Review Infrared spectroscopy of proteins Andreas Barth ��� Department of Biochemistry and Biophysics, The Arrhenius Laboratories for Natural Sciences, Stockholm University, S-106 91 Stockholm, Sweden Received 5 January 2007 received in revised form 18 June 2007 accepted 19 June 2007 Available online 28 June 2007 Abstract This review discusses the application of infrared spectroscopy to the study of proteins. The focus is on the mid-infrared spectral region and the study of protein reactions by reaction-induced infrared difference spectroscopy. �� 2007 Elsevier B.V. All rights reserved. Keywords: Infrared spectroscopy FTIR Protein Protein structure Protein function Secondary structure Water Amide I Amino acid side chain Enzyme activity Difference spectroscopy 1. Introduction Infrared spectroscopy is one of the classical methods for structure determination of small molecules. This standing is due to its sensitivity to the chemical composition and architecture of molecules. The high information content in an infrared spectrum carries over also to biological systems. This makes infrared spectroscopy a valuable tool for the investigation of protein structure [1���11] of the molecular mechanism of protein reactions [2,10���35] and of protein folding, unfolding and misfolding [7,10,36���44]. The wealth of information in the infrared spectrum can be exploited even for biological systems that are larger than proteins [45���49]. A striking example is the possibility to identify bacterial strains from the infrared spectrum and to differentiate and classify microorganisms [45]. Further advantages of infrared spectroscopy are a large application range from small soluble proteins to large membrane proteins, a high time resolution down to 1 ��s with moderate effort, often a short measuring time, the low amount of sample required (typically 10���100 ��g) and the relatively low costs (top class spectrometers for 40000 ��� or $). These advantages for protein research are widely recognised in the academic world, but, surprisingly, the breakthrough of infrared spectroscopy in commercial protein analysis has still to come. This review discusses the application of infrared spectro- scopy to the study of proteins. The focus is on the mid-infrared spectral region and the study of protein reactions by reaction- induced infrared difference spectroscopy. 2. Absorption of infrared light The absorption of infrared radiation excites vibrational transitions of molecules. In the mid- and far-infrared spectral regions this is generally the case when the frequencies of light and vibration are equal and when the molecular dipole moment changes during the vibration. Since vibrational frequency and probability of absorption depend on the strength and polarity of the vibrating bonds, they are influenced by intra- and inter- molecular effects. The approximate position of an infrared ab- sorption band is determined by the vibrating masses and the type of bond (single, double, triple), the exact position by electron withdrawing or donating effects of the intra- and intermolecular environment and by coupling with other vibrations. The strength of absorption increases with increasing polarity of the vibrating bonds. In protein science, the effect of the environment on vibrational frequencies is often a telltale of how proteins work. Basically all polar bonds contribute to the infrared absorp- tion. This is at the same time the crux and the strength of infrared Biochimica et Biophysica Acta 1767 (2007) 1073���1101 www.elsevier.com/locate/bbabio Abbreviations: ATR, attenuated total reflection ��, in plane bending vibration FTIR, Fourier transform infrared IR, infrared ��w, wagging vibration ��t, twisting vibration ��r, rocking vibration ��, extinction coefficient NMA, N- methylacetamide ��, stretching vibration ��s, symmetric stretching vibration ��as, antisymmetric stretching vibration TDC, transition dipole coupling TDM, transition dipole moment ��� Tel.: +46 8 162452 fax: +46 8 155597. E-mail address: Andreas.Barth@dbb.su.se. 0005-2728/$ - see front matter �� 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.bbabio.2007.06.004

spectroscopy���a crux, because the spectrum of larger molecules is composed of many overlapping bands with the consequence that much information can be hidden under broad, featureless absorption bands a strength, because nearly all biomolecules absorb infrared radiation. The latter brings with it that there is no need to label biomolecules to make them detectable. This notion of infrared spectroscopy being a marker-free technique has recently received an interesting twist in a study that reported a genetically encoded CN infrared label which was used to probe ligand binding to myoglobin [50]. The infrared spectrum is plotted against the inverse of the wavelength, the wavenumber v, �� which is proportional to the transition energy and has the unit cm���1. The horizontal coordinate of the spectrum runs from high wavenumbers to low wavenumbers according to a recommendation of the International Union of Pure and Applied Chemistry (IUPAC) [51,52]. This is equivalent to running from small wavelength to large wavelength