2.1. FTIR

Spectra were acquired using a single-reflection attenuated total reflectance (ATR) apparatus with a diamond internal reflection element (Specac, Kent, England or ASi/SensIR, Warrington, England). The ATR apparatus was installed in either a BioRad FTS-40 or FTS-575c FTIR spectrometer (Randolph, MA, USA). Both spectrometers were equipped with a deuterated tryglycine sulfate (DTGS) detector and were purged with dry air (dew point 373 K) from a Balston (Haverhill, MA, USA) air dryer to minimize spectral contributions from atmospheric water vapour. The temperature at the surface of the diamond IRE was 295-298 K. All figures were prepared with GPLOTC (NRC, Ottawa, ON, Canada) and Microsoft Office 97 (Microsoft).

2.2. Lipid-protein standard curve

Standard solutions with different ratios of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC; Avanti Polar Lipids, Alabaster, AL, USA) to lysozyme (Sigma, Oakville, ON, Canada) were prepared in 100 mM NaCl, 10 mM Tris-HCl pH 7.5, 0.1 mM EDTA and 0.02%(w/v) NaN₃ in H₂O. The final concentration of lysozyme in each standard was 30 mg ml^-1 (BioRad protein assay, BioRad, Mississauga, ON, Canada). After thorough vortexing, a 10 µl drop of each standard solution was placed on the diamond internal reflection element and dried completely with a stream of N₂ gas (the absorption of ¹H₂O at 3500 cm^-1 was monitored to ensure complete evaporation). 512 scans at 4 wavenumber resolution were then collected and co-added. For each FTIR spectrum, the area under both the lipid ester carbonyl (1710-1765 cm^-1) and the protein amide I (1580-1710 cm^-1) bands were determined by drawing a line between the minima on either side of each band and then integrating the area between the line and the spectrum using the data-analysis software GRAMS 32 (Galactic Industries, Salem, NH, USA). The areas used in the integration procedure are denoted by the shaded regions in Fig. 1(a).

Figure 1 (a) Dry FTIR spectra of lysozyme and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) at different lipid:protein ratios. The lipid:protein ratio (mol:mol) for a 300 kDa protein is shown to the right of each spectrum. The shading denotes the areas for both lipid and protein that were integrated to calculate the standard curve. (b) Lipid-protein standard curve calculated from the spectra in (a).

Lipid:protein (mol:mol) ratios were calculated from the relative weights of lipid and protein in each solution assuming a M_r of 760.1 for POPC and a MW of 300 kDa for the protein. The latter is justified because the intensity of the amide I band is proportional to the number of peptide C=O functional groups, regardless of the molecular weight of the protein. A MW of 300 kDa is relevant to studies of the nicotinic acetylcholine receptor (nAChR; see below).

2.3. Detergent standards

All detergent standards were made in 10 mM Tris-HCl ¹H₂O pH 7.5. The n-octyl--D-glucoside (-OG), N-dodecyl-N,N-dimethyl-3-ammonio-1-propanesulfonate (Zwittergent 3-12) and cholate were from Sigma (Oakville, ON, Canada), while the n-decyl--D-maltopyranoside (C₁₀M) and octaethylene glycol monotridecyl ether (C₁₂E₈) were from Calbiochem (La Jolla, CA, USA).

For the octylglucoside standard curve, a 10 µl drop of each standard was placed on the surface of the diamond ATR. 512 scans at 4 wavenumber resolution were co-added for each spectrum, which took approximately 10.5 min using a DTGS detector. This length of time is ideal in that the drop does not dry out appreciably during spectral acquisition. The number of scans is sufficient to obtain a very high signal-to-noise ratio. Comparable spectra have also been recorded with a mercury cadmium telluride (MCT) detector, which acquires 512 scans in about 1.5 min. Solvent (10 mM Tris-HCl ¹H₂O pH 7.5) subtraction was performed with GRAMS 32 software so that the region encompassing the detergent bands approximated a flat baseline. The area under the spectra corresponding to the detergent bands (947-1187 cm^-1) was integrated (GRAMS 32) and plotted against the known octylglucoside concentration (Fig. 3).

