Insight into polyproline II helical bundle stability in an antifreeze protein denatured state

Abstract

The use of polyproline II (PPII) helices in protein design is currently hindered by limitations in our understanding of their conformational stability and folding. Recent studies of the snow flea antifreeze protein (sfAFP), a useful model system composed of six PPII helices, suggested that a low denatured state entropy contributes to folding thermodynamics. Here, circular dichroism spectroscopy revealed minor populations of PPII like conformers at low temperature. To get atomic level information on the conformational ensemble and entropy of the reduced, denatured state of sfAFP, we have analyzed its chemical shifts and {¹H}-¹⁵N relaxation parameters by NMR spectroscopy at four experimental conditions. No significant populations of stable secondary structure were detected. The stiffening of certain N-terminal residues at neutral versus acidic pH and shifted pK_a values leads us to suggest that favorable charge-charge interactions could bias the conformational ensemble to favor the formation the C1-C28 disulfide bond during nascent folding, although no evidence for preferred contacts between these positions was detected by paramagnetic relaxation enhancement under denaturing conditions. Despite a high content of flexible glycine residues, the mobility of the sfAFP denatured ensemble is similar for denatured α/β proteins both on fast ps/ns as well as slower μs/ms timescales. These results are in line with a conformational entropy in the denatured ensemble resembling that of typical proteins and suggest that new structures based on PPII helical bundles should be amenable to protein design.

Significance

Polyproline II (PPII) helical bundles are an emerging protein fold showing promise as a new LEGO brick for protein design. Here, we use NMR spectroscopy to characterize the conformation and dynamics of the H. harveyi “snow flea” antifreeze protein, a model PPII helical bundle under unfolding conditions. The reduced, denatured protein shows no detectable secondary structure under three different temperature/pH values or in the presence or absence of an N-terminal His-tag. Despite a high content of glycine residues (almost 50%), the backbone mobility is similar to those of typical globular proteins with less glycine and composed of α-helices and β-sheets. These results represent a significant advance in our understanding of the conformational stability and folding of PPII helical bundle proteins.

Introduction

Great advances have been made in recent years in protein design. Novel protein folds and switches as well as new enzymes catalyzing reactions, which are unknown in nature, have been achieved (1,2). Nevertheless, almost all are based on combinations of α-helices or β-sheets or both, but largely ignore polyproline II (PPII) helices (3). In nature, PPII helices are common as they are the main component of collagen, the most abundant human protein, and most globular proteins contain at least one turn of the PPII helix (4,5,6). Moreover, a small protein class, recently reviewed (7), which includes snow flea antifreeze proteins (sfAFP) (8), bacteriophage tail spike proteins (9), the enzyme acetophenone carboxylase (10), the universally conserved translation factor Obg (11), and the recently discovered anaplastic lymphoma kinase (12) consist of or contain glycine-rich PPII helical bundle domains. These glycine-rich PPII helical bundles can give rise to remarkably flat or honeycomb structures, which could find novel applications in protein design and biotech applications. Nevertheless, their use in protein design is hampered by a lack of understanding regarding the bases of their conformational stability.

Recently, we characterized the stability and dynamics of the Hypogastrura harveyi sfAFP by NMR spectroscopy and computational methods (13). Discovered in the snows of Ontario (14), this protein’s 3D structure, which was determined by X-ray crystallography (8), consists of six long PPII helices connected by short turns, as shown in Fig. S1. The results showed that sfAFP is as well folded and rigid as globular proteins composed of α-helices or β-sheets or both despite a very high content of glycine residues. In fact, the composition of sfAFP is almost 50% glycine residues. The presence of two disulfide bonds and 28 weak Cα-H···O=C contribute significantly to the conformational stability of sfAFP (13,15). However, glycine residues have very small side chains—just H—and can sample broad regions of the Ramachandran space. This means that the conformational entropy change for folding is expected to be much larger and unfavorable for glycine residues relative to the other 18 common amino acid residues and proline. Nevertheless, the residue level dynamics and conformational entropy of the unfolded sfAFP are still unknown.

