Spectroscopic studies of protein folding: Linear and nonlinear methods

Abstract

Although protein folding is a simple outcome of the underlying thermodynamics, arriving at a quantitative and predictive understanding of how proteins fold nevertheless poses huge challenges. Therefore, both advanced experimental and computational methods are continuously being developed and refined to probe and reveal the atomistic details of protein folding dynamics and mechanisms. Herein, we provide a concise review of recent developments in spectroscopic studies of protein folding, with a focus on new triggering and probing methods. In particular, we describe several laser-based techniques for triggering protein folding/unfolding on the picosecond and/or nanosecond timescales and various linear and nonlinear spectroscopic techniques for interrogating protein conformations, conformational transitions, and dynamics.

Introduction

Despite extensive study over the past few decades, the protein folding problem¹ continues to inspire new research directions. This is due not only to the sheer complexity of the process of protein folding, which makes it extremely difficult to achieve a mechanistic, quantitative, and predictive understanding of how proteins fold, but also to the fact that it provides an excellent test-bed for the development and validation of new computational and experimental methods for biology, biochemistry, and biophysics. Herein, we present a concise review of recent advances in the application of spectroscopic methods to investigations of protein folding, with a focus on fluorescence and vibrational spectroscopies, as well as conformational triggering methods. Although dynamic information can sometimes be revealed or inferred by static spectroscopic measurements, for most ensemble kinetic experiments, a suitable triggering event is required to start the “reaction” of interest; in addition, one or more physical observables or probes are used to follow the progression of this reaction. For protein folding studies, it is apparent that the choice of the triggering method determines the time resolution of the observable kinetics, whereas the conformational probes used dictate the structural resolution of the experiment. Below, we will first discuss available protein folding and conformational triggering methods, with a focus on those that offer nanosecond or faster time resolutions, and then will review spectroscopic methods that offer varying degrees of structural resolution and sensitivity when applied to probe protein folding/unfolding equilibria and kinetics.

Triggering Methods

The free energy of a given protein in solution depends on the temperature, pressure, solvent composition, and other external constraints if they exist. Thus, the thermodynamic equilibrium between its folded and unfolded states can be manipulated by simply changing one of these factors. As such, all current protein folding triggering methods rely on this principle, whereby the system in question is rapidly (compared with the kinetic events of interest) shifted to a nonequilibrium state. The subsequent relaxation kinetics towards the new thermodynamic equilibrium, which could be probed by a wide variety of techniques, can be dominated by either folding, unfolding, or both, depending on the final condition of the protein solution. For example, a folding or unfolding process could be initiated by rapid mixing of two solutions via either the stopped-flow or continuous-flow techniques.^{2, 3} Although the majority of previous folding kinetic studies used such mixing-based triggering methods, a comprehensive review of these subjects is beyond the scope of this paper. Below we only discuss triggering methods that rely on the application of short (picosecond–nanosecond) laser pulses, which may involve direct excitation of chemical groups in the protein, resulting in chemical changes that leave the protein backbone in a nonequilibrium state (e.g., by removal of a conformational constraint), or photoinduced changes of physical and/or solution properties (e.g., temperature or pH) via excitation of solvent or solute molecules.

Laser-induced temperature-jump

The ultrafast (picosecond or nanosecond) temperature-jump (T-jump) method relies on the fact that in condensed phases vibrational relaxations typically take place on the picosecond timescale,⁴ thus allowing a burst of photons to rapidly heat up the solution of interest through excitation of solvent or solute^{5, 6} vibrations. For aqueous solutions, a rapid T-jump can be achieved by exciting water's near-infrared (IR) overtone transitions in the vicinity of either 1.5 μm (for H₂O solution) or 2 μm (for D₂O solution) using either nanosecond (ns) or picosecond (ps) laser pulses.^7–12 The amplitude of the T-jump depends on the pump energy and excitation volume and can be determined from solvent absorbance using a calibration curve.⁷ Because the thermal energy initially concentrated in the laser excitation volume will gradually dissipate into the surroundings, the useful time window of this technique in which the temperature remains approximately constant ranges from a few to tens of milliseconds (ms), depending on the optical setup and sample holder. Because the T-jump technique does not require any specific protein modifications, it is, therefore, the most commonly used triggering method in the study of ultrafast protein folding events.^13–23

