Proteomimetic Zinc Finger Domains with Modified Metal-binding β-Turns

Abstract

The mimicry of protein tertiary folds by chains artificial in backbone chemical composition leads to proteomimetic analogues with potential utility as bioactive agents and as tools to shed light on biomacromolecule behavior. Notable successes toward such molecules have been achieved; however, as protein structural diversity is vast, design principles must be continually honed as they are applied to new prototype folding patterns. One specific structure where a gap remains in understanding how to effectively generate modified backbone analogues is the metal-binding β-turn found in zinc finger domains. Literature precedent suggests several factors that may act in concert, including the artificial moiety used to modify the turn, the sequence in which it is applied, and modifications present elsewhere in the domain. Here, we report efforts to gain insights into these issues and leverage these insights to construct a zinc finger mimetic with backbone modifications throughout its constituent secondary structures. We first conduct a systematic comparison of four turn mimetics in a common host sequence, quantifying relative efficacy for use in a metal-binding context. We go on to construct a proteomimetic zinc finger domain in which the helix, strands, and turn are simultaneously modified, resulting in a variant with 23% artificial residues, a tertiary fold indistinguishable from the prototype, and a folded stability comparable to the natural backbone on which the variant is based. Collectively, the results reported provide new insights into the effects of backbone modification on structure and stability of metal-binding domains and help inform the design of metalloprotein mimetics.

Graphical Abstract

Introduction

In proteins, amino acid sequence encodes function of a given chain by leading that chain to adopt a precise 3-dimensional folded shape. The structure of proteins is defined by a hierarchy where primary sequence of covalently connected amino acids enables the formation of locally ordered motifs (secondary structure), arrangement of these motifs into defined unimolecular folds (tertiary structure), and the association of multiple folded chains (quaternary structure). Due to their essential role in life and countless possible functions, proteins are appealing targets for mimicry by designed synthetic agents; like natural proteins, these artificial protein mimetics exist in a spectrum of structural complexity. Research on peptidomimetic analogues of primary and secondary structure is a vibrant field that now spans decades.^[1-2] In recent years, chemists have begun to address the question of how to mimic protein architectures at the level of tertiary structure and beyond; the successful development of these proteomimetic agents has been driven by complementary strategies including modification of typical polypeptide chain topology and backbone composition.^[3]

One strategy applied to generate artificial peptidic oligomers that adopt complex folding patterns is the systematic alteration of backbone covalent connectivity in biologically derived sequences.^[4] The result of such modification, when made judiciously, is a heterogeneous backbone in which artificial building blocks are interspersed among natural α-peptide, yet the resulting chain is able to reproduce the sophisticated fold of a given prototype sequence. Such heterogeneous-backbone protein mimics have been used to probe assembly behavior in amyloidogenic sequences,^[5] reproduce diverse tertiary folding patterns,^[6-11] and create agents capable of native-like molecular recognition.^{[6, 8,11]} As various classes of artificial monomers have different conformational propensities, creating heterogenous-backbone mimics of complex and diverse folds found in nature relies on collective application of many different types of building blocks. The choice of which building block to use in a given sequence and where is guided by precedent involving peptidomimetics^[1-2] as well as isolated modifications in larger proteins.^[12]

Design rules used to construct heterogeneous-backbone proteomimetic tertiary structures described to date have proven general to a number of different prototype sequences and folds, but there are exceptions. Gaining insights into the origin of unsuccessful designs is an important area of inquiry, as it leads to more robust and generalizable strategies applicable toward mimicry of a larger fraction of the myriad tertiary folds found in nature. We encountered one such case in a recent effort to develop proteomimetic analogues of a metalloprotein domain, zinc finger domain 3 from the transcription factor Sp1 (Sp1-3).^[9] Backbone modifications could be introduced throughout the helix and strands in Sp1-3 without compromising folded structure or stability; however, supplementing these with any change in composition at the turn near the zinc binding site resulted in a variant unable to fold or bind metal.^[9] This result was puzzling, as the artificial moieties attempted for modification of the turn (D-Pro, aminoisobutyric acid [Aib], side-chain linked ornithine [δ-Orn]) were all well-precedented as turn mimetics in short peptides^[13-17] as well as larger proteins.^{[7, 10-11, 18-19]}

