The structure of tri-proline in water probed by polarized Raman, Fourier transform infrared, vibrational circular dichroism, and electric ultraviolet circular dichroism spectroscopy

The Structure of Tri-Prolinein Water Probed by PolarizedRaman, Fourier TransformInfrared, Vibrational CircularDichroism, and ElectricUltraviolet Circular DichroismSpectroscopy

Reinhard Schweitzer-Stenner1,2

Fatma Eker3

Alejandro Perez1

Kai Griebenow1

Xiaolin Cao4

Laurence A. Nafie4

1 Department of Chemistry,University of Puerto Rico,

Rıo Piedras Campus,San Juan,

Puerto Rico 00931

2 Department of Chemistry,Drexel University,

Philadelphia, PA 19104

3 Department of Biology,University of Puerto Rico,

Rıo Piedras Campus,San Juan,

Puerto Rico 00931

4 Department of Chemistry,Syracuse University,Syracuse, Syracuse,

NY 13244

Received 15 June 2003;accepted 18 July 2003

Abstract: Tripeptides serve as model systems for understanding the so-called random-coil state ofpeptides and proteins. While it is well known that polyproline or proline-rich polypeptides adopt thevery regular polyproline-II (PPII) or left-handed 31-helix conformation, it was thus far not clearwhether this is also the predominant structure adopted by proline-containing tripeptides. To clarifythis issue, we have investigated the amide I� band profile in the ir, isotropic, and anisotropic Raman,and vibrational circular dichroism (VCD) spectrum of cationic and zwitterionic tri-proline in D2O.The data were analyzed by modifying a recently developed algorithm, which allows one to obtainthe central dihedral angles of tripeptides from the amide I� band intensities (R. Schweitzer-Stenner,Biophysical Journal, 2002, Vol. 83, pp. 523–532). Our analysis revealed that the peptide adopts anearly canonical PPII structure in water with � and � values in the range of 175°–165° and�70°–(�80°), respectively. This is fully confirmed by the respective electronic ultraviolet-CD

Correspondence to: Reinhard Schweitzer-Stenner; email:[email protected]

Contract grant sponsor: Center for Research in Protein Struc-ture, Function and Dynamics (CRPSFD), NIH-SCORE, and Fon-dos Institucionales para la Investigacion of the University of PuertoRico (UPR)

Contract grant number: P20 RR16439-01 (NIH-COBRE II,CRPSFD), S06 GM008102-3052 (NIH-SCORE), and 20-02-2-78-514 (UPR)Biopolymers (Peptide Science), Vol. 71, 558–568 (2003)© 2003 Wiley Periodicals, Inc.

558

spectra. Our result indicates that the strong PPII propensity of trans proline results from localinteractions between the pyrrolidine ring and the backbone and is not due to any long-rangeinteractions. © 2003 Wiley Periodicals, Inc. Biopolymers (Pept Sci) 71: 558–568, 2003

Keywords: tripeptides; random-coil state; proteins; polyproline; proline-containing tripeptides;amide I�

INTRODUCTION

For a long period of time it has been believed that theunfolded state of proteins and peptides is structurallycompletely disordered, because the respective dihe-dral angles can, in principle, sample the entire allowedregion of the Ramachandran space. This view wastheoretically corroborated by the random coil modelintroduced by Brant and Flory,1 who argued that anunfolded polypeptide could be treated like a syntheticflexible polymer. During the last fifteen years, how-ever, this notion has been substantially modified inthat it was recognized that in many cases some resid-ual native and non-native structure persists.2 In thiscontext the polyproline II conformation has emergedas one of the structural motifs in the disordered stateof peptide fragments as well as of proteins.2,3 Thecanonical type II polyproline (PPII) is a left-handedhelix with an axial translation of 3.2 A composed ofthree prolyl residues per turn.4 Based on the similarityof electronic ultraviolet CD (ECD) spectra, Tiffanyand Krimm suggested 35 years ago that this confor-mation is not only adopted by polyproline but con-tributes also to the coil state of polylysine and poly-glutamic acid.5,6 This view was later confirmed byvibrational circular dichroism spectroscopy (VCD).7,8

PPII segments have been identified in the x-ray struc-ture of many proteins9 and CD and NMR data indicatethat it is adopted by a variety of cell receptor bindingpeptides.3,10 Based on Raman optical activity mea-surements Barron and co-workers have hypothesizedthat for some proteins PPII may constitute an inter-mediate state in the process of amyloid fibril forma-tion.11

While classical secondary structures like helices,�-sheets, and turns are stabilized by a combination oflocal, nonlocal and peptide–solvent interactions, it isgenerally thought that the PPII conformation of non-proline residues can exist only in water and reflectsthe local propensity of a given residue.12,13,14,15 Fromthis it follows that even short peptides should be ableto adopt the PPII conformation, in contrast to thecommon believe that their structure is completelyrandom.16 Indeed, a (temperature-dependent) mixtureof PPII and extended �-strand conformation havebeen obtained for the tri-alanine17,18,19 and variousalanine-based oligopeptides.20,21 Even the classicalalanine dipeptide seems to be predominantly PPII.13

