INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

Mammalian pulmonary surfactant is a highly specialized substance that contains approximately 85% phospholipid, 7-10% protein, and 4-8% neutral lipid (Creuwels et al., 1997; Notter, 2000). The most abundant phospholipid class is phosphocholine, whereas anionic phospholipids including phosphoglycerols make up about 10-15% of lung surfactant phospholipids (Hunt et al., 1991; Kahn et al., 1995; Holm et al., 1996). Lung surfactant also contains four apoproteins, including two small hydrophobic proteins (pulmonary surfactant protein SP-B and SP-C) that are known to enhance the adsorption and dynamic film behavior of phospholipids (Hawgood and Schiffer, 1991; Johansson et al., 1994a; Creuwels et al., 1997; Notter, 2000). The lack of surfactant in the under-developed lungs of premature infants is the root cause of respiratory distress syndrome (Avery and Mead, 1959), whereas disruption of surfactant activity is linked to the pathophysiology of clinical lung damage seen in acute respiratory distress syndrome (Pison et al., 1989; Lewis and Jobe, 1993; Notter and Wang, 1997).

Researchers have long used insoluble, monomolecular films spread at the airwater (A/W) interface as models for pulmonary surfactant (Notter, 1984, 2000). Surface balance techniques have been used to study the monolayer properties of the hydrophobic proteins SP-B and SP-C and their mixtures with lipids at the A/W interface (Oosterlaken-Dijksterhuis et al., 1991; Taneva and Keough, 1994a,b,c; Wang et al., 1996). In addition, a variety of biophysical techniques have been used to study surfactant model systems, including electron microscopy (Tchoreloff et al., 1991), Brewster angle microscopy (Discher et al., 1999), fluorescence microscopy (Krüger et al., 1999; Ding et al., 2001; Takamoto et al., 2001), near-field microscopy (Kramer et al., 2000), and scanning probe microscopy (Krol et al., 2000). Although these microscopic techniques provide important biophysical information, they cannot give the same detailed molecular-level information about lipid-protein interactions that can be obtained using vibrational spectroscopic methods.

The use of external reflection Fourier transform infrared spectroscopy (FTIR) to study the structure of monomolecular films directly at the A/W interface was originally developed in the mid 1980s, and progress in this field has recently been reviewed (Mendelsohn et al., 1995; Dluhy, 2000). This infrared reflection-absorption spectroscopy (IRRAS) technique has been applied to the study of monolayer films of extracted lung surfactant preparations (Dluhy et al., 1989) and, more recently, to investigate the roles that SP-B and SP-C play in the function of lung surfactant (Pastrana-Rios et al., 1995; Gericke et al., 1997; Flach et al., 1999). In addition to lipid phase information, it has been shown that amide vibrations can be observed in pure or highly enriched lipid-protein films and that structural information concerning the surfactant proteins can be obtained. Although these studies have contributed significant information to the study of surfactant systems, including the nature of surfactant protein structure and orientation, difficulties remain. In particular, the low band intensities inherent in IRRAS at the A/W interface, and the highly overlapping nature of the amide region makes assignment of protein conformational intermediates problematic.

Recently, an approach using statistical correlation analysis known as two-dimensional infrared spectroscopy (2D IR) has been described to uncover spectral features not readily observable using traditional IR spectroscopy (Noda, 1990, 1993b; Harrington et al., 2000; Noda et al., 2000). Two-dimensional IR spectroscopy is based on the correlation of dynamic spectral variations induced by an external sample perturbation. The effect of these perturbation-induced changes in the local molecular environment is manifested as pseudo time-dependent changes in IR spectral parameters. These resulting dynamic spectra are subject to a cross-correlation analysis that produces 2D maps that can enhance spectral information by spreading out the IR band intensities along two orthogonal axes. 2D IR spectroscopy has particular advantages in simplifying complex spectra, identifying inter- and intramolecular interactions, and facilitating band assignments (Ozaki and Noda, 2000).

Literature references to 2D IR correlation analysis have predominately been in the area of polymer structure, an application for which the method was first developed (Noda et al., 1999). However, the last few years has seen increasing application of this methodology to biological problems, in particular, the use of 2D IR for the study of macroscopic properties of proteins in aqueous solutions. For example, the thermal transitions of a number of proteins has been studied using 2D IR, including cytochrome c (Filosa et al., 2001; Paquet et al., 2001), cytosine monophosphate kinases (Schultz et al., 2000), ovalbumin (Wang et al., 1998), -lactoglobulin (Sefara et al., 1997), avidin (Ismoyo et al., 2000), and synthetic helix-forming peptides (Graff et al., 1997). Several studies have been published that use pH gradients or H-D exchange to enhance the amide spectral region and assign conformations to the underlying band components (Nabet and Pezolet, 1997; Murayama et al., 2001a,b). Studies using 2D hetero-spectral correlations have appeared that enable comparisons to be made among a number of spectral techniques (Kubelka et al., 1999; Pancoska et al., 1999).

Two-dimensional IR correlation analysis has also been used to analyze structure in monomolecular films. The phase behavior of phospholipid monolayers using 2D IR have been studied, and it was shown how these methods could distinguish bands resulting from coexisting phases in a disorder-order phase transition in the monolayer (Elmore and Dluhy, 2000a,b).

Although 2D IR has successfully been used to study structural changes and to make band assignments in proteins, it can also be used in determining the temporal order of events that occur during the external sample perturbation, albeit qualitatively. To more quantitatively describe the degree of coherence between spectral intensity changes and the sequence of molecular events in a set of dynamic spectra, we have recently developed a modified 2D IR correlation method called correlation analysis (Elmore and Dluhy, 2001). This method is a variation of asynchronous cross-correlation, in which dynamically varying spectra are correlated against a mathematical function with varying phase angle. We recently applied correlation analysis to surface pressure-induced changes in the IRRAS spectra of phospholipid monolayers at the A/W interface, and showed how the relative rates of acyl chain and methyl group reorientation could be quantitatively determined (Elmore et al., 2002).