as usual in ultraviolet-visible spectroscopy. The convention is particularly important for the near-infrared spectral range where some spectra are plotted against wave- length and others against wavenumber. The infrared spectral region is adjacent to the visible spectral region and extends from 0.78 ��m to about 1000 ��m. It can be further subdivided into the near-infrared region from 780 nm to 2.5 ��m, the mid-infrared region from 2.5 ��m to 50 ��m and the far-infrared region from 50 ��m to 1000 ��m. The latter region is also called terahertz frequency regime. The mid-infrared spectral range extending from 2.5 to 50 ��m corresponds to 4000 to 200 cm���1, which corresponds to frequencies of 1013 to 1014 Hz. Thermal energy kT at room temperature corresponds to ���200 cm���1 implying that absorption in the mid-infrared spectral range is generally from the vibrational ground state the first excited vibrational state. Infrared spectroscopy is one variant of vibrational spectro- scopy. Other variants are Raman spectroscopy, reviewed in refs. [53���61], and photoacoustic spectroscopy which provide essentially the same information. Thus, some of the examples in the next section are from Raman studies. 3. Information that can be derived from the infrared spectrum 3.1. Chemical structure of the vibrating group The chemical structure of a molecule is the dominating effect that determines vibrational frequencies via the strengths of the vibrating bonds and the masses of the vibrating atoms. In spite of that, the chemicalstructureof a protein cannotbededuced fromthe infrared spectrum because of many overlapping bands. Changes in chemical structure however can be detected and an important example is a change of protonation state of side chains which is often essential for protein function. Here, infrared spectroscopy seems to be the method of choice since the protonation state of most side chains is reflected in the spectrum. Some examples are: protonation of Asp and Glu residues accompanies proton pumping by bacteriorhodopsin [14���17,23], electron transfer reactions [12,13], Ca2+ release from the Ca2+ ATPase [62] and seems to provide a mechanism of charge compensation when the negatively charged ATP binds to the Ca2+-ATPase [63]. As for Asp and Glu residues, the protonation state of other catalytically active side chains can be characterised by infrared spectroscopy, as done for example for His and Tyr residues of photosystem II [64,65] and bacteriorhodopsin [66���68]. Other examples for an alteration of chemical structure that can be monitored by infrared spectroscopy are protein modi- fications like phosphorylation [69,70] and chemical reactions catalysed by enzymes [71���78], reviewed in Ref. [27]. 3.2. Chemical properties of neighbouring groups in a molecule The electron density of a given bond in a molecule is influenced by neighbouring groups within the molecule via inductive and mesomeric effects. The latter is the reason for amide groups absorbing at lower wavenumbers than carbonyl groups of carboxylic acids [79]. 3.3. Redox state Redox reactions are the basis of the energy delivering processes in living organisms. They affect the electron density distribution of a given molecule and thus will alter its vibrational spectrum. This is illustrated well by model studies on isolated biological cofactors involved in photosynthesis [80���83] that allowed the assignment of signals in the protein spectra to specific molecular groups of the cofactors and in consequence statements about their protein environment [12,13, 21,22]. 3.4. Bond parameters As pointed out by Deng and Callender [53], vibrational spectroscopy is exceptionally sensitive to changes in bond strength since a change of 0.02% can be easily detected. As bond strength and bond length are directly related [59,84���86], bond distortions smaller than 1 pm in the course of a catalytic reaction can be monitored with astonishing accuracy. Deng and Callender concluded: ���although an oversimplification, it can be said that the resolution of vibrational spectroscopy picks up where diffraction and multidimensional nuclear magnetic resonance (NMR) techniques leave off, at ca. 0.2 ��, and extends down to much lower lengths���. The sensitivity to bond parameters has been exploited in quantitative studies of, for example, chymotrypsins [87,88], subtilisins [87], myosin [89], Ras [90���92], and Ca2+-ATPase [93,94]. 3.5. Bond angles and conformation Vibrations in adjacent parts of a molecule are often coupled and this coupling depends on details of the molecular geometry. Therefore, coupling can provide insight into the three-dimen- sional structure of molecules. Examples are (i) the two coupled CO vibrations in the COO��� group which make the carboxylate absorption sensitive to the mode of cation chelation [95���97] and has been used in studies of several Ca2+ binding proteins [98,99] (ii) the sensitivity of P\O or V\O stretching vibrations of phosphates or vanadates on the bond angle [100], applied to 1074 A. Barth / Biochimica et Biophysica Acta 1767 (2007) 1073���1101