To address this issue, the main goal of this study is to characterize the dynamics of unfolded sfAFP at the level of single residues using NMR spectroscopy. Fluorescence and circular dichroism spectroscopies are also used to assess the presence of hydrophobic clusters and tendencies to adopt secondary structure. Previous NMR characterizations of unfolded proteins have made use of chemical denaturants, such as urea (16,17). On the other hand, disulfide bond reduction is perhaps a more physiologically relevant way to denature proteins. In contrast to glycine residues, intact disulfide bonds act to stabilize proteins by limiting the conformational freedom of the denatured state (18). In this way, they decrease the unfavorable change in conformational entropy that accompanies folding. Using theory, it is possible to calculate the stabilizing contribution based on the number of residues within the loops closed by the disulfide bonds (19) and is 7.2 kcal/mol at 25.0°C in the case of sfAFP (as detailed in the supporting material). Considering that the overall conformational stability of sfAFP is significantly lower than 7.2 kcal/mol (13,15), disulfide bond reduction is expected to denature this protein, as was observed by low-/medium-resolution spectroscopic methods (15). Here, we confirm by multidimentional NMR spectroscopy that sfAFP is indeed unfolded by reduction. The chief result of this study is that ps/ns and μs/ms backbone dynamics of sfAFP, as measured by ¹H-¹⁵N-based NMR relaxation measurements, retains a moderate rigidity akin to other disordered or unfolded proteins with a much lower glycine content.

Materials and methods

sfAFP isotopically labeled with ¹³C,¹⁵N was produced as described previously (13). In brief, transformed BL21 star (DE3) E. coli bacteria containing the appropriate vector codifying sfAFP were grown in 2 L LB at 37°C until the OD₆₀₀ reached 0.8 units. Afterward, they were centrifuged and transferred to 0.5 L of minimal medium with ¹³C-glucose and ¹⁵NH₄Cl as the sole carbon and nitrogen sources, respectively. After 1 h at 37°C, expression was induced with 0.5 mM of IPTG and the culture was kept at 25°C overnight. The protein was purified by Ni²⁺ affinity chromatography followed by anionic-exchange chromatography. Following purification, the His-tag followed by an enterokinase cleavage site (sequence: MAHHHHHHVGTGSNDDDDK) was not cleaved, except for one set of experiments. The purified protein was then reduced with tris(2-carboxyethyl)phosphine (TCEP) (obtained from Sigma, Saint Louis, MO, USA), a strong, phosphine-based reducing agent. Unlike β-mercaptoethanol, glutathione, or dithiothreitol, TCEP does not form adducts with protein sulfhydral groups (20).

Ammonium 8-anilinonaphthalene-1-sulfonate acid (ANS), used for fluorescence experiments was from Fluka (>98% pure). Stock solutions of ANS were prepared in 4.5 mM Na₂HPO₄/0.5 NaH₂PO₄ (pH 6) and its concentration was measured by absorbance using an extinction coefficient of 5000 cm⁻¹ M⁻¹ at 350 nm. Bovine serum albumin was from Sigma.

The NMR samples contained approximated 0.7–1.4 mM ¹³C,¹⁵N sfAFP, 4.5 mM Na₂HPO₄/0.5 NaH₂PO₄, 3.0 mM TCEP, and 0.2 mM sodium trimethylsilylpropanesulfonate (DSS) as the internal chemical shift reference as well as 10% D₂O for the spectrometer lock. The concentration of sfAFP, which lacks aromatic residues and produces anomalous results in BCA assays due to the presence of reducing agents, was estimated by integration of NMR signals compared with the integral of the trimethyl moiety of a known concentration of DSS. The uncertainty of this approach is estimated to be <10%. All samples were placed in a water-matched Shigemi NMR tube and sealed with parafilm to reduce exposure to atmospheric oxygen.

The conformation and dynamics of reduced sfAFP were studied by NMR spectroscopy under three different sets of conditions, namely: 1) pH 2.5 and 25°C, 2) pH 2.5 and 5°C, and 3) pH 6.15 and 5°C. The low pH conditions were chosen to optimize spectral quality for assignment, first at 25°C and then at 5°C, which is more physiologically relevant (15). The third set of conditions match those used to characterize oxidized, folded sfAFP (13). For a fourth sample, the His-tag was cleaved and measurements were recorded at pH 6.57 and 5°C.

Fluorescence spectroscopy

Emission fluorescence spectra of sfAFP and bovine serum albumin, alone or in the presence of ANS, were recorded on a Jobin-Yvon Fluoromax-4 spectrofluorimeter fitted with a Peltier temperature control unit. Spectra were recorded at 25.0°C using 2 nm excitation and emission bandwidths and a 2 nm s⁻¹ scan speed. The excitation wavelength was 380 nm and emission was scanned over 400–600 nm.

