Molecular discrimination between two conformations of sphingomyelin in plasma membranes

Summary

Sphingomyelin and cholesterol are essential lipids that are enriched in plasma membranes of animal cells where they interact to regulate membrane properties and many intracellular signaling processes. Despite intense study, the interaction between these lipids in membranes is not well understood. Here, structural and biochemical analyses of Ostreolysin A (OlyA), a protein that binds to membranes only when they contain both sphingomyelin and cholesterol, reveals that sphingomyelin adopts two distinct conformations in membranes when cholesterol is present. One conformation, bound by OlyA, is induced by stoichiometric, exothermic interactions with cholesterol, properties that are consistent with sphingomyelin/cholesterol complexes. In its second conformation, sphingomyelin is free from cholesterol and does not bind OlyA. A point mutation abolishes OlyA’s ability to discriminate between these two conformations. In cells, levels of sphingomyelin/cholesterol complexes are held constant over a wide range of plasma membrane cholesterol concentrations, enabling precise regulation of the chemical activity of cholesterol.

Graphical Abstract

ETOC summary

The plasma membrane lipid sphingomyelin has two distinct conformations depending on the presence or absence of cholesterol

Introduction

A special relationship between sphingomyelin (SM) and cholesterol in the plasma membranes (PMs) of animal cells has been proposed to modulate many cellular signaling processes (Simons and Ikonen, 1997, Simons and Toomre, 2000). One such process is cholesterol homeostasis, which ensures optimal cholesterol levels in cellular membranes by precise regulation of its synthesis and uptake (Brown and Goldstein, 2009). When cholesterol levels drop below a threshold set-point, transcription factors called SREBPs are activated and upregulate expression of cholesterol biosynthetic enzymes, which leads to increased cholesterol synthesis, and the low-density lipoprotein receptor, which leads to increased uptake of exogenous cholesterol (Brown et al., 2018). As a result, cholesterol levels rise until an optimum concentrate on is reached, at which point SREBP activation is terminated and cholesterol synthesis and uptake declines.

The PM contains ~80% of total cellular cholesterol, however, SREBPs and the sensors that regulate SREBP activation are located in the endoplasmic reticulum (ER) membrane, which contains only ~1% of total cellular cholesterol (Lange and Steck, 2016). A carefully regulated lipid transport pathway between PM and ER allows cholesterol sensors in ER to monitor the cholesterol content of PM (Infante and Radhakrishnan, 2017). Regulation of PM-to-ER cholesterol transport is determined by the organization of cholesterol in PMs, which critically depends on the interaction of cholesterol with SM (Simons and Ikonen, 2000). The link between SM levels and cholesterol synthesis was initially revealed by a pair of studies where i) SM addition to cells increased cholesterol synthesis (Gatt and Bierman, 1980), and ii) SM removal from PMs of cells by treatment with sphingomyelinase (SMase), an enzyme that degrades SM, reduced cholesterol synthesis (Slotte and Bierman, 1988). Both of these effects were explained by a model where SM sequesters cholesterol in PMs, preventing cholesterol transport to ER membranes to shut down activation of SREBPs (Slotte et al., 1990, Scheek et al., 1997). Depletion of SM by SMase released some sequestered PM cholesterol, which then was free for transport to ER to shut down activation of SREBPs.

More recently, we have quantified distinct pools of PM cholesterol using two soluble bacterial toxins, Perfringolysin O (PFO) and Anthrolysin O (ALO), which bind PM cholesterol when it is accessible at the membrane surface but not when it is rendered inaccessible due to sequestration by SM and other phospholipids (Das et al., 2014, Chakrabarti et al., 2017). PM cholesterol is accessible to toxins only after the cholesterol concentration surpasses a threshold of 30-35 mole% of PM lipids (Das et al., 2013). Cholesterol in excess of this threshold is transported to ER to signal that the cholesterol requirements of the cell have been met and to terminate SREBP activation (Das et al., 2014, Infante and Radhakrishnan, 2017). About half of the inaccessible cholesterol, ~15 mole% of PM lipids, is sequestered by sphingomyelin (SM), and can be liberated by treating cells with SMase. The remaining inaccessible PM cholesterol is sequestered by other membrane factors. Our understanding of the role of accessible PM cholesterol has steadily improved, but we know less about how inaccessible cholesterol is sequestered. Here, we focus on how SM sequesters PM cholesterol to make it inaccessible to sensors and cellular signaling processes.

Results

OlyA senses SM/cholesterol complexes in membranes

Our strategy for probing the SM/cholesterol interaction was based on previous success in identifying PFO and ALO as sensors for accessible, but not SM-sequestered, cholesterol. We speculated that there may be other toxins that have the opposite ability, to bind SM-sequestered but not accessible cholesterol. A literature search identified several candidate SM-binding toxins, some of which had been reported to also require cholesterol to bind membranes (Anderluh and Macek, 2002, Bernheimer and Avigad, 1979, Bhat et al., 2013, Makino et al., 2017, Skocaj et al., 2014, Tomita et al., 2004, Yamaji et al., 1998). We overexpressed and purified three toxins from this group, namely Lysenin (Lys), Equinatoxin II (Eqt), and Ostreolysin A (OlyA), for analysis of their respective lipid specificities (Fig. 1a). In these initial experiments, we used a version of SM containing a C16 ceramide base and an amide-linked oleoyl (18:1) acyl chain (18:1 SM, see Fig. S1a). Pelleting assays showed that Lys and Eqt bound to liposomes containing SM, and their binding did not change in the presence of either cholesterol or epicholesterol, a diastereomer of cholesterol (Figs. S1a, 1b, lanes 1-3). Binding was eliminated when SM was replaced with dioleoyl-phosphatidylcholine (DOPC), another choline-containing phospholipid (Figs. S1a, 1b, lanes 4, 5). In contrast, OlyA bound to membranes only when they contained both SM and cholesterol (Fig. 1b, lanes 1-5). Thus, we focused on OlyA to gain insight into the SM/cholesterol interaction.

Figure 1. Lipid binding specificity of SM-sensing proteins.

a, SM sensors. Coomassie staining of purified recombinant Lys-His₆, His₆-Eqt, and OlyA-His₆ (2 μg each) b, Lipid specificity. After incubation of SM sensors with liposomes of the indicated compositions for 1 h at room temperature, liposome-bound proteins were measured using a pelleting assay. c, Hemolysis inhibition assay to measure solution binding of SM and cholesterol to OlyA and ALOFL. In each reaction, the indicated lipids were dried on the sides of a tube, after which buffer A containing either OlyA (3 μM) or ALOFL (30 nM) was added. After overnight incubation, PlyB (10 nM) was added to tubes containing OlyA, and then buffer C containing ~3 × 10⁸ RBCs was added to all tubes. Following incubation for 30 min, hemolysis was assayed (n=3). For testing sequential addition, after overnight incubation of proteins with either SM or cholesterol, the contents of assay tubes were transferred to another tube containing dried cholesterol or SM, respectively, and the remainder of the assay was carried out as above. d-e, Dependence of OlyA binding on membrane lipid mobility. d, Supported bilayers of the indicated lipid compositions containing TR-DHPE (fluorescent lipid) were generated in glass-bottom 96-well plates and lateral fluidity of lipid molecules was measured by FRAP before and after fixation with osmium tetroxide. Gray bar denotes the 30s photobleaching step. Shown are averages of fluorescence values from three different regions in a well. e, Unfixed and osmium tetroxide-fixed supported bilayers of the indicated lipid composition were incubated with 5 μg of fOlyA for 30 min, after which membrane-bound fOlyA and TR-DHPE was measured (n=4). f-g, Dependence of OlyA binding on SM:cholesterol ratio. f, Binding to liposomes containing SM and varying molar concentrations of the indicated sterol was measured as in b, except that the incubation time was 4 h (n=3). g, Binding to lipid films. Ethanolic solutions of varying molar ratios of SM and the indicated sterol were prepared and 80 nmol of each mixture was deposited on nitrocellulose membranes and allowed to dry for 5-10 min. Each membrane strip was then subjected to dot blot analysis to measure bound OlyA (n=3). Chol., cholesterol; Epi., epicholesterol.

We considered two models to explain the dual specificity of OlyA for SM and cholesterol in membranes. In one model, OlyA contains distinct binding sites for SM and cholesterol, respectively, and these two sites may or may not be allosterically linked. In another model, OlyA contains a single binding site for a pre-existing complex of SM and cholesterol in membranes. We used two approaches to distinguish between these models. In one approach, we used a competition assay to test whether pre-incubation with SM alone, cholesterol alone, or mixtures of the two lipids, would prevent the binding of OlyA to red blood cells (RBCs), whose membranes contain high concentrations of SM and cholesterol (de Gier and Van Deenen, 1961). While OlyA binding does not disrupt RBC membranes, membrane-bound OlyA can recruit a cofactor, pleurotolysin B (PlyB), to form a transmembrane pore and cause hemolysis (Fig. S1b,c). Using PlyB-mediated hemolysis as a readout for OlyA binding, we observed that pre-incubation with either SM or cholesterol did not block OlyA binding to RBCs (Fig. 1c). In contrast, pre-incubation with a 1:1 mixture of SM and cholesterol resulted in a complete block of OlyA binding to RBCs (Fig. 1c). Sequential pre-incubation, first with SM and then with cholesterol, or vice versa, did not block OlyA binding. A different result was obtained with cholesterol-binding and pore-forming full-length ALO (ALOFL), hemolysis by which was inhibited by pre-incubation with any of the conditions that included cholesterol irrespective of the order of addition (Fig. 1c). In a second approach, we tested the effect of immobilization of membrane lipids on OlyA binding. If OlyA was sequentially binding SM and cholesterol at two distinct binding sites or was inducing SM/cholesterol complex formation, one might expect that restricting lipid motion would disrupt the membrane binding of OlyA. We generated a fluorescently labeled version of OlyA (designated as fOlyA, Fig. S1d, lanes 1,2) and also prepared supported lipid bilayers on glass surfaces. The lipid molecules in both SM and SM/sterol bilayers were fluid, as judged by fluorescence recovery after photobleaching (FRAP) measurements (Fig. 1d). Recovery was slower in sterol-containing bilayers compared to bilayers without sterols. Fixation with osmium tetroxide completely blocked the fluidity of all bilayers (Fig. 1d), a result consistent with previous studies (Jost et al., 1973). We then measured fOlyA binding and found that fOlyA only bound to supported bilayers containing both SM and cholesterol and this binding was unaffected by osmium tetroxide fixation (Fig. 1e).

While not conclusive, both of the above experiments favor a model where OlyA contains a single binding site for a pre-formed SM/cholesterol complex in membranes. To further define this complex, we measured the cholesterol concentration dependence for OlyA binding. Binding of OlyA increased in a sigmoidal fashion as cholesterol content of liposomes increased up to 60 mole% (Fig. 1f), which is close to the solubility limit of cholesterol in bilayers containing choline-phospholipids (Huang et al., 1999). No binding was observed when cholesterol was absent or replaced with epicholesterol. To investigate higher cholesterol concentrations, we deposited ethanolic mixtures of SM and cholesterol as lipid films on nitrocellulose membranes. This approach allowed us to assay the entire range of SM/cholesterol ratios, and we found that OlyA binding showed a sharp maximum at 65 mole% cholesterol (Fig. 1g).

Crystal structure reveals lipid binding site in OlyA

To gain further insight into OlyA’s interaction with SM and cholesterol, we performed structural analysis. A modified version of the OlyA construct used for the above biochemical studies, designated as OlyA(WT) (Fig. S2a), was purified (Fig. S2b, c) and crystallized in the absence or presence of saturating amounts of an equimolar mixture of SM and cholesterol. Crystals obtained from both conditions were used to determine the structure of OlyA(WT) at resolutions of 1.15 Å (no ligand, Fig. S2d) and 1.33 Å (with ligand, Fig. 2a) (Table 1). Both OlyA(WT) structures showed a β-sandwich fold, similar to that observed for a related toxin, Pleurotolysin A (Lukoyanova et al., 2015). The β-sandwich is comprised of nine β-sheets and 1 short α-helix that are connected by 4 flexible loops on the COOH-terminal end and 3 flexible loops on the NH₂-terminal end. Crystals obtained with lipid ligands contained additional electron density in a shallow channel bounded by the sidechains of W28 and K99, residues that lie on two separate loops on the NH₂-terminal end (Fig. 2a,b). This density was strong enough to be modeled in only one of the four monomers in the crystallographic asymmetric unit. Part of the SM ceramide base (Fig. 2c, red) was tentatively assigned to this observed density, and a possible orientation of the rest of SM is shown for reference (Fig. 2a). No other parts of SM, cholesterol, or crystallization buffer components could adequately account for the observed density. This assignment of a portion of SM is still not definitive as the elongated electron density could be another hydrophobic molecule that co-purified with OlyA(WT). Even with these caveats, it is noteworthy that the only clear bound density may correspond to the ceramide portion of SM, which is the main structural difference between SM and PC (Fig. S1a), lipids for which OlyA(WT) shows all-or-none specificity.