While D-Pro, Aib, and δ-Orn are well known turn mimetics, none of that precedent involved modification of a β-turn that was also a binding site for a metal ion (as is the case in Sp1-3). In nature, metal binding β-turns are found in proteins involved in calcium binding^[20] as well as heavy metal homeostasis;^[21] however, their most common roles are as structural elements in zinc finger domains.^[22-23] In addition to facilitating local reversal of the backbone, such turns must precisely position nearby side chains in order to create a geometry compatible with high-affinity metal coordination. Among the broader literature on modification of β-turn backbone composition in peptides and proteins, only one other example involved a metal-binding context. A β-turn dipeptide analogue (BTD), another well-precedented turn mimetic,^[24-25] was incorporated in zinc finger domain 3 from the protein YY1 (YY1-3) and shown to be structurally accommodated;^[26] however, the effects on folded stability were not assessed. The qualitatively different observations from the Sp1-3 and YY1-3 systems in terms of their tolerance to modification of the metal-binding turn suggested several possible factors at play: (1) the artificial moiety being incorporated in the metal-binding turn (i.e., BTD vs. D-Pro, Aib, δ-Orn), (2) the sequence context (i.e., Sp1-3 vs. YY1-3), or (3) the density of artificial building blocks in the modified backbone (i.e., isolated to the turn vs. incorporated throughout the domain). Here, we report our efforts to gain insights bearing on the above hypotheses and leverage these insights to construct a proteomimetic zinc finger analogue with backbone modifications throughout its constituent secondary structures and high folded stability. These results have implications related to a broader field of researchers working to develop strategies toward artificial backbone analogues of metal-binding peptides and proteins.^[27-32]

Materials and methods

Peptide synthesis and purification.

Peptides (1-6) were synthesized via microwave assisted Fmoc solid phase methods on a CEM MARS 5 microwave. NovaPEG rink amide resin (0.05 mmol) was allowed to swell in DMF overnight. In a standard coupling step, Fmoc-protected amino acid (4 equiv) was activated by treatment with HCTU (3.9 equiv) and DIEA (6 equiv) in NMP (2 mL) for 2 min. The activated monomer solution was added to resin and the reaction heated to 70 °C with a 2 min ramp to the target temperature followed by a 4 min hold. Modifications were made to the standard above method in the following cases. Coupling of Fmoc-His(Trt)-OH or Fmoc-Cys(Trt)-OH was carried out at room temperature for 40 min. Coupling of Aib and residues immediately following Aib and N-Me residues used PyAOP (4 equiv) in place of HCTU. Coupling of Fmoc-BTD-OH (sourced from ChemImpex) was performed twice at room temperature for 40 min (3 equiv Fmoc-BTD-OH, 3 equiv PyAOP, 4.5 equiv DIEA for the first reaction followed by 2 equiv Fmoc-BTD-OH, 2 equiv PyAOP, 3 equiv DIEA for the second reaction). In peptide 3, every residue subsequent to the introduction of BTD was coupled twice (5 equiv Fmoc-amino acid, 5 equiv PyAOP, and 7.5 equiv DIEA). Fmoc deprotections were carried out using 20% v/v 4-methyl piperidine (2 mL) in DMF at 80 °C with a 2 min ramp to the target temperature followed by a 2 min hold. After each coupling and deprotection step, the resin was washed 3 times with DMF. After the final Fmoc deprotection, the resin was washed 3 times each with DMF, DCM and MeOH and was dried under vacuum for 30 min. Peptides were cleaved from resin by treatment with TFA/H2O/EDT/thioanisole/phenol (87 : 3.5 : 2.5 : 3.5 : 3.5 v/v/v/v/w) for 4 h. Crude peptides were precipitated by addition of cold diethyl ether to the filtered cleavage mixture, and the resulting pellets collected by centrifugation. Peptides were purified by preparative RP-HPLC on a Phenomenex Jupiter C₁₈ column using gradients between 0.1% TFA in water (solvent A) and 0.1% TFA in acetonitrile (solvent B). Pellets were dissolved in HPLC solvents (80% solvent A, 20% solvent B) prior to purification. Pure fractions were combined and lyophilized. The identity of each peptide was confirmed by MALDI-TOF MS (Table S1; Figures S6-S11) using super dihydrobenzoic acid as the matrix. Purity was confirmed by analytical HPLC (Figure S12). Peptides were stored as lyophilized powders under vacuum to minimize contact with oxygen.

