Stable picodisc assemblies from saposin proteins and branched detergents

Abstract

Methods for maintaining membrane proteins in their native state after removal from the lipid bilayer are essential for the study of this important class of biomacromolecules. Common solubilization strategies range from the use of detergents to more complex systems that involve a polypeptide working in concert with lipids or detergents, such as nanodiscs, picodiscs and peptidiscs, in which an engineered protein or synthetic peptide surrounds the membrane protein along with a lipid sheath. Picodiscs employ the protein saposin A, which naturally functions to facilitate lipid degradation in the lysozome. Saposin A-amphiphile complexes therefore tend to be most stable at acidic pH, which is not optimal for most membrane protein applications. In search of new picodisc assemblies, we have explored pairings of saposin A or other saposin proteins with a range of detergents, and we have identified a number of combinations that spontaneously co-assemble at neutral pH. The resulting picodiscs are stable for weeks and have been characterized by size-exclusion chromatography, native mass spectrometry and small-angle x-ray scattering. The new assemblies are formed by double-tail detergents rather than more traditional single-tail detergents; the double-tail detergents can be seen as structurally intermediate between single-tail detergents and common lipids. In addition to saposin A, an engineered variant of saposin B (designated saposin B^W) forms picodisc assemblies. These findings provide a framework for future efforts to solubilize membrane proteins with multiple picodisc systems that were previously unknown.

Graphical Abstract

Introduction

Characterization of membrane proteins is challenging because some sort of molecular "escort" is required to maintain the native conformation in aqueous solution after protein removal from the lipid bilayer. A local membrane-mimetic environment can be maintained with detergent micelles or more sophisticated systems in which an engineered protein, peptide or synthetic polymer encloses the protein of interest along with a pool of lipids. Most widely studied among the latter approaches are the "nanodisc" assemblies pioneered by Sligar et al.^1,2 Nanodiscs employ a membrane scaffold protein (MSP) to support a belt of phospholipid molecules around the target membrane protein. This technique has supported elucidation of several membrane protein structures via NMR^3,4 and cryo-EM studies.^5-14 The success of this approach has inspired studies of alternative escort proteins, such as saposin A,^15-22 as well as the use of designed peptides^23,24 or synthetic polymers (styrene maleic acid co-polymer lipid particles, SMALPs).²⁵ Continued development of these systems is necessary because many membrane proteins remain difficult or impossible to solubilize in a manner that enables structural and/or functional characterization.

Saposin A is one of a family of sphingolipid activator proteins.¹⁵ At neutral pH, saposin A exists in a “closed” conformation that buries hydrophobic surfaces away from the aqueous environment²⁶ (Figure 1). Under acidic conditions native to the lysosome (pH = 4.8), and in the presence of lipid and detergent saposin A accesses an “open” conformation that displays hydrophobic surfaces and enables binding to lipid molecules (Figure 1).¹⁵ Saposin A can also access this open conformation and assemble into lipoprotein discs at neutral pH in the presence of DDM. After assembly, the detergent is removed using detergent-binding beads¹⁸ or the sample is diluted to below the CMC value of DDM and further purified.¹⁹ Saposin A-lipid assemblies have been used to encapsulate membrane proteins and facilitate structural analysis via NMR¹⁸ and cryo-EM.^19-22 These "picodisc" assemblies contain two or more saposin A monomers, depending on the ratio of protein to lipid molecules and the size of the membrane protein encapsulated.^16-20,27 The crystal structure of a picodisc in which two saposin A monomers surround 40 lauryl dimethylamine-N-oxide (LDAO) molecules, reported in 2012,¹⁵ offers the prospect that this solubilization method might ultimately be useful for obtaining crystallographic insights on membrane protein structure. However, no subsequent crystal structures of picodisc assemblies have been reported (neither nanodisc assemblies nor peptidisc assemblies have been crystallized). One potential problem with currently known picodisc assemblies is limited stability under conditions required for maintaining membrane protein structure, which usually include neutral pH. Therefore, we were motivated to seek saposin-amphiphile assemblies that would form at neutral pH and display high stability under these conditions. New picodisc systems with these properties might ultimately prove useful for structural and functional characterization of membrane proteins.

Figure 1.

In solution at neutral pH, saposin A adopts a “closed” conformation (left, PDB = 2DOB). Upon incubation with lipids at acidic pH or detergent-solubilized lipids at neutral pH, saposin A adopts an “open” conformation (right, PDB = 4DDJ) and binds to detergents or lipids to form lipoprotein picodiscs.

Our experimental design was inspired by inspection of the detergent component in the saposin A-LDAO crystal structure (PDB 4DDJ). The LDAO molecules display a bilayer arrangement, with the polar amine oxide groups clustered on two surfaces in the middle of the disc and facing the aqueous solvent. The hydrophobic lauryl tails project toward the center of the disc, where they are protected from hydration.¹⁵ This mode of assembly is unusual for LDAO and other conventional detergents, which generally prefer assemblies with a high radius of curvature, such as a spherical micelle. In contrast, lipid molecules, which often have two long hydrophobic tails rather than the single hydrophobic tail common among detergents, favor assemblies with a low radius of curvature, a preference that leads to bilayer formation. These divergent behaviors have been rationalized based on amphiphile packing parameters:²⁸ the cone-like shape of single-tail surfactants leads to assemblies with high curvature, whereas the rod-like shape of double-tail surfactants leads to assemblies with low curvature.