The research described in this paper represents the first study that uses correlation analysis for the study of protein structure. We use this method to probe the conformational intermediates in the surface pressure-resolved IRRAS spectra of two lipid-protein samples: 1,2-dipalmitoyl-sn-glycero-3-phosphoglycerol (DPPG)/SP-C and DPPG/SP-B at the A/W interface. Although these proteins have been previously studied using IRRAS at the A/W interface, the detailed study of their conformational intermediates and reorientation in response to increasing surface pressure has not been completely described.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

Surface chemistry

The synthetic phospholipid DPPG was purchased from Avanti Polar Lipids (Alabaster, AL) and was used as received. ACS grade NaCl and HPLC grade methanol and chloroform were obtained from J.T. Baker (Phillipsburg, NJ). Subphase H₂O was obtained from a Barnstead (Dubuque, IA) ROpure/Nanopure reverse osmosis/deionization system having a nominal resistivity of 18.3 M cm.

Purification of SP-B and SP-C

SP-B and SP-C were purified from calf lung surfactant extract (CLSE) by isocratic normal phase liquid chromatography on Silica C8 as described elsewhere (Baatz et al., 2001). Briefly, ~700 mg CLSE (total lipid plus protein) was initially reduced in volume by evaporation under nitrogen to ~4 ml (~1% of column bed volume). Small amounts of chloroform were added to the concentrated CLSE if the solution became cloudy. The concentrated CLSE was then applied to a 450-ml bed volume LC column that had been pre-equilibrated with 7:1:0.4 MeOH/CHCl₃/5% 0.1 M HCl. The added CLSE was allowed to completely adsorb to the column, and was then eluted with 7:1:0.4 MeOH/CHCl₃/5% 0.1 M HCl at a flow rate of 0.4 ml/min and UV detection at 254 nm. SP-B eluted from the column as the first peak, whereas SP-C eluted as the second peak as determined by SDS PAGE and protein sequencing (see below). SP-B fractions were combined and, as determined by the method of Shin, the final SP-B solution was free of phospholipid (Shin, 1962), whereas SDS PAGE and protein sequencing indicated the absence of SP-C (see below). Fractions containing SP-C were combined, concentrated via N₂ gas stream and run on a second C8 LC column. Fractions containing purified SP-C from this second column were combined. Phospholipid content of purified SP-C using the method of Shin (1962) was determined to be <4 mole lipid/mole SP-C. Protein sequencing analysis (see below) indicated that the purified SP-C was free of SP-B.

SDS PAGE and amino acid analysis

For SDS PAGE, 20 ml of the appropriate column fractions were suspended in NuPage sample buffer (Invitrogen, Carlsbad, CA) and applied to 4-10% gradient acrylamide Tris/Bis gels (NuPage gels, Invitrogen) under nonreducing conditions. Electrophoresis was performed at a constant voltage of 200 V for 40 min with a morpholinoethanesulphonic acid (MES) buffer containing 50 mM MES/50 mM Tris Base/3.5 mM SDS/1 mM EDTA, pH 7.7 as the running buffer. Silver staining for detection of protein bands was performed according to Morrissey (1981). N-terminal amino acid sequence analysis was performed for seven to ten cycles using an Applied Biosystem Procise sequence analyzer. SP-B and SP-C were identified by the N- terminal sequences of Phe-Pro-Ile-Pro-Ile-Pro and Leu-Ile-Pro-???-???-Pro-Val, respectively, where positions 4 and 5 in the SP-C sequence were both assumed to be Cys. The concentrations of SP-B and SP-C were determined by the BCA total protein assay (Sigma Chemical Co., St. Louis, MO).

Preparation of samples

Stock solutions of DPPG (~2.5 mg/ml in 4:1 CHCl₃:MeOH) were prepared and concentrations verified by inorganic phosphorus assay (Chen et al., 1956). Solutions of the lipid and proteins containing 20:1 lipid-protein mol:mol (for SP-C) or 40:1 lipid-protein mol:mol (for SP-B), i.e., ~22 wt% for both SP-B and SP-C, were prepared by mixing the appropriate amounts of the phospholipid stock solution together with the stock solutions of SP-B or SP-C in 1:1 CHCl₃:MeOH. The subphase used for all experiments was 150 mM NaCl in deionized H₂O (pH 5.6).

FTIR external reflectance measurements

Infrared external reflection-absorbance spectra of monolayers at the A/W interface were acquired using a Perkin-Elmer Spectrum 2000 FTIR spectrometer equipped with an external sample beam. A 60° gold-coated, off-axis parabolic mirror (Janos Technology Inc., Townshend, VT) reflected the beam coming from the spectrometer onto the surface of a Nima 601M film balance (Coventry, U.K.) at an incidence angle of 30° to the surface normal. The beam reflects off of the subphase, sampling the film, and a second parabolic mirror collects the beam and directs it into a collection mirror and then onto the sensing chip of a liquid N₂-cooled HgCdTe detector. The film balance, optical components, and detector are housed in a sealed, Plexiglas chamber that allows humidity control of the local trough environment, thus improving water vapor background subtraction. A schematic diagram of the experimental set-up has been previously published (Dluhy, 2000).

The subphase was first cleaned by aspiration, and a single beam spectrum was collected for use as the IR background spectrum. The subphase temperature was held constant at 22 ± 1°C by flowing thermostatted water through the hollow body of the trough. The temperature in the enclosed chamber was typically 24°C and the relative humidity remained fairly constant at 70%. Typically 5-10 µl of sample was spread via syringe onto the trough surface. The film was allowed to equilibrate for a period of 30 min and then was compressed intermittently and spectra collected over a range of surface pressures from ~5 mN/m to a maximum of 45-65 mN/m depending on the nature of the film.

External reflection-absorption spectra were collected with 1024 scans at 16-cm¹ resolution, apodized with a Norton-Beer (medium) function, and were Fourier transformed with one level of zero filling. A resolution of 16 cm¹ was chosen for several reasons: 1) time of collection is minimized and SNR is maximized when spectra are collected at lower resolution, 2) residual water vapor bands are easier to subtract at low resolution when the relative humidity varies slightly during the course of the experiment, and 3) valid statistical correlation analyses can be constructed for IR spectra collected at lower resolutions, as has recently been discussed (Berry and Ozaki, 2001).