myosin [89] and Ras [90,91] studies and (iii) transition dipole coupling of the amide groups of the protein backbone which makes this mode sensitive to the 3D structure of the backbone. 3.6. Hydrogen bonding Hydrogen bonds stabilize the structure of proteins and are essential for catalysis. Vibrational spectroscopy is one of the few methods that directly report on the strength of hydrogen bonds. As a general rule, hydrogen bonding lowers the frequency of stretching vibrations, since it lowers the restoring force, but increases the frequency of bending vibrations since it produces an additional restoring force [79]. Formation of a single hydrogen bond to a C_O group leads to a 20 cm���1 down shift for the methylacetate���water complex in an argon matrix [101,102] and of 25 cm���1 for propionic acid in an ethanol/ hexane (1:200) mixture [103]. Hydrogen bonding to PO2 ��� groups lowers the observed band position of the symmetric stretching vibration by 3 to 20 cm���1 and of the antisymmetric stretching vibration by 20���34 cm���1 [104���107] with a single hydrogen bond contributing 16 cm���1 in a nitrogen matrix [107]. This is due to two effects [106]: (i) less electron density in the P\O bonds and (ii) a relaxation of phosphate geometry towards the ideal tetrahedron form due to the reduced Coulomb repulsion. For the quantitative evaluation of band shifts induced by hydrogen bonding Badger and Bauer relationships [108] are often used. These are linear dependencies between the enthalpy of hydrogen bond formation ��H and the frequency of stretching vibrations [53,59,109,110]. For the carbonyl group of methyl acetate, a model for Asp and Glu, a downshift of 1 cm���1 corresponds to a favourable binding enthalpy of approximately 1.7 kJ/mol [53,59,110]. In Raman studies [54], favourablebinding enthalpies with respect to water as large as ���60 kJ/mol have been detected for C_O groups of substrates bound to enzymes. 3.7. Electric fields Similar to hydrogen bonding, the electric field produced by the environment modifies the electron density distribution of a vibrating group. A strong electric field has been detected for example in the active site of dehalogenase where it polarises the product of the catalytic reaction [55]. For carboxyl groups in the absence of hydrogen bonding (bands above 1740 cm���1), there is an inverse correlation of the C_O stretching frequency with the dielectric constant �� [103]. 3.8. Conformational freedom Besides band position and band intensity, the third spectral parameter, the band width, is also informative. Due to its short characteristic time scale on the order of 10���13 s, vibrational spectroscopy provides a snapshot of the sample conformer population. As the band position for a given vibration usually is slightly different for every conformer, heterogeneous band broadening is the consequence. Flexible structures will thus give broader bands than rigid structures and the band width is a measure of conformational freedom. It is possible to relate band width with entropy and thus to quantify entropic effects in catalysis [53]. Motional narrowing can counteract the effects of a heterogeneous environment for amide groups that are hydrogen bonded to bulk water [111���113]. For molecules that bind to proteins, the restriction of conformational freedom is a natural consequence of binding. This often reduces the band width by a factor of two [27]. The band width has been interpreted in studies of nucleotide binding to Ras [90,114], ubiquinone binding to cytochrome bo3 [115], substrate analogue binding to lactate dehydrogenase [53], nicotinamide dinucleotide binding to transhydrogenase [116], and of the phosphorylated Asp residue of the sarcoplasmic reticulum Ca2+-ATPase [69]. 4. Infrared spectrometers 4.1. Fourier transform infrared spectrometers Modern infrared spectrometers are usually Fourier transform infrared (FTIR) spectrometers. Their name originates from the fact that the detector signal of these spectrometers is related by a Fourier transformation to the measured spectrum. The heart of an FTIR spectrometer is an interferometer, like the Michelson interferometer shown in Fig. 1. It has a fixed and a movable mirror. The latter generates a variable optical path difference between two beams which gives a detector signal that contains the spectral information. Light emitted from the light source is split by a beam splitter: about half of it is reflected towards the fixed mirror and from there reflected back towards the beam- splitter where about 50% passes to reach the detector (black arrows in Fig. 1). The other half of the initial light intensity passes the beam splitter on its first encounter, is reflected by the movable mirror back to the beamsplitter where 50% of it is reflected towards the detector (grey arrows in Fig. 1). When the two beams recombine they interfere and there will be constructive or destructive interference depending on the optical path difference d. The instrument measures the light intensity relative to the position of the movable mirror and this is called an interfero- gram. It turns out that the interferogram is the Fourier transform of the spectrum. A second Fourier transform performed by a computer converts the measured data back into a spectrum. In