites which may lead to exciting new catalytic transformations. However, designing novel metalloproteins is a challenging and complex task, especially if one desires to prepare asymmetric metal environments.

Numerous metalloprotein systems have been designed over the past fifteen years, typically using unassociated peptides that assemble into three- stranded coiled coils or helix-loop-helix motifs that form anti-parallel four- stranded bundles. In terms of metal ion binding these systems have been functionalized with heme,[3, 4] and non-heme mononuclear,[5] and binuclear centers.[6, 7] It is often difficult to prepare non-symmetrical metal sites through these strategies due to the symmetry of the systems relying on homo-oligomerization. Thus, the preparation of a single polypeptide chain capable of controlling a metal coordination environment is a key objective.

Previously, we have designed soft, thiol-rich metal binding sites into the interior of parallel, three-stranded α-helical coiled coils involving cysteine and/or penicillamine as the ligating amino acid residues.[8, 9] These systems have served as hallmarks for understanding the metallobiochemistry of different heavy metals such as Cd^II, Hg^II, As^III, and Pb^II.[8-11] We have shown how to control the geometry and coordination number of metals such as Cd^II and Hg^II at the protein interior as well as fine-tuning the physical properties of the metals, which has led to site-selective molecular recognition of Cd^II.[12-14] While these homotrimeric assemblies have been very useful, the production of heterotrimeric systems that are capable of fine tuning metal environments controllably or introduction of a site specific H-bond has been elusive.[15] Therefore, we have chosen an alternative strategy to satisfy this objective using a single polypeptide chain instead of multiple self associating peptides.

Existing designed heteromeric helical bundles and coiled-coils show energetic preferences of several kcal/mol for the desired heteromeric versus homomeric assemblies.[16, 17] However, the energy gap between a hetero and homomeric assembly often depends critically on ionic strength, pH, and other environmental parameters. Moreover, the objective of many studies in de novo protein design is to impart an energetically sub-optimal coordination geometry on the metal ion, and the degree to which this strategy will be successful depends on the size of the energy gap between the desired heteromeric assembly and other homomeric or misfolded states. Also, even when heterooligomeric bundles have been used to successfully identify specific environmental effects that influence substrate binding or metal ion cofactor reactivity,[18] the non-covalently assembled complexes have often been difficult to characterize structurally, possibly due to small populations of alternately assembled species. In this case, combining the active site residues into a construct with linked helices greatly facilitated structural analysis and catalytic characterization.[19]

An attractive starting scaffold to meet our objectives is the de novo designed three-helix bundle α₃D. The NMR structure of this protein is known and it has been proven that the helices are oriented in a counter-clockwise topology.[20] Although the α₃D protein originated from a coiled-coil, its helices were shortened to the point where it might be better considered as a globular protein whose repetitive structure makes each of the heptads very similar to one another (in the absence of end-effects). The stability of α₃D is similar to that of natural proteins and thus α₃D should be tolerant to mutations, allowing this protein to serve as an excellent framework to engineer specific metal binding sites. Additionally, with this protein scaffold we can study the effect of the ligating residue located on the second helix which is antiparallel to the first and third helices of the bundle.

Before attempting to prepare asymmetric metal coordination environments or site specific H-bonds, we felt it was important to first redesign α₃D so as to introduce symmetric metal binding sites involving cysteine residues. This approach would allow us to exploit the extensive body of work defining heavy metal complexation using the TRI and Coil Ser peptides,[8, 15] to assess whether a specific metal structure could be achieved in the modified α₃D construct. It would also allow us to probe the physical properties of metals such as Cd^II, Hg^II and Pb^II in a more natural antiparallel helical system. A significant amount of literature exists where metal binding sites have been designed in existing four-helix bundles and in a mixture of α/β protein structural frameworks.[21, 22] However, there are no such examples where novel metal binding site(s) have been engineered within an antiparallel single chain three-helix bundle by rational design. The three-helix bundle occurs ubiquitously in nature as a versatile and robust scaffold ranging from the helical IgG-binding domains,[23] to DNA-binding proteins, structural proteins, and enzymes..[24] Despite its widespread occurrence in nature only a few attempts have been made to prepare single chain three-helix bundles. [25, 26]

