Transparent window 2D IR spectroscopy of proteins

Abstract

Proteins are complex, heterogeneous macromolecules that exist as ensembles of interconverting states on a complex energy landscape. A complete, molecular-level understanding of their function requires experimental tools to characterize them with high spatial and temporal precision. Infrared (IR) spectroscopy has an inherently fast time scale that can capture all states and their dynamics with, in principle, bond-specific spatial resolution. Two-dimensional (2D) IR methods that provide richer information are becoming more routine but remain challenging to apply to proteins. Spectral congestion typically prevents selective investigation of native vibrations; however, the problem can be overcome by site-specific introduction of amino acid side chains that have vibrational groups with frequencies in the “transparent window” of protein spectra. This Perspective provides an overview of the history and recent progress in the development of transparent window 2D IR of proteins.

INTRODUCTION

Proteins are complex, dynamic macromolecules. The chemistry is heterogeneous throughout their structures, and distinct locations can differently contribute to functional mechanisms. Moreover, proteins are dynamic ensembles of states differentiated in scale from large global domain rearrangements to small fluctuations of side chains and protein-associated water. The corresponding dynamics range from the second and longer to the picosecond time scales. This spatial and temporal heterogeneity impedes generation of a complete, molecular-level description of protein function. Infrared (IR) spectroscopy is well suited to tackle both challenges. IR chromophores can be as small as a bond to achieve spatially selective characterization. The technique has no size limitation and is applicable to any protein in any state. The inherent, picosecond time scale of IR spectroscopy prevents band coalescence and motional narrowing to ensure that most functionally relevant states of a protein can be captured. Moreover, kinetic processes can be directly monitored as fast as the femtosecond time scale.

Although IR spectroscopy, in principle, can interrogate a single bond, in practice, the spectral congestion from thousands of overlapping bands in protein spectra prevents the detection of any single vibration. To tackle this issue, specific parts of a protein can be characterized by the introduction of IR labels with frequencies within the “transparent window” of a protein spectrum (∼1800–2300 cm⁻¹), where no native protein vibrations absorb (Fig. 1).^1–4 This region has historically been utilized for the study of hemeproteins by analysis of the vibrations of small, diatomic ligands to the heme.⁵ Any location in a protein, in principle, can be accessed by the introduction of frequency-resolved labels at protein side chains. The approach has developed over the past two decades, and many noncanonical amino acids containing frequency-resolved labels now have been applied for the study of protein function by IR spectroscopy.^3,4

FIG. 1.

Example protein (myoglobin) IR spectrum highlighting the transparent frequency window with purple shading and the structures of ligands and noncanonical amino acids with frequency-resolved vibrations discussed in this Perspective.

The development of transparent window two-dimensional IR spectroscopy (2D IR) for site-selective study of proteins has been underway for the last decade.^6–9 2D IR informs about connections between absorptions and their time-dependent behavior that reveal richer information about the underlying molecular structures and dynamics. The theory, methodology, and advantages of 2D IR have been extensively described.^10–15 The two axes, respectively, reflect the frequencies excited and then probed at the same or later times (Fig. 2). The absorption bands in the linear spectrum appear along the diagonal, while off-diagonal bands indicate that the corresponding species are connected, either because they are electronically or mechanically coupled or undergo conformational or chemical exchange. An important advantage of 2D IR of vibrations in proteins is the ability to deconvolute inhomogeneous broadening due to the ensemble of microenvironments from homogeneous broadening due to dynamics fast on the IR time scale and the finite population and orientational lifetime via the diagonal/off-diagonal broadening of the 2D bands, respectively. The time-dependent 2D band shape evolution reflects spectral diffusion, the dynamics among the frequency inhomogeneity, yielding insight into underlying molecular dynamics influencing the vibration.

FIG. 2.

(a) Inhomogeneous (homogeneous) broadening results in elongation of 2D bands along the diagonal (anti-diagonal). (b) Interconversion among the underlying states with increasing time between excitation and detection reduces the elongation. (c) Analysis of the 2D line shapes can yield a frequency–frequency correlation function that describes spectral diffusion within the inhomogeneous frequency distribution.

As the field of 2D IR matures, several excellent reviews covering applications to proteins are available.^10,16–19 This Perspective provides an overview of the history and recent progress in the development of transparent window 2D IR of proteins. A brief discussion of several routes for further advancement of the field of transparent window 2D IR follows.

LIGAND VIBRATIONAL PROBES

The earliest transparent window 2D IR experiments directed at proteins employed small molecule ligands.^20–23 Di/triatomic ligands to metal centers in metalloproteins and substrate analogs with frequency-resolved vibrations have been used to probe the environments at their binding sites. Use of this approach is limited to the study of proteins that bind a particular ligand and the one location at the binding site in the protein. However, when a suitable ligand is available, its location within a protein is typically key to function.

