Heteromeric three-stranded coiled coils designed using a Pb(ii)(Cys)₃ template mediated strategy

Abstract

Three-stranded coiled coils are peptide structures constructed from amphipathic heptad repeats. Here we show that it is possible to form pure heterotrimeric three-stranded coiled coils by combining three distinct characteristics: (1) a cysteine sulfur layer for metal coordination, (2) a thiophilic, trigonal pyramidal metalloid (Pb(ii)) that binds to these sulfurs and (3) an adjacent layer of reduced steric bulk generating a cavity where water can hydrogen bond to the cysteine sulfur atoms. Cysteine substitution in an a site yields Pb(ii)A₂B heterotrimers, while d sites provide pure Pb(ii)C₂D or Pb(ii)CD₂ scaffolds. Altering the metal from Pb(ii) to Hg(ii) or shifting the relative position of the sterically less demanding layer removes heterotrimer specificity. Because only two of the eight or ten hydrophobic layers are perturbed, catalytic sites can be introduced at other regions of the scaffold. A Zn(ii)(histidine)₃(H₂O) centre can be incorporated at a remote location without perturbing the heterotrimer selectivity, suggesting a unique strategy to prepare dissymmetric catalytic sites within self-assembling denovo-designed proteins.

De novo protein design studies a wide range of protein systems in simplified peptidic environments^1–4. Bioinorganic constructs have been used to study the relationship between the protein scaffold and the function of a metal cofactor^5–12. A wide range of metal geometries and binding sites have been modelled in three-stranded coiled coils (3SCCs)^13–20. 3SCCs consist of amphipathic heptad repeats of amino acids (residues sequentially labelled abcdefg). Hydrophobic residues in the first (a) and fourth (d) positions form the core of the coiled coil, which is further stabilized by salt bridging between exterior residues in the fifth (e) and seventh (g) positions. Metal-binding sites result when any of the interior residues (a or d sites) are substituted with metal-binding residues. In the 3SCC these positions are designated a or d layers. Substitution of a Leu for a Cys generates a sulfur-rich heavy metal-binding site^{13,18,20–22} while a His substitution generates a catalytic site like that of carbonic anhydrase (CA) or copper nitrite reductase (CuNiR)^7,11,23,24.

Our lab has previously reported the best aqueous peptidic models of both CA and CuNiR within these scaffolds^10,23. The active site of CA consists of Zn(ii) bound pseudo-tetrahedrally to three His ligands and an exogenous solvent molecule²⁵. The previously reported CA model performs to within 300-fold efficiency of the most active human enzyme, but does not reproduce critical acid/base interactions due to the symmetry limitations imposed by the homotrimeric scaffold^{11,23,26–28}. Similarly, our CuNiR model accurately represents the first coordination sphere of the native enzyme, Cu bound to three His ligands and a water, but cannot model the second coordination sphere accurately^24,29–32. Attempts to include individual second coordination sphere residues are thwarted by the three-fold symmetry of the TRI scaffolds²⁴. Therefore, dissymmetric structures of the type A₂B or ABC are vital to the advancement of self-assembling de novo designed metalloproteins as many catalytic centres are dissymmetric in either the first or second coordination spheres, if not both. More generally, methods to prepare quantitatively and reliably dissymmetric scaffolds for different stoichiometries of self-assembling designed proteins are in their infancy.

Several groups have designed self-assembling, parallel, dissymmetric 3SCCs, but these constructs have either required substitution at each layer within the hydrophobic core or non-natural amino acids to ensure asymmetry^33–35. Herein, we report a metal templating strategy using only coded amino acids that provides the first parallel, dissymmetric 3SCC modified within a single heptad. Although discussed specifically for 3SCCs, this approach underlines the general importance of metal template protein-protein interactions (MTPPIs) as a mechanism to achieve dissymmetric assemblies^36,37. This new set of scaffolds provides a significant advance in metalloprotein design, allowing the preparation of dissymmetric metal-binding sites within the assembly, without costly non-natural residues. Our scaffolds require a cysteine layer for Pb(ii) binding and a mixture of leucine and alanine residues in an adjacent layer (Table 1).

Table 1 |.