Packing parameter considerations led us to explore the possibility of picodisc formation with recently developed detergents that feature two alkyl tails and have been demonstrated to promote membrane protein solubilization.^29-31 Although these branched detergents form micelles, we hypothesized that their behavior might be intermediate between those of conventional detergents and lipids. While we predict that the branched detergents will form a bilayer-like arrangement similar to lipids in the presence of saposin protein, unlike lipids these molecules rapidly exchange in solution. The ability to exchange in solution could allow saposin+detergent complexes to each a thermodynamic energy minimum. Our results show that several double-tail detergents form very stable picodisc assemblies with saposin A and with another member of this family, saposin B. Collectively, these findings offer a range of new strategies for future exploration of membrane protein solubilization and crystallization.

Materials and Methods

Protein Expression and Purification.

Codon optimization and cloning of the saposin A gene into the expression vector, pET-15b, between XhoI and BamHI enzyme digestion sites was performed by GenScript. Codon optimization and cloning of the saposin B, C, and D genes mutated to contain a single tryptophan (see SI for details) into the expression vector, pET-15b, between NdeI and XhoI enzyme digestion sites was performed by GeneUniversal. The modified saposins are designated saposin B^W, etc., because of the added Trp residue. The pET-15b vector includes an Amp-resistant gene as well as an N-terminal 6X-His tag followed by a thrombin cleavage site for purification. The resulting plasmids were transformed into E. coli strain SHuffle cells (New England Biolabs Inc.). The growth and purification of saposin A, B^W, C^W, and D^W follows literature precedent.¹⁵ Briefly, cells were grown at 37°C in Terrific broth until the samples reached an OD600 of 0.4-0.8. Induction was performed by addition of isopropyl β-D-1-thiogalactopyranoside to a concentration of 1 mM. After 4 hours of growth at 37°C, the cells were harvested by centrifugation at 5,000 rpm for 10 min.

Cells were resuspended in Ni affinity wash buffer (25 mM Tris-HCl, pH = 7.5, 150 mM NaCl) and subjected to lysis by sonication. Lysates were centrifuged at 10,000 rpm for 10 min and the supernatants collected. Isolated supernatants were heated to 85°C for 10 min prior to an additional centrifugation at 10,000 rpm for 45 min. Samples were subjected to affinity purification by injection onto a 5 mL His-Trap column (GE Lifesciences) and were eluted with a linear gradient of elution buffer (elution buffer = 25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 500 mM imidazole; gradient = 0-50% elution buffer, 100-50% wash buffer) performed using a NGC Quest FPLC system (Bio-Rad) followed by cleavage of the 6X-His tag by thrombin (thrombin-agarose resin, Sigma-Aldrich). Cleaved protein was first purified by affinity and further purified by size-exclusion chromatography (Superdex 75 16/600, 120 mL, GE Lifesciences) with 25 mM Tris-HCl, pH 7.5, 150 mM NaCl elution buffer.

Picodisc Formation.

Saposin A- and saposin B^W-detergent co-assemblies were generated from concentrated stock solutions of purified protein and detergent in 25 mM Tris-HCl, pH 7.5, 150 mM NaCl buffer. Once the protein and detergent were combined, samples were incubated at 37°C for 45 min and then cooled to room temperature.

Size-Exclusion Chromatography (SEC) Studies.

The relative size and stability of saposin+detergent assemblies were assessed by SEC. Elution profiles of saposin+detergent samples were determined with two size-exclusion columns (Superdex 75 16/600 and Superdex 200 16/600, GE Lifesciences). The Superdex 75 column is optimal in the molecular weight range 3-70 kDa, while the Superdex 200 is optimal for 10-200 kDa (Superdex 75 16/600, 120 mL, GE Lifesciences) with 25 mM Tris-HCl, pH 7.5, 150 mM NaCl elution buffer at a flow rate of 1 mL/min. The elution profiles of molecular weight standards (Low Molecular Weight Gel Filtration Kit, GE Lifesciences) were acquired using the Superdex75 column with 25 mM Tris-HCl pH 7.5, 100 mM Na2SO4, 2% glycerol elution buffer (Figure S2).

Tryptophan Fluorescence Spectroscopy.

Samples were incubated at a 1:10 ratio of protein (20 μM) to detergent (200 μM). After incubation at 37°C for 45 min, the samples were either assessed directly or diluted 10X with buffer (25 mM Tris-HCl, pH 7.5, 150 mM NaCl) prior to data collection in a 1.0 cm x 1.0 cm Quartz cuvette. The excitation wavelength was 295 nm, and fluorescence emission was collected between 305 and 500 nm with excitation and emission slits of 1.0 mm. All data were collected on an ISS PC1 photon counting fluorometer (ISS Instruments Inc., Champaign, IL). Data were also collected for buffer samples, and data obtained with protein-containing solutions were corrected by subtracting the data from the buffer.

Native Mass Spectrometry.