Figure 3 (a) Selected aqueous unsubtracted FTIR spectra of various n-octyl--D-glucoside standards (top trace, 100 mM n-octyl--D-glucoside; bottom trace, 10 mM Tris-HCl pH 7.5. (b) Aqueous buffer-subtracted FTIR spectra of n-octyl--D-glucopyranoside standards (baseline-corrected between 1187 and 947 cm-1). The concentration of n-octyl--D-glucopyranoside in each standard is (from top to bottom): 200, 100, 50, 25, 12.5 and 6.25 mM. (c) n-Octyl--D-glucopyranoside standard curve calculated using the spectra in (b).

2.4. Protein secondary structure

10 µl of 10, 30 and 60 mg ml^-1 lysozyme (Sigma, Oakville, ON, Canada) in 100 mM NaCl, 10 mM Tris pH 7.5, 0.1 mM EDTA and 0.02%(w/v) NaN₃ in ¹H₂O was placed on the diamond ATR surface and 512 scans were acquired at 4 wavenumber resolution. For the denatured sample, lysozyme was incubated at 373 K for 20 min prior to spectral acquisition. Aqueous buffer and residual H₂O vapour was subtracted using GRAMS 32 software (Reid et al., 1996).

2.5. Ultrafiltration experiment

The nicotinic acetylcholine receptor was affinity purified as described elsewhere (daCosta et al., 2002), except that it was eluted in cholate elution buffer. The eluate was pooled and concentrated at 1000g using a 100 kDa molecular-weight cutoff (MWCO) concentrator (Millipore, Ottawa, ON, Canada). Aliquots were removed at 5 min intervals and the nAChR concentration was determined by the BCA protein assay (Pierce). Cholate concentration was determined using FTIR by integrating cholate spectral features (992-1135 cm^-1) and comparison with a cholate standard curve (not shown).

3.1. FTIR data-acquisition accessories

Our primary goal was to establish a method for measuring both lipid:protein ratios and detergent concentrations in solutions containing detergent-solubilized integral membrane proteins. FTIR spectroscopy, which is sensitive to bond vibrational frequencies, could be used to measure both parameters (Goormaghtigh et al., 1990; Pistorius et al., 1994). FTIR can also be used to investigate protein secondary structure and could thus shed light on how lipid:protein ratios and detergent concentrations influence protein structural stability (He et al., 1991). The problem, however, is that the accessories (either transmission or ATR) commonly used to acquire spectra from biological samples are not amenable to rapid screening of many small aliquots. We required a sampling accessory that allows rapid spectral acquisition from small aliquots at all stages of our sample preparation.

We tested the utility of a diamond ATR microsampling accessory for acquiring spectra from detergent-solubilized membrane-protein samples (Figs. 1, 2, 3 and 4). The advantages of this ATR accessory are the ease of sample preparation and the small sample requirements. 5-10 µl of solution are simply placed on the surface of the ATR accessory and spectra are acquired. The total time for sample preparation and spectral acquisition can be less than 1 min, although longer times were used here to increase the number of scans and thus obtain a higher signal-to-noise ratio. In addition, the diamond ATR crystal is easily cleaned for repeated spectral analyses.

3.2. Lipid:protein ratios

The utility of the ATR accessory for assessing lipid:protein ratios is demonstrated in Fig. 1. Spectra recorded from dried solutions (see §2) containing a constant concentration of the protein lysozyme and increasing levels of the lipid POPC exhibit two main infrared vibrations in the 1760-1550 cm^-1 region (Fig. 1a). The broad amide I band between 1600 and 1700 cm^-1 arises primarily from the C=O stretching vibration of the polypeptide backbone, whereas the relatively narrow band between 1700 and 1760 cm^-1 arises from the C=O stretching vibration of lipid ester carbonyl groups (Jackson & Mantsch, 1995; Mendelsohn & Mantsch, 1986). The relative area of the lipid C=O and amide I band is directly related to the lipid:protein ratio as shown in the standard curve (Fig. 1b). Note that the molar lipid:protein ratio has been calculated for a 300 kDa protein (i.e. nAChR). As seen in both Figs. 1(a) and 1(b), the technique can accurately detect lipid:protein ratios down to roughly five molecules of lipid per molecule of 300 kDa protein.