On the contrary, tri-valine mostly adopts an extended�-sheet conformation.17,18 First measurements on dif-ferent AXA peptides indicate that most side chainsfluctuate between PPII and �-strand in the absence ofnonlocal interactions (Eker, Cao, Nafie, Schweitzer-Stenner, unpublished results). The respective molarfractions most likely reflect side chain–solvent andside chain–backbone interactions.15

It is surprising that even though PPII has becomean important issue for the structure analysis of evenshort peptides, the solution structure of tri-proline hasnot yet been determined. The above-cited resultsclearly lead to the prediction that it should adopt astructure close to the canonical PPII. Dukor and Kei-derling reported VCD spectra of a series of (P)n

peptides including P3 and interpreted their result asindicating a PPII structure.8 Helbecque and Lou-cheux–Lefebvre used electronic ECD to investigate aseries of GPn peptides and found that n�3 is theminimal number of proline residues required to obtaina PPII signal.22 None of these studies ruled out anadmixture of other conformations. To our best knowl-edge the dihedral angles of P3 have not yet beendetermined. In the present study we measured andanalyzed the amide I� band in the FTIR, polarizedRaman, and VCD spectra of tri-proline to fill this gap.To obtain the dihedral angle between the two peptidegroups from the amide I� band profiles in these spec-tra, we employed a modified and mathematicallymore consistent version of a recently developed algo-rithm.23 Our results were checked by measuring theECD signal in the far-uv region.

THEORETICAL BACKGROUND

The theory used to obtain the dihedral angles oftripeptides from the amide I� bands in their visibleRaman and ir spectra has been described in detailelsewhere.23 In what follows, this theory will bebriefly reviewed and slightly modified to extend itsapplicability. We assume a two-oscillator model todescribe the mixing between the two amide I� modesof tripeptides by transition dipole and through bondcoupling.24 The corresponding excitonic states arewritten as

The Structure of Tri-Proline in Water 559

���� � cos���1� � sin���2�(1)

���� � sin���1� � cos���2�

The parameter � describes the degree of mixing be-tween the unperturbed states ��1� and ��2�, which ismaximal for ��45°. This requires the unperturbedmodes to be accidentally degenerate. ���� and ����are the excitonic states of the in-phase (ip) and out-of-phase (oop) combination of the interacting modes.

The mixing parameter � can be determined fromthe intensity ratio Riso � Iiso

� / Iiso� of the two amide I�

bands in the spectrum of isotropic Raman scattering(Iiso

� and Iiso� are the isotropic intensities of ���� and

����) by utilizing the equation:

� �1

2arcsin�1 � Riso

1 � Riso� (2)

The corresponding ratio Raniso � Ianiso� / Ianiso

� dependson the mixing ratio and on the dihedral angles � and�, which determine the relative orientation of the twopeptide groups. In order to extract this informationfrom the experimentally determined Raniso value, onehas to calculate the Raman tensor of the excitonicstates ���� and ����:

�� � cos� � �1 � sin� � �2

(3)�� � sin� ��1 � sin� � �2

where �1 and �2 are the amide I� Raman tensors of thetwo peptide groups. In order to calculate �1 and �2

one has to transform one peptide tensor into the co-ordinate system of the other one. Thus, the Ramantensors of the excitonic states become dependent onthe relative orientation of the peptide groups. In ourearlier study this coordinate transformation wasachieved by rotating the coordinate system of �1 bythe tilt angle � between the peptide normals (i.e., thez axes in Figure 1) and subsequently by an angle �,which is the azimuthal angle between the now copla-nar peptide groups (i.e., between the carbonyl groupsof the peptides in a planar conformation, cf. Figure1).23 While this procedure is in principle correct, ithas the disadvantage that the real value for � dependson the choice of the coordinate system for the rotationaround �. Thus, it is conformationally dependent andcannot generally be identified with the angle betweenthe peptide groups in a coplanar configuration, asdone in Ref. 23. Moreover, this procedure makes theselection of the coordinate system for the Ramantensor dependent of the actual dihedral angles � and�. This is problematic because it obfuscates the ori-entation of the coordinate system of the Raman tensorwith respect to the principal axes of the Raman tensor

FIGURE 1 Reference structure of a tripeptide (��180°, ��0°, tri-glycine has been chosen forthe sake of simplicity) and representation of the coordinate system chosen for the transformation ofthe Raman tensor and the transition dipole moment. The structure was obtained by using theprogram TITAN from Wavefunction, Inc.

