Raman Optical Activity of a Cyclic Dipeptide Analyzed by Quantum

Publication Date (Web): May 29, 2014 ... An important concern for the explicit solvent models is the number of MD snapshots that lead to a good agreem...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/JPCB

Raman Optical Activity of a Cyclic Dipeptide Analyzed by Quantum Chemical Calculations Combined with Molecular Dynamics Simulations Hiroyasu Urago, Torao Suga, Taiki Hirata, Hiroaki Kodama, and Masashi Unno* Department of Chemistry and Applied Chemistry, Graduate School of Science and Engineering, Saga University, Saga 840-8502, Japan S Supporting Information *

ABSTRACT: Raman optical activity (ROA) measures the different intensity of right- and left-circularly polarized Raman scattered light and provides information on chirality associated with vibrational modes. Because of a high sensitivity to subtle structural and environmental changes, interpretations of ROA spectra usually rely on quantum chemical simulations. Recent advances in computational chemistry allow us to consider explicit solvent models that are derived from molecular dynamics (MD) simulations to compute the Raman and ROA spectra. An important concern for the explicit solvent models is the number of MD snapshots that lead to a good agreement between the observed and calculated spectra. In the present study, we measured the Raman and ROA spectra of cyclo(L-Ala-Gly) and then simulated the spectra using density functional theory combined with MD simulations. Although cyclo(L-Ala-Gly) is a relatively rigid cyclic molecule, boatup and boat-down conformations were found from the MD calculations. Because the Raman spectra of the two conformations are similar except for a lower frequency region, ∼10 MD snapshots are capable of reproducing the main features of the observed Raman spectra. In contrast, a larger number of MD snapshots was required to reproduce the ROA spectra. In the middle freqency region of 800−1580 cm−1, an average of ∼40 spectra led to good agreement between the observed and calculated spectra. On the other hand, the low (0−800 cm−1) and high (1580−1800 cm−1) frequency regions require more than 60 and 120 MD snapshots, respectively. The Raman and ROA spectra in the low frequency region are relatively broad, and such spectral features require a larger number of averaged spectra. The high frequency region of the spectra consists of an amide I band, which is primarily a CO stretching vibration. Since both the ROA intensity and frequency of the amide I band are highly sensitive to structural and environmental differences, a large number of the spectra need to be averaged to reproduce the small negative features in the observed ROA spectra.

■

INTRODUCTION In recent years, Raman optical activity (ROA) has emerged as a promising approach for gathering information on molecular conformations.1,2 ROA measures the different intensity of rightand left-circularly polarized Raman scattered light and provides information on chirality associated with vibrational modes. Since most of the biological samples are chiral, this technique is considered an ideal tool for the study of biological molecules in aqueous solution.1,3 Because ROA is very sensitive to details of the molecular structure and its environment, interpretations of ROA spectra usually rely on quantum chemical simulations.4−7 One of the issues that we should consider for the simulations of the vibrational spectra is the fact that the observed spectra represent a sum of contributions from individual conformers.4,8−18 Another important issue is the aqueous environment of a solute molecule, as it stabilizes the conformation and charge states.12,14,19−21 In addition, a direct interaction with solvent, such as a hydrogen bond, will affect the vibrational spectra.22,23 A practical approach to including the solvent environment in the quantum chemical calculation is the use of a polarizable continuum model (PCM).24 In this implicit solvent model, the © 2014 American Chemical Society

solvent is represented as an infinite, homogeneous, and polarizable dielectric medium and the solute molecule is placed in a molecule-shaped cavity in this medium. Although PCM is a popular solvent model that is computationally inexpensive, recent advances in computational chemistry allow us to use quantum mechanical/molecular mechanical (QM/MM) methods, where the MM region comprises explicit solvent models derived from molecular dynamics (MD) simulations.25−28 In fact, several studies have shown that the explicit solvent model leads to an excellent level of agreement between experimental and calculated spectra.25,26,28 A practical concern for the explicit solvent models is the number of MD snapshots required for a reasonable agreement with the experimental spectra. Cheeseman et al.26 used methylβ-D-glucose, which shows two stable conformations with ratios of 60:40, as a model system. A satisfactory result was achieved by 60%:40% weighted averages of the spectra generated from the 16 MD snapshots for each of the two conformations. Received: April 21, 2014 Revised: May 25, 2014 Published: May 29, 2014 6767