Isothermal Titration Calorimetry (ITC).

A Malvern Microcal iTC200 instrument was used for ITC experiments. All solutions were purged with argon and experiments were performed at 25 °C. Concentrations of peptide stock solutions were determined by Ellman’s assay in 6 M guanidine hydrochloride, 0.1 M sodium phosphate buffer at pH 8.6 (ε₄₁₂ = 13,700 M⁻¹ cm⁻¹ for reduced Ellman’s reagent). Concentrations of zinc chloride stock solutions were determined by colorimetric methods using Zincon.^[33] To the sample cell, 300 μL of solution containing 150 μM peptide, 50 mM HEPES, 50 mM NaCl at pH 7.4 was added. The syringe was filled with a stock solution of ZnCl₂ (0.9-1.2 mM) in 50 mM HEPES, 50 mM NaCl, pH 7.4. Each titration experiment consisted of 18 injections of 2 μL volume each. Data were analyzed using the Origin software supplied with the instrument. Heats of zinc dilution obtained from control experiments were subtracted from the experimental data and the data was then fit to a 1:1 binding model to obtain the binding enthalpy (ΔH_ITC). Errors reported for ΔH values in Table 1 are the parameter uncertainties from the fits; errors for ΔΔH were determined by error propagation.

Table 1.

Thermodynamic parameters for the interaction of peptides 1-6 with Zn²⁺.^a

#	β-turn	ΔHITC (kcal mol−1)	ΔΔH relative to 1 (kcal mol−1)
1	Ser-Tyr	−18.4 ± 0.4
2	Aib-Gly	−23.4 ± 0.2	−5.0 ± 0.4
3	BTD	−23.0 ± 0.4	−4.6 ± 0.6
4	d-Pro-Gly	−19.3 ± 0.3	−0.9 ± 0.5
5	δ-Orn	−19.5 ± 0.3	−1.1 ± 0.5
6	Aib-Glyc	−22.2 ± 0.2	−3.8 ± 0.4

^aDetermined by ITC experiments performed in 50 mM HEPES, 50 mM NaCl, pH 7.4 at 25°C.

^bIn addition to the turn, peptide 6 also contains backbone modifications in the hairpin and helix.

NMR spectroscopy.

All NMR experiments were performed on Bruker Avance 700 MHz spectrometer at 298 K. A solution was prepared of lyophilized peptide (1.2-1.5 mM) in 10 mM deuterated Tris buffer in 9:1 H2O:D2O with 0.05 mM DSS. The pH was adjusted to 7.5, ZnCl₂ (1.5 equiv from a stock solution in water) was added, and the pH adjusted to 6.0. NOESY spectra were acquired with 512 t₁ increments and 4096 data points in the direct dimension. COSY and TOCSY spectra were acquired with 512 t₁ increments and 2048 data points in the direct dimension. The excitation sculpting method was used to suppress water signal, and parameters O1, P1 and SP1 were manually optimized before each set of experiments. Mixing times were 200 ms for NOESY experiments and 80 ms for TOCSY experiments. The relaxation delay (D1) was set to 3 seconds, and SPNAM1 was set to Sinc1.1000 for water suppression. Raw FIDs in both dimensions were multiplied by a phase-shifted sine bell function, zero filled, and Fourier-transformed to yield 2048 by 2048 matrices. All spectra were processed using Topspin and chemical shifts referenced to internal DSS. For H/D exchange experiments, NMR samples described above were lyophilized, reconstituted in D2O, and monitored by ¹H NMR. The first spectrum, acquired ~15 min after sample preparation, was considered as t = 0. Integrated peak area (I) for signals of interest were normalized to the DSS methyl peak and plotted as a ratio to the corresponding peak area at t = 0 (I_o). The resulting time-dependent exchange was fit to a single-phase exponential decay using GraphPad prism to obtain the reported half-life values.