560 Schweitzer-Stenner et al.

(PART). Even though a check yielded a negligiblesmall error for the earlier investigated tripeptides,17 itis preferable to use a more rigorous formalism, whichcan be obtained by using the coordinate systems inFigure 1 (this coordinate system was also used in Ref.23, but not for the rotation of the Raman tensor, incontrast to what is stated in the article). The x axis ofthe coordinate system coincides with the NC� bondsof the two peptides and thus with the rotational axis ofthe dihedral angle �. The tensor �2 can be rotatedfrom S1(x1, y1, z1) into S2(x1, y1, z1) by the matrixoperation:

�2(S1) RT���RT����RT���

� �RT���)�2�S2�R�����R����R����R��(4)

which can be understood as follows. First, S1 has to berotated by an angle ����� . Subsequently, a rota-tion by x in the xy plane is necessary so that the ycoordinate coincides with the C�C bond, which is therotational axis for y. � is the angle formed by the y1

axis and the C�C bond. Next, the system is rotated bythe dihedral angle �. The fourth step involves therotation by an angle , which is formed by the C�Cbond and the y2 axis. This causes the abscissa tocoincide with the NC� bond. This rotation causes they axis to coincide with the NC� bond. Note that for acoplanar arrangement of the peptide groups as de-picted in Figure 1, and � can be obtained fromtextbooks on peptide structure with 96° and 20°, re-spectively).

The tensors calculated by means of Eqs. (3) and (4)can be used to calculate the isotropic and anisotropicscattering for the two excitonic states:

�s�2 � 1

9(Tr ��)2�aniso�

2 � 12

[(�xx,� � �yy,�)2

� (�yy,� � �zz,�)2 � (�zz,� � �xx,�)2]

�34

[(�xy,� � �yx,�)2�(�yz,� � �zy,�)2

�(�zx,� � �xz,�)2]

(5)

Eq. (5) can used to calculate Riso��s�2/�s�2 andRaniso��anisso�2/�aniso�2 as function of the mixingparameters � and the dihedral angles � and �.

In the next step we use the mixing parameter andthe intensity ratio RIR�IIR

� /IIR� in the FTIR spectrum to

obtain the angle � between the transition dipole mo-ments of the amide I� mode. The � can be calculatedas function of � and �. The related algorithm isdescribed in Ref. 23. Only those � and � values thatreproduce the experimentally obtained Raniso and RIR

are considered consistent with the experimental data.Thus, one generally obtains up to eight solutions. Inmost cases six of them can be ruled out because theyrepresent sterically forbidden conformations.

The physical, dihedral, and oriental parametersthus obtained were finally used to simulate the VCDsignal of amide I� as described in detail earlier.17

Thus, VCD serves a check of our analysis and is usedto discriminate between the different solutions ob-tained from the ir and Raman data. Generally, thisprocedure yields a single pair of values for � and �.

MATERIAL AND METHODS

Materials

L-prolyl-L-alanine and L-prolyl-L-prolyl-L-proline were pur-chased from Bachem Bioscience, Inc. (�98% purity), andused without further purification. NaClO4 and the quencherKI were obtained from Sigma-Aldrich Chemical Company(St. Louis, MO). All chemicals were of analytical grade.The peptides were dissolved in D2O at concentrations of0.15M (for ir and Raman), 0.075M (VCD) and 1 mM(ECD). The pD of the solutions were adjusted by addingsmall aliquots of DCl or NaOD, respectively, to obtain thecationic, zwitterionic, and anionic state of the peptides. ThepD values were determined by utilizing the method ofGlasoe and Long25 to correct the values obtained from pHelectrode measurements. For the Raman experiments thesolvent contained 0.25–0.1M NaClO4 whose 934 cm�1

Raman band was used as an internal standard.26 Addition-ally, we added 20 �L of KI M of in order to partially quenchthe fluorescence of the tri-proline samples.

Methods

Raman Spectroscopy. The 442 nm excitation (70 mW)was obtained from a HeCd laser (Model IK 4601R-E, Kim-mon Electric US). The polarized exciting laser beam wasfocused onto the sample with a lens of 100 mm focal length.The Raman scattered light was collected in a 135° back-scattering geometry. The scattered radiation was imagedonto the entrance slit (width adjusted to 100 �m) of atriple-grating spectrometer (T64000, Jobin Yvon, Inc.). Apolarization analyzer followed by a appropriately oriented�/2 plate between collimator and the entrance slit of thespectrometer were employed to measure the Raman inten-sity polarized parallel (Ix) and perpendicular (Iy) to thescattering plane. The scattering light was dispersed by thespectrometer and then detected by a liquid nitrogen cooledcharge-coupled device (CCD) with 256 1024 pixels in thechip. The spectral resolution was 4.0 cm�1. The frequencycalibration of the recorded Raman spectra was checked bymeans of the 934 cm�1 band of the internal standard, thefrequency of which had been determined earlier with highaccuracy.26 For tri-proline the very intense fluorescencebackground (even after partial quenching by 20 �L of KI)

The Structure of Tri-Proline in Water 561

deteriorated the signal to noise ratio of the Raman spectra.In order to improve the latter, we recorded multiple spectrafor both polarization directions and added them up to aver-age the noise. This yielded spectra of sufficient quality forthe zwitterionic peptide. For cationic PPP the Raman spec-tra were further smoothed by averaging a total of ten spec-tra, which were differently shifted along the wavenumberaxis in an interval of �5 cm�1. The thus caused deteriora-tion of spectral resolution could be tolerated in view of themuch larger bandwidth of the individual amide I� bands.