dx.doi.org/10.1021/jp503874z | J. Phys. Chem. B 2014, 118, 6767−6774

The Journal of Physical Chemistry B

Article

Hopmann et al.25 analyzed the Raman and ROA spectra for (R)-(+)-lactamide, which is a flexible small molecule. MD calculations were also carried out to consider an explicit solvent environment. They examined the spectra generated based on 10−1000 MD snapshots and estimated that 50 and 500 MD snapshots are needed for a reliable simulation of the Raman and ROA intensities, respectively. Although their study provides an important clue regarding the minimum number of structures needed for the averaging, we must keep in mind that the explicit solvent model using the MD simulations involves not only different hydrated structures but also conformational fluctuations of the solute molecule. Since lactamide exhibits several stable conformations, the effects of fluctuations of both solvent and solute structures are likely to be included in the averaged spectra. Thus, it is possible that 500 MD snapshots were required for lactamide because of its high conformational flexibility. In addition, the number of necessary MD snapshots could be different between the low and high frequency regions. This consideration is based on the fact that normal modes in the lower frequency region mainly consist of delocalized skeletal deformations, whereas higher frequency modes are relatively localized. In this study, we used cyclo(L-Ala-Gly) as a model to examine the minimum number of MD snapshots required to reproduce realistic Raman and ROA spectra. Cyclic dipeptides are also known as 2,5-diketopiperazines, and the structure of cyclo(L-Ala-Gly) is shown in Figure 1. Although a cyclic

MeOH−ether−petroleum ether (7.3 g, 62%). Hydrogen gas was bubbled through a suspension of the product (4.41 g, 15 mmol) and Pd-black in MeOH at room temperature and atmospheric pressure for 8 h. The catalyst was removed and filtrate was refluxed for 10 h. The solvent was evaporated, affording a cyclic dipeptide (1.85 g, 96%): mp 231−234 °C; 1H NMR (DMSO-d6) δ 8.11 (d, 1H), 7.93 (d, 1H), 3.83 (q, J = 6.3 Hz, 1H), 3.71 (s, 2H), 1.25 (d, J = 6.9 Hz, 3H). The melting point was taken on a Thomas-Hoover capillary melting point apparatus and is uncorrected. The NMR spectrum was taken on a JEOL AL-300 spectrometer, and tetramethylsilane was used as an internal reference. Sample Preparations. We used 0.50 M cyclo(L-Ala-Gly) aqueous solutions for the ROA measurements. N-Deuterated isotopomer was prepared from solutions in D2O (Cambridge Isotope Laboratories Inc.; 99.9 atom % D). ROA Spectroscopy. The ROA instrument used in this study is based on an incident circular polarization scheme, and its basic design is identical to our near-infrared ROA spectrometer.29−31 A 532 nm light from a diode-pumped solid-state laser (Ventus532; Laser Quantum, Ltd., U.K.) was used to measure the Raman and ROA spectra. Circularly polarized light was generated by a Glan-laser polarizer (extinction ratio, 105:1) and a liquid crystal variable retarder (LRC-200-VIS; Meadowlark Optics Inc.). The laser beam was focused by a lens (F/15, f = 150 mm) into samples contained in a 3 × 3 × 48 mm quartz cuvette, and two 45° rod mirrors (Edmund Optics Inc.) were used to steer the laser light to the sample. The backscattered light from the sample was collected by an aspheric glass condenser lens (F/0.8, f = 20 mm) and refocused by an achromatic lens (F/1, f = 30 mm) to the end of a fiber bundle with a core diameter of 100 μm. A long-pass edge filter (OD > 6, CVI Laser, LLC) rejected the laser light before the refocusing lens. The light was directed into a spectrometer (SpectraPro 2300i; Princeton Instruments) equipped with a liquid nitrogen cooled UV-coated CCD detector (Spec10:256E; Roper Scientific Inc., Trenton, NJ). This spectrometer included an error-correction scheme32 that used a set of half-wave plates to reduce deterministic spurious signals. All spectra were taken at room temperature (∼25 °C), and custom-written software eliminated the noise spikes in the spectra caused by cosmic rays. All Raman spectra were calibrated using neat fenchone. The acquisition times for cyclo(L-Ala-Gly) samples were about 12 h. Figure S1 in the Supporting Information shows Raman and ROA spectra for standard samples, including a small organic molecule ((−)-α-pinene and (+)-α-pinene), a small biological molecule (L-alanine and D-alanine), and a protein molecule (lysozyme). The observed ROA spectra agree well with the reported spectra.1,4,32,33 An artifact in the ROA spectra can be expressed as the ratio defined as the intensity of an artificial signal divided by its parent Raman intensity. This ratio was estimated to be less than 1 × 10−4 for L- and D-alanine solutions. MD Calculations. Both the initial setup and the MD runs were performed with the AMBER11 program34 using an explicit representation of solvent molecules and the ff99SB allatom force field.35 Water was modeled by the TIP3P potential.36 A peptide molecule was surrounded by a periodic octahedral box of TIP3P water molecules measuring 18.8 × 18.8 × 18.8 Å. This led to a total of 486 waters. All LennardJones interactions were cut off at 8 Å, and a particle mesh Ewald method37 was applied to calculate the long-range