Figure 2. Structural analysis of SM binding by wild-type OlyA.

a, Overall structure of OlyA (teal) bound to a portion of SM (yellow sticks, see c). Possible orientations for the rest of OlyA-bound SM are shown as yellow ovals. b, Close-up view of a surface representation of the shallow channel formed by K99 and W28 (boxed region in a) which houses the observed electron density (gray mesh, see c). Superimposed on the density is the modeled portion of SM (yellow sticks, see c). c, Shown in gray mesh is the ∣mFo – DFc∣ electron density calculated after omitting the ligand from the model and contoured at 2.5σ. A portion of 18:1 SM (highlighted in red) was superimposed on this electron density. d, Close-up view of the region of OlyA with bound SM. Side-chains of amino acids that lie within 5 Å of the modeled portion of SM (yellow) are shown as sticks (teal) against a semi-transparent main chain backbone (light teal). e, Binding properties of OlyA mutants. Indicated His₆-tagged mutant versions of OlyA were overexpressed, purified, and their lipid specificities were measured using liposome dot blot assays. The mean value (n = 3) for binding of OlyA(WT) to 1:1 SM:cholesterol liposomes was set to 1. All other mean binding values (n = 3) were normalized relative to this set-point and converted to a green-to-red color scale. Standard errors for all measurements were less than 10%. Chol., cholesterol; Epi., epicholesterol; N, NH₂-terminus; C, COOH-terminus.

Table 1.

Data collection and refinement statistics for OlyA

Data collection
Crystal	OlyA(WT)	OlyA(WT) + lipids	OlyA(E69A) + lipids
PDB accession code	6MYI	6MYJ	6MYK
Space group	P21	P21	C2
Cell constants	46.43 Å, 100.33 Å, 59.02 Å, 106.43°	46.43 Å, 100.56 Å, 58.81 Å, 106.29°	69.86 Å, 86.68 Å, 92.88 Å, 99.13°
Wavelength (Å)	0.97926	0.97926	0.97903
Resolution range (Å)	33.30 – 1.15 (1.17 – 1.15)	44.56 – 1.33 (1.36 – 1.33)	49.27 – 1.80 (1.83 – 1.80)
Unique reflections	181,403 (8,891)	117,866 (5,878)	47,399 (1,701)
Multiplicity	6.4 (6.0)	7.5 (7.2)	5.2 (3.9)
Data completeness (%)	98.6 (96.8)	99.4 (99.7)	93.4 (68.4)
Rmerge (%)a	4.4 (43.0)	4.9 (46.2)	3.9 (43.8)
Rpim (%)b	1.8 (31.3)	1.8 (17.0)	1.8 (23.3)
CC1/2 (last resolution shell)	0.78	0.91	0.89
I/σ(I)	37.7 (2.4)	41.6 (4.0)	38.2 (2.2)
Wilson B-value (Å2)	15.1	13.5	18.1
Refinement statistics
Resolution range (Å)	33.30 – 1.15 (1.19 – 1.15)	44.56 – 1.33 (1.36 – 1.33)	49.27 – 1.800 (1.86 – 1.80)
No. of reflections Rwork/Rfree	166,128/1,451 (7,433/66)	117,716/2,000 (8,086/140)	41,217/1,401 (2,181/77)
Data completeness (%)	90.7 (41.0)	99.3 (97.0)	81.4 (44.0)
Atoms (non-H protein/metalions/ligand/solvent/waters)	4,272/4/NA/44/626	4,253/4/10/80/554	4,204/4/NA/66/503
Rwork (%)	14.2 (20.5)	15.5 (20.8)	19.4 (25.4)
Rfree (%)	16.1 (26.9)	16.8 (22.8)	21.9 (31.3)
R.m.s.d. bond length (Å)	0.018	0.016	0.004
R.m.s.d. bond angle (°)	1.15	0.95	0.62
Mean B-value (Å2) (chains A/B/C/D/metalions/ligand/solvent/waters)	17.3/20.6/20.9/20.0/28.4 NA/28.7/29.8	19.7/19.9/19.9/17.1/32.5/63.7/29.8/27.0	15.7/27.1/25.3/27.9/31.740.4/36.1/29.1
Ramachandran plot (%) (favored/additional/disallowed)c	97.8/2.3/0.0	97.6/2.4/0.0	96.6/3.2/0.2
Clashscore/Overall scorec	2.44/1.06	1.62/0.91	1.66/1.13
Maximum likelihood coordinate error	0.09	0.11	0.21
Missing residues	A:1. B:1-3. C:1-2, 66. D:137-138.	A:1-2, 138. B:1-2, 138. C:1-3. D:1.	A:1, 137-138. B:1. C:1, 137-138. D:1, 137-138.

Data for the outermost shell are given in parentheses.

^aR_merge = 100 ∑_h∑_i∣I_h,i— ⟨I_h⟩∣/∑_h∑_i ⟨I_h,i⟩, where the outer sum (h) is over the unique reflections and the inner sum (i) is over the set of independent observations of each unique reflection.

^bR_pim = 100 ∑_h∑_i [1/(n_h - 1)]^1/2∣I_h,i— ⟨I_h⟩∣/∑_h∑_i⟨I_h,i⟩, where n_h is the number of observations of reflections h.

^cAs defined by the validation suite MolProbity (Chen, V.B., Arendall, W.B.A., Headd, J.J., Keedy, D.A., Immormino, R.M., Kapral, G.J., Murray, L.W., Richardson, J.S., Richardson, D.C. (2010) MolProbity: all-atom structure validation for macromolecular crystallography. Acta Cryst. D66, 12-21.).

To test the relevance of this structural model, we generated mutant versions of OlyA(WT) where all residues within 5A of the putative ligand density (13 in total) were individually mutated to alanines (Fig. 2d). Dot blot assays for protein-lipid binding (Fig. S2e,f) showed that replacement of D93, S94, T100, G70, T71, or T72 with alanines retained specific binding of OlyA to liposomes containing SM and cholesterol, while replacement of Q5, W6, W28, P95, or W96 with alanines completely abolished binding of OlyA (Fig. 2e). Mutation of K99 to alanine partially reduced OlyA binding, while mutation to oppositely-charged glutamate completely abolished binding. We also tested these OlyA mutants for their ability to lyse RBCs in a PlyB-dependent manner (Fig. S1b,c) and found that mutations that disrupted OlyA binding to membranes in the dot blot assay also disrupted their PlyB-mediated hemolytic capacity, whereas mutations that did not disrupt OlyA binding to membranes did not affect hemolysis (Fig. S3). It is worth emphasizing that OlyA binding is abolished by mutation of either W28 or K99, residues that form the boundaries of the shallow channel housing the putative SM density (Fig. 2b).

Unexpectedly, mutation of E69 to alanine eliminated the cholesterol specificity of OlyA(WT). This mutant protein, designated as OlyA(E69A), bound to liposomes containing SM whether or not they also contained cholesterol or epicholesterol (Fig. 2e). The specificity for SM was still intact, since no binding was observed to membranes containing DOPC. To further understand the role of E69, we mutated this residue to every other amino acid, purified each mutant protein, and tested for their lipid specificity. The cholesterol dependence for OlyA(WT) binding was abolished only when E69 was replaced by alanine or serine, both of which have similar small side-chains (Fig. 3a). However, replacement with even smaller glycine did not affect OlyA’s cholesterol specificity. Replacement with oppositely charged arginine or similarly charged but smaller aspartate completely eliminated OlyA binding. To understand how both the size and charge of E69 combine to confer cholesterol specificity on OlyA(WT)’s binding pocket, we performed structural analysis of OlyA(E69A), one of the two mutations that abolished cholesterol specificity. OlyA(E69A) was purified (Fig. S2g-i) and crystallized in the presence of lipids, using similar procedures as for OlyA(WT). In contrast to OlyA(WT), which only formed crystals in the presence of ammonium chloride, OlyA(E69A) only formed crystals in bis-tris buffers (pH 5.5) containing lithium sulfate. These crystals were used to determine the structure of OlyA(E69A) at a resolution of 1.80 Å (Table 1).

Figure 3. Structural analysis of SM binding by OlyA(E69A).

a, Effects of mutation of E69 on OlyA’s lipid specificity. Indicated His₆-tagged mutant versions of OlyA were overexpressed, purified, and their lipid specificities were measured using liposome dot blot assays. The mean value (n = 3) for binding of OlyA(WT) to 1:1 SM:cholesterol liposomes was set to 1. All other mean binding values (n = 3) were normalized relative to this set-point and converted to a green-to-red color scale. Standard errors for all measurements were less than 10%. b, Overall structure of OlyA(E69A) (purple) bound to bis-tris (yellow sticks, see d). c, Close-up view of a surface representation of the shallow channel formed by K99 and W28 (boxed region in b) which houses the observed electron density (gray mesh, see d). Superimposed on the density is the modeled bis-tris structure (yellow sticks, see d). d, Shown in gray mesh is the ∣mFo – DFc∣ electron density calculated after omitting the ligand from the model and contoured at 2.5σ. The structure of bis-tris (highlighted in red) was superimposed on this electron density. e, Overlay of the regions containing bound ligands in OlyA(WT) (teal) and OlyA(E69A) (purple). Side-chains of amino acids that have different orientations in the two structures are shown as sticks against a semi-transparent main chain backbone. Chol., cholesterol; Epi., epicholesterol; N, NH₂-terminus; C, COOH-terminus.

A single amino acid (E69) determines SM/cholesterol specificity of OlyA

The structures of OlyA(E69A) (Fig. 3b) and OlyA(WT) (Fig. 2a) were virtually superimposable (root-mean-squared deviation of just 0.26 Å for 137 aligned C-α atoms). We once again observed extra electron density in one of the four OlyA(E69A) monomers in the crystallographic asymmetric unit. The density was located in the same shallow channel bordered by W28 and K99 where we had previously observed density in the structure of OlyA(WT) (Fig. 3c). In the case of OlyA(WT), the density was assigned to a portion of SM containing the ceramide base (Fig. 2c). However, in the case of OlyA(E69A), the observed density was fit best by a molecule of bis-tris (Fig. 3d), which was present at high concentrations (100 mM) in the crystallization buffers for OlyA(E69A). Bis-tris was not used in the purification or crystallization buffers for OlyA(WT) crystals. Inasmuch as the structure of bis-tris (Fig. 3d) shares some common features with the structure of the ceramide base portion of SM (Fig. 2c), we speculate that the shallow channels in both OlyA(WT) and OlyA(E69A) constitute a binding “hot spot” that accommodates SM. When we compared the location and orientation of amino acids within 5Å of the center of the shallow channel, we observed that the E69A mutation resulted in three significant differences between the two structures (overlay in Fig. 3e). Two of the differences were in the orientation of the side chains of channel-bordering K99 and W28. The third difference was in the orientation of the side chain of Q5, which lies on the protein surface on the side opposite to the shallow channel and was previously shown to be important for membrane binding of OlyA(WT) (Fig. 2e).

To better understand the altered lipid binding surface of OlyA(E69A), we compared the binding affinities of OlyA(WT) and OlyA(E69A). A dose curve analysis showed that OlyA(E69A) had an ~100-fold greater concentration sensitivity than OlyA(WT) for binding to SM-containing membranes (Fig. 4a). When compared to OlyA(WT), the association rate of OlyA(E69A) for SM/cholesterol membranes was only ~3-fold higher (Fig. 4b), suggesting that a much slower dissociation rate from the membrane likely determines the increased binding affinity of OlyA(E69A).

