Hydrogen Bonding and Solvent Polarity Markers in the UV Resonance Raman Spectrum of Tryptophan: Application to Membrane Proteins

Abstract

Ultraviolet resonance Raman (UVRR) spectra of tryptophan compounds in various solvents and a model peptide are presented and reveal systematic changes that reflect solvent polarity, hydrogen bond strength, and cation–π interaction. The commonly utilized UVRR spectral marker for environment polarity that has been based on off-resonance Raman data, the tryptophan Fermi doublet ratio I_~1360/I_~1340, exhibits different values in on- and off-resonance Raman spectra as well as for different tryptophan derivatives. Specifically, the UVRR Fermi doublet ratio for indole ranges from 0.3 in polar solvents to 0.8 in nonpolar solvents, whereas the respective values reported here and previously for off-resonance Raman spectra are 0.5–1.3. UVRR Fermi doublet ratios for the more biologically relevant molecule, N-acetyl tryptophan ethyl ester (NATEE), are in a smaller range of 1.1 (polar solvent) to 1.7 (nonpolar solvent) and correlate to the solvent polarity/polarization parameters π* and $E_{\mathrm{T}}^{\mathrm{N}}$. As has been reported previously, several UVRR modes are also sensitive to the hydrogen bond strength of the indole N–H moiety. Here, we report a new unambiguous marker for H-bonding: the ratio of the W10 (~1237 cm⁻¹) intensity to that of the W9 (~1254 cm⁻¹) mode (R_W10). This ratio is 0.7 for NATEE in the absence of hydrogen bond acceptors and increases to 3.1 in the presence of strong hydrogen bond acceptors, with a value of 2.3 in water. The W8 and W17 modes shift more than +10 and approximately −5 cm⁻¹ upon increase in hydrogen bond strength; this range for W17 is smaller than that reported previously and reflects a more realistic range for proteins and peptides in solution. Finally, our data provide evidence for change in the W18 and W16 relative intensity in the presence of cation–π interactions. These UVRR markers are utilized to interpret spectra of model membrane-bound systems tryptophan octyl ester and the peptide toxin melittin. These spectra reveal the importance of intra- and intermolecular hydrogen bonding and cation–π interactions that likely influence the partitioning of membrane-associated biomolecules to lipid bilayers or self-associated soluble oligomers. The UVRR analysis presented here modifies and augments prior reports and provides an unambiguous set of spectral makers that can be applied to elucidate the molecular microenvironment and structure of a wide range of complex systems, including anchoring tryptophan residues in membrane proteins and peptides.

Introduction

Tryptophan is the least abundant residue in soluble proteins, accounting for only 1.1% of the amino acids expressed in cytoplasmic proteins,¹ but is more prevalent in membrane proteins, with an abundance of 2.9% in transmembrane α-helical domains.² This aromatic residue typically plays key functional roles in proteins because of its unique properties among the 20 natural amino acids: tryptophan exhibits the largest accessible nonpolar surface area that is highly polarizable, possesses an indole N–H moiety that is capable of hydrogen bond donation, and displays the greatest electrostatic potential for cation–π interactions.^3,4 These important physical properties render tryptophan an ideal amphiphilic residue with the greatest propensity to reside in the interfacial region of a membrane protein as compared to any other naturally occurring amino acid.⁵

Tryptophan has been found to stabilize membrane spanning proteins and peptides by acting as anchors along the interface of the bilayer.^6,7 For example, replacement of tryptophan residues with phenylalanine in the 325-residue integral membrane protein, outer membrane protein A, destabilizes the protein relative to wild-type when folded into lipid bilayers.^8,9 Tryptophan residues in membrane-associated antimicrobial peptides also play important functional roles in hemolytic and bactericidal activity.^10,11 In the antibiotic channel peptide gramicidin A, substitution of tryptophan for phenylalanine residues results in reduction of antibacterial activity.^12,13 These and other examples illustrate that the presence, location, and environment of tryptophan residues are critical in the study of folding and insertion of membrane proteins and membrane-associated peptides.

Tryptophan displays environment-sensitive photophysical properties. Tryptophan fluorescence has been widely utilized to reveal its local environment in proteins, accessibility to solvent, and location with respect to another biological (e.g., heme) or exogenous (fluorescence acceptor) moiety. UV resonance Raman (UVRR) spectroscopy is a vibrational technique that selectively enhances signal from absorbing chromophores and has also been applied to the study of tryptophan and proteins. One of the biggest advantages of UVRR over fluorescence and other electronic spectroscopy techniques is that UVRR inherently reports on environment and structure, including hydrogen bonding states and orientation with respect to backbone, with great selectivity and sensitivity.^14–16 UVRR may also be applied to probe different protein moieties, including other residues, such as tyrosine and proline, as well as the amide backbone to obtain a more comprehensive picture of the biomolecule.

The commercial availability of tunable deep-UV lasers has resulted in a vast number of UVRR studies of tryptophan residues in proteins over the past two decades.^17–20 However, interpretation of these data has generally relied on data of l-tryptophan and model tryptophan derivatives that were acquired under off-resonance conditions.^21–23 Detailed UVRR studies on tryptophan and its model compounds have focused primarily on topics such as UVRR excitation profiles,^24–26 excited-state relaxation rates and saturation spectra,^27,28 and photoinduced transient radical species.²⁹ Here, we present a systematic analysis of UVRR spectra of tryptophan derivatives to illustrate spectral markers for hydrogen bonding, environment polarity, and cation–π interactions of this unique amphiphilic residue under resonance conditions; these results are compared with previous off-resonance Raman data. In addition to this comparison, we also report new spectral markers that can be utilized to interpret UVRR spectra.

The significance of the current work is the expansion of UVRR to membrane-associated proteins. Because of the enrichment of aromatic amino acids in membrane proteins and the large changes in environment and structure that are expected upon protein insertion into a bilayer, UVRR is an ideal tool for studies of membrane protein folding. Data and analyses presented here help differentiate tryptophan residues that are solvent-exposed, bound to lipid headgroups, or buried in the hydrophobic core of a lipid membrane with or without a hydrogen bonding partner. Results for the widely studied membrane toxin melittin as well as a membrane-associated molecule, tryptophan octyl ester, are presented to illustrate the successful application of UVRR spectroscopy in studying membrane-associated biomolecules.