List of abbreviations of representative peptide sequences with mutations shown in bold

Peptide name	Abbreviation	Full sequence
T		Ac-G WKALEEK LKALEEK LKALEEK LKALEEK G NH2
G		Ac-G WKALEEK LKALEEK LKALEEK LKALEEK LKALEEK G-NH2
C		Ac-E WEALEKK LAALESK LQALEKK LEALEHG-NH2
GC		Ac-E WEALEKK LAALESK LQALEKK LQALEKK LEALEHG-NH2
G-L12AL16CL30H	G-A	Ac-G WKALEEK LKAAEEK CKALEEK LKALEEK HKALEEK G-NH2
G-L16CL30H	G-B	Ac-G WKALEEK LKALEEK CKALEEK LKALEEK HKALEEK G-NH2
G-L12CL16A	G-C	Ac-G WKALEEK LKACEEK AKALEEK LKALEEK LKALEEK G-NH2
G-L12C	G-D	Ac-G WKALEEK LKACEEK LKALEEK LKALEEK LKALEEK G-NH2
T-L12AL16C	T-A	Ac-G WKALEEK LKAAEEK CKALEEK LKALEEK G-NH2
T-L16C	T-B	Ac-G WKALEEK LKALEEK CKALEEK LKALEEK G-NH2
GC-L12AL16C	GC-A	Ac-E WEALEKK LAAAESK CQALEKK LQALEKK LEALEHG-NH2
GC-L16CL30H	GC-B	Ac-E WEALEKK LAALESK CQALEKK LQALEKK HEALEHG-NH2
GC-L12CL16A	GC-C	Ac-E WEALEKK LAACESK AQALEKK LQALEKK LEALEHGH2
C-L12C	C-D	Ac-E WEALEKK LAACESK LQALEKK LEALEHG-NH2
G-L16CL26TL30H	G-BT	Ac-G WKALEEK LKALEEK CKALEEK LKATEEK HKALEEK G-NH2
G-L16CL26DL30H	G-BD	Ac-G WKALEEK LKALEEK CKALEEK LKADEEK HKALEEK G-NH2

In the text, a-site peptides are designated A or B corresponding to whether they contain an alanine (A) or a leucine (B) above the cysteine, regardless of the remainder of the sequence. Similarly, d-site peptides are designated C or D corresponding to whether they contain an alanine (C) or a leucine (D) below the cysteine, regardless of the remainder of the sequence.

The interior of coiled-coil peptides contains hydrophobic residues in a desolvated core. Substitution of these residues for a metal-binding residue (for example, Cys, His, Asp) generates a 3SCC of reduced stability. Similarly, reducing the size of the side chain (for example, alanine) disrupts optimal packing and allows water to enter the core^38,39. Although building a metal-binding site in the centre of the coiled coil and allowing water into the hydrophobic core are both destabilizing, it is possible that when these modifications occur in tandem at adjacent layers, new stabilizing forces may result. Specifically, a symbiotic relationship between a hydrated alanine layer and metal-binding ligands can potentially lead to new hydrogen-bonding interactions that stabilize the system. We will show that this previously unappreciated cooperative interaction can be exploited to generate A₂B type 3SCCs. The optimization of stability of this heterotrimeric scaffold will be shown to depend both on the hydrophobe, leucine, and the number of water molecules hydrogen-bonded to sulfur atoms within an internal cysteine layer that forms a trigonal pyramidal Pb(ii)-S₃ binding site. We will demonstrate that unique A₂B-type 3SCCs can be obtained with a templating Pb(ii) ion either by utilizing a cysteine substitution in a d site, with an alanine placed at the subsequent a site (that is, C-XXX-A), or with an alanine in a d site preceding a cysteine substitution at an a site (that is, A-XXX-C). Reversing the order of the alanine/cysteine modifications from these arrangements (that is, A-XX-C or C-XX-A) or substituting Hg(ii) for Pb(ii) no longer allows access to distinct A₂B type structures. Through crystallographic and computational analysis, we explain the mechanics behind this strategy for heterotrimer formation. We then demonstrate that the addition of a remote Zn(ii)-His₃ site performs catalytic functions without perturbing heterotrimer formation.