Based on visual inspection of the α₃D structure, four potential sites along the bundle were identified where three cysteines, one from each helix, could be introduced. Out of these four mutants, α₃DIV (Figure 1), located at the C-terminal end of the bundle, seemed to be optimal based on the properties of the starting protein.[20, 27-29] Previous NMR structural and dynamic investigations showed a gradient in the dynamic behavior and malleability of the protein, with the C-terminal end of the bundle being most amenable to amino acid substitutions. The selected location has a well-ordered backbone conformation, but the side chains of the residues to be mutated are less well ordered than residues in other locations of the bundle. The 3-Cys site, which is largely sequestered from solvent, occupies a “box”; the sides are formed by the backbone of the helices and the bottom by the apolar side chains of Phe 31, Ile 14, and Ile 63. The aromatic residue of Phe 31 lies directly over the predicted metal-binding site and lines most of the bottom of the box. The top is formed by the main-chain and side chains of residues in the non-helical loops including Leu 21, as well as the apolar portion of Tyr 70 at the terminus of helix 3. His 72, which was entirely disordered in the NMR solution structure, also lies proximal to the site. Moreover, after introduction of the Cys side chains in one of the two preferred rotamers for Cys, the thiol S^G atoms formed a nearly equilateral triangle with the side chains well oriented to form the desired site (inter S^G distances = 3.5 − 4.5 Å). Overall, the location is ideal to explore the effects of hydrophobic sequestration in the present work. The sequence of α₃DIV is shown in Table 1. After expression of a synthetic gene of α₃DIV in E. coli followed by purification by HPLC (see Materials and Methods) the MW of α₃DIV was determined to be 7945.1 Da by Electrospray Ionization Mass Spectrometry (ESI), which corresponds to α₃DIV with the deletion of the N-terminal Met (calc MW = 7946.9 Da). The folding behavior of α₃DIV was studied in solution by CD and NMR spectroscopy. The CD spectrum of 5 μM α₃DIV has double minima at 208 and 222 nm at pH 8 with the molar ellipticity [θ] values characteristic of a well-folded α-helical construct (97% folded based on [θ]₂₂₂) (Figure S1A). Furthermore, α₃DIV remained well-folded between pH 3 and 9. The ¹H-¹H NOESY spectrum of 3 mM α₃DIV shows chemical shift dispersions characteristic of a well-folded α-helical structure at pH 6 (Figure S1B). GuHCl-induced unfolding of α₃DIV was performed by monitoring the change in ellipticity at 222 nm as a function of GuHCl concentration at pH 8. The resulting titration curve was plotted as concentration of folded protein vs. GuHCl concentration (Figure S2) and was fit to a two-state equilibrium,[30-32] yielding the free energy of unfolding (ΔG_u) to be 2.5 kcal mole⁻¹, with the degree of cooperativity (m) being 1.4 kcal mole⁻¹ M_GuHCl⁻¹. The midpoint of the transition (C_m) occurred at a 1.8 M GuHCl concentration. These results indicate that replacing three Leu residues of the WT α₃D with cysteines resulted in a loss of unfolding free energy of ~ 2.5 kcal mole⁻¹.[20]

Figure 1.

PyMol model of α₃DIV generated from the NMR structure of α₃D. Cys residues, located at the C-terminal end of the bundle are shown as spheres. The protein backbone is shown in orange. The Cys site can be considered to be located in a hydrophobic “box” formed by the hydrophobic residues F31, I14, I63, L21, and Y70, shown as sticks.

Table 1.

Sequence of α₃DIV.^a

Peptide	Sequence
α3DIV	MGS WAEF K QR LAAIKTR CQAL GG SEAECAAF E KE IAAFESE LQAY KGKGNPE
	V E AL R KE AAAIRDE CQAY RHN

^aResidues in red represent mutations to the WT α₃D

Having established that α₃DIV is well folded and stable in solution, we investigated the complexation of Hg^II, Cd^II, and Pb^II to this peptide. Metal ion binding titrations were performed by adding aliquots of stock solutions of metals to peptide solutions at pH values where each metal fully coordinated to α₃DIV (pH 8.6 for Hg^II, and pH 8 for Cd^II and Pb^II). The progress of the titrations was monitored by the appearance of characteristic absorption bands due to Ligand to Metal Charge Transfer (LMCT) transitions at characteristic wavelengths,[15, 33, 34] upon formation of metal-thiolate bonds in the complexes Hg^II(α₃DIV)⁻, Cd^II(α₃DIV)⁻, and Pb^II(α₃DIV)⁻. The resulting UV/Vis absorption spectra (Figure S3) and the molar absorption coefficients (Δε), at various wavelengths (Table 2) are consistent with the assignment that all three Cys thiolates of α₃DIV are incorporated into the first coordination sphere of the metal ions.[15, 33, 34] Cd^II and Pb^II binding constants were determined to be 2.0×10⁷ M⁻¹ and 3.1×10⁷ M⁻¹, respectively, from the analysis of the titration data.[35] Due to the high affinity binding, the association constant of Hg^II could not be determined. Next, we examined pH dependent complexation of these metals to α₃DIV by monitoring changes in the LMCT band as a function of pH. Titration data of Hg^II were fit to the release of one thiol proton upon forming a HgS₃ complex from linear HgS₂(SH) complex of α₃DIV,[15] resulting in a pK_a of 7.1±0.1 (Table 2, Figure S4). For Cd^II and Pb^II the titration curves (Figure S4) were fit to simultaneous dissociation of two Cys thiols,[15, 33, 34, 36] yielding pK_a2 of 10.6±0.1 and 10.2±0.1 (Table 2), respectively. These pK_a values are slightly more acidic than TRI peptides.[15, 33, 36] Nonetheless, these pK_a values are consistent with the coordination modes of Hg^II as trigonal HgS₃, Cd^II as pseudotetrahedral CdS₃O (O being an exogenous water molecule), and Pb^II as trigonal pyramidal PbS₃ complexes.