Materials and Methods

Chemicals

N-Acetyl tryptophan ethyl ester (NATEE) was purchased from TCI America, tryptophan octyl ester (TOE) was obtained from Chem Impex International, 1,3-dimethyl-3,4,5,6-tetrahydro-2(1H)-pyrimidinone (DMPU) was purchased from Sigma Aldrich, and deuterated water (>99%) was obtained from Spectra Stable Isotopes. All other tryptophan derivatives, solvents, and potassium phosphate salts were purchased from Fisher Scientific. Melittin was purchased from Axxora (San Diego, CA), and neutral lipid 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) and anionic lipid 1-palmitoyl-2-oleoyl-sn-glycero-3-[phospho-rac(1-glycerol)] (sodium salt, POPG) were obtained from Avanti Polar Lipids as chloroform solutions. Compounds were used as received without further purification. If necessary, solvents were dried with molecular sieves. Tryptophan derivative concentrations were 10–50 mM for all Raman experiments. TOE and melittin UVRR samples contained potassium phosphate buffer at pH 7.3. TOE concentrations were 40 µM in phosphate buffer and 15 µM in POPC vesicles. Melittin concentrations were 40 µM in each sample.

Vesicle Preparation

To make 2:1 molar ratio POPC/POPG vesicles, a chloroform solution containing 5 mg of POPG was added to a chloroform solution containing 10 mg of POPC. The resulting solution of anionic lipid and a solution of 10 mg neutral POPC lipid were dried under a stream of argon. Dried lipids were resuspended in 20 mM potassium phosphate buffer using a bath sonicator. The vesicles were formed by extruding the lipid suspension eleven times through a polycarbonate filter with pore size 200 nm (100 nm for TOE experiments) using a liposome extruder. This vesicle solution was filtered through a 0.45 µm filter and passed through a gravity-driven desalting column (BioRad). The first 3 mL elution was discarded, and the second 3 mL elution containing vesicle was collected and allowed to equilibrate for 2 h at 37 °C. This aliquot was confirmed to contain a majority (>70%) of ~200 nm vesicles via dynamic light-scattering measurements. The final lipid concentration used in the experiments was 1 mg/mL.

Steady-State Fluorescence Spectroscopy

Tryptophan fluorescence spectra were obtained on a Jobin Yvon Horiba Fluorolog-3 spectrofluorometer. The excitation wavelength was 280 nm for tryptophan derivatives and 290 nm for melittin and TOE, and the entrance and exit bandpass was 3 nm. Fluorescence spectra were recorded at a constant temperature of 20 °C.

Raman Spectroscopy

The off-resonance Raman spectra were acquired with 514 nm excitation from a mixed-gas Kr–Ar ion laser. The 50 mW beam was focused onto a 1.8 mm o.d., 1.4 mm i.d. capillary tube and collected using a 90° scattering geometry as described previously.³⁰ Scattered light was focused onto a 100 µm entrance slit, and Rayleigh light was rejected using a 514 nm long-pass edge filter (Semrock RazorEdge). Raman scattered light was dispersed in an f/6.9, 0.75 m spectrograph equipped with a 1200 grooves/mm grating and imaged onto a Peltier-cooled CCD. The spectral bandpass was 10 cm⁻¹, and accuracy as determined by ethanol calibration was ± 1 cm⁻¹. The precision based on repeatability was ± 1 cm⁻¹.

UV Resonance Raman Spectroscopy

The UVRR setup has been described elsewhere.³¹ Briefly, vibrational spectra of melittin and tryptophan derivatives were obtained by setting the fundamental laser wavelength to 920 nm to generate a 230 nm excitation beam. For tryptophan derivatives, a typical sample volume of 4 mL was flowed through a 200 µm i.d., 350 µm o.d., vertically mounted, fused-silica capillary at a rate of 0.40 mL/min to ensure fresh sample for each laser pulse. The capillary size for melittin and TOE experiments was 100 µm i.d., 160 µm o.d., and the flow rate was 0.16 mL/min. The UV power was ~4 mW at the sample. Ten 1-min spectra were collected and summed for all samples with the exception of TOE, which required 30–60 min of acquisition time. UVRR spectra of all appropriate blank solutions were also collected and subtracted from the corresponding tryptophan derivative, TOE, and melittin spectra. Accuracy was determined using standard ethanol peaks and was found to be ± 2 cm⁻¹. The bandpass for the Raman experiment was ~8 cm⁻¹. Overlapping peaks were decomposed into Gaussian bands using a least-squares fitting routine. The precision based on repeatability was ± 2 cm⁻¹.

Correlation Analysis

The correlation coefficient, r, is defined as³²

$$r=\frac{{\displaystyle \sum (x_i-\overline{x})(y_i-\overline{y})}}{\sqrt{{\displaystyle \sum {(x_i-\overline{x})}^2{(y_i-\overline{y})}^2}}}$$

where the x_i are the values of the solvent parameter of interest (e.g., π*), x̄ is the average of the x_i, the y_i are the values of the spectroscopic parameter of interest (e.g., R_FD), and ȳ is the average of the y_i. The correlation coefficient, r, is a value between −1 and 1; values close to −1 indicate negative correlation, and values close to +1 indicate positive correlation. Values near 0 indicate no correlation. Probability is defined as the likelihood that the same number of measurements of two uncorrelated variables of x and y would produce a correlation coefficient with |r_uncorrelated| ≥ |r|.³²

Results

Raman Spectra of Model Compounds with 230 and 514 nm Excitation Wavelengths

Raman spectra of 10–50 mM l-tryptophan (l-Trp), N-acetyl l-tryptophan ethyl ester (NATEE), N-acetyl l-tryptophanamide (NATA), 3-methylindole (skatole), and indole were acquired with incident wavelengths 230 and 514 nm. These derivatives, illustrated in the Supporting Information, were chosen because of the abundance of prior experimental data (l-Trp and NATA), solubility in organic solvents (NATEE, skatole, and indole), and extensive vibrational mode analysis (skatole and indole). NATEE, l-Trp, and NATA serve as models of the biologically relevant amino acid, whereas skatole and indole exhibit the greatest solubility in membranelike organic solvents. Concentration-dependence experiments revealed no changes in peak positions or relative intensities over the relevant range of concentrations in methanol. Raman spectra of each derivative in methanol are shown in Figure 1; a spectrum of l-Trp is shown in the Supporting Information and is consistent with prior reports. NATEE and NATA exhibit nearly identical spectra in terms of peak positions and relative intensities; skatole also displays similar bands, but indole spectra deviate substantially from these other tryptophan derivative spectra. Comparison of the on- and off-resonance Raman spectra for a given tryptophan derivative reveals that peak positions are similar, with less than 4 cm⁻¹ difference for corresponding peaks. However, as expected, there are variations in relative band intensities between on- and off-resonance Raman spectra.³³

Figure 1.