Circular dichroism spectroscopy

A JASCO 810 circular dichroism spectrophotometer equipped with a Peltier temperature control system was used to record far UV circular dichroism (CD) spectra (200–260 nm) on reduced, denatured sfAFP at pH 2.5 and pH 6.1 at 0.0, 20.0, 40.0, 60.0, and 80.0°C using a scan speed of 50 nm s⁻¹ and a 1.2 nm bandwidth. Ten scans were averaged per spectrum. Samples contained 22 μM sfAFP (pH 2.5) or 25 μM sfAFP (pH 6.1) in 5.0 mM sodium (di)hydrogen phosphate buffer. In addition, the signal at 220 nm was monitored from 0.0 to 80.0°C using a heating rate of 1°C min⁻¹.

NMR spectral assignment

A Neo Avance 800 MHz (¹H) NMR spectrometer, equipped with a ¹H,¹³C,¹⁵N cryoprobe and Z-gradients, was utilized for most NMR experiments. The program Topspin (versions 2.1 and 4.0.8, Bruker BioSpin, Billerica, MA, USA) was utilized to record, transform, and analyze the spectra. As sfAFP is disordered under the conditions employed here and since its sequence contains a high proportion of Gly and Ala residues, it is challenging for NMR spectral assignment. Since ¹³CO and ¹⁵N nuclei retain more dispersion in IDPs, we have used a nonconventional strategy based on a pair of 3D ¹³C-detected spectra which provide consecutive (i, i+1) ¹³CO and ¹⁵N backbone connectivities (21,22). Even with this approach, some sequential residues with identical ¹³CO and ¹⁵N chemical shift values were observed. Therefore, another 3D ¹H-detected spectrum which yields consecutive ¹HN_i-¹HN_i+1 connectivities was registered (23,24). The analyses of all these data led to the essentially complete spectral assignment at pH 2.5 and 25°C. These results, and a ¹H-¹⁵N HSQC spectrum recorded at 15°C, were used to transfer the assignments to pH 2.5, 5°C. Following the acquisition of spectra at pH 2.5, 5°C, the pH was increased to 6.15 by adding small volumes of 0.1 M Na₂CO₃ dissolved in D₂O. A schematic diagram summarizing the assignment strategy is shown in Fig. S2. The initial analysis of the pH 6.15, 5°C data set revealed that the protein had been cleaved between residues T16/A17. This sample was also characterized to determine the effect of said cleavage on any residual structure and rigidity. Afterward, another sample was prepared and the measurements at pH 6.15, 5°C were repeated.

Dynamic measurements

{¹H}-¹⁵N longitudinal relaxation rates (R₁) and relaxation rates in the rotating frame (R_1ρ) as well as the {¹H}-¹⁵N heteronuclear NOE (hNOE) were recorded to assess dynamics on the ps/ns and μs/ms timescales. hNOE values measured for backbone amide groups as the ratio of resonance intensities registered in the presence or absence of saturation in an interleaved mode using a recycling delay of 11 s. Peaks were integrated with Topspin 4.0.8 (Bruker BioSpin) and uncertainties were estimated as the standard deviation of the integrals of peak-sized spectral regions that lack signals and contain just noise.

Sets of seven ¹H-¹⁵N correlation spectra with relaxation delays of 20, 1720, 1500, 860, 220, 1100, and 500 ms for R₁ and 10 ¹H-¹⁵N correlation spectra with relaxation delays at 8, 300, 36, 76, 900, 100, 500, 156, 200, and 700 ms for R_1ρ were recorded. The R₁ and R_1ρ relaxation rates and their estimated uncertainties were calculated using KaleidaGraph (version 3.6) by least-squares fitting of the exponential decay function I_t = I₀ exp^(−kt), where I_t is the integral at time t, I₀ is the integral at time zero, and k is the rate, to peak integral data measured with Topspin 4.0.8. In all cases, Pearson’s correlation coefficient, R, was >0.99.

