1.

Introduction

Protein molecules are among the most interesting biological structures since they are involved in functioning of living organisms. Complicated spatial structure of a protein macromolecule determines its functional activity. Even minor structural changes may cause substantial variations in enzymatic activity.

Computer simulations based on molecular dynamics show that protein macromolecules may exhibit low-frequency molecular vibrations presumably related to function-related motions of large domains of molecules.^1–4 Most computations were performed for normal modes with frequencies of $<100{\mathrm{cm}}^{-1}$. A monotonic decrease in the number of vibrational modes with an increase in frequency to $300{\mathrm{cm}}^{-1}$ was demonstrated in Ref. 2. Such calculations stimulate the spectroscopic study at frequencies of several terahertz (several hundreds of wavenumbers). The recent interest in such experiments is due to the progress in THz technologies.^5,6 Amide VI and VII bands in protein spectra at frequencies of about 600 and $200{\mathrm{cm}}^{-1}$, respectively, must be conformation sensitive bands.^7–9

Raman and IR spectroscopic techniques are efficient tools for the study of molecular structures. Measurements in the fingerprint range have shown that several bands in protein spectra can be used to characterize protein structure. In particular, amide I and III bands can be used to determine secondary structure.^10–13 The bands of disulphide bridges and tyrosine doublet are sensitive to structural changes, induced by various external effects. Raman spectroscopy was used in Refs. 1415.–16 to study the influence of chemical reagents on protein molecules. The results show intensity redistribution of the bands of disulphide bridges related to their conformational transitions. Such transitions may cause changes of a protein molecule as a whole.

Rapid progress in THz spectroscopy necessitates the analysis of the low-frequency spectral bands that can presumably be used to additionally characterize protein structure. Note that the existing assignments based on IR and Raman data are tentative and insufficient. In one of the first works using low-frequency Raman spectroscopy, the spectra of crystalline lysozyme were compared with the spectra of the protein solution.¹⁷ The bands at 25, 75, 115, and $160{\mathrm{cm}}^{-1}$ were detected in the crystalline sample. The first band was missing in the spectrum of solution. The authors assigned the bands at 25 and $75{\mathrm{cm}}^{-1}$ to intermolecular and torsional vibrations, respectively. Several proteins were studied in Ref. 18 using low-frequency Fourier transform infrared (FTIR) spectroscopy. An attempt was made at the assignment of spectral bands with the aid of approximation of generally similar protein spectra with Lorentzian components. $N$-methylacetamide was used in Ref. 19 to simulate a peptide bond of protein at several temperatures. The bands peaked at 190 and $280{\mathrm{cm}}^{-1}$ were assigned to torsional motion and in-plane bending, respectively. The analysis of a few H-bonded systems was used to assign the band at about $100{\mathrm{cm}}^{-1}$ to the vibrations of H-bonds. A further analysis of the low-frequency IR and Raman of $N$-methylacetamide, lysozyme, and aprotinin using theoretical models proved the assignment.²⁰ The band at $290{\mathrm{cm}}^{-1}$ appears to be sensitive to the structure of water in protein solutions. Low-frequency spectral range was also discussed in Refs. 21 and 22, in particular, the bands at 230, 280, and $310{\mathrm{cm}}^{-1}$ were assigned to intramolecular vibrations, and the band at $100{\mathrm{cm}}^{-1}$ was assigned to intermolecular vibrations. Low-frequency Raman spectra of DNA films and lysozyme were studied in Ref. 23.

Far-IR spectra of native and structurally modified proteins were compared in Ref. 24. Factors that lead to denaturation (pH and temperature) cause spectral changes in the low-frequency range of several proteins. Different combinations of experimental parameters lead to different spectral changes that can be due to conformational changes of the protein or its dimerization. Two proteins with substantially different secondary structures [chymotrypsin and bovine serum albumin (BSA)] were studied in Ref. 25 using low-frequency FTIR and Raman spectroscopic measurements. The spectra of native proteins were compared with the spectra of thermally and chemically denatured molecules. Spectral data show that the effect of thermal denaturation is stronger than the effect of chemical agents that cause the cleavage of disulphide bonds in proteins.

Generally, protein structures are predominantly studied using spectral data of fingerprint range, whereas low-frequency data are poorly discussed in the literature. As distinct from the fingerprint range, the low-frequency range of vibrational spectra does not contain bands of single amino acids. Summarizing the existing data, we conclude that the low-frequency bands can be assigned to NH bending and out-of-plain vibrations, CN torsions, C=O out-of-plain vibrations, CCC bending vibrations, and CNC and CCN deformations. Thus, vibrations can be classified as skeletal or as vibrations that belong to almost all amino acids. Alternatively or additionally, the low-frequency bands can be assigned to correlated vibrations in large fragments (domains) of protein molecules.

The purpose of this work is to compare and analyze low-frequency Raman and FTIR spectra of several proteins with different secondary structures.

2.