Mass spectrometry analysis was performed on a maXis II ETD Q-TOF (Bruker Daltonics). SEC purified icodisc assemblies (in 200 mM ammonium acetate) were infused via electrospray ionization at a flow rate of 3 μL/min using a syringe pump. Instrument parameters were as follows: positive ionization mode, capillary voltage (3.8 V), end plate voltage (500), dry gas flow (5 l/min), source temperature (150 °C), funnel 1 RF (400 Vpp), multipole RF (600 Vpp), quadrupole energy (4 ev), collision cell energy (4 ev), transfer time (140 μs), and pre pulse storage (42 μs). Bruker DataAnalysis V4.3 software was used to process the data.

SAXS Data Collection.

SAXS data were collected at beamline 12-ID-B of the Advanced Photon Source at Argonne National Laboratory. The wavelength, λ, of X-ray radiation was set to 0.9322 Å. Scattered X-ray intensities were measured using a Pilatus 2M detector. The sample-to-detector distance was set such that the detecting range of momentum transfer q [=4π sinθ/λ, where 2θ is the scattering angle] was 0.005-0.85 Å⁻¹. The flow cell was a cylindrical quartz capillary 1.5 mm in diameter and 10 μm wall thickness. The exposure time was set to 1 second to reduce radiation damage, and data were collected at every other second. The 2-D scattering images were converted to 1-D SAXS (I(q) vs q) curves through azimuthally averaging after solid angle correction and then normalizing with the intensity of the transmitted X-ray beam flux, using the beamline software.

SEC-SAXS data were acquired by inclusion of an in-line AKTA micro FPLC setup with a Superdex 75 Increase 5/150 GL size-exclusion column (GE Lifesciences). The sample passed through the FPLC column and was fed to a flow cell for SAXS measurements. See SI for data analysis details.

Results

Identification of Saposin A-Detergent Picodiscs That Assemble at Neutral pH.

We surveyed a range of commercially available detergents (Table S1) in search of new saposin A+detergent assemblies that form and remain stable at neutral pH. Association of detergent with saposin A was assessed via changes in tryptophan fluorescence emission as previously described.^15,18 Saposin A contains a single tryptophan residue. In the crystal structure of monomeric saposin A (closed conformation; PDB 2DOB), the tryptophan residue is exposed to solvent,²⁶ while in the saposin A+LDAO picodisc structure (PDB 4DDJ), the Trp residue is oriented toward the detergent molecules that fill the disc’s interior¹⁵ (Figure S3). Tryptophan fluorescence is highly sensitive to the local environment.³¹ We therefore expect a change in tryptophan fluorescence emission (blue shift in the maximum and increase in intensity) upon formation of a picodisc assembly, because the local environment of the tryptophan side chain is of lower polarity in the picodisc than in the monomeric form of saposin A.

Initial studies involved saposin A and a homologous series of alkyl-β-D-maltopyranosides,³³ conventional detergents that have been widely used for membrane protein solubilization. Detergent concentration was held constant at 0.2 mM. No fluorescence change was observed when saposin A was combined with n-decyl-β-D-maltopyranoside (DM), but a slight blue shift in the fluorescence emission maximum and a small intensity increase were evident when saposin A was incubated with n-dodecyl-β-D-maltopyranoside (DDM, Figure 2; Figure 3A). The DDM data are consistent with previous observations for saposin A+DDM combinations at neutral pH.^18,19 Nietlispach and coworkers have hypothesized that DDM promotes a closed-to-“open-like” conformational change of saposin A, which is necessary for picodisc formation.¹⁸ Use of a longer-chain analogue, n-tridecyl-β-D-maltopyranoside (TriDM, Figure 2) or n-tetradecyl-β-D-maltopyranoside (TetraDM, Figure 2), led to larger changes in tryptophan fluorescence emission (Figure 3A).

Figure 2.

Chemical structures of the detergents that were observed to assemble with saposin A or saposin B^W.

Figure 3.

Tryptophan fluorescence emission spectra of monomeric saposin A (20 μM, black) overlaid with data collected from samples of saposin A (20 μM) incubated in a 1:10 ratio of detergent (200 μM) in pH 7.5 conditions (buffer = 25 mM Tris-HCl, pH 7.5, 150 mM NaCl). Panel A shows the emission spectra of saposin A combined with detergent DM (orange), DDM (red), TriDM (purple) and TetraDM (blue). Panel B shows the emission spectra of saposin A combined with MNA C12 (royal blue), MNA C13 (dark red) and LMNG (green).

The CMC values of the alkyl maltoside detergents become smaller as chain length increases. The detergent concentration used for these exploratory studies (0.2 mM) is below the CMC of DM (1.8 mM (ref. 34)), slightly above the CMC of DDM (0.17 mM (ref. 34)) and well above the CMC values for TriDM, and TetraDM (0.024 and 0.01 mM (ref. 35)), respectively. These results suggest that micelle formation may be required for saposin A-detergent co-assembly. The relationship between CMC and saposin A-detergent co-assembly was further explored through experiments involving 0.02 mM TriDM and TetraDM; this detergent concentration is below the CMC of TriDM but above the CMC of TetraDM. At this lower detergent concentration, the tryptophan fluorescence of the saposin A+TriDM combination was identical to that of monomeric saposin A, but the saposin A+TetraDM combination displayed a blue-shift and increase in intensity (Figure S4). These data support the hypothesis that micelle formation is necessary for saposin A-detergent co-assembly.