Measurement of Asp and His pK_a values

The pH titration of ¹³C,¹⁵N labeled sfAFP charged residues over a pH range of 3–9 was monitored by recording 12 ¹H-¹³C HSCQ spectra at 800 MHz (¹H) and 5°C. Since the His-tag contains six His and four Asp residues, for this sample the tag was cleaved with enterokinase (New England Biolabs, Ipswich, MA, USA) to facilitate pK_a measurements. This ¹³C,¹⁵N sfAFP His-tag-less sample contained 0.7 mM protein, 5 mM sodium hydrogen phosphate, 10 mM NaCl, and 5 mM deuterated sodium acetate, and was titrated by adding small aliquots of DCl or KOH. As a control, a 3D HNCA spectrum and {¹H}-¹⁵N NOE experiment were measured and analyzed to assess the impact of the His-tag on the conformational dynamics. A 3D CBCAcoNH spectrum was also recorded and used to confirm the assignment of the ¹³Cβ signals of His and Asp residues, which are overlapped in the 2D ¹H-¹³C HSQC spectra. To obtain the collective pK_a value for each type of residue, the Henderson-Hasselbach equation was fit to the ¹³Cβ chemical shift versus pH data using a least-squares algorithm.

Paramagnetic relaxation enhancement measurements

For paramagnetic relaxation enhancement (PRE) experiments, a triple variant was prepared in which three of the four cysteine residues (C13, C28, and C43) were mutated to serine. The sole remaining cysteine residue (C1) was linked to the spin label (S-(1-oxyl-2,2,5,5-tetramethyl-2,5-dihydro-1H-pyrrol-3-yl)methyl methanesulfonothioate (MTSL), which was purchased from Enzo Life Sciences, Farmingdale, NY, USA, by previously described methods (25). This sample: sfAFPC13S,C28S,C43S,C1-MTSL, abbreviated as: sfAFP-C1^∗, was labeled with ¹⁵N. The PRE effect was measured as the ratio between 2D ¹H-¹⁵N HSQC resonance intensities sfAFP-C1^∗ with MTSL in the paramagnetic and diamagnetic states. These resonance intensities were measured as peak integrals using Topspin 4.0.8 (Bruker BioSpin). Ascorbic acid (Sigma, buffered to pH 7) was added to a final concentration of 5.0 mM to reduce MTSL to the diamagnetic state. Afterward, a 3D ¹H-¹⁵N-¹H HSQC-NOESY spectrum (mixing time = 100 ms) was recorded to confirm resonance assignments.

Effective transverse relaxation

The effective transverse relaxation rates, R_2eff, were assessed as a function of an applied B1 radiofrequency field using Carr-Purcell-Meiboom-Gill (CPMG) relaxation dispersion experiments. Ten interleaved ¹⁵N B1 fields of 0, 67, 133, 200, 257, 400, 533, 733, 866, and 1000 Hz were used with a total relaxation time of 60 ms. These experiments were acquired at 800 MHz (¹H frequency) with 2048 and 256 points in the direct and indirect dimensions, respectively, with spectral widths of 12 and 23 ppm centered at 4.5 and 117.5 ppm, for ¹H and ¹⁵N, respectively, on a His-tagged, ¹⁵N-sfAFP sample at pH 6.1 and 5.0°C.

Results

ASN fluorescence suggests that reduced, denatured sfAFP lacks hydrophobic clusters

Native sfAFP has a flat hydrophobic face, which is thought to drive dimerization (13) and bind ice (8). To test if some nonpolar residues form a hydrophobic cluster in reduced, denatured sfAFP, we recorded fluorescence spectra of sfAFP alone or in the presence of ANS. As a positive control, this experiment was also performed with BSA, a protein known to have exposed hydrophobic patches (26). Whereas BSA strongly enhanced and blue shifted the fluorescence emission spectrum of ANS, no such spectral changes were produced by reduced, denatured sfAFP (Fig. S3). This suggests that reduced, denatured sfAFP does not sport exposed hydrophobic patches and is not a molten globule.

Circular dichroism spectroscopy of reduced sfAFP reveals a tendency for denatured state ensemble to modestly populated PPII conformations

The CD spectra recorded on reduced sfAFP at pH 2.5 and pH 6.1 and over a range of temperatures (Fig. S4, A and B) lack spectral features typical of α-helices and β-sheets, as well as folded sfAFP protein (15), and resemble the reported spectrum for chemically denatured sfAFP (15). However, a decrease in the signal around 220 nm is observed as the temperature increases. This is inconsistent with a loss of α-helical or β-sheet conformations but is suggestive of a loss of PPII helical structure (Fig. S4, C and D) as difference spectra resemble those reported for PPII helices (15,27).