total, a Fourier transform spectrometer performs two Fourier transformations: one by the interferometer, one by the computer. That the interferometer produces the Fourier transform of the spectrum is best seen when a monochromatic source is considered. Its spectrum S(��) is described by a delta function located at ��0. Depending on the position of the movable mirror one obtains constructive or destructive interference at the detector and the detector signal I(d) varies as a cosine function with the mirror position which determines the optical path difference d. This cosine function and the delta function describing the monochromatic spectrum are related by a Fourier transformation. Thus, a second Fourier transformation per- formed by the computer generates the spectrum S(��). The main advantage of Fourier transform spectrometers is the rapid data collection and high light intensity at the detector 1075 A. Barth / Biochimica et Biophysica Acta 1767 (2007) 1073���1101

and in consequence the high signal to noise ratio. Therefore a spectrum can be recorded in as few as 10 ms. Different types of detectors, light sources and other optical components are used in different regions of the infrared spectral range. 4.2. Dispersive infrared spectrometers Dispersive infrared spectrometers are no longer used for routine measurements but are still valuable for special applica- tions like time-resolved measurements. Briefly, a dispersive infrared spectrometer consists of a light source, optics (often mirrors) that focus the light on the sample and then on the entrance slit of a monochromator, a monochromator, another mirror to focus the light on a detector and a detector. The positioning of the monochromator behind the sample avoids the detection of heat radiation from the sample. The disadvantage of these instruments is that the incident light is split into its spectral components. Therefore the light intensity reaching the detector is low, which results in a signal to noise ratio that is much worse than that of Fourier transform infrared spectrometers. 5. Sampling techniques 5.1. Transmission measurements In transmission measurements, the infrared light passes a cuvette containing the sample before it reaches the detector. If the sample is homogeneous, it absorbs light according to the Beer���Lambert law A������ v �� log��I0=I�� �� ce������d v where A is the absorbance, I0 the light intensity before the sample, I the light intensity after the sample, c the concen- tration of the absorbing substance, �� the molar absorption index or extinction coefficient and d the pathlength. ZnSe, BaF2 and CaF2 cuvettes are often used, the latter two have the advantage that excitation with UV light is possible. A simple demountable infrared cuvette consists either of two plane windows separated by a spacer that defines the pathlength or of one plane window and a window with a trough. Recently, a number of flow cells have been developed that enable mixing experiments [36,117���119]. One drawback of infrared spectroscopy of aqueous solutions is the strong absorbance of water in the mid-infrared spectral region (near 1645 cm���1) [120] which overlaps the important amide I band of proteins and some side chain bands (see below). When these protein bands are of interest, the strong water ab- sorption demands a short path length for aqueous samples, which is typically around 5 ��m, and in turn relatively high concentrations. Using 2H2O, the pathlength can be increased to 50 ��m and the concentration lowered because the water band is downshifted to ���1210 cm���1. Typical concentrations in infrared spectroscopy are 0.1 to 1 mM for proteins and 1 to 100 mM for small molecules. However, concentrations as low as 0.3 mg/ml have been reported [121] and the amount of protein required is low because of the small sample volume, typically in the 10��� 100 ��g range, and can be as low as 50 ng [122]. A promising technique to further reduce the amount of material is surface- enhanced infrared spectroscopy [123]���the infrared analog of surface enhanced Raman spectroscopy. It enables studies of protein monolayers because protein absorption is enhanced by typically a factor of 100 when the protein is absorbed on a film or suspension of metal particles. Possible disadvantages are the sensitivity of band intensities on the structure of the metal surface and structural modifications due to protein interaction with the metal surface. Fig. 1. Scheme of a Fourier transform infrared spectrometer. See text for further explanation. Republished from [11]. �� 2006 Nova Science Publishers. 1076 A. Barth / Biochimica et Biophysica Acta 1767 (2007) 1073���1101

