Non‐additive stabilization by halogenated amino acids reveals protein plasticity on a sub‐angstrom scale

Abstract

It has been a long‐standing goal to understand the structure‐stability relationship of proteins, as optimal stability is essential for protein function and highly desirable for protein therapeutics. Halogenation has emerged as a minimally invasive strategy to probe the physical characteristics of proteins in solution, as well as enhance the structural stabilities of proteins for therapeutic applications. Although advances in synthetic chemistry and genetic code expansion have allowed for the rapid synthesis of proteins with diverse chemical sequences, much remains to be learned regarding the impact of these mutations on their structural integrity. In this contribution, we present a systematic study of three well‐folded model protein systems, in which their structural stabilities are assessed in response to various hydrogen‐to‐halogen atom mutations. Halogenation allows for the perturbation of proteins on a sub‐angstrom scale, offering unprecedented precision of protein engineering. The thermodynamic results from these model systems reveal that in certain cases, proteins can display modest steric tolerance to halogenation, yielding non‐additive consequences to protein stability. The observed sub‐angstrom sensitivity of protein stability highlights the delicate arrangement of a folded protein core structure. The stability data of various halogenated proteins presented herein should also provide guidelines for using halogenation as a strategy to improve the stability of protein therapeutics.

Introduction

The ability to adapt and maintain an optimal structure is essential for proteins to carry out their physiologic functions. As such, a longstanding goal of protein biochemistry has been centered around achieving structure prediction of proteins based on their sequences alone.1 Toward this goal, it is necessary to gain a detailed understanding of the factors, which contribute to protein stability.2 While much knowledge has been collected on the roles of hydrophobic effects, hydrogen bonding, and stereoelectronic effects in protein folding, more remains to be elucidated toward the full account of protein stability, particularly on the interdependence of various factors that have been individually well characterized.3, 4, 5 For example, the strength of backbone hydrogen bonds was found to be heavily dependent on the extent of hydrophobic burial,6, 7 leading to the fact that the hydrophobic packing of side chains and backbone hydrogen bonding synergistically stabilize a protein structure. In other words, the energetic gains of hydrophobic packing and backbone hydrogen bond formation afford non‐additive consequences to protein stability. Better understandings of such non‐additive behaviors are expected to yield improved models for protein structure prediction, as well as aid in the design of protein therapeutics.

Optimization of steric packing is believed to be crucial for the maintenance of a stable protein fold, as the mutation of a smaller to a larger amino acid usually causes detrimental effects to protein stability.8 However, recent studies of fluorinated proteins show that native protein structures can accommodate multifluorination of both aliphatic and aromatic amino acids. Incorporation of perfluorinated valines and phenylalanines into the core of protein structures often leads to improved stability despite the larger size of these amino acids,9, 10, 11, 12, 13, 14, 15 indicating that the hydrophobic core of proteins can accommodate some degree of steric expansion. Ideally, this plasticity can be exploited using site‐specific mutagenesis (e.g., by inserting fluorinated amino acids) to improve the stability of engineered proteins. However, the extent of allowed steric expansion has not yet been systematically assessed, since sterically‐demanding mutations often result in protein unfolding, as the steric bulk is too dramatic to be tolerated by a protein core. A series of non‐interacting, individually stabilizing point mutations within the same protein has been shown to have additive effects on stability. However, these mutations are spatially separated and act independently of one another, suggesting that structural perturbations remain localized to the site of substitution.16 Herein, we report a systematic investigation of steric expansion in three model proteins through the chemical incorporation of various halogenated amino acids in place of a native Phe residue (Fig. 1). The site‐specific incorporation of a given amino acid allows for steric expansion with sub‐angstrom increments at a single site of the protein. Thermodynamic analysis of the mutant proteins in our study reveals that proper halogenation can stabilize the protein fold substantially. However, the stabilizing effects of individual halogenation substituents are often non‐additive due to the increasing resistance of steric expansion by folded proteins. These results give unprecedented resolution to our understanding of the plastic nature of protein structures.