All monolayer spectra are presented as reflection-absorption spectra, i.e., A = log(R/R₀), where R is the IR reflectivity of the monolayer surface and R₀ is the IR reflectivity of the bare water subphase background. The reflectance IR spectra used in the analyses presented here were baseline corrected using the GRAMS/32 (Galactic Industries, Salem, NH) software package before determination of peak positions and band intensities; in addition, residual water vapor bands have been subtracted. Adjustments for changes in surface density (i.e., intensity normalization) were also performed using GRAMS/32. Other than baseline correction and intensity normalization, the spectra have not been smoothed or further processed. Vibrational frequencies were calculated using a 5-point center of gravity algorithm (Cameron et al., 1982) written in our laboratory for the Grams/32 environment.

Calculation of 2D IR correlation spectra

The 2D IR synchronous spectrum, (₁, ₂), and the asynchronous spectrum, (₁, ₂), were calculated using Eqs. 1 and 2. These algorithms use the most recent mathematical formalism in which a Hilbert transform is utilized for calculating the asynchronous spectrum rather than the more commonly used Fourier transform (Noda, 2000). In all cases, the average spectrum was subtracted from each sequentially obtained surface pressure-dependent IRRAS spectrum to produce a set of dynamic IR spectra. The dynamic spectra were then used in the correlation analysis

(1)

(2)

In Eqs. 1 and 2, ₁ and ₂ represent two independent frequencies, n_j represents the number of the spectrum in the ordered sequence where the first spectrum number is zero, N represents the total number of spectra used in the calculation, and M_jk is the Hilbert transform matrix, which is defined in

(3)

The 2D plots presented in this article were calculated using 2D IR correlation analysis algorithms written in our laboratory for the METLAB programming environment (Version 6, The MathWorks, Inc., Natick, MA).

Correlation analysis

A correlation analysis is a mathematical asynchronous cross correlation performed on a set of dynamically varying IR spectra against a set of sinusoidal functions that differ only by their phase angle . A full description of the details of the correlation analysis has been presented elsewhere (Elmore and Dluhy, 2001). This type of correlation analysis is mathematically described using

(4)

The correlation intensity  at some point (, ) represents the correlation of the measured IR spectral intensity y(, n_j) with the mathematical function sin(k + ). In Eq. 4, y is the IR intensity;  is the frequency or wavenumber; n_j is the number of the spectrum in the ordered sequence where the first spectrum number is zero;  is the phase angle of the respective sine function; N is the total number of spectra used in the calculation;  is a constant value in degrees (or radians) chosen based upon the total number of dynamic spectra used in the calculation, and M_jk is the Hilbert transform matrix previously defined in Eq. 3.

In this study, all correlations were performed with  = 10^o, so that sin(k10° + ) describes ~1/4 of the cycle of a sine function, or the approximate form of a commonly observed variation in spectral band intensities upon sample perturbation. Only the asynchronous correlation algorithm is used in the correlation analysis presented here, since asynchronous 2D IR correlations are more sensitive to differences in the form of the signal variation than are synchronous correlations (Noda, 1990). Note also that the computational algorithm for the correlation analysis uses the most recent mathematical formalism, in which a Hilbert transform is used for calculating the asynchronous spectrum, rather than the more computationally cumbersome Fourier transform (Noda, 2000).

The effective phase angle, _e, is defined by

(5)

In Eq. 5,  is the point of maximum positive correlation intensity in the plot of  versus  as defined by Eq. 4. The value of _e is defined in this fashion so that the phase angle  and the effective phase angle _e are the same for a sinusoidal signal variation with constant frequency. In this article, the contour levels are evenly spaced in the plots from 0 to the maximum value for positive correlations. Negative correlations are not displayed because they simply differ from the positive correlations by 180^o. The plots of effective phase angles versus wavenumber were calculated using correlation analysis algorithms written in our laboratory for the MATLAB programming environment.

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

IRRAS spectra of lipid:protein monolayer films

Infrared external reflectance-absorption spectra were obtained at the A/W interface for monolayer films of DPPG plus the surfactant proteins SP-B and SP-C. These spectra are shown in Fig. 1. These IR spectra were acquired while the monolayer was held at specific surface pressure values from 4.0 mN/m to 40.0 mN/m during step-wise compression of the film balance. External reflection-absorption IR spectra of monomolecular films at the A/W interface differ substantially from spectra obtained by conventional transmission IR spectroscopy. In particular, the complex refractive indices of both sample and substrate contribute to the observed IRRAS spectra of A/W monolayers. Therefore, these spectra are functions of the wavelength, state of polarization, thin film thickness, angle of incidence of the incoming light, and the optical constants of the three phases involved (Dluhy, 1986). The physical basis for the method and the use of IR spectroscopy to study monolayer films at the A/W interface has recently been reviewed (Mendelsohn et al., 1995; Dluhy, 2000).

View larger version (28K): [in this window] [in a new window]   FIGURE 1   IR external reflection-absorption spectra of a lipid:protein monolayer showing the amide I and amide II spectral regions collected at surface pressures from 4.0 to 40.0 mN/m. The spectra were collected as a function of increasing monolayer surface pressure and have been normalized for changes in surface density. (A) DPPG/SP-C (20:1, mol:mol, ~22 wt% SP-C). (B) DPPG/SP-B (40:1 mol:mol, ~22 wt% SP-B).

Monolayer IRRAS spectra were collected using an unpolarized incident IR source and have been normalized to account for changes in trough area. Normalization refers to the adjustment of IR spectral intensities to take into account changes in surface density as the trough area available to the monomolecular film decreases during compression. Intensity changes in area-normalized monolayer spectra more accurately reflect conformational changes in the monolayer, as opposed to merely reflecting an increase in number density. We have previously shown that intensity normalization is important for understanding 2D IR correlation maps calculated from IRRAS monolayer spectra (Elmore and Dluhy, 2000b). In addition, the spectra used in these analyses were collected at 16-cm¹ resolution. The issue of resolution is particularly relevant to monolayer IRRAS at the A/W interface, because it is a low-intensity reflectance technique with inherently weak (10³-10⁴ AU) band intensities. Due to the very broad amide I proteins bands (bandwidths > 30 cm¹), a spectral resolution of 16 cm¹ was used to minimize collection time and maximize signal to noise. A recently published paper has addressed instrumental issues in the calculation of 2D IR spectra and has shown that lower resolution spectra may be appropriate for use in 2D calculations under certain circumstances (Berry and Ozaki, 2001).