3.3. Detergent concentration

The applicability of the diamond ATR accessory for measuring detergent concentrations in solubilized membrane-protein samples is illustrated in Figs. 2 and 3. Five representative spectra of detergents commonly used in membrane-protein purification and crystallization are compared with the absorption spectra of DPPC and lysozyme in Figs. 2(a)-2(e). Each detergent exhibits spectral features in the 1200-900 cm^-1 region that do not overlap with the absorption bands of proteins (Fig. 2g). Although lipid does absorb in the 1200-900 cm^-1 region (Fig. 2f), the spectral contributions of lipids are negligible at the lipid:protein ratios normally found in our detergent solubilized membrane-protein samples (usually less than 30:1 mol:mol lipid:protein). In cases with very high lipid and very low detergent, the contributions of lipid can be scaled and subtracted from the spectra in order to quantify the intensities of the detergent bands (not shown).

Figure 2 Comparison of 1H2O buffer-subtracted FTIR spectra of various detergents with spectra of lipids and protein. (a) n-Octyl--D-glucopyranoside (-OG), (b) n-decyl--D-maltopyranoside (C10M), (c) octaethylene glycol monotridecyl ether (C12E8), (d) Zwittergent 3-12, (e) cholic acid. Spectrum (f) is a dry spectrum of the phospholipid 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) and spectrum (g) is an 1H2O-subtracted spectrum of the protein lysozyme.

Spectra recorded from solutions with increasing concentrations of the detergent n-octyl--D-glucoside are presented in Figs. 3(a) and 3(b). Fig. 3(a) shows the intensity of the n-octyl--D-glucoside peaks relative to the solvent ¹H₂O bending vibration (1640 cm^-1) before the buffer is subtracted. Fig. 3(b) shows the buffer-subtracted and baseline-corrected spectra used to calculate the standard curve (Fig. 3c). The area of the detergent vibrations in the 1200-950 cm^-1 region correlates in a linear fashion with the concentration of the detergent (Fig. 3c), showing that the infrared technique is a viable approach for measuring detergent concentrations in aqueous solutions. Note that accurate spectra of this and other detergents have been recorded to well below their critical micellar concentrations (for octylglucoside this is 19-25 mM, but for dodecylmaltoside the CMC is only 0.18 mM; Le Maire et al., 2000). In fact, accurate infrared spectra have been obtained at detergent concentrations as low as 0.08 mM (data not shown). We have found that our concentrated detergent-solubilized membrane-protein samples often exhibit detergent concentrations in the range 10-50 mM, which is a highly accurate concentration range for spectral analysis.

It is also important to note that both the shape of the detergent bands and the linearity of their absorption intensities with increasing detergent concentration are unaffected by the CMC. This shows that association of n-octyl--D-glucoside into micelles does not greatly influence either the shape or intensity of the infrared absorption bands. This result is expected as the absorption bands result from fundamental infrared vibrations, the intensities of which are not dramatically influenced by environmental factors (for example, the infrared absorption bands of acetylcholine are not dramatically changed upon binding to the nAChR even though acetylcholine binds tightly with a nanomolar affinity (Baenziger et al., 1993; other unpublished observations). In addition, the head-group vibrations should have similar hydration states in both the monomeric and micellar forms.

For similar reasons, detergent head-group vibrational bands should not be influenced strongly by the formation of protein-detergent complexes. In support of this assertion, it has been shown that the rate of removal of detergent by dialysis in solubilized N⁺/K⁺-ATPase samples is equivalent whether monitored using FTIR or radiolabelling techniques (Pistorius et al., 1994).

3.4. Protein secondary structure

The utility of the diamond ATR accessory for monitoring protein secondary structure was assessed by recording spectra of lysozyme in ¹H₂O. The main difficulty in such an analysis is that the amide I band, which is the vibration whose shape is characteristic of protein secondary structure, overlaps with a strong ¹H₂O vibration in the 1600-1700 cm^-1 region (see Fig. 4a). Analysis of the protein amide I band shape thus requires subtraction of the overlapping ¹H₂O vibration.

Figure 4 (a) Unsubtracted aqueous FTIR spectra of lysozyme. (b) Spectra of lysozyme. Spectrum (i), dry lysozyme. Spectra (ii), (iii) and (iv) are 1H2O-subtracted spectra of 10, 30 and 60 mg ml-1 lysozyme. Spectrum (v) is an 1H2O-subtracted spectrum of thermally denatured lysozyme (10 mg ml-1). The spectra in (b) have been scaled relative to their amide I band to facilitate band-shape comparison.