IR Spectroscopy. FTIR spectra were measured with aNicolet Magna-IR System 560 optical bench as describedelsewhere.27 A total of 256 scans at 2 cm�1 resolution usingHapp–Ganzel apodization were averaged to obtain eachspectrum. For all experiments, a Spectra Tech liquid cellequipped with CaF2 windows and 6-�m thick mylar spacerswere used. The peptide sample was put between CaF2

windows. Each peptide sample was measured at least fourtimes. Spectra were corrected for the solvent background inan interactive manner using Nicolet OMNIC 3.1 software.

VCD Spectroscopy. VCD spectra were measured with aChiralir FT-VCD spectrometer from Bomem/BioTools,modified to the setup of dual polarization modulation(DPM). The DPM setup includes two ZeSe photoelasticmodulators (PEM), one of which is to create left and rightcircularly polarized radiation and the other is used to sup-press the linear birefringence and the associated VCD arti-facts. The spectrometer is equipped with a liquid nitrogencooled HgCdTe detector having a cutoff at 800 cm�1. VCDspectra were measured in D2O with resolution of 8cm�1using a CaF2 cell with a pathlength of 30 �. The VCDspectra were collected in blocks for a total collection time ofapproximately 9 h depending on the peptide sample inves-tigated. The VCD spectra of solvent D2O were also mea-sured at the identical conditions for the purpose of VCDbaseline correction and ir spectral solvent subtraction. Forall measurements, the PEM was optimized for maximumquarter-wave response at 1400 cm�1, which is around themidpoint of spectral range of interest. Other experimentalconditions are provided in the figures captions referring tothe VCD spectra.

CD Spectroscopy. Far-uv CD spectra (250–190 nm)were measured with an OLIS DSM-10 UV/Vis CD spec-trophotometer in a 1.0-mm quartz cell with 2 nm resolution.The samples were placed in a nitrogen gas purged OLIS CDmodule. The temperature at the cuvette was controlled bymeans of a Peltier-type heating system (accuracy �1°C).For each measurement, the sample in the cuvette was al-lowed to equilibrate for 5 min at the adjusted temperatureprior to data acquisition. For all experiments reported in thisarticle �A(�,T) was measured by increasing the temperaturein increments of 5°C. The integration time was set as afunction of high volts to obtain an appropriate signal-to-noise ratio. The room temperature spectra were obtained byaveraging 5 scans. The solvent reference spectra were used

as baselines, which were automatically subtracted from thepeptide CD spectra. For the final presentation in this articlethe original �A(�,T) spectra were converted to the ��(�,T)representation by using the above sample concentrationsand the path length of the cuvette.

Spectral Analysis. All ir and Raman spectra were ana-lyzed using the program MULTIFIT.28 They were normal-ized to the internal standard, i.e., the ClO4

� band at 934cm�1. To eliminate solvent contributions, we measured thesolvent reference spectra for both polarizations, which werethen subtracted from the corresponding peptide spectra. Theintensities of the normalized polarized Raman bands werederived from their band areas. These and the correspondingir spectrum were self-consistently analyzed in that theywere fitted with a set of identical frequencies, halfwidths,and band profiles. The isotropic and anisotropic Ramanintensities and the depolarization ratios were calculated asfollows:

Iiso � Ix �4

3Iy

(6)Iansio � Iy

� �Ix

Iy

It should be mentioned that in principle Ianiso should bewritten as 8Iy/3. As mentioned in earlier articles,17 we preferto identify it with Iy in the depicted figures so that thepolarization properties of different lines can be better in-ferred.

Density Functional Theorem Calculations. We usedthe TITAN program from Wavefunction, Inc., to perform astructural optimization of L-alanyl-L-proline in vacuo withthe Becke–Lee–Young–Parr composite exchange functional(B3LYP) at the B3LYP/6-31g level of theory. All harmonicfrequencies of the optimized geometry were found to bereal, showing that it corresponds to a stable energy mini-mum.