Figure 1.

Side chains of halogenated phenylalanine derivatives incorporated into VHP35, α₂D and PinWW mutants. Dipole moments were obtained with Gaussian 09 in gas phase for the respective toluene derivatives after geometry optimization at the M06L/LANL2DZ level

Results and Discussion

Halogenation is becoming an increasingly popular strategy in protein research and engineering for several reasons. First, the inertness of halogenated amino acids allows them to be easily incorporated into synthetic and recombinant proteins. Second, halogenation primarily changes the steric size and hydrophobicity of the target amino acids, although perfluorination of aromatic amino acids can significantly change their electronic properties.17 We envisioned that the plasticity of proteins could be assessed at sub‐angstrom resolutions via halogenation, which allows us to incrementally alter the steric size of particular side chains of interest (Fig. 1). We have assembled a set of halogenated phenylalanines with varied halogenation patterns to provide a range of steric sizes and geometries that differ by sub‐angstrom increments. Most of the amino acids are commercially available with a few exceptions including F₂₆F, Z_p, and Z_pI. We synthesized Z_p and Z_pI according to the early protocols reported by our group.18, 19 F₂₆F was synthesized as the Fmoc‐protected L‐amino acid by using a similar procedure (see Supporting Information for details).

Halogenation of VHP35

With these amino acids in hand, we first examined the effects of fluorination with the villin headpiece subdomain (VHP35) as a model system. VHP35 is an α‐helical miniprotein with 35 amino acids. It has been widely utilized for protein folding studies due to its small size and well‐characterized two state folding behavior.13, 18 VHP35 contains a hydrophobic core consisting of three interacting Phe residues with the H‐4 of F6 packing against the phenyl ring of F17, while the H‐6 atom of F10 packs against the π cloud of F6 [Fig. 2(A)].20, 21 This tightly packed core of aromatic residues makes them particularly suitable for studying protein plasticity with halogenated aromatic amino acids. We have synthesized a panel of VHP35 mutants that incorporate various fluorinated phenylalanines to replace the native Phe10 residue (Table 1). The mutants were analyzed via thermal denaturation and chemical denaturation, from which the T _m and ΔG_f (folding free energy) values were obtained, respectively.

Figure 2.

Structure and double mutant cycle analysis of VHP35. Mutant side chains are incorporated at position Phe10 (a) Cartoon representation of VHP35 (PDB: 1YRF) showing the aromatic residues Phe6, Phe10, and Phe17 in the hydrophobic core. Hydrogen atoms thought to be engaging in favorable interactions have been colored in blue; (b) Analysis of VHP35 variants show a decreased gain in stability upon para‐fluorination of F₃₅F. Calculated dipole moments (in Debye) for each side chain are denoted in purple

Table 1.

Thermodynamic Parameters for Selected VHP35, α₂D, and PinWW Mutants^a

Protein	T m (°C)b	C m (M)c	ΔG f (kcal mol−1)d	ΔΔG f (kcal mol−1)e
VHP35‐WT	64 ± 1.0	3.2± 0.1	−2.6 ± 0.1	–
VHP35‐F2F	70 ± 1.0	3.7± 0.1	−3.1 ± 0.1	−0.5 ± 0.1
VHP35‐F26F	62 ± 1.0	2.9± 0.1	−2.2 ± 0.1	+0.4 ± 0.1
VHP35‐F4F	74 ± 1.0	3.9± 0.1	−3.2 ± 0.1	−0.6 ± 0.1
VHP35‐F35F	74 ± 1.0	4.1± 0.1	−3.8 ± 0.1	−1.2 ± 0.1
VHP35‐F345F	70 ± 1.0	3.4± 0.1	−3.0 ± 0.1	−0.4 ± 0.1
α2D‐WT	29 ± 1.0	N/A	−5.9 ± 0.1	–
α2D‐ F4F	49 ± 1.0	N/A	−7.1± 0.1	−1.2 ± 0.1
α2D‐Zp	52 ± 1.0	N/A	−8.7 ± 0.1	−2.8 ± 0.1
α2D‐Z	78 ± 1.0	N/A	−12.6 ± 0.4	−6.7 ± 0.4
α2D‐F4I	65 ± 1.0	N/A	−11.2 ± 0.2	−5.3 ± 0.2
α2D‐ZpI	76 ± 1.0	N/A	−12.6 ± 0.2	−6.7 ± 0.2
PinWW‐WT	46 ± 1.0	2.1 ± 0.1	−2.2 ± 0.1	–
PinWW‐F4I	52 ± 1.0	3.3 ± 0.1	−2.7 ± 0.1	−0.5 ± 0.1
PinWW‐Zp	55 ± 1.0	4.2 ± 0.1	−3.4 ± 0.1	−1.2 ± 0.1
PinWW‐ZpI	52 ± 1.0	3.9 ± 0.1	−3.2 ± 0.1	−1.0 ± 0.1