Figure 4. Comparison of affinities and lipid specificities of OlyA(WT) and OlyA(E69A).

a, Concentration dependence. Binding of OlyA(WT) and OlyA(E69A) to liposomes of the indicated compositions was measured by dot blot assays (1 μg/ml of OlyA = 57.6 nM). The mean value (n = 3) for binding of each protein to liposomes composed of 1:1 SM:Chol. at the highest protein concentration was set to 100% and all other mean binding values (n = 3) were normalized relative to these set-points. b, Association rates. After incubation of fOlyA(WT) and fOlyA(E69A) with liposomes of the indicated compositions for various times, liposome-bound fOlyA was measured. 100% of control values for fraction of bound proteins were 0.33 for fOlyA(WT) and 0.604 for fOlyA(E69A) (n=3). c-d, Phospholipid and sterol specificity. Dot blot assays were used to measure the binding of OlyA(WT) and OlyA(E69A) to liposomes composed of equimolar mixtures of cholesterol and phospholipids with 18:1 acyl chains and the indicated headgroup (c, left), cholesterol and SM with the indicated amide-linked acyl chain length (c, right), or 18:1 SM and the indicated sterol (d). The mean value (n = 3) for binding of each protein to liposomes composed of cholesterol and 18:1 SM was set to 100%. All other mean binding values (n = 3) were normalized relative to this set-point. e, Temperature dependence. After incubation of OlyA(WT) and OlyA(E69A) with liposomes of the indicated compositions for 4 h at various temperatures, liposome-bound OlyA was measured using a pelleting assay (centrifugation steps were also carried out at various temperatures). 100% of control values for fraction of bound proteins were 0.604 for OlyA(WT) and 0.704 for OlyA(E69A) (n=3). Epi., epicholesterol; Dchol., dihydrocholesterol; Chol., cholesterol.

We then carried out a detailed analysis of the varying ligand specificities of OlyA(WT) and OlyA(E69A). First, we tested their specificity for various structural features of SM. Both OlyA(WT) and OlyA(E69A) bound to membranes containing cholesterol and ceramide-based SM (Fig. S1a), but showed no binding when the membranes contained glycerol-based phospholipids with choline, ethanolamine, serine, inositol, glycerol, or phosphatidic acid headgroups (Fig. 4c, left). No binding was observed to membranes containing PCs with 16:0-16:0, 16:0-18:1, 18:0-18:1, or 18:1-18:1 acyl chains (Fig. S4a), suggesting that the absolute specificity for SM over PC, both of which contain choline, was likely determined by SM’s ceramide base structure. We then examined the effect of varying the length of the amide-linked acyl chain of SM (Fig. 4c, right). Both OlyA(WT) and OlyA(E69A) bound strongly to membranes containing SM with 18:1, 24:0, or 24:1 amide-linked chains. Binding was reduced when membranes contained SM with shorter amide-linked chains and was completely abolished when the amide-linked chain was eliminated as in lyso-SM (Fig. S4b). The headgroup of SM likely plays other important roles in determining OlyA’s specificity since binding was eliminated when SM was replaced with erythro-sphingosine or naturally-derived mixtures of ceramides, gangliosides, or cerebrosides (Fig S4b).

To study the effects of PCs on SM/cholesterol interactions in membranes, we measured OlyA binding to three-component liposomes comprised of cholesterol, 18:1 SM, and PC with 16:0-18:1 acyl chains (POPC) (Fig. S4c). When cholesterol was absent or present at a low concentration of 20 mole%, OlyA(WT) showed no binding to POPC/SM membranes whereas binding of OlyA(E69A) increased in a linear fashion as the fraction of SM increased. When the cholesterol content was raised to 40 mole% or 60 mole%, we observed increased binding of OlyA(WT) as the fraction of SM increased and this binding reached a plateau value when the ratio of cholesterol to SM was between 1:1 and 2:1. At these higher cholesterol concentrations, the binding of OlyA(E69A) retained its linear dependence on the fraction of SM and did not show sharp plateaus. These results suggest the presence of SM/cholesterol complexes in membranes containing high amounts of competing phospholipids such as POPC, consistent with earlier studies showing that cholesterol interacts better with SM than other phospholipids (Mattjus and Slotte, 1996, Phillips et al., 1987).

We then tested the specificity of OlyA(WT) and OlyA(E69A) for various structural features of cholesterol (Fig. S1a). As expected, OlyA(E69A) showed no sterol specificity and bound equally well to SM-containing membranes without or with any of the panel of sterols tested here (Fig. 4d). In contrast, membrane binding of OlyA(WT) required distinct structural features of cholesterol. OlyA(WT) bound to membranes containing cholesterol, but not epicholesterol where the 3-hydroxyl is in the opposite α-orientation. Binding was reduced by > 80% when the structure of the steroid nucleus was altered (dihydrocholesterol, lanosterol, ergosterol). Modification of the iso-octyl side chain with a double bond (desmosterol) did not affect OlyA(WT) binding whereas a polar group (25-hydroxycholesterol) eliminated binding. Adding an ethyl group (sitosterol) increased binding by ~2-fold. The structural specificity is consistent with previously proposed requirements for optimal interactions of SM with cholesterol: i) 3-hydroxyl group in the β-orientation allowing for hydrogen bond formation with the amide nitrogen on SM (Slotte, 2016), ii) “smooth” steroid nucleus that can pack optimally with SM acyl chains (Bloch, 1983, Miao et al., 2002), and iii) non-polar iso-octyl side chain that anchors cholesterol in membranes and interacts with the acyl chain of SM (Bittman, 1997, Bloch, 1983).

We next probed the temperature dependence for binding to SM/cholesterol membranes and observed that binding of OlyA(WT), but not that of OlyA(E69A), was reduced by ~70% when the temperature was raised from 23°C to 37°C (Fig. 4e). In contrast, OlyA(WT) showed no binding to SM/epicholesterol membranes at any temperature, whereas OlyA(E69A) showed strong binding that was not affected by temperature. The circular dichroism spectra for OlyA(WT) and the cholesterol concentration dependence for OlyA(WT) binding were similar at 23°C and 37°C (Fig. S4d,e). This suggests that OlyA (WT)’s lowered binding to SM/cholesterol membranes at higher temperatures is not simply due to denaturation of the protein or changes in SM/cholesterol complex stoichiometries, but rather due to dissociation of SM/cholesterol complexes in the membrane. The strong temperature dependence of OlyA(WT) binding is consistent with previous scanning calorimetry analysis of model cholesterol/phospholipid membranes (Anderson and McConnell, 2001, Hinz and Sturtevant, 1972, McMullen et al., 1993) and measurements of membrane cholesterol accessibility to cyclodextrins at various temperatures (Radhakrishnan and McConnell, 2002), the results from which were accounted for in terms of cholesterol/phospholipid complexes with exothermic heats of formation of 6-9 kcal/mol of phospholipid.

Binding pockets in OlyA(WT) and OlyA(E69A) accommodate distinct conformations of SM

Our results so far suggest that OlyA(WT) binds SM/cholesterol complexes, but not uncomplexed SM, and that mutation of E69 to alanine abolishes this specificity. However, we have not observed any electron density in the OlyA(WT) structure that could be attributed to any part of cholesterol. This raises the possibility that OlyA(WT) directly binds only to SM and that cholesterol specificity arises indirectly because OlyA(WT) distinguishes SM’s conformation when bound to cholesterol from SM’s conformation when free from cholesterol. To address this possibility, we performed comparative docking simulations of a portion of SM containing part of the ceramide base (Fig. 5a, red) with portions of OlyA(WT) and OlyA(E69A) centered on the shallow lipid-binding channel (Fig. 2b, 3c).

Figure 5. Docking simulations for binding of SM to OlyA(WT) and OlyA(E69A) and chemical modification assays.

a, Chemical structure of 18:1 SM with the fragment used in simulations highlighted (red). b – e, Top-scoring models are shown for binding of SM (yellow spheres) to OlyA(WT) (b) and OlyA(E69A) (d). Plausible schematic models for binding of OlyA(WT) (c) and OlyA(E69A) (e) to SM in membranes are shown with proteins depicted as cartoons with transparent surfaces and the top scoring docking poses of bound SM (yellow sticks) oriented on the membrane surface with acyl chains extrapolated into membrane bilayer (yellow ovals). f, g, Chemical modification of OlyA(WT) and OlyA(E69A). OlyA(WT)-His₆ (f) or OlyA(E69A)-His₆ (g) were incubated with indicated liposomes for 3h at room temperature, after which liposome-bound OlyA proteins were subjected to modification with mPEG-MAL-5000 followed by immunoblot analysis. As controls, OlyA(WT)-His₆ or OlyA(E69A)-His₆ in solution were also subjected to mPEG-MAL-5000 modification. M, modified form of OlyA; U, unmodified form of OlyA. h, Schematic description of mPEG-MAL-5000 modification results from f and g.

The docking simulations generated distinct models for binding of the SM fragment to OlyA(WT) and OlyA(E69A). For OlyA(WT), the top 100 binding poses clustered to yield a binding model where the positively-charged choline of SM forms an ionic interaction with E69 and the rest of the SM fragment is stably nestled in the shallow channel bordered by W28 and K99 (top 10 poses in Fig. S5, topmost pose in Fig. 5b). Binding is further stabilized by an interaction between the nitrogen of the K99 sidechain and the carbonyl oxygen on SM’s amide-linked acyl chain. Using this model for binding of SM to OlyA(WT) at the membrane surface, we extrapolated the rest of SM’s acyl chains into the core of a hypothetical lipid bilayer. The result is schematically shown in Fig. 5c where OlyA(WT) is oriented such that its N-to-C axis forms an angle of 26° with the bilayer surface. In this orientation, the choline group of OlyA(WT)-bound SM is parallel to the bilayer surface, possibly interacting with the cholesterol headgroup and steroid nucleus.

For OlyA(E69A), the top 100 binding poses of the SM fragment once again clustered in the same shallow channel, but in an orientation opposite to that observed for OlyA(WT) (top 10 poses in Fig. S5, topmost pose in Fig. 5d). This difference arises due to elimination of E69 and alterations in the lipid-binding channel of OlyA(E69A) caused by the different configurations of W28 and K99 (Fig. 3c,e). The carbonyl oxygen on the amide-linked acyl chain of SM can still interact with the nitrogen of the K99 sidechain, however the choline of SM is no longer stabilized by E69 and flips to the opposite end of the channel. Using this different model for binding of SM to OlyA(E69A) at the membrane surface, we extrapolated the rest of SM’s acyl chains into a hypothetical bilayer. In this model, OlyA(E69A) is oriented such that its N-to-C axis is almost perpendicular to the bilayer surface, and the choline group of OlyA(E69A)-bound SM is also perpendicular to the bilayer in a conformation that may be adopted by SM when free from cholesterol (Fig. 5e). Steric clashes prevent fitting of the OlyA(E69A)-bound SM conformation into the binding channel of OlyA(WT). In contrast, there is no steric hindrance to fitting the OlyA(WT)-bound SM conformation into the binding channel of OlyA(E69A). We speculate that OlyA(E69A) can recognize both the cholesterol-free and cholesterol-bound conformations of SM whereas OlyA(WT) only detects cholesterol-bound SM.

To detect the distinct membrane-bound orientations of OlyA(WT) and OlyA(E69A) predicted by the docking simulations, we used a cysteine modification assay to probe for the accessibility of a sole engineered cysteine residue at OlyA’s COOH-terminus when bound to membranes. In solution, treatment with cysteine-reactive mPEG-MAL-5000 led to modification of both OlyA(WT) and OlyA(E69A) (Fig. 5f, g, lanes 1-4). However, when bound to SM/cholesterol liposomes, mPEG-MAL-5000 did not react with either OlyA(WT) or OlyA(E69A) (Fig. 5f,g, lanes 5-8). In contrast, when bound to 100% SM liposomes, OlyA(E69A) was modified by mPEG-MAL-5000 (Fig. 5g, lanes 9-12) to a degree similar to that observed for OlyA(E69A) in solution. Our interpretation of these results is shown by the schematic model in Fig. 5h. OlyA(E69A) can adopt two different orientations on membranes to recognize both cholesterol-free and cholesterol-bound SM conformations whereas OlyA(WT) only detects the cholesterol-bound conformation of SM. We did not observe any modification of OlyA(E69A) when it was bound to SM/cholesterol liposomes because most of the SM in these bilayers was likely complexed to cholesterol and not in a free form.