RESULTS AND DISCUSSION

Amide I� of Prolylalanine

As reported in detail below, the Raman spectra oftri-proline (PPP) indicate that the Raman tensor ofamide I� depends on the protonation state of theterminal groups, in contrast to what has been foundfor tri-alanine. This requires a modification of themathematical algorithm reported in Ref. 23, which isbased on the assumption that the amide I� Ramantensors of the two peptide groups are identical. Thus,

562 Schweitzer-Stenner et al.

in order to fix as many parameters as possible for ouranalysis, we measured and analyzed the amide I�region of the FTIR as well as of the isotropic andanisotropic Raman spectra of the dipeptide L-prolyl-L-alanine (PA) in D2O at acid, neutral, and alkalinepD (Figure 2). In order to allow for comparison, theRaman spectra were normalized by means of the 934cm�1 line of sodium perchlorate, which thus served asan internal standard. All ir spectra are displayed asextinction coefficient spectra. The amide I� bandscould be satisfactorily fitted by single Voigtian bands

with the spectral parameters listed in Table I. Smalldeviations between the fit and experimental profileindicate a slight band asymmetry due to fluctuationsof the hydrogen bonds between the peptide and thewater molecules in the solvation shell.29 From the fitswe also obtained the integrated band intensities anddepolarization ratios listed in Table I. The depolariza-tion ratios of the three protonation states are signifi-cantly different, i.e., 0.17, 0.11, and 0.06 for thecationic, zwitterionic, and anionic state of the dipep-tide, respectively. This reflects nearly exclusively dif-ferences between the anisotropic contribution to theRaman cross section. In contrast, we obtained practi-cally identical ir extinctions for the three protonationstates. The obtained variations are in the limit ofaccuracy. However, a comparison with the ir spectraof di- and tri-alanine revealed a substantial increase ofthe amide I� oscillator strength by a factor of 3.4. Thisreflects a 1.8-fold increase of the transition dipolemoment. Besides its relevance for our analysis, thisresult is also important for the ir spectroscopy onpeptides and proteins, for which it is generally as-sumed that the amide I� oscillator strength is indepen-dent on the side chain composition.30

In order to understand the depolarization ratio westart with the coordinate system visualized in Figure1. Its x axis coincides with the NC� bond and its y axislies in the peptide plane to form an angle of 117° withthe carbonyl bond. Thus, z becomes the out-of-planecoordinate. The peptide nitrogen is chosen as the zeropoint of the coordinate system. In our earlier studieswe assumed that the amide I� intensity reflects vi-bronic coupling to electronic transitions in the peptideplane (mostly 3 1

* (NV1) and 3 2*(NV2), Fig-ure 3), so that its Raman tensor can be written as23:

FIGURE 2 Amide I� region of the FTIR, isotropic, andanisotropic Raman spectra of L-prolyl-L-alanine in D2Omeasured at pD 1 (cationic, black), 6 (zwitternionic, red),and 12 (anionic, blue). The Raman spectra were recordedwith 457.9 nm excitation (laser power: 150 mW, slit width:100 �m).

Table I Spectral Parameters and Raman TensorElements of the Amide I� Mode of the ThreeProtonation States of L-Prolyl-L-Alanine

Parameter Cationic Zwitterionic Anionic

�A1 (cm�1)a 1669 1667 1631�L (cm�1)b 10.0 10.0 10.0�G (cm�1)c 27 21 29.4

� 0.14 0.11 0.07ad 0.65 0.65 0.65bd 1 1 1cd �0.07 �0.07 �0.07dd 0 0.07 0.25

aWavenumber position.bLorentzian halfwidth of the Voigtian profile.cGaussian halfwidth of the Voigtian profile.dRelative tensor elements of the amide I mode.

The Structure of Tri-Proline in Water 563

� � � a c 0c b 00 0 0

� (7)

The depolarization ratio varies between 0.75 (a�-b)and 0.125 (a�b, c�0). The values generally obtainedfor amide I� with excitation in the visible region are inthe range between 0.14 and 0.17, indicating that a andb have the same sign and similar values. In order toobtain depolarization values smaller than 0.125 a zzcomponent has to contribute to the Raman tensor inthe PART frame, so that

� � � a c 0c b 00 0 d

� (8)

Now the depolarization ratio can reach the value zero,if a�b�d. We disregard contributions from xz, zx, yz,and zy, since they would all increase rather thandecrease the depolarization ratio. One possible candi-date for this is the thus far neglected n(COO�)4 (peptide) charge transfer transition to which amide I�is indeed weakly coupled, since this transition in-volves expansions of the CO as well as of the CNbond.31,32 As shown in Ref. 30 for di-glycine thedipole transition moment points from the carboxylate

carbon to the peptide nitrogen (Figure 3). Its orienta-tion with respect to our coordinate system and thePART frame depends on the dihedral angle �. For aPPII conformation (�� 60°–80°) it would have asubstantial z component, so that vibronic coupling ofamide I� to this transition would add a zz element tothe Raman tensor. However, the amide I� Ramantensor obtained for di-glycine crystals by Pajcini etal.32 suggest that this contribution is very weak with alimited influence on the depolarization ratio. There-fore, we are led to disregard the above charge transfertransition as a possible explanation for the low depo-larization values. It must result from vibronic cou-pling to a yet unidentified electronic transition in thefar uv, the dipole moment of which has a substantialz component. It might be associated with the pyrroli-dine ring. A DFT calculation on neutral PA in vacuo(Figure 3) revealed an optimized geometry in which itis nearly perpendicular to the peptide plane. The cor-responding � angle (165°) is close to the canonicalPPII value. We compared the calculated and experi-mental wavenumbers of the high frequency peptidemodes. As expected, the wavenumber of the amide Iwas substantially overestimated, owing to the neglectof hydrogen bonding to solvent molecules. Since thedepolarization ratio of the cationic peptide (��0.14)is in the range known from tri-alanine, tri-glycine, andother tripeptides, we assume that the respective Ra-