^aMutations occur at position Phe10 for VHP35, Phe10 and Phe29 for α₂D, and Phe25 for PinWW respectively. The uncertainties of T _m, C _m, and ΔG _f were obtained through curve fitting (see below and experimental section for details). The uncertainties of ΔΔG _f were estimated based on the uncertainties of ΔG _f. Key experimental conditions used for thermal and chemical denaturation: 20 mM NaPi, 150 mM NaCl, pH 7.4 buffer; 20 μM (α₂D), 40 μM (PinWW), and 50 μM (VHP35) for T _m determination; chemical denaturation experiments were performed at 2°C with 10 μM VHP35 and 20 μM PinWW.

^bMelting curves were generated by plotting ellipticity against temperature; the T _m values were extracted via curve fitting according to the procedure described in Ref. [25].

^cConcentration of denaturant at the midpoint of the sigmoidal denaturant‐induced unfolding transition, determined by plotting the ellipticity data against GdmCl concentrations to generate denaturation curves, then fitting the curves using the standard baseline extrapolation method as described in reference 18. Chemical denaturation experiments were not run for α₂D variants.

^dThe ΔG _f values were calculated via equilibrium chemical denaturation experiments for VHP35 and PinWW at 2°C, and via variable temperature CD experiments for α2D (calculated for 37°C). Details of curve fitting and analysis can be found in the Materials and Methods Section.

^eΔΔG_f = ΔG_{f Mutant} – ΔG_{f WT}.

Fluorination of Phe10 at one of the two ortho positions is well tolerated by the VHP35 structure, resulting in a more stable protein VHP35‐F₂F (by −0.5 kcal/mol), presumably due to the increased hydrophobicity caused by fluorination. Interestingly, fluorination of both ortho positions results in decreased stability of the protein (VHP35‐F₂₆F), suggesting that double ortho‐fluorination reaches beyond the limit of VHP35 structural plasticity. However, the comparison of VHP35‐F₂F and F₂₆F is complicated by the CH‐π interactions, which we previously showed to contribute to the stability of VHP35 significantly.18 Consistent with this notion, double fluorination of Phe10 at the meta positions, which are not involved in CH‐π interactions, results in stabilized protein VHP35‐F₃₅F, which is more stable than VHP35‐WT by −1.2 kcal/mol. Similar to VHP35‐F₂F, the incorporation of F₄F gives a more stable protein (by −0.6 kcal/mol) as expected for the added size and hydrophobicity. Surprisingly, the triple fluorinated mutant, VHP35‐F₃₄₅F, gives comparable stability to that of VHP35‐F₄F. This presents a sharp contrast to VHP35‐F₃₅F, which is significantly more stable than VHP35‐WT. Clearly, the stabilizing effect afforded by meta fluorination of Phe10 depends on the para substituent. In other words, both meta and para fluorination of Phe10 does not give additive energetic gains in VHP35 stability [Fig. 2(B)]. We submit that the non‐additive behavior can be best explained by invoking the plasticity of the protein core, which appears to accommodate meta or para fluorination individually, but not in combination. The small size difference between fluorine and hydrogen suggests that the core structure of VHP35 only displays modest plasticity to accommodate the size expansion of F₃₅F or F₄F, but not F₃₄₅F.

