Solvent-Exposed Salt Bridges Influence the Kinetics of α-Helix Folding and Unfolding

Salt bridges are known to play an essential role in the thermodynamic stability of the folded conformation of many proteins, but their influence on the kinetics of folding remains largely unknown. Here, we investigate the effect of Glu-Arg salt bridges on the kinetics of α-helix folding using temperature-jump transient-infrared spectroscopy and steady-state UV circular dichroism. We find that geometrically optimized salt bridges (Glu– and Arg+ are spaced four peptide units apart, and the Glu/Arg order is such that the side-chain rotameric preferences favor salt-bridge formation) significantly speed up folding and slow down unfolding, whereas salt bridges with unfavorable geometry slow down folding and slightly speed up unfolding. Our observations suggest a possible explanation for the surprising fact that many biologically active proteins contain salt bridges that do not stabilize the native conformation: these salt bridges might have a kinetic rather than a thermodynamic function.


Steady-state CD and FTIR spectroscopy
Circular dichroism (CD) spectra were collected using an Olis DSM-1000 CD Spectrophotometer equipped with a Julabo (CF31) temperature controller. All CD measurements were obtained at peptide concentrations of 40 µM in 20 mM phosphate buffer (pH= 6.8 or pH= 2.5) using 2-mm quartz cuvettes. The ellipticity is reported as mean residue ellipticity (θ ), and the relative helicity in each peptide is determined by monitoring the ellipticity at 222 nm. The temperature-dependency of the ellipticity was monitored for each of the peptides between 190−257.5 nm with incremental steps of 0.75 nm and averaging over 5 s of data acquisition time (0.15 nm/min scanning speed).
The set point of the water bath is software-controlled and the actual sample temperature was continuously measured using an external temperature probe inserted into the cuvette holder. Global fitting with singular value decomposition (SVD) of the datasets collected as function of temperature were performed using the Olis GlobalWorks software package. All datasets were fitted with a two-state model (F U). We find that we can describe the data by assuming no change in the specific heat upon folding (∆C p = 0), as is to be expected since the solvent exposure does not change significantly upon helix formation. 1 The fit function used to determine ∆H UF and T m from the SVD of the UV-CD data is given by 1 where

Molecular Dynamics Simulations
The sequence of the (i + 4)ER peptide was modeled as an ideal α-helix using Chimera, 2 and capped with acetate at the N-terminus and methylamide at the C-terminus. Glutamate residues were deprotonated and arginine residues were protonated. This structure was solvated in a periodic dodecahedron box with a volume of 97.29 nm 3 , followed by the addition of 3143 water molecules. To meet experimental conditions and ensure electrostatic neutrality of the system, 1 Na + ion and 1 Cl − ion were added, representing a concentration of 20 mM NaCl. In total, the system contained 9670 atoms. Interactions between atoms were described by the AMBER99-SB force field, 3  between the glutamate and arginine within the same repeat and with a distance between the Glu-C δ and Arg-C ζ atom less than 0.8 nm. Excluding the first 10 ns of each simulation, two-dimensional probability distributions were calculated of hhb and nsb, with a binsize of 1, and contoured at intervals of 1 %. Snapshots were visualized with VMD. 10

Time-resolved T -jump IR experiments
Laser-induced T -jumps of 6 to 8 K were obtained by nanosecond excitation of the OD-stretch overtone of D 2 O using the same optical setup as described previously. 11 In brief, the T -jump pulse is generated using a beta barium borate ( To obtain transient IR spectra, we determine absorption changes where T and T 0 are the transmission of the IR-probe pulse in the presence and absence of the T -jump pulse, respectively, as described previously. 12 We correct for small pulse-to-pulse fluctuations in the intensity of the IR-probe pulses by simultaneously measuring the intensity of a reference pulse (split off from the probe pulse before the sample) that passes through the sample in an area that is not influenced by the T -jump pulse. Transient absorption changes are measured using frequency-dispersed detection of the probe and reference pulses using a liquid-nitrogen-cooled 2×32 HgCdTe (MCT) array detector (Infrared Associates).
Each delay value is the average of approximately 1000 T -jumps. Given the repetition frequency of the T -jump laser (20 Hz), care is taken to ensure that the sample has relaxed back to its initial temperature before the subsequent T -jump. The initial temperature of the sample was controlled using a thermostatted cell-holder with a circulating heat bath and was calibrated with an IR camera (FLIR ThermaCAM E2 linear response with temperature in the amide I' spectral region. 13 2 Additional Data