The overlaid spectra in Fig. 1 are reflection-absorbance spectra displaying the amide I (1700-1600 cm¹) and amide II (1600-1500 cm¹) regions of the surfactant proteins. Also observed is the C==O vibration from the phospholipid between 1750 and 1700 cm¹. These IRRAS absorption bands exhibit negative IR intensities, in agreement with the experimental conditions (Dluhy, 1986). Figure 1 A illustrates IRRAS spectra at the A/W interface for a DPPG/SP-C monolayer at a protein concentration of ~22 wt%. Clearly evident in the spectra are the lipid C==O band at 1738 cm¹, two protein amide I bands at ~1654 and 1625 cm¹, and the protein amide II band at ~1550 cm¹. Figure 1 B illustrates the same spectral regions as in Fig. 1 A for a DPPG/SP-B monolayer also at a protein concentration of ~22 wt%. The identical carbonyl and amide bands are evident in the spectra of the SP-B monolayer as in the spectra of the SP-C monolayer, although relative band heights for the amide vibrations differ.

Identification of unique surface pressure regimes using windowed autocorrelation analysis

Before analysis of the lipid-protein monolayer spectra using 2D IR correlation spectroscopy, we used a windowed autocorrelation method to identify the surface pressure regimes that encompass the greatest variation in amide spectral intensity. These regions can be located by adapting a dynamic filtering technique that autocorrelates IR intensities within a defined surface pressure window and plots them against the average surface pressure, thereby locating the regions of maximal spectral variation. This technique has been previously applied to the temperature-dependent 2D IR spectra of liquid crystals (Thomas and Richardson, 2000).

The windowed autocorrelation analysis method begins by separating the monolayer IRRAS spectra into a smaller data set containing only the first four spectra of lowest surface pressure (e.g., P₁ through P₄). An average spectrum is calculated from the spectra in this window; the individual spectra in the window are mean-centered by subtraction of the average. Next, 2D IR analysis is used to calculate the correlation intensities for each spectral frequency in this windowed, mean-centered data set. The resulting autocorrelation intensities represent the amount of spectral variation that occurs at each frequency as a function of the average surface pressure of the windowed data set. This autocorrelation window is then swept over the entire data set, translating it one surface pressure-resolved spectrum at a time (i.e., the second window contains P₂ through P₅, etc.).

The autocorrelation spectrum that results from this process is plotted as a function of the average surface pressure of the autocorrelation window. The spectral frequencies that have the largest autocorrelation intensities will be those frequencies at which the largest spectral variations occur. In this manner, surface pressure regions that contain maximal spectral variations may be identified. With this information at hand, 2D IR or correlation analysis can be performed solely within these regions to establish the structural or temporal relationships that contribute to the spectral variations.

We have applied this windowed autocorrelation analysis method to the monolayer IRRAS spectra of DPPG/SP-C and DPPG/SP-B. The results are shown in Fig. 2. In analyzing these spectra, we were particularly concerned with the protein structural components, hence we concentrated on autocorrelation of the amide I spectral region. Figure 2 A shows a contour plot of the autocorrelation spectra for the DPPG/SP-C sample. This contour plot is dominated by a large band at ~1650 cm¹ that shows two intensity maxima: one between ~12 and 18 mN/m and one above ~27 mN/m. Based on these autocorrelation spectra, we can reliably partition the surface pressure-resolved spectra of the DPPG/SP-C monolayer into two pressure regions, low (4-25 mN/m) and high (25-40 mN/m). A similar contour plot of the autocorrelation intensities for the DPPG/SP-B spectra is shown in Fig. 2 B. In this case, an intense autocorrelation maxima at ~1650 cm¹ is observed at high surface pressure (25 mN/m), however, lower intensity contours occur below 25 mN/m. Therefore, we have also divided the surface pressure-resolved spectra of the DPPG/SP-B monolayer into two pressure regimes: low (4-25 mN/m) and high (25-40 mN/m). The approach that we have taken here isolates the surface pressure regions that contain maximal spectral variations. Using these newly defined surface pressure regimes, we have performed 2D IR and correlation analyses on the IRRAS spectra of both the SP-C- and SP-B-containing monolayers to better understand the surface pressure-induced protein rearrangements in the monolayer.

View larger version (46K): [in this window] [in a new window]   FIGURE 2   Surface pressure-wavenumber contour plot of autocorrelation intensities based on a dynamic filtering technique using a windowed autocorrelation analysis method. Contours result from the autocorrelation intensities calculated as a function of a translating surface pressure window. (A) DPPG/SP-C (20:1, mol:mol, ~22 wt% SP-C). (B) DPPG/SP-B (40:1 mol:mol, ~22 wt% SP-B).

2D IR correlation analysis of the SP-C amide I region

Synchronous and asynchronous 2D IR correlation maps were calculated for the SP-C-containing monolayer to investigate protein conformational changes that occur upon monolayer compression. Figure 3 shows the 2D contour map for the synchronous correlation of the DPPG/SP-C IRRAS spectra, in both the low surface pressure regime below 25 mN/m (Fig. 3 A) and the high surface pressure region above 25 mN/m (Fig. 3 B). Synchronous 2D spectra are characterized by on-diagonal auto peaks and off-diagonal, symmetric cross-peaks. Synchronous 2D auto peaks simply reflect how the spectral intensity responds to the external perturbation; synchronous cross-peaks, in contrast, develop when two separate transition dipole moments are significantly coupled, or if they reorient in phase to the external perturbation (Noda, 1993a).

View larger version (54K): [in this window] [in a new window]   FIGURE 3   Synchronous 2D IR correlation plots of the monolayer IRRAS spectra of DPPG/SP-C (20:1, mol:mol, ~22 wt% SP-C). Solid lines indicate regions of positive correlation intensity; dashed lines indicate regions of negative correlation intensity. Spectra were divided into two pressure regions based on their autocorrelation intensities. 2D IR synchronous correlations were calculated on spectra contained within the pressure region. The one-dimensional spectrum shown in this figure is the calculated average of the spectra used in the 2D analysis. (A) 4.0-25.0 mN/m. (B) 25.0-40.0 mN/m.

