Conformation of the synaptobrevin transmembrane domain

CD Spectroscopy.

Previous studies have shown that the cytoplasmic domain of synaptobrevin is unstructured (9, 11, 12). The synaptobrevin transmembrane domain was proposed to contain a mixture of α-helix and β-sheet, based on IR and CD studies of a 22-aa peptide containing residues 97–112 of the synaptobrevin transmembrane domain flanked by tyrosine and lysine residues (34). To examine the effect of the complete transmembrane domain on the structure of the full-length protein, we analyzed our three synaptobrevin constructs using far-UV CD spectroscopy (Fig. 1 A and B). Spectra were collected in 50 mM β-octyl glucoside (βOG) at identical protein concentrations. Data are reported as molar ellipticity ([θ]_ME) rather than mean residue ellipticity because of the disparate lengths of the constructs. The spectrum of the full-length protein showed negative ellipticity at 222 nm and positive ellipticity at 190 nm, indicating some α-helix content (35). The spectrum of the cytoplasmic domain in detergent showed slight negative ellipticity at 222 nm, indicating some α-helical content in agreement with a previous report (11). The transmembrane peptide showed a strong double minima at 208 and 222 nm, both of which are characteristic of α-helical conformation.

The CD spectra were analyzed to provide information on secondary structure content using cdpro, which has been shown to be successful in estimating the secondary structure of membrane proteins (36, 37). The cytosolic domain contained only 9%, or eight residues, in an α-helical conformation. The 116-residue, full-length protein contained 23% α-helix, which corresponds to 26 residues. The 43-residue peptide contained 63% α-helix, which also corresponds to 26 residues.

The sensitivity of the secondary structure to detergents was tested by comparing the spectra of the transmembrane peptide in anionic (SDS), bile salt (cholate), nonionic (βOG), and lipid-like (dodecylphosphocholine) detergents (Fig. 1C). The mean residue ellipticity, [θ]_MRE, of the synaptobrevin transmembrane construct was fairly insensitive to the nature and concentration of the detergent used. The secondary structure of the transmembrane construct was also insensitive to pH over a range of 4–9 (data not shown). To determine whether the peptide could be forced into helical conformation, the CD spectrum of the peptide was taken in the presence of increasing amounts of 2–2-2-trifluoroethanol (TFE), which promotes α-helix formation (Fig. 1D). The α-helical content of the peptide increased to 72% in 5% TFE, 77% in 10% TFE, and 80% in 30% TFE. Thus, even in 30% TFE, an estimated three to four residues remained unstructured.

IR Spectroscopy in Lipid Bilayers.

To examine the conformation of the synaptobrevin transmembrane construct in phospholipid bilayers, we used ATR-FTIR. This technique relies on the correlation between the amide I absorption frequency (predominantly C=O stretching) and the protein secondary structure (26, 27). Four synaptobrevin transmembrane constructs, each containing ¹³C at the carbonyl carbon of a different leucine residue, were individually reconstituted into 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) liposomes at a protein-to-lipid ratio of 1:100. The ATR-FTIR absorption spectra of all four constructs show a predominant absorption centered at 1,655 cm⁻¹, indicative of α-helical conformation (28) (Fig. 2A). The broad amide I absorbance could not be fit to a single peak, so automated peak picking combined with nonlinear least-squares fitting was used to resolve the overlapping peaks in the spectra. All four samples were fit to a four-peak model, resulting in goodness of fit values (X²) ranging from 10⁻⁵ to 10⁻⁷. Inclusion of additional peaks did not decrease X². The four peaks are the 1,655 cm⁻¹ band, the corresponding isotopically shifted band, and two absorbances centered at 1,675 cm⁻¹ and 1,635 cm⁻¹ that correspond to extended peptide conformations. A simple comparison of the peak areas averaged over the four peptides predicts 69 ± 3% α-helix (1,655 cm⁻¹), 17 ± 4% anti-parallel β-sheet/β-turn (1,675 cm⁻¹), and 10 ± 3% β-sheet/extended (1,635 cm⁻¹). This estimate is qualitative because the molar absorptivity is thought to differ for different secondary structure types (38). The CD spectra of the peptide indicated the presence of unstructured residues (17% for the peptide), which was not explicitly fit in the IR data analysis. This broad absorbance, 1640–1660 cm⁻¹ (28, 39), overlaps with the helical peak and could therefore contribute to the estimated area of this peak. However, because our data were well fit by the four-peak model described above, the contribution from the unstructured residues cannot be large.