Figure 1. Structure and atom numbering for cyclo(L-Ala-Gly). Dihedral angles that characterize the peptide backbone conformation (ψ, ω, ϕ) are also shown.

dipeptide is relatively rigid, the MD simulations found two conformations in cyclo(L-Ala-Gly). Furthermore, the present study demonstrates that 10 MD snapshots are sufficient to provide a reasonable agreement between the observed and calculated Raman spectra. In contrast, many of the ROA bands exhibit different intensities and signs between the two conformations. Because of this high structural sensitivity, a larger number of MD snapshots (40−120) are required to obtain a good agreement between the observed and calculated spectra. The present study provides an important basis for the analysis of Raman and ROA spectra using explicit solvent models derived from MD simulations.

■

EXPERIMENTAL SECTION Synthesis of Cyclo(L-Ala-Gly). To a solution of Cbz-GlyOH (8.37 g, 40 mmol) and N-methylmorpholine (4.4 mL, 40 mmol) in THF (200 mL) was added ethyl chloroformate (5.2 mL, 40 mmol) at −15 °C. After 10 min a chilled solution of HAla-OMe HCl (6.43 g, 46 mmol) and N-methylmorpholine (5.1 mL, 46 mmol) in CHCl3 (90 mL) was added to the reaction mixture. The solution was stirred for 1 day at room temperature and evaporated. The residual oil was dissolved in CHCl3 (300 mL), and the solution was washed successively with 4% NaHCO3, 10% citric acid, and H2O, then dried over Na2SO4, and evaporated. The residue was recrystallized from 6768

dx.doi.org/10.1021/jp503874z | J. Phys. Chem. B 2014, 118, 6767−6774

The Journal of Physical Chemistry B

■

RESULTS AND DISCUSSION Experimental Raman and ROA Spectra. The upper four traces in Figure 2 show the observed Raman and ROA spectra

electrostatic interactions. The MD run was set up using the following protocol. First, the system was subjected to 1000 steps of minimization to remove close van der Waals contacts and to allow the formation of hydrogen bonds between solvent molecules and the peptide. In this stage, we kept the peptide molecule fixed and simply minimized the positions of the water. In the second step, 2500 steps of minimization were performed without the restraints, i.e., the entire system was minimized. As a third step, the system was then gradually heated to 300 K over 20 ps of constant volume dynamics. The temperature was controlled via Langevin dynamics38 using a collision frequency of 1.0 ps−1. The SHAKE algorithm39 was used to constrain bonds between hydrogen and heavy atoms, and the time integration step was set to 2 fs. After heating, the obtained system was simulated for 12.8 ns at 300 K and 1 atm, with the time integration step being set to 2 fs. DFT Calculations. All quantum chemical calculations were performed using the program Gaussian 09.40 For cyclo(L-AlaGly) with implicit hydration, geometry optimizations, the harmonic vibrational frequencies, and Raman and ROA intensities with 532 nm excitation were computed at the B3LYP/aug-cc-pVDZ level of density functional theory (DFT) by employing the PCM24 aqueous solvent correction. For the explicit solvent model, a two-layer ONIOM41 method was used to perform the QM/MM calculations of cyclo(L-Ala-Gly) surrounded by explicit water molecules. Initial geometries were obtained from snapshots of the MD simulation. The number of the surrounding water molecules was reduced to 200. In this process, water molecules closer to the solute molecule were selected by custom-written software. The QM region consisted of cyclo(L-Ala-Gly), and the MM region consisted of all of the water molecules. All MM calculations were performed using the parm96 parameters of the Amber force field42 and TIP3P parameters for water. The QM part of the system was computed at the B3LYP/aug-ccpVDZ level of theory. An electronic embedding scheme,43 which incorporates the partial charges of the MM region into the QM Hamiltonian, was used throughout. For geometry optimizations, the positions of the MM water molecules were frozen. The simulated spectra were generated assuming a Gaussian band shape with a half-width of 10 cm−1. The calculated vibrational frequencies were uniformly scaled to best match the experimental spectra. The calculated frequencies were scaled using a factor of 0.9617.44 The errors of the limited number of spectra to be averaged were evaluated as performed by Hopmann et al.25 First we calculated the sum of the intensities for the average of 250 spectra using eq 1: S=