Experimental Methods

In the experiments, we use chymotrypsin from Samson-Med, bovine serum albumin (BSA) from MP Biomedicals (Cat. no. 160069), lysozyme from MP Biomedicals (CAS 9001-63-2), fibrinogen from Sigma (CAS 9001-32-5), ovalbumine from Worthington (Cat. no. 3056), concanavalin from MP Biomedicals (Cat. no. 150710), human serum albumin from Sigma (CAS 70024-90-7), lactalbumin from MP Biomedicals (Cat. No. 102128), collagen from MP Biomedicals (Cat. No. 160084). Table 1 shows the average content of elements of secondary structure in proteins under study. Note that fibrinogen and collagen differ from the remaining proteins, since the former can be classified as superhelical (or coiled-coil) proteins.

Table 1

The content of elements of secondary structure in proteins under study.

Lyophilized protein powders are used in Raman measurements. Thin pressed tablets of the powders with masses ranging from 2.5 to 10 mg and thicknesses of about $200\mu \mathrm{m}$ are employed in the FTIR measurements.

Raman spectra are measured with the aid of a Thermo Scientific DXR Raman confocal microscope with a $10\times$ objective at an excitation wavelength of 532 nm, a maximum power of 10 mW, and a spectral resolution of $5{\mathrm{cm}}^{-1}$.

FTIR measurements are performed using a Thermo Scientific Nicolet-6700 FTIR spectrometer with a spectral resolution of $2{\mathrm{cm}}^{-1}$. Each spectrum results from averaging over 500 scans.

We employ background subtraction based on rolling-circle filter.⁴⁰ The filter parameters are identical for the spectra in each figure, and subtraction is performed for each spectral interval. A typical radius of the filter with respect to wavenumbers is about $\mathrm{10,000}{\mathrm{cm}}^{-1}$. The method of Brandt et al.⁴¹ makes it possible to approximate a single spectrum using a weighted sum of several spectra and a polynomial. The low-frequency Raman spectra are processed using $\mathrm{R}(\nu )$ representation.⁴² We choose the method or combination of methods of processing to most clearly present spectral variations.

3.

Results and Discussion

We analyze FTIR spectra of proteins in three spectral intervals to reveal conformation-sensitive bands.

Figure 1 compares FTIR spectra of the proteins in a spectral interval of 70 to $200{\mathrm{cm}}^{-1}$. The spectral curves result from the following sequential processing. First, background contributions are equalized in a spectral interval of 70 to $600{\mathrm{cm}}^{-1}$ using the method for comparison of Ref. 41. Then, almost identical linear backgrounds are subtracted in a spectral interval of 70 to $200{\mathrm{cm}}^{-1}$. Finally, each spectrum is normalized by its maximum in this interval. Hereafter, averaged spectra are presented and the intensity error is 5% to 7%. Measured optical densities for the proteins are close to each other, which are seen in the inset to Fig. 1, where we compare spectra without background subtraction and normalization.

Fig. 1

Processed FTIR spectra of (1) lysozyme, (2) BSA, (3) chymotrypsin, (4) ovalbumine, and (5) concanavalin (see text for details). The inset shows the FTIR spectra of (triangles) BSA and (solid line) concanavalin without background subtraction and normalization.

It is seen that the spectra of $\alpha$-helical proteins (BSA and lysozyme) exhibit a developed shoulder at a frequency of $150{\mathrm{cm}}^{-1}$, which has significantly lower intensity in the spectra of $\beta$-sheet proteins (concanavalin, chymotrypsin, and ovalbumine). Concanavalin has the lowest content of $\alpha$-helical fragments, and the corresponding spectrum has the lowest intensity in an interval of 130 to $190{\mathrm{cm}}^{-1}$. However, the percentage of secondary structure elements cannot be determined using the intensities in this spectral interval.

Figure 2 presents FTIR spectra in an interval of 200 to $280{\mathrm{cm}}^{-1}$. A rolling circle filter is used to subtract backgrounds. We clearly observe a decrease in the spectral intensity in an interval of 220 to $260{\mathrm{cm}}^{-1}$ in a series concanavalin, chymotrypsin, ovalbumine, lysozyme, and BSA. Peak intensity (regardless of the frequency) and integral intensity (with integration over a spectral interval of 200 to $280{\mathrm{cm}}^{-1}$) can be determined for each spectrum. Both peak (squares in the inset) and integral (circles in the inset) intensities of the band monotonically increase with an increase in the content of $\beta$-sheets. Significant intensity difference for concanavalin can be related to the absence of both $\alpha$-helical fragments and disulphide bridges.

Fig. 2

Processed FTIR spectra of (1) concanavalin, (2) chymotrypsin, (3) ovalbumine, (4) lysozyme, and (5) BSA (see text for details). The inset shows plots of normalized (squares) peak and (circles) integral intensities versus relative content of $\beta$-sheets.

Rolling circle filter is also used to obtain spectral curves in an interval of 275 to $375{\mathrm{cm}}^{-1}$ (Fig. 3). Each spectrum is normalized by the intensity at $318{\mathrm{cm}}^{-1}$, which is the peak position for the $\alpha$-helical proteins. These proteins exhibit almost symmetric bands, whereas complicated band shapes are typical of $\beta$-sheet proteins. Under the above normalization, the signal intensity at $333{\mathrm{cm}}^{-1}$ almost monotonically increases with an increase in the content of $\beta$-sheets (see inset to Fig. 3).