DI/TRIATOMIC METAL LIGANDS

Carbon monoxide (CO) ligated to myoglobin served as a model system during the development of ultrafast IR spectroscopy and since has been extensively subject to study by 2D IR.^{20,21,24–27} The advantage of the CO ligand is its intense absorbance. In addition, the solvatochromism is well described by a Stark effect from electrostatic interactions, so spectral interpretation is straightforward. A CO ligand since has been employed for characterization of many hemeproteins: hemoglobin, neuroglobin, horseradish peroxidase, cytochrome c552, cytochrome P450 et al.^28–37 The probe continues to be useful for our studies of cytochrome P450 activity.^35,37 Additional small di/triatomic ligands, such as nitric oxide, thiocyanide, selenocyanate, and azide, have been explored for the characterization of hemeprotein environments as well.^38–41 CO and cyanide (CN⁻) ligands of non-heme metals also have provided vibrational probes of proteins. For example, a CO ligand to a copper center yielded insight into the structure and dynamics of a de novo metalloprotein, reported by Ross et al.⁴² Another example is a recent study of the structure and dynamics of a hydrogenase by analysis of the native CO and CN⁻ ligands of the iron center by Horch et al.⁴³

SUBSTRATE AND TRANSITION STATE ANALOGS

In an early application of 2D IR for the study of protein recognition, Fang et al. analyzed the two cyano groups of a non-nucleoside inhibitor of HIV reverse transcriptase (Fig. 1).²² Unlike the overlapping absorbances observed in solution, the two cyano groups have distinct frequencies when bound to HIV reverse transcriptase, reflecting the different environments sensed within the enzyme active site. Characterization of the spectral diffusion of one of the cyano groups indicated that the distribution of its environments is sampled in tens of picoseconds.²² Later studies of the inhibitor bound to drug resistant mutant enzyme captured faster dynamics that revealed that the drug maintained binding potency through acquisition of water-mediated interactions.⁴⁴

2D IR with a ligand probe is being applied to address a longstanding question regarding the role of dynamics in hydrogen tunneling during enzymatic catalysis by Bandaria et al.⁴⁵ They have taken advantage of azide, a transition state analog inhibitor of formate dehydrogenase, to probe the active site. 2D IR has been used to compare the dynamics of the binary complex with only the azide ligand to ternary complexes with a nicotinamide cofactor bound, whereas the dynamics sensed by the ligand probe in the binary complex are essentially static over the 5 ps time window, and they occur on a 2–5 ps time scale distinctly in ternary complexes with the nicotinamide cofactor that is structurally primed for catalysis.²³ Consistent with the model developed using the azide probe, an azido-functionalized NAD⁺ analog reports large static inhomogeneity of the binary complex of formate dehydrogenase with the cofactor.⁴⁶ A later analysis of the azide ligand with an improved spectroscopic methodology and purer sample revealed underdamped oscillatory dynamics for the ternary complex with the charged, oxidized NADP⁺ cofactor [Fig. 3(a)].⁴⁷ Since these oscillations are not observed for the complex with reduced NADPH, they likely reflect sensitivity to the motions of the charged cofactor. Such underdamped motions indicate insulation of the azide within the active site that is consistent with rigid packing that would be conducive to hydrogen tunneling. Further supporting the interpretation that the azide is sensitive to motions involving NADP⁺, mutation of a valine residue that packs against the cofactor to decrease the side chain bulk leads to increased amplitude of the oscillations [Fig. 3(b)].⁴⁸ The changes in dynamics due to mutation are correlated with increased temperature dependence of the kinetic isotope affect, supporting their involvement in hydrogen tunneling. Another impactful study showed via the azide probe that the dynamics, along with the thermal stability, are perturbed by ¹⁵N and ²H isotopic labeling, an observation that could have wide ranging implications to common biophysical approaches.⁴⁹

FIG. 3.

(a) Center line slope (CLS) decays showing spectral diffusion of azide in the ternary complexes of formate dehydrogenase with nicotinamide cofactors. Adapted with permission from Pagano et al., J. Phys. Chem. Lett. 7(13), 2507–2511 (2016). Copyright 2016, American Chemical Society. (b) CLS decays of ternary complexes for WT (black), V123A (red), and V123G (blue) formate dehydrogenase (upper left), real part of the FFT of the CLS decay (lower left), structural model of the active site showing azide (blue), NADP⁺ cofactor (purple), and V123 (yellow) (upper right), and the plot of intensity of the 9 cm⁻¹ peak in the FFT of the CLS against the difference in activation energies between H and T isotopes in the hydride transfer reaction. Figures adapted with permission from Pagano et al., ACS Catal. 9(12), 11199–11206 (2019). Copyright 2019, American Chemical Society.

A second application of a nitrile substrate analog by ultrafast IR spectroscopy was directed at understanding the mechanism of ketosteroid isomerase by Jha et al.⁵⁰ Excitation of a nitrile photoacid mimics the change in charge density of the native steroid substrates during the catalytic cycle. When the photoacid is in aqueous solution, evolution of the cyano vibrational frequency after excitation occurs upon rearrangement of water–nitrile hydrogen bonding interactions on a 12 ps time scale in response to the change in excited state charge distribution. This rapid frequency response disappears upon binding the enzyme. The essentially static response indicates rigid hydrogen bonding interactions between the photoacid and waters in the active site that would provide an organized electrostatic environment to support catalysis.