∫ν̃

νf̃

|I250(ν)̃ | dν ̃

Figure 2. Observed and calculated Raman and ROA spectra of cyclo(L-Ala-Gly) in H2O (black) and D2O (blue) solutions. The observed Raman (a, b) and ROA (c, d) spectra are shown. The Raman (e, f) and ROA (g, h) spectra determined using QM/MM calculations with 250 MD snapshots are also shown. The ROA spectra are magnified by a factor of 3000.

of cyclo(L-Ala-Gly) with 532 nm excitation. Since the observed ROA spectra below 500 cm−1 contain large noise due to quartz cell walls, the spectra are only shown in the 550−1850 cm−1 region. The Raman spectrum (trace a) is characterized by bands at 1673, 1526, 1447, 1317, 1097, 890, and 762 cm−1, and these spectral features are consistent with those reported previously.45 The figure also displays the Raman spectrum for the N-deuterated derivative (trace b), and some of the observed bands are affected by the H/D exchange. In Figure 2, we also show the ROA spectra of cyclo(L-Ala-Gly) in both H2O and D2O solutions (traces c, d). The ROA spectrum exhibits some negative bands at ∼1450, 1317, and 890 cm−1, whereas positive bands are observed at 1300 and 766 cm−1. In addition, broad negative features are seen near 1670 and 700 cm−1. As shown in the figure, most of the ROA bands are shifted upon Ndeuteration. Molecular Dynamics Simulations. The conformations of cyclic dipeptides can be characterized by the dihedral angles ψ, ω, and ϕ, which are defined in Figure 1. Although the crystal structure of cyclo(L-Ala-Gly) has not been reported, a previous study using DFT calculations as well as vibrational spectroscopy reported the ψ and ϕ values as −18.9 and 34.6°.45 In the present study, we performed MD calculations to examine the structure in an aqueous environment. We placed a cyclo(L-AlaGly) molecule in an 18.8 Å octahedral box containing TIP3P water molecules, and a 12.8 ns MD run was executed using the AMBER11 MD package.34 Although a total of 25,600 snapshots

(1)

i

where I250(ν̃) is the 250 averaged Raman or ROA intensity at ν̃ cm−1. The errors for the averaged spectra In(ν̃) based on the number of n MD snapshots are approximately expressed as the deviations: ν̃

deviation =

∫ν ̃ f |In(ν)̃ − I250(ν)̃ | dν ̃ i

S

Article

(2) 6769

dx.doi.org/10.1021/jp503874z | J. Phys. Chem. B 2014, 118, 6767−6774

The Journal of Physical Chemistry B

Article

with a 0.5 ps interval were analyzed, 250 snapshots of the simulation were taken every 50 ps for the subsequent QM/MM calculations. The conformational distribution of the cyclic dipeptide for the selected 250 snapshots is represented as a Ramachandran plot46 in Figure 3. The temporal changes in the

Figure 4. Optimized geometries of cyclo(L-Ala-Gly). (A) Representative snapshot of the explicitly hydrated cyclo(L-Ala-Gly) with 200 water molecules. The QM region of the QM/MM calculation is illustrated by a ball and stick model. (B) Conformation 1. (C) Conformation 2. In panels B and C, only the solute molecule is shown. Black, blue, and red represent carbon, nitrogen, and oxygen atoms, respectively. Figure 3. Plot of the ψ and ϕ angles (indicated as black dots) for 250 MD snapshots of cyclo(L-Ala-Gly), superimposed on a Ramachandran plot. The data for the PCM model (models 1 and 2) and the conformations 1 and 2 are also plotted. See the text for details.