Fig. 4 shows spectra recorded from solutions containing 10, 30 and 60 mg ml^-1 concentrations of lysozyme (spectra ii, iii and iv). In each case, the amide I band shape achieved after subtraction of the overlapping ¹H₂O vibration (Fig. 4b) is similar to that obtained in a spectrum of lysozyme dried on the ATR accessory (Fig. 4b, spectrum i). Specifically, each amide I band exhibits an intense maximum near 1655 cm^-1, characteristic of -helical secondary structure, and less intense shoulders near 1630 and 1670 cm^-1, characteristic of -sheet (Jackson & Mantsch, 1995). As expected, all four spectra are indicative of a predominantly -helical protein with substantial -sheet.

In contrast, the amide I band shapes differ substantially from that observed in a spectrum recorded from lysozyme after thermal denaturation (Fig. 4, spectrum v). Upon denaturation, the amide I band exhibits an increase in intensity at both 1620 and 1680 cm^-1, features that are observed in the spectra of numerous denatured proteins (Methot & Baenziger, 1998; Young et al., 1995). Similar, although less pronounced, changes in amide I band shape have also been detected in spectra of lysozyme and the nicotinic acetylcholine receptor (nAChR) recorded in the presence of high concentrations of detergent (data not shown), although subtle changes in the amide I band shape can be difficult to detect owing to the subjectivity involved in the subtraction of the overlapping water vibration. We have found that FTIR can provide a qualitative estimate of the structural stability of a membrane protein in the presence of detergent. A more definitive analysis can be performed by recording spectra in ²H₂O buffer, as ²H₂O does not exhibit vibrations overlapping the 1600-1700 cm^-1 region (He et al., 1991). In addition, spectra can be obtained after drying the sample on the ATR surface to eliminate the ¹H₂O vibrations in the spectra. It should be noted, however, that drying concentrates the detergent in the sample, which could result in protein denaturation.

3.5. Changes in detergent levels during sample manipulations

To demonstrate the utility of the ATR technique for monitoring detergent levels during the preparation of a membrane-protein stock solution that is suitable for crystallization trials, we examined the changes in detergent levels that occur during concentration of affinity-purified nAChR. The nAChR was first solubilized from its native membranes and affinity purified in the presence of the detergent cholate. The cholate-solubilized nAChR was then concentrated by ultrafiltration in an Amicon protein concentrator (100 kDa MWCO).

Protein assays show that there is a very slow increase in protein concentration during the first 35 min of the spin, but thereafter the protein concentration increases at a dramatic rate. In contrast, the detergent concentration increases only slightly during the first 50 min of the spin (Fig. 5). We believe that the lag time between the increases in detergent and protein concentration may be because of the fact that excess protein-free detergent micelles exist in solution and initially spin through the 100 kDa MWCO Amicon filters. When the detergent is no longer in excess, it concentrates at the same rate as the protein because they are tightly associated.

Figure 5 A comparison of how detergent and protein concentration change during the concentration of cholate-solubilized nAChR by ultrafiltration. Solid circles correspond to [nAChR]. Open circles correspond to [cholate].

Using our FTIR techniques, we have found that the initial lipid:protein ratio in our nAChR samples can influence the degree to which the detergent concentrates during protein concentration. This may occur because excess lipid forms protein free lipid-detergent mixed micelles that may be too large to spin through the 100 kDa MWCO filters.

In addition, properties of the detergent (i.e. micellar size) as well as the size of the filter pores may influence how much the detergent concentrates. With ultrafiltration devices that have a MWCO that is large enough to allow the free passage of detergent micelles, the concentration of detergent monomers and micelles stays the same over the course of a typical concentration experiment, while the ratio of micelles and monomers to protein-detergent complexes (PDCs) decreases (Fig. 6). Therefore, it is best to minimize the initial amount of excess detergent in a solubilized membrane-protein sample before ultrafiltration.

Figure 6 Conceptual depiction of how detergent:protein ratios vary with protein concentration during an ultrafiltration experiment. At the start of the experiment, there is a low protein concentration and excess detergent is present as protein-free micelles. If the molecular-weight cutoff of the protein concentrator is large, the micelles spin through the filter of the concentrator and do not concentrate during the experiment. Although the absolute concentration of micelles and monomers does not change, their molar ratio relative to protein-detergent complexes decreases. Therefore, the final ratio of protein-free detergent micelles and monomers to protein-detergent complexes depends on both the starting concentration of the detergent and the final concentration of protein. Initial detergent concentrations close to the critical micellar concentration of the detergent are likely to be ideal.