Azobenzene isomerization

The reversible cis/trans photoisomerization of azobenzenes has also been demonstrated to have the capability to control peptide and protein conformations and thus has found utility in folding kinetic studies.^24–29 A thorough review on the use of azobenzene and derivatives as photoswitches in biomolecules can be found elsewhere.³⁰ Simple azobenzene compounds photoisomerize on the ps timescale and, therefore are suitable for studying the early events in protein folding. Additionally the azobenzene cis/trans isomers exhibit well separated spectral features, allowing for independent excitation of either isomer and thus reversible switching of the trigger. As shown first by Kumita et al.,²⁶ a reversible change in helicity of up to 36% can be obtained on photoisomerization of the azobenzene cross-linker in an alanine-based peptide. More recent work has been done to understand how such linkers affect peptide backbone conformations,^{31, 32} and to probe the dynamics and mechanism of the helix–coil transition,^33–35 among other applications.^36–38

Photodissociation of a cage compound or disulfide bond

Another method of conformational switching is to use a relatively large molecular moiety (i.e., a cage) to force the protein of interest to unfold under native conditions; on cleavage of the cage by excitation with an appropriate laser pulse, the protein will fold to its native conformation. Thus, several photocages, such as benzoinyl cages, have been used to trigger protein folding kinetics.^39–43 Photocaged molecules can also be used in a cyclization scheme to produce a photocleavable intramolecular linker. In some cases, it is desirable to impose a well defined structural constraint on the protein molecule, enabling the folding/unfolding reaction to start from a unique position on the folding energy landscape on removal of the conformational constraint. For example, this approach has been applied to the study of a mutant of the small α-helical villin headpiece subdomain (i.e., cVHP-34 M12C), in which an internal cysteine residue and a 3′-(carboxymethoxy)benzoin (CMB) molecule, which is covalently attached to the N-terminus of the polypeptide, have been linked to restrict the conformation in the unfolded state. On illumination with a 355 nm laser pulse, this linker molecule is irreversibly cleaved with a yield of 0.6–0.7; the loop is therefore broken to yield the linear protein that is free to refold. The time resolution of this method is determined by the rate of photolysis of the linker, which is faster than one nanosecond.⁴⁴ However, it is worth mentioning that if the photocage has to diffuse out of its initial position for the protein to fold, the time resolution of such cage cleaving methods would be determined by the rate of this diffusion process.

One convenient structural constraint commonly found in proteins is the disulfide bond formed between two thiol-containing amino acid sidechains. Lu et al.⁴⁵ have shown that a disulfide cross-linker can be transiently broken by an ultraviolet (UV) laser pulse, making it potentially useful as a phototriggering method in protein folding studies. However, in practice this method is less useful as the resultant thiyl radicals undergo rapid geminate recombination, which greatly reduces the population of protein molecules that will eventually undergo a conformational transition. Recently, Hamm and coworkers^{46, 47} have shown that for small peptides this problem can be alleviated to some extent due to the increased strain exerted on the disulfide bond by cross-linking two cysteines that are close in sequence. A more promising method, which is currently under investigation in our laboratory,⁴⁸ however, is to use a photoinduced electron transfer (PET) reaction to permanently break the disulfide cross-linker of interest. Cleavage of disulfide bonds via light-induced electron transfer from adjacent tryptophan residues has been observed in various proteins^49–54 suggesting that it could be engineered to control protein/peptide conformations using light.

Other phototriggering methods

Besides the methods discussed above, other avenues for generating high free energy conformations in proteins, such as electron transfer^{55, 56} and ligand dissociation,^{57, 58} have also found important applications, some of which have been thoroughly reviewed in the past.^{59, 60} More recently, Tucker et al.⁶¹ have shown that disulfonyl-tetrazine could be used as an ultrafast phototrigger of protein conformations because its photodissociation, which yields N₂, occurs on the ps timescale with a reasonably high yield.