The evidence of possible saposin A co-assembly with TriDM and TetrDM from tryptophan fluorescence measurements led us to investigate these systems via size-exclusion chromatography (SEC). The saposin A+LDAO picodisc characterized via crystallography would have a molecular weight of approximately 28 kDa, the molecular weight of two saposin A monomers plus forty LDAO molecules. Species resulting from the co-incubation of saposin A (105 nM) with TriDM or TetraDM (2.1 mM, corresponding to a 1:20 ratio of protein to detergent) had a lower elution volume than that of monomeric saposin A; the observed elution volumes suggest formation of species that are consistent in apparent size with a co-assembly containing two saposin A molecules and multiple detergent molecules (Figure S5). However, the Superdex 75 elution profiles of these same samples showed elution volumes only slightly lower than that of monomeric saposin A and are consistent with a species the molecular weight of which is between that of monomeric saposin A and the saposin+LDAO disc (Figure 4A, B), suggesting dissociation on the Superdex 75 column of the higher molecular weight species that was observed in the Superdex 200 elution profile.

Figure 4.

The SEC (Superdex75) elution volumes of saposin A incubated with different protein:detergent ratios were compared with monomeric saposin A (black traces) to detect assembly. Panel A shows the traces for saposin A-TriDM assembly; panel B, saposin A-TetraDM assembly; panel C, saposin A-MNA C12 assembly; panel D, saposin A-MNA C13 assembly; and panel E, saposin A-LMNG assembly. The colors of the traces indicate the protein:detergent ratio, with green indicating 1:5, blue indicating 1:10, and pink indicating 1:20. For reference, the elution volumes of molecular weight standards are indicated by black triangles. The elution buffer used was 25 mM Tris-HCl, pH 7.5, 150 mM NaCl.

SEC studies of saposin A+DDM mixtures suggested that even if a detergent micelle is formed, co-assembly to form a picodisc is not guaranteed at neutral pH. We examined solutions containing saposin A (105 nM) and DDM (3.9 mM, 7.8 mM, and 15.7 mM; corresponding to protein:detergent ratios ranging from 1:37 to 1:148), in which the DDM concentration was roughly 23- to 92-fold higher than the DDM CMC value. None of these solutions provided SEC data consistent with a picodisc-sized assembly (Figure S6). Taken together, the SEC studies with conventional single-tail detergents indicate that picodisc assemblies may form between saposin A and TriDM or TetraDM at neutral pH, but these picodiscs are not very stable, and that DDM does not co-assemble with saposin A at neutral pH to form picodiscs.

We turned to detergents with branched structures that have recently been developed for membrane protein applications, lauryl maltose neopentyl glycol (LMNG)²⁹ and mannitol-based amphiphiles MNA-C12 and MNA-C13³⁰ (Figure 2). As with DDM and its longer analogues, the polar groups in the new detergents are derived from carbohydrates, either glucose or maltose (a dimer of glucose). However, in contrast to DDM and other classical detergents, the new molecules contain two hydrophobic tails, a feature that is reminiscent of lipid structures. The unusual architectures of these detergents give rise to superior performance in membrane protein solubilization and stabilization.^29-31,36,37 We speculated that the presence of two hydrophobic tails might alter the packing parameters^28,38 and/or other properties of these detergents in a way that would be conducive to the bilayer-like association necessary for picodisc formation with saposin A.

The tryptophan fluorescence spectra for saposin A combined with LMNG, MNA C12 or MNA C13 in pH 7.5 buffer were blue-shifted and showed increased intensity in comparison with that of the monomeric saposin A (Figure 3B). The CMC values of these detergents are comparable to or lower than those of TriDM and TetraDM, and the CMC values of LMNG, MNA C12 and MNA C13 are lower than the concentrations of these detergents used in the tryptophan fluorescence emission studies. Thus, the tryptophan fluorescence data are consistent with the hypothesis that micelle formation is required for saposin A-detergent co-assembly.

We used SEC to examine co-assemblies formed by saposin A and each of the double-tail detergents. As many biophysical methods require sample purification via SEC we used SEC as one benchmark for our stability assessment. In each case we sought to identify the smallest detergent:saposin A ratio that would support picodiscs containing two saposin A molecules. As our tryptophan studies suggested CMC values may be required for picodisc formation, in each sample the detergent concentration was higher than its CMC. Comparison of the elution volumes determined in these experiments with results for a set of standard proteins with known molecular weights and for monomeric saposin A allowed us to identify the minimal protein:detergent ratios required to achieve full conversion of monomeric saposin A to saposin A-detergent assemblies (Figure 4). The presence of a peak with an elution volume corresponding to a species whose molecular weight is consistent with saposin A dimer+detergent indicated to us the formation of a thermodynamically and kinetically stable complex. We discuss the MNA C12 co-assembly as an example. The elution profile of the species produced upon incubation of saposin A with MNA C12 at a 1:5 protein:detergent ratio suggests that a fraction of saposin A forms a picodisc assembly, and the remaining protein is monomeric. At a 1:10 protein:detergent ratio, the elution profile suggests formation of a single species with a mass consistent with that of a picodisc. Increasing the detergent proportion (1:20 protein to detergent ratio) does not appear to lead to increased sample homogeneity; this observation suggests that a 1:10 saposin A:MNA C12 ratio is necessary and sufficient to form a homogeneous picodisc assembly. Repeating this analysis with MNA C13 and LMNG led us to conclude that a 1:10 or 1:15 protein:detergent ratio is required to produce full conversion of saposin A to a picodisc when co-incubated with MNA C13 or LMNG, respectively.