The presence of ¹³C shifts the amide I absorbance frequency by ≈40 cm⁻¹ for the labeled peptide bond (29, 31, 32). Shifted satellite peaks are indeed present in the spectra of the four constructs (Figs. 2 A and B). For the construct labeled at L84, this satellite peak is centered at 1,629 cm⁻¹, a 40-cm⁻¹ shift from a nonhelical region of the spectrum. The other three constructs (L93, L99, and L107) each show a satellite peak centered near 1,615 cm⁻¹, indicative of a 40-cm⁻¹ shift from the α-helical region of the spectrum. Thus, the transmembrane α-helix continues into the polybasic, juxtamembrane region, but this helix does not extend into the cytoplasmic domain in agreement with previous results (22).

We next examined the orientation of the α-helical portion of the transmembrane construct relative to the bilayer using polarized ATR-FTIR. Samples were reconstituted in 1,2 dimyristoyl-sn-glycero-3-phosphocholine (DMPC), which forms a gel-phase bilayer at room temperature. DMPC is commonly used in ATR-FTIR because increased acyl-chain order results in clearer assessment of bilayer orientation with respect to the crystal surface. Representative polarized ATR-FTIR spectra of the constructs illustrating the phospholipid acyl-chain and protein amide I regions of the spectrum are shown in Fig. 3.

The lipid antisymmetric methylene and symmetric methylene stretches were found to be centered at 2,917 cm⁻¹ and 2,849 cm⁻¹, respectively, which is consistent with the deposited bilayer being in the gel phase (40) (Fig. 3A). The dichroic ratio, R^ATR, of the symmetric methylene stretch, determined from the ratio of absorbance at 0° and 90° polarizations, was 1.1, which results in a calculated lipid order parameter, S_L of 0.7. This value is consistent with lipids in the gel phase forming a well ordered deposited bilayer parallel to the crystal surface (28, 41).

The spectra of all four peptides in DMPC were also dominated by the absorption at 1655 cm⁻¹ (Fig. 3B). DMPC samples contained more residual water than samples prepared from POPC, which widened the amide peaks and obscured the ¹³C-shifted peaks. Estimation of secondary structure by direct comparison of peak area suggests slightly less α-helix in DMPC than POPC, 58 ± 9% at 1,655 cm⁻¹, and similar β-sheet/extended conformation, 12 ± 8% at 1,635 cm⁻¹; whereas the amount of antiparallel β-sheet/β-turn has increased to 29 ± 2% at 1,675 cm⁻¹. These values could indicate the presence of some aggregated peptide.

Only the component of the amide I absorption at 1,655 cm⁻¹ was analyzed for orientation because the nonhelical regions could not be assumed to have a continuous conformation. The R^ATR of the transmembrane domain was 2.86 ± 0.16 in DMPC. Assuming that the angle between the helix axis and the amide I vibration mode is 39° (42, 43), we calculated a mean value of 0.49 within a range of 0.42–0.56 for the peptide order parameter, S_P. If one assumes a continuous α-helix and that the low value for S_P results exclusively from a tilt away from the bilayer normal rather than sample disorder, the insertion angle of the transmembrane domain can be derived from S_P (Eq. 2). This calculation gives a mean tilt angle of 36° for the synaptobrevin peptide, with a range of 33–39° considering the error in R^ATR. It is possible that sample disorder (i.e., defects in the deposited bilayer) contributes to the apparent peptide order parameter, resulting in overestimation of the tilt angle. Sample disorder can be estimated from S_L, which would approach 1 if the acyl-chains were perfectly ordered, but perfect acyl-chain order would not be expected for a deposited DMPC bilayer at room temperature. Reported S_L values for DMPC in the literature are in the range of 0.5–0.8 (28), which is similar to the observed S_L of 0.7. Explicit analysis of sample disorder in similar DMPC/peptide samples concluded that sample disorder was not significant (44), but this analysis may not be directly comparable. In any case, our data would not be compatible with a helix perpendicular to the membrane.