UVRR (230 nm excitation) and off resonance Raman spectra (514 nm excitation) of NATEE (A), NATA (B), skatole (C), and indole (D) in methanol. Tryptophan derivative UVRR spectra were normalized to the W18 band intensity; off-resonance data were arbitrarily scaled. All spectra are offset for clarity. Strong methanol peaks obscured signal in the regions of the off-resonance data marked by an asterisk.

The Fermi doublet intensity ratios also vary with excitation wavelength and tryptophan derivative. The Fermi doublet ratio is defined as R_FD = I_{~1360 cm⁻¹}/I_{~1340 cm⁻¹} and was determined by directly obtaining intensity values from the spectra as well as by Gaussian decomposition of the overlapping bands. The R_FD values from direct intensities are summarized in Table 1 for NATEE and indole on- and off-resonance and in different solvents; skatole R_FD values are not resolved and therefore not presented here. Solvent properties^34,35 of polarity/polarizability (π*), normalized polarity $(E_{\mathrm{T}}^{\mathrm{N}})$, and dielectric constant (ε_r) as well as the wavelengths of maximum fluorescence emission for indole are also listed. Fermi doublet values in off-resonance spectra of NATEE and indole are similar to those reported previously.²¹

TABLE 1.

Solvent Polarity/Polarization Scales π* and $E_{\mathrm{T}}^{\mathrm{N}}$, Dielectric Constants ε_r,³⁵ and Spectroscopic Parameters R_FD and Emission Maxima (λ^fluo) of NATEE and Indole^a

solvent	π*	$E_{\mathrm{T}}^{\mathrm{N}}$	εr	UV RFDb	UV RFDc	Vis RFDb	λfluo (nm)c
water	1.09	1.00	78	1.11	0.29	0.97	352
DMPU	1.08	0.35	36	1.42	0.67	n.a.	n.a.
DMSO	1.00	0.44	46	1.28	0.46	1.21	334
formamide	0.97	0.78	110	1.33	0.39	n.a.	340
DMF	0.88	0.39	37	1.49	0.42	n.a.	329
cyclohexanone	0.68	0.28	16	1.68	0.62	n.a.	n.a.
acetonitrile	0.66	0.46	36	1.47	0.36	1.03	326
methanol	0.60	0.76	33	1.56	0.46	1.33	335
benzene	0.55	0.11	2.3	1.41	0.78	1.21	311
ethanol	0.54	0.65	25	1.52	0.62	n.a.	335
toluene	0.49	0.10	2.4	n.a.	0.78	n.a.	323
dioxane	0.49	0.16	2.2	1.50	0.46	0.98	322
cyclohexane	0.00	0.01	2.0	n.a.	0.81	n.a.	300
hexanes	−0.11	0.01	1.9	n.a.	0.75	n.a.	300

^aR_FD is defined as the ratio of intensities of the Fermi doublet at ~1360 to ~1340 cm⁻¹ (I_~1360/I_~1340).

^bValues are for NATEE.

^cValues are for indole.

DMPU, 1,3-dimethyl-3,4,5,6-tetrahydro-2(1H)-pyrimidinone; DMSO, dimethyl sulfoxide; DMF, dimethyl formamide; UV, ultraviolet (230 nm excitation); Vis, visible (514 nm excitation); n.a., not acquired or not available due to solvent interference.

Graphs of UVRR R_FD as a function of the solvent properties π* and $E_{\mathrm{T}}^{\mathrm{N}}$ along with representative data are shown in Figure 2. Linear fits to the data are also included. Decomposition of the Fermi doublet region resulted in slightly different R_FD values; with the exception of dimethyl formamide (DMF) and dimethyl sulfoxide (DMSO), the R_FD values obtained via decomposition of the UVRR spectra for NATEE in various solvents were within 10% of the values obtained when direct intensities were utilized. The peak positions in the NATEE UVRR Fermi doublet shifted in some solvents; the low-energy peak varied from 1340 to 1346 cm⁻¹, and the high-energy peak varied from 1363 to 1365 cm⁻¹.

Figure 2.

UVRR Fermi doublet R_FD values and linear regression lines for indole (●) and NATEE (○) in nonhalogenated solvents as a function of solvent Kamlet–Taft π* and polarity $E_{\mathrm{T}}^{\mathrm{N}}$ scale. Representative spectra and Gaussian decomposition are shown as insets.

Correlation coefficients between spectroscopic observables (UVRR R_FD and emission λ_max) and solvent parameters (π*, $E_{\mathrm{T}}^{\mathrm{N}}$, and hydrogen bond acceptor value β) are summarized in Table 2. The correlations between UVRR R_FD and the solvent parameters π* and $E_{\mathrm{T}}^{\mathrm{N}}$ are significant, with three of four of the probability values less than 2.1%. The correlation between R_FD and $E_{\mathrm{T}}^{\mathrm{N}}$ in the case of NATEE is not as strong (probability 12.9%). In contrast, analysis of R_FD and β show that these parameters are unlikely to be correlated, with probabilities of 28.0% or 45.7%. In addition, the fact that the r values are of the opposite sign for NATEE and indole supports a lack of correlation between R_FD and β. Analogous analysis of the fluorescence maximum λ_max and solvent parameters indicate that λ_max is strongly correlated to all three parameters: π*, $E_{\mathrm{T}}^{\mathrm{N}}$, and β. Graphs of R_FD vs β as well as of λ_max values are included as Supporting Information.

TABLE 2.