Further insight into OlyA’s SM-binding pocket was gained by comparison to structures of Sticholysin II (Stn), an ~20 kDa soluble protein that belongs to the equinatoxin family, members of which bind SM in membranes but show no requirement for cholesterol (Fig. 1b). Structures of Stn have been determined in its ligand-free state (PDB 1O71) and in a state bound to phosphocholine (POC), the headgroup of SM (PDB 1O72) (Mancheno et al., 2003). Although the primary sequences of Stn and OlyA are only 11% identical, their structures superimpose well with a root-mean-squared deviation of just 2.2 Å for 115 aligned C-α atoms. Moreover, the POC binding pocket in Stn is also bounded by two loops connecting β-strands, in a structurally equivalent location as the SM binding pockets in OlyA(WT) and OlyA(E69A). Despite these similarities, several key differences provided clues into the differential lipid specificity of these pockets (Fig. S6). The binding pocket in Stn is larger than in OlyA(WT) or OlyA(E69A), and is more electropositive, primarily due to an arginine residue (R51). Removal of the negatively-charged sidechain in OlyA(E69A) does not drastically change the electrostatic properties of the interior of the pocket, but it does enlarge the pocket in this mutant. It is possible that the pocket in OlyA, which is precisely formed to bind to the very specific conformation of SM found in SM/cholesterol complexes, loses this specificity when enlarged in the E69A mutant, and can accommodate a wider range of SM conformations, both bound to and free from cholesterol. Mutation of residues in or near the binding pocket of OlyA (Q5, W6, W28, P95, W96) likely deform the pocket enough to entirely abolish binding of SM (Fig. 2e).

SM/cholesterol complexes in plasma membranes are maintained at constant levels over a wide range of cholesterol concentrations

We next probed PMs of cultured cells for SM/cholesterol complexes using fluorescently-labeled versions of complex-binding OlyA(WT), cholesterol-insensitive OlyA(E69A), and nonbinding OlyA(W6A) (Fig. S1d). Binding of OlyA(WT) and OlyA(E69A) to CHO-K1 cells rose in a time-dependent manner, reaching saturation after ~30 min (Fig. S7a). Maximal binding of OlyA(WT) was ~3-fold higher at 4°C than at 37°C, co nsistent with its earlier observed temperature dependence (Fig. 4e). Maximal binding of OlyA(E69A) was temperature-independent and higher than that of OlyA(WT), while OlyA(W6A) showed no binding at either temperature (Fig. S7a). Dose curve analysis showed saturable binding of OlyA(WT) and OlyA(E69A) at 4°C, whereas OlyA(W6A) showed minimal binding (Fig. S7b). Binding of OlyA(WT) was undiminished when PM lipids were immobilized by osmium tetroxide fixation (Fig. 6a), suggesting that SM/cholesterol complexes may be present in PMs under normal conditions without inducement by OlyA(WT). As expected, binding of OlyA(E69A) was unaffected by PM fixation, and no binding was observed for OlyA(W6A) to unfixed or fixed PMs (Fig. 6a).

Figure 6. Organization of SM and cholesterol in PMs of CHO-K1 cells.

a, Dependence of OlyA binding on membrane lipid mobility. Lateral fluidity of lipid molecules in CHO-K1 cell membranes was measured by FRAP before and after fixation with osmium tetroxide (left). Gray bar denotes the 30s photobleaching step. Shown are averages of fluorescence values from three different regions in a well. Unfixed and osmium tetroxide-fixed wells were incubated with the indicated fOlyA (3 μM) for 30 min, after which membrane-bound fOlyA and TR-DHPE was measured (right) (n=4). b-d, On day 0, CHO-K1 cells were set up in medium B at a density of 2.5 × 10⁵ cells/60-mm dish (b, 6 dishes/replicate/condition; c-d, 12 dishes/replicate/condition). b, SMase treatment. On day 2, cells were switched to fresh medium B containing the indicated concentrations of SMase. After incubation for 30 min at 37°C, cells were washed and 2 dishes from each cond ition were incubated with 3 μM of the indicated sensor protein. After incubation for 1 h at 4°C, cells were harvested and PM-bound proteins were quantified. (n=3) c, Cholesterol modulation by serum and compactin. On day 1, cells were switched to either fresh medium B, medium C, or medium D containing the indicated serum without or with compactin. On day 2, cells were washed and binding of the indicated sensor proteins was carried out as in b for 6 dishes/replicate/condition. Cells from the remaining 6 dishes were pooled and used for PM purification and cholesterol quantification. (n=3) d, Cholesterol modulation by cyclodextrin. On day 2, cells were switched to fresh medium B (one group) or fresh medium C containing 0.01% - 2% (w/v) HPCD. After incubation for 1 h at 37°C, cells were washed and b inding of the indicated sensor proteins and quantification of PM cholesterol was carried out as in c (n=3). 100% of control values for bound ALOD4, OlyA(WT), and OlyA(E69A) were 5.7, 3.7, and 5.4 μg/mg protein, respectively. e, On day 0, the indicated cell lines were set up at a density of 2.5 × 10⁵ cells/60-mm dish (media described in Methods; 6 dishes/replicate/condition). On day 1, cells were switched to media containing the indicated serum without or with compactin. On day 2, some cells were further treated with SMase (100 mU/mL) or HPCD (1% w/v) as described above, after which all cells were washed and binding of 3 μM of the indicated sensor protein was carried out as described above. 100% of control values for bound ALOD4, OlyA(WT), and OlyA(E69A) for each cell line was as follows: SV-589 (4.9, 3.5, and 6.8 μg/mg protein); Neuro-2A (1.1, 1.8, and 2.6 μg/mg protein); ST88-14 (6.3, 14.3, and 18.2 μg/mg protein); MDCK (4.3, 3.8, and 5.2 μg/mg protein); Caco-2 (4.5, 8.3, and 10.2 μg/mg protein) (n=3).

We next tested the effects of modulating the cellular lipid composition on binding of OlyA(WT) and OlyA(E69A) to PMs. Binding assays were carried out at 4°C to eliminate internalization of OlyA, as has been reported previously (Skocaj et al., 2014), and allow for assaying the disposition of SM on just the PM. For comparison, we also measured the binding of ALOD4, a domain of ALO, which binds accessible cholesterol in PM but not SM-sequestered cholesterol (Chakrabarti et al., 2017, Infante and Radhakrishnan, 2017). We first altered SM levels in PMs of CHO-K1 cells by incubation with SM-degrading SMase. This treatment abolished binding of both OlyA(WT) and OlyA(E69A) (Fig. 6b), consistent with the inability of these proteins to bind membranes containing ceramide, the product of SMase treatment (Fig. S4b). In contrast, binding of ALOD4 increased after SMase treatment (Fig. 6b), a result in line with previous observations where cholesterol sequestered by SM was liberated upon SMase treatment and made accessible for binding to ALOD4 (Chakrabarti et al., 2017).

We then tested the effects of changing cellular cholesterol levels on PM cholesterol content and PM binding of the three lipid-sensing proteins (Fig. 6c). In lipoprotein-rich serum, PMs of CHO-K1 cells contained 44.1 mole% cholesterol and showed robust binding of ALOD4, which binds accessible cholesterol. When cells were switched to lipoprotein-deficient serum, their PM cholesterol content dropped to 35.2 mole% and ALOD4 binding decreased. Incubation with compactin, an inhibitor of cholesterol synthesis, further lowered PM cholesterol to 33.7 mole% and ALOD4 binding was diminished even more. In contrast, OlyA(E69A), which binds both SM/cholesterol complexes and free SM, showed no change in binding to PMs containing different amounts of cholesterol. Surprisingly, OlyA(WT), which only binds SM/cholesterol complexes, also showed no reduction in binding even when PM cholesterol content was reduced from 44.1 to 33.7 mole%. To further test the cholesterol dependence of OlyA(WT), we treated cells with a cholesterol-extracting cyclodextrin reagent (HPCD), which depletes PM cholesterol to a greater extent than the milder treatments of Fig. 6c (Das et al., 2014). HPCD treatment depleted PM cholesterol from 42.4 mole% down to 22.7 mole% (Fig. 6d), resulting in the expected sharp sigmoidal decline of ALOD4 binding. Binding of cholesterol-insensitive OlyA(E69A) was not affected even by the most severe cholesterol reduction. Similar to what was observed in Fig. 6b, the binding of OlyA(WT) was stable as PM cholesterol decreased from 42.4 mole% to 32.4 mole%. However, the binding of OlyA(WT) declined sharply as PM cholesterol declined even further to 22.7 mole%, likely due to dissociation of SM/cholesterol complexes under these conditions of extreme cholesterol depletion.

To test the generality of the OlyA binding results obtained with CHO-K1 cells, we also tested binding to five other cell types – human fibroblasts (SV-589), mouse neuroblasts (Neuro-2A), human schwannoma cells (ST88-14), canine kidney epithelial cells (MDCK), and human colon epithelial cells (Caco-2) (Fig. 6e). In all cases, binding of ALOD4 was reduced by more than 80% when cells were switched from lipoprotein-rich serum to lipoprotein-deficient serum along with compactin, whereas binding of OlyA(WT) and OlyA(E69A) was not reduced. Further depletion of cholesterol using HPCD lowered ALOD4 binding by more than 90% and also lowered OlyA(WT) binding by more than 80%, while leaving binding of cholesterol-insensitive OlyA(E69A) unaffected. Treatment of cells with SMase eliminated binding of both OlyA(WT) and OlyA(E69A) and increased binding of ALOD4. In the case of ST88-14 schwannoma cells, the cholesterol released by SMase treatment was consistently higher (~270% of control) compared to that observed in the other five cell lines tested (150-190% of control). Combined, these results suggest that SM/cholesterol complexes are present and maintained at stable levels in all six of these cell lines.

Discussion

The current studies highlight how distinct conformations of lipids in membranes can modulate their interactions with sensor proteins in an all-or-none fashion. In particular, we show here that SM adopts two conformations in PMs, one of which is induced by complex formation with cholesterol and the second of which is adopted when SM is free from cholesterol. We are able to discriminate between these two conformations of SM in membranes using a soluble lipid-sensing protein, OlyA. Our biochemical and structural analysis reveal that OlyA binds specifically to SM/cholesterol complexes but not to free SM, and that this specificity is controlled by a single glutamic acid residue near the binding pocket.

The idea of SM/cholesterol complexes in membranes is not new (Finean, 1953, Slotte, 1992, Radhakrishnan et al., 2001, McConnell and Radhakrishnan, 2003), however they have been resistant to direct detection and remain controversial (Huang and Feigenson, 1999). The difficulty in detecting SM/cholesterol complexes likely stems from their small sizes and short lifetimes in fluid bilayers of cellular PMs (Sezgin et al., 2017). (It is worth noting that many examples of molecular complexes in liquids with relatively well-defined structures but very short lifetimes have been described (Fayer, 2009)). In our study, these limitations are overcome by OlyA possibly trapping and stabilizing these complexes, allowing for direct measurement of their levels in PMs. Previous structural studies of SM-bound sensor proteins did not reveal cholesterol-induced changes in SM conformation since they focused on lysenin (De Colibus et al., 2012) and sticholysin (Mancheno et al., 2003), both of which do not require cholesterol for binding SM-containing membranes (Fig. 1b). There have also been elegant structural studies of Nakanori (Makino et al., 2017) and Pleurotolysin A (Lukoyanova et al., 2015), two proteins that may discriminate between free and cholesterol-bound SM in a manner similar to OlyA. Unfortunately, these earlier studies did not contain bound lipids in their reported structures. There have also been several studies of protein sensors such as PFO and ALO that discriminate between conformations of another lipid, cholesterol, at membrane surfaces (Flanagan et al., 2009, Gay et al., 2015), however there is no structural information on the interaction of these proteins with cholesterol.

Using OlyA, we find that SM/cholesterol complexes are present in PMs of six different cultured cell lines (Fig. 6b-e). The complexes are present at relatively constant levels even when cholesterol amounts fluctuate by ~10 mole% (Fig. 6b,c). This invariance is in line with a previous observation that levels of a SM-sequestered pool of cholesterol in PMs of human fibroblasts were maintained at constant levels (Das et al., 2014). Quantification of PM lipids in this previous study showed that total SM and SM-sequestered cholesterol constituted ~10 mole% and ~15 mole% of total PM lipids, respectively. This ratio of 1:1.5 is consistent with the SM:cholesterol complex stoichiometries inferred from our current studies (ranging from 1:1 to 1:2 SM:cholesterol in Figs. 1, 5, S4c). Sequestration of a fixed pool of PM cholesterol in complexes with SM could allow for precise, switch-like control over the levels of accessible PM cholesterol that can travel to the ER to ensure cholesterol homeostasis.