Probing Methods

The greatest challenge in the study of the molecular mechanism of protein folding is that none of the existing experimental techniques are capable of capturing conformational snapshots of the protein in question along its “folding coordinate” with both high spatial and temporal resolutions. As such, a wide variety of spectroscopic techniques have been developed and used, some of which are continuously being refined, to probe the fine detail of certain regions or features of the underlying folding free energy landscape. Although all of these techniques are worthy of discussion, below we will focus on linear and nonlinear methods based on the application of fluorescence and vibrational spectroscopies. The spectroscopic properties of a particular sample are determined by its polarization, , induced by the electric field of an incoming light beam, E(t), which can be expressed as,⁶²

(1)

where is the nth order response function of the system. If the electric field is sufficiently weak so that only the first term (i.e., ) in Eq. (1) needs to be considered to describe the underlying light-matter interaction, various linear spectroscopies, such as absorption spectroscopy and fluorescence excitation, result. In contrast, nonlinear spectroscopic techniques require consideration of higher order terms (i.e., n > 1) and typically involve application of ultrafast laser pulses.

Native fluorescence

Three aromatic amino acids, that is, tryptophan (Trp), tyrosine (Tyr), and phenylalanine (Phe), are fluorescent and thus could be used to probe protein conformational changes. In practice, however, only Trp fluorescence is extensively used in protein folding studies due to its sensitivity to the environment and its larger quantum yield. Because many reviews^63–67 on Trp fluorescence and its application to protein conformational and folding studies exist in the literature, below we focus the remainder of our discussion on fluorescent methods utilizing nonnative fluorophores.

Nonnative fluorescence

Recently, we have shown that the unnatural amino acid, p-cyanophenylalanine (Phe_CN), is a very useful fluorescent probe of protein folding and binding.⁶⁸ The absorption spectrum of Phe_CN exhibits two broad bands, one centered around 240 nm and the other one centered around 280 nm, whereas its fluorescence spectrum peaks at ∼295 nm and is insensitive to excitation wavelength and solvent.⁶⁹ However, the fluorescence quantum yield (Q_F) of Phe_CN is sensitive to its immediate environment (e.g., in water Q_F = 0.11, whereas in acetonitrile Q_F = 0.03), making it an ideal fluorescent probe of protein folding and binding transitions.^{70, 71} For example, using this probe, Tang et al.⁷² have investigated the binding/insertion/dimerization kinetics of a transmembrane peptide. As shown (Fig. 1), their results showed that when placed at the same position Phe_CN is a more sensitive fluorescent reporter of the insertion and dimerization processes than Trp.

Figure 1.

Stopped-flow kinetics of a transmembrane peptide on association with a model membrane probed via Trp (black) and Phe_CN (blue) fluorescence, with the latter exhibiting greater sensitivity to the key kinetic events. (Adapted from Tang J, et al., (2009) J Am Chem Soc 131:3816–3817, with permission from American Chemical Society.)⁷²

Moreover, Serrano et al.⁶⁹ have shown that the fluorescence of Phe_CN exhibits single-exponential decay kinetics with a lifetime (τ_F) of 5–8 ns, depending on the solvent. In particular, they found that for a series of protic solvents τ_F increases with increasing hydrogen bonding strength of the solvent, due to a decrease in the nonradiative rate of the excited state. These findings further indicate that when combined with time-resolved lifetime measurements (see below), Phe_CN fluorescence could be used to probe the conformational distribution of proteins under native or nonnative conditions.

Fluorescence resonance energy transfer

Modulation of the fluorescence intensity of a fluorophore of interest by another molecular moiety (or quencher) through various mechanisms⁷³ has been widely used as a means to measure the proximity of the fluorophore to the quencher. For protein molecules, such modulations have been used to report on conformations and conformational changes as a function of various physical variables (e.g., time, temperature, or denaturant concentration). When fluorescence quenching arises solely from fluorescence resonance energy transfer (FRET), distance information may be obtained. According to Förster theory,⁷⁴ the mean FRET efficiency of a fluctuating biopolymer, , can be expressed as:

(2)

where r is the separation distance between the donor and acceptor, P(r) is the equilibrium distribution function of the donor–acceptor separation distances, and R₀ is the Förster radius⁷⁵ defined as:

(3)

where N_A, η, and Q_D are the Avogadro constant, the refractive index of the solvent, the quantum yield of the donor, respectively, whereas κ describes the relative orientation between the transition dipole moments of the donor and acceptor. J is the overlap integral of the donor fluorescence spectrum and acceptor absorption spectrum.