Correlation Coefficients (r) That Indicate Strength of Correlation between Spectral Markers and Solvent Parameters^a

spectral marker	solvent parameter	correlation coefficient (r)	probability (%)
indole RFD	π*	−0.65	1.1
	$E_{\mathrm{T}}^{\mathrm{N}}$	−0.77	<0.5
	β	−0.31	28.0
NATEE RFD	π*	−0.68	2.1
	$E_{\mathrm{T}}^{\mathrm{N}}$	−0.49	12.9
	β	0.25	45.7
indole λmax	π*	0.89	<0.05
	$E_{\mathrm{T}}^{\mathrm{N}}$	0.92	<0.05
	β	0.61	3.9

^aProbability is defined as the likelihood that the same number of measurements of two uncorrelated variables of x and y would produce a correlation coefficient with |r_uncorrelated| ≥ |r|. See main text for further details.

UVRR Spectra of Tryptophan and Derivatives in H₂O and D₂O

UVRR spectra of NATEE, skatole, and indole in H₂O and D₂O are shown in Figure 3. These spectra are similar to previously published spectra^15,36–38 and illustrate peaks that involve large displacement of the N–H moiety on the indole ring. Significant changes are observed for the W17, W8, and W4 peaks, with downshifts of 19, 19, and 10 cm⁻¹, respectively, for skatole; the W17 and W4 peaks have been shown to be sensitive markers for hydrogen-bonding strength.^22,23 Other significant spectral changes are observed in the W12 and W10 regions upon deuteration. Bands that are minimally perturbed upon deuteration include the intense W3 and W16 peaks. In D₂O, the UVRR Fermi doublet shifts to lower frequency and is less resolved than in H₂O. This behavior of the Fermi doublet in the UVRR spectra is consistent with previous on- and off-resonance spectra.^15,21

Figure 3.

UVRR spectra of NATEE (A), skatole (B), and indole (C) in H₂O and D₂O. Spectra were normalized to the W18 band intensity and offset for clarity. Main NATEE peaks are indicated.

UVRR Spectra of Skatole with Solvent Titrations

The effects of hydrogen bonding of the indole N–H moiety on the resonance Raman spectrum was probed with skatole. Because skatole does not possess the backbone amide group, these solvent titrations reflect localized changes of the indole ring. Figure 4 shows UVRR spectra of skatole in mixtures of cyclohexane and dioxane to illustrate spectral changes associated with hydrogen bonding in a nonpolar environment; both solvents are nonpolar and aprotic but exhibit different hydrogen bond acceptor basicities based on the Kamlet-Taft β scale (cyclohexane β = 0, dioxane β = 0.37).³⁹ This variation in hydrogen bond capability is illustrated in FTIR spectra that show skatole N–H peaks at 3500 and 3342 cm⁻¹ in cyclohexane and dioxane, respectively (data not shown). Significant shifts in UVRR relative intensity and peak positions are observed upon enhancement of hydrogen bond strength between skatole N–H and solvent. The W17 and W14 peaks grow more intense upon addition of dioxane, and their positions shift −2 and 0 cm⁻¹, respectively. Growth of the W17 peak is quantified relative to the W18 and W16 peaks. These ratios, referred to as R_W17:W18 = I_W17/I_W18 and R_W17:W16 = I_W17/I_W16, are indicated in Table 3.

Figure 4.

UVRR spectra of 50 mM skatole in cyclohexane/dioxane mixtures: 100% cyclohexane (A), 25 mM dioxane (B), 50 mM dioxane (C), 100 mM dioxane (D), 500 mM dioxane (E), 5 M dioxane (F), and 100% dioxane (G). Spectra were normalized to the W18 band intensity and offset for clarity.

TABLE 3.

UVRR Intensity Ratios, W17 Frequencies, and Fluorescence Maxima, λ^fluo, of Skatole in Cyclohexane with Increasing Amounts of Dioxane^a

dioxane (mM)	RW10	W17 (cm−1)	RW17:W18	RW17:W16	λfluo (nm)
0	0.5	884	0.15	0.24	307
25	0.5	884	0.17	0.26	316
50	0.5	884	0.16	0.26	319
100	0.6	884	0.18	0.28	320
500	0.8	882	0.20	0.32	327
5000	1.1	882	0.25	0.45	331
pure	1.2	882	0.26	0.50	335

^aR_W10 is defined as the ratio of intensities at ~1237 (W10) to ~1254 (W9) cm⁻¹ (I_~1237/I_~1254); R_W17:W18 is defined as the ratio of intensities at ~882 (W17) to ~763 (W18) cm⁻¹ (I_W17/I_W18); R_W17:W16 is defined as the ratio of intensities at ~882 (W17) to ~1014 (W16) cm⁻¹ (I_W17/I_W16).

In contrast, the W1 mode decreases in intensity relative to W18 upon addition of dioxane (Supporting Information). The W12 and W10 regions show changes in position and relative intensity; the sharp W12 peak at 1154 cm⁻¹ in cyclohexane shifts toward 1173 cm⁻¹ in dioxane and simultaneously loses intensity. The W10 region transforms from an intense peak (1247 cm⁻¹) with a shoulder (1231 cm⁻¹) in cyclohexane to a clearly resolved doublet (1254 and 1237 cm⁻¹) with nearly equal intensities in dioxane. This doublet consists of the W9 (~1254 cm⁻¹) and W10 (~1237 cm⁻¹) peaks, and the relative intensity of this W10 doublet, R_W10 = I_W10/I_W9 = I_{~1237 cm⁻¹}/I_{~1254 cm⁻¹}, shifts from 0.5 in 100% cyclohexane to 1.2 in 100% dioxane (Table 3). A similar trend was observed for NATEE in chloroform (β = 0.1) and dioxane with R_W10 values of ~0.7 and 3.1, respectively. Other spectral shifts include a decrease in intensity of an out-of-plane (OOP) peak at 854 cm⁻¹ and an 8 cm⁻¹ upshift in W8 position to 1307 cm⁻¹ in dioxane. Skatole fluorescence maxima in this mixture are also included in Table 3.