SIDE CHAIN LABELS

Attachment of frequency-resolved labels to protein side chains, in principle, enables the study of any location in any protein. Ideally, the noncanonical amino acid is incorporated selectively at a single location to achieve site-selective analysis and avoid creating spectral congestion. Analysis of labels placed at varying locations can yield a comprehensive view of protein structure and function. Preferable vibrations are decoupled from other vibrations in the molecule and, thus, have local mode character, which provides spatial selectivity and simplifies interpretation. For example, Fermi resonances complicate spectral interpretation for some labels, particularly azido probes, which is discussed further below.^51–56 Smaller labels minimize the chance their incorporation will perturb protein structure or function.

A major challenge in implementing transparent window 2D IR is achieving sufficient sensitivity to detect the absorption of a single oscillator in a protein at low concentration with non-negligible background absorbance of water. For example, the native S–H bond of reduced cysteines absorbs at ∼2500 cm⁻¹ and has been detected by 2D IR, but since has not been under investigation due to its weak absorption.⁵⁷ Noncanonical amino acids with side chains functionalized with frequency-resolved labels that generate stronger signals have been sought as probes of proteins by 2D IR. The intensity of absorption bands depends on the vibration’s transition dipole strength. Linear absorbance depends on the square of the transition dipole strength, while 2D IR absorbance has a quartic dependence. The quartic dependence on transition dipole strength is an advantage because the 2D IR absorptions are spectrally narrowed and accentuated above the solvent background absorbance. Nonetheless, the solvent response must be subtracted to uncover the absorptions of most labels. The transition dipole strength is among the most important properties in selecting a frequency-resolved label because the attainable concentrations of protein samples are limited, so it determines the range of proteins that can be studied by 2D IR.

Another important consideration for measurement of equilibrium dynamics by 2D IR is the vibrational lifetime. The vibrational lifetime determines the decay of the 2D IR absorbance as the time between exciting and probing is increased, so it controls the experimental time window. Extensive effort has been directed at addressing this limitation by identification and design of new frequency-resolved labels of proteins, as described further below.

EARLY DEVELOPMENT OF SIDE CHAIN LABELS

The initial demonstrations of 2D IR with a frequency-resolved label focused primarily on establishing the approach by investigating the aromatic nitrile of p-cyanophenylalanine (CNF) incorporated by solid phase synthesis into a model system, the 35-residue peptide HP35 that folds into a subdomain of the villin headpiece protein (Fig. 4). Urbanek et al. first reported 2D IR of a single CNF label in HP35 and observed two absorption bands [Fig. 4(a)].⁶ Shortly later, Chung et al. reported 2D IR of the same HP35 peptide [Fig. 4(b)].⁷ Their study employed two CNF labels for a greater signal but observed a single absorption band, finding that the second band reported earlier likely arises from aggregation or dimerization. This and subsequent studies followed the unfolding of HP35 by 2D IR of the CNF probes.^7,58,59 The CNF revealed increasingly rapid dynamics upon unfolding, reflective of the loss of the structured environment within the core of the folded state. A further study of the label in mutant HP35 with increased stability observed slower fluctuations indicative of improved packing. These observations later were shown to be reflective of CNF of the singly labeled HP35.⁵⁸ Importantly, these first studies established the use of frequency-resolved labels with 2D IR to measure dynamic characteristics of the solution and protein environments.

FIG. 4.

(a) Example time-dependent 2D IR spectra of cyanophenylalanine (CNF) probe in HP35. Figure adapted with permission from Urbanek et al., J. Phys. Chem. Lett. 1, 3311–3315 (2010). Copyright 2010, American Chemical Society. (b) Structural model of HP35 showing residues replaced by CNF (left) and center line slope decays that report the spectral diffusion of the CNF under varying conditions. Figures adapted with permission from Chung et al., Proc. Natl. Acad. Sci. U. S. A. 108(9), 3578–3583 (2011). Copyright 2011, National Academy of Sciences.

Characterization of a single vibrational label in a full-sized protein by 2D IR was demonstrated by Thielges et al. by incorporation of azidophenyalanine (AzF) by amber suppression into the model protein, myoglobin (Fig. 5).⁸ An advantage of azido vibrations is that the transition dipole strengths are relatively strong; thus, their detection requires lower protein concentrations. The AzF was directed into the heme pocket so that it would sample the same environment as a CO ligand, a well-established 2D IR probe of myoglobin. While the shorter lifetime of AzF limits the experimental time window compared to CO, the two probes reported the same time scales of dynamics. In addition, they were sensitive to the presence of the other. The consistency of the information obtained by 2D IR with two distinct probes in the same environment validated the approach for characterization of complex systems such as proteins.

FIG. 5.

(a) Structural model of myoglobin with azidophenylalanine (AzF) and carbon monoxide (CO) probes. (b) Center line slope (CLS) decays that show spectral diffusion of CO (upper panel) and AzF (lower panel) in the presence (blue) and absence (red) of the other probe. Figure adapted with permission from Thielges et al., J. Phys. Chem. B 115, 11294–11304 (2011). Copyright 2011, American Chemical Society.