Probing Saposin A+Detergent Picodisc Composition via Native Mass Spectrometry.

Lipid-containing saposin A picodiscs have previously been characterized using native mass spectrometry.^27,39-42 We used this method to evaluate the compositions of assemblies formed between saposin A and the detergents in Figure 2. After co-assembly, saposin A-detergent samples were dialyzed and re-purified in a buffer compatible with native mass spectrometry (200 mM ammonium acetate, pH 6.8). Assemblies formed with the conventional detergents TriDM or TetraDM proved too unstable for native MS analysis. SEC purification of the TriDM assemblies dialyzed into the mass spectrometry-compatible buffer showed an elution profile similar to that of the monomeric saposin A, which suggests that the picodisc dissociated in the new buffer. The SEC elution profile of the TetraDM assembly suggested that the picodisc was intact after dialysis; however, the native mass spectrum of the TetraDM assembly did not show any peaks with m/z values that were consistent with an intact picodisc. Based on these data, we conclude that the TetraDM assembly is not stable enough to survive electrospray ionization. Assemblies formed by saposin A with LMNG, MNA C12 or MNA C13 were amenable to native mass spectrometry, in contrast to the assemblies formed with conventional detergents. In each case, signals were observed with m/z values corresponding to two saposin A molecules and multiple detergent molecules. The data obtained with MNA C12 suggest formation of picodiscs containing 7 to 12 detergent molecules (Figure 5A-B). For MNA C13, the picodiscs contain 11 to 14 detergent molecules (Figure 5C), and for LMNG, the picodiscs contain 15 to 18 detergent molecules (Figure 5D). As shown in Figure 5A, the signal intensities for m/z values corresponding to monomer detergent or saposin A are much stronger than those assigned to intact picodisc assemblies; however, because Q-TOF mass spectrometers with microchannel plate (MCP) detectors are biased⁴³ toward species of small molecular weight, these relative intensities cannot be interpreted to reflect proportions of species in solution. In addition, we cannot know whether some or all dissociation of picodisc assemblies occurred during the ionization process.

Figure 5.

Native mass spectra of saposin A+MNA C12 (A and B), saposin A+MNA C13 (C), and saposin A+LMNG (D) assemblies. The full spectrum of the MNA C12 sample (A) is shown to indicate the region and intensity of the picodisc assembly peaks relative to those of monomeric saposin A and of isolated detergent molecules. We note that Q-TOF mass spectrometers with microchannel plate (MCP) detectors are biased toward species of low molecular weight; therefore, the relative intensities of signals for isolated detergent, monomeric saposin A and dimeric saposin A+MNA C12 assembly cannot be interpreted to represent relative populations of these species in solution. Panels B, C and D show only the data for the assemblies formed by saposin A with MNA C12, MNA C13 or LMNG, respectively. The green box in panel B shows a zoom-in of the 2685-2687 m/z region, highlighting the resolved charge state envelope. The red text indicates the charge and the black text indicates the number of detergents per assembly corresponding to the identified m/z value.

These native mass spectrometry data allow comparison with analogous studies of lipid-containing assemblies involving saposin A reported by others.^27,39-42 While SEC studies have suggested that saposin A lipoprotein discs can maintain their stability for up to 70 days,⁴⁴ native mass spectrometry data have shown the integrity of these systems is variable under pH-neutral conditions.²⁷ Composition and sample homogeneity are maintained for three hours after sample preparation at acidic pH, but at neutral pH the sample is heterogeneous after three hours.³⁹ In contrast, our data were collected up to five days after sample preparation, which indicates comparatively high stability of the saposin A-detergent picodisc assemblies. As SEC purification was performed as part of the sample preparation, we do not expect excess detergent was present in these samples and therefore not necessary for the retention of the picodisc composition. Even after several weeks, native mass spectrometry data were consistent with the picodisc compositions evident in earlier measurements (Figure S7).

Saposin B^W-Detergent Co-Assembly.

The studies outlined above identify several saposin A+detergent combinations that form stable co-assemblies at neutral pH. We wondered whether the co-assembly range could be further broadened by using alternative saposin modules. Saposins A, B, C and D are all of similar length (~80 residues), and they have similar isoelectric points (~4.2). Saposin A has been crystallized as a monomer in the absence of detergent, but saposins B (PDB 1N69),⁴⁵ C (PDB 2QYP )⁴⁶ and D (PDB 5U85)⁴⁷ have all been crystallized as dimers under detergent-free conditions. The structures of dimeric saposins B, C and D overlay well with saposin A in its open, LDAO-associated state (Figure S1), despite the absence of detergent in the structures of saposins B-D. The hydrophobic surfaces that line the central cavities of the saposin B-D dimers are similar to the inward-oriented hydrophobic surfaces of the saposin A dimer co-assembled with LDAO (Figure S8). It has previously been shown that saposin B, C and D can each form picodiscs with lipids, although assembly varies as a function of lipid composition and protein:lipid ratio.¹⁶ These considerations suggested that saposin B, C and/or D might co-assemble with detergents.