Halogenation of α2D

To further probe the tolerance of steric expansion by folded protein structures, we applied halogenation to another model helical protein called α₂D. Originally developed by DeGrado and coworkers, α₂D is a soluble, 35‐residue protein that folds into a dimeric, four‐helix bundle [Fig. 3(A)].22 Upon dimerization, the Phe10 residue of one monomer interacts with the Phe29 residue of a second monomer in a face‐face stacked manner. In addition to this structural feature, α₂D exhibits a highly cooperative, reversible two‐state folding behavior that allows for facile thermodynamic analysis. We have synthesized a panel of α₂D double mutants with both the Phe10 and Phe29 residues replaced with one of the fluorinated analogs (Table 1). Since the folding of α₂D is a dimerization event, the folding free energies can be reliably determined by the concentration‐dependent thermal melting experiments and using the van't Hoff equation.15 No chemical denaturation experiments were performed.

Figure 3.

Structure and double mutant cycle analysis of α2D. Mutant side chains are incorporated at positions Phe10 and Phe29 (a) Cartoon representation of α2D (PDB: 1QP6) homodimer highlighting the two stacked Phe pairs in the core of the folded dimer (colored orange). The Phe29 residue of one monomer stacks against the Phe10 residue of another monomer; (b) Analysis of α₂D variants shows that para‐fluorination of Z_p is better tolerated compared to c) the more sterically demanding para‐iodination, hinting at the limited plasticity of α2D. Calculated dipole moments (in Debye) for each side chain are denoted in purple

Incorporation of F₄F into α₂D gives a moderate stabilization of −1.2 kcal/mol, which is consistent with the results of F₄F in VHP35, considering the double mutation (Phe10 and Phe29) of α₂D. The incorporation of Z_p into α₂D affords improved stability as well, with the folding free energy lower by −2.8 kcal/mol. Combining the fluorination of F₄F and Z_p gives the amino acid Z. Interestingly, the α₂D‐Z mutant gives significantly greater energetic stabilization than the sum of F₄F and Z_p (–6.7 kcal/mol vs. −4.0 kcal/mol). The positive non‐additivity [Fig. 3(B)] observed here contrasts to what we observed with the VHP35 mutants [Fig. 2(B)]. We submit that this can be rationalized by the greater dipole of the Z side chain, which stacks up in an antiparallel fashion in α₂D. The favorable dipole‐dipole coupling has been previously shown as a significant contributor to the stability of α₂D.15 Nevertheless, the steric expansion by perfluorination of Phe appears to be well accommodated by α₂D. Considering α₂D‐Z incorporates four Z residues, this designed protein fold appears to exhibit greater plasticity than the naturally occurring protein VHP35.

To further probe the tolerance of steric expansion of α₂D, we have incorporated para‐iodinated phenylalanine analogs (F₄I and Z_pI) into this model protein [Fig. 3(C)]. Para iodination of Phe (F₄I) gives a comparable dipole to that of F₄F. However, the F₄F to F₄I mutation dramatically stabilizes the protein (by −4.2 kcal/mol) despite the larger size of the iodo substituent. The effect of this single atom substitution is rather remarkable indicating that the steric size of iodine is well accommodated by the protein core of α₂D. In sharp contrast, replacing the para‐fluorine atom of Z with an iodine gives essentially no benefit to protein stability. The lack of stabilization by Z_pI in comparison to Z is likely due to Z_pI reaching the limit of structural plasticity of the protein core although Z and F₄I are both well accommodated. In other words, the presence of four fluorine substituents makes the protein core more resistant to additional steric expansion. We note that Z_pI does show a lower dipole moment than Z, which we expect to contribute to the lack of stabilization by the F‐to‐I substitution as well. However, it is difficult to dissect out the individual contributions of the dipole and steric factors.