Like AzF, azidohomoalanine (Aha) has a large transition dipole strength that enables detection at μM protein concentrations. Another advantage of Aha is the possibility for its introduction into larger proteins at native methionine residues via expression in methionine auxotrophs. Although specific labeling requires the methionine residue be unique, methionine is an infrequent amino acid residue, so minimal protein engineering would often be needed to generate a protein labeled with a single Aha. Bloem and co-workers have developed the amino acid analog as a label for 2D IR, focusing on illuminating molecular recognition of the PDZ domain of human tyrosine phosphatase.^9,60–62 They first evaluated Aha as a label with 2D IR when incorporated at four sites within a peptide ligand consisting of the C-terminal sequence of the guanine nucleotide exchange factor (Fig. 6).⁹ Optimization of 2D IR data acquisition and careful subtraction of the solvent enabled isolation of the 2D IR bands. Aha at three of the residues was sensitive to binding. When the peptide is associated with the PDZ domain, one Aha indicated two populations, and two Aha indicated increased heterogeneity. While the short ∼1 ps vibrational lifetime of the azido group limits the temporal window for measurement of equilibrium dynamics, time-dependent 2D IR was performed to measure spectral diffusion and reveal that the increased inhomogeneity upon association with the PDZ domain undergoes dynamics on time scales slower than the experimental window. Notably, the study highlights the complexity of protein–ligand recognition and the need for experimental approaches that provide high spatial and temporal details.

FIG. 6.

(a) Structural model of the PDZ domain of human tyrosine phosphatase with a bound peptide ligand. (b) Overlay of 2D IR spectra of each azidohomoalanine (Aha) probe when in solution (red) and bound to the PDZ domain (blue). Figure adapted with permission from Bloem et al., J. Phys. Chem. B 116(46), 13705–13712 (2012). Copyright 2012, American Chemical Society.

METAL CARBONYL SIDE CHAIN LABELS

The key advantage of metal carbonyl groups is their massive transition dipole strengths that result in intense IR absorptions.^63–68 This feature makes metal carbonyl groups viable as probes of proteins at μM concentrations. While potential perturbation arising from their relatively large size hinders their incorporation within the cores of proteins, metal carbonyl vibrations are useful for characterization of surface hydration and dynamics. Because the carbonyl vibrations are decoupled from the rest in the molecule, intramolecular relaxation is disfavored. Rather, intermolecular interactions determine vibrational relaxation, making the vibrational lifetime sensitive to the probe’s hydration. King et al. introduced ruthenium carbonyl labels for 2D IR by attachment to distinct histidine residues in lysozyme homologs (Fig. 7).⁶⁴ They showed how the hydration of the probes influences the vibrational relaxation. H₂O/D₂O exchange changed the vibrational lifetime when the metal carbonyl group was placed at a solvent-exposed, flexible location of the protein but not when placed at a water-restricted location within a crevice between two helices. Additionally, the label at the water-restricted location was distinctly impacted by trifluoroethanol addition, showing how interaction with cosolvent depends on the detailed chemical nature of the protein surface.

FIG. 7.

(a) Overlay of structural models of lysozyme homologs labeled with a metal carbonyl probe. (b) Vibrational relaxation of the metal carbonyl probe in H₂O and D₂O for the solvent-exposed (left) and buried (right) locations. (c) Correlation functions showing spectral diffusion of buried probe in solutions of varying PEG400 content (left), and overlay of the time scale of hydration dynamics (blue) and static inhomogeneity (red) of buried metal carbonyl probe in solutions of varying PEG400/water content. (a) and (b) adapted with permission from King et al., J. Phys. Chem. B 116(19), 5604–5611 (2012). Copyright 2012, American Chemical Society. Panel (c) adapted with permission from King et al., J. Am. Chem. Soc. 136(1), 188–194 (2014). Copyright 2014, American Chemical Society.

Later investigations were directed at metal carbonyl vibrations in the water-restricted crevice upon increase in viscosity by the addition of glycerol or induction of molecular crowding by the addition of PEG400 or increase in protein concentration.^65,68 The vibrational lifetime was insensitive to the addition of the cosolvents, indicating that they do not disrupt the local hydration. However, the increase in viscosity due to addition of glycerol led to slower picosecond spectral diffusion, interpreted as hydration dynamics, and an increase in the static inhomogeneity, attributed to protein dynamics. Upon addition of PEG400 or an increase in protein concentration, the time scale of spectral diffusion and static contribution showed an abrupt transition to a state with slower hydration and coupled protein dynamics. Molecular dynamics simulations indicated no structural change in the protein, indicating that the transition involves changes only to the hydration and coupled protein dynamics. A key observation in these studies was that the changes observed in hydration and protein dynamics correlated, suggesting that they are coupled.

Woys et al. reported the attachment of a rhenium carbonyl probe with alkyl linkers to cysteine residues.⁶⁶ Cysteine is a less frequent amino acid residue than histidine that could serve as a site of unique attachment in a broader range of proteins. The label was compared in solvents, attached at two locations of ubiquitin, and attached to alpha-synuclein in aqueous solution and lipid environments. As observed for ruthenium carbonyl vibrations, the vibrational lifetime was found to inform about the hydration. The group reported another strategy for incorporating rhenium carbonyl probes selectively into proteins.⁶⁷ Azidohomoalanine is incorporated at native methionine residues by expression in methionine auxotrophs. The azido group then serves as a handle for the attachment of the metal carbonyl label via click chemistry.