Following the signal at 220 nm while heating at 1°C/min revealed a gradual decrease in signal at both pH values (Fig. S4 E). This gradual decrease could be approximately fit by a linear function (Fig. S4 F) and does not have a sigmoidal shape as is frequently observed for the cooperative denaturation of well folded proteins, including sfAFP (15). Following recooling to 0°C, a final spectrum was recorded. The initial CD spectrum was recovered completely at pH 2.5 and approximately at pH 6.1 (Fig. S4 G).

Taken together, these circular dichroism spectroscopic results are consistent with a lack of stable secondary structure under all the conditions studied. A minor population of the PPII may be present at low temperature and decreases upon heating. This population might be slightly higher at pH 6.1 than pH 2.5.

Reduced sfAFP is unfolded and lacks significant populations of secondary structure

The 2D ¹H-¹⁵N HSQC and ¹³C-¹⁵N CON spectra of ¹³C,¹⁵N-labeled sfAFP at pH 2.5 and 25°C are shown in Fig. 1. These spectra are strikingly different from those of oxidized, folded sfAFP observed at pH 6.15 and 5°C (13). In particular, the excellent ¹HN chemical shift dispersion observed for the folded protein is absent following reduction. Despite the low sequence complexity and poor ¹HN chemical shift dispersion, by following the nonconventional assignment strategy described in the materials and methods section, it was possible to obtain the essentially complete backbone ¹HN, ¹⁵N, ¹³CO, ¹³Cα, and side chain ¹³Cβ chemical shift assignments for reduced, denatured sfAFP. The only unassigned resonances are ¹H,¹⁵N for Cys1, ¹³Cα and ¹³Cβ for residues preceding proline residues (C13, G25, N40, TT70, A77, A80) and also ¹³Cα, ¹³Cβ, ¹³CO for P81. Assignments were also obtained at pH 2.5, 5°C (only backbone ¹H-¹⁵N; 83% complete) and at pH 6.15, 5°C where 97% of ¹HN, 97% of ¹⁵N, 96% of ¹³CO, 86% of ¹³Cβ, and 98% of ¹³Cα signals are assigned. All three sets of assignments have been deposited in the BMRB databank under access number 51324.

Figure 1.

Assigned 2D ¹H-¹⁵N HSQC and 2D CON spectra and ¹³Cα conformational chemical shifts show that reduced ¹³C,¹⁵N-sfAFP is thoroughly denatured. Assigned 2D ¹H-¹⁵N HSQC (red) (A) and 2D ¹³CO¹⁵N (blue) (B) of reduced ¹³C,¹⁵N sfAFP at pH 2.5, 25°C. The ¹⁵N of each residue are labeled. Residues of the His-tag are labeled with negative numbers. (C) Conformational ¹³Cα chemical shifts for sfAFP at pH 2.5, 25°C (top panel), and at pH 6.15 and 5°C with (middle panel) or without (bottom panel) a proteolytic cleavage between residues T16/A17. To see this figure in color, go online.

The observation of signals of reduced TCEP in 1D, ¹H spectra (Fig. S5) means that the Cys will be reduced since TCEP is a potent reducing agent and quantitatively reduces disulfide bonds (20). The ¹³Cβ chemical shift values of the Cys residues are observed to be <30 ppm. Taking into account reported ¹³Cβ chemical shift values for reduced (<32 ppm) and oxidized (>35 ppm) Cys (28), these values indicate that the four side chains are reduced to -CH₂-SH and that the two disulfide bonds are broken. sfAFP contains six proline residues, four of which paradoxically lie in turns outside the six PPII helices. Whereas two of the Pro residues adopt the cis conformation in folded sfAFP, here all ¹³Cβ proline chemical shifts range from 32.0 to 32.3 ppm and are consistent with the trans conformation (29). This concords with unfolding as the trans conformation normally dominates denatured state ensembles. Following the example of Bax and co-workers (30), we analyzed minor peaks but none arising from Xaa-Pro cis peptide bonds could be rigorously assigned.

