Direct localization of detergents and bacteriorhodopsin in the lipidic cubic phase by small-angle neutron scattering

Using small-angle neutron scattering with contrast-matching of the lipidic cubic phase (LCP) bilayer, we determined that bacteriorhodopsin (bR) and detergents are largely dissociated upon incorporation into the LCP. This implies that the crystallization of membrane proteins in LCP is relatively insensitive to the choice of detergent, unlike the solution case, and this was experimentally verified with several detergents for bR.

-Detergent micelle dimensions. SANS data were collected for detergent solutions. The data were modeled as solid ellipsoids of revolution (top) with dimensions determined by fitting. Plots are in order of calculated micelle volume from smallest to largest. Dimensions are specified in nm. See Table  S1 for additional fit parameters and experimental details. Error bars are plus or minus one standard deviation and represent the uncertainty due to counting statistics. SurfactAmps X-100  Table S3. The maximum dimension measurable by SANS (under the present instrument configuration) was about 100 nm. In most cases, especially when the diameter and length are within roughly 2-3 fold, both prolate and oblate solutions usually exist with very similar fit quality. In these cases, the solution with the best χ 2 is shown above, but it should be understood that small changes in data or fitting procedures can easily result in selection of the alternative solution. Total ellipsoid volume, however, is generally comparable whether the prolate or oblate solution is selected. Modeling of structure factor effects was omitted in certain cases, as discussed in Fig. S2, in which case the volume fraction was fixed at 0.03. Parameter uncertainties are 68% confidence intervals.

Fig. S2 -Fitting procedures for detergent micelle scattering.
A solid, monodisperse ellipsoid model was used to fit the scattering data of all detergents. This was done for simplicity and uniformity in quantifying micelle size/shape, even though in some cases, other models may have been more appropriate (e.g., the very long micelle of AX-114 at 295 K is better fit by a flexible cylinder model, which better captures the low-q power law; χ 2 = 7 for a flexible cylinder, as opposed to 119 for the solid ellipsoid; data not shown). Fits were performed using SasView with the DREAM optimizer (Vrugt et al, 2009), which provides an estimate of (A) parameter correlations and (B) uncertainties (OG is shown above as an example). The axis limits of (A) and (B) are the 95% confidence intervals, which are listed in (B). Structure factor effects were included using a "hard sphere" model, with the effective hard sphere interaction radius constrained during fitting to the average radius of curvature of the ellipsoid. In most cases, the volume fraction and SLD of the micelle were fit independently. For "long" micelles (see Table S1), the hard sphere structure factor model was not suitable and was omitted, with volume fractions simply fixed at 0.03. This was also done in certain cases where the apparent structure factor effects were judged to be too small to reasonably constrain the fit (as determined by the SLD vs. volume fraction correlation plot). The maximum dimension measurable by SANS (under the present instrument configuration) was about 100 nm. In most cases where the length and diameter are similar (within 2-3 fold of each other), both prolate and oblate solutions can be found. In this case, the solution with the best χ 2 was reported in Table S1. However, it should be understood that these solutions can be nearly indistinguishable, such that small errors or changes in fitting procedures can easily shift the choice from one to the other. Given the model simplifications, the choice of prolate or oblate solution should not be taken as certain, at least when the length and diameter are similar.

A.
OG scattering fit -parameter correlations Parameter uncertainty B.

Figure S3 -Contrast matching of (A) LCP and (B) OG micelles.
In 100% D 2 O, scattering from the LCP is reduced by two orders of magnitude, making it possible to subtract the remaining LCP scattering when bR is included. LCP scattering is not completely eliminated since the lipid head and tail groups are not perfectly matched to each other. Scattering from cmOG micelles is, however, entirely absent at the match point of 100% D 2 O (the small amount of scattering at very low-q is typical of, e.g., cuvette imperfections, and is mainly visible due to the otherwise very low amount of scattering from cmOG). It is also possible to contrast-match the non-deuterated compounds ("hMO" and "hOG") using mixtures near 100% 1 H 2 O, but this leads to high incoherent backgrounds from 1 H that mask the signal of interest from bR. Error bars are plus or minus one standard deviation and represent the uncertainty due to counting statistics.