RECENT 2D IR OF LABELS IN PROTEINS

Azidohomoalanine probes

Recent 2D IR studies have been leveraging the key advantage of frequency-resolved labels to probe and compare specific local sites at different parts of proteins [Fig. 8(a)].⁶⁰ Aha was incorporated at six sites in a PDZ domain engineered with an azobenzene-based linker by Johnson et al. Photoisomerization of the linker induces a structural change akin to ligand binding. 2D IR difference spectroscopy upon photoisomerization detected only for one Aha a change in amplitude, attributed to direct interaction with the linker itself, demonstrating the need for residue-specific analysis of protein function [Fig. 8(b)]. In a subsequent study of Aha incorporated at the surface of the PDZ domain, the frequency was found sensitive to binding the wild-type and the mutated sequence of the peptide ligand.⁶² In addition, 2D IR performed at μM concentrations with Aha inserted into the peptide ligand was demonstrated as a viable method to characterize affinity with residue-selective precision.⁶¹

FIG. 8.

(a) Structural model of the PDZ domain of human tyrosine phosphatase modified with an azobenzene-based linker with locations of incorporation of azidohomoalanine (Aha) shown as green spheres. (b) 2D IR difference spectra of each Aha upon photoisomerization of the linker. Figure adapted with permission from Stucki-Buchli et al., J. Phys. Chem. A 121(49), 9435–9445 (2017). Copyright 2017, American Chemical Society.

Another interesting application of the Aha with ultrafast spectroscopy is use as a vibrational reporter to measure the vibrational energy flow between specific sites in proteins.^69–71 Most efforts to investigate vibrational relaxation through proteins rely on large visible chromophores for energy deposition. Müller-Werkmeister and Bredenbeck employed Aha with an azulene-based amino acid, an excellent structural analog of native tryptophan that undergoes ultrafast internal conversion upon excitation. After excitation of the azulene, Aha provided a local reporter of vibrational energy transport. This approach first was demonstrated to detect vibrational energy transfer between the two amino acid analogs when separated in sequence in a peptide.⁶⁹ Vibrational energy transfer within a protein was accomplished through an amber codon expression system for selective introduction of the azulene-based amino acid at two locations in the tyrosine phosphatase PDZ domain.⁷⁰ An Aha placed in the peptide ligand served as a thermal reporter. The combined labeling approaches enable more precise investigation of the anisotropic energy flow that could mediate subtle conduction pathways underlying protein allostery. A quantitative analysis of vibrational energy transfer was accomplished by placing the Aha donor in varying locations with respect to the azulene donor within a Trp hairpin.⁷¹ By combining the spectroscopy with non-equilibrium molecular dynamics simulations, the study quantified vibrational energy transfer along strands, between strands, and through-space contacts. Interestingly, the diffusive dynamics of the energy flow leads to substantial between-strand energy transfer.

A complication with the use of azido probes is that they often show Fermi resonances, which can be erroneously interpreted as the presence of a second state. Side bands attributed to Fermi resonances, for example, have been observed for AzPhe, azidoalanine, azidopyridine, and azido-functionalized sugar of a NAD⁺ analog.^51–56 Those observed for azidoalanine are weak, and none have been reported for Aha. Fermi resonances are less problematic for 2D IR spectroscopy than linear spectroscopy, as they result in predictable cross bands and are suppressed relative to the fundamental due to their dependence on a transition dipole strength of a weak or dark mode.^56,72 However, the efficient relaxation pathway through the low frequency mode decreases the vibrational lifetimes, limiting the 2D IR signal duration. Fortunately, in many cases, the Fermi resonances can be ablated by isotopic substitution.^53,56

Cyanophenylalanine probes

We have pushed forward CNF as a frequency-resolved label for 2D IR. The noncanonical amino acid is minimally perturbative as a tyrosine or phenylalanine analog. CNF can be site-selectively incorporated into proteins by an established amber codon expression system.⁷³ Furthermore, the aromatic nitrile has a sufficiently large transition dipole strength to enable characterization of proteins by 2D IR at ∼1 mM concentration.⁷⁴ We have applied CNF for the site-specific study of the conformational heterogeneity and dynamics involved in protein recognition.^74–76

We employed site-selective 2D IR using CNF to characterize the recognition of a Src homology three domain and a proline-rich peptide ligand.⁷⁴ CNF was introduced at six distinct sites in the protein domain [Fig. 9(a)]: each of three conserved Tyr residues that form the binding surface, a Tyr residue within a flexible loop, a Phe residue in the protein core, and a Tyr residue that is distant from the ligand binding surface. Upon binding the ligand, the frequency of the three CNF placed along the binding surface and the buried site changed sightly. In comparison, 2D IR revealed substantial and varying impact to the inhomogeneous broadening and the spectral diffusion [Fig. 9(a)], indicating that the heterogeneity of the environments of the probes and the rates of interconversion among the underlying states are differentially affected by complexation. Even probes placed at the adjacent conserved Tyr residues that form the binding surface responded differently to binding. As side chain dynamics play a key role in entropy changes of molecular recognition by other proteins,⁷⁷ we speculate that the site-specific changes in the heterogeneity and dynamics uncovered in our study of recognition by an Src homology three domain may make important contributions to binding thermodynamics and specificity within a cellular interaction network.

FIG. 9.