Reduced, denatured sfAFP might retain some partially structured segments as has often been observed in intrinsically disordered (31) or chemically denatured proteins (17,32). Since ¹HN chemical shift dispersion arises mainly from >C=O···¹H-N H-bond formation, the loss of dispersion observed for all the conditions studied here indicates that the interhelical H-bond network has broken down following reduction. To further test for residual structure, the conformational chemical shifts (Δδ) for ¹³Cα plotted are in Fig. 1 C for reduced sfAFP at three conditions; i.e., 1) pH 2.5 and 25°C, 2) pH 6.15 and 5°C for the sample containing a proteolytic cleavage between T16 and A17, or 3) pH 6.15 and 5°C without that cleavage. The mean Δδ ¹³Cα values are close to zero for all three data sets; namely −0.07 ± 0.14, −0.08, ±0.12, and −0.09 ± 0.15 ppm, respectively. Moreover, there is no statistical difference between the three sets of Δδ¹³Cα values as judged by p values > 0.40 for each pairwise comparison. Based on these findings, we conclude that reduced, denatured sfAFP is essentially disordered following reduction.

sfAFP retains some rigidity on ps/ns timescales following reduction

To assess the ps/ns flexibility of reduced, denatured sfAFP, we measured the {¹H}-¹⁵N NOE ratios for residues whose ¹H-¹⁵N signals are well resolved in the 2D ¹H-¹⁵N HSQC spectra. Whereas these ratios are high in the folded protein and even approach values indicative of complete rigidity (13), the values in the reduced, denatured protein are much lower (Fig. 2, A–C). Overall they are about halfway between values expected for fully stiff (0.85) and high flexibility. Similar {¹H}-¹⁵N NOE ratios were observed in three different conditions: pH 2.5, 25°C; pH 2.5, 5°C; and pH 6.15, 5°C. The N-terminal His-tag also appears to be somewhat more rigid than the rest of denatured sfAFP and the residues corresponding to the last PPII helix seem to be more flexible. Within the rest of the protein, i.e., residues 1–70, there seems to be a slight trend toward increased flexibility moving from the N-terminus to the C-terminus. This trend might be related to a lower content of potentially stabilizing charge-charge interactions in the C-terminal half of sfAFP.

Figure 2.

Fast ps/ns dynamics of reduced, denatured sfAFP. The per-residue {¹H}-¹⁵N ratios are shown for reduced, denatured sfAFP at: (A) pH 2.5, 25°C, (B) pH 2.5, 5°C, (C) pH 6.15, 5°C, and (D) pH 6.15, 5°C with an internal proteolytic cleavage. Shaded regions show the approximate positions of polar (cyan) and nonpolar (yellow) PPII helices in the folded protein. The gray line marks the trend as calculated with a weighting function. The uncertainties in these values are about ±0.015. To see this figure in color, go online.

In the presence of the proteolytic cleavage between residues 16 and 17, a dramatic decrease in the {¹H}-¹⁵N NOE values is seen in the nearby residues corresponding to the second PPII helix, i.e., the helix which contains the cleavage site, and a moderate decrease is also seen in the residues corresponding to the fourth PPII helix (Fig. 2 D). Considering that both PPII helices 2 and 4 harbor nonpolar residues, these results suggests that hydrophobic contacts in the denatured state might reduce the flexibility of the uncleaved, reduced, denatured sfAFP. Similar hydrophobic clusters in denatured proteins been reported previously (32).

Rigidity on μs/ms timescales increases upon cooling and is highest in the N-terminal polar segments

The longitudinal relaxation rates (R₁) and relaxation rates in the rotating frame (R_1ρ) were measured for reduced, denatured sfAFP under three different conditions: pH 2.5, 25°C; pH 2.5, 5°C; and pH 6.15, 5°C to assess mobility on the μs/ms timescales and the results are shown in Fig. 3. Overall, the rates are low, which is in line with the denatured character of the chain under reducing conditions. Some interesting trends can be observed. First, the residues of the N-terminal His-tag show relatively high R_1ρ rates, which may be due to a dearth of Gly residues. Secondly, the mobility, gauged most clearly by R_1ρ values, diminishes as the temperature is lowered from 25 to 5°C (Fig. 3, A versus B) and when the pH is increased from 2.5 to 6.15 (Fig. 3, B versus C). This increase is especially noticeable for the residues that correspond to PPII helices 1 and 3 in the folded protein. This suggests that the balance of electrostatic interactions may favor certain conformations at neutral pH, stiffening the chain. Thirdly, under all three conditions, the residues corresponding to the first PPII helix have higher R_1ρ rates, meaning they are more rigid, whereas the rates for the C-terminal residues are lower. Moreover, the residues which would constitute the polar PPII helices in the folded protein (Figs. S1 and 3), tend to be more rigid than those corresponding to the nonpolar PPII helices. The results obtained at pH 6.15, 5°C, which are approximately the physiological conditions that the nascent sfAFP chain would experience after synthesis, is suggestive of some innate rigidity in the residues corresponding to PPII helices 1 and 3 which may help guide nascent folding.