Because the gel-phase bilayers formed by DMPC are not likely representative of biological membranes, we also characterized peptide orientation in fluid POPC bilayers. Representative polarized ATR-FTIR spectra of the peptide in POPC showing the phospholipid acyl-chain and protein amide I regions of the spectrum are shown in Fig. 3. The lipid antisymmetric methylene and symmetric methylene stretches of POPC were centered at 2,922 cm⁻¹ and 2,852 cm⁻¹, respectively (Fig. 3C). The positions of these peaks are consistent with the acyl region of the deposited bilayer being in the fluid phase (40), as is expected for POPC at room temperature. The R^ATR for the symmetric methylene stretch was 1.4, which results in a value of 0.4 for S_L. This S_L value would be considered low for DMPC but is consistent with liquid–crystalline lipids having an average bilayer orientation perpendicular to the supporting crystal and is in good agreement with S_L values reported previously for gel-phase lipids taken above the phase transition temperature (41) and previous studies of POPC (45).

Analysis of the peptide orientation was again limited to the transmembrane α-helix, which was assumed to be continuous (Fig. 3D). We found an average R^ATR for the transmembrane domain of 1.99 ± 0.04. This gives a value of −0.009 for the S_P, which results in a maximum tilt angle of 55° for the peptide relative to the bilayer, assuming that the membranes are perfectly oriented and that the transmembrane domain is a continuous helix. It should be noted that, mathematically, an S_P value of zero would also be compatible with an isotropic, random distribution of the peptide orientation relative to the crystal surface. However, given that the bilayers have a mean orientation parallel to the surface and that the hydrophobic peptide is reconstituted in the bilayer, an isotropic distribution is unlikely. The transmembrane domain contains a glycine at residue 100, so it is possible that the transmembrane helix may not be continuous. The ¹³C-labeled L99 residue, immediately adjacent to G100, and the other transmembrane leucine residues showed strong IR absorption peaks at 1,615 cm⁻¹ (Fig. 2A), indicating a helical conformation. Thus, any helical break would be quite limited. Nonetheless, two helical segments at different orientations could appear as a random distribution of orientation.

Constructs.

Rat synaptobrevin 2 (residues 1–116) and the isolated cytoplasmic domain (1–96) were expressed by using pet28a (Novagen) in Escherichia coli strain BL21 (DE3) (Novagen). Synthetic peptides corresponding to residues 74–116 of rat synaptobrevin 2 were synthesized by John Flory (W. M. Keck Protein Chemistry Facility, Yale University) using standard N-tert-butoxycarbonyl chemistry. Individual leucine residues were selectively labeled through the incorporation of l-[1-¹³C]leucine (Cambridge Isotope Laboratories, Andover, MA) at indicated positions. The four peptides behaved identically in purification, electrophoresis, CD, and UV/visual light spectroscopy (data not shown).

Purification.

Purification of full-length synaptobrevin was as described in ref. 64. Briefly, cells were resuspended in 50 mM sodium phosphate/0.5 M NaCl, pH 6.5, with 10 mM 2-mercaptoethanol/20 mM imidazole/Complete EDTA-free tablets (Boeringer Mannheim) added as per manufacturer’s instructions. Thesit was added from a 25% stock to a final concentration of 2%. The supernatant was applied to Ni-NTA agarose (Qiagen, Valencia, CA) and washed with 0.2% Thesit and 40 mM imidazole. For exchange into other detergents, washes of 20 column volumes containing the appropriate detergent were used. Elution was carried out with 250 mM imidazole. Fractions were assessed for purity by SDS/PAGE. Immunoblotting for synaptobrevin was performed from 20% Phast gels (GE Healthcare) according to standard methods with mouse monoclonal antibody clone CL-69.1 (Synaptic Systems, Goettingen, Germany) and developed with ECL (GE Healthcare). The cytoplasmic domain was purified as described in refs. 9 and 33.