Results and discussion

The objective of this study is to prepare parallel, heterotrimeric 3SCCs of the type A₂B by using a metal templating strategy that allows for further elaboration of the peptide assembly at locations remote from the templating metal region. With this aim, we have used three key components: Pb(ii), which is known to form trigonal pyramidal structures within this class of 3SCCs^39–41; a layer of three cysteines incorporated either in an a type or d type layer to sequester the Pb(ii); either Leu or Ala placed at a layer adjacent to cysteine-containing residues to effectuate hydrogen bonding between the cysteine sulfurs and water localized within the hydrophobic cavities of the scaffold. The driving force for attaining the desired A₂B constructs is the energetic competition between optimal packing of hydrophobes, such as Leu, and the stability obtained when water internalized within a hydrophobic cavity, formed with Ala, can hydrogen-bond to cysteine sulfur atoms coordinated to Pb(ii).

X-ray crystal structures of the homotrimeric Pb(ii) constructs provide insight into the formation of heterotrimeric constructs. Lead binds to both C-L12C (C-D) and GC-L16CL30H (GC-B) in endo conformations, meaning that both the Pb(ii) and methane carbons of cysteine are on the same side of the three-sulfur-atom plane (Table 1)³⁹. In the d site (for example, G-D), the Pb(ii) sits toward the N termini, above the cysteine plane, while in the a site (GC-B), the lead is displaced toward the C termini, below the cysteine plane. The introduction of an alanine layer adjacent to the cysteine layer, below a d site or above an a site, does not reorient the Pb(ii) into a more spacious cavity; rather, three water molecules fill the void space allowing hydrogen bonding to each of the cysteine sulfurs (Fig. 1). In the a-site system, GC-L12AL16C (GC-A), water molecules are at hydrogen-bonding distances, ranging from 3.0 to 3.4 Å, but are still too far to interact directly with the Pb(ii). This generates a water cavity that creates a stabilizing network of hydrogen bonds unique to the Ala-containing structure. Stability of the 3SCC is lost on removal of hydrophobic leucine packing in the absence of Pb(ii) (Supplementary Fig. 1), but this hydrogen-bond network probably stabilizes the scaffold in the presence of Pb(ii). When considering cysteines in d sites, GC-L12CL16A (GC-C) also has water present in the cavity left by introducing an alanine residue below the cysteine plane, but in a much different orientation (Fig. 1). Instead of the Cys-S bonding to two different water molecules aligned between each peptide strand, the space left from the Leu to Ala mutation allows for the association of a water 3.4 Å directly below each Cys residue (Fig. 1). Thus, in moving from an a-site Cys to a d-site Cys, a hydrogen-bonding network is still present but the network is constructed differently. The d-site system has fewer sulfur-water interactions and at different angles, explaining the difference in selectivity of the a- and d-site heterotrimers (vide infra). Once again, apo-G-C₃ is destabilized with respect to apo-G-D₃; however, Pb(ii) binding to apo-G-C₃ provides both strong Pb(ii)-S bonds to the aggregate and stabilizing hydrogen bonds, which should both strengthen the aggregate. Given these parent structures, we predicted that heterotrimeric structures could be obtained by balancing the stabilities achieved through knob-in-hole packing of leucines versus stabilizing the water hydrogen bonds to sulfur atoms coordinated to Pb(ii) found in alanine-modified scaffolds.

Fig. 1 |. Ribbon diagrams of Pb(ii)A₃ (PDB: 6EGP) and Pb(ii)C₃ (PDB:6MCD) 3SCCs with coordinated water residues.

a,b, Pb(ii) is shown as a large, dark grey sphere and water residues are shown as small, red spheres. Pb(ii)C₃ is shown from the side (a) and top (b) with the water-Cys distance of 3.4 Å shown as a solid line. c,d, Pb(ii)A3 is shown from the side (c) and top (d) with water-Cys distances of 3.0 Å and 3.4 Å. The difference in the hydrogen-bonding networks of Pb(ii)C₃ and Pb(ii)A₃ are apparent in the position of the water and the number of hydrogen bonds to sulfur atoms formed by each. The region showing Zn(ii) bound to the histidines has been omitted for clarity.