UVRR spectra of skatole in a mixture of the polar, aprotic solvents acetonitrile and DMPU were also acquired to probe spectral modifications associated with hydrogen bonding in a polar environment (Figure 5). As with cyclohexane and dioxane, these solvents have different hydrogen-bond-accepting capability, with acetonitrile being a weaker hydrogen bond acceptor (β = 0.31) than DMPU (β = 1.05).^40,41 Because DMPU was found to be an excellent substitute for HMPA (hexamethylphosphoramide) ⁴¹ the β value of HMPA was used here for DMPU. The strong hydrogen bond between skatole N–H and DMPU was confirmed with FTIR measurements that showed that the skatole N–H stretch was lower in DMPU (3228 cm⁻¹) than in acetonitrile (3396 cm⁻¹) (data not shown). Despite the difference in hydrogen bond strength in these two solvents, no significant changes in peak position or intensity were observed for the W17 or W4 peaks. However, complex changes were observed in the W12 region, with a broad peak centered near 1177 cm⁻¹ in acetonitrile shifting to lower energy, ~1150 cm⁻¹, in DMPU. A systematic increase in the relative intensity of the skatole W10 doublet, R_W10, from 0.9 to 1.9 also accompanied the enhancement of the hydrogen bonding in DMPU. A similar increase in R_W10 of approximately a factor of 2 was observed for NATEE in acetonitrile (R_W10 = 1.5) and DMPU (R_W10 = 2.8), suggesting that these systematic changes are not limited to skatole. These and other results are summarized in Table 4, along with fluorescence maxima of the mixtures. Fermi doublet ratios were not calculated for either cyclohexane/dioxane or acetonitrile/DMPU titrations because the doublet bands are unresolved in the skatole UVRR spectra.

Figure 5.

UVRR spectra of 50 mM skatole in acetonitrile/DMPU mixtures: 100% acetonitrile (A), 25 mM DMPU (B), 50 mM DMPU (C), 100 mM DMPU (D), 500 mM DMPU (E), 5 M DMPU (F), and 100% DMPU (G). Spectra were normalized to the W18 band intensity and offset for clarity.

TABLE 4.

UVRR Intensity Ratios, W17 Frequencies, and Fluorescence Maxima, λ^fluo, of Skatole in Acetonitrile with Increasing Amounts of DMPU^a

DMPU (mM)	RW10	W17 (cm−1)	RW17:W18	RW17:W16	λfluo (nm)
0	0.9	879	0.26	0.53	346
25	0.9	879	0.29	0.53	347
50	1.0	879	0.29	0.53	348
100	1.0	879	0.24	0.53	348
500	1.1	879	0.25	0.53	351
5000	1.8	879	0.29	0.56	351
pure	1.9	879	0.30	0.56	352

^aR_W10 is defined as the ratio of intensities at ~1237 (W10) to ~1254 (W9) cm⁻¹ (I_~1237/I_~1254); R_W17:W18 is defined as the ratio of intensities at ~882 (W17) to ~763 (W18) cm⁻¹ (I_W17/I_W18); R_W17:W16 is defined as the ratio of intensities at ~882 (W17) to ~1014 (W16) cm⁻¹ (I_W17/I_W16).

UVRR Spectra of TOE and Melittin

We now compare these tryptophan derivative data to the UVRR spectra of a membrane-associated compound TOE and the well-studied peptide melittin. TOE is a simple model for interfacial tryptophan, ⁴² and melittin is a toxic peptide that contains a single native tryptophan residue (Trp-19). At low concentration (<100 µM), melittin assumes a monomeric random coil structure when in water or in a low ionic strength buffer system.⁴³ When melittin concentration or ionic strength is increased, the peptide folds into a tetrameric α-helical structure that results in a buried tryptophan residue in a hydrophobic core.^43,44 Figure 6 presents UVRR spectra of TOE in phosphate buffer and in the presence of neutral POPC vesicles and melittin in phosphate buffer, in the presence of neutral POPC and anionic POPC/POPG vesicles, and in 2 M NaCl.

Figure 6.

UVRR spectra of TOE in phosphate buffer (A) and POPC lipid vesicles (B) and melittin in phosphate buffer (C), 2:1 POPC/POPG lipid vesicles (D), POPC lipid vesicles (E), and 2 M NaCl (tetramer) (F) are shown. Spectra were normalized to the W18 band intensity and offset for clarity. Top insets are the W18 and W16 regions of the TOE spectrum in phosphate buffer (solid thin line) and POPC lipid vesicles (solid thick line). Same spectral regions for melittin in various environments (bottom inset), including phosphate buffer (solid thin line), 2 M NaCl (dotted line), 2:1 POPC/POPG lipid vesicles (dashed line), and POPC lipid vesicles (solid thick line). Insets were normalized to the W18 band intensity.

In phosphate buffer, TOE exhibits a fluorescence maximum of 359 nm with R_FD and R_W10 values of 1.1 and 2.4, respectively. TOE bound to POPC vesicles exhibits a fluorescence emission maximum of 343 nm with R_FD and R_W10 values of 1.4 and 2.6, respectively. The W17 peak position varies from 879 to 882 cm⁻¹ for TOE and melittin in various environments. The W3 band (~1550 cm⁻¹) has been found to be sensitive to the torsion angle χ^2,1 of the tryptophan side chain.⁴⁵ The W3 band is peaked at 1548 cm⁻¹ and is broad for TOE in vesicle, with a fwhm of ~20 cm⁻¹; this peak transforms to a sharp peak centered at 1555 cm⁻¹ and a shoulder (~1545 cm⁻¹) for TOE in buffer. The torsion angle corresponding to the sharp peak is 105°.

The Fermi doublet R_FD value is 1.7 for melittin in 2 M NaCl, 1.3 (POPC) and 1.7 (POPC/POPG) in the presence of vesicles, and 1.2 in phosphate buffer. The R_W17:W18 values are 0.29 for melittin in 2 M NaCl, 0.31 for melittin in POPC vesicles, 0.32 in POPC/POPG vesicles, and 0.21 for melittin phosphate buffer. The R_W10 values for melittin in 2 M NaCl, inserted in POPC and POPC/POPG vesicles, and in phosphate buffer are 0.8, 1.3, 1.2, and 1.5, respectively. TOE and melittin intensity ratios, W17 frequencies, and fluorescence maxima are summarized in Table 5. The W3 band frequency in the UVRR spectra of melittin ranges from 1552 to 1554 cm⁻¹, corresponding to torsion angles from 98 to 103°.

TABLE 5.