We used tryptophan fluorescence for initial evaluation of detergent interactions with engineered variants of saposin B (Figure 6), C (Figure S9) and D (Figure S10). Because none of these saposin proteins contains tryptophan, we designed and expressed mutant forms with a single tryptophan. The new proteins, saposin B R38W, saposin C K38W, and saposin D F38W, are designated saposin B^W, C^W and D^W. The position of the tryptophan mutation was chosen by overlaying each dimeric saposin structure (B, C or D) with saposin A in its open, LDAO-associated state (Figure S1). The residue that most closely overlaid with the position of the saposin A tryptophan residue was chosen as the tryptophan mutation site in saposin B, C and D (see SI for details). Tryptophan fluorescence emission spectra were collected for saposin B^W, C^W and D^W in the absence of detergent and for 1:10 protein:detergent mixtures involving the alkyl-β-D-maltopyranoside detergents, MNA C12, MNA C13, and LMNG (Figure 2). Detergent binding was assessed by comparing the emission spectra collected for the isolated protein and 1:10 protein:detergent protein mixtures. There was no significant change in the tryptophan fluorescence profiles of saposin C^W or D^W incubated with any of the detergents (Figures S8 and S9 respectively).

Figure 6.

Tryptophan fluorescence emission spectrum of saposin B^W (2 μM, black) overlaid with data collected from samples of saposin B^W (2 μM) incubated in a 1:10 ratio with detergent (20 μM) in pH 7.5 conditions (buffer = 25 mM Tris-HCl, pH 7.5, 150 mM NaCl). Panel A shows the emission spectra of saposin B^W combined with detergent DM (orange), DDM (red), TriDM (purple) or TetraDM (blue). Panel B shows the emission spectra of saposin B^W combined with MNA C12 (royal blue), MNA C13 (dark red) or LMNG (green).

Tryptophan fluorescence of saposin B^W mixed with MNA C12, MNA C13 or LMNG showed a pronounced blue shift in maximum and increase in intensity, relative to detergent-free conditions, suggesting that saposin B^W can co-assemble with each of these detergents (Figure 6). While the tryptophan fluorescence emission data for saposin A presented above suggest that micelle formation is required for picodisc assembly, the data for saposin B^W suggest that picodisc assembly for this protein can occur even in the absence of micelle formation by the detergent. The maximum blue shift and intensity increase in the tryptophan fluorescence emission of saposin B^W was observed even when the protein was incubated with DDM (CMC = 0.17 mM (ref. 34)) or TriDM (CMC = 0.024 mM (ref. 35)) at a detergent concentration below either CMC value (0.02 mM, Figure 6). As saposin B^W exhibits a dimeric conformation at neutral pH, we hypothesize that micelle formation is not necessary for detergent picodisc formation in this case because saposin B^W presents an “open” structure even in the absence of detergent. This proposed behavior of saposin B^W is distinct from the behavior of saposin A suggested by Nietlispach and coworkers, who proposed that DDM promotes a closed-to-“open-like” conformational change. We suggest that micelle formation is required for this saposin A conformational change.¹⁸ SEC elution profiles of saposin B^W incubated with MNA C12, MNA C13 or LMNG confirmed the formation of a species larger than that of the saposin B^W dimer (Figure S11), as expected for a dimeric assembly that includes multiple detergent molecules.

Structural Analysis via Small Angle X-Ray Scattering (SAXS).

SAXS has previously provided low-resolution structural insights on nanodiscs¹ and saposin A+lipid assemblies.¹⁶ Here, we employed SEC-coupled SAXS to evaluate the shapes and relative sizes of the saposin A monomer, saposin A+detergent assemblies involving MNA C12, MNA C13 or LMNG, the saposin B^W dimer, and saposin B^W+detergent assemblies involving MNA C12, MNA C13 or LMNG. The SEC-SAXS experimental protocol consists of SEC purification of the isolated protein or saposin+detergent prior to SAXS data collection. The eluent from the SEC column flowed into the SAXS sample chamber, and data were collected every other second. The SEC elution profiles of the samples indicate the presence of protein in the eluent though the presence of detergent could not be verified as the detergent does not absorb in the 280 nm region. By selecting the data that correspond to the top of the elution peak for averaging and analysis, we hoped to minimize the influence of possible heterogeneity in sample preparation. SEC-SAXS data provide a basis for characterization of the saposin+detergent structures (Figure 7).

Figure 7.