Halogenation of PinWW domain

The results presented above demonstrate that, in at least in two model systems, a helical protein fold displays limited steric tolerance or plasticity, which gives rise to non‐additive energetic consequences by halogenation. To examine the generality of these observations, we enlisted the PinWW domain as a model β‐sheet protein. Similar to VHP35 and α₂D, the PinWW domain has also been widely utilized in protein folding studies due to its small size and well‐characterized folding behavior.6 The PinWW domain adopts a three‐stranded β‐sheet with its hydrophobic core hosted on the concave face of the sheet.23 The hydrophobic core features a central phenylalanine residue [Fig. 4(A)], which should be ideal for performing halogenation studies as we did for the helical proteins.

Figure 4.

Structure and double mutant cycle analysis of PinWW. Mutant side chains are incorporated at position Phe25 (a) PinWW sequence and structure (PDB: 1F8A) showing the location of Phe25 in the flexible loop region of PinWW; (b) Analysis of PinWW mutants reveals a similar trend to α₂D despite the different local environments of the peptides. Calculated dipole moments (in Debye) for each side chain are denoted in purple

Both para‐iodination and tetrafluorination of the native F25 residue stabilize the folded state of PinWW compared to the WT peptide [Fig. 4(B)], though to different extents (–0.5 kcal mol⁻¹ for F₄I and −1.2 kcal mol⁻¹ for Z_p). Interestingly, the para‐iodinated version of Z_p, namely Z_pI, is +0.2 kcal mol⁻¹ less stable than Z_p (–1.0 kcal mol⁻¹ vs −1.2 kcal mol⁻¹ for Z_pI vs. Z_p, respectively), despite the significant increase in hydrophobicity afforded by the Z_pI side chain. Similar to the results for the previously discussed helical model proteins, the β‐sheet fold of PinWW domain appears unable to benefit from the stabilizing effect of F₄I and Z_p in combination, again presumably due to Z_pI breaching the limit of the plasticity of the protein.

Conclusion

Earlier studies of fluorinated proteins indicate that the core structure of folded proteins can accommodate the steric expansion of some fluorinated amino acids, through which improved stability can be acquired for the host proteins. This contribution describes our use of various halogenated amino acids to perform a systematic analysis of such protein plasticities. The results of three model protein systems consistently show that multifluorination and iodination can be accommodated by the core structure of proteins, leading to substantial improvements of protein stability. However, the thermodynamic benefit of halogenation cannot always be enjoyed in combination. We submit that the non‐additive effect of protein halogenation can be best rationalized by the plasticity of host proteins, which appears to be modest and leads to a “stabilization plateau” where the increased steric bulk associated with a side chain mutation is no longer tolerated. The fact that these results are consistent across all model peptide systems lends credibility to the fact that the observed toleration limits are general effects of site‐specific mutagenesis. More importantly, the results indicate that although certain mutations may be favorable on their own, a combination may sometimes be detrimental to overall protein stability. The results should provide some guidelines for using halogenation as a stabilizing mechanism in protein engineering.