NMR structure determination and analysis.

Resonance assignments for each peptide were performed by standard methods using NMRFAM-SPARKY.^[34] A minor conformer in slow exchange on the NMR time scale was observed in the native domain and variants 1-5 corresponding to the cis isomer of the Arg¹-Pro² amide. Structure determination for those variants, detailed below, was conducted for the major trans amide species. NMR structures were calculated by simulated annealing with experimental restraints using the software package ARIA (Ambiguous Restraints for Iterative Assignment, version 2.3)^[35] in conjunction with CNS (Crystallography & NMR System, version 1.2).^[36] Parameter and topology definitions for unnatural residues were based on analogous atom types for canonical α-residues. Settings for each ARIA run were modified from program defaults, as previously described.^[37] Backbone φ dihedral restraints were generated based on ³J_Hα-HN coupling constants measured from resolved amide doublets in the 1D ¹H NMR spectrum (φ = −65° ± 25° for J ≤ 6.0 Hz and φ = −120° ± 40° for J ≥ 8.0 Hz). Restraints were included for tetrahedral Cys₂His₂ coordination of Zn²⁺ (2.3 Å for S_γ-Zn, 2.0 Å for N_ε-Zn).^[38] NOE-based distance restraints were generated automatically by ARIA in iterative fashion over the course of the calculation, starting from an unassigned set of integrated NOESY peaks and a list of ¹H resonance assignments. The ARIA structure calculation for each peptide was performed in two sequential runs. In the first run, a set of H-bond restraints for the helical segment was generated based on qualitative inspection of the NOESY spectrum for i→i+3 H_α→H_β and/or H_α→H_N correlations. As published structures of the prototype domain (PDB 1SP2, 1VA2) show tighter winding of the helix near the C-terminus,^[39-40] H-bond restraints for this region were omitted at this stage. The ensemble of 10 lowest energy structures resulting from the first run was used as the input structure for a second ARIA run. Parameters for the second run were the same as above, save the addition of new restraints for H-bonds observed in the first ensemble. The final set of 10 lowest energy structures from the second run was taken as the NMR ensemble for that peptide (Table S2). Ensemble coordinates and supporting experimental data are deposited in the PDB (1: 6UCP, 2: 6PV1, 3: 6UCO, 4: 6PV0, 5: 6PV2, 6: 6PV3) and BMRB (1: 30674, 2: 30642, 3: 30673, 4: 30641, 5: 30643, 6: 30644).

Results and Discussion

Impact of β-turn modification on folded structure and thermodynamic stability of a zinc finger domain.

Both Sp1-3 and YY1-3 are canonical Cys₂His₂ zinc fingers, defined by the consensus sequence (F/Y)-X-C-X_2-5-C-X₃-(F/Y)-X₅-ψ-X₂-H-X_3-5-H (metal binding residues underlined; X and ψ represent any amino acid and a hydrophobic residue, respectively).^[41] The length of the loop between the metal-binding Cys residues, which contains the ββα motif β-turn, is variable; two-and four-residue segments are most common. The length of this inter-Cys loop is four residues in YY1-3 and two residues in Sp1-3 (Figure 1). We hypothesized this difference might affect the ability of these two domains to tolerate modification in the β-turn. In order to test this hypothesis, we selected a third sequence—zinc finger 2 of the Sp1 protein (Sp1-2)—as a host for evaluation of artificial turn inducers in a metal-binding structural context. Sp1-2 is highly similar to Sp1-3 in both sequence and structure (48% sequence identity, 1.2 Å C_α rmsd);^[39-40] however, Sp1-2 bears an extended 4-residue hairpin loop as found in YY1-3. We designed host peptide 1, based on Sp1-2, and four variants (2-5) in which different artificial turn inducers (Aib-Gly, BTD, D-Pro-Gly, and δ-Orn) were incorporated in place of residues Ser⁸-Tyr⁹, at positions i+1 and i+2 of the metal-binding β-turn in the native sequence.^[39] Our goal in preparing and characterizing 1-5 was to understand the optimal turn modification with respect to recreating the native folded structure while maintaining or improving on folded stability. We synthesized peptides 1-5 by Fmoc solid phase methods, purified the products by preparative reverse phase HPLC, and confirmed identity and purity of isolated material by analytical HPLC and mass spectrometry. In both the natural backbone prototype and variants, Met⁴ was replaced by norleucine (Nle) to prevent complications arising from oxidation during synthesis, purification, and storage.