FIGURE 3 Structure of neutral L-prolyl-L-alanine in vacuo as obtained from the DFT calculationdescribed in the Material and Methods section. The figure also contains the direction of electronictransition dipole moments obtained from Ref. 30. It should be noted that the CT transition occursonly if the C-terminal group is deprotonated.

564 Schweitzer-Stenner et al.

man tensor of AA has no zz element and can thereforebe described by Eq. (7). Consistency with the di-glycine PART tensor obtained by Pajcini et al.32 it hasbeen shown to require that23

c �b � a

9.3(9)

The depolarization ratio can be calculated by combin-ing Eqs. (8) and (9) with Eqs. (12), (14), and (15) inRef. 23. Since it depends on the ratios a/b and c/brather than on the absolute values of the tensor ele-ments, we have to choose an arbitrary value for one of

the tensor elements. In accordance with the findingsof Pajcini et al.32 we identified the y axis as the majoraxis of the Raman tensor in the chosen coordinatesystem and assumed b�1 as a reference value. Thus,� can be reproduced by a�0.65 and c��0.04. In thenext step, we attributed the somewhat lower depolar-ization ratio of 0.11 observed for the zwitterionic stateto an additional zz element d and thus obtainedd�0.07. To account for the even lower depolarizationratio of the anionic peptide, d�0.25 has to be inserted.In the following, we will use the thus derived Ramantensors (listed in Table I) to analyze the amide I� bandprofile of cationic and zwitterionic PPP.

FIGURE 4 Amide I� region of the FTIR, isotropic Raman, anisotropic Raman, and VCD spectra ofL-prolyl-L-prolyl-L-proline in D2O measured at pD 1 (cationic) and 7.5 (zwitterionic). The Raman spectrawere recorded with 441 nm excitation (laser power: 65 mW, slit width: 100 �m). The line profiles in their and Raman spectra result from a global spectral decomposition as described in the Material andMethods section. The solid lines in the VCD spectra result from a calculation based on the dihedralangles obtained from the analysis of the amide band intensities as described in the text.

The Structure of Tri-Proline in Water 565

Structure Analysis of Tri-Proline

Figure 4 depicts the isotropic Raman, ir, and VCDspectra of the amide I� region of cationic and zwitte-rionic tri-proline. In a first step, we analyzed thespectra of the zwitterionic peptide, because we ob-tained the spectra with the best signal to noise ratio forthis species. Fluorescence and a lower Raman crosssection made the recording of good spectra difficult.The ir and Raman spectra were then analyzed asdescribed under Material and Methods. The obtainedintensity ratios Riso, Raniso, and RIR are listed in TableII. To simulate the Riso and Raniso as a function of theexcitonic mixing parameter �, the orientational angle�, and the dihedral angles � and �, we started byusing the Raman tensor obtained for anionic andzwitterionic PA for the N and C-terminal amide I�,respectively. This did not allow us to reproduce theexperimental Raniso and the depolarization ratios(���0.06, ���0.17) because the theoretical orienta-tional dependencies of these parameters were not pro-nounced enough. Therefore, we allowed for slightchanges of the d values and eventually obtained asatisfactory solution for dN�0.15 and dC�0.1,namely the dihedral angles ��165° and ��70°. Fig-ure 5 shows RIR, Raniso, �� and �� as a function of �for ��165°. In the limit of accuracy, all experimentalvalues are reproduced in the interval ���70°�10°.Hence, we obtained nearly the canonical PPII struc-ture from our analysis. We checked for other solutionsand found them all in sterically forbidden regions.

The corresponding amide I� VCD signal (Figure 4)was reproduced with the above structural parameters

by considering that the PA ir spectra indicate a muchlarger dipole transition moment. It should be notedthat the negative peak at 1590 cm�1 results from theCOO� antisymmetric stretch. It has been heuristicallymodeled by a Gaussian profile. We rule out excitoniccoupling between this mode and the amide I� of theC-terminal based on the observation that the band ofthe corresponding COO�1 is totally depolarized in theRaman spectrum. Excitonic coupling would give riseto some isotropic intensity.