Table 2.

Physical Parameters of Cd^II, Hg^II and Pb^II complexes of α₃DIV.

Complex	UV/Vis λnm (Δε M−1cm−1)	NMR δ (ppm)		Apparent pKa	Binding Constant (Kb) (M−1)a
		113Cd	199Hg
Cd(α3DIV)−	232 (18, 200)	583		10.6±0.1b	2.0×107
		595
Hg(α3DIV)−	247 (12,500)		−244	7.1±0.1c
	265 (8,400)
	295 (3,900)
Hg(α3DIV)d	240 (850)		−938e
Pb(α3DIV)−	236 (18,000)			10.2±0.1b	3.1×107
	260 (14,400)
	278 (9,100)
	346 (3,150)

^aModel used to obtain binding constants is M^II + (α₃DIV)³⁻ ⇌ M^II(α₃DIV)⁻ (K_b). These values represent the lower limit of K_b.

^bModel used to obtain pK_a2 values for Cd^II and Pb^II is M^II(α₃DIVS(SH)₂)⁺ ⇌ M^II(α₃DIV)⁻ + 2H⁺ (K_a2).

^cModel used to obtain the pK_a for Hg^II is Hg^II(α₃DIVS₂(SH)) ⇌ Hg^II(α₃DIV)⁻ + H⁺ (K_a).

^dLinear HgS₂ complex of α₃DIV.

^e199Hg NMR δ of linear HgS₂ complex at pH 5.8.

¹⁹⁹Hg NMR and ^199mHg PAC (Perturbed Angular Correlation) spectroscopies were used to probe the coordination environment around Hg^II bound to α₃DIV at pH 5.8, 8.6, and 7.4. Based on the pK_a value, Hg^II is expected to form a linear HgS₂ complex at pH 5.8 and a trigonal HgS₃ complex at pH 8.6, with a mixture of linear and trigonal complexes at the intermediate pH. ¹⁹⁹Hg NMR chemical shifts (Figure 2) and ^199mHg PAC spectral parameters (Figure S5, Table S1) of α₃DIV at pH 5.8, 8.6, and 7.4 confirm that α₃DIV forms a linear HgS₂ complex at pH 5.8 and a trigonal HgS₃ complex at pH 8.6, with the formation of both dithiolate- and trithiolate-Hg^II complexes at pH 7.4 in distorted geometries.[10, 37-43] ¹¹³Cd NMR spectrum of α₃DIV has two resonances at δ = 595 and 583 ppm at pH 8 (Figure S6) indicating the presence of two Cd^II species with chemical shift values similar to what has been observed for 4-coordinate pseudotetrahedral CdS₃O species.[36] ^111mCd PAC spectroscopy was used to confirm the coordination environment and geometry of Cd^II complexes of α₃DIV. ^111mCd PAC spectrum of α₃DIV has three Nuclear Quadrupole Interactions (NQIs), at ω₀ = 0.35, 0.27 and 0.17 rad/ns (Figure S7, Table S2) at pH 8.1. The first two peaks at 0.35 rad/ns and 0.27 rad/ns agree strikingly well with a typical CdS₃O signal in exo and endo conformations, respectively, as observed for TRI peptides.[36] The lowest frequency NQI at 0.17 rad/ns can be best assigned to a CdS₃N species where N is His 72. Even though the chemical shift of 595 ppm in the ¹¹³Cd NMR spectrum (Figure S6) is lower than what has been reported for a CdS₃N species,[44] Quantum chemical calculations show that a change in Cd-S bond length of 0.01 Å can cause a change in chemical shift of ~20 ppm (Hemmingsen et.al. unpublished results). Thus, tentatively, the ¹¹³Cd NMR resonance at δ = 595 ppm can be best assigned to the CdS₃N species.

Figure 2.

¹⁹⁹Hg NMR spectra of solutions containing 2.93 mM α₃DIV and 0.8 equivalent of ¹⁹⁹Hg(NO₃)₂ at pH 5.8 (A), 7.4 (B), and 8.6 (C).