Figure 3.

R₁ and R_1ρ relaxation rates of reduced, denatured sfAFP. The per-residue {¹H}-¹⁵N longitudinal relaxation rates (R₁, open squares) and longitudinal relaxation rates in the rotating frame (R_1ρ, filled circles) are shown for reduced, denatured sfAFP at: (A) pH 2.5, 25°C, (B) pH 2.5, 5°C, and (C) pH 6.15, 5°C. Shaded regions show the approximate positions of polar (cyan) and nonpolar (yellow) in PPII helices in the folded protein. The gray line marks the trend as calculated with a weighting function. Error bars correspond to the per-residue uncertainties of the least-squares fits of an exponential equation to the data. To see this figure in color, go online.

Shifted pK_a values of Asp and His residues evince attractive electrostatic interactions in unfolded sfAFP

The measurement of pK_a values in denatured proteins can provide insight into residual structure (33). Here, the pK_a values of Asp and His residues were measured at 5°C by monitoring their ¹³Cβ chemical shifts in 2D ¹H-¹³C HSQC spectra at different pH values using a sample of reduced, denatured ¹³C-¹⁵N-sfAFP which lacked the N-terminal His-tag. Analysis of a 3D HNCA spectrum and an {¹H}-¹⁵N NOE experiment and ps/ns dynamics revealed a lack of the His-tag did not impact the backbone conformation and slightly increased the flexibility of the N-terminal residues (Fig. S6). The pK_a values were found to be 3.43 ± 0.09 for D5, D29, and D75 and 6.83 ± 0.03 for H8 and H32 (Fig. 4 A). These values are lower and higher, respectively than the values reported for Asp (3.9) and His (6.5) in short structureless peptides (34). This is evidence for the attractive electrostatic interactions between the Asp and His side chains in the unfolded ensemble (33).

Figure 4.

Electrostatic and long-range contacts in sfAFP. (A) Determination of His (left y axis, blue circles) and Asp (right y axis, red squares) pK_a values. The curves represent the least-squares fit of the Henderson-Hasselbach equation to the pH versus ¹³Cβ chemical shift data. (B) Per residue ¹H-¹⁵N peak integral ratios for paramagnetic: diamagnetic sfAFP-C1^∗ (green bars). Negative residue numbers (x axis) correspond to the N-terminal His-tag. To see this figure in color, go online.

No significant long-range contacts are detected by a PRE experiment

The perturbed pK_a values led us to consider the hypothesis that favorable electrostatic interactions among H8⁺···D5⁻···H32⁺···D29⁻ might position C1 at the N-terminus near C28 (in the loop between PPII helices 2 and 3) to favor nascent folding. To test this possibility, we performed a PRE experiment. Previously, low populations of close contacts in diverse protein systems have been detected by PRE experiments (35). For this test, a His-tagged sfAFP variant with C1 labeled with MTSL and the other three cysteines mutated to serine. As expected for the mutated and modified sites, the ¹H¹⁵N chemical shifts of this modified variant are similar to the WT protein (Fig. S7). The decreased signal intensity due to the PRE effect was found to vary with the distance along the sequence of protein and no special weakening is observed for residues corresponding to the loop between PPII helices 2 and 3 (Fig. 4 B). No alternative conformational states were detected by relaxation dispersion NMR spectroscopy. Relaxation dispersion NMR spectroscopy is a powerful approach for characterizing conformational exchange processes (36) and has been utilized to identify atoms involved in early folding processes (37). Therefore, it could provide evidence for or against the participation of the N-terminus of sfAFP and the loop connecting PPII helices 2 and 3 in preferred conformations. The R_2eff rates for all assigned, nonproline residues in sfAFP are near 5 s⁻¹ and show little dependence on the CPMG rate, as illustrated in Fig. S8 for nine representative residues and summarized in the last plot showing the R_2eff rate:CPMG rate slope for each analyzed residue. These measurements strongly suggest that the protein does not undergo conformational exchange between distinct states on μs/ms timescales.