5.2. Attenuated total reflectance The attenuated total reflectance (ATR) technique [26,124��� 126] is illustrated in Fig. 2. Compared to transmission ex- periments, it avoids the handling problems which are caused by the required short pathlength. In an ATR experiment a sample is placed on a crystal with an index of refraction that is larger than that of the sample and typically larger than 2. Infrared light is coupled into the crystal and directed towards the sample inter- face at such an angle of incidence that it is totally reflected at the interface between sample and crystal. After one or several reflections the measuring light leaves the crystal and is focused on the detector. Upon reflection at the interface between sample and crystal, light penetrates into the sample. This so-called evanescent wave has the same frequency as the incoming light but the amplitude of the electric field decays exponentially with the distance from the interface. The evanescent wave may be absorbed by the sample and thus the light reaching the detector carries the information about the infrared spectrum of the sample. The penetration depth is on the order of the wavelength, which means that the optical thickness of the sample is small enough for measurements of aqueous solutions. The penetration depth depends also on the angle of incidence and the ratio of the indices of refraction of crystal and sample. For protein samples, usually a protein film is prepared on the crystal surface [121,125,127,128], often simply by drying, and a buffer is placed on top of the film. The thickness of the buffer layer does not influence the measured spectrum. The advantage of the method is that often the buffer can be exchanged without disturbing the film. This makes sample manipulations relatively simple and makes the method very flexible. The disadvantage is that the preparation of a stable film can be difficult and some- times impossible. The film might also detach when the buffer is exchanged. For these difficult cases and for soluble proteins it is advan- tageous to separate a sample compartment close to the ATR crystal from a reservoir by a dialysis membrane. In this way the medium in the reservoir can be altered without disturbing the protein sample in the sample compartment [129���132]. 6. Time-resolved infrared spectroscopy 6.1. Overview Three of the most common time-resolved infrared techniques that have been applied to study proteins will be briefly discussed here: the rapid scan technique, the step scan technique and single wavelength measurements. The former two are performed on FTIR spectrometers, while the latter is done on a dispersive instrument. Not discussed here will be the pump- probe technique and stroboscopic Fourier transform infrared spectroscopy. All techniques mentioned and biological or photochemical applications thereof have been reviewed before [20,23,133���143]. 6.2. The rapid scan technique In the rapid scan technique the movable interferometer mirror is moved at a maximum speed of 10 cm/s. From one complete forward and backward movement of the mirror, up to 4 spectra can be obtained which results in a maximum time resolution of about 10 ms at 12 cm���1 optical resolution. A single experiment can yield a full series of time-resolved spectra. 6.3. The step scan technique Data recording in an FTIR spectrometer is not continuous but done at discrete positions of the movable mirror. In the step scan technique the movement of the movable mirror is stopped at these positions, a time-resolved experiment performed and the time course of intensity at the detector recorded. Then the mirror is moved to the next position and the experiment is repeated. One obtains a series of time-resolved intensity measurements at the different mirror positions. When intensity data of all mirror positions at a given time are combined, one obtains the interferogram at that time which can be transformed into a spectrum. Accordingly all data are reshuffled to obtain a time- resolved series of interferograms which is then transformed into a series of time-resolved spectra. The time resolution of this technique can be of the order of nanoseconds and is limited by the response times of detector and electronics. A requirement for the step scan technique is that the experiment can be accurately reproduced at least several hundred times, since kinetic traces at typically about 600 mirror positions have to be sampled at 4 cm���1 optical resolution [144]. This requirement is either met by cyclic or reversible systems where the reaction of interest can be repeated many times with the same sample, by repeating the experiment at different small sample spots for each interfero- meter position [145], or by using a flow cell to conveniently refill the infrared cuvette with fresh material [146]. 6.4. Single wavelength measurements Single wavelength measurements are done on a dispersive instrument. The monochromator is set to a wavenumber of interest and the intensity change at the detector during the time- resolved experiment is recorded. The advantage of this method is that a kinetic trace can be obtained from a single experiment. However, the relatively low signal to noise ratio of dispersive instruments has the consequence that only relatively large absorbance changes (��AN0.002���0.01) can be resolved in a single experiment. The lower limit of sensitivity applies when the infrared band of interest is broad which allows one to increase the spectral width that is detected. The signal to noise Fig. 2. A typical ATR setup. Republished from [11]. �� 2006 Nova Science Publishers. 1077 A. Barth / Biochimica et Biophysica Acta 1767 (2007) 1073���1101