Note that, due to the symmetric properties of synchronous correlation cross-peaks with respect to the diagonal, it is necessary to describe only the cross-peaks above the diagonal line. This property also holds for the antisymmetric properties of the cross-peaks in the asynchronous correlation plots (see below). In this article, correlation peaks are described as ₁ versus ₂, where ₁ refers to the wavenumber value of the x axis and ₂ refers to the wavenumber value of the y-axis.

Mature SP-C consists of 35 amino acids and is an extremely hydrophobic, predominately -helical protein of approximately 4.2 kD with charged amino acids (K10 and R11) near its N terminus. The cysteines C4 and C5 are acylated in bovine SP-C. The NMR structure of SP-C in apolar solvent essentially describes the protein as a rigid rod in which only a few residues near the N terminus (L1-P7) and the C terminus itself are not helical (Johansson et al., 1994b). The length of this helix (V8-G34) is ~39 Å with a diameter of ~12 Å.

The 2D IR synchronous map of the low-pressure region of the DPPG/SP-C monolayer (Fig. 3 A) is dominated by a strong auto peak at 1652 cm¹ with less intense auto peaks at 1672 and 1625 cm¹. Positive cross-peaks are observed at 1652 versus 1625 cm¹ and 1685 versus 1625 cm¹. Positive synchronous cross-peaks indicate a coordinated spectral response in which the functional groups are reorienting in the same direction. In addition, there is a negative cross-peak between the 1676 versus 1630 cm¹ bands. Negative synchronous cross-peaks also indicate significantly coupled molecular reorientation, albeit one where the spectral intensity of one component increases while the second decreases.

The high surface pressure region (>25 mN/m) of the DPPG/SP-C monolayer is also dominated by the major auto peak at 1656 cm¹ (Fig. 3 B) with additional auto peaks observable at 1685, 1671, 1666, and 1622 cm¹. Cross-peaks associated with the 1622-cm¹ band can be observed at wavenumber values 1656 versus 1622 cm¹, 1671 versus 1622 cm¹ and 1685 versus 1622 cm¹. Additional cross-peaks at 1671 versus 1654 cm¹ and 1685 versus 1654 cm¹ are also observed.

The wavenumber location of the cross-peaks in the 2D IR synchronous map can be identified with protein secondary structure conformations (e.g., -helices, -sheets, and turns, unordered structure, etc.) using previously published IR correlations, see e.g., Byler and Susi (1986), Surewicz et al. (1993), Goormaghtigh et al. (1994), and Jackson and Mantsch (1995). Caution is advisable when making conformational assignments using IR amide I spectra because secondary structure correlations with a specific wavenumber are not unique, and a wavenumber range exists in the amide I region for any particular protein conformation. Nevertheless, using the most well-established associations, the cross-peak at 1654 cm¹ can be considered characteristic of -helices, whereas the bands at 1630 and 1676 cm¹ are assigned to -sheet and -turn/loop structures, respectively.

In both the raw spectra (Fig. 1) and the synchronous and asynchronous (see below) 2D maps for DPPG/SP-C, we also observe bands and cross-peaks at wavenumber values both high (1687 cm¹) and low (1620 cm¹) in the amide I range. Using ab initio calculations, this pair of peaks has been assigned to a structure of extended, multistranded, antiparallel -sheet aggregates (Kubelka and Keiderling, 2001); these peaks have also been experimentally observed for a model  -sheet peptide (Khurana and Fink, 2000). This association is strengthened in the case of SP-C, because the 2D synchronous maps for both low and high pressures show a cross-peak between 1687 and 1620 cm¹, indicating a coordinated response between the two bands. The implication for SP-C, a predominately -helical protein (Johansson et al., 1994b), is that some portion of the molecule undergoes intermolecular hydrogen bonding resulting from aggregation of unfolded protein segments.

Taking these structural assignments into account, the synchronous 2D map for SP-C shows that the -helix is the predominate structural motif in both the low and high pressure regions, as expected. However, the cross-peaks at 1625, 1630, 1672, and 1685 cm¹ also demonstrate that SP-C contains a more varied secondary structure containing a higher degree of aggregated -strands than previously reported.

Asynchronous 2D IR correlation maps were also calculated from the DPPG/SP-C IRRAS monolayer spectra. In a similar fashion to the synchronous plots of Fig. 3, Fig. 4 shows the 2D contour map for the asynchronous correlation of the DPPG/SP-C spectra, in both the low surface pressure regime below 25 mN/m (Fig. 4 A) and the high surface pressure region above 25 mN/m (Fig. 4 B). Asynchronous 2D spectra are antisymmetric with respect to the diagonal in the correlation map and contain no auto peaks; the spectrum consists only of off-diagonal cross-peaks with two intensity maximaone positive and one negative. Peaks appear in asynchronous 2D correlation maps if the transition dipole moments are significantly decoupled, or if the dipole moments reorient out-of-phase or at different rates in response to the external perturbation. This attribute is used to unmask the differential response of functional groups in the molecule, and makes asynchronous 2D correlation plots particularly useful for resolution enhancement (Noda, 1993a).

View larger version (51K): [in this window] [in a new window]   FIGURE 4   Asynchronous 2D IR correlation plots of the monolayer IRRAS spectra of DPPG/SP-C (20:1, mol:mol, ~22 wt% SP-C). Solid lines indicate regions of positive correlation intensity; dashed lines indicate regions of negative correlation intensity. Spectra were divided into two pressure regions based on their autocorrelation intensities. 2D IR asynchronous correlations were calculated on spectra contained within the pressure region. The one-dimensional spectrum shown in this figure is the calculated average of the spectra used in the 2D analysis. (A) 4.0-25.0 mN/m. (B) 25.0-40.0 mN/m.

The asynchronous 2D IR map of SP-C also indicates the presence of a varied, heterogeneous secondary structure for SP-C, in agreement with the synchronous correlation spectra. Four prominent cross-peaks are observed in the low-pressure region (Fig. 4 A) at 1652 versus 1634 cm¹ (), 1663 versus 1652 cm¹ (+), 1675 versus 1663 cm¹ () and 1685 versus 1675 cm¹ (+). Less intense, broad cross-peaks are observed for the association of the band attributed to a -sheet intermolecular aggregation (~1615 cm¹) with the bands 1634, 1652, and 1663 cm¹. The presence of asynchronous cross-peaks at 1615, 1634, 1675, and 1685 cm¹ confirms the presence of these bands in the synchronous spectrum, and indicates that the SP-C protein conformation is composed of extended -sheet structure.