Discussion

Research on the sfAFP is adding valuable understanding to the folding and conformational stability of proteins composed by PPII helical bundles. Our study builds on investigation from the Sosnick lab, who used folding kinetics, SAXS, and computational modeling of disulfide-bond intact (i.e., -S–S-) sfAFP to shed light on the energetics and folding mechanism of sfAFP (15), and on our NMR and computational energetics study of disulfide-bond intact, folded sfAFP (13), which elucidated the native state dynamics and additional stabilizing contributions. These two studies drew a few different conclusions; namely, whereas Gates et al. (15) reported that conformational entropy of sfAFP’s folded state is higher than that of typical proteins composed of α-helices or β-sheets, Treviño et al. (13) reported NMR relaxation measurements suggesting that folded sfAFP is just as stiff as other folded proteins. Moreover, while Gates et al. (15) reasoned that the stabilizing contribution of the hydrophobic effect is low, Treviño et al. (13) uncovered that sfAFP buries some nonpolar surface area through dimerization. On the other hand, and more importantly, both studies concurred that Cα-H···O=C H-bonds also contribute to native state stability. Gates et al. found that the high Gly content does not make the unfolded ensemble of sfAFP especially condensed but that this ensemble is biased toward PPII conformation (15). This conformational bias might be related to the PPII CD spectral features and the moderate stiffness found here for the reduced, denatured state on ps/ns timescales. Finally, Gates et al. reasoned that the conformational entropy change during folding (ΔS_conf·F) is higher for glycine residues due to its more flexible backbone. However, since glycine residues do not have a side chain, they pointed out that this decreases the ΔS_conf·F (15). The sum of these two effects is that glycine residues’ contribution to the ΔS_conf·F is similar to that of non-Gly residues.

Our results, obtained under physiologically relevant reducing conditions corroborate the model for sfAFP folding with disulfide bonds present proposed by Gates et al. (15). As they advanced, proline residues could favor turn forming, preorganizing the chain into segments corresponding roughly to the PPII helices. Whereas aspartic acid residues are neutral at pH 3, they are predominantly charged at pH 6 where our relaxation measurements found increased rigidity. These findings and the perturbed Asp and His pK_a values measured here, lead us to additionally suggest that electrostatic interactions formed by K2, D5, H8, D29, and H32 are present at pH 6 and might aid the juxtaposition of C1 and C28, favoring disulfide bond formation. However, the PRE and relaxation dispersion NMR experiments revealed no detectable populations of conformers that place the N-terminus and the loop connecting PPII helices 2 and 3 close together under the reductive denaturing conditions used here. Future NMR experiments under conditions that favor the denatured chain to fold are required to better characterize the folding pathway.

Here, we found a moderate stiffness seen in reduced, denatured sfAFP, which strongly suggests that the unfavorable ΔS_conf·F will be lower than if the protein were more flexible under denaturing conditions. It would be interesting to precisely measure this ΔS_conf·F for sfAFP and compare it with values for typical proteins with an average glycine residue content and composed of α-helices and/or β-sheets. However, a quantitative calculation for ΔS_conf·F based on NMR relaxation data faces important limitations, as discussed previously (38,39). Nevertheless, the hNOE, R₁ and R_1ρ values measured here are similar to those reported previously under denaturing conditions for representative proteins, such as hen egg white lysozyme (16), CheY (40), and Staphylococcal nuclease (41) or protein domains, such as an SH3 domain (42). Taken together with the similar rigidities of the folded states of typical glycine-poor, α, β, α+β proteins, and sfAFP (13), this suggests that the ΔS_conf·F of all these proteins would be similar despite the high glycine content of the latter. This means that the free energy “cost” paid by the glycine-rich PPII helical bundle proteins for losing ΔS_conf·F is likely similar to that paid by typical proteins. Therefore, glycine-rich PPII helical bundle proteins should not require an exceptionally robust set of stabilizing interactions to fold and be stable. A “normal” set typical of an “average” α/β protein should do. This finding should facilitate the de novo design of PPII helical bundles.

Conclusion

In conclusion, our results show that reduced sfAFP is fully denatured upon reduction of its two disulfide bonds. Despite a high proportion of glycine residues, its denatured ensemble is not more flexible than those of typical proteins composed of α-helices and/or β-sheets. The insight will help guide the design of new proteins composed of PPII helical bundles.

Author contributions

M.Á.T., M.M., and D.V.L. designed the research. M.Á.T., M.R.M., R.L.S., and D.V.L. performed the research. M.Á.T. and D.P.-U. contributed analytical tools. M.R.M., R.L.S., D.P.-U., and D.V.L analyzed the data. M.Á.T., R.L.S., D.P.-U., M.M., and D.V.L. wrote paper. R.L.S. acknowledges the FPI contract.

Editor: Bernd Reif.

Supporting citations

References (19,43,44,45) appear in the supporting material.