FRET measurements can be used to determine folding/unfolding thermodynamics. For example, for a simple two-state system, it can be easily shown that the mean energy transfer efficiency as a function of a denaturation variable, x (e.g., temperature, urea concentration, etc.), is

(4)

where and are the denaturant dependent transfer efficiency for the folded and unfolded state, respectively, and K_eq is the equilibrium constant for unfolding. However, in practice caution must be taken when applying FRET to extract any quantitative information from the experimental data as the R₀ of the FRET pair used may show a strong dependence on x and, in many cases, the accurate determination of is rather difficult.⁷⁶

Recently, we suggested a simpler approach⁷⁷ to determine folding/unfolding thermodynamics from FRET denaturation data based on the ratio of the integrated areas of the donor and acceptor fluorescence spectra in the case where the donor quantum efficiency is invariant as a function of unfolding, that is,

(5)

where r is the donor-to-acceptor separation distance, Q_D is the quantum efficiency of the donor, and Q_A,f and Q_A,u are the quantum efficiencies of the acceptor in the folded and unfolded states, respectively. The above expression can be further simplified if it is assumed that P(r) = δ(r) holds for both the folded and unfolded state ensembles, yielding two distance parameters r_f and r_u. Provided that the acceptor can be selectively excited in the presence of the donor, the normalized fluorescence quantum yield of the acceptor as a function of the denaturant (e.g., urea) can be determined:

(6)

where F₀ is the integrated fluorescence intensity of the acceptor in the absence of denaturant. The quantum yield of the acceptor and the folding free energy are furthermore assumed to be linear functions of denaturant concentration:

(7)

(8)

The six fitting parameters (i.e., ΔG°(0), m, , r_f, r_u, and q) in Eqs. (5–(8) are used to globally fit the measured curves of F_DA/F_AD and F_Denaturant/F₀ as a function of urea concentration x. As shown (Fig. 2), using this approach Glasscock et al.⁷⁷ showed that the free energy of unfolding as well as the donor-to-acceptor separation distance in the folded state of the LysM domain were in good agreement with those found by alternative spectroscopic methods,^{77, 78} even under the simplifying assumption that P(r) is a delta function in the unfolded state.

Figure 2.

(a) Phe_CN-Trp FRET data obtained with the LysM domain (pdb code: 1E0G) and (b) the corresponding F_DA to F_AD ratios at different urea concentrations. The locations of the Phe_CN (blue) and Trp (orange) are shown in the inset of (b). Also shown in (b) is the normalized fluorescence intensity of Trp, measured on direct excitation at 290 nm. Smooth lines are global fits to these data according to the methods discussed in the text, yielding the following parameters: r_f = 13.0 Å, r_u = 16.4 Å, and ΔG⁰(0) = 1.4 kcal/mol. The latter is comparable to that (1.2 kcal/mol) determined via CD spectroscopy.⁷⁷ (Adapted from Glasscock JM, et al., (2008) Biochemistry 47:11070–12076, with permission from American Chemical Society.)

Another interesting application of FRET is its use in extracting chain dynamics of biopolymers. For such studies, however, it is most useful to choose a donor with a long fluorescence lifetime, as backbone motions that move the FRET pair closer or farther apart modulate the energy transfer rate, k_T, according to:

(9)

where τ_D is the donor fluorescence lifetime in the absence of acceptor. Thus, long-range intrachain diffusive motions lead to complex, nonsingle-exponential decay of the donor fluorescence. The time dependent probability of finding a donor in its excited state at a distance r from the acceptor can be described by the following equation^{79, 80}:

(10)

with The first term in Eq. (10) represents the distance dependence of the FRET signal, whereas the second term accounts for intrachain diffusive motions with a diffusion coefficient D_AD. To probe both the equilibrium distance distribution, P(r), and D_AD for flexible biopolymers, one can measure the donor fluorescence decay while varying one or more properties such as τ_D or the solvent viscosity. For example, by using the naphthalene/dansyl and pyrene/dansyl FRET pairs, Kiefhaber and coworkers⁸¹ used the latter approach to probe the end-to-end distance distribution and chain dynamics of poly(Gly–Ser) chains. The fluorescence lifetimes of naphthalene (∼36.9 ns) and pyrene (∼225.5 ns) dyes in water are significantly different, but their Förster radii are very similar, with 23.3 and 20.5 Å for the naphthalene/dansyl and pyrene/dansyl FRET pairs, respectively. Thus, by fitting the experimental data to Eq. (10) and using a model for P(r), information about both the distance distribution and D_AD were determined.

