Specular neutron reflectivity and the structure of artificial protein maquettes vectorially oriented at interfaces

Artificial peptides can be designed to possess a variety of functionalities. If these peptides can be ordered in an ensemble, the functionality can impart macroscopic material properties to the ensemble. Neutron reflectivity is shown to be an effective probe of the intramolecular structures of such peptides vectorially oriented at an interface, key to ensuring that the designed molecular structures translate into the desired material properties of the interface. A model-independent method is utilized to analyze the neutron reflectivity from an alkylated, di-$\alpha$-helical peptide, containing perdeuterated leucine residues at one or two pre-selected positions, in mixed Langmuir monolayers with a phospholipid. The results presented here are more definitive than prior work employing x-ray reflectivity. They show explicitly that the di-helical peptide retains its designed $\alpha$-helical secondary structure at the interface, when oriented perpendicular to the interface at high surface pressure, with the helices projecting into the aqueous subphase without penetrating the layer of phospholipid headgroups.

Figure 1

Schematic representation of the di-$\alpha$-helical peptide BBC16, synthesized as a mostly $\alpha$-helical 31-mer with a flexible $\mathrm{Gl}{\mathrm{y}}_3$ linker ending in an $N$-terminal Cys with a palmitoyl $\left({\mathrm{C}}_{16}\right)$ chain attached and then dimerized in air under basic conditions to form BBC16. Heme dissolved in dimethyl sulfoxide added to the peptide can bind at bis-His ligation sites incorporated in the design. When the peptide is prepared with ${}^2\mathrm{H}$-labeled Leu-${\mathrm{d}}_{10}$ at selected sites in the sequence, the contrast of the individual residue relative to the rest of the helix is enhanced. Dimensions indicate the average diameter and rise/residue of an $\alpha$ helix.

Figure 2

At low surface pressure (a), the helices of BBC16 lie within the plane of the air/water interface and all residues are at the same position $z$ relative to the interface. At high surface pressure (b), the helices orient with the helical axis approximately perpendicular to the interface and each residue is mapped onto a unique position $z$ relative to the interface. Neutron reflectivity can then be used to determine the position of ${}^2\mathrm{H}$-labeled residues, such as Leu-${\mathrm{d}}_{10}$.