The original Raman spectra of the amide I� regionof cationic PPP were very noisy due to a strong

Table II Spectral Parameters of and GeometricValues Derived from the Amide I� Mode of the ThreeProtonation States of L-Prolyl-L-Prolyl-L-Proline

Parameter Cationic Zwitterionic

�A1� (cm�1)a 1628 1611�A1� (cm�1)a 1646 1630�L� (cm�1)b 10.0 10.0�L� (cm�1)b 10.0 10.0�G� (cm�1)c 16.9 20.0�G� (cm�1)c 17.5 23.0�� 0.28 0.11�� 0.08 0.06Riso 0.37 0.47Raniso 1.89 0.96RIR 0.92 1.17� (°) 85 92� (°) �80 �60� (°) 175 150

aWavenumber position.bLorentzian halfwidth of the Voigtian profile.cGaussian halfwidth of the Voigtian profile.

FIGURE 5 The � Raniso, ��, and �� as function of thedihedral angle � for ���165° by using the algorithmdescribed in the theory section and the Raman tensorsderived in the Results and Discussion section. The horizon-tal lines depict the respective experimental values. The solidvertical line labels the � values for which all experimentalvalues could be reproduced in the limit of their accuracy.

566 Schweitzer-Stenner et al.

fluorescence background and a weak relative amide I�intensity. While the spectral averaging described inMaterial and Methods yielded at least an isotropicspectrum of sufficient quality for the spectral analysis,the eventually obtained anisotropic scattering spec-trum must be considered with great caution since itcannot be ruled out that the smoothing procedure hasaffected the overall band profile. The spectrum dis-played in Figure 4 is indicative of a much larger Raniso

value than that obtained for the zwitterionic species.We believe that this does at least qualitatively reflectthe real spectral distribution. It is also theoreticallyexpected owing to the relatively large depolarizationratio observed for cationic PA. We subjected theRaman and ir spectra to a self-consistent analysis,which yielded identical bandwidths but slightly dif-ferent wavenumber positions for the respective amideI� bands in the ir and Raman spectra. This discrepancymost likely results from the smoothing procedure. Theobtained spectral parameters were analyzed as de-scribed above to yield � � 175°�5°–15° and���80°�20°. The relatively large errors reflect theuncertainties of the spectral analysis. This coordinatescan still be regarded as reflecting a PPII structure. Inthe limit of accuracy, one can argue that the centraldihedral angles of the zwitterionic and cationic spe-cies are identical. We checked the coordinates of thecationic species by calculating the VCD spectrum. Inorder to account for the asymmetry of the amide I�couplet (Figure 4), we have to assume a magneticdipole moment for the C-terminal amide I�, in accor-dance with what was obtained for other tripep-tides.14,17 Our calculation yielded a nearly perfectreproduction of the experimental signal.

As a final control of our analysis, we also measuredthe electronic uv-CD spectra of cationic and zwitteri-onic PPP. The spectra are shown in Figure 6. Forcationic PPP, we obtained a clear and intense PPIIsignal, which is practically independent on tempera-ture. The variations on the high energy side of thenegative signal at 195 nm most likely stem from atemperature dependent overlapping band at lowerwavenumbers. The spectra of zwitterionic PPP are inprincipal similar to that of the cationic species, but areindicative of a weak temperature dependence of thePPII signal. However, that could well result fromcontributions assignable to the above mentionedcharge transfer transition.31,32 We can therefore con-clude that the CD spectra fully confirm the notion thateven PPP is locked into a single conformation, whichshows a nearly canonical PPII structure. The absenceof any significant temperature dependence of the PPIICD signal argues against the presence of additionalconformers even at high temperatures. Thus, our re-sults are fully in line with the behavior of longerpolyproline peptides investigated by Dukor and Kei-derling.8

SUMMARY AND CONCLUSIONS

We have measured and analyzed the amide I� bandregion in the FTIR and Raman spectra of tri-proline inwater. By utilizing a modified version of a recentlydeveloped algorithm we were able to determine thedihedral angles between the peptide groups. The thusobtained “secondary” structure of the cationic andzwitterionic peptide is close to the canonical polypro-

FIGURE 6 UV-CD spectra of L-prolyl-L-prolyl-L-proline at pD 1 and 7 measured at 20°C (blue),40°C (green), and 60°C (red).

The Structure of Tri-Proline in Water 567

line II conformation. Any admixtures from other con-formations cannot be inferred from our data. Thisdemonstrates that the preference for this structure isdue to local interactions between the pyrrolidine ringand the backbone and not due to any long-rangeinteractions. From this it follows that proline residuesshould also predominantly (nearly 100%) adopt PPIIin proline rich peptides. The PPII contents of a seriesof AcP3XP3GY–NH2 peptides (X represents a set ofdifferent residues) reported by Kelly et al. corroboratethis notion.33 For X�P obtained a PPII content of66%, which one would indeed expect, if one assumeslow PPII propensity for G and Y, some reducedstability for the C-terminal P and a nearly 100% PPIIoccupation for the remaining prolines. In generalterms, our study adds another piece of evidence to thenotion that tripeptides are ideal tools to understand theso-called random coil state of peptides and proteins.