Although FRET pairs based on fluorescent dyes have been extensively used in biological and biophysical studies, applying them to studies involving natively folded proteins could be problematic as they tend to be large, thus greatly destabilizing and possibly altering their folded states. Thus, we and others have developed a range of FRET pairs that are amino acid-like in nature, and thus minimally perturb the energetic and structural properties of the protein and peptide systems in question.^{68, 71, 82–85} As shown (Table I), these FRET pairs are mostly analogs of amino acid sidechains, making them ideally suited for probing conformational distributions and transitions of peptides and small proteins, as well as conformational changes occurring over a relatively short distance. Moreover, and perhaps more importantly, these FRET pairs could be used as a starting point for constructing multi-FRET systems, allowing more structural information to be extracted from a single FRET experiment. For example, Rogers et al.⁸⁴ have shown that Phe_CN, Trp, and 7-azatryptophan (7AW) can be used together to form a two-step FRET system and have utilized it to investigate the denaturant induced unfolding of the WW domain.

Table I.

Unnatural Amino Acid-Based FRET Pairs

Donor	Acceptor	R0 (Å)	Ref.
p-Cyanophenylalanine	Tryptophan	16.0	68
p-Cyanophenylalanine	Tyrosine	12.0	71
p-Cyanophenylalanine	5-Hydroxytryptophan	18.5	84
p-Cyanophenylalanine	5-Cyanotryptophan	n.d.	82
p-Ethynylphenylalanine	Tryptophan	n.d.	83
p-Cyanophenylalanine	7-Azatryptophan	18.5	84
p-Cyanophenylalanine	Thioamide	15.6	85

Triplet state and fluorescence quenching

Similar to FRET, other types of quenching mechanisms that lead to shortening of either the excited singlet or triplet state lifetimes have also found very important applications in protein folding studies. Unlike FRET, however, such quenching processes occur only over a very short distance, that is, typically only on formation of van der Waals contacts between the chromophore and quencher.^{86, 87} For example, for a simple loop closure process (i.e., open (O) to closed (C) state), the kinetics of quenching can be described by the following scheme:

where k_TT is the quenching rate in the closed state, whereas k_op and k_cl are the intrinsic opening and closing rates of the loop. It can be shown that if the rate of quenching is much larger than the rate of opening (i.e., ) then the observed rate constant corresponds to the rate of closing. Several research groups have utilized such quenching methods,^88–90 especially quenchers of the triplet state of Trp (Table II), to investigate peptide loop closure rates,^89–91 native state conformational fluctuations,⁹² and global cooperative folding/unfolding transitions.⁹³

Table II.

Tryptophan Triplet State Quenchers

Quencher	Quenching rate constant (M−1 S−1)	Ref.
3-Nitro-tyrosine	7 × 109	88
Tryptophan	2.1 × 108	89
Cysteine	1.3 × 108	89
Cystine	1.9 × 108	89
Lipoate	1.4 × ·09	89
Naphtylalanine	3 × 109	90

Recently, fluorescence correlation spectroscopy (FCS) has become increasingly used as a tool to investigate the dynamics of protein folding and conformational changes.^94–102 FCS is based on correlating the fluorescence intensity fluctuations of the molecular species in question induced by either a fluorescence quenching event or by molecular diffusion in and out of a small confocal laser excitation volume. For example, Sauer and coworkers^{103, 104} have shown that the fluorescence of oxazine dyes can be effectively quenched by Trp through the mechanism of photoinduced electron transfer (PET), making them ideal for use as probes of peptide and protein dynamics by means of FCS^{105, 106} More recently, Rogers et al.¹⁰⁷ have applied the PET-FCS technique, wherein the fluorescence of the dye Atto 655 is effectively quenched by nearby Trp residues, to reveal the conformational dynamics of a truncated version (M2TM) of the Influenza A M2 proton channel in model membranes and found that the conformational transition between the N-terminally open and C-terminally open states of the M2TM channel occurs on a timescale of about 500 μs.