The experimental SEC-SAXS profiles of saposin A (A), saposin A+MNA C12 (B), saposin A+MNA C13 (C), saposin A+LMNG (D), dimeric saposin B^W (E), saposin B^W+MNA C12, (F) saposin B^W+MNA C13 (G), and saposin B^W+LMNG (H) are shown in black. Comparison of the saposin A monomer SEC-SAXS curve with the simulated data for the monomeric saposin A (PDB 2DOB, purple) is shown in panel A. The saposin A and B^W+MNA C12 profiles are compared with the simulated data of the saposin A+LDAO picodisc (PDB 4DDJ, teal) in panels B and F respectively. The profile of dimeric saposin B^W is compared with the simulated data of dimeric saposin B (PDB 1N69³⁵, light green) in panel E. All simulated datasets were generated by CRYSOL.

The SAXS signature of saposin A (Figure 7A) is shown overlaid with a simulated dataset (Figure 7A, purple) generated with CRYSOL⁴⁸ based on the crystal structure of saposin A (PDB 2DOB), which shows a monomeric protein molecule in the closed form. The experimental data for saposin A in absence of detergents (Figure 7A) overlay very well with the simulated data for monomeric saposin A, and these experimental data are clearly different from the experimental data corresponding to the saposin A+detergent assemblies (Figure 7B-D). The signatures of the saposin A+detergent assemblies show oscillation in the q region between 0 and 0.3 Å⁻¹. This profile resembles that of micelle X-ray scattering,⁴⁹ indicating that detergent is present. The saposin A+MNA C12 assembly (Figure 7B) is shown overlaid with the simulated curve generated from the crystal structure of the saposin A+LDAO assembly (PDB 4DDJ), which represents the open, dimeric form of the protein (Figure 7B+F, teal). The saposin A+detergent assemblies (Figure 7B-D) all give rise to signatures that correspond more closely to the simulated data for the saposin A+LDAO picodisc than the simulated data for monomeric saposin A. These observations suggest that the picodiscs formed by saposin A in the presence of MNA C12, MNA C13, and LMNG are of similar size and shape to the saposin A+LDAO picodisc. However, because of the limitations of SEC-SAXS as well as the limitations of CRYSOL’s ability to predict the hydration shell of a system, these data do not allow us to generate high-resolution structural models for the new saposin A+detergent assemblies. The SEC-SAXS data for the saposin A+MNA C12 combination displays a minimum at a slightly higher q value relative to the minimum in the data for the MNA C13 and LMNG combinations, which suggests that the picodisc formed with MNA C12 is slightly smaller than those formed with the other two detergents. These comparative analyses support the conclusion that assemblies containing two saposin A molecules surrounding a bilayer of detergent molecules, similar in size and shape to the saposin A+LDAO picodisc characterized via crystallography, are generated with each of the three double-tail detergents.

A simulated data set was generated from CRYSOL for the dimeric saposin B crystal structure (PDB 1N69; no detergent) consisting of Chains A and B in the asymmetric unit. This data compared with experimental data to discern whether SAXS can be used to distinguish between an 'empty' dimer and a dimeric protein+detergent assembly. The simulated data generated for saposin B dimer (Figure 7E, light green) provides a reasonably close overlay with experimental SAXS data for saposin B^W in the absence of detergent. However, the experimental SAXS data for saposin B^W suggests that the species in solution is slightly larger than the crystal structure of the saposin B dimer would suggest. This difference may reflect a slight conformational change induced by the non-native tryptophan residue, or this difference could indicate that the dimeric structure in solution differs slightly from the dimer in the crystal. The SAXS signatures obtained for the saposin B^W+detergent assemblies appear more similar to the simulated data for the saposin A+LDAO picodisc (Figure 7B and F, teal) than to the simulated data for the detergent-free saposin B^W dimer. Comparing SEC-SAXS data for detergent-free saposin B^W with data for saposin B^W in the presence of MNA C12, MNA C13 or LMNG clearly shows that a significant change in shape and size results from the addition of each detergent (Figure S12).

Overall, the three analytical techniques we have employed provide a self-consistent picture of assemblies formed by saposin B^W with the double-tail detergents. The tryptophan fluorescence data (Figure 6) demonstrate that MNA C12, MNA C13 and LMNG associate with saposin B^W, SEC data (Figure S11) show that this association leads to an increase in apparent molecular weight, and the SAXS data indicate a change in overall shape upon assembly with the detergent. Because the detergent-free saposin B^W dimer has an interior space lined by a hydrophobic surface (PDB 1N69), we conclude that the saposin B^W+detergent assemblies are structurally similar to the saposin A+LDAO picodisc (PDB 4DDJ).