UVRR Intensity Ratios, W17 Frequency Values, and Fluorescence Maxima, λ^fluo, of TOE and Melittin^a

sample	RFD	W17 (cm−1)	RW10	λfluo (nm)
TOE in buffer	1.1	881	2.4	359
POPC	1.4	879	2.6	343
melittin in buffer	1.2	880	1.5	355
POPG/POPC	1.7	882	1.2	336
POPC	1.3	882	1.3	343
2 M NaCl	1.7	880	0.8	339

^aR_FD is defined as the ratio of intensities of the Fermi doublet at ~1360 to ~1340 cm⁻¹ (I_~1360/I_~1340); R_W10 is defined as the ratio of intensities at ~1237 (W10) to ~1254 (W9) cm⁻¹ (I_~1237/I_~1254).

Normalized and expanded regions of the W18 and W16 bands of TOE and melittin are shown in Figure 6. The W16 peak of TOE in POPC and melittin in 2 M NaCl decrease in intensity relative to the W18 peak. This change is similar to that reported previously in a protein with cation–π interaction⁴⁶ and of a model cation–π system of skatole in a solution with tetramethylammonium salt (data not shown).

Discussion

UVRR has developed into a powerful and sensitive technique to probe the local environment, structure, and dynamics of aromatic amino acids and the peptide backbone and is preferable to off-resonance Raman spectroscopy because of selective, >10⁷-fold signal enhancement. UVRR has become particularly useful when studying micromolar concentrations of aromatic amino acids and amide groups found in proteins and peptides.^14–16 Until now, the majority of UVRR spectra of tryptophan in proteins and peptides have been interpreted based on off-resonance Raman spectra of tryptophan model compounds. As expected, peak positions remain constant in on- and off-resonance Raman spectra of tryptophan derivatives, whereas relative peak intensities vary with excitation wavelength because of the resonance enhancement factor. Comparison of the Raman spectra of the various tryptophan derivatives utilized in past studies also show significant differences for a number of key modes and thereby suggest that peaks in appropriate model compounds must be utilized to interpret spectra of proteins (see Figure 1). Lack of detailed, systematic studies on the effects of hydrogen bonding and environment on pertinent UVRR spectra motivates the current study; results presented here may be generally applied to other UVRR studies of complex systems and reactions.

On- and Off-Resonance Raman Spectra: Fermi Doublet and Solvent Polarity

The W7 Fermi doublet is one of the most widely utilized UVRR observable that reports on the local environment of tryptophan residues in proteins. With off-resonance 488 or 514 nm excitation, the Fermi doublet R_FD values of indole are similar (data not shown). However, these values differ considerably from UVRR indole R_FD values and also differ from the UVRR R_FD values for the more physiologically relevant molecule NATEE. The difference between on-and off-resonance R_FD values is attributed to the change in relative intensities of the out-of-plane combination band and W7 fundamental mode that constitute the Fermi doublet when the excitation wavelength coincides with an absorption band.²¹ In off-resonance spectra, R_FD values for indole ranged from 0.54 to 1.32; in UVRR spectra of indole, R_FD values were considerably lower and ranged from 0.29 to 0.81. UVRR R_FD values for NATEE were within a smaller range of 1.1–1.7. These differences indicate that previous quantitative applications of indole off-resonance R_FD values to interpret protein UVRR data were not appropriate, and analysis of the spectra of the appropriate molecule and excitation condition is required. A recent UVRR study did not attempt to interpret observed R_FD values because of the lack of relevant systematic studies, further confirming the need for the current report.⁴⁷

The UVRR R_FD values increase with hydrophobicity of solvent, and this trend is qualitatively consistent with previous off-resonance data. Figure 2 and Table 2 show the correlations of indole and NATEE UVRR R_FD in nonhalogenated solvents with two common solvent parameters: the Kamlet–Taft π* scale for solvent polarity/polarizability and the normalized solvent polarity parameter $E_{\mathrm{T}}^{\mathrm{N}}$ which has excellent correlation to the earlier Kosower Z values.^34,35 Data in Table 2 suggest in the case of NATEE, π* may have a more significant correlation to R_FD than $E_{\mathrm{T}}^{\mathrm{N}};$ however, a more thorough analysis is needed. In contrast to π* and $E_{\mathrm{T}}^{\mathrm{N}}$, there was minimal correlation of R_FD with the Kamlet–Taft hydrogen bond acceptor basicity β, suggesting that R_FD is a sensitive reporter of the polarity of the environment, as opposed to hydrogen-bonding capability. The emission maxima of indole in the various solvents showed significant correlation to all three parameters, π*, $E_{\mathrm{T}}^{\mathrm{N}}$, and β, with the strongest correlations to π* and $E_{\mathrm{T}}^{\mathrm{N}}$. These correlations indicate that an important advantage of UVRR over fluorescence is the ability to independently report on environment polarity and hydrogen-bonding structure, whereas shifts in fluorescence maxima may reflect a combination of both effects.

We propose the following general trends to be applied toward interpretation of UVRR spectra: R_FD for solvent-exposed tryptophan is the lowest value measured at ≤ 1.1; R_FD values for hydrophilic, polar environments, such as those of DMSO and formamide, are in the range of 1.2–1.4; R_FD values for less polar environments similar to acetonitrile and DMF are 1.4–1.5; and R_FD values for least polar environments, similar to ethanol and dioxane, are >1.5. These findings are summarized in Table 6.

TABLE 6.

Range of Values Observed for UVRR Markers of Hydrogen Bonding and Polarity and Corresponding Values in Aqueous Solution (H₂O)^a

	properties
UVRR markerb	no H-bond	H-bond	H2O
RW10	0.7	3.1	2.3
W17 (cm−1)	884	879	879
W8 (cm−1)	1299	1309	1307
W4 (cm−1)	1492	1496	1494
	Nonpolar	Polar	H2O
RFD	1.7	1.1	1.1

^aR_W10 is defined as the ratio of intensities at ~1237 (W10) to ~1254 (W9) cm⁻¹ (I_~1237/I_~1254); R_FD is defined as the ratio of intensities of the Fermi doublet at ~1360 to ~1340 cm⁻¹ (I_~1360/I_~1340).

^bR_W10 and R_FD values are for NATEE in various solvents and buffer (H₂O), W17 values are for indole in various solvents and buffer, and W8 and W4 values are for skatole in solvent mixtures and buffer.