Fig. 3

Processed FTIR spectra of (1) concanavalin, (2) chymotrypsin, (3) ovalbumine, (4) lysozyme, and (5) BSA (see text for details). The inset shows plot of signal intensity at $333{\mathrm{cm}}^{-1}$ versus relative content of $\beta$-sheets.

Several conformation sensitive intervals can also be selected in Raman spectra of the proteins under study.

The spectra in Fig. 4 are obtained using the following procedure. First, $\mathrm{R}(\nu )$ representation is employed.^19–21,42 Then, the background contributions are equalized in a spectral interval of 70 to $600{\mathrm{cm}}^{-1}$ using the method for comparison of Ref. 41. Finally, each spectrum is normalized by its integral intensity in an interval of 40 to $200{\mathrm{cm}}^{-1}$. The spectra of $\beta$-sheet proteins are close to each other [panel (a)] and substantially differ from the spectra of proteins with significant contents of $\alpha$-helices. Recall that collagen contains left-hand $\alpha$-helices.

Fig. 4

Processed Raman spectra of (a) (1) ovalbumine, (2) chymotrypsin, and (3) concanavalin, and (b) (4) fibrinogen, (5) lysozyme, (6) collagen, and (7) BSA (see text for details).

Each spectral curve in Fig. 5 results from background subtraction using the rolling circle filter and normalization by the integral intensity in an interval of 180 to $470{\mathrm{cm}}^{-1}$. The spectra are sorted from top to bottom in ascending order with respect to the content of $\alpha$-helices and descending order with respect to the content of $\beta$-sheets. The spectral differences correlate with neither contents of secondary structure elements nor molecular weights (sizes) of the protein molecules. However, a general low-frequency shift of the spectra with an increase in the content of $\alpha$-helices may be noted.

Fig. 5

Processed Raman spectra of (1) concanavalin, (2) chymotrypsin, (3) lactalbumin, (4) ovalbumin, (5) fibrinogen, (6) lysozyme, (7) BSA, (8) human serum albumin, and (9) collagen.

As was mentioned in Sec. 1, the low-frequency bands can be assigned to skeletal vibrations or vibrations of identical fragments of different amino acids. Therefore, low-frequency spectra must be insensitive to variations in amino acid composition of proteins. However, we observe significantly different Raman spectra of the proteins (Fig. 5). The spectral bands can hardly be assigned to correlated vibrations of large fragments of molecules or subglobular vibrations due to the absence of dependences on molecular weights of proteins. At the same time, the differences of Raman spectra cannot be interpreted using differences of the contents of the secondary structure elements. In this regard, we assume that such differences can be due to the differences of tertiary and/or quaternary structures.

Finally, we compare FTIR and Raman spectra of globular and nonglobular (superhelical) proteins. Figure 6 shows developed low-frequency spectral features of the nonglobular proteins (fibrinogen and collagen) that are absent in the spectra of all globular proteins under study (only the spectrum of chymotrypsin is presented). Almost identical backgrounds in Fig. 6 are obtained using the method of Ref. 41.

Fig. 6

FTIR spectra of (1) fibrinogen, (2) collagen, and (3) chymotrypsin.

The Raman spectra of nonglobular (superhelical) proteins (fibrinogen and collagen) generally differ from the spectra of globular proteins although the spectrum of ovalbumine is similar to the spectrum of fibrinogen.

It is known that amide bands, which can be used to estimate contents of secondary structure elements, dominate in the FTIR spectra of proteins in the fingerprint range. Only amide I and amide III bands are observed in Raman spectra of proteins. However, Raman spectra contain many bands assigned to single amino acids or atomic groups. Such features of the two spectroscopic techniques are manifested in this work, since the FTIR results can be interpreted in terms of contents of secondary structure elements.

Finally, note that the FTIR and Raman spectra in this work are in general agreement with the results for few proteins.^18,43–46

4.

Conclusions

Differences of the vibrational spectra of proteins under study in several spectral intervals can be interpreted using the differences of the secondary structures.

In particular, $\alpha$-helical proteins exhibit an increase in spectral intensity in an interval of 130 to $190{\mathrm{cm}}^{-1}$ in both FTIR and Raman spectra.

For the FTIR spectra, integral intensity in an interval of 220 to $260{\mathrm{cm}}^{-1}$ and intensity at a frequency of $333{\mathrm{cm}}^{-1}$ almost monotonically increase with an increase in the content of $\beta$-sheets.

Low-frequency FTIR spectra of fibrinogen and collagen (nonglobular proteins) exhibit developed spectral features that are absent in the spectra of globular proteins. Such spectral features can be related to superhelical structure of the proteins.

Noticeable differences of the Raman spectra of proteins in an interval of 190 to $470{\mathrm{cm}}^{-1}$ do not correspond to the differences of secondary structures or molecular weights (sizes) of the protein molecules (on the assumption on linear dependences of Raman intensities on contents of secondary structure elements).