The transmembrane construct was purified by reverse-phase HPLC. The aqueous phase was 0.1% (wt/vol) trifluoroacetic acid, and the organic phase was 40% isopropanol, 60% acetonitrile, and 0.1% trifluoroacetic acid. Crude peptides were resuspended in TFE, diluted with 10% organic phase, and loaded onto a C18 semipreparative column (Vydac, Hesperia, CA) equilibrated in 10% organic phase. The peptide was eluted with a linear gradient from 65–85% organic phase. Peptide fractions were assessed by MALDI-TOF MS. Pure fractions were lyophilized and resuspended in TFE. Aqueous samples were prepared by adding transmembrane construct in TFE to detergent solution in water followed by lyophilization and resuspension in the desired buffer.

Reconstitution.

Twelve milligrams of POPC or DMPC in chloroform (Avanti Polar Lipids) was dried to a thin film under argon followed by redrying from pentane to remove any residual chloroform. The lipid was resuspended in 1 ml of deionized water using three rounds of bath sonication (Laboratory Supplies, Hicksville, NY) interspersed with freeze/thaw cycles in liquid nitrogen. The resuspended lipid was added to a lyophilized peptide/detergent mixture resulting in a final peptide lipid ratio of 1:100 and 5% βOG. This synaptobrevin concentration is higher than that reported for synaptic vesicles but does not disrupt membranes (16). This solution was then extensively dialyzed against water containing BioBead SM2 (Bio-Rad) to remove residual detergent. Final dialysis was carried out in 5 mM sodium phosphate/10 mM NaCl at pH 7. Insoluble material was removed by centrifugation for 30 min at 16,000 × g.

CD Spectroscopy.

Far-UV CD spectra were collected on a model 62DS CD spectrometer (Aviv Associates, Lakewood NJ) with 0.1-mm quartz cuvettes at 25°C. All CD spectra represent the average of five scans collected at 1-nm intervals using a 5-s integration time. CD spectra were collected in 10 mM sodium phosphate/50 mM NaCl/1 mM 2-mercaptoethanol at the pH and detergent conditions indicated. Detergents were from Anatrace (Maumee, OH). For determination of molar ellipticity, protein concentrations were determined from the absorbance at 280 nm using a calculated extinction coefficient because all constructs contain the same two tryptophan residues. Spectra were fit with all three methods contained in cdpro, which all gave similar results indicating that the fitting was successful (37). cdpro estimates both regular and distorted α-helix, which were summed to give an estimation of the total amount of α-helix in each construct.

IR Spectroscopy.

For each peptide, deposited bilayers were formed on germanium by spontaneous condensation from reconstituted proteoliposomes in aqueous solution. Reconstituted samples were dried onto the surface of a trapezoidal germanium crystal (dimensions: length, 50 mm; width, 20 mm; depth, 2 mm) using dry air. Spectra were collected at room temperature on a Nicolet Protégé 420 spectrometer equipped with a cooled mercury–cadmium–telluride (MCT/A) detector and extensively purged with nitrogen. The spectrometer was equipped with a 25-reflection variable-angle ATR accessory. For polarized IR spectroscopy, a 0.25-μm wire grid polarizer was used (Grasby Specac, Kent, U.K.). Interferograms (1,000) with a spectral resolution of 1 cm⁻¹ were averaged for each spectrum. Interferograms were processed with omnic 3.1 (Nicolet) by using blank subtraction, 1-level zero-fitting and Happ–Genzel apodization. Overlapping peaks in the spectra were resolved by using the peak-fitting routines within grams/32 (Galactic Industries). Automated peak picking limited to widths of >13 cm⁻¹ was used to identify starting positions for iterative fitting. Fitting was done to mixed Gaussian/Lorentzian line shapes and constrained to positive values within 5 cm⁻¹ of the initial identified peak position. The spectrum of each peptide was recorded three times using separate reconstitutions for each trial. No variation was seen between trials or between different peptides, excluding the wave number of the ¹³C absorbance. The value of R^ATR was obtained from the ratio of the areas of the parallel to perpendicular peaks from the fit described above. Assuming the sample to be thicker than the penetration depth of the evanescent wave (thick film), R^ATR was converted to an order parameter using the following relationship (65):

The values for the electric field components of the evanescent wave were E_x = 1.398, E_y = 1.516, and E_z = 1.625 (65). The angle α of the vibrational dipole relative to the molecular axis was 39° for the amide I (42, 43) and 90° for the symmetric methylene (28). The order parameter was used to derive a tilt angle θ by inverting the following relationship (65):