Linear IR spectroscopy

Vibrational transitions in condensed phase occur on fs and/or ps timescales, thus IR spectroscopy can be used to investigate ultrafast conformational events of biological molecules. For proteins and peptides, IR bands arising from backbone vibrations are particularly useful as they show varying degrees of sensitivity to conformations.¹⁰⁸ Due to having the largest intensity, however, the amide I band, which originates mostly from the stretching vibrations of backbone carbonyls, is the most widely used as a probe of protein folding.^109–116 The conformational sensitivity of the backbone amide vibrational transitions arises from anharmonic interactions between the different vibrators located on different amide groups.^{117, 118} The underlying theoretical framework is often represented in terms of an N-dimensional Hamiltonian written in the basis of the N number of free floating vibrators located at the positions of the corresponding backbone amide groups:

(11)

where ε_i is the unperturbed frequency of the ith oscillator, and Δ_ij is the interaction energy between the ith and jth vibrators. Although experimentally determining Δ_ij is not straightforward, it is generally assumed that long range electrostatic interactions play a dominant role, while through-bond contributions may be important for nearest neighbor terms.^{119, 120} Because both ε_i and Δ_ij are sensitive to various environmental and structural factors, the resulting IR band shows a dependence on the protein conformation and hydration status.^{121, 122} For example, in D₂O the amide I′ band of solvated α-helices is centered at ∼1630 cm⁻¹,¹¹ whereas that of antiparallel β-sheets consists of two bands at ∼1620 cm⁻¹ (strong) and ∼1680 cm⁻¹ (weak),¹¹⁷ respectively.

In conjunction with laser-induced T-jump techniques, time-resolved IR measurements have significantly enhanced our understanding of the folding dynamics of protein secondary structures in isolation^{123, 124} and in various mini-proteins.^125–127 For example, Dyer and coworkers⁷ are among the first to show that, following a T-jump, the conformation relaxation of short alanine-based α-helical peptides, probed by IR spectroscopy, occurs on the ns timescale. In addition, application of the T-jump IR technique to the study of β-hairpin folding dynamics has allowed us to substantiate the idea that turn formation is the rate limiting step in β-hairpin folding^128–130 and helix–turn–helix formation.^{131, 132} More recently, we have shown, using the mini-protein Trp-cage^{133, 134} as an example (Fig. 3), that T-jump IR measurements carried out using different probing frequencies within the amide I′ band of the protein of interest can reveal important information regarding the individual folding rates of the constituent structural elements.¹²⁷

Figure 3.

Conformational relaxation kinetics of a Trp-cage in response to a T-jump obtained with different probing frequencies, showing that folding intermediates or kinetic heterogeneity could be revealed by recording transient IR kinetics over the entire amide I′ band region. (Adapted from Culik RM, et al., (2011) Angew Chem Int Ed 50:10884–10887, with permission from John Wiley and Sons.)¹²⁷

Strategies to enhance the structural resolution of IR spectroscopy

Although the aforementioned couplings among different amide vibrators provide the intrinsic sensitivity of the amide I vibration to protein conformation, it also causes the transition to broaden. In addition, the amide I frequencies of different secondary structures are not well separated. As a result, the amide I band of proteins is often broad, featureless, and congested with overlapping peaks. Thus, to gain site-specific structural information, certain strategies need to be used. One such strategy is to alter the reduced mass and hence the vibrational frequency of the vibrator of interest using isotopic substitution.^135–138 For example, assuming that the backbone C=O stretching vibration is harmonic, simply changing ¹²C to ¹³C would lead to a decrease in its vibration frequency by approximately 40 cm⁻¹, making it possible to separate the ¹³C=O vibration from a background arising from unlabeled (i.e., ¹²C=O) vibrators (Fig. 4). Therefore, the technique of isotope editing has been widely used in protein folding and aggregation studies to gain a better structural resolution when used in conjunction with IR spectroscopy.^{113, 139}

Figure 4.