To test this idea, the stabilities of the trimeric combinations of A-D type peptides were determined by quantum mechanics/molecular mechanics (QM/MM) calculations. For both a- and d-site constructs, the most stable scaffold is the 2 Ala:1 Leu heterotrimer (A₂B and C₂D forms). The structures represent the energy minima achieved through a combination of both hydrophobic packing and hydrogen bonding. In agreement with the different hydrogen-bonding networks observed by crystallography, there is a notable difference in the relative complexation energies of the a- and d -site constructs. In a-site constructs, where only one heterotrimer forms selectively, the energy differences between each trimer combination are much smaller than that for the d-site constructs (Table 2). The A₂B heterotrimer requires the least energy to complex, while the non-selective AB₂ heterotrimer complexes with 8.5 kcal mol⁻¹ more energy. The more favoured homotrimer, B₃, complexes with an additional 8.5 kcal mol⁻¹ more energy than AB₂. The 3 Ala (A₃) homotrimer is the least energetically favourable with complexation energy 15 kcal mol⁻¹ higher than the B₃ homotrimer. In the d-site constructs, however, the complexation energy differences are much greater between each scaffold. Again, the 2 Ala:1 Leu (C₂D) 3SCC is the most favourable, while the 1 Ala:2 Leu (CD₂) complexes with 12.9 kcal mol⁻¹ more energy, which is slightly less favourable than AB₂, the a-site 1 Ala:2 Leu complex. Importantly, the formation of the 3 Leu (D₃) trimer has been significantly destabilized in the d site, with a complexation energy 53.9 kcal mol⁻¹ higher than the most stable heterotrimer (C₂D) while the 3 Ala (C₃) trimer is 33 kcal mol⁻¹ higher, comparable to the A₃ trimer. As shown below, this negative design forces full selectivity in both heterotrimers as there is no competing structure with a small enough complexation energy to destabilize the desired assembly.

Table 2 |.

QM/MM calculations of complexation energies, ²⁰⁷Pb NMR chemical shifts and Pb-S₃ pK_a values of all constructs

Construct	Complexation energy (kcal mol−1)a	207Pb NMR shift (ppm)b	pKac
A3	32.0	5,574	10.88 ± 0.03
A2B	0.0	5,597	10.55 ± 0.07
AB2	8.5	5,627, 5,595 (minor)	10.93 ± 0.11
B3	17.0	5,645	11.65 ± 0.04
C3	33.0	5,762	11.23 ± 0.19
C2D	0.0	5,875	11.6 ± 0.2
CD2	12.9	5,803	11.9 ± 0.2
D3	53.9	5,789	11.7 ± 0.5

^aCalculated from GC peptides.

^bDetermined with G peptides.

^cDetermined with T peptides for A and B, and G peptides for C and D.

The energy differences between the a- and d-site complexes are due to the substantial structural differences in the Pb(ii) forms of A₂B and C₂D heterotrimers. Again, both A₂B and C₂D bind lead in endo conformations, positioning the lead below the Cys plane in A₂B and above for C₂D. These coordination environments cause structural changes to the neighbouring mutation sites due to the differences in side-chain orientation and spacing between amino-acid layers (Fig. 2). For A₂B, the B-strand leucine reorients within the cavity to allow the Pb(ii) coordination to remain fully symmetric. C₂D, however, maintains the orientation of the leucine but has a dissymmetric Pb(ii) coordination. These differences in the coordination environments cause tighter packing and greater orbital overlap for the C₂D variant when compared with the neighbouring mutated residues for A₂B and C₂D.

Fig. 2 |. Ball-and-stick models of the Cys and adjacent Ala/Leu layers for the 3 Ala and 2 Ala:1 Leu trimers.

a-d, The hydrogen-bonding interactions of A₃ (a), A₂B (b), C₃ (c) and C₂D (d) with the coordinated water molecules are shown in both side and top views. The Pb(ii) remains fully symmetric in A₂B, while the leucine of the D strand forces the Pb(ii) to be asymmetric in C₂D. This leucine also disrupts the hydrogen bonding to the D strand Cys-S.

The three water molecules present in the Pb(ii) cavity also interact differently with the A and C peptides. In both A₃ and A₂B, the water molecules associate with the Cys-S and the backbone of Ala/Leu layer above and are oriented close to the helical interface. In C₃, however, the water molecules are oriented towards the interior of the helix and the hydrogen bond with each other and the Cys-S. In C₂D, the leucine residue disrupts this hydrogen bonding, resulting in the loss of a hydrogen bond to the D strand Cys-S. The stabilization of B₃ relative to A₃ is due to these water molecules interacting only with the side chains and backbones of the Cys residues and orienting closer to the helical interface. As a result, this structure contains more hydrogen bonds and provides more space to the bulky Leu residues than A₃. The side chains of the Leu residues also minimize the steric clash by pointing down or to the exterior of B₃, making it more energetically favourable than A₃ (Supplementary Fig. 2). In contrast, the water molecules of both C₃ and D₃ are oriented more towards the interior of the helix and hydrogen-bond to each other and the Cys side chains. However, in D₃, these water molecules create a layer between Cys and Leu that sterically constrain the side chains of the Leu residues. Consequently, all 3 Leu residues are closely packed and their side chains are oriented towards the interior in this structure. This arrangement can destabilize the formation of D₃ over C₃ and further favours heterotrimer formation via negative design.