Temperature-dependent UV-CD measurements
The helix-coil transition of the peptides was analyzed in thermal equilibrium using CD spectroscopy. UV CD spectra were collected between 190−257.5 nm in a temperature range of 260−346 K under both neutral and acidic pH conditions (Fig. 1). The dependence on temperature of the mean residue ellipticity detected at 222 nm for each of the peptides at neutral and acidic pH is shown in Fig. 2. The thermal unfolding curves have a sigmoidal shape, and from a singular value decomposition of the temperature-dependent CD spectra and global fitting we find that the folding is well described by a two-state model. The melting temperatures T m and the unfolding enthalpy changes ∆H are listed in Table 1. The observed helicity and thermodynamic stability obtained at neutral pH ( Fig. 2a) follows the trend indicating that the α-helix-stabilizing effect is largest for ER-oriented salt bridges in which E and R are spaced four peptide units apart, as was reported previously. 14 To confirm that the salt bridges provide (de)stabilization to the α-helical structure, the saltbridge formation was inhibited by protonating the Glu side-chains at acidic pH (Fig. 2b). Under these conditions, the relative helicity and stability of the peptides is (i + 4)ER ≈ (i + 4)RE > (i + 3)ER ≈ (i + 3)RE, which agrees with previously reported results. 14 The dichotomy observed between the (i, i + 4) and (i, i + 3) spaced peptides in absence of salt bridges probably arises from the fact that the helix-forming propensity of amino acids may depend slightly on their position in the sequence of a peptide. Comparison of the melting curves at neutral and acidic pH ( Fig. 2a and   b) shows that the relative α-helix content and thermodynamic stability of the ER-type peptides decrease due to disruption of the favorable salt-bridge interaction at acidic pH. In contrast, the RE-

Unfolding in thermal equilibrium monitored by UV-CD and FTIR spectroscopy
In Fig. 3 we compare the temperature-dependency of the mean residue ellipticity at 222 nm monitored using UV-CD and the temperature-dependent FTIR response at 1630 cm −1 (decrease of the α-helical population, (a)) and 1658 cm −1 (increase of the random coil population, (b)) of peptide (i + 4)ER at neutral pH. Both the UV-CD and FTIR thermal unfolding curves show a sigmoidal transition, reflecting the temperature-induced conformational changes of the peptide in thermal equilibrium. From least-squares fits of the temperature-dependent FTIR data at both frequencies to a two-state model, 1 we obtain a melting temperature T m of 310.5 ± 2.5 K, which is in quantitative agreement with the thermal melting point of 312.9 ± 1.2 K (see Table 1) obtained from SVD and global fitting analysis of the temperature-dependent CD data. Therefore, our static temperature-dependent CD and IR data further support that the thermal unfolding of this peptide can be described by an effective two-state model. 1658 cm −1 (increase of the random coil population). The solid curves represent least-square fits to a twostate model. 1 In order to compare the UV-CD melting curve with the temperature-dependent FTIR response detected at 1658 cm −1 , the y-axis representing the θ 222 response in (b) is reversed compared to (a).

Effect of intrinsically preferred side-chain rotamers on folding thermodynamics
The differences in thermodynamic stability (the free-energy difference ∆G UF between the unfolded and folded states) of the four investigated peptides are in agreement with previous results, 14 and probably arise mostly from sterical effects. The preferred Glu and Arg rotameric states (which depend on the backbone conformation) may allow salt-bridge formation in one orientation of Glu and Arg, but not in the other, causing a difference in the helix stability of the peptides with ER and RE oriented residues. To gain more insight into the effect of rotameric states on the thermodynamic stability of the investigated peptides, we determine the number and relative probability of the accessible side-chain χ rotamers that allow the formation of Glu−Arg salt bridges using Richardson's rotamer library 21 implemented in Chimera. 22 For this purpose, we have performed a systematic search to determine which intrinsically preferred side-chain conformations (χ angles) of Glu and Arg in an α-helical backbone conformation are such that salt-bridge formation is possible, i.e., are such that the distance between at least one Glu oxygen and one Arg nitrogen atom is between 2.70 and 3.00 Å. The center value of this distance range was chosen based on X-ray measurements on a Glu − −Arg + model compound, from which it was found that the distances between the oxygen and nitrogen atoms of the two Glu − −Arg + hydrogen bonds in an optimal salt bridge are 2.81 Å and 2.89 Å. 23,24 Sterically forbidden conformations were excluded, and only rotameric states with probability ≥ 0.5% were taken into account.
For the α-helical backbone conformation of peptide (i+4)ER we find three different accessible rotameric isomers in which the relative orientation of the intrinsically preferred side-chain conformations of Glu and Arg is such that salt-bridge formation is possible (Fig. S6). For the α-helical backbone conformation of the thermodynamically less stable peptides (i + 3)ER, (i + 4)RE, and (i + 3)RE, no (sterically unhindered) rotameric states that allow the formation of a salt bridge are significantly populated according to the rotameric library, 21 suggesting that the formation of salt bridges in the folded states of these peptides requires deviation of the χ angles from their intrinsically preferred χ angles, and/or of the backbone φ , ψ angles from their ideal α-helical values. In either case, the helix-stabilizing effect of the salt bridges will be less. Thus, the observed differences in thermal stability between the different peptides probably originate from differences in their geometric ability to form (intrinsically preferred) low-energy side-chain rotamers that allow saltbridge formation. In addition, the salt-bridging residue pairs in the folded state of peptide (i+4)ER possibly enjoy a larger degree of freedom due to the ability to occupy multiple intrinsically preferred salt-bridging Glu−Arg rotamer pairs, and contributes entropically to the free-energy balance of the peptide. 16 . We find that both the α-helical conformation and native Glu−Arg salt bridges are highly populated at low temperature (the distribution peaks at nsb = 2). On the other hand, at high temperature both the α-helical structure and Glu−Arg interactions are hardly populated (nsb peaks at ≈ 0). We therefore conclude that the salt bridges indeed break upon thermal unfolding.