(a) FTIR spectra of a ¹³C labeled (solid lines) in D₂O solution measured at different temperatures and (b) the corresponding difference FTIR spectra. Also shown in (a) is the FTIR spectrum (dashed line) of the unlabeled peptide. (Reprinted from Huang CY, et al., (2001) J Am Chem Soc 123:12111–12112, with permission from American Chemical Society.)¹³⁹

In addition to backbone vibrations, many sidechain vibrations are also expected to be useful IR markers of protein conformations. However, sidechain vibrations are much less used in protein folding studies due in part to their smaller molar extinction coefficients. Exceptions include the carboxylate asymmetric stretching vibrations of the Glu and Asp sidechains, which possess strong absorptions appearing near 1567 and 1584 cm⁻¹, respectively.¹¹⁴ On protonation of these sidechains at low pH, the corresponding vibrations shift to near 1700 cm⁻¹. Thus, these sidechain absorptions could be used as sensitive IR probes of local pH as well as salt bridge and ion pair formations. Moreover, we and others have recently demonstrated a novel method to enhance the structural resolution in IR studies by using unnatural amino acids as site-specific environmental and conformational reporters. Interested readers can find more details of this topic in a recent review.¹⁴⁰

UV resonance Raman

Protein vibrational transitions can also be probed via Raman spectroscopy. It has been shown that excitation of an electronic transition of the polypeptide using deep UV photons, for example the π → π* transition of the amide backbone or the ¹L_b transition of Trp residues, allows resonantly enhanced Raman spectra to be collected.^141–145 In comparison to IR measurements, which often suffer from interference of solvent vibrational bands in the amide II and III regions, the UV resonance Raman (UVRR) technique can greatly reduce solvent interferences, thus allowing for more easy recording of protein vibrational bands below 1600 cm⁻¹, for example, the amide III band and the C_α—H stretching vibration. The ability to reliably record protein vibrational spectrum in the lower wavenumber region makes UVRR a very useful tool of reporting on protein conformations. For example, Asher et al.¹⁴⁶ have proposed that the amide III peak frequency shows a roughly sinusoidal dependence on the backbone Psi angle, making a more quantitative interpretation of the vibrational spectrum possible. In addition, UVRR has also been used in conjunction with the laser-induced T-jump technique to study protein and peptide folding kinetics.^147–151 Further details on the application of UVRR to protein folding studies can be found in a recent review by Asher and coworkers.¹⁴⁵

Nonlinear IR methods

Probing the nonlinear response of a protein sample induced by a series of ultrashort laser pulses can provide additional dynamic and structural information that cannot be directly acquired via linear spectroscopic measurements. In particular, 2-dimensional infrared (2D-IR) spectroscopy has proven to be very useful in generating new insights into the ultrafast conformational and solvation dynamics of proteins and peptides^152–154 and is expected to play a growing role in studying the protein folding problem.

Because a detailed description of the 2D-IR technique can be found elsewhere,¹⁵⁵ below we offer only a brief description. As shown (Fig. 5), one common method for generating 2D-IR spectra makes use of four ultrashort (<100 fs), phase locked, IR laser pulses that may have different polarizations and are delayed in time with respect to each other. The first pulse (k₁) is used to promote the sample vibration into a superposition of the ground and excited states (or to produce a coherence), which is intersected by a second pulse (k₂) at a certain time delay, τ, generating a transient population grating that is then allowed to interact with a third pulse (k₃) after a delay time T, stimulating emission of a photon echo signal.¹⁵⁶ The magnitude and phase of the photon echo are then determined through interference with a fourth pulse, the local oscillator (LO), along the detection time axis, t. Finally, Fourier analysis of this signal along the two time axes, τ and t, produces the 2D-IR spectrum, which is a function of two frequencies, ω_τ and ω_t, and the mixing time T. One of the advantages of 2D-IR spectroscopy is that it allows one to spread the spectrum of interest into two dimensions, thus effectively increasing the spectral resolution by revealing features that overlap, cancel, or simply do not exist in linear IR spectra.¹⁵⁷ As shown (Fig. 6), a 2D-IR spectrum can also reveal cross correlations between frequencies that go undetected in linear IR spectroscopy, thus providing additional structural information.¹⁵⁶ Moreover, by following the evolution of the spectral line-widths in 2D spectra as a function of T, one can measure ultrafast (ps) dynamics due to solvent or peptide motions.¹⁵⁸ Additionally, 2D-IR signals show a fourth-power dependence on the underlying vibrational transition dipole moment (i.e., proportional to μ⁴), allowing for enhanced discrimination of overlapping IR bands.¹⁵⁹

Figure 5.