Hydrogen Bonding in Polar and Nonpolar Environments

Tryptophan derivative UVRR peaks that are sensitive to H/D exchange are identified and shown in Figure 3. Peak assignments for tryptophan derivatives in D₂O were determined by comparison to previous studies and vibrational mode analysis.^15,48 Consistent with previous reports, the W17 band undergoes a large downshift in D₂O. The frequencies of the W12, W10, W8, W4, and W1 bands shift for each derivative in D₂O, indicating that their normal modes involve displacement of the indole N–H group. The positions of these bands, with the exception of the W1 peak, are also found to be sensitive to hydrogen bonding in the current study. The W7 Fermi doublet transforms from a distinct doublet in H₂O to a primarily single peak in D₂O, consistent with other reports.^15,38

Solvent titrations of skatole in Figure 4 and Figure 5 illustrate key spectral changes that occur in polar and nonpolar environments upon enhancement of hydrogen bond strength. The choice of skatole reveals vibrational changes due to hydrogen bonding of the indole N–H moiety, as opposed to changes in hydrogen bonding of backbone atoms. The cyclohexane/dioxane titration models the interior of the lipid bilayer (ε_r ~ 2),⁴⁹ whereas the acetonitrile/DMPU titration models the polar headgroup region of the lipid bilayer (ε_r ~ 35).⁵⁰ These skatole titrations can therefore be applied to reveal the hydrogen bonding condition of tryptophan residues that interact with different regions of membranes.

A previous FTIR/off-resonance Raman study utilizing crystalline tryptophan derivatives and dissolved skatole reported the sensitivity of the W17 band frequency (871–883 cm⁻¹) to hydrogen bond strength.²² Our results are consistent with these prior data of dissolved skatole but suggest a more realistic range of W17 frequencies from ~879 to 884 cm⁻¹ for strong and weak hydrogen bonds in solution, respectively. The low frequencies reported in the previous study are likely applicable only to strongly hydrogen bonded crystalline species. In contrast to this relatively small range for W17, we have found that W8 is a sensitive indicator of hydrogen bond strength. The W8 mode consists of a C₉–N–C₂ symmetric stretch combined with an N–H bend; this peak upshifts 10 cm⁻¹ for skatole, from 1299 to 1309 cm⁻¹, upon transfer from cyclohexane to DMPU. A smaller 4 cm⁻¹ upshift, from ~1492 to 1496 cm⁻¹, was observed for the W4 band. This finding suggests that W8 and W4 are reliable indicators of hydrogen bonding and may be utilized in conjunction with the W17 mode to probe the hydrogen bonding environment. The range of W17, W8, and W4 values are summarized in Table 6.

The W17 band intensity also successfully reports on hydrogen bonding, particularly in nonpolar environments. The W17, W18, and W16 modes are resonant with the B_b absorption (~220 nm).^24–26 Solvatochromic shifts occur in this absorption band and are reflected in UVRR spectra by changes in relative intensity.⁵¹ However, additional enhancement of the W17 mode is observed in the presence of a strong Lewis base. This additional effect is quantified relative to other B_b-coupled modes W18 and W16 for skatole in the form of R_W17:W18 and R_W16:W18 in Table 3 and Table 4. In the nonpolar solvent cyclohexane, R_W17:W18 and R_W17:W16 for skatole increase systematically by factors of 1.7 and 2.1, respectively, upon addition of the hydrogen bond acceptor dioxane. This result is consistent with a prior qualitative report.⁵² Similar, but more subtle, changes were observed in the polar solvent acetonitrile; addition of the strong hydrogen bond acceptor DMPU enhanced R_W17:W18 and R_W16:W18 by factors of 1.2 or 1.1, respectively. The subtle nature of this trend in the polar solvent is likely due to the fact that acetonitrile is a hydrogen bond acceptor (β value of 0.31), albeit weaker than DMPU (β value of 1.05).^39,40 NATEE exhibits similar ratios, with a maximum R_W17:W18 value of 0.29 in both DMPU and dioxane, and maximum R_W17:W16 values of 0.35 and 0.39 in DMPU and dioxane, respectively. These values for NATEE in a non-hydrogen bonding environment, such as chloroform, are 0.19. These data suggest that large intensity ratios R_W17:W18 and R_W17:W16 are sensitive indicators of enhanced hydrogen bonding in nonpolar environments in the absence of possible cation–π interactions (see below).

The ratio of the W9–W10 doublet found at ~1250 cm⁻¹ in both titrations, R_W10, is an unambiguous indicator of N–H hydrogen bonding. The values found for skatole (Table 3 and Table 4) differ from those obtained for NATEE. For example, NATEE R_W10 values range from 0.7 to 3.1 in chloroform and dioxane, respectively, with a value of 2.3 in water (Table 6). The W10 band is also resonant with the B_b electronic transition,²⁴ and the relative intensity in the UVRR spectrum therefore likely correlates to shifts observed in the absorption spectrum. In contrast to the significant change in relative intensities, the positions of the W9 and W10 bands shift a small amount of ~7 cm⁻¹. This subtle shift is attributed to the N–H bend component in the W9 and W10 modes.

UVRR Markers for Cation–π Interactions

In contrast to the numerous UVRR studies that report on the hydrogen bonding, structure, and environment of amino acids and peptide backbone, only a few UVRR experiments have aimed to reveal cation–π interactions involving tryptophan residues, despite the relatively common occurrence of this interaction. To the best of our knowledge, there are only four such reports.^46,53–55 Alterations in UVRR intensities and peak positions in these previous publications focus on the W18, W16, and W7 modes. Modifications of delocalized modes, such as W18 (indole ring-breathing mode) and W16 (benzene ring-breathing mode) are expected, given the strength of cation–π interactions of up to a few kilocalories/mole stabilization.^3,14,56 The W18 (762 cm⁻¹) and W16 (1014 cm⁻¹) peak positions remain constant within error in the solvent titrations and in the various solvents. The intensity of the W16 peak relative to W18, R_π = I_W16/I_W18, is 0.77 ± 0.05 for NATEE, 0.56 ± 0.05 for indole, and 0.56 ± 0.07 for skatole, where the error describes the range of R_π values for the compounds in a variety of solvents. These values for R_π indicate that the ratio of I_W16 to I_W18 changes less than 7% for the biologically relevant tryptophan compound, NATEE, in response to solvent environment. The observations that W16 and W18 are relatively insensitive to hydrogen bonding and environmental modifications suggest that these modes may be reliable markers for cation–π interactions.