In the high pressure region above 25 mN/m, the asynchronous 2D correlation plot for DPPG/SP-C is characterized by a number of different cross-peaks (Fig. 4 B), with a more complicated cross-peak structure than is observed in the asynchronous spectrum of the low-pressure region. Major positive asynchronous correlations are observed between 1654 versus 1647 cm¹, 1675 versus 1647 and 1665 cm¹, with an elongated correlation intensity band at 1688 versus 1647, 1665 and 1680 cm¹. Major negative asynchronous correlations are observed between 1665 versus 1654 cm¹ and 1680 versus 1654 and 1675 cm¹. In addition, there is an elongated negative correlation intensity band between ~1647-1680 versus 1625 cm¹.

The most relevant features of the asynchronous spectrum for the high pressure region are the split -helix peak at 1654/1647 cm¹, the elongated correlation intensity maxima associated with the 1680 cm¹ band, and the elongated correlation intensity minima associated with the 1625 cm¹ band. The ability of asynchronous 2D IR to resolve overlapped peaks is demonstrated in Fig. 4 B because the low-pressure asynchronous -helix cross-peak at 1652 cm¹ has divided into two components at high pressure, one at 1654 cm¹ and one at 1647 cm¹. This is the first report of two coexisting -helix conformations in SP-C lipid-protein monolayers at high surface pressure. The 1647-cm¹ helix band, in particular, shows asynchronous cross-peaks with a number of other bands, including 1654, 1675, and 1680 cm¹, indicating an out-of-phase response of this conformation with the other protein conformations. These large elongated asynchronous correlation features at 1620 and 1680 cm¹ indicate that the motion of the extended -structure is also significantly decoupled from the rest of the protein and reorients independently of the main helix structure at high surface pressures.

The asynchronous correlation spectrum for SP-C (Fig. 4) presents a more complex band structure than does the synchronous correlation spectrum (Fig. 3). The multiple cross-peaks observed in the asynchronous correlation plot demonstrate that SP-C is composed of a varied secondary structure. The major result of the asynchronous 2D correlation analysis is that the main -helix band splits into two peaks at high surface pressures, indicating two different coexisting helix conformations for SP-C. The peaks at ~1634 and 1675 cm¹ indicate that a -sheet structure exists at both low and high surface pressures. Also, the cross-peaks at ~1620 and 1680 cm¹ demonstrate that an intermolecular aggregation of extended -sheet structure exists in the monolayer. This is possibly due to the relatively high amount of SP-C in the monolayer (~22 wt%), as compared to the physiologically relevant concentration of ~1 wt%. A previous IRRAS study has documented the presence of protein aggregation in highly enriched lipid/SP-C monolayers (Pastrana-Rios et al., 1995). However, other more recent research suggests that monomeric -helical SP-C is thermodynamically metastable, with the peptide irreversibly forming -sheet aggregates resembling amyloid fibrils, which may be implicated in pulmonary alveolar proteinosis (Szyperski et al., 1998; Gustafsson et al., 1999). Therefore, the presence of several -helix conformations and the -aggregate bands in the 2D spectra may indicate an intermediate in the -to- conversion, which is known to be a slow, kinetically controlled process. This is the first time that specific IR evidence has demonstrated that multiple -helix conformations and -structure conformational intermediates exist for SP-C-containing monolayers at the A/W interface.

Correlation analysis of the SP-C amide I region

In addition to its use in studying structural changes and making band assignments, 2D IR correlation spectroscopy has also been used to determine the temporal order of events that occur during the external sample perturbation. The basis for this determination is the relative signs of the asynchronous and synchronous cross-peaks (Noda, 1993a). A positive asynchronous cross-peak at (₁, ₂) indicates that the intensity change at ₁ occurs before ₂; A negative cross-peak at ₁ is observed if the change occurs after ₂. This rule, however, is reversed if the corresponding synchronous peak at (₁, ₂) has a negative sign, i.e., (₁, ₂) < 0. Although it is possible to determine the relative sequence of molecular rearrangements based on comparison of the signs of the cross-peaks in the asynchronous versus synchronous correlation maps, this procedure is somewhat cumbersome, inherently qualitative in nature, and leads to uncertainties for highly overlapped spectra.

To more quantitatively describe the degree of coherence between the observed spectral intensity changes and the sequence of molecular events in a discrete set of dynamic spectra, we have recently developed a modified 2D IR correlation method called correlation analysis (Elmore and Dluhy, 2001). In this method an asynchronous cross-correlation is performed using a set of dynamically varying spectra, i.e., y(, n_j), against a mathematical function that approximates the functional form that the external perturbation induces on the IR spectral intensities. To date, we have used a sine function, e.g.,. sin(k + ). The resulting correlation intensities are a function of the spectral frequency () and the phase angle () of the mathematical function. The maximum correlation intensity will be observed at one point (, ) in the correlation plot for the range 360 > >0; this point is used to define a new parameterthe effective phase angle _e of f(, ). The _e value quantitatively reveals the degree of coherence between the experimental intensities and the sequence of molecular events in a discrete set of dynamic spectra. We recently applied correlation analysis to surface pressure-induced changes in the IRRAS spectra of phospholipid monolayers at the A/W interface, and showed how the relative rates of acyl-chain and methyl-group reorientation could be quantitatively determined (Elmore et al., 2002).

The correlation plot for the amide I region of SP-C at low surface pressure (<25 mN/m) is shown in Fig. 5 A. In addition, the values for the effective phase angle (_e) and the band assignments for the peaks in this plot are presented in Table 1. It is immediately obvious from Fig. 5 A that the most intense peak in the plot is observed at 1650 cm¹ and corresponds to the -helix of SP-C. However, peaks due to  strands (1663 cm¹) and to extended, aggregated  structures (1620 and 1682 cm¹) are also apparent. (The peak at 1606 cm¹ is likely due to a side-chain vibration and is not included in this analysis.) It is also apparent from Fig. 5 A and Table 1 that each of the amide I peaks for SP-C at low surface pressures have nearly identical _e values, with a standard error of <1% relative to the mean. These data confirm that all segments of the SP-C protein reorient at the identical relative rate when the surface pressure is increased up to 25 mN/m.