The sigmoidicity of the OlyA binding curves (Fig. 1f, S4e) suggests cooperativity in formation of SM/cholesterol complexes in membranes. Studies of phospholipid/cholesterol complexes in model membranes have yielded cooperativity values in the range of 3-12 (McConnell and Radhakrishnan, 2003). Using an average molecular area of ~45 Ų/lipid obtained in these earlier studies, the size of such cooperative complexes would range from ~500 – 1600 Ų, which is below optical resolution. In PMs, these complexes could coalesce in the presence of specific proteins to form larger microdomains that may be observed by OlyA staining (Skocaj et al., 2014, Kishimoto et al., 2016). However, large-scale phase separation is not necessary for SM to control cholesterol’s accessibility (chemical activity) through complex formation (McConnell and Radhakrishnan, 2003). The maintenance of a constant pool of SM/cholesterol complexes in PMs in the face of significant cholesterol depletion (Fig. 6c-e) raises the possibility of an active mechanism mediated by one or more proteins in the PM. Lipid conformation-specific probes such as OlyA and ALOD4 should allow for the identification of such proteins that may regulate lipid homeostasis and other cellular signaling events.

STAR METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Arun Radhakrishnan (arun.radhakrishnan@UTSouthwestern.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Bacterial strains

Recombinant protein overexpression was carried out in the E. coli BL21 (DE3) pLysS strain.

Cell culture

Hamster CHO-K1 cells (female) were cultured in Medium B and were maintained at 37°C in 8.8% CO₂. Human fibroblast SV-589 cells were cultured in Medium F and were maintained at 37 in 5% CO₂. Canine kidney epithelial (MDCK) cells (female) and human Schwannoma (ST88-14) cells were cultured in Medium G and were maintained at 37°C in 5% CO ₂. Mouse neuroblast (Neuro-2A) cells were cultured in Medium I and were maintained at 37°C in 5% CO₂. Human colon epithelial (Caco-2) cells (male) were cultured in Medium J and were maintained at 37°C in 5% CO ₂. If not specified, the sex of the animal from which the cell line was derived is not known.

METHOD DETAILS

Buffers and media

Buffer A contains 50 mM Tris-HCl (pH 7.5), 150 mM NaCl and 1 mM TCEP. Buffer B is buffer A supplemented with 1% (w/v) SDS. Buffer C is Dulbecco’s phosphate buffered saline (PBS) supplemented with 2% (v/v) LPDS and 1 mM EDTA. Buffer D contains 25 mM HEPES-KOH (pH 7.4), 150 mM NaCl and 0.2% (w/v) bovine serum albumin. Washing buffer is phosphate buffered saline supplemented with 0.05% Tween 20 (Sigma). Blocking buffer is washing buffer supplemented with 5% (w/v) non-fat dry milk. Medium A is a 1:1 mixture of Ham’s F-12 and Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 100 units/ml penicillin and 100 μg/ml streptomycin sulfate. Medium B is medium A supplemented with 5% (v/v) FCS. Medium C is medium A supplemented with 5% (v/v) LPDS (prepared as described in (Das et al., 2013)). Medium D is medium C supplemented with 50 μM compactin and 50 μM sodium mevalonate. Medium E is DMEM-low glucose supplemented with 100 units/ml penicillin and 100 μg/ml streptomycin sulfate. Medium F is medium E supplemented with 5% (v/v) FCS. Medium G is medium E supplemented with 10% (v/v) FCS. Medium H is DMEM-high glucose supplemented with 100 units/ml penicillin and 100 μg/ml streptomycin sulfate. Medium I is medium H supplemented with 10% (v/v) FCS. Medium J is medium H supplemented with 20% (v/v) FCS.

Plasmids

The gene encoding lysenin from Eisinia fetida (amino acids 1-297) with a COOH-terminal hexahistidine tag and flanking BamHI and EcoRI restriction sites was synthesized by GenScript (Piscataway, NJ) with a codon sequence optimized for efficient bacterial overexpression, and provided to us in the pUC57 cloning vector. This lysenin gene was excised and ligated into the pRSET B expression vector, and this construct is hereafter referred to as pLys-His₆. Expression plasmids encoding equinatoxin II from Actinia equina with a NH₂-terminal hexahistidine tag in the pET8c vector (pHis₆-Eqt) and ostreolysin A (OlyA) from Pleurotus ostreatus with a COOH-terminal hexahistidine tag in the pET21c+ vector were kindly provided to us by Dr. Kristina Sepcic (University of Ljubljana, Slovenia) (Skocaj et al., 2014). Using OlyA as template, we generated a derivative where the two native cysteines of OlyA were mutated to serines and a new cysteine was introduced at the COOH-terminus (C62S C94S S151C), hereafter designated as pOlyA-His₆. For crystallographic studies, we generated a derivative of OlyA where its two native cysteines were mutated to serines (C62S C94S), the COOH terminal His₆ tag was removed, and a NH₂-terminal His₆ tag was introduced followed by a Tobacco etch virus (TEV) protease cleavage site (ENLYFQG), hereafter designated as pHis₆-TEV-OlyA. An expression plasmid encoding pleurotolysin B (PlyB) from Pleurotus ostreatus in the pET3a vector was kindly provided to us by Dr. Michelle Dunstone (Monash University, Australia) (Lukoyanova et al., 2015). Using PlyB as template, we generated a derivative where an octahistidine tag was appended to the NH₂-terminus, hereafter designated as pHis₈-PlyB. A plasmid encoding TEV protease with a NH₂-terminal hexahistidine tag (pHis₆-TEV) was kindly provided to us by Jing Yang (University of Texas Southwestern Medical Center). Plasmids encoding full-length anthrolysin O (ALOFL) and domain 4 of anthrolysin O (ALOD4) have been described previously (Gay et al., 2015). Mutations in all constructs were generated by site-directed mutagenesis using a QuickChange II XL Site-Directed Mutagenesis Kit (Agilent). The integrity of each plasmid was verified by DNA sequencing of its entire open reading frame. All plasmids were transformed into BL21 (DE3) pLysS Escherichia coli-competent cells (Invitrogen). For recombinant protein production, a single colony from a freshly transformed plate was grown in Luria Broth media (Research Products International) containing 100 μg/ml ampicillin.

Protein Purification and Labeling.

ALOFL and ALOD4 were purified and fluorescently labeled as described previously (Gay et al., 2015). His₆-TEV overexpression was induced with 1 mM IPTG at 37°C for 3 h, and the recombinant protein was purified by nickel chromatography using the same procedures as described for ALOFL (Gay et al., 2015). Protein-rich fractions were pooled, concentrated, and then diluted with buffer A to lower the imidazole concentration to 35-55 mM. After addition of glycerol (20% v/v final concentration), aliquots of His₆-TEV were flash frozen in liquid nitrogen and stored at −80°C. Lys-His₆ overexpression was induced with 1 mM IPTG at 22°C for 20 h, and the recombinant protein was purified by nickel chromatography followed by gel filtration using the same procedures as described for ALOD4 (Gay et al., 2015) with one exception, post-lysis centrifugation was carried out at 25,000 × g. His₆-Eqt overexpression was induced with 1 mM IPTG at 37°C for 3 h, and the recombinant protein was purified by nickel chromatography followed by gel filtration using the same procedures as described for ALOFL (Gay et al., 2015) with one exception, post-lysis centrifugation was carried out at 20,000 × g.

OlyA-His₆ and His₆-TEV-OlyA overexpression was induced with 1 mM IPTG at 18°C for 16 h, and the recombinant protein was purified by nickel chromatography followed by gel filtration using the same procedures as described for ALOD4 (Gay et al., 2015) with one exception, post-lysis centrifugation was carried out at 25,000 × g. After purification, some aliquots of OlyA-His₆ were labeled with Alexa Fluor 488 C₅-maleimide dyes as described previously (Gay et al., 2015). Free dye was separated from labeled OlyA-His₆ by passing the reaction mixture over a hand-packed nickel column (~1 ml), and eluting labeled OlyA-His₆ using buffer A containing 300 mM imidazole. The eluted fluorescently-labeled OlyA-His₆ (fOlyA) was then subjected to gel filtration chromatography on a Superdex 200 column preequilibrated with buffer A. Protein labeling was verified by SDS-PAGE followed by imaging using a BioSpectrum scanner (Analytik Jena, Germany). After purification, His₆-TEV-OlyA was subjected to cleavage by TEV protease. In each cleavage reaction, 500 μg of His₆-TEV-OlyA was incubated with 50 μg of His₆-TEV at 4°C. After 16 h, the cleavage reaction was loaded onto a prepacked 1 mL nickel column (GE Healthcare), and cleaved OlyA was collected in the flow through. OlyA was concentrated in a 3,000 MWCO concentrator and subjected to gel filtration chromatography using a Superdex 200 column pre-equilibrated with Buffer A. Proteinrich fractions were pooled and concentrated to 40-50 mg/ml for crystallization studies. His₈-PlyB was overexpressed and purified as described (Lukoyanova et al., 2015). After the final gel filtration chromatography step in buffer A, all purified proteins that were not used for crystallographic studies were concentrated to 2-5 mg/ml, stored at 4°C, and used within one week.

Assays for binding of proteins to model liposome membranes.

Liposomes of the indicated compositions at final concentrations of 800 μM or 1600 μM (total lipid) were prepared as described previously (Gay et al., 2015, Sokolov and Radhakrishnan, 2010). Binding of Lys-His₆, His₆-Eqt, and OlyA-His₆ (unlabeled and fluorescently labeled versions) to liposomes was measured using a pelleting assay where centrifugation was used to separate liposome-bound proteins from unbound proteins. In these assays, reaction mixtures (200 μL of buffer A) containing 1560 μM liposomes (total lipid) and the indicated proteins (0.1 μg (6 pmol) of fluorescently labeled OlyA-His₆ or 1 μg of Lys-His₆ (29 pmol), His₆-Eqt (47 pmol), or unlabeled OlyA-His₆ (57 pmol)) were set up in 1.7 mL low-retention microcentrifuge tubes (Fisher Scientific). After incubation at room temperature for 4 h (except for the experiment in Fig. 1b, where the incubation time was 1 h), the reaction mixtures were subjected to centrifugation at 21,000 × g for 1h at room temperature. The resulting supernatants (200 μL) were collected and the pellets were resuspended in 200 μL buffer A. For assays with labeled proteins, equal fractions (100 μL each) of supernatants and pellets were transferred to a 96-well plate (Greiner Bio-One; black, flat-bottom, non-binding). 100 μL of buffer B was added to each well and the plate was placed on a shaker for 1h at room temperature, following which fluorescence from liposomes and labeled protein was measured using a microplate reader (Tecan M1000 Pro) using the following parameters – Texas Red (excitation wavelength, 595 nm; emission wavelength, 617 nm; band pass, 5 nm for each); Alexa Fluor 488 (excitation wavelength, 495 nm; emission wavelength, 516 nm; band pass, 5 nm for each). For assays with unlabeled proteins, aliquots (20 μL) of supernatants and resuspended pellets were subjected to immunoblot analysis. In some cases, the immunoblot LICOR intensities from the supernatants (unbound OlyA) and pellets (bound OlyA) were used to calculate the percentage of OlyA bound to liposomes. For some batches of purified OlyA, we occasionally observed significant pelleting in the absence of liposomes. In these cases, the microcentrifuge tubes were pre-incubated with 1 mL of a blocking solution (0.25 mg/ml bovine serum albumin in buffer A) for 30 min at room temperature, after which the blocking solution was removed, and OlyA-liposome binding reactions were carried out as above.