(a) Structural model of a Src homology three domain with a bound peptide ligand showing locations of incorporation of cyanophenylalanine (CNF) probes (left), and center line slope (CLS) decays reflecting the spectral diffusion of each CNF for the free domain (colored) and the complex with the ligand (black). Reproduced with permission from Ramos et al., Phys. Chem. Chem. Phys. 21, 780–788 (2019). Copyright 2019, The PCCP Owner Societies. (b) Structural model of the complex of plastocyanin and cytochrome f showing locations of incorporation of CNF (left), and CLS decays reflecting spectral diffusion of each CNF for free plastocyanin (colored) and the complex with cytochrome f (black). Figures adapted with permission from Ramos et al., J. Phys. Chem. B 123(9), 2114–2122 (2019). Copyright 2019, American Chemical Society. (c) Structures of camphor and norcamphor substrates and a structural model of cytochrome P450cam highlighting F/G helices (dark gray) and locations of incorporation of CNF (left panel). CLS decays reflecting the spectral diffusion of each CNF for the free enzyme (black) and complexes with camphor (blue) or norcamphor (green) (right panel). Figure adapted with permission from Ramos et al., Biophys. J. 120(5), 912–923 (2021). Copyright 2021, Elsevier.

A second application of CNF with 2D IR was directed at understanding the dynamic complexes formed by electron transfer proteins, plastocyanin and cytochrome f [Fig. 9(b)].⁷⁵ We generated plastocyanin selectively labeled with CNF at a location in the hydrophobic patch at the “head,” within the electrostatic patch on the “side,” and at the edge of the hydrophobic patch. Upon association with cytochrome f, the inhomogeneity of all labels increased, suggestive of greater heterogeneity of interactions. While the extent of the increase varies among probes, 75% of the increased heterogeneity in the complex interconverts on a fast 1–2 ps time scale [Fig. 9(b)]. This observation implies that the interactions of the CNF in the complex are highly dynamic; thus, the interface between the electron transfer partners is highly mobile. Altogether, the CNF probes suggest that the head more substantially than the side of plastocyanin interacts with cytochrome f but that the complex remains highly dynamic with a mobile interfacial water layer likely present for the majority of the ensemble.

Transparent window 2D IR was recently applied to elucidate how conformational flexibility and dynamics contribute to the regioselectivity of catalysis by the cytochrome P450 superfamily of heme oxidases.⁷⁶ Regioselectivity is likely to depend on the rigidity with which the enzymes bind and orient substrates with respect to a reactive oxy-heme intermediate. To evaluate this possibility, 2D IR with CNF labels was applied to compare the recognition by the archetypical P450cam of its native substrate camphor, which is hydroxylated with 100% regioselectivity, and another substrate, norcamphor, which yields two major hydroxylation products [Fig. 9(c)]. The CNF probes illuminate how binding norcamphor does not equivalently induce a larger scale conformational transition to the closed state of P450cam as does binding camphor. For instance, a label directed toward the F/G helices (CNF87) showed an asymmetric line shape for the norcamphor complex that evinces a second populated state; slow spectral diffusion measured for this probe is consistent with relatively slow interconversion. A label directed at the substrate (CNF96) also shows an asymmetric line shape indicating two bands for the norcamphor complex, but spectral diffusion is more rapid than measured for a single band found for the camphor complex. Corresponding with this observation, a CO heme ligand that also contacts the bound substrate similarly reports faster dynamics for the norcamphor than the camphor complex.³⁵ The rapid dynamics are consistent with the time scale of transitioning orientations of norcamphor that would explain formation of multiple hydroxynorcamphor products.

Cyanocysteine probes

Thiocyanates can be incorporated via straightforward chemistry under mild conditions at cysteine residues.^78,79 Because cysteine is a relatively rare amino acid, incorporation at a single site requires minimal protein modification. The drawback of the cyano vibration of cyanocysteine (CNSC) compared to CNF is the smaller transition dipole strength. However, the heavy sulfur atom decouples the vibration from intramolecular relaxation pathways, leading to a vibrational of lifetime of 30–160 ps compared to ∼4.5 ps for CNF. The extended vibrational lifetime enables measurement of equilibrium dynamics by 2D IR over a longer time window. Moreover, like metal carbonyl vibrations, vibrational relaxation is dominated by intermolecular interactions, so the vibrational lifetime provides a measure of solvent exposure.^80,81

Proof of feasibility for the use of CNSC as a label for 2D IR was demonstrated through labeling two cysteine residues in dihydrofolate reductase by Rock et al.⁸² This study reported the 2D IR spectrum with the enzyme at ∼4.6 mM, but it was directed primarily at the evaluation of implementations of 2D IR. Shortly later, van Wilderen et al. labeled the single cysteine in the alpha subunits of the heterotetramer hemoglobin.⁸⁰ This study showed how the lifetime of CNSC was solvent dependent due to disruption of intramolecular vibrational relaxation by the heavy atom insertion.