Table 1. Selected Dihedral Angles (deg) and Relative Free Energies (kJ mol−1) of Cyclo(L-Ala-Gly) conformation 1 conformation 2 model 1 (PCM)b model 2 (PCM)b model 3 (vac)d model 4 (vac)d

dihedral angles ψ, ω, and ϕ and their probability distributions are also shown in Figure S2 in the Supporting Information. As illustrated in Figure 3 and Figure S2 in the Supporting Information, the three dihedral angles mostly adopt values around 0°. However, a closer inspection of the ψ−ϕ plot shown in Figure 3 indicates the presence of two conformations: a major conformation with ψ ≈ −20°, ϕ ≈ 20° and a minor conformation with ψ ≈ 20°, ϕ ≈ −20°. On the basis of the 25,600 MD snapshots, the populations for the main (ψ < 0, ϕ > 0) and minor (ψ > 0, ϕ < 0) conformations are 66% and 14%, respectively. Geometries of Cyclo(L-Ala-Gly). To compute the Raman and ROA spectra, hydrated clusters with 200 water molecules were extracted from the 250 MD snapshots. For quantum chemical calculations, we used the QM/MM method to simulate the Raman and ROA spectra. The QM part comprises the solute molecule, whereas the water molecules were treated as the MM region. The positions of the MM water molecules were frozen during geometry optimizations, and Figure 4A displays a sample structure. In Figures 4B and 4C, we also show the representative structures for the main and minor conformations, which we denote as conformations 1 and 2, respectively. The dihedral angles ψ, ω, and ϕ for the two conformations are summarized in Table 1. As shown in Figure 4B, the conformation 1 is the boat-up conformation, where the methyl group of L-Ala is in an axial position. In contrast, the methyl group is in an equatorial position for conformation 2. In addition to the explicit solvent modeling, the optimized structures with an implicit solvent model were also determined. We used the conformations 1 and 2 as initial structures, and geometry optimizations with the PCM solvent correction led to similar structures denoted models 1 and 2, respectively. Their dihedral angles ψ, ω, and ϕ are compared with those for conformations 1 and 2 in Table 1. An energetic consideration indicates that model 1 is more stable than model 2 by 0.62 kJ mol−1 (Table 1). As shown in Figure 3, the MD calculations

ψa

ωa

ϕa

ΔG

−21.3 15.0 −15.6 18.1 −14.0 18.4

−1.6 14.7 −8.7 9.7 −10.7 11.7

23.9 −36.2 26.6 −29.5 27.1 −32.6

0 0.62c 0 −0.62e

a Definitions in Figure 1. bCyclo(L-Ala-Gly) with the PCM solvent model. cThe free energy relative to that for model 1. dCyclo(L-AlaGly) in a vacuum. eThe free energy relative to that for model 3.

produced additional conformations with ψ > 0°, ϕ > 0° and ψ < 0°, ϕ < 0°. Although these structures were also used as initial structures, geometry optimizations resulted in a structure similar to that of model 2. This indicates that cyclo(L-Ala-Gly) adopts only two stable conformations, and model 1 is the most stable structure. Here we note the results of geometry optimizations without the PCM model, i.e., cyclo(L-Ala-Gly) in a vacuum. As listed in Table 1, two optimized structures, models 3 and 4, which are similar to models 1 and 2, respectively, are present in a vacuum. However, their relative energies are opposite, i.e., the boat-down structure (model 4) is energetically lower than the boat-up structure (model 3). Calculated Raman and ROA Spectra. The Raman and ROA spectra calculated on the basis of 250 hydrated clusters were averaged, and Figure 2 compares the simulated and observed spectra. The figure also includes the calculated spectra for a deuterated sample, in which exchangeable protons were deuterated (N1−D, N4−D). As shown in the comparison in Figure 2, the predicted Raman and ROA spectra are very similar to their corresponding experimental spectra. In addition, the observed H/D exchange effects on the spectra are well reproduced in the simulated spectra. The reasonable agreement between the experimental and calculated spectra allows us to assign the Raman and ROA bands as summarized in Table 2. The present band assignments are consistent with those made by previous studies on cyclic dipeptides.17,18,45 Figures S3 and 6770

dx.doi.org/10.1021/jp503874z | J. Phys. Chem. B 2014, 118, 6767−6774

The Journal of Physical Chemistry B

Article

Table 2. Observed and Calculated Vibrational Frequencies (cm−1) of Cyclo(L-Ala-Gly) νobsa

νcalb

assignmentc

1673 (−22) 1526 (−15) 1447 (−9) 1398 (−16) 1317 (+7) 1247 (0) 1175 (na) 1097 (−12) 1048 (na) 998 (0) 968 (+4) 890 (−30) 806 (−17) 766 (−5) 755 (−36) 611 (−15)

1678 (−15) 1497 (−20) 1431 (−7) 1361 (−8) 1286 (+10) 1218 (−4) 1141 (na) 1071 (−8) 1021 (na) 979 (−6) 942 (−3) 864 (−30) 779 (−14) 736 (−2) 713 (−20) 591 (−19)