Given the reported X-ray structures of the relevant homotrimers and the subsequent computations for the relative energies of different mixed peptide species in the presence of Pb(ii), we next considered whether one obtains exclusively the desired A₂B or C₂D heterotrimeric scaffolds experimentally. For the purposes of these studies, both T and G peptides were examined, giving comparable results, so only the G systems will be described in detail. First Pb(ii) binding was tested for all homo- and heterotrimeric combinations using UV-vis spectroscopy. Spectra were obtained yielding the classic signatures for Pb(ii)S₃ chromophores (Supplementary Fig. 3). Unfortunately, there are only small shifts in the UV band at ~340 nm, so determination of the presence and ratios of different stoichiometric assemblies of peptides could not be determined. A more sensitive measure of species distribution can be obtained using ²⁰⁷Pb NMR, which has a large chemical shift range, with Pb-S₃ signals spread over greater than 300 ppm (ref.⁴¹). Using this highly sensitive technique, Pb(ii) binding to binary mixtures of A and B peptides (or C and D) was investigated to assess whether pure A₂B or C₂D species could be formed.

Homotrimeric peptides containing a-site Cys (Pb(ii)A₃ or Pb(ii)B₃) are separated by ~75 ppm, with pure Pb(ii)B₃ at 5,645 ppm and pure Pb(ii)A₃ at 5,574 ppm (Fig. 3). The 2:1 mixtures of these peptides have chemical shifts evenly spaced within the range formed by these homotrimers. Pb(ii)A₂B has a resonance at 5,597 ppm, 25 ppm downfield of Pb(ii)A₃. Pb(ii)AB₂ has two resonances. The major peak at 5,627 ppm is centred 25 ppm between the Pb(ii)A₂B and Pb(ii)B₃ assemblies. The minor resonance at 5,595 ppm corresponds to the Pb(ii)A₂B peptide. Thus, formation of the Pb(ii) A₂B is fully selective, while the 1 Ala:2 Leu heterotrimer is biased (we estimate ~80%) towards formation of the desired heterotrimer but lacks full selectivity. The low calculated energy gap between the desired and non-desired heterotrimer in this mixture explains the lack of selectivity. The non-desired Pb(ii)A₂B complex is still able to form, but in small amounts due to the ratio of peptides in solution.

Fig. 3 |. ²⁰⁷Pb NMR of T and G peptides with a-site Cys residues.

Left: ²⁰⁷Pb NMR of T peptides A₃ (1), A₂B (2), AB₂ (3) and B₃ (4), showing the full selectivity of T-A₂B. Right: ²⁰⁷Pb NMR spectra of G peptides A₃ (1), A₂B (2), AB₂ (3), B₃ (4) and A₅B (5) with 2 equiv. of ²⁰⁷Pb(ii), showing the full selectivity of G-A₂B as well as the coexistence of G-A₃ and G-A₂B to demonstrate the presence of asymmetric scaffolds rather than coalescence of symmetric signals.

An alternative explanation for these observations is that the ²⁰⁷Pb NMR is simply reporting a chemical shift for an average distribution of species that are coalescing to a single resonance. The two peaks seen for the Pb(ii)AB₂ conditions speak against this possibility; however, a more rigorous test is to perturb an existing signal by addition of one of the peptide components. Figure 3 shows how the spectra of a pure Pb(ii)A₂B system is altered by addition of more A peptide to yield a stoichiometry of 5 A and 1 B peptide strands in the presence of two equivalents of lead. If coalescence were responsible for the observed intermediate signals, one would expect a single resonance shifted 12 ppm downfield of the Pb(ii)A₂B peak, corresponding to a mixture that is 5/6 one homotrimer and 1/6 the other homotrimer. Instead, two peaks are observed that correspond to the resonances (5,597 ppm and 5,574 ppm) expected for Pb(ii)A₂B and Pb(ii)A₃ existing in a slow exchange equilibrium on the NMR timescale.