Figure 1.

(A) Comparison of the NMR structures of zinc finger domains Sp1-2 (PDB 1SP2) and Sp1-3 (PDB 1SP1).^[40] The metal-binding β-turn is boxed in the overlay on the left and shown in zoomed views from the individual structures on the right. (B) Sequences of Sp1-2, Sp1-3, YY1-3, and peptides 1-5 alongside chemical structures of unnatural amino acid residues used in variants 2-5. In the sequences, metal-coordinating Cys and His residues are underlined and positions i+1 and i+2 in the metal-binding β-turn are highlighted in gray; B = norleucine. BTD and δ-Orn are treated as dipeptides in sequence numbering.

We subjected prototype zinc finger 1 and turn-modified variants 2-5 to NMR spectroscopy to assess the impact of backbone modification on folded structure. We acquired homonuclear ¹H/¹H NOESY, TOCSY and COSY spectra for each sequence (9:1 H₂O/D₂O, 10 mM deuterated Tris, pH 6), assigned proton resonances, and determined high resolution structures from simulated annealing with NMR-derived restraints. The ensemble obtained for peptide 1 (Figure 2) showed excellent agreement with previously reported NMR structures of the native Sp1-2 sequence (Figure S1),^[39-40] confirming the innocuous nature of the Met⁴→Nle substitution. The data and structure for 1 were used as the basis for comparison with the turn-modified variants. With the exception of residues at or near the turn, H_α chemical shifts for the variants were virtually identical to those for 1 (Figure S2). Comparison of the high-resolution structures for the variants shows that each adopts an overall fold similar to the native backbone (Figure 2A). Subtle structural differences were apparent in the metal-binding hairpin loop (Figure 2B), which we sought to better understand through additional analyses detailed below.

Figure 2.

NMR structures of parent peptide 1 and turn variants 2-5. (A) Ensemble of 10 lowest energy structures for each sequence. (B) Zoomed view of the metal-binding β-turn from the lowest energy conformer from each ensemble. BTD and δ-Orn are treated as dipeptides in sequence numbering.

We first categorized the turn type present in each metal-binding loop based on the backbone dihedrals at positions i+1 and i+2 (Figure S3).^[42] Because the backbone chemical structure of δ-Orn deviates from canonical polypeptide, its turn type cannot be categorized, and it was omitted from this analysis. The β-turn in prototype peptide 1 was type II, as noted previously for native Sp1-2.^[39-40] The Aib-Gly motif in 2 also favored a type-II turn, while the d-Pro-Gly and BTD turn replacements in 3 and 4 adopted mirror image type-II′ turns. The turn type preferences for the artificial turn mimetics seen in Sp1-2 are in accord with prior observations in non-metal-binding proteins,^{[7, 10, 18-19, 25]} suggesting the sequence context is not influencing their behavior.

We next sought a residue-level comparison of the NMR structures between the parent domain and variants. Our goal was a quantitative assessment of each turn mimetic for its ability to recreate a native like structure in the metal-binding loop. In order to account for uncertainty inherent to structures determined by NMR, we applied reported methods to make use of full ensembles in the analysis rather than a single representative structure from each.^[43] For each pairwise comparison (Figure 3), inter-ensemble backbone rmsd along the sequence after overlay is plotted alongside the corresponding intra-ensemble rmsd values for the two peptides involved in the comparison. Localized increases in intra-ensemble rmsd values compared to inter-ensemble rmsd in the same region indicates a localized difference between sequences that is significant with respect to structural uncertainty in the NMR ensembles. Structural differences among the variants were isolated to the metal-binding turn (Figure 3). The magnitude of the local distortion from the parent domain varies with the identity of the artificial turn moiety. The Aib-Gly turn was most native-like, while the δ-Orn turn led to the greatest distortion in the metal binding loop. Interestingly, differences among the turn mimetics were also apparent in the propagation of the local backbone distortion beyond the central (i+1, i+2) sites of the turn to metal-binding Cys¹⁰ at the i+3 position. In the case of the d-Pro-Gly and δ-Orn turn variants, the inter-ensemble rmsd at this site deviates considerably from the corresponding intra-ensemble values.