Figure 3

(Color online) Model SLD profile structures for Langmuir monolayers of BBC16 based on the electron density profile structures obtained from x-ray reflectivity data. The ${\mathrm{H}}_2\mathrm{O}$ subphase extends to the right for $z>0\mathrm{\AA }$; air extends from $z<-65\mathrm{\AA }$ (consistent with Fig. 4; other figures use a different convention). The perdeuterated hydrocarbon chains contribute to a region of high SLD $(-65<z<-45\mathrm{\AA })$ while the peptide has uniform, relatively low SLD, except at the label site, where Leucine is replaced by Leu-${\mathrm{d}}_{10}$ (here in the 14 position of the sequence). A series of model profile structures in which the label site was moved in 3 Å increments (every other sequence position) was used to generate a series of simulated Fresnel-normalized reflectivity curves, $R_{\mathit{\text{label}}}\left(Q_z\right)∕R_F\left(Q_z\right)$ (not shown). The differences between these simulated reflectivities and the simulated reflectivity generated from the unlabeled model profile, $R_{\mathit{\text{nolabel}}}\left(Q_z\right)∕R_F\left(Q_z\right)$, $\mathrm{\Delta }R\left(Q_z\right)∕R_F\left(Q_z\right)=R_{\mathit{\text{label}}}\left(Q_z\right)∕R_F\left(Q_z\right)-R_{\mathit{\text{nolabel}}}\left(Q_z\right)∕R_F\left(Q_z\right)$, (b), exhibit a sinusoidal variation over the experimentally accessible range of momentum transfer and its frequency is highly sensitive to the position of the deuterated residue within the monolayer profile structure. Such selectively deuterated minus fully hydrogenated difference data are routinely employed in structure analysis via neutron scattering, as they exploit the difference in the deuterium versus hydrogen atomic scattering factors for neutrons to the maximal extent (see Ref. 14). (c) Experimental neutron reflectivity data $R\left(Q_z\right)$ collected from Langmuir monolayers spread from mixtures of the apo form of the synthetic peptide BBC16 and the phospholipids DLPE on a ${\mathrm{H}}_2\mathrm{O}$ subphase at high surface pressure with the hydrocarbon chains perdeuterated and the peptide either undeuterated (black), with Leucine 09 deuterated (green), with Leucine 14 deuterated (blue), or with Leucine 28 deuterated (red). (d) The same data after both Fresnel and integral normalization, norm $R\left(Q_z\right)∕R_F\left(Q_z\right)$. The latter integral normalization of the Fresnel-normalized reflectivity data places the data on the same, but arbitrary scale (see Methods section). For clarity, error bars have been omitted, but this noise level, which increases with larger $Q_z$, is readily apparent from the point-to-point fluctuations in the data along $Q_z$ as shown [see, for example, Fig. 5b]. (e) Difference data computed from the data in (d) between the experimental data for the peptide with a selected deuterated Leucine residue and that for the fully protonated peptide, norm $\mathrm{\Delta }R\left(Q_z\right)∕R_F\left(Q_z\right)$, again on the same arbitrary scale. The sensitivity of these data to the sequence position of the selected deuterated Leucine residue is clearly evident. The “sinusoidal” variation in the difference data arises from the position of the single deuterated residue in the monolayer profile structure, as predicted by the simulated data shown in Fig. 3b, irrespective of the different ordinate scales.

Figure 4

Real-space analysis for fully protonated BBC16 peptide (all-$\mathrm{H}$; left-column) and BBC16 peptide containing a perdeuterated leucine at sequence position 14 (Leu-${\mathrm{d}}_{10}$-14; right column). Top row left: Experimental gradient SLD profile, norm $d\rho \left(z\right)∕dz$, from box refinement (dotted) vs the best model 4-Gaussian model from nonlinear fitting (solid) for the all-$\mathrm{H}$ case. Top row right: Experimental gradient SLD profile, norm $d\rho \left(z\right)∕dz$, from box refinement (dotted) vs the same best 4-Gaussian model of the gradient SLD profile for the all-$\mathrm{H}$ case plus the additional best derivative of an arbitrary Gaussian representing the deuterium labeled residue from nonlinear fitting (solid). Second row: The corresponding residuals from the nonlinear fitting. Third row: numerical integral of the experimental gradient SLD profiles (dotted) vs the analytic integral of the corresponding best fitting model representations of the gradient SLD profiles, namely the SLD profiles themselves, norm $\rho \left(z\right)$, containing four error functions in the all-$\mathrm{H}$ case and the same four error functions plus a Gaussian function in the Leu-${\mathrm{d}}_{10}\text{−}14$ case. The perdeuterated hydrocarbon chains are localized within the $-60<z<-40\mathrm{\AA }$ region in these SLD profiles. The small differences in the SLD profiles between the all-$\mathrm{H}$ and Leu-${\mathrm{d}}_{10}$-14 cases, readily apparent in the region occupied by the peptide, namely $-40<z<0\mathrm{\AA }$, arise exclusively from the position of the perdeuterated leucine-14 residue in the SLD profile represented by a Gaussian function whose position is thereby determined by the nonlinear fitting to an accuracy of ±0.5 Å [compare with the model profile structure on an absolute scale in Fig 3a.] [Simple models, such as that shown in Fig. 3a were useful in planning the experiment, but required modification in several respects before they could be used to approximately account for the experimental neutron reflectivity. The water content of the monolayer decreased the contrast between the peptide and the subphase, the disorder of the perdeuterated hydrocarbon chains reduced their SLD, and the experimental resolution made features less prominent. Even with these modifications, the model-independent box-refinement method for obtaining the profile gradient, and its analytic integration providing the profile itself (third row of Fig. 4) did a much better job of explaining the observed data (as compared with Fig. 7(c), Ref. 14).] Fourth row: The gradient of these SLD profiles in the top row clearly account for their corresponding Fresnel and integral normalized reflectivity data, norm $R\left(Q_z\right)∕R_F\left(Q_z\right)$, to within the counting statistics for $Q_z∕2\mathrm{\pi }<0.025{\mathrm{\AA }}^{-1}$. The experimental data show the counting statistics in the point-to-point fluctuations in the data along $Q_z$ as shown and the obvious effect of the perdeuterated leucine at sequence position −14, while the smooth curves simply demonstrate the convergence of the box refinement to the experimental data. Note that all of the ordinate scales in this figure are arbitrary, as they are linked to the integral normalization of the experimental Fresnel-normalized reflectivity as described in the text, and so indicated via the prefix “norm.” Nevertheless, these arbitrary scales for norm $R\left(Q_z\right)∕R_F\left(Q_z\right)$, and therefore also norm $d\rho \left(z\right)∕dz$ and norm $\rho \left(z\right)$, are the same for the fully hydrogenated vs the selectively deuterated cases allowing their direct comparison.