The d-site Cys-containing peptides also showed unique resonances for heterotrimeric mixtures of peptides, with all d-site resonances shifted significantly downfield (Supplementary Fig. 4). The homotrimeric peptides (in this case Pb(ii)C₃ and Pb(ii)D₃) are separated by only 30 ppm, with Pb(ii)C₃ at 5,762 ppm and Pb(ii)D₃ at 5,789 ppm. Interestingly, the resonances from heterotrimeric peptide mixtures fall downfield of the range of these two homotrimers. Pb(ii)CD₂ is ~15 ppm downfield of Pb(ii)D₃, at 5,803 ppm. Pb(ii) C₂D is the most downfield-shifted at 5,875 ppm. As these resonances cannot have occurred from a mixture of homotrimeric signals, and because single peaks are observed in both cases, they must both correspond to fully selective heterotrimers. Heterotrimers therefore represent the peak stability from a mixture of hydrophobic packing of leucine residues and hydrogen bonding present from alanine residues when a d-site cysteine is employed. This is a consequence of the steep decrease in stability of the corresponding d-site homotrimers compared to the heterotrimeric scaffolds (Table 2). In the a-site constructs, full selectivity is only observed for Pb(ii)A₂B, indicating that only this mixture has an energy minimum significant enough for complete formation of the desired heterotrimer.

UV spectra monitoring the pH titration of homo- and heterotrimeric systems provides further support for heterotrimer formation and illustrates the change in chemical properties associated with alanine substitution. The equilibrium examined is the release of two protons, in a single step, to convert a Pb(ii)S(SH)₂ peptide into the fully complexed Pb(ii)S₃. Using T-L12AL16C (T-A) and T-L16C (T-B) for the series of Pb(ii)A_nB_{3 −n}, both heterotrimers have lower effective pK_a values than the calculated 2:1 mixtures of homotrimers (Fig. 4 and Table 2), demonstrating that a structure distinct from the homotrimers exists. The lower pK_a (10.55) for Pb(ii)A₂B illustrates that this species most easily adapts the preferred Pb(ii)S₃ structure. Interestingly, when Pb(ii)C_nD_{3 −n} peptides are titrated, the complexes containing leucine have pK_a values approximately equivalent to Pb(ii)D₃, while Pb(ii)C₃ has a slightly lower pK_a (Supplementary Fig. 5).

Fig. 4 |. pH titration curves of Pb(ii)A_nB₃ − n scaffolds at λ_max for homo- and heterotrimers.

Symbols are data points and solid lines are fits to the equation in the Supplementary Information showing the acid-shifted pK_a values for heterotrimers based on the calculated ratios of homotrimeric signals. Pb(ii)A₃ is represented by open squares, Pb(ii)A₂B by open circles, Pb(ii)AB₂ with × symbols and Pb(ii)B₃ with + symbols. pK_a values, calculated for each fit assuming an equilibrium Pb(ii)(H₂A_nB_{n − 3}) → Pb(ii) A_nB_{n − 3} + 2H⁺, are provided in Table 2.

Mercury(ii) has high affinity for thiolate ligation and will form trigonal planar structures in these 3SCCs^18,39. Nevertheless, selective formation of Hg(ii)A₂B type heterotrimers cannot be achieved, as demonstrated by ¹⁹⁹Hg NMR of these systems (Supplementary Fig. 6). This is probably a consequence of the different numbers of waters in an alanine-generated cavity and the different hydrogen-bonding patterns these molecules form with the bound sulfur atoms. This also suggests that the apopeptides will not segregate in the absence of metal into these heterotrimeric assemblies, indicating that a trigonal pyramidal-type metal is essential. Additionally, positioning the alanine below an a-site cysteine does not lead to selectivity (Supplementary Fig. 7), indicating that proper water location relative to the lead is necessary for heterotrimer formation.