νCO, δNH νCN, δNH CH2 (scissor), CH3 antisym deform. CH3 sym deform. CH wag CH2 twist (Gly) νCα−N, δCαH, δNH νCα−Cβ (Ala), δNH νCα−Cβ (Ala), CH2 rock (Gly) CH2 rock ring deform. (ip) ring deform. (ip) ring deform. (ip) ring deform. (ip) ring deform., γNH γNH

a

Observed vibrational frequencies. The numbers in parentheses are the N-deuteration shifts. bCalculated vibrational frequencies. cApproximate descriptions of the calculated normal modes. Abbreviations: ν, stretching; δ, in-plane bending; γ, out-of-plane bending. Figure 5. Observed and calculated Raman and ROA spectra of cyclo(L-Ala-Gly) in H2O solutions. (a, b) Experimental spectra. (c, d) Calculated spectra based on the MD + QM/MM method. The averaged spectra are shown in green, and the calculated spectra for 250 conformations are shown in black. The ROA spectra are magnified by a factor of 3000.

S4 in the Supporting Information present atomic displacements of selected normal modes for conformations 1 and 2, respectively. The Raman band at 1673 cm−1 is characteristic of the amide I band, which arises from the carbonyl CO stretching vibrations. This mode only shows a small negative feature in the ROA spectra (Figure 2, traces c and d). The sharp Raman band at 1526 cm−1 is due to the CN stretching mode. These high frequency modes involve the N−H bending motions, explaining their large D2O-induced shifts of ca. 20 cm−1. The most intense negative ROA band at 1447 cm−1 is assigned to a coupled CH2 bending (scissor) and CH3 deformation mode. The Raman band at 1398 cm−1 is mainly attributable to the symmetric CH3 deformation mode. The diminished Raman intensity of the N-deuterated derivative is well reproduced in the simulated spectra (Figure 2, traces e and f). There are several normal modes that are ascribed to CH2 and CH wagging vibrations near 1320 cm−1. The presence of several normal modes accounts for the negative/positive couplet in the ROA spectra (Figure 2, traces e−h). We assign the Raman band at 1097 cm−1 to the ring−methyl stretching with an N−H bending character. The N−H bending motion also contributes to the Raman and ROA bands at 1175 cm−1. The normal modes belonging to the ring deformation are expected below 1000 cm−1. The Raman and ROA bands at 890 cm−1 are mainly allocated to the N−C−Cα bending coordinate. The involvement of a motion for the NH moiety accounts for the large Ndeuteration shift of 30 cm−1. The sharp Raman and ROA signals at 762 cm−1 mainly originate from a ring-breathing vibration. In the low frequency region below ∼700 cm−1, a cyclic dipeptide molecule exhibits bands that can be assigned to out-of-plane bending vibrations. Effects of Spectral Averages. The effects of the spectral averaging on the Raman and ROA spectra are explored in Figure 5, where all the calculated spectra of 250 conformations as well as the averaged spectra are compared to the experimental spectra. Black lines in traces c and d represent the calculated spectra for the 250 conformations, whereas their

averaged spectra are shown in green lines. In the case of the Raman spectra, the spectral averaging is especially important for the low (1600 cm−1) frequency regions. For the amide I band around 1670 cm−1, each conformer exhibits an intense Raman band. However, because the amide I frequency is sensitive to hydrogen bonding interactions at the CO carbonyl moieties, the spectral averaging makes it a weaker broad Raman band. The broad spectral features below 800 cm−1 are also well reproduced by the averaged spectrum. Figure 5 demonstrates that the effect of the spectral average is more important for the ROA spectra than for the Raman spectra. Many of the ROA bands show different signs, when the conformations and environments of cyclo(L-Ala-Gly) are different. The amide I region is a clear example of the high structural sensitivity of the ROA spectra. As illustrated in trace d of Figure 5, the positive and negative bands are largely canceled to give a net small negative ROA band around 1678 cm−1. Similar effects of the spectral average are also observed in the other spectral regions. In order to examine the effect of the spectral average in more detail, Figure S5 in the Supporting Information compares the calculated spectra for conformations 1 and 2. The two conformations exhibit the most intense negative (conformation 1) and positive (conformation 2) ROA bands in the amide I region. We also show the calculated spectra of models 1 and 2 in Figure S6 in the Supporting Information. Figure S5 in the Supporting Information indicates that many of the ROA bands, such as methyl deformation (∼1420 cm−1), ring deformations (950, 860 cm−1), and NH out-of-plane bending vibrations (