Lung Surfactant Protein SP-B Promotes Formation of Bilayer Reservoirs from Monolayer and Lipid Transfer between the Interface and Subphase

Abstract

We investigated the possible role of SP-B proteins in the function of lung surfactant. To this end, lipid monolayers at the air/water interface, bilayers in water, and transformations between them in the presence of SP-B were simulated. The proteins attached bilayers to monolayers, providing close proximity of the reservoirs with the interface. In the attached aggregates, SP-B mediated establishment of the lipid-lined connection similar to the hemifusion stalk. Via this connection, a lipid flow was initiated between the monolayer at the interface and the bilayer in water in a surface-tension-dependent manner. On interface expansion, the flow of lipids to the monolayer restored the surface tension to the equilibrium spreading value. SP-B induced formation of bilayer folds from the monolayer at positive surface tensions below the equilibrium. In the absence of proteins, lipid monolayers were stable at these conditions. Fold nucleation was initiated by SP-B from the liquid-expanded monolayer phase by local bending, and the proteins lined the curved perimeter of the growing fold. No effect on the liquid-condensed phase was observed. Covalently linked dimers resulted in faster kinetics for monolayer folding. The simulation results are in line with existing hypotheses on SP-B activity in lung surfactant and explain its molecular mechanism.

Introduction

Lung surfactant is a thin film lining the gas exchange interface in the lung alveoli. It consists of lipids (∼90% wt) and proteins (∼10% wt) (1). The lipid fraction consists mainly of phosphatidylcholine (∼80% of total), half of which is dipalmitoyl phosphatidylcholine (DPPC) (∼40%); other components are anionic phosphatidylinositol and phosphatidylglycerol (∼8–15% of lipids), neutral lipids (mainly cholesterol, ∼5–10%), palmitic acid, and other components in small amounts, with comparable contents of saturated and unsaturated lipids (2). Two hydrophilic proteins, SP-A and SP-D, constitute a larger fraction and are mainly involved in the innate immune response, whereas two hydrophobic proteins, SP-B and SP-C (<3% of the total), participate in surface activity (3–5). This mixture forms a monolayer at the air/water interface with associated bilayer reservoirs in water (6–8).

The main function of lung surfactant is maintaining the surface tension at the interface at low values (9). This facilitates breathing and prevents alveolar collapse due to Laplace pressure. To achieve this function, lung surfactant must be able to adsorb rapidly to the interface, sustain near-zero surface tension on interface compression at exhalation, and prevent surface tension increase at inhalation (5,10,11). To meet the first and third of these requirements, efficient transfer of lipids to the interface is essential. This transfer is mediated by SP-B and SP-C proteins (see Rugonyi et al. (11) and Serrano and Pérez-Gil (12) and references therein) and is believed to be achieved via a highly curved stalk-like intermediate (13). The proteins are also known to facilitate monolayer collapse, leading to formation of bilayer reservoirs (14–18). Collapse occurs close to the equilibrium surface tension, at which a monolayer coexists with bilayers (19). Bilayer reservoirs not only provide the source of lipids for the