Figure 5

(Color online) (a) Specular neutron reflectivity data collected from Langmuir monolayers composed of 2:1 D-DLPE:BBC16, in which the BBC16 had perdeuterated chains and was otherwise either unlabeled (엯), or doubly labeled with ${\mathrm{D}}_{10}$-Leucine either at positions 9 and 21 (red ▵), or at positions 14 and 28 (blue ▿), all on a ${\mathrm{D}}_2\mathrm{O}$ subphase (b). The same data after Fresnel-normalization (symbols) as well as the best fits obtained from box refinement (curves) (bottom). These data are on an absolute scale, as described in the text. Note that the point-to-point variations in the data, e.g., as for $R\left(Q_z\right)∕R_F\left(Q_z\right)$ in Fig. 3b, are the same magnitude as the error bars of the counting statistics, because the data were collected at intervals of $Q_z$ much smaller than the widths of any of the maxima/minima in the data arising from features in the gradient SLD profile with maximal separation, the gradient SLD profile being of finite extent [8].

Figure 6

(Color online) Comparison of SLD profile structures on an absolute scale from the doubly labeled peptide experiment. Solid black (dashed red): analytically integrated SLD profile from the unlabeled peptide (peptide labeled at Leu21,Leu09) with perdeuterated chains on a ${\mathrm{D}}_2\mathrm{O}$ subphase. In both cases, the sum of five Gaussians was used to fit $d\rho ∕dz$. Solid green: the best fit to the labeled peptide SLD profile (dashed red) using the sum of the unlabeled SLD profile (black) plus a Gaussian with floating parameters. The best-fit Gaussian, shown in dashed green, is centered at $z=-28.36\pm 1.0\mathrm{\AA }$ with a width of $21.63\pm 1.0\mathrm{\AA }$. The residuals appear as a black dashed curve. The ordinate is normalized by the SLD of ${\mathrm{D}}_2\mathrm{O}$ $(6.25\times {10}^{-6}{\mathrm{\AA }}^{-2})$. Thus the Gaussian has an integrated area corresponding to $20.0\pm 0.8\times {10}^{-6}{\mathrm{\AA }}^{-1}$. Assuming an area of $100{\mathrm{\AA }}^2∕\alpha$ helix, the observed net change in total scattering length is $20\pm 0.8\times {10}^{-12}\mathrm{cm}$, in good agreement with expectations for the double label (which exchanges 20 ${}^1\mathrm{H}$ for ${}^2\mathrm{H}$; $\mathrm{\Delta }{b}_{\mathit{\text{total}}}=20\times 1.028\times {10}^{-12}\mathrm{cm}=20.56\times {10}^{-12}\mathrm{cm}$.