Schmidt-Engler and co-workers have introduced CNSC at varying locations to study the photocycle of the photoactive yellow protein (PYP) photoreceptor.^81,83,84 The labels show site-specific responses to photoisomerization of the chromophore of PYP. Step scan FT IR spectroscopy was first used to follow the kinetics of the structural evolution upon excitation. The dynamics for one residue deviated from the global protein response reported by the amide backbone, illustrating the importance of characterizing all parts of protein to gain a comprehensive view. The solvent exposure of labels was assessed by comparison of the vibrational lifetimes in D₂O and H₂O because vibrational relaxation of CNSC is dominantly intermolecular.⁸¹

Characterization of five CNSC labels by 2D IR has provided a comprehensive view of the site-specific heterogeneity and dynamics of PYP and the response to photoexcitation [Fig. 10(a)].⁸⁴ Incorporated labels spanned the structure of PYP: on an N-terminal helix proposed to detach upon excitation (L23), on two beta strands that make up the protein core (A30 and V122), and on two helixes predicted to be responsive to chromophore photoisomerization (A44 and V57). 2D IR spectroscopy revealed the spatial variation among the inhomogeneity and dynamics and response to chromophore photoisomerization. For example, labels at A44 and V57 at solvent-exposed positions report substantial inhomogeneity sampled with rapid dynamics. Labels at A30 and V122 are inhomogeneously broadened but show narrow linewidths, reflecting their locations in a homogeneous protein core. Photoisomerization leads to slower dynamics for labels at all residues but L23, consistent with partial unfolding proposed in the photocycle. In addition, labels located in adjacent environments within the beta strand core, A30 and L122, show the same response. In comparison, faster dynamics for the label at L23 suggest detachment of the helix during the photocycle.

FIG. 10.

(a) Structural model of changes upon photoisomerization of PYP binding with residues replaced by thiocyanocysteine (CNSC) probes labeled (left panel). Center line slope decays reflecting the spectral diffusion of each CNSC probe for the initial (black) and photoisomerized states (gray) (right panels). Reproduced with permission from Schmidt-Engler et al., Phys. Chem. Chem. Phys. 22(40), 22963–22972 (2020). Copyright 2020, The PCCP Owner Societies. (b) Structural model of conformational changes of calmodulin upon binding Ca²⁺ and a peptide ligand with residues replaced by thiocyanocysteine (CNSC) probes labeled (left panel). Center line slope decays reflecting the spectral diffusion of each CNSC probe for the apoprotein (black), after binding Ca²⁺ (red), and after binding the ligand (gray) (right panel). Data for the model compound methyl thiocyanate are shown in blue. Reproduced with permission from Schmidt-Engler et al., Phys. Chem. Chem. Phys. 22(10), 5463–5475 (2020). Copyright 2020, The PCCP Owner Societies.

CNSC has also been applied to site-specifically investigate successive conformational changes in the Ca²⁺ sensor calmodulin domain in response to binding Ca²⁺ and a peptide ligand [Fig. 10(b)].⁸⁵ The probes were placed at a conserved native isoleucine in the Ca²⁺-binding site, conserved native methionine residues that contact the peptide ligand, and a solvent-exposed control site in the ion recognition domain. Characterization by 2D IR revealed how the sensitivity of the dynamics to binding Ca²⁺ or the peptide ligand varies among locations in calmodulin. The CNSC near the ion-binding site responded to binding Ca²⁺ but was only modestly impacted by peptide binding. In contrast, CNSC at locations involved in ligand recognition were sensitive to binding the peptide but not Ca²⁺. Altogether, the data illuminate the varying involvement of different locations in two aspects of the functional mechanism of calmodulin signaling. Due to the lower transition dipole strength of CNSC, the samples were at relatively high concentration (∼7.5 mM), but the minimally sized probes enabled characterization of spectral diffusion throughout a protein over the longest duration reported to date.

Outlook

Efforts are ongoing to expand and improve the tool set of labels for 2D IR of proteins. Cyanotryptophan can serve as both an IR and a fluorescence probe and has yet to be fully leveraged for the study of protein function.^86,87 Isocyanoalanine and isocyanotryptophan are attractive due to their moderately strong transition dipole strengths and vibrational lifetimes of 4–6 ps.^54,88,89 Azidoproline, azidoalanine, azidomethylphenylalanine, and a cyanamide side chain have been introduced as probes but not yet utilized for protein studies.^90–92 Carbon deuterium bonds can serve as frequency-resolved labels with no complications from perturbation but are prohibitively weak for 2D IR at attainable protein concentrations.^1,93

Since probes with longer lifetimes would enable characterization of equilibrium dynamics over longer time scales, their development has been extensively pursued. Isotopic labeling has been explored to detune vibrations from accepting modes. This strategy resulted in a modest (∼2-fold) increase in lifetime for CNF.^94,95 An alternate strategy to increase the vibrational lifetime is insertion of a heavy atom to decouple the probe vibration from others within the molecule and disfavor intramolecular energy relaxation. For example, vibrational decoupling underlies the longer lifetime of the cyano vibration of CNSC compared to CNF. Cyanoselenophenylalanine has an extended lifetime through heavy atom decoupling, but the instability of the amino acid hindered its incorporation into a protein.⁹⁶ Methyl thiocyanate and selenocyanate phenylalanine analogs with long cyano vibrational lifetimes (100 and 300 ps, respectively) have also been prepared.⁹⁷ Heavy atom insertion similarly resulted in longer vibrational lifetimes for cyano groups attached to proline.⁹⁸ Insertion of silicon to decouple an alkyne vibration increased the vibrational lifetime and transition dipole strength.⁹⁹ The next challenge for practical use of the many amino acid analogs now available is their selective incorporation into proteins. Alternately, a spectroscopic approach to extending the experimental time scale is vibrationally promoted electronic resonance (VIPER) spectroscopy.¹⁰⁰ Vibrational relaxation that would otherwise diminish the 2D IR signals is prevented by excitation to a longer lived electronic excited state. Specific molecules in a mixture can also be selectively enhanced if the frequencies of their electronic transitions are sufficiently distinct.