First of all, we have to thank Dr. Timothy Keiderling fromthe University of Illinois, Chicago, for strongly suggestingto apply our method to tri-proline. Financial support wasprovided from the NIH-COBRE II grant for the Center forResearch in Protein Structure, Function and Dynamics (P20RR16439-01), from the NIH-SCORE grant (S06 GM008102-3052) and from the Fondos Institucionales para la Investiga-cion of the University of Puerto Rico (20-02-2-78-514).

REFERENCES

1. Brant, D. A.; Flory, P. J. J Am Chem Soc 1965, 87,2791–2800.

2. Bochicchio, B.; Tamburro, A. M. Chirality 2002, 14,782–792.

3. Shi, Z.; Woody, R. W.; Kallenbach, N. R. Adv ProteinChem 2002, 62, 163–240.

4. Cowan, P. M.; McGavin, S. Nature 1955, 176, 501–503.5. Tiffany, M. L.; Krimm, S. Biopolymers 1968, 6, 1379–

1382.6. Tiffany, M. L.; Krimm, S. Biopolymers 1968, 1767–1770.7. Paterlini, M. G. Freedman, T. B.; Nafie, L. A. Biopoly-

mers 1986, 25, 1751–1765.8. Dukor, R.; Keiderling, T. Biopolymers 1991, 31, 1747–1761.9. Seerama, N.; Woody, R. W. Protein Sci 2003, 12,

384–388.10. Siligardi, G.; Drake, A. F. Biopolymers 1995, 37, 281–

292.

11. Blanch, E. W.; Morozowa-Roche, L. A.; Cochran,D. A. E.; Doig, A. J.; Hecht, L.; Barron, L. D. J MolBiol 2000, 301, 553–563.

12. Han, W. G.; Jalkanen, J.; Elstner, M.; Suhai, S. J PhysChem B 1998, 102, 2587–2602.

13. Weise, C. F.; Weisshaar, J. C. J Phys Chem B 2003,107, 3265–3277.

14. Eker, F.; Cao, X.; Nafie, L.; Huang, Q.; Schweitzer-Stenner, R. J Phys Chem B 2003, 107, 358–365.

15. Pappu, R. V.; Rose, G. D. Protein Sci 2002, 11, 2437–2455.

16. Dill, K. A. Protein Sci 1999, 8, 1166–1180.17. Eker, F.; Cao, X.; Nafie, L.; Schweitzer-Stenner, R.

J Am Chem Soc 2002, 124, 14330–14341.18. Eker, F. Griebenow, K.; Schweitzer-Stenner, R. J Am

Chem Soc 2003, 125, 8178–8185.19. Mu, Y.; Kosov, D. S.; Stock, G. J Phys Chem B 2003,

107, 5064–5073.20. Park, S.-H.; Shalongo, W. Stellwagen, E. Protein Sci

1997, 6, 1694–1700.21. Shi, Z.; Olson, C. A.; Rose, G. D.; Baldwin, R. L.;

Kallenbach, N. R. Proc Natl Acad Sci USA 2002, 99,9190–9195.

22. Helbecque, N.; Loucheux-Lefebvre, M. H. Int J PeptideProtein Res 1987, 19, 94–101.

23. Schweitzer-Stenner, R. Biophys J 2002, 83, 523–532.24. Torii, H.; Tasumi, M. J. Raman Spectrosc 1998, 29,

81–86.25. Glasoe, P. K.; Long, F. A. J Phys Chem 1960, 64,

188–193.26. Sieler, G.; Schweitzer-Stenner, R. J Am Chem Soc

1997, 119, 1720–1726.27. Griebenow, K.; Diaz Laureano Y.; Santos A. M.; Mon-

tanez Clemente I.; Rodriguez L.; Vidal M.; Barletta G.J Am Chem Soc 1999, 121, 8157–8163.

28. Jentzen, W.; Unger, E.; Karvounis, G.; Shelnutt, J. A.;Dreybrodt, W.; Schweitzer-Stenner, R. J Phys Chem1996, 100, 14184–14191.

29. Gnanakaran, S.; Hochstrasser, R. M. J Am Chem Soc2001, 123, 12886–12898.

30. Al-Azzam, W.; Pastrana, E.A.; Ferrer, Y.; Huang, Q.;Schweitzer-Stenner, R.; Griebenow, K. Biophys J 2002,83, 3637–3651.

31. Chen, X. G.; Li, P.; Holtz, J. S. W.; Chi, Z.; Pajcini, V.;Asher, S. A.; Kelly, S. A. J Am Chem Soc 1996, 118,9705–9715.

32. Pajcini, V.; Chen, X. G.; Bormett, W.; Geib, S. J.Asher, S. A. J Am Chem Soc 1996, 118, 9716–9726.

33. Kelly, M. A.; Chellgreen, B. W.; Rucker, A. L.; Trout-man, J. M.; Fried, M. G.; Miller, A.-F.; Creamer, T. P.Biochemistry 2001, 40, 14376–14383.

568 Schweitzer-Stenner et al.