The addition of Zn(ii) to the remote binding His₃ site of G-A and G-B (at the C-terminal end of the peptide) does not affect metal binding to the Cys₃ site, nor does it change the stoichiometry of the desired 3SCC (Supplementary Fig. 8)¹¹. Therefore, this retention of A_nB_{3 − n} stoichiometry in the presence of a remote, catalytic metal ion allows for the study of catalysis in asymmetric coordination environments. The formation of heterotrimers was also found to be pH-independent, in the presence and absence of Zn(ii), from 7.5 to 9.5, allowing for studies of p-nitrophenyl acetate (pNPA) hydrolysis at pH 9.5 (Supplementary Fig. 9). Lengthening the peptide from T to G resulted in a decrease in catalytic efficiency (~20%) of pNPA hydrolysis (Table 3) for the control G-B₃ homotrimer due to the increased stability of the helical bundle⁴², which decreased substrate access. The addition of a threonine residue to G-B in the d layer above the His residue was done to model T199 in native CAII. The homotrimer Pb(ii)Zn(ii)G-B_T3 decreased the catalytic efficiency by a further 20%. We believe that this decreased activity is due to the methyl group of threonine rotating to the centre of the bundle and the hydroxy group pointing towards the helical interface (Supplementary Fig. 10). By including only a single T in a heterotrimeric Pb(ii)Zn(ii)G-A₂B_T, the rate increases to a value exceeding the Pb(ii)Zn(ii)G-B_T3 by ~30% and is superior to the parent Pb(ii)Zn(ii)G-B₃ homotrimer. This increase shows the modest, but significant effect that asymmetry can have on catalysis. To combat the complication of side-chain orientation with T residues, we instead examined the effect of aspartate coordination on catalysis. We reasoned that the aspartate side chain was too short to directly bind to the Zn(ii) when three histidines were coordinated to the metal, but would provide an oxygen that could serve as an acid-base catalyst, similar to T199. With three additional Asp residues Pb(ii) Zn(ii)G-B_D3, a 25% increase in catalytic efficiency is seen over Pb(ii) Zn(ii)G-B₃. Although a slight, though statistically insignificant, increase is seen with just a single Asp residue Pb(ii)Zn(ii)G-A₂B_D as compared to Pb(ii)Zn(ii)G-B_D3, both of these peptides match or exceed the catalytic efficiency of the original Pb(ii)Zn(ii)T-B₃ and are over 40% better than Pb(ii)Zn(ii)G-B_T3. Although only modest changes in catalytic efficiency are illustrated with these examples, one can imagine larger effects with more drastic modifications that could be achieved in future designs.

Table 3 |.

Kinetic values for pNPA ester hydrolysis at pH 9.5 for Pb(ii) and Zn(ii) bound peptides

Peptide	kcat/KM (M−1 s−1)	kcat (s−1)	KM (mM)
T-B3a	23.3 ± 0.3	0.040 ± 0.012	1.7 ± 0.5
G-B3	18.5 ± 0.6	0.064 ± 0.007	3.5 ± 0.5
G-BT3	14.8 ± 1.4	0.056 ± 0.026	3.8 ± 2.1
G-A2BT	20.6 ± 2.1	0.030 ± 0.007	1.5 ± 0.7
G-BD3	24.2 ± 2.5	0.024 ± 0.004	1.0 ± 2.0
G-A2BD	25.1 ± 2.5	0.021 ± 0.003	0.8 ± 0.2

^aFull-sequence T L9CL23H¹¹.

We have presented a novel approach to achieve selectivity for supramolecular assembly in designed proteins by combining three distinct features: (1) a cysteine sulfur layer for metal coordination, (2) a thiophilic, trigonal pyramidal metalloid (Pb(ii)) that binds to these sulfurs and (3) an adjacent layer of reduced steric bulk within the hydrophobic core allowing a cavity that contains waters that hydrogen-bond to the cysteine sulfur atoms. When combined properly, peptides with a-site cysteine substitution yield A₂B heterotrimers exclusively, and when cysteine is in the d site, either pure C₂D or CD₂ scaffolds are obtained. Altering the metal from Pb(ii) to Hg(ii) or shifting the relative position of the sterically less demanding side chain removes heterotrimer selectivity. These studies illustrate the importance of solvent-based molecules for stabilizing protein-protein interactions that define the appropriate scaffold assembly. Finally, incorporating a bulky side chain, such as histidine, at a remote location within the coiled coil, and even binding a metal such as Zn(ii) at that site, does not perturb the ability to obtain the desired heterotrimers. Thus, these studies provide a unique strategy to prepare dissymmetric catalytic sites within self-assembling de novo designed proteins. Further studies will explore the behaviour of metals in these remote sites and examine whether this strategy can be extended to other coiled coils of different stoichiometries.