Low sensitivity due to the weak signals and limited sample concentrations remains a principal challenge for 2D IR of single vibrations in proteins. The background absorbance from the aqueous solvent is relatively low and suppressed in 2D IR spectra relative to absorbance of most labels, but it becomes significant with lower protein concentrations and vibrations with weaker transition dipole strengths. In addition, use of higher excitation energy has improved detection limits, but the greater energy deposition into solvent causes a thermal response that can overwhelm or distort the spectra of the label. Spectrally narrowing the excitation pulses to minimize solvent excitation can mitigate this issue.⁸¹ In addition, a number of approaches can be applied to remove the solvent contribution. A 2D IR spectrum acquired at a waiting time several times longer than the lifetime of the label, when its signal has decayed, provides a crude background spectrum. Pulse shaping to remove the resonance frequencies from the laser spectra can generate 2D IR difference spectra that account for the background heat response.¹⁰¹ Another approach, albeit with increased sample demands, is to acquire spectra for the sample and reference sequentially, cycling in situ with a flow cell.⁹ Non-equilibrium 2D IR is increasingly being developed for the study of kinetic processes but also yields better detection limits due to the accurate background subtraction.^60,61 For example, >1 mM sample concentration is typically required for the detection of absolute 2D IR spectra for single vibrations, whereas a difference 2D IR spectrum of an Aha probe can be obtained at a ten-fold lower concentration of 0.1 mM.¹⁴ Substantial effort over the past decades has been directed to the optimization of 2D IR methodology generally that has facilitated protein studies.^14,101–108

Non-equilibrium spectroscopy also enables the study of kinetic processes on longer time scales than attainable for equilibrium dynamics and accesses larger scale protein motions that are frequently fundamental to function. Upon triggering a process, 2D IR spectra can be acquired successively with timing controlled by conventional delay states or through pulse shaping, respectively, accessing nanoseconds or up to days of time scales.^21,109–114 A scheme to access intermediate time scales of microseconds to milliseconds based on high repetition Yb-based lasers has been recently described.¹¹⁵ The requirement to rapidly trigger processes, often repeatedly if signal averaging is necessary, presents a technical challenge for analysis of fast kinetics. Rapid reaction initiation can be achieved by phototriggering, either via photoisomerization or a temperature jump.^21,109–112 Solvent heating by excitation of water vibrations can be applied to any protein, but globally impacts the protein, and the solvent heating causes challenges with signal distortion that must be addressed.¹¹¹ Photoisomerization by introducing azobenzene handles within proteins or ligands can impart local conformational control. The approach, however, does require significant protein engineering effort. Non-equilibrium 2D IR has been extensively applied to the amide vibrational region to provide mechanistic insight into protein folding, aggregation, and ligand binding.^21,109–114 The non-equilibrium dynamics of a transparent window probe, CO, ligated to myoglobin has been followed by 2D IR upon photodissociation.²¹ Aha probes in PDZ domains engineered with azobenzene phototriggers have been analyzed by 2D IR difference spectroscopy,⁶⁰ and CNSC-labeled PYP with a native chromophore for phototriggering are systems already well primed for non-equilibrium 2D IR spectroscopy.⁸⁴ Extended analyses of the time-resolved responses are likely forthcoming.

A promising future application of 2D IR is to characterize local conformational changes by analyzing the coupling of two labels. The strength of coupling depends on the distance and orientation of the two probes, so the amplitude and polarization dependence of cross bands can provide constraints for modeling the protein structure. This approach was demonstrated by Remorino and Hochstrasser to determine the structures of small peptides.¹¹⁶ Chalyavi et al. measured the coupling between two side chain probes for a trimer model of a Trp cage peptide.⁸⁷ As described earlier, dual labeling with donor visible azulene and acceptor Aha vibrational labels to quantify vibrational energy transport exemplifies the potential for gaining fundamental insight into protein biophysics.^70,71 Incorporation into proteins of two labels with similar but distinct frequencies within the transparent window is possible, but substantially increases the protein engineering demands. In addition, the vibrational couplings are often weak and challenging to detect. However, extension of the method to larger proteins would establish a powerful approach to characterize local and global conformational changes.

CONCLUSION

Generating a complete molecular-level description of protein function requires the ability to experimentally characterize them with residue-specific precision on all time scales. Transparent window 2D IR is an experimental approach that provides sufficient spatial and temporal resolution to tackle the complexity of proteins and measure even rapid dynamics at multiple specific locations in a protein to test if and how each contributes to the function. Although sensitivity remains a challenge, an assortment of frequency-resolved labels now are available for characterizing a diversity of proteins. The foundation of transparent window 2D IR has been laid, and the approach is primed for expanded applications to advance understanding of protein biophysics.

DATA AVAILABILITY

Data sharing is not applicable to this article as no new data were created or analyzed in this study.