Materials and Methods

General procedures

All chemicals were obtained from Thermo Fisher Scientific (Waltham, MA) or Sigma‐Aldrich (St. Louis, MO) unless otherwise noted. Fmoc‐Gly Wang and Fmoc‐Phe‐Wang resins were purchased from Novabiochem (San Diego, CA). All Fmoc‐protected amino acids, HBTU, and piperidine were purchased from Advanced Chemtech (Louisville, KY) or Chem‐Impex International (Wood Dale, IL). N‐methylmorpholine (NMM) was purchased from TCI AMERICA (Portland, OR). Fmoc‐N‐pentafluorophenylalanine (Fmoc‐Z) was purchased from PepTech Corporation (Burlington, MA). Fmoc‐N‐2,3,5,6‐tetrafluoro‐4‐iodophenylalanine (Z_pI) was synthesized according to a previously reported protocol.19, 24 Peptide synthesis was carried out on a Tribute® peptide synthesizer (Protein Technologies, Tucson, AZ). All crude peptides were purified by RP‐HPLC using a Jupiter C18 300 Å (250 × 20.00 mm, 10 μm) prep column (Phenomenex, Inc. Torrance, CA) on a Waters PrepLC system (Waters Corporation Milford, MA). Size exclusion chromatography was conducted with a HiLoad Superdex^TM 30 column on the ÅKTA FPLC system (GE Healthcare, Piscatawat, NJ). LC‐MS analyses of the peptides were performed on an Agilent LC‐MS (Agilent Technologies, Santa Clara, CA) in house. All peptide concentrations were determined spectrophotometrically by measuring Trp absorbance at 280 nm (ɛ = 5690 M⁻¹ cm⁻¹ per Trp residue) using a Nanodrop 2000c Spectrophotometer (Thermo Fisher Scientific, Waltham, MA). ¹H and 13C NMR measurements for the amino acids were performed on a Varian INOVA 500 MHz spectrometer and 19F NMR measurements of the peptides were taken on a Varian INOVA 600 MHz spectrometer. Circular dichroism (CD) measurements were performed on an AVIV CD spectrometer (Aviv Biomedical Inc. Lakewood, NJ).

Peptide synthesis

All peptides were synthesized via automated solid phase peptide synthesis (SPPS) on a Tribute^® peptide synthesizer (Protein Technologies, Tucson, AZ) using standard Fmoc/tBu chemistry (Supporting Information Table S1). Syntheses were performed on a 0.05mmol scale using Fmoc‐Phe‐Wang resin (VHP35) and Fmoc‐Gly‐Wang resin (α₂D and PinWW) and 5 equivalents of Fmoc‐protected amino acids. All peptides were synthesized with an N‐terminal amine and a C‐terminal carboxylic acid. The resin was treated twice with a mixture of 20% piperidine v/v in DMF for 10 minutes each at room temperature to remove the Fmoc protecting group. For standard coupling steps, 5 equivalents of Fmoc‐protected amino acids were mixed with 5 equivalents of HBTU and subsequently dissolved in a 0.4 M N‐methylmorpholine (NMM) in DMF solution. After a 2‐minute activation time, the amino acid solution was delivered to the resin under N₂ protection and mixed for 30 minutes at room temperature. For coupling of unnatural residues, 3 equivalents of the amino acid were manually activated and mixed with resin under N₂ protection for 1 hour at room temperature (HBTU‐mediated activation conditions are identical to standard coupling). After a final Fmoc‐deprotection step, the peptide was cleaved and globally deprotected with reagent K (80% TFA, 7.5% phenol, 5% thioanisole, 5% H₂O and 2.5% 1,2,ethanedithiol) for 2 hours at room temperature. The crude products were precipitated with cold diethyl ether (3x), centrifuged and vacuum‐dried overnight.

Peptide purification

Peptides were purified by reverse‐phase HPLC on a Waters PrepLC using a Phenomenex Jupiter C18 prep column (250 × 20.00 mm, 10 micron) and a water/acetonitrile gradient eluent system. Eluent A: 95% H₂O, 5% CH₃CN, 0.1% TFA; Eluent B: 5% H₂O, 95% CH₃CN, 0.1% TFA. Crude peptides were dissolved in 100% Eluent A, filtered, and purified using focused gradients. After RP‐HPLC purification, the purest fractions were collected, combined and lyophilized, then re‐dissolved in 2 mL of 8.0 M guanidinium chloride (GdmCL). Size exclusion chromatography was performed using a HiLoad Superdex^TM 30 column eluted with a 0.22 μm‐filtered phosphate buffer (20 mM NaPi, 150 mM NaCl, pH 7.4). Desired fractions were combined and concentrations were determined by measuring Trp absorbance (ɛ = 5690 M⁻¹ cm⁻¹ per Trp residue). Purities and identities were confirmed using LC‐MS analysis (Supporting Information Tables S1 and S2).