In some studies, binding of OlyA-His₆ to lipid membranes were measured using a dot blot assay, which allows for more rapid evaluation of protein-lipid binding than the pelleting assay. In these assays, 4 nmol of the indicated liposomes (5 μL of stocks prepared at 800 μM total lipid concentration in buffer A) were deposited on nitrocellulose membranes and allowed to dry for 5-10 min. The membranes were incubated in blocking buffer for 30 min at room temperature and then switched to blocking buffer containing the indicated His₆-tagged versions of OlyA (0.5 μg/ml of OlyA(E69A), 1 μg/ml of all other versions of OlyA). After incubation for 1 h, membranes were switched to washing buffer for 5 min, incubated with anti-His antibody in blocking buffer (1 μg/ml) for 1 h, and processed for immunoblot analysis to quantify liposome-bound OlyA-His₆. A detailed example of this assay is shown in Fig. S2e, f.

For measurement of association rates, we used liposomes containing 1 mole% 18:1 Biotinyl CAP PE for quick separation of bound and unbound OlyA using magnetic streptavidin beads (NanoLink^TM Streptavidin Magnetic Beads 1.0 μm). Aliquots containing 780 nmoles of indicated liposomes were incubated with 300 μg washed magnetic streptavidin beads for 30 min at room temperature. Preliminary experiments indicated that this amount of streptavidin beads was sufficient to capture all the added liposomes. Reaction mixtures (100 μL of buffer A) containing 1560 μM liposomes bound to streptavidin beads from above and 0.1 μg (6 pmol) of fluorescently labeled OlyA-His₆ (fOlyA) were set up in 1.7 mL low-retention microcentrifuge tubes. After incubation for indicated times at room temperature, the reaction mixtures were subjected to magnetic pull-down (DynaMag−2 magnet). The resulting supernatants (100 μL), containing unbound proteins were collected and the beads were washed once with 100 μL buffer A. The washed beads containing bound liposomes and proteins were resuspended in 100 μL buffer A. The supernatants and resuspended beads (100 μL each) were transferred to a 96-well plate (Greiner Bio-One; black, flat-bottom, non-binding). 100 μL of buffer B was added to each well and the plate was placed on a shaker for 1h at room temperature, following which fluorescence from liposomes (Texas Red) and labeled protein (Alexa Fluor 488) was measured using a microplate reader as described above for the pelleting assay.

Preparation of supported lipid bilayers in 96-well plates.

The glass surface of each well of a 96-well glass-bottom plate (Greiner Bio-One) was cleaned and processed by the following steps: i) wells were washed three times with water (350 μL each); ii) wells were treated with isopropanol (300 μL) for 30 min, followed by three washes with water (350 μL each); iii) wells were treated with 1 M NaOH (250 μL) for 1 hour, followed by five washes with water (350 μL each); iv) wells were dried under a stream of compressed air. Plates were then covered with lids, wrapped in aluminum foil, and used within two days. Supported lipid bilayers were generated on the processed glass surfaces of these wells by the following steps: i) wells were filled with water (80 μL) followed by addition of liposomes (30 μL of 1.6 mM stock) containing 0.2 mole% of TR-DHPE, a fluorescently labeled phospholipid; ii) after incubation for 1 hour, undeposited liposomes were removed by three successive washing steps, each of which consisted of adding buffer A (200 μL) to each well followed by removing a fraction of the well contents (110 μL of 310 μL total); iii) wells were then treated with a blocking agent (150 μL of 0.5 mg/mL BSA in buffer A) to prevent nonspecific binding; iv) after incubation for 1 hour, 200 μl of buffer A was added and unbound BSA was removed by three successive washing steps, each of which consisted of removing a fraction of the well contents (150 μL out of 350 μL total) followed by adding more buffer A (150 μL); v) after the last wash, 300 μL of the well contents was removed, leaving behind a supported lipid bilayer in a total volume of 50 μL of buffer A. All of the preceding steps were carried out at room temperature. Once liposomes were added to wells, caution was exercised to ensure that the glass surface and supported bilayer was not disturbed or exposed to air. For fixation studies, we added 200 μl of a 1:1 mixture of osmium tetroxide (4% stock in water) and PBS to each well. After incubation for 30 min in a chemical hood, 100 μl of buffer A was added to the well, after which the osmium tetroxide fixative was removed by serial dilution with ten successive washing steps, each of which consisted of removing a fraction of the well contents (200 μl of 350 μl total) followed by adding more buffer A (200 μl). Unfixed wells received the same treatment, except that 200 μl of a 1:1 mixture of water and PBS were added to the well during the 30 min fixation step. Membrane fluidity was measured by fluorescence recovery after photobleaching (FRAP). The success rate for generating fluid supported bilayers was ~75% for bilayers without sterols and ~20% for bilayers with sterols. Epifluorescence microscopy (Nikon Ti-E microscope, 60x objective) was used to monitor TR-DHPE from a circular region (~10 μm in diameter) for 30 s, after which a focused beam from a 561 nm laser source was used to photobleach TR-DHPE molecules in that region. After 30 s, the laser was turned off and TRDHPE fluorescence of the bleached region was monitored. The fluorescence before bleaching was set to 1 and the fluorescence after the 30 s bleaching step (~40-50% reduction) was normalized to 0.

Assay for binding of fluorescently labeled OlyA (fOlyA) to supported lipid bilayers.

Binding reactions were carried out in 96-well plates, the well surfaces of which were covered with supported lipid bilayers and processed as described above. To each well containing 150 μl of buffer A, we added 5 μg of fOlyA proteins (90 pmoles) in a total volume of 10 μl of buffer A. After incubation for 30 min, buffer A (190 μL) was added to increase the total reaction volume to 350 μL. Unbound fOlyA was then removed by serial dilution with ten successive washing steps, each of which consisted of removing a fraction of the well contents (200 μL of 350 μL total) followed by adding more buffer A (200 μL). Membrane-bound fOlyA and fluorescence from supported bilayers was measured using a microplate reader (Tecan M1000 Pro) using the following parameters – Texas Red (excitation wavelength, 595 nm; emission wavelength, 617 nm; band pass, 5 nm for each); Alexa Fluor 488 (excitation wavelength, 495 nm; emission wavelength, 516 nm; band pass, 5 nm for each).

Chemical modification of OlyA

We probed for accessibility of OlyA’s sole engineered COOH-terminal cysteine using mPEG-MAL-5000, a high molecular weight membrane-impermeable reagent that forms a covalent bond with exposed cysteines and causes proteins to migrate with higher molecular weight on SDS/PAGE (Le Gall et al., 2004). OlyA-liposome binding reactions were carried out as described earlier with the following modifications. Reaction mixtures (400 μL of buffer A) containing 640 nmol of liposomes and 2 μg (120 pmol) of OlyA(WT)-His₆ or OlyA(E69A)-His₆ were set up in 1.7 ml low-retention micro centrifuge tubes. After incubation for 3h at room temperature, the reaction mixtures were subjected to centrifugation at 21,000 × g for 30 min at room temperature. The resulting pellets were resuspended in 240 μL buffer A. Chemical modification reaction mixtures, in a final volume of 100 μL of buffer A, containing either resuspended pellets from the above step (60 μL) or 500 μg of soluble OlyA(WT)-His₆ or OlyA(E69A)-His₆ (in a final volume of 60 μL of buffer A), and the indicated amounts of mPEG-MAL-5000 (in a final volume of 40 μL of buffer A) were incubated in 1.7 ml low-retention microcentrifuge tubes at room temperature. After 2 hours, reactions were quenched by addition of 10mM DTT (1 M stock). Aliquots of the reaction mixtures (20 μL) were subjected to immunoblot analysis.

Hemolysis assays

Fresh rabbit blood was obtained from the Animal Resource Center at UT Southwestern Medical Center and hemolysis assays were carried out as described previously (Gay et al., 2015).

Hemolysis inhibition assays

These assays were used as an indirect measure of solution binding of SM and cholesterol to OlyA and ALOFL. In each reaction, 0.6 μmol of either 18:1 SM or cholesterol, or 1.2 μmol of an equimolar mixture of 18:1 SM and cholesterol, were dried on the sides of a tube, after which 50 μl of buffer A containing either OlyA (3 μM) or ALOFL (30 nM) was added. Following overnight incubation, tubes containing OlyA were supplemented with PlyB (10 nM) and then all tubes received 450 μl of buffer C containing ~3 × 10⁸ RBCs. After incubation for 30 min, each reaction was subjected to 2000 × g centrifugation for 15 min and an aliquot of the supernatant (100 μl) was assayed for released hemoglobin (absorbance at 540 nm). For testing sequential addition, after overnight incubation of proteins with either 18:1 SM or cholesterol, the contents of the assay tubes were transferred to another tube containing dried cholesterol or 18:1 SM, respectively, and the remainder of the assay was carried out as above.

Immunoblot analysis

After indicated incubations, samples from liposome pelleting assays were mixed with 5x loading buffer, heated at 95ΌC for 10 min, and subj ected to 10% SDS-PAGE (lysenin) or 15% SDS-PAGE (Eqt, OlyA). Samples containing lysenin were additionally incubated with 4M urea to break down membrane-bound oligomers. The electrophoresed proteins were transferred to nitrocellulose membranes using the Bio-Rad Trans Blot Turbo system. These membranes and nitrocellulose membranes from dot blot assays were subjected to immunoblot staining with anti-His antibody (1:1000 dilution). Bound antibodies were visualized by chemiluminescence (Super Signal Substrate; Thermo Fisher) by using a 1:5000 dilution of donkey anti-mouse IgG conjugated to horseradish peroxidase (Jackson ImmunoResearch, West Grove, PA). Filters were exposed to Phoenix Blue X-Ray Film (F-BX810; Phoenix Research Products, Pleasanton, CA) at room temperature for 1-30 s or scanned using an Odyssey FC Imager (Dual-Mode Imaging System; 2 min integration time) and analyzed using Image Studio ver. 5.0 (LI-COR, Lincoln, NE).

Crystallization, data collection, structure determination and refinement.

Crystals of OlyA(WT) were grown using the sitting-drop vapor-diffusion method from drops composed of 0.75 μl of buffer A containing 30 mg/ml OlyA(WT) and 1 μl of reservoir solution (100 mM ammonium chloride and 16% (w/v) PEG 3350) and equilibrated over reservoir solution at 20°C. Cryoprotection was performed by transferring the crystals to a final solution of 100 mM ammonium chloride, 50 mM Tris pH 7.5, 150 mM NaCl, 19% (w/v) PEG 3350 and 30% (v/v) ethylene glycol, and were flash-cooled in liquid nitrogen. OlyA(WT) crystals diffracted to a minimum Bragg spacing (d_min) of 1.1 – 1.3 Å and exhibited the symmetry of space group P2₁ with cell dimensions of a = 46.4 Å, b = 100.3Å, c = 79.7 Å, b = 106.4° and contained four OlyA(WT) molecules per asymmetric unit.

To generate crystals of ligand-bound OlyA(WT), we first dried down 80 μmol of an equimolar mixture of 18:1 SM and cholesterol on the sides of a 1.7 ml tube. We then added 200 μl buffer A containing 30 mg/ml OlyA(WT) and incubated the mixture overnight at room temperature on a rotator. We also dried down 50 μg each of 18:1 SM and cholesterol on a sitting-drop vapor-diffusion tray and set up crystals using the SM/cholesterol-saturated OlyA(WT) generated above.

Crystals of OlyA(E69A) were grown by the same method described above for ligand-bound OlyA(WT), but the reservoir solution was 0.1 M Bis-Tris (pH 5.5), 125 mM lithium sulfate and 20-23% (w/v) PEG 3350, and the final cryoprotectant solution was 0.1 M Bis-Tris (pH 5.5), 125 mM lithium sulfate, 25% (w/v) PEG 3350 and 25% (v/v) ethylene glycol. OlyA(E69A) crystals diffracted to a minimum Bragg spacing (d_min) of 1.8 Å and exhibited the symmetry of space group C2 with cell dimensions of a = 69.9 Å, b = 86.7Å, c = 92.9 Å, b = 99.1 ° and contained four OlyA(E69A) molecules per asymmetric unit.