Figure 3.

Per-residue backbone rmsd values from the overlay of the NMR ensembles for turn variants 2-5 with that of parent peptide 1. For each pairwise comparison, the corresponding intra-ensemble rmsd plots for the two individual sequences are shown. The four-residue β-turn is highlighted in gray, and the rmsd value for Cys¹⁰ at position i+3 labeled. Data for the disordered termini (residues 1, 29-31) are not shown.

Protein backbone alteration can have profound effects on backbone dynamics, both in the folded and unfolded states.^{[18-19, 44]} Chemical shift effects attributable to long-range tertiary contacts provide one measure of such dynamics in the folded state and have been applied in the study of zinc finger domains.^[45-46] In the metal-binding loop of Sp1-2, the indole of Trp⁷ (position i of the four-residue β-turn) packs against the side chain of His²⁷, resulting in a significant upfield shift in one of the His²⁷ H_β signals (δ = 0.99 ppm for peptide 1). In each of the variants, this His²⁷ H_β signal shows a similar upfield shift; however, the shielding at this position is attenuated in the δ-Orn turn variant (δ = 1.55 ppm) compared to the others (δ = 0.97-1.07 ppm). Corresponding differences in the packing interactions between Trp⁷ and His²⁷ are not apparent in the NMR ensemble, suggesting the possibility of altered dynamics in the turn region. To gain some additional data bearing on this hypothesis, we measured H/D exchange rates for each sequence. In the case of parent peptide 1 and δ-Orn variant 5, no backbone amide signal was visible 15 min after reconstitution of lyophilized material in deuterated buffer. This result suggests the possibility of significant dynamics not apparent in the NMR structure ensembles—behavior that has been noted previously in some zinc finger domains.^[45-46] The high intrinsic flexibility of the parent sequence 1 prevents an incisive test of the hypothesis that the δ-Orn turn enhances that flexibility from these data. Modification with Aib-Gly, d-Pro-Gly, or BTD in variants 2-4 led to protection of several backbone amides in the hairpin, indicating these turn promoting moieties had an overall rigidifying effect on the domain or that subtle local structural changes in the turn led to protection from solvent. Potentially most informative was the impact on Phe¹⁴ (Figure S4), which is removed from the metal-binding turn, forms an inter-strand hydrogen bond with Phe³, and has a side chain packed in the hydrophobic core. The Phe¹⁴ amide proton showed an exchange half-life of almost 1 hour in Aib-Gly variant 2 and d-Pro-Gly variant 4. The protection at this site afforded by BTD in the turn was more modest, with a half-life of ~10 minutes.

Collectively, the above NMR results show that all four artificial backbone moieties examined lead to a Sp1-2 zinc finger domain variant that adopts a tertiary fold that is qualitatively similar to the prototype native backbone. The fact that three of the four turn motifs examined failed to support a native-like fold when incorporated in the closely related domain Sp1-3 supports the hypothesis that hairpin loop length in a Cys₂His₂ ββα zinc finger sequence is an important determinant of its ability to tolerate modification in the metal-binding turn. Domains with four-residue loops between metal-binding Cys side chains can readily accommodate such changes, while those in which the Cys residues are at positions i and i+3 of the turn do not. The NMR results also indicate differences among the various turn mimetics employed in the structure and dynamics of the fold bearing the modified β-turn. These observations suggest potential impacts on folding thermodynamics; however, the NMR results do not bear directly on this question. Understanding the effect of backbone modifications on the stability of a domain is imperative if such modifications are to be used in construction of proteomimetics with highly modified backbones. Even slightly destabilizing substitutions can easily overwhelm the modest folding free energy of a typical small protein when incorporated at high densities (i.e., 20-30%) in heterogeneous backbones.^[7]