T -jump control measurements
To verify that the origin of the kinetic phase arises from the folding/unfolding re-equilibration of the peptides, we measure the transient absorption changes of an aqueous solution of N-methylacetamide (NMA, which contains a single trans-peptide group) in response to a nanosecond T -jump (Fig. 6).
As can be seen in Fig. 6b, the transient absorption changes of NMA shows an instantaneous Tjump response, which is very similar to the instantaneous response as observed for the different peptides (see Fig. 2). However, the T -jump relaxation of NMA does not exhibit a kinetic phase, and we conclude that the observed kinetic phase as observed for the peptides arises from the conformational redistribution process of the peptides during re-equilibration.

Infrared spectrum as a probe of the Glu protonation state
To probe the α-helix content we examine the IR response of the peptides, see Fig. 7. The amide I' absorbance of proteins and peptides, mainly originating from backbone C=O stretch vibrations, is a sensitive probe of secondary structure and can be directly related to the backbone conformation. 25 In particular, amide residues in an α-helix absorb at ∼1635 cm −1 , whereas those in a random coil absorb at 1645 cm −1 . 26 The differences in helicity between the four peptides as derived from the amide I' maximum in the low-temperature equilibrium FTIR spectra (Fig. 7) are in agreement with those derived from the CD results (see Fig. 1). The smaller peaks at 1585 and 1607 cm −1 originate from the symmetric and antisymmetric stretch vibration of the guanidinium (CN 3 H 5 + ) group of Arg + . 26 The band observed at 1565 cm −1 in the FTIR spectra at neutral pH arises from the COstretching mode of the COO − group of Glu − (Fig. 7a), and shifts to 1705 cm −1 upon protonation 26 (Fig. 7b). The absence of the COO − peak at pH = 2.5 confirms that the carboxylate groups of the Glu side chains are completely protonated at this pH, making salt-bridge formation impossible. Figure 7: Normalized low-temperature equilibrium FTIR spectra in the amide I' region of each of the peptides obtained at 278 K under (a) neutral and (b) acidic pH conditions. The spectral feature arising from the ν as (C=O) mode of the COO − group of Glu − observed at 1565 cm −1 in the FTIR spectra at neutral pH (a) shifts to higher frequency upon protonation and appears at 1705 cm −1 (the ν(C=O) mode of the Glu COOH group) 26 in the FTIR spectra obtained under acidic pH conditions (pH * = 2.5). This unambiguously demonstrates that salt-bridge formation at a pH value of 2.5 is prevented due to protonation of the carboxylate groups of Glu. The insets compare the amide I' center frequencies.

Two-state folding behavior at all temperatures, pH values, and IR frequencies
The T -jump-induced re-equilibration between α-helical and random coil ensembles was probed at a wide range of temperatures in the amide I' spectral region. As can be seen in Fig. 8, all transients show single-exponential T -jump relaxation kinetics. The detection of single-exponential T -jump relaxation kinetics with equal relaxation rates observed at 1630 cm −1 and 1658 cm −1 at all final T -jump temperatures and at both the investigated pH values indicates that the dynamics of the helix-coil transition can be described by an effective two-state model, 27,28 and that the αhelical and random-coil state ensembles can be described as two broad free-energy minima in the conformational free-energy landscape, which are separated by a single kinetic barrier. [29][30][31] This is in agreement with previous studies that showed that the ensemble dynamics of the helix-coil transition can be well approximated using two-state analysis. 29-37

Complete data set underlying Figures 5 and 6 of article
We find that the effective folding and unfolding rates (k F,eff and k U,eff ) of each of the peptides exhibit Eyring temperature-dependence, and from least-squares fits of the data to the Eyring equation we obtain the effective transition enthalpy (∆H ‡ eff ) and entropy (∆S ‡ eff ), both for the folding (U → F) and unfolding (F → U) transitions. The results are listed in Table 1 (folding) and Table 2 (unfolding). (d) (i + 3)RE, measured at neutral and acidic pH. The effective folding and unfolding rate constants (k F,eff and k U,eff ) were obtained from the T -jump relaxation rates (k R ) and the folding equilibrium constants (K eq ) according to a two-state analysis. The straight lines are linear fits.

Correlation between free-energy barriers and equilibrium free-energy difference
We find a correlation between the equilibrium free-energy difference (∆G UF ) and the free-energy barriers for the folding and unfolding processes (∆G ‡ U→F and ∆G ‡ F→U , respectively), see the Figure   below.