All diffraction data were collected at beamline 19-ID (SBC-CAT) at the Advanced Photon Source (Argonne National Laboratory, Argonne, Illinois, USA) and processed in the program HKL-3000 (Minor et al., 2006) which applied corrections for i) effects resulting from absorption in a crystal and for radiation damage (Borek et al., 2003), ii) calculation of an optimal error model, and iii) compensating for the phasing signal for radiation-induced increase of non-isomorphism within the crystal (Borek et al., 2013). Crystals of OlyA(E69A) displayed anisotropic diffraction and while 99% complete to 2.2 Å resolution, the diffraction intensity fell off towards the high-resolution limit of 1.8 Å. Phases for OlyA(WT) were obtained by molecular replacement in the program Phaser (McCoy et al., 2007). The crystal structure of Pleurotus ostreatus PleurotolysinA (PDB ID 4OEB) (Lukoyanova et al., 2015) was modified for use as a search model by truncating non-identical residues to the last common atom. Completion of the protein model was performed by manual rebuilding in the program Coot (Emsley et al., 2010), and model refinement was performed in the program Phenix (Adams et al., 2010). For the OlyA(WT) data to 1.1 Å resolution, fully anisotropic atomic displacement parameters for all non-hydrogen protein atoms were refined. Data collection and structure refinement statistics are summarized in Table 1.

Docking Simulations

The binding of the SM ligand to OlyA receptors was studied using molecular docking simulations. For the docking ligand, we used a portion of SM containing the phosophocholine headgroup and part of the ceramide base and N-linked acyl chain (Fig. 5a, red). We did not include the rest of the ceramide or N-linked acyl chains to minimize non-specific associations of these hydrophobic, membrane-embedded flexible chains with soluble, membrane surfacescanning OlyA proteins. For the docking surface, we used portions of OlyA(WT) and OlyA(E69A) centered on the shallow lipid-binding channel (Fig. 2b, 3c) and extended out by 5 Å in all directions. The Maestro platform (version 10.3) was used to access modules of the Schrodinger software package (version 3.1, Schrodinger) for structure preparation and docking. Structures of OlyA(WT) and OlyA(E69A) were prepared using Protein Preparation Wizard (version 11.2, Schrodinger) and the PROPKA module to set the protonation state of the protein at pH 7.0. The center of the binding pocket in both proteins was defined by residues that lie within a 5Å radius of the putative SM density in the structure of OlyA (WT). 3D coordinates of the hydrophilic part of SM (ligand) were generated with LigPrep (version 3.5) using the EPIK module (version 3.3) to set the pH to 7.0 and the OPLS_2005 force field option, and the resulting SM structure was then docked to the structural models of OlyA(WT) and OlyA(E69A) using Glide standard precision (SP) scoring function (version 6.8, Schrodinger). The docking procedure yielded a single cluster of poses for each protein. The poses with the highest docking scores were chosen as representatives of the binding model.

Fixation of cells and measurement of FRAP and OlyA binding.

For fixation studies, CHO-K1 cells were set up on day 0 in medium B in 96-well glass-bottom plates at a density of 5000 cells/well. When cells had reached full confluency (usually day 2), media was removed and replaced with 200 μl medium B containing 1 μg/ml of a fluorescent lipid (TR-DHPE) (1 mg/ml stock in ethanol). After incubation for 1 h for incorporation of TR-DHPE, media was removed, cells were washed five times with PBS, and then incubated with 200 μl of a 1:1 mixture of osmium tetroxide (4% stock in water) and PBS. After incubation for 30 min in a chemical hood, the fixing solution was removed, cells were washed 5 times with buffer D, after which 200 μl of PBS was added to each well. Unfixed wells received the same treatment, except that 200 μl of a 1:1 mixture of water and PBS were added to the well during the 30 min fixation step. Membrane fluidity was measured by fluorescence recovery after photobleaching (FRAP). Epifluorescence microscopy (Nikon Ti-E microscope, 60x objective) was used to monitor TR-DHPE from a circular region (~5 μm in diameter) for 30 s, after which a focused beam from a 561 nm laser source was used to photobleach TR-DHPE molecules in that region. After 30 s, the laser was turned off and TR-DHPE fluorescence of the bleached region was monitored. The fluorescence before bleaching was set to 1 and the fluorescence after the 30 s bleaching step (40-50% reduction) was normalized to 0.

To measure binding of OlyA to unfixed and fixed wells, we used fluorescently-labeled OlyA (fOlyA) proteins. After the above fixation or control treatments, cells were subjected to two 10-min washes with buffer D at room temperature. After these washes, each well was incubated with 3 μM of fOlyA proteins in a total volume of 100 μl of buffer D. After incubation for 30 min, cells were washed six times with PBS to remove unbound fOlyA and membrane-bound fOlyA and lipid fluorescence from cell membranes was measured using a microplate reader (Tecan M1000 Pro) using the following parameters – Texas Red (excitation wavelength, 595 nm; emission wavelength, 617 nm; band pass, 5 nm for each); Alexa Fluor 488 (excitation wavelength, 495 nm; emission wavelength, 516 nm; band pass, 5 nm for each).

Quantification of PM cholesterol and bound fluorescently-labeled OlyA and ALOD4 proteins

After indicated treatments, PM membranes were purified and their cholesterol content was quantified as described previously (Das et al., 2013). For measurement of binding of fluorescently-labeled OlyA or ALOD4, cells were first washed after indicated treatments as follows to remove surface-bound lipoproteins or HPCD: three rapid washes with buffer D at room temperature, followed by two 10-min washes with ice-cold buffer D at 4°C (for binding measurements at 4°C) or warm buffer D at 37C (for binding measurements at 37°C). After these washes, each dish was incubated at 4°C or 37° C with 2 ml of buffer D containing fluorescent sensors as described in the Figure Legends. After indicated times, cells were washed three times with ice-cold or warm PBS, solubilized in 1 ml of buffer B, and shaken on a rotator at room temperature. Cell-bound fluorescently-labeled proteins were quantified as described previously (Infante and Radhakrishnan, 2017). Total cell protein was quantified using a BCA colorimetric assay.

QUANTIFICATION AND STATISTICAL ANALYSIS

All experiments were repeated on at least three different days with different batches of proteins and liposomes. Cell assays were repeated at least three different times with cells set up on different days. Number of replicates for each experiment (n) is indicated in the figure legends. All data were analyzed using Microsoft Excel and are presented as mean ± standard error of measurement (SEM). When not visible, error bars are smaller than the size of the data symbols.

DATA AVAILABILITY

The accession numbers for the data reported in this paper are PDB: 6MYI, 6MYJ, 6MYK.

Supplementary Material

1

Figure S1. Chemical structures of lipids, assay for OlyA-mediated hemolysis, and fluorescent labeling of OlyA proteins, Related to Figure 1

a, Chemical structures of phospholipids and sterols used in this study. Various structural features of sphingophospholipids (18:1 sphingomyelin) and glycerophospholipids (di(18:1) phosphatidylcholine) are labeled. The three carbons of the glycerol backbone are also indicated. b, OlyA-mediated hemolysis. Schematic of dual requirement of OlyA and PlyB for pore formation in RBC membranes. c, Hemolysis assays. Recombinant OlyA-His6 and His8- PlyB were overexpressed and purified as described in Methods. Each reaction, in a total volume of 500 μl of buffer C, contained 450 μl of rabbit RBCs washed and diluted as described in Methods and the indicated concentrations of OlyA, PlyB, or Triton X-100 detergent. After incubation on a rotator for 30 min at room temperature, each reaction was subjected to 2000 × g centrifugation for 15 min at room temperature, following which an aliquot of the supernatant (100 μl) was assayed for released hemoglobin (absorbance at 540 nm) (n=3). d, Fluorescent labeling of OlyA proteins. Recombinant OlyA(WT), OlyA(W6A), and OlyA(E69A) were purified and labeled with Alexa Fluor 488 maleimide as described in Methods. Aliquots (2 μg each) were collected before and after the labeling reaction, subjected to 15% SDS-PAGE, and proteins were visualized with Coomassie stain (top) or by fluorescence scanning (450 nm channel) (bottom).

2

Figure S2. Purification of OlyA(WT) and OlyA(E69A) for crystallography, structure of lipid-free OlyA(WT), and dot blot assay to measure OlyA binding to liposomes, Related to Figure 2 and Figure 3

a, Schematic of recombinant wild-type OlyA plasmid overexpressed in E. Coli. Starting at the NH₂-terminus, pHis₆-TEV-OlyA(WT) contains a His₆ epitope tag, a TEV protease cleavage site, and OlyA (aa 1-138). b, His₆-TEV-OlyA was purified and cleaved with TEV protease as described in Methods to generate tag-free OlyA, an aliquot (5 μg) of which was subjected to 15% SDS-PAGE, after which the protein was visualized with Coomassie stain. c, gel filtration chromatography of tag-free OlyA on a Superdex 200 column. d, Overall structure of wild-type OlyA (green) in the absence of lipid ligands. e, f, Dot blot assay for OlyA binding to liposomes. Liposomes of the indicated compositions, each containing 0.02 mole% of Texas Red DHPE, were deposited on nitrocellulose membranes and allowed to dry for 5-10 min. Each membrane strip was then incubated for 1h at room temperature with OlyA-His₆ (1 μg/ml) and subjected to anti-His immunoblot analysis as described in Methods. LICOR imaging was used to visualize deposited liposomes (Texas Red, 600 nm channel) (e, top) and liposome-bound OlyA-His6 (chemiluminescence channel) (e, middle). An overlay of both signals is shown in the bottom panel of e. f, Quantification of LICOR signal for bound OlyA. The signal intensity from OlyA-His₆ bound to 1:1 SM:Chol liposomes was set to 100% and all other binding signal intensities were normalized to this set-point (n=3). g, Schematic of recombinant OlyA(E69A) plasmid overexpressed in E. Coli. Starting at the NH₂-terminus, pOlyA(E69A)-TEV-His₆ contains OlyA (aa 1-137 with the E69A point mutation), a TEV protease cleavage site, a 10-aa linker, and a His₆ epitope tag. h, OlyA(E69A)-TEV-His₆ was purified and cleaved with TEV protease as described in Methods to generate tag-free OlyA(E69A), an aliquot (2 μg) of which was subjected to 15% SDS-PAGE, after which the protein was visualized with Coomassie stain. i, gel filtration chromatography of tag-free OlyA(E69A) on a Superdex 200 column. SM, 18:1 sphingomyelin; Chol., cholesterol; Epi., epicholesterol; DOPC, di(18:1) phosphatidylcholine; N, NH₂-terminus; C, COOH-terminus.

3

Figure S3. Effect of mutations on OlyA’s PlyB-mediated hemolytic activity, Related to Figure 2

(top and bottom) Hemolysis assays. Recombinant OlyA-His₆, the indicated mutant versions of OlyA-His₆, and His₈-PlyB were overexpressed and purified as described in Methods. Each reaction, in a total volume of 500 μl of buffer C, contained 450 μl of rabbit RBCs washed and diluted as described in Methods, 10 nM of PlyB, and various concentrations of the indicated mutant version of OlyA. After incubation on a rotator for 30 min at room temperature, each reaction was subjected to 2000 × g centrifugation for 15 min at room temperature and an aliquot of the supernatant (100 μl) was assayed for released hemoglobin (absorbance at 540 nm) (n=3).

4

Figure S4. Lipid structural specificity and temperature dependence for OlyA binding, Related to Figure 4

a, Phospholipid headgroup and acyl chain specificity. Liposomes composed of either 100 mole% of the indicated phospholipid without cholesterol or 50 mole% of the indicated phospholipid and 50 mole% cholesterol were deposited on nitrocellulose membranes and allowed to dry for 5-10 min. Each membrane strip was then incubated for 1h at room temperature with 1 μg/ml OlyA(WT)-His₆ or 0.5 μg/ml OlyA(E69A)-His₆, after which immunoblot analysis and LICOR imaging was used as described in Methods to quantify liposome-bound OlyA. Signal intensities from OlyA(E69A)-His₆ bound to liposomes containing 18:1 SM (left) or an equimolar mixture of 18:1 SM and cholesterol (right) were set to 100% and all other binding signal intensities were normalized to this set-point (n=3). b, Sphingolipid specificity. Liposomes composed of 50 mole% cholesterol, 25 mole% DOPC, and 25 mole% of the indicated sphingolipid were deposited on nitrocellulose membranes and allowed to dry for 5-10 min (a fixed amount of DOPC (25 mole%) was included in all cases to ensure liposome stability). Binding of OlyA(WT)-His₆ and OlyA(E69A)-His₆ was measured as in a. Signal intensities from OlyA-His₆ bound to liposomes containing sphingomyelin (18:1) were set to 100% and all other binding signal intensities were normalized to this set-point (n=3). c, OlyA binding to three component liposomes. Each reaction, in a final volume of 200 μl of buffer A, contained 1 μg of either OlyA(WT)-His₆ or OlyA(E69A)-His₆ and 1560 μM liposomes with the indicated mole fractions of sphingomyelin (18:1), cholesterol, and POPC. After incubation for 4h at room temperature, liposome-bound proteins were measured using a pelleting assay as described in Methods (n=3). d, Circular dichroism spectra. Recombinant OlyA-His₆ was overexpressed and purified as described in Methods, and spectroscopic measurements of 110 μg of OlyA-His₆ in a final volume of 220 μl of buffer A were carried out in a JASCO J-815 CD spectrometer using a 1-mm path length cuvette. The spectra at each indicated temperature represents the average of ten measurements. e, Cholesterol dependence of OlyA binding. Each reaction, in a final volume of 200 μl of buffer A, contained 1 μg of OlyA-His₆ and 1560 μM of liposomes of the indicated compositions. After incubation for 4h at either 23°C or 37°C, liposome-bound OlyA was measured using a pelleting assay as described in Methods (centrifugation was also carried out at the indicated temperatures). The 100% of control value for fraction of bound OlyA-His₆ was 0.62 (n=3).