In conclusion, we have been successful in engineering metal binding sites containing cysteine residues in an existing de novo antiparallel three-helix bundle, yielding α₃DIV, a protein which is well folded and stable in solution, and capable of binding heavy metals with high affinity (>10⁷ M⁻¹). Spectroscopic properties of Hg^II, Cd^II and Pb^II complexes of α₃DIV are very similar to existing parallel three-stranded coiled coils, showing that we have achieved our objective of preparing a single polypeptide chain capable of binding metal ions with high affinity and predefined coordination geometry. Understanding the biochemistry of binding of heavy metals to a single polypeptide chain can be potentially useful for developing peptide-based water purification systems or specific heavy metal ion sensors. Clearly, having a single peptide chain rather than self-associating helical peptides makes these goals more achievable. Further studies will explore the possibilities of preparing similar constructs containing asymmetric metal binding sites such as those in Type I blue copper proteins, and to explore the effects of the electronic structure of the aromatic residue, Phe 31, or second-shell effects by replacing any of the surrounding residues as well as making catalytic metalloproteins.

Experimental Section

A synthetic DNA of α₃DIV was cloned into pET-15b (Celtek Genes) vector and expressed in E. coli BL21(DE3) competent cells (Stratagene) grown in M9 media. After sonication and heat denaturation at 55°C the lyophilized powder was purified on a C18 preparative reverse phase HPLC column over a linear gradient of 100% H₂O/0.1% TFA to 10% H₂O/90% Acetonitrile/0.1% TFA over 50 minutes. The MW of the pure peptide was determined to be 7945.1 Da by ESI which corresponds to α₃DIV without the first Met (calc MW 7946.9 Da). Yield of the pure protein was 17 mg/L. Concentration of the protein was determined by A_280nm using the known extinction coefficient ε₂₈₀ = 8.61 mM⁻¹.[45]

CD spectra were collected on an Aviv model 202 CD spectrometer using rectangular open top quartz cuvettes at 25°C. GuHCl titration experiments were carried out using a Microlab 500 series syringe pump automatic titrator controlled by Aviv software. Titrations were carried out by mixing two separate solutions of 10 μM peptide containing 0.0 and 7.63 M GuHCl in 10 mM phosphate buffer at pH 8. Observed ellipticities in millidegree were converted to molar ellipticities (deg cm² dmole⁻¹ res⁻¹) as described previously,[33] using 59 amino acids in the helical region of the protein. GuHCl titration data were fit to an equation derived based on a two-state model. [30-32] A ¹H-¹H NOESY experiment was performed using standard procedures.[46]

Cd^II, Hg^II, and Pb^II binding titrations were performed at room temperature on a Cary 100 Bio UV/Vis spectrometer using anaerobic cuvettes (Starna Inc.) of 1-cm path length by adding aliquots of stock solutions of different metals. Peptide samples of 20 – 30 μM were prepared in 50 mM of appropriate buffers (TRIS for pH 8 and CHES for pH 8.6) and 40 – 60 μM TCEP inside an inert atmosphere box (Vacuum Atmospheres Co., model OMNI-LAB). Stock solutions of 8 mM CdCl₂, 7.37 mM HgCl₂, and 5.16 mM PbCl₂ were also prepared inside the inert atmosphere box. In each case, difference spectra were obtained by subtracting the background spectra of samples containing peptide, buffer, and TCEP. Direct titration data were analyzed by non-linear least squares fits to an equation used previously.[47] The difference molar extinction coefficients (Δε) were determined based on the total metal concentrations after subtracting the background spectra.

pH titrations were performed as described previously,[34, 36] by adding small aliquots of KOH to solutions containing 20 – 30 μM of α₃DIV and 1 equivalent of CdCl₂, HgCl_2′ or PbCl₂. In the cases of Cd^II and Pb^II, the titration data were analyzed using the model: simultaneous two proton dissociation as described previously.[34, 36] For Hg^II, the data were analyzed using the model: dissociation of one thiol proton of Cys.[15]

¹¹³Cd NMR and ¹⁹⁹Hg NMR experiments were performed according to standard procedures.[48] An exponential line broadening of 200 Hz was applied prior to Fourier transformation while processing ¹⁹⁹Hg NMR data.

Samples for ^111mCd PAC measurement contained 300 μM α₃DIV, 1/12 equivalent of Cd^II, and 55% sucrose (w/w) in 20 mM TRIS buffer at pH 8.1. Sample preparation and data collection were performed at the University of Copenhagen.[36, 48] Samples for ^199mHg PAC experiments contained 200 μM α₃DIV, 80 μM Hg^II, and 55% sucrose in 100 mM of an appropriate buffer (phosphate for pH 5.8, 7.4 and CHES for pH 8.6). Sample preparation and data collection were performed at CERN.[41]