Schematic representation of the 2D IR photon echo experiment. Three ultrashort IR pulses (k₁, k₂ and k₃) interact in succession with a protein sample to produce a photon echo signal (depicted in blue), which is analyzed along the time axis t by overlapping with a reference pulse (LO).

Figure 6.

A representation of the 2D-IR photon echo spectrum (bottom) for two interacting anharmonic vibrators, demonstrating characteristic features that distinguish it from the linear IR spectrum (top) such as the appearance of well-defined homogeneously or inhomogeneously broadened peaks, as well as cross peaks arising from coupling.

Taken together, these added spectral and dynamic details contained in the nonlinear response of a protein makes 2D-IR spectroscopy a more sensitive structural tool for studying protein folding, conformational transitions, and local hydration status, among many other applications. For example, Hochstrasser and coworkers¹⁶⁰ have used the 2D-IR technique to study the conformational heterogeneity of HP35 via the C≡N (nitrile) stretching vibration of Phe_CN incorporated into the hydrophobic core of the protein. It was found that the 2D-IR spectra could only be described by two nearly overlapping distributions of vibrators that do not interconvert on a timescale similar to the range of T (0–3 ps). Because the nitrile stretching vibration is sensitive to hydration, this finding provided strong evidence that in the native potential well of HP35 two or more slowly interconverting subpopulations exist. A similar conclusion was also arrived at by a triplet–triplet state quenching study.⁹²

Although much progress is left to be made in the implementation of nonlinear spectroscopic methods such as 2D-IR spectroscopy to protein folding studies, several previous works stand out so far as showing their potential usefulness. For example, a large amount of effort has been made to characterize the 2D-IR spectral features of different secondary structural motifs,^161–165 laying the foundation for interpreting 2D-IR data of more complex protein systems. In addition, a small number of protein folding kinetic studies^{46, 166–168} have demonstrated that when combined with other resolution enhancement strategies, such as isotope-editing, 2D-IR spectroscopy is capable of revealing details regarding the folding/unfolding dynamics of the protein system of interest that are inaccessible from linear spectroscopy methods. With the development of other light-induced triggering methods,⁶⁴ and continuing refinement of the experimental methods,¹⁶⁹ we expect that multidimensional nonlinear spectroscopies will serve as critical tools in providing new mechanistic insights into protein folding dynamics.

Summary

The problem of protein folding has inspired a vast variety of studies by scientists from different disciplines in the past and continues to stimulate the development of new experimental methods that can measure and report, with as much structural detail as possible, on how and on what range of timescales specific and/or nonspecific conformational transitions occur. In particular, the development and application of various ultrafast triggering methods, such as the laser-induced T-jump technique, have made it possible to directly characterize the folding kinetics and mechanism of protein secondary structures and proteins that fold on the nanosecond and microsecond timescales. In addition, and perhaps more importantly, the recent advances in vibrational and optical spectroscopies hold great promise for disentangling and even “imaging” the complex network of molecular interactions involved in protein folding. Herein, we provide a concise summary of several recent developments in spectroscopic studies of protein folding, with the hope that it will provide useful information to those interested in applying some of the aforementioned methods to protein conformational studies. Due to page limit, however, we are unable to discuss several other deserving spectroscopic methods, such as those based on single-molecule fluorescence and/or FRET measurements. Finally, it is worth noting that many of those spectroscopic methods discussed are not limited to protein folding studies, they can also be applied to investigate other biophysical and biological questions, such as aggregation,^170–173 membrane protein conformation and dynamics,^174–178 peptide–peptide interactions, RNA and DNA dynamics,^179–182 and structure determination.¹⁵⁴