A.
B.

Contrast-Matching
Incoherent Incoherent

Fig. S4 -SANS curves for LCP with and without incorporated detergent.
Scattering intensities on an absolute scale, without normalization for detergent concentration, are shown after subtraction of the incoherent background. Error bars are plus or minus one standard deviation and represent the uncertainty due to counting statistics. (A) Subtraction of the LCP-only curve (black triangles) from the LCP + detergent curves (solid blue lines), and division by the detergent concentration in mg/mL, yields the blue curves in Fig. 2. Because the scattering at very low-q (i.e., less than about 0.02 Å -1 ) is mainly due to air bubbles (see Fig. S5) and differs slightly from sample to sample, the subtraction is not meaningful in this q range.  Although the low-q scattering (q ≾ 0.02 Å -1 ) was not needed for our conclusions, we have analyzed it in order to fully understand the sample, and have concluded that it was due to air bubbles. Because of the small volumes (usually around 25 µL each of lipid and aqueous components) and manual syringe mixing process used to produce the LCP, it is very difficult to purge all air from the system prior to mixing. In addition, to prevent denaturation, protein solutions must be kept cold prior to mixing. This results in an increased solubility for dissolved air, which then precipitates from solution upon mixing of the LCP at room temperature. In D 2 O, entrapped air bubbles result in a large scattering signal at very low-q (see A, blue circles, q ≾ 0.02 Å -1 ). Degassing after mixing was always attempted, and did slightly reduce the low-q scattering compared to non-degassed samples, but did not effectively remove small bubbles from the viscous cubic phase. Several lines of evidence indicate that this low-q scattering is from air bubbles, and not some other unidentified source: (1) The low-q scattering of dMO LCP in 100% D 2 O, near its match point, can be compared to that of hMO LCP at its match point of 7% D 2 O. Because the SLD of 7% D 2 O (-0.07×10 -6 A -2 ) is very near that of air (SLD of 0), any air in hMO LCP would be nearly contrast matched, and indeed the low-q scattering is greatly reduced under this condition (A, red curve).
(2) The slope of the low-q power law is very near to -4.0, as would be expected for a smooth interface from air.
(3) Because the data are on an absolute scale, it is also possible to fit a more detailed model to explain the observed scattering intensity. For instance, fitting a 20% polydisperse sphere model with an assumed volume fraction of 1%, a sphere SLD of 0 (for air), and solvent SLD of 6.38×10 -6 Å -2 (for D 2 O) gives a bubble diameter of about 22 µm, consistent with air bubbles often seen when LCP is examined by optical microscopy (see B, arrows). The bubble radius and volume fraction are completely correlated as fit parameters; at a given intensity, smaller volume fractions of air imply smaller bubbles. We do not believe more than 2% air could have been introduced during mixing, based on microscopy. Fortunately, scattering from air is well-separated from the signals of interest from bR and detergent micelles, which is in the Q range of 0.02-0.2 Å -1 . Error bars are plus or minus one standard deviation and represent the uncertainty due to counting statistics.  The lattice parameters are tabulated for the different mixtures. The same samples were used as for SANS (Fig. 2), but the Bragg peaks are largely absent in SANS due to contrast matching. Intensities have not been put on an absolute scale. Expected peak positions for the Pn3m cubic phase are marked for the two extremes ("No detergent" and "OG"). After solubilization in OG, SEC was used to separate monomeric bR from aggregates. Peak fractions were collected and concentrated for incorporation into LCP. Protein was never frozen after performing SEC. Because bR was highly concentrated, absorbance was monitored at 620 nm (away from the bR absorption maximum at 550 nm) in order to remain in the linear range of the detector. Standards were used to determine the apparent molecular weight of the monomeric bR/OG complex, which was found to be 71. were brought away from the LCP match point (near 100% D 2 O) by removing 1/3 of the D 2 O buffer from the cuvette, and replacement with H 2 O buffer, resulting in a solvent containing 67% D 2 O. This allowed for the determination of which Bragg peaks were due to the lipidic matrix, since these would greatly increase in intensity away from the LCP match point. Conversely, this brought the solvent nearer to the match point of the protein (about 42% D 2 O). Therefore, if any Bragg peaks originated from ordering of the protein, these would decrease in intensity. All of the residual Bragg peaks seen near the LCP match condition were found to increase greatly in intensity when H2O was added, indicating an origin in lipid structures rather than protein. Peak positions in samples with no bR or 0.66 mg/mL bR were consistent with the Pn3m phase (ratios of the q positions of the first peak to its shoulder, marked with vertical lines, were near 2: 3). At the highest bR concentration of 4.16 mg/mL, a new peak appears at q ≈ 0.053 Å -1 , indicating a phase change, although this could not be confidently indexed due to the poor q-resolution and small number of Bragg peaks. The onset of this behavior can also be seen in an additional shoulder near q ≈ 0.05 Å -1 in the scattering from the 1.65 mg/mL bR sample.  Figure S9 -Comparison of bR/LCP scattering to theoretical scattering from bR monomer and trimer. The experimental scattering curve from 1.65 mg/mL bR in LCP is shown, after subtraction of the residual LCP scattering (as in Fig. 3B). CRYSON was used to compute theoretical scattering curves from atomistic models. The bR monomer and trimer structures used in the calculation were from PDB code 1c3w with lipids deleted. Calculated scattering from CRYSON was not scaled to the experimental data, but was placed on an absolute scale using I 0 values from the SASSIE Contrast Calculator; thus, the good agreement between the calculated and experimental scattering intensity is notable. The experimental bR/LCP scattering was consistent with the calculated monomeric scattering plus a small amount of fractal aggregate, as determined by fitting using SasView with the built-in "fractal" model (see Fig. S2 for additional details on SasView fitting procedures). The volume fraction of aggregate was 4.91×10 -5 , which corresponds to 0.066 mg/mL bR, or 4.0% of the total protein mass, present as aggregate. In contrast to the monomeric case, the calculated scattering for trimeric bR was inconsistent with the experimental data.

Parameter Fit Value
Scale 1 (fixed) Background 0 (fixed) Volume Fraction (4.9 ± 0.3)×10 -5 Radius (  Anapoe-20 Anapoe X-114 (n = 7-8) Anapoe NID-P40 (n = 8.5-9.5) Triton X-100 (n = 9-10) Anapoe X-305 Single-phase LCP mixtures are difficult to prepare consistently, since the LCP has a narrow range where the Pn3m phase is formed (near full hydration) and the location of the phase boundary is sensitive with respect to conditions such as temperature, salt, and the amount of detergent added. Instead, we prepare two-phase dispersions of LCP in an excess of buffer. This ensures that the LCP is always exactly at full hydration. Since the buffer and LCP are near the contrast-matched condition, scattering from the LCP grain/solution interfaces is minimal. Dispersion of the LCP also facilitates degassing after insertion into cuvettes, which is important for reducing the large amount of low-q scattering often present from small air bubbles.          1. Glycerol-d 8 (98% deuterated) was purchased from Sigma-Aldrich; it is assumed that no H/D exchange would occur during the subsequent synthesis. 2. The percentage deuteration of the tail was calculated assuming that the deuteration of the head remained as 98% and having determined the overall deuteration of the molecule to be 93.2% (±2%) via mass spectrometry: