Renaturing Membrane Proteins in the Lipid Cubic Phase, a Nanoporous Membrane Mimetic

Membrane proteins play vital roles in the life of the cell and are important therapeutic targets. Producing them in large quantities, pure and fully functional is a major challenge. Many promising projects end when intractable aggregates or precipitates form. Here we show how such unfolded aggregates can be solubilized and the solution mixed with lipid to spontaneously self-assemble a bicontinuous cubic mesophase into the bilayer of which the protein, in a confined, chaperonin-like environment, reconstitutes with 100% efficiency. The test protein, diacylglycerol kinase, reconstituted in the bilayer of the mesophase, was then crystallized in situ by the in meso or lipid cubic phase method providing an X-ray structure to a resolution of 2.55 Å. This highly efficient, inexpensive, simple and rapid approach should find application wherever properly folded, membrane reconstituted and functional proteins are required where the starting material is a denatured aggregate.

Direct, chromatographic kinase assay. To measure kinase activity in meso in Acidic Urea, the formic acid and urea components of which were incompatible with PK and LDH in the coupled assay above, the direct ATP-dependent production of lyso-phosphatidic acid (lyso-PA) from monoolein was monitored using TLC 1 . Typically, 20 L of the protein-laden mesophase in a 1.5 mL Eppendorf tube was incubated at RT with 1 mL of either Acidic Urea or Refolding Buffer. After shaking (800 rpm, Thermomixer) for 30 min at RT, the bathing solutions were replaced with fresh bathing solution, and the process repeated twice. The kinase reaction was initiated by adding to the bathing solution magnesium acetate and ATP to 10 mM and 60 mM, respectively. The reaction was allowed to run for 1 h at 30 °C with shaking (800 rpm, Thermomixer) after which the bathing solution was removed and 0.8 mL chloroform:water (1:1 by vol.) was added to solubilize the lipid. Centrifugation at 20,000 g for 10 min at RT separated the phases and the upper aqueous phase was removed. 2 L of the chloroform solution containing the extracted lipid were loaded onto TLC plates (Cat. No. 1.05554.0001, HX068423. Merck, Darmstadt, Germany) that had been pre-run in chloroform. The plates were developed in chloroform:methanol:acetone:acetic acid:water (10:2:4:2:1 by vol.) at RT. After drying on a heat block at 40 °C under nitrogen for 15 min and staining with 20 %(w/v) phosphomolybdic acid in ethanol, the plate was placed on a hot plate at 150 °C for stain development.
Protein concentration was determined by measuring A 280 (ε 1 mg/ml = 2.1) 5, 6 in a NanoDrop 1000 spectrometer (Thermo Fisher Scientific Inc., Wilmington, DE). UV-visible spectroscopic analysis was performed at 0.64 mg protein/mL in a 1 cm-pathlength quartz cuvette (Sigma Aldrich, St. Louis, MO, USA) with a UVIKON XL spectrophotometer (Northstar Scientific, Leeds, UK). Spectra were recorded from 340-275 nm at a scanning speed of 200 nm/min. For comparisons, all spectra were scaled to the same A 280 . Difference spectra were obtained by subtracting the spectrum recorded in DM from that recorded in Acidic Urea with Detergent, SDS Buffer and Acidic Urea.
Fluorescence measurements were carried out at 0.1 mg protein/mL in a 3 mm pathlength quartz cuvette (Hellma, Jena, Germany) with a FluoroMax-3 spectrofluorometer (Horiba, Kyoto, Japan). Emission spectra were recorded from 375-335 nm at 10 nm/s with an excitation wavelength of 295 nm and slits corresponding to a spectral width of 2 nm. Circular Dichroism (CD). CD analysis was carried out at 0.57 mg protein/mL in a 0.1 mm pathlength quartz cuvette (Starna, Hainault, UK) with a Jasco J-815 spectrometer (Jasco, Easton, MD, USA) at 20 ºC. Spectra from 260-190 nm were recorded at 50 nm/min in 1 nm steps with a band width setting of 1 nm. Spectra were smoothed using the binomial function included in the Jasco spectra analysis software package (version 1.54.03). As reported previously 7 , high concentrations of urea give noisy CD data in the 190-215 nm range. This region was omitted from spectra of urea-containing samples in the current study.

Rounds of reconstitution.
The solubility of DgkA* in acidic urea was only 1 mg/mL. After reconstitution under standard conditions, the protein concentration in the cubic phase would only reach 0.4 mg/mL, 12 -15 times below the concentration used typically for crystallization (4.8 mg/mL in monoolein; 6 mg/mL in 7.8 MAG) 9 . To incrementally raise protein concentration in the mesophase a method referred to as 'rounds of reconstitution' was implemented, as outlined below (Fig. S2)).
Sixteen microliters of DgkA* at 1 mg/mL in acidic urea was homogenized with 24 L of monoolein (Step 1, Fig. S2) by using a coupled syringe mixer consisting of two 0.1 mL Hamilton syringes (Syringes A and B) and a narrow-bore coupler 10 . After mixing, the freshly formed protein-laden mesophase was transferred to Syringe A. Syringe B was replaced with a 0.5 mL Hamilton syringe (Syringe C) containing 0.45 mL Refolding Buffer (1 mM TCEP, 0.1 M NaCl, 0.1 M HEPES pH 7.5) and the mesophase in Syringe B was transferred to Syringe C. Empty Syringe A was replaced with an empty 0.5 mL syringe (Syringe D) and the contents of the coupled syringe device were mixed at RT to dilute out the denaturant and to reconstitute and refold the protein (Step 2). After 30 min of mixing, the contents were transferred to Syringe D and the coupled syringe was left to sit at RT for 10 min in a vertical position with Syringe D on top. The mesophase rises in the barrel of Syringe D to the Teflon tip of the plunger where it naturally sticks together. As a result, the lower, phase-separated buffer fraction could be transferred from Syringe D into Syringe C with very little loss of mesophase. Syringe C was detached and excess buffer discarded (Step 3). The process of removing excess buffer was repeated until the volume of mesophase with residual buffer was less than 0.1 mL. At this point, the contents were transferred to coupled 0.1 mL syringes (Syringes E, F) to further reduce excess buffer. The volume recorded at this final stage was ~38 L with a calculated protein concentration in the mesophase of 0.4 mg/mL. The entire contents were transferred to Syringe F in preparation for rounds of reconstitution.
Rounds of reconstitution began by mixing 50-60 L of DgkA at 1 mg/mL in acidic urea with the ~38 L of protein-laden mesophase obtained from the initial reconstitution step above (Step 4). This volume ratio ensures that the urea concentration does not drop below 6.5 M at which point the protein may come out of solution. For this purpose, the protein solution and mesophase were contained in coupled 0.1 mL syringes (Syringes F, G). Mixing was carried out for 30 min after which the contents were divided equally between the two syringes. The contents of each were washed three times with Refolding Buffer, as outlined above, and finally combined in one syringe (Steps 5 and 6). Rounds of reconstitution (Steps 4 -6) were repeated until a total of 200-320 L (0.20 -0.32 mg) of DgkA in acidic urea solution had been used. It is estimated that 1.5-2 L of mesophase was lost at each round. The proteinladen mesophase prepared in this way contained a small volume of excess buffer and was cloudy (Step 7). To absorb excess buffer and to convert the system entirely to the optically clear mesophase ready for crystallization and functional assays, 2-3 mg of monoolein was mixed with the sample (Step 8).
After 3-5 rounds of reconstitution, ~32 L of mesophase with an estimated DgkA concentration of 6-9 mg/mL was recovered. It was washed three times with 0.45 mL Refolding Buffer and then three times with Washing Buffer (1 mM TCEP, 0.1 M NaCl, 10 mM Tris pH 7.8) to remove residual formic acid and urea and to switch from HEPES to Tris-HCl buffer in preparation for enzyme assays and for crystallization trials.
The rounds of reconstitution method was implement also with 7.8 MAG in which case equal volumes of lipid and protein solution were used to form the mesophase initially, following an established protocol 9 . The entire process results in ~28 and ~22 L of protein-laden mesophase for monoolein and 7.8 MAG, respectively.
X-ray diffraction. Diffraction data were collected on GM/CA CAT beamline 23ID-B, the Advanced Photon Source (APS), beamline I24, the Diamond Light Source (DLS), and PX II at the Swiss Light Source (SLS). At the APS, data were collected with a 1° oscillation and a 1 s exposure per image, a collimated beam size of 10 × 10 μm 2 and a sample-to-detector distance of 350-500 mm, with a MAR 300 CCD detector using 1.033 Å wavelength X-rays. At the DLS, data were collected with a 0.2 ° oscillation and a 0.2 s exposure per image, a micro-focus beam size of 10 × 10 μm 2 and a sample-to-detector distance of 400-650 mm, with a Pilatus 6M detector using 0.978 Å wavelength X-rays. At the SLS, data were collected with a 0.1 ° oscillation and 0.1 s exposure per image, a collimated beam size of 10 or 30 × 15 μm 2 and a sample-to-detector distance of 300-490 mm, with a Pilatus 6M detector using 1.033 Å wavelength X-rays. Diffraction images, recorded with a 10-fold attenuated beam, were used to locate crystals in the mesophase and to center on highly ordered regions of the crystal. Complete data sets for wild-type and thermostable DgkA* were collected with single crystals grown in the 7.8 MAG mesophase. Data was reduced with xia2 13 using XDS 14 , XSCALE and SCALA 15 .
Structure solution and refinement. Initial phases for thermostable and wild-type DgkA* were obtained by molecular replacement using Phaser 16 with the protein component of a published structures (PDB 3ZE3 and 3ZE4, respectively) 2 as the search model. In subsequent cycles of iterative model building and refinement, the program Coot 17 was used for model building; the program Phenix 18 was used for refinement. For the low resolution wild-type DgkA* structure at 3.8 Å, non-crystallographic symmetry restraints were applied to torsion angles during refinement. Structures were visualised with Pymol 19 .  Step 1: Monoacylglycerol lipid and solubilized membrane protein in denaturant solution are mixed, spontaneously forming the bicontinuous cubic mesophase.
Step 2: Denaturant-free buffer is added which enables denaturant to flood out of the nanoporous mesophase and the protein to reconstitute and renature in the bilayer of the mesophase.
Step 3: Most of the excess denaturant-containing buffer, that naturally separates from the mesophase, is removed. When enough protein has been reconstituted into the mesophase for end use, the process moves to Step 8 where excess buffer is absorbedd by mixing in a small amount of lipid to produce homogenous, optically clear cubic phase.
Step 4: When insufficient protein has been incorporated into the mesophase, rounds of reconstitution begins by mixing in fresh denaturant solution that contains solubilized membrane protein. The protein preferentially associates with the mesophase.
Step 5: Denaturant-free buffer is added to dilute out the denaturant thereby facilitating reconstitution and renaturation of solubilized membrane protein.
Step 6: Most of the excess denaturant-containing buffer that naturally separates from the mesophase is removed leaving behind mesophase enriched in membrane protein.
Step 7: Steps 4, 5 and 6 are repeated sequentially until sufficient protein has been reconstituted and renatured in the mesophase.
Step 8: Excess buffer is absorbedd by mixing in a small amount of lipid to produce homogenous, optically clear protein-laden cubic phase suitable for end use. The reference spectrum was recorded with untreated WT DgkA in detergent micelles. Sample spectra were recorded with detergent-free WT DgkA* in SDS (green), in acidic urea (blue), and DgkA in acidic urea containing detergent (red) and subtracted from the reference spectrum. The minimum at 294 nm is characteristic of denatured DgkA 6 . b, Fluorescence emission of untreated WT DgkA in detergent micelles (black) and in acidic urea containing detergent (red), and of detergent-free WT DgkA* in SDS (green) and in acidic urea (blue). A red shift, coupled with a reduction in fluorescence yield, is indicative of tryptophan exposure to a polar environment and to denaturation. c, Circular dichroism spectra of untreated WT DgkA in detergent micelles (black) and in acidic urea containing detergent (red), and of detergent-free WT DgkA* in SDS (green) and in acidic urea (blue). In urea, data are only shown to 215 nm due to strong absorbance at lower wavelengths. A strong negative ellipticity at 222 nm correlates with -helical secondary structure and the folded state. All data shown were collected with WT DgkA. The corresponding data for thermostable DgkA are shown in Fig. 2. Both constructs behaved similarly.