In order to quantify the impact of backbone alteration in the metal-binding turn on folding energetics, we performed isothermal titration calorimetry (ITC) experiments of peptides 1-5 with Zn²⁺. Metal binding is tightly coupled to folding in zinc finger domains, and the thermodynamics of the overall process has two components: metal-peptide binding and peptide folding.^[47] For sequences with identical metal-coordinating residues measured under identical experimental conditions, differences in the overall measured energetics measured by ITC are dominated by changes to folding thermodynamics.^[48] Thus, ITC results for peptides 1-5 obtained in 50 mM HEPES, 50 mM NaCl, pH 7.4 at 25 °C (Figure S5, Table 1) show a number of trends that are informative with respect to understanding the effects of backbone modification on folding. As analysis of zinc binding affinities by ITC can in some cases lead to underestimation of the true association constant,^[23] we limited our analysis to comparing enthalpies among prototype peptide 1 and turn-modified variants 2-5. The enthalpy determined by ITC ΔH_ITC) was measured under identical experimental conditions, peptides 1-5 have highly similar folded structures (vide supra), and the metal-binding residues in the zinc finger are the same (Cys₂His₂) across the series. Thus, enthalpies associated with protonation of buffer, deprotonation of metal-binding side chains, and metal-peptide interactions are reasoned to be similar,^[46] and the differences in observed enthalpies interpreted as arising primarily from the impact of backbone modification on folding.

The enthalpy for the interaction of parent zinc finger peptide 1 with Zn²⁺ is comparable to that reported previously for related sequence Sp1-3 under similar experimental conditions.^[49] Analyzing the difference in enthalpies for each variant compared to the prototype (ΔΔH), a few trends are apparent. Incorporation of Aib or BTD has a large favorable effect on folding enthalpy (ΔΔH ~ −5 kcal mol⁻¹). In contrast, variants with d-Pro or δ-Orn in the turn (peptides 4 and 5) were stabilized only slightly compared to the native backbone (ΔΔH ~ −1 kcal mol⁻¹).

In total, detailed analysis of structure and stability in Sp1-2 and variants yielded insights into the comparative behavior of the turn inducers examined in a metal-binding loop context. Aib-Gly proved the most effective turn mimetic among the series, in both recreating the native tertiary structure as well as dramatically improving folding enthalpy. BTD ranked next, also improving folding enthalpy but distorting the metal-binding loop subtly from the prototype. The d-Pro-Gly and δ-Orn modifications, while tolerated, were least effective among the turn replacements employed. Both led to deviation in turn geometry, backbone displacement of one of the metal-binding Cys residues, and minimal enthalpic stabilization of the domain.

Design, synthesis, and characterization of a proteomimetic analogue of Sp1-2 with modified helix, hairpin, and turn.

Having identified an optimal moiety for generating analogues of Sp1-2 with modified β-turns, we next sought to establish whether such modification was compatible with backbone alterations in the remainder of the domain. Thus, we targeted a variant of Sp1-2 in which all the individual secondary structures (hairpin, turn, and helix) were simultaneously modified in backbone chemical composition. We applied Aib-Gly in the turn based on the variant containing it showing the most-native-like structure among the artificial moieties examined above and maintaining the folded stability of the prototype domain. This modification was supplemented with six additional changes throughout the prototype guided by our prior work on Sp1-3 and other systems.^[4] Two α→N-Me-α residue substitutions were made at non-hydrogen bonding positions in the hairpin; this enables formation of a native-like hairpin fold in which the side chain display is retained and cross-strand hydrogen bonding is not interrupted. Four α→β³ residue substitutions were made at solvent-exposed positions from each turn of the helix; this approach has proved a versatile means for helix mimicry in both isolated secondary structure and more complex folds.^[50] Each newly introduced N-Me-α-orβ³-residue displays the same side chain as the α-residue in the prototype sequence that it replaces in order to retain the essential sequence-encoded folding characteristics of the prototype sequence. In one other change relative to prototype peptide 1, Arg¹ from the disordered N-terminus of the domain was omitted to avoid complications in NMR analysis resulting from cis/trans isomerization about the tertiary Arg¹-Pro² amide (seen in data for 1-5). The combination of the above modifications led to Sp1-2 variant 6 (Figure 4A), which was prepared, purified, and subjected to characterization of folded structure by NMR and folded stability by ITC, following methods described above.

Figure 4.

(A) Sequences of Sp1-2 peptide 1 and variant 6 alongside chemical structures of unnatural amino acid residues used in the variant. An R group, when present, matches that of the α-amino acid denoted by the corresponding single letter code in the sequence. Metal-coordinating Cys and His residues are underlined in the sequence; B = norleucine. (B) Ensemble of 10 lowest energy coordinate sets from the NMR structure of variant 6. (C) Overlay of the NMR structure of prototype domain 1 and variant 6. A single representative member of each ensemble is shown. Coloring of carbons in the structure of 6 matches the color scheme from panel (A). (D) Per-residue backbone rmsd values from the overlay of the NMR ensembles of variant 6 and prototype 1; the corresponding intra-ensemble rmsd plots for the two individual sequences are shown. Data for the disordered termini (residues 1, 29-31) are not shown.

The structural ensemble for peptide 6 obtained from simulated annealing with NMR-derived restraints shows good internal agreement (Figure 4B). Similar to the parent peptide 1 and variant 2, the hairpin loop in peptide 6 contains a type-II turn (Figure S3). The effective mimicry of the prototype fold by the heterogeneous backbone extends beyond the turn—both the β³-residue-modified helix and N-Me-α-residue-modified hairpin overlay closely with the coordinates of the natural backbone prototype (Figure 4C). Residue level comparison of the native backbone and variant shows that local backbone variation between the ensembles are similar in magnitude as the corresponding intra-ensemble uncertainty (Figure 4D).

Collectively, the NMR results above show that variant 6 is able to recreate the tertiary fold of the native domain. Analysis of the binding of Zn²⁺ by peptide 6 as monitored by ITC (Figure S5, Table 1) provides some insights into the effects of the backbone modifications on folding energetics. Similar to the Aib-Gly variant 2, the interaction of peptide 6 with zinc is more enthalpically favorable than prototype peptide 1 (ΔΔH −3.8 kcal mol⁻¹). This result indicates the artificial backbone units employed, which constitute about a quarter of the overall sequence, have a collective stabilizing effect on folding enthalpy. Comparison of result for variant 6 with variant 2, in which only the turn is modified, shows that the enthalpic effects of the various backbone modification present in 6 are dominated by the Aib-Gly moiety in the turn. This underscores the important role the metal-binding turn plays in the folding of zinc finger domains.^[51]

Conclusions

In summary, we have shown here that the metal-binding β-turn folding motif found in canonical Cys₂His₂ ββα zinc finger domains can accommodate a variety of artificial-backbone turn replacements; however, the length of a hairpin loop is an important determinant of its ability to tolerate modification. Comparative analysis of the impact of various turn modifications on domain folded structure and folding thermodynamics reveal subtle differences with respect to compatibility of different turn-promoting moieties within the delicate context of a metal-binding hairpin. Aib-Gly proved most effective at recreating the native fold, followed by BTD, d-Pro-Gly, and δ-Orn as the least effective. Incorporation of additional artificial residues in the helix and hairpin alongside Aib in the turn yields a proteomimetic zinc finger domain with a quarter of the backbone modified and a tertiary fold that experiences enthalpic benefit and is indistinguishable from the wild type in structure. Collectively, these results provide new insights to help inform the design of metalloprotein mimetics. Open areas of inquiry for future efforts include development of β-turn replacements that can be accommodated in the more demanding sequence context of a compressed hairpin loop or in a functional metalloprotein. Other unanswered questions surround the molecular origins of the large magnitude enthalpy / entropy compensation seen in folding thermodynamics accompanying backbone modification in zinc fingers^[9] as well as other systems.^[44] Work to address the above questions is ongoing, alongside continued broader efforts to establish scope and demonstrate applications of heterogeneous backbone proteomimetics.