5

Figure S5. Docking simulations for binding of SM to OlyA(WT) and OlyA(E69A), Related to Figure 5

Docking simulations of a fragment of SM (Fig. 5a) to both OlyA(WT) (top) and OlyA(E69A) (bottom) were carried out as described in Methods. The top 10 scoring poses of SM (yellow sticks) bound to OlyA(WT) (transparent teal) and OlyA(E69A) (transparent purple) are overlaid with key residues represented as sticks (left). On the right, the same top 10 scoring poses of SM are shown bound to surface representations of OlyA(WT) and OlyA(E69A) colored according to electrostatic potential, from negative (red) to positive (blue) (scale ranges from −1 k_BT/e to +1 k_BT/e). Electrostatic potentials were calculated with the APBS module implemented in PyMOL (version 1.8.2.0., Schrodinger, LLC).

6

Figure S6. Sphingomyelin binding pockets occur in structurally similar locations in OlyA(WT), OlyA(E69A), and Stn, Related to Figure 5

a, Ribbon diagrams (top) and electrostatic surface potentials (bottom) of OlyA(WT) (left) and OlyA(E69A) (right). Models for a portion of sphingomyelin (SM) bound to OlyA(WT) (left) and for bis-tris (B-T) bound to OlyA(E69A) (right) are represented as yellow sticks. b, Ribbon diagrams (top) and electrostatic surface potentials (bottom) of Stn (left) and Stnphosphocholine (POC) complex (right). The model for POC is represented as yellow sticks. Key sidechains and bound ligands are shown in stick representation, and sodium ions and waters are shown as purple and red spheres, respectively. All electrostatic potentials were calculated with the APBS module implemented in PyMOL (version 1.8.2.0., Schrodinger, LLC), and the gradients shown ranged from −10k_BT/e (red) to +10 k_BT/e (blue). N, NH₂-terminus; C, COOH-terminus.

7

Figure S7. Time course and dose dependence of binding of fluorescently-labeled OlyA to CHO-K1 cells, Related to Figure 6

a-b, Recombinant OlyA(WT), OlyA(W6A), and OlyA(E69A) were purified and labeled with Alexa Fluor 488 maleimide as described in Fig. S1d. On day 0, CHO-K1 cells were set up in medium B at a density of 6 × 10⁴ cells/well of 48-well plates. On day 1, media was removed, and cells were washed with 500 μl of PBS followed by addition of 200 μl of medium C containing either 3 μM (a) or varying concentrations (b) of the indicated version of OlyA. For quantification purposes, ~5% of OlyA proteins were labeled with Alexa Fluor 488 dyes. After incubation for various times at the indicated temperature (a) or 1 h at 4°C (b), cells were harvested and subjected to fluorescence analysis of bound OlyA as described in Methods (n=3). Chol., cholesterol; Epi., epicholesterol.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse Anti-Histag antibody	EMD Millipore	Cat. # 05-949
donkey anti-mouse secondary	Jackson ImmunoResearch	Cat. # 715-035-150
Bacterial and Virus Strains
BL21 (DE3) pLysS Escherichia coli competent cells	Invitrogen	Cat. # C606003
Chemicals, Peptides, and Recombinant Proteins
Coomassie Brilliant Blue R-250 staining solution	Bio-Rad	Cat. # 161-0436
Dulbecco’s phosphate buffered saline	Corning	Cat. # D0632-25G
Hydroxypropyl beta cyclodextrin (HPCD)	CTD holdings	Cat. # THPB-P
Isopropyl-1-thio-B-d-galactopyranoside (IPTG)	Life Technologies	Cat. # I56000-25
Alexa Fluor 488 C5 maleimide	Life Technologies	Cat. # A10254
Texas Red 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine (TR-DHPE)	Life Technologies	Cat. # T1395MP
EDTA-free protease inhibitor tablets	Roche	Cat. # 05056489001
1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC)	Avanti Polar Lipids	Cat. # 850375
1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE)	Avanti Polar Lipids	Cat. # 850725
1,2-dioleoyl-sn-glycero-3-phospho-L-serine (sodium salt) (DOPS)	Avanti Polar Lipids	Cat. # 840035
1,2-dioleoyl-sn-glycero-3-phospho-(1'-myo-inositol) (ammonium salt) (DOPI)	Avanti Polar Lipids	Cat. # 850149
1,2-dioleoyl-sn-glycero-3-phospho-(1'-rac-glycerol) (sodium salt) (DOPG)	Avanti Polar Lipids	Cat. # 840475
1,2-dioleoyl-sn-glycero-3-phosphate (sodium salt) (DOPA)	Avanti Polar Lipids	Cat. # 840875
N-acetyl-D-erythro-sphingosylphosphorylcholine (2:0 SM)	Avanti Polar Lipids	Cat. # 860580
N-hexanoyl-D-erythro-sphingosylphosphorylcholine (6:0 SM)	Avanti Polar Lipids	Cat. # 860582
N-palmitoyl-D-erythro-sphingosylphosphorylcholine (16:0 SM)	Avanti Polar Lipids	Cat. # 860584
N-stearoyl-D-erythro-sphingosylphosphorylcholine (18:0 SM)	Avanti Polar Lipids	Cat. # 860586
N-oleoyl-D-erythro-sphingosylphosphorylcholine (18:1 SM)	Avanti Polar Lipids	Cat. # 860587
N-lignoceroyl-D-erythro-sphingosylphosphorylcholine (24:0 SM)	Avanti Polar Lipids	Cat. # 860592
N-nervonoyl-D-erythro-sphingosylphosphorylcholine (24:1 SM)	Avanti Polar Lipids	Cat. # 860593
18:1 Biotinyl CAP PE	Avanti Polar Lipids	Cat. # 870273
1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC)	Avanti Polar Lipids	Cat. # 850355
1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC)	Avanti Polar Lipids	Cat. # 850457
1-stearoyl-2-oleoyl-sn-glycero-3-phosphocholine (SOPC)	Avanti Polar Lipids	Cat. # 850467
Sphingosylphosphorylcholine (18:1 lyso SM)	Avanti Polar Lipids	Cat. # 860600
D-erythro-sphingosine	Avanti Polar Lipids	Cat. # 860490
Ceramide (Egg, Chicken)	Avanti Polar Lipids	Cat. # 860051
Ganglioside (Brain, Pig)	Avanti Polar Lipids	Cat. # 860053
Cerebroside (Brain, Pig)	Avanti Polar Lipids	Cat. # 131303
cholesterol	Sigma-Aldrich	Cat. # C8667
ergosterol	Sigma-Aldrich	Cat. # 45480
sitosterol	Sigma-Aldrich	Cat. # S1270
lanosterol	Sigma-Aldrich	Cat. # L5768
Dithiothreitol (DTT)	Sigma-Aldrich	Cat. # D0632-25G
Phenylmethanesulfonyl fluoride (PMSF)	Sigma-Aldrich	Cat. # P7626-25G
Lysozyme from chicken egg white	Sigma-Aldrich	Cat. # L6876-25G
Tris (2-carboxyethyl) phosphine (TCEP)	Sigma-Aldrich	Cat. # C4706-50G
Fetal calf serum (FCS)	Sigma-Aldrich	Cat. # F2442
epicholesterol	Steraloids	Cat. # C6730-000
25-hydroxycholesterol	Steraloids	Cat. # C6510-000
dihydrocholesterol	Steraloids	Cat. # C4350-015
Newborn calf lipoprotein-deficient serum (LPDS)	Das et al., 2014	N/A
Compactin	Das et al., 2014	N/A
Sodium mevalonate	Das et al., 2014	N/A
mPEG maleimide, MW 5000	Nanocs Inc.	Cat. # PG1-ML-5k-1
NanoLinkTM Streptavidin Magnetic Beads 1.0 μm	TriLink Biotechnologies	Cat. # M-1002
Penicillin and streptomycin sulfate	Corning	Cat. # 30-002-CI
1:1 mixture of Ham’s F-12 and Dulbecco’s modified Eagle’s medium	Sigma Aldrich	Cat. # 51445C
Dulbecco’s modified Eagle’s medium (low glucose)	Sigma Aldrich	Cat. # D5796
Dulbecco’s modified Eagle’s medium (high glucose)	Sigma Aldrich	Cat. # D5671
Osmium tetraoxide (4% aqueous solution)	Electron Microscopy Sciences	Cat. # 19150
QuickChange II XL Site-Directed Mutagenesis Kit	Agilent	Cat. # 200522
Bovine Serum Albumin Standard (2 mg/ml)	Thermo Fisher Scientific	Cat. # 23210
Experimental Models: Cell Lines
CHO-K1 cells	ATCC	CCL-61
SV-589 cells	NIGMS Human Genetic Cell Repository	N/A
Neuro-2A cells	ATCC	CCL-131
ST88-14 cells	Gift from F. S. Collins, University of Michigan, Ann Arbor in 1993	N/A
MDCK cells	ATCC, 1991	N/A
Caco-2 cells	ATCC, 1990	N/A
Recombinant DNA
pUC57-lysenin	This paper (synthesized by GenScript)	N/A
pLys-His6	This paper	N/A
pHis6-Eqt	Skocaj et al., 2014	Gift from Dr. Kristina Sepcic, University of Ljubljana, Slovenia
pET21c+-OlyA-His6	Skocaj et al., 2014	Gift from Dr. Kristina Sepcic, University of Ljubljana, Slovenia
pOlyA-His6	This paper	N/A
pHis6-TEV-OlyA	This paper	N/A
pET3a-PlyB	Lukoyanova et al., 2015	Gift from Dr. Michelle Dunstone, Monash University, Australlia
pHis8-PlyB	This paper	N/A
pALOD4	Gay et al., 2015	N/A
pALOFL	Gay et al., 2015	N/A
pHis6-TEV	This paper	N/A
Software and Algorithms
HKL-3000	Minor et al., 2006	http://www.hkl-xray.com/
Phaser	McCoy et al., 2007	https://www.phenix-online.org/
Coot	Emsley et al., 2010	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Phenix	Adams et al., 2010	https://www.phenix-online.org/
Pymol	Pymol	https://pymol.org/2/
Maestro	Schrodinger	Version 10.3
Protein Preparation Wizard	Schrodinger	Version 11.2
LigPrep	Schrodinger	Version 3.5
EPIK	Schrodinger	Version 3.3
Glide	Schrodinger	Version 6.8
Other
1.7 mL low-retention microcentrifuge tubes	Fisher Scientific	Cat. # 02-681-320
DynaMag−2 magnet	Life Technologies	Cat. # 12321D
Tricon 10/300 Superdex 200 Increase column	GE healthcare	Cat. # 28-9909-44
1-ml HisTrap HP Ni column	GE Healthcare	Cat. # 17524701
Black 96-well flat bottom non-binding plate	Greiner Bio-One	Cat. # 655900
Black 96-well glass bottom plate	Greiner Bio-One	Cat. # 655891

Highlights:

OlyA detects sphingomyelin/cholesterol complexes, but not free sphingomyelin

A shallow channel in OlyA binds ceramide, but not glycerol, phospholipid backbones

OlyA’s cholesterol specificity is determined by a single glutamic acid residue

Plasma membranes maintain constant levels of sphingomyelin/cholesterol complexes