Results in Figure 6 (discussed below) support prior reports that cation–π interactions may cause a decrease in W16 intensity.⁴⁶ In experiments when internal standards may not be utilized to obtain absolute intensities because of interactions with the protein or undesired alterations in ionic strength, we propose that R_π may be a reliable marker for cation–π interactions. However, it is clear that the excitation wavelength is a critical factor for quantitative analysis: although 229 nm excitation results in a decrease, excitation with 238 and 244 nm causes the W16 intensity to increase.⁴⁶ In addition, the intensities of W16 and W18 may vary slightly in different environments, and the W16 peak will be affected by the presence of a large number of phenylalanine residues in the protein. Therefore, only the relative change of R_π may be utilized as a marker, where the reference value of R_π in the absence of cation–π interaction is defined as 100% and determined under solvent-exposed conditions where water may prevent these interactions.

UVRR Spectra of Membrane-Associating Compounds

The observed trends are applied to two membrane-associating systems, TOE and melittin, to gain molecular insight into tryptophan interactions of these systems. On the basis of fluorescence quenching measurements,⁵⁷ in TOE, the tryptophan moiety is known to lie at the bilayer-water interface at a distance of ~11 Å from the hydrocarbon core. Figure 6 and Table 5 highlight the spectral changes associated with TOE binding to POPC vesicles. In phosphate buffer, the UVRR indicators R_FD and R_W10 suggest that TOE is in a polar environment and strongly hydrogen bonded, presumably to solvent. In the presence of POPC vesicles, TOE is localized to the less polar environment of the bilayer interface (R_FD = 1.4) but remains hydrogen bonded to the lipid head groups or water (R_W10 = 2.6). This decrease in environment polarity is consistent with the fluorescence blue shift from 359 nm (TOE in buffer) to 343 nm (TOE in vesicles); however, UVRR provides evidence for strong hydrogen bonding of the tryptophan moiety in TOE bound to vesicles that is not discernible in fluorescence spectra. Additional insight into molecular interactions is obtained from R_π. This ratio decreases 14% upon binding to POPC vesicles, providing support for cation–π interaction between indole and perhaps the lipid choline headgroup. Finally, the UVRR spectra reveal torsion angles, β ^2,1, about the tryptophan backbone in the different environments. The change in shape and breadth of the W3 peak indicates heterogeneity of the TOE structures in solution and bound to vesicles.

The single native tryptophan residue in melittin (Trp-19) is crucial for the folding and function of this toxic peptide.¹⁰ A wide range of spectroscopic studies, including UVRR, has been reported on melittin.^58–60 We aim to apply our current findings to elucidate in-depth molecular details that distinguish the α-helical structures in NaCl and bound to neutral (POPC) and anionic (POPC/POPG) vesicles. As expected, unfolded melittin in buffer exhibits a UVRR signal that supports a polar, hydrogen bonded environment for Trp-19. In 2 M NaCl, however, the UVRR spectrum suggests that Trp-19 is significantly buried in a nonpolar environment with little to no hydrogen bonding. This conclusion is based on values for R_FD and R_W10. Additionally, the W8 band frequency downshifts at least 2 cm⁻¹ in 2 M NaCl, supporting the loss of hydrogen bonding upon formation of tetrameric α-helical melittin. The loss of hydrogen bonding and enhancement of environmental hydrophobicity is accompanied by addition of a cation–π interaction, evidenced by the decrease in R_π by 17%. The most likely candidate for this additional stabilization in an α-helix is a cation at position i + 4 relative to the aromatic residue.⁶¹ Therefore, it is likely that Lys-23 from the same peptide unit is the cation partner for Trp-19, as has been suggested previously.⁶² However, it is possible that an arginine residue from a different peptide unit of the tetramer is sufficiently close to Trp-19 and provides cation–π interaction.

Comparison of the melittin α-helical structures in NaCl and in the presence of charged and neutral vesicles reveals variation in these conformations. Trp-19 of melittin in anionic vesicles is deeply buried in a hydrophobic core, similar to the local environment of melittin in NaCl; however, unlike in NaCl, there is evidence that Trp-19 remains hydrogen bonded to the solvent or a lipid headgroup when bound to anionic vesicles. This increase in hydrogen bonding is evident in the growth of R_W10 and R_W17:W18 values relative to melittin in NaCl. Melittin in neutral vesicles is in a more polar environment as compared to anionic vesicles and also retains hydrogen bonded structure. In addition to enhanced hydrogen bonding, melittin loses cation–π interaction upon interaction with vesicles; the R_π values in vesicle are within 5% of the values in buffer. The loss in intramolecular stabilization of the soluble, tetrameric form resulting from Trp-19 and Lys-23 cation–π interaction appears to be offset by enhanced hydrogen bonding with vesicle or vesicle-embedded solvent moieties. This comparison of the UVRR spectra of the important functional residue of Trp-19 in melittin provides molecular insight into subtle, yet significant, differences that likely contribute to the association of melittin to membrane bilayers. The melittin and TOE examples illustrate new and important spectral observations about tryptophan in membrane proteins and peptides when utilizing the UVRR markers presented in this study.

Summary

Tryptophan is a key residue in membrane-associated proteins and peptides, largely because of its high thermodynamic driving force to localize on the bilayer–water interface. Here, we explore tryptophan UVRR markers for hydrogen bonding, environment polarity, and cation–π interactions. The current study extends prior work conducted under off-resonance conditions, and the results propose relevant and more realistic values for commonly utilized UVRR parameters, such as the Fermi doublet and W17 frequencies. In addition to these well-established markers, a new, unambiguous signature for hydrogen bonding based on W10 and W9 intensities is discussed. Finally, spectral markers that report on cation–π interactions are explored. These UVRR markers are utilized to help interpret spectra of model membrane-associated biomolecules. The results for tryptophan octyl ester and melittin support a picture in which the conversion from an oligomeric α-helical melittin structure to one that is bound to a lipid bilayer is accompanied by important modifications of hydrogen bonding and cation–π interactions as well as environment polarity.