View larger version (53K): [in this window] [in a new window]   FIGURE 5   Correlation plots derived from the monolayer IRRAS spectra of DPPG/SP-C (20:1, mol:mol, ~22 wt% SP-C). Spectra were divided into two pressure regions based on their autocorrelation intensities. correlations were calculated on spectra contained within the pressure region. (A) 4.0-25.0 mN/m. (B) 25.0-40.0 mN/m.

View this table: [in this window] [in a new window]   TABLE 1   Effective phase angle, e, values calculated for DPPG/SP-C monolayer IRRAS spectra in amide I region

The relative reorientation of SP-C secondary structure becomes more complicated at high surface pressures (25-40 mN/m), as illustrated in Fig. 5 B. As seen in Table 1, the _e values at high pressures divide into three identifiable groups. The largest effective phase angle (indicative of the most rapid reorientation) is that of the 1656 cm¹ peak with _e = 341.9, attributable to the high wavenumber -helix conformation. The second group to reorient includes the peaks at 1687 cm¹ (_e = 273.4), 1636 cm¹ (_e = 270.9), and 1625 cm¹ (_e = 263.5), all of which can be attributed to extended -sheet structures. Last, the third group to reorient at a much slower relative rate includes the peaks at 1678 cm¹ (_e = 40.5), 1665 cm¹ (_e = 43.0), and 1647 cm¹ (_e = 53.2). The cross-peaks in this group are associated with  turn/loop structures and the low wavenumber -helix conformation. The plot at high surface pressures (Fig. 5 B) shows that most of the correlation intensity of the two -helix peaks is concentrated in the lower wavenumber peak at 1647 cm¹, which is also the helix conformation that reorients the slowest.

A consideration of the 2D IR and _e values for SP-C leads to the following model for protein reorientation. At low surface pressures, the protein exists in a variety of secondary structure conformations, most noticeably the predominate -helix, but also including extended -sheet structures. All conformations of the SP-C molecule react identically to increasing surface pressure and reorient in phase with each other. Above 25 mN/m, however, the increasing surface pressure selectively affects the coexisting protein conformations and leads to an independent reorientation of these protein subunits. The high wavenumber -helix conformation reorients first, closely followed by the reorientation of the extended  structures, including that of the aggregated protein strands. The last protein reorientational motion to occur originates from the low-wavenumber -helix conformation and from  turn/loop and unordered structures, which are most likely due to very short protein fragments that link together the more ordered protein segments. The reorientation motion of this second helix conformation and the  turn/loop fragments lags significantly behind the reorientation rates of the other ordered ( and ) protein segments. As the majority of the correlation intensity is concentrated in this second helix conformation, it is presumably this conformation that orients its helix axis toward the surface normal, as previously described (Gericke et al., 1997).

It is possible that the selective reorientation of the two -helix conformations plays a role in the SP-C-mediated formation of three-dimensional, surface-associated, lipid-protein structures at high surface pressures. Using microscopic techniques methods, these structures have been observed to be formed at high surface pressures when human recombinant or synthetic SP-C is incorporated in lipid monolayers (Galla et al., 1998; Bourdos et al., 2000; Kramer et al., 2000). Although models have been proposed for the formation of these structures, the details of how SP-C facilitates their development is still unknown. However, reorientation of the -helix is postulated to play a pivotal role.

2D IR correlation analysis of the SP-B amide I region

We have also used 2D correlation methods to study protein conformational changes in monolayers of DPPG/SP-B. In contrast to the DPPG/SP-C monolayers, the DPPG/SP-B samples were prepared at a 40:1 ratio (lipid:protein, mol:mol). Due to the higher molecular weight of SP-B, however, this mol ratio still equates to ~22 wt% protein in the monolayer. As described above, we have used a windowed autocorrelation method to divide the DPPG/SP-B monolayer spectra into two separate regions for detailed analysis (Fig. 2 B). The autocorrelation method defines a low-pressure regime (4-25 mN/m) and a high pressure regime (25-40 mN/m) for DPPG/SP-B monolayers, similar to the case of DPPG/SP-C.

The mature form of SP-B is a highly charged protein containing 79 amino acids of ~18 kD. A high percentage of these amino acids are cysteine, basic, or hydrophobic residues. The protein exists as a disulfide-linked homodimer (Hawgood et al., 1987) and is expected to have several amphipathic -helical segments on both the amino and carboxy terminal ends (Gustafsson et al., 1999). In addition, each SP-B monomer contains three intramolecular disulfide bridges linking cysteines residues; a fourth disulfide bridge is responsible for the intermolecular dimerization. Evidence suggests that SP-B is not a transmembrane or transmonolayer protein. IR results of lipid-protein vesicles (Vandenbussche et al., 1992) demonstrated that domains of SP-B are associated with the phospholipid headgroups, whereas other domains are located inside the bilayer. Fluorescence anisotropy also determined that SP-B was not a transmembrane protein, but was associated with the membrane surface (Baatz et al., 1990). The polypeptide motif of the SP-B monomer is characterized by amphiphatic -helices with solvent-associated hydrophilic side chains, whereas other hydrophobic conformations form a protein core stabilized by intramolecular disulfide bonds (Hawgood et al., 1998).

Figure 6 shows the 2D contour map for the synchronous correlation of the DPPG/SP-B IRRAS spectra, in both the low surface pressure regime below 25 mN/m (Fig. 6 A) and the high surface pressure region above 25 mN/m (Fig. 6 B). The synchronous map of the low-pressure region (Fig. 6 A) is dominated by strong auto peaks at 1653 and 1686 cm¹ with less intense auto peaks at 1640 and 1625 cm¹. Positive cross-peaks are formed between the -helix band at 1653 versus 1640 and 1625 cm¹. In addition, a number of positive cross-peaks are formed between the -structure bands at 1685 and 1676 cm¹ versus 1653, 1640, and 1625 cm¹. Negative cross-peaks occur with the band at 1611 cm¹. However, as in the case of SP-C, the peak at this wavenumber is most likely due to side-chain vibrations unrelated to protein secondary structure.