Generating models of the picodisc assemblies from the SAXS data is complicated by the presence of the detergent molecules. Ab initio models generated using DAMMIF⁵⁰ based on the SEC-SAXS data display interesting characteristics. Twenty independent models were generated for each SEC-SAXS dataset, and averaged models were acquired by the program suite DAMAVER which produces a frequency map based on the alignment of the independent models. The program suite DAMFILT considers the frequency map in order to generate a compact, more probable structure. The DAMAVER and DAMFILT models for the saposin A monomer, saposin A+detergent assemblies, saposin B^W dimer, and saposin B^W+detergent assemblies are shown in Figure S13 alongside the crystal structures of saposin A (PDB 2DOB), saposin B (PDB 1N69), and the saposin A+LDAO picodisc (PDB 4DDJ). The DAMFILT models of the saposin A+detergent and B^W+detergent assemblies all display a lack of density at the center of the model. As the electron density or X-ray scattering length of alkyl chains is expected to be lower than that of the solvent, this absence of density could be explained by the presence of detergent molecules. A comparable phenomenon was observed in SAXS-based models of saposin A+lipid picodiscs.¹⁶ These saposin+detergent models appear to have additional electron density surrounding the empty center than would be explained by the presence of the saposin proteins. As the scattering length of glucoside groups is expected to higher than that of protein motifs, it is possible that the DAMFILT models are displaying regions occupied by the sugar head groups of the detergent molecules. While we cannot obtain high-resolution structural data from these SEC-SAXS datasets, the models are consistent with our hypothesis that saposin A and saposin B^W form picodisc assemblies in which a protein dimer surrounds a region populated by segments with low X-ray scattering, such as the alkyl chains of the detergents.

Discussion

We have shown that both saposin A and saposin B^W can associate with the branched detergents MNA C12, MNA C13 and LMNG at neutral pH to form very stable picodisc assemblies. In contrast, only weak and transient association occurs with the conventional detergent DDM, which is widely employed in the membrane protein community, or with homologues of DDM. We predict that the new picodisc systems established here will prove to be useful tools for the study of membrane proteins because the use of detergents as the amphiphilic component allows rapid equilibration in aqueous solution. We also expect that these detergent systems will retain the abilities of their lipoprotein counterparts to assemble discs of variable sizes,¹⁹ by addition of saposin A monomers, when incubated with guest membrane proteins though we can’t predict their exact stoichiometries based on our empty disc studies. In contrast, we predict saposin-lipid assemblies should be kinetically trapped entities, since the lipids have very low intrinsic aqueous solubility and liposomes present kinetically trapped states.⁵¹ Poor lipid solubility may explain why saposin A+lipid assemblies display relatively low stability at neutral pH.²⁶

We were surprised by the pronounced difference between conventional detergents, DDM, TriDM and TetraDM, and those with unusual structural features such as double hydrophobic tails and branched headgroups. Empirical evidence suggests that branched detergent architectures can lead to improved utility with membrane proteins,^29-31,36,37 although the origin of the improvement is not yet clear. Very recent molecular dynamics simulations comparing LMNG and DDM suggest that branching in the LMNG structure restricts mobility within the detergent molecule, and the resulting rigidity leads to more favorable hydrophobic interactions with a membrane protein, relative to DDM.³¹ More favorable interactions between the LMNG alkyl chains and the interior hydrophobic surface of a saposin dimer could at least partially account for the observed differences in stability between the DDM- and LMNG-based picodiscs.

The improved stability of saposin picodiscs formed with MNA C12, MNA C13 or LMNG relative to those formed with conventional detergents may arise in part because the structures of the branched detergents are better suited to bilayer formation than are conventional detergents. A detergent’s preference for forming a micelle, cylinder, or bilayer upon aggregation can be inferred from the packing parameter.²⁸ The packing parameter is given by the expression v_o/l_ca_e, where v_o represents the volume of the hydrophobic tail(s), l_c represents the length of the hydrophobic tail(s), and a_e is the equilibrium area per molecule at the aggregate surface. A micelle is the preferred aggregation state of amphiphiles with a packing parameter ≤1/3, a cylinder is the preferred aggregation state of amphiphiles with packing parameter between 1/3 and 1/2, and a bilayer is the preferred aggregation state of amphiphiles with packing parameter >1/2.^28,38 For most amphiphiles, v_o/l_c is approximately 21 Ų if there is a single hydrophobic tail and 42 Ų if there are two hydrophobic tails.³⁸ The parameter a_e depends on the size of the head group and on interactions between head groups.^28,38 Inspection of the LMNG and DDM structures (Figure 2) might suggest that a_e for LMNG is twice a_e for DDM, which would mean that the two have identical packing parameters. However, recent molecular dynamics simulations of detergent micelles³¹ suggest that intramolecular H-bonding between the two maltoside groups in the LMNG head group occurs to a greater extent relative to intermolecular H-bonding between maltoside head groups of neighboring DDM molecules. The internal H-bonding allows the LMNG molecules to pack more tightly relative to DDM molecules.³¹ These simulations therefore suggest that a_e of LMNG less than twice a_e of DDM, which would mean that the packing parameter of LMNG is higher than that of DDM, and that LMNG has a greater propensity to form a bilayer assembly relative to DDM.

The work reported here establishes new strategies for formation of picodiscs that are very stable at neutral pH. These assemblies are based on either saposin A, which has been widely studied for this purpose, or an engineered variant of saposin B (saposin B^W), which has received little attention in this context. Most surprising among our findings is the discovery that detergents with branched structures (MNA C12, MNA C13 and LMNG) display very marked superiority to conventional detergents (DDM, TriDM and TetraDM) in terms of propensity for and stability of picodisc assemblies. These results highlight the benefits of ongoing exploration of non-traditional amphiphile architectures.^29-31,36,37 More broadly, this work should encourage the application of the new saposin-based picodisc systems for solubilization and functional and structural characterization of membrane proteins.