Figure 7

(Color online) Comparison of SLD profile structures on an absolute scale from the doubly labeled peptide experiment. Solid black (dashed blue): analytically integrated SLD profile from the unlabeled peptide (peptide labeled at Leu28,Leu14) with perdeuterated chains on a ${\mathrm{D}}_2\mathrm{O}$ subphase. For the unlabeled peptide (black), the sum of five Gaussians fit $d\rho ∕dz$, while for the labeled peptide (dashed blue), six Gaussians were necessary. Solid green: the best fit to the labeled peptide SLD profile (dashed blue) using the sum of the unlabeled SLD profile (black) plus a Gaussian with floating parameters. The best-fit Gaussian, shown in dashed green, is centered at $z=-32.87\pm 0.65\mathrm{\AA }$ with a width of $18.37\pm 0.65\mathrm{\AA }$. The residuals appear as a black dashed curve. The ordinate is normalized by the SLD of ${\mathrm{D}}_2\mathrm{O}$ $(6.25\times {10}^{-6}{\mathrm{\AA }}^{-2})$. Thus the Gaussian has an integrated area corresponding to $40.4\pm 1.2{\mathrm{\AA }}^{-1}$. Assuming an area of $100{\mathrm{\AA }}^2∕\alpha$ helix, the observed net change in total scattering length is $40.4\pm 1.2\times {10}^{-12}\mathrm{cm}$, about twice as large as expected for the double label.

Figure 8

(Color online) Comparison of the same SLD profiles on an absolute scale as in Fig. 7, unlabeled peptide (solid black) and Leu28,Leu14 doubly labeled peptide (dashed blue). Here, the solid green curve shows the best fit to the dashed blue curve by the sum of the unlabeled profile (solid black) plus two Gaussians, shown as dashed green curves. The Gaussian on the left is centered on $z=-37.52\pm 0.29\mathrm{\AA }$, with a width of $12.56\pm 0.32\mathrm{\AA }$ and an integrated area of $31.1\pm 0.6{\mathrm{\AA }}^{-1}$, while the Gaussian on the right is centered at $z=-6.84\pm 0.40\mathrm{\AA }$, with a width of $6.90\pm 0.39\mathrm{\AA }$ and an integrated area of $8.99\pm 0.53{\mathrm{\AA }}^{-1}$. Residuals are shown as a dashed black curve.

Figure 9

(Color online) Specular x-ray reflectivity data collected from Langmuir monolayers composed of 2:1 D-DLPE:BBC16, in which the BBC16 had perdeuterated chains and was otherwise either unlabeled (엯), or doubly labeled with ${\mathrm{D}}_{10}$-Leucine either at positions 9 and 21 (red ▵), or at positions 14 and 28 (blue ▿), before (a) and after (b) Fresnel normalization, the latter on an absolute scale. The reproducibility of the results leads to the isomorphic electron density profile structures shown in Fig. 10.

Figure 10

(Color online) Electron density profile structures on an absolute scale for 2:1 D-DLPE:BBC16 obtained by numerically integrating the profile gradients derived by box refinement from the x-ray reflectivity data shown in Fig. 9b. Black: unlabeled; red: doubly labeled at Leu21,09; blue: doubly labeled at Leu28,14. The solid vertical lines show the centers of the excess SLD due to the double labels as determined from the neutron results in Figs. 6, 8, that is, they should correspond to the positions of residues 21 and 15, as indicated. The dashed vertical lines show the corresponding positions of the individual labels that can be deduced by assuming the peptide is perfectly $\alpha$ helical, while the dotted vertical lines indicate where residues 5 and 31, the ends of the helical portion of the peptide, would be based on the same reasoning. (The electron dense feature to the right of position 5 is due to the headgroups of D-DLPE.)