View larger version (55K): [in this window] [in a new window]   FIGURE 6   Synchronous 2D IR correlation plots of the monolayer IRRAS spectra of DPPG/SP-B (20:1, mol:mol, ~22 wt% SP-B). Solid lines indicate regions of positive correlation intensity; dashed lines indicate regions of negative correlation intensity. Spectra were divided into two pressure regions based on their autocorrelation intensities. 2D IR synchronous correlations were calculated on spectra contained within the pressure region. The one-dimensional spectrum shown in this figure is the calculated average of the spectra used in the 2D analysis. (A) 4.0-25.0 mN/m. (B) 25.0-40.0 mN/m.

The high surface pressure region (25 mN/m) of the DPPG/SP-B monolayer presents a simpler correlation map than does the low-pressure region because it is dominated by the major -helical auto peak at 1653 cm¹ (Fig. 6 B). There are additional auto peaks at 1686, 1672, and 1621 cm¹, whereas positive cross-peaks can be observed at wavenumber values 1653 versus 1621 cm¹, 1676 versus 1653 and 1621 cm¹, and 1685 versus 1653 cm¹.

The cross-peaks seen in the 2D synchronous spectra of DPPG/SP-B are very similar in wavenumber to the synchronous cross-peaks calculated for SP-C, indicating a high helical content for SP-B with contributions from -sheet and unordered structure. These types of secondary structures have previously been observed in bulk phase IR studies of SP-B or the truncated N-terminal peptide of SP-B (Vandenbussche et al., 1992; Gordon et al., 2000). Also evident in the 2D spectra are the 1620/1685 cm¹ bands attributable to extended -aggregated structures. These extended -structures bands have not been seen in either the previous bulk phase studies or the IRRAS studies of SP-B mentioned above. The synchronous 2D IR spectrum of SP-B mainly differs from that of SP-C in that, for SP-B, the 2D correlations become less numerous at high surface pressures and are dominated by the -helix band at 1653 cm¹.

Asynchronous 2D IR correlation maps were also calculated from the DPPG/SP-B IRRAS monolayer spectra. Figure 7 shows the 2D contour map for the asynchronous correlation of the DPPG/SP-B spectra, in both the low surface pressure regime below 25 mN/m (Fig. 7 A) and the high surface pressure region above 25 mN/m (Fig. 7 B). The asynchronous 2D IR map of SP-B also indicates the presence of a heterogeneous secondary structure for this protein in the DPPG/SP-B monolayer film.

View larger version (55K): [in this window] [in a new window]   FIGURE 7   Asynchronous 2D IR correlation plots of the monolayer IRRAS spectra of DPPG/SP-B (40:1, mol:mol, ~22 wt% SP-B). Solid lines indicate regions of positive correlation intensity; dashed lines indicate regions of negative correlation intensity. Spectra were divided into two pressure regions based on their autocorrelation intensities. 2D IR asynchronous correlations were calculated on spectra contained within the pressure region. The one-dimensional spectrum shown in this figure is the calculated average of the spectra used in the 2D analysis. (A) 4.0-25.0 mN/m. (B) 25.0-40.0 mN/m.

In the low-pressure region (Fig. 7 A) several prominent cross-peaks are observed that demonstrate the ability of asynchronous 2D IR to resolve overlapped peaks. Most noticeable is the fact that the prominent -helix peak that occurs at 1653 cm¹ in the low-pressure synchronous spectrum of SP-B (Fig. 6 A) splits into two components in the low-pressure asynchronous spectrum (Fig. 7 A), one at 1656 and one at 1649 cm¹. Both of these -helix components generate asynchronous cross-peaks at wavenumber values characteristic of other secondary-structure conformations. For example, asynchronous cross-peaks are seen between 1649 cm¹ and (1640, 1628, 1690, 1676, and 1656 cm¹). The higher wavenumber -helix component at 1656 cm¹ also generates cross-peaks with 1682 and 1615 cm¹. In addition, the extended -sheet band observed at 1686 cm¹ in the low-pressure synchronous spectrum also splits into lower and higher frequency components in the asynchronous spectrum (at 1682 and 1690 cm¹, respectively). A number of additional prominent cross-peaks are seen at 1640 cm¹, characteristic of unordered structures. Cross-peaks are observed between the extended -sheet split components and other bands, including 1682 cm¹ and (1676, 1640, 1628, and 1690 cm¹) and between 1690 cm¹ and (1615, 1624, and 1637 cm¹).

The presence of two -helix components in the 2D IR correlation spectrum for SP-B is consistent with previous IR results based on curve-fitting of the amide I region in the ATR spectra of bulk phase lipid/SP-B vesicles (Vandenbussche et al., 1992). This previous study attributed two amide I band components to two separate populations of -helix in SP-B, a hydrophobic fraction associated with the lipid chains and a hydrophilic fraction parallel to the membrane surface. Presumably, the more hydrophilic fraction would encounter stronger H-bond potential with the aqueous solvent, thus slightly reducing its amide I frequency (i.e., the 1649-cm¹ peak is likely associated with the hydrophilic fraction). In Fig. 7 A, the distribution of correlation intensity between the two -helix cross-peaks indicates that the protein exists primarily in the 1656-cm¹ hydrophobic fraction at low surface pressures. This current study is the first time that two -helix fractions have been observed for SP-B in monomolecular films at the A/W interface.

The asynchronous map for DPPG/SP-B reveals that the splitting of the -helix band into two cross-peaks persists in the high-pressure region above 25 mN/m (Fig. 7 B). However, the distribution of correlation intensity is reversed at high surface pressures. Above 25 mN/m, the main correlation intensity (and the sign of this cross-peak) has shifted and is now concentrated in the 1649-cm¹ component. The asynchronous correlation of the two amide I components results in a number of cross-peaks. The 1649-cm¹ -helix band results in cross-peaks with (1640 and 1628 cm¹), whereas the 1656-cm¹ helix component shows cross-peaks with 1649, 1634, and 1620 cm¹. The extended -sheet structure at 1690 cm¹ results in a long correlation-intensity minimum with cross-peaks at 1670, 1648 and 1620 cm¹. Other prominent cross-peaks are seen between 1676 cm