CD spectra

All peptides samples for CD analysis were dissolved in Native Buffer A (20 mM NaP_i. 150 mM NaCl, pH 7.4). Wavelength scans in the Far‐UV region were collected using an AVIV model 420 spectrometer (Aviv Biomedical, Inc. Lakewood, NJ) equipped with a temperature‐controlled cell holder and autotitrator. For VHP35 variants, spectra were recorded at 2°C at 50 μM peptides concentrations, using a 2 mm path length quartz cuvette and monitored at a wavelength between 260 nm to 190 nm. Spectra for PinWW and α₂D variants were collected at both 25°C and 2°C and several concentrations per peptide, scanning from 260 nm to 200 nm at 1 nm intervals. All VHP35 and α₂D variants displayed the characteristic α‐helical structures with ellipticity minima at 222 nm and 208 nm. All PinWW variants displayed a maximum at 227 nm, which is characteristic of PinWW domain and suggests a well‐folded structure despite point mutations.

Thermal denaturation experiments

For the PinWW series, thermal denaturation experiments were monitored at 226 nm and data were collected from 2°C to 98°C (at 2°C intervals). VHP35 and α₂D variants were monitored at 222 nm and collected from 2°C to 110°C (at 2°C intervals). All peptides samples were prepared in Native Buffer A (20 mM NaP_i. 150 mM NaCl, pH 7.4) at concentrations of 50 μM for VHP35 and several concentrations per peptide for PinWW and α₂D in a 2 mm path length quartz cuvette. For each temperature point, the samples were allowed to equilibrate for 90 seconds prior to data collection. Signals were averaged for 30 seconds and then averaged. Upon completion, samples were cooled downed to 25°C and spectra were acquired to confirm the highly reversible folding/unfolding behavior. Plotting raw ellipticity data against temperature generated thermal denaturation curves. All thermal denaturation curves display sigmoidal transitions, consistent with a two‐state folding model. Curve fitting according to a reported procedure25 yielded the T_m values shown in Table 1. For α₂D variants, whose folding involves dimerization, melting curves were collected at 6 different concentrations (10 μM, 20 μM, 30 μM, 40 μM, 50 μM, 100 μM). The ΔG_f values were extracted from the concentration dependence of T_m (defined by the van't Hoff equation (lnK = –ΔH/RT + ΔS/R)) as we described previously.15 The specific equations used to calculate these values can be found in the Supporting Information as well.

Chemical denaturation experiments

Guanidinium chloride (Gdm·Cl) chemical denaturation experiments were performed on all PinWW and VHP35 variants to assess their thermodynamic stabilities. Experiments were carried out at 2°C on a circular dichroism (CD) spectrometer equipped with an automated titrator at peptide concentrations of 10 μM (VHP35) and 20 μM (PinWW). Sample A was prepared in Native Buffer A (20 mM NaP_i. 150 mM NaCl, pH 7.4) and sample B in Native Buffer B (20 mM NaP_i. 150 mM NaCl, 8M Gdm·Cl, pH 7.4). Sample B was added in 0.2 M fractions and before each addition, an equal volume of the sample (solution A) in the cuvette was taken out to maintain a constant volume. After each addition, the sample was mixed for 30 seconds, followed by a 2 minutes equilibration period. The data was collected for 30 seconds and averaged for each data point. The ellipticity values were recorded at 225 nm (VHP35) and 226 nm (PinWW) and plotted against GdmCl concentration to yield the denaturation curves. Thermodynamic fitting of these denaturation curves was carried out by following a previously published protocol.18 Briefly, the pre‐ and post‐transition baselines were generated with the standard baseline extrapolation method and linear fitting after plotting folding free energies against GdmCl concentration gave the thermodynamic parameters (C_m and ΔG_f) listed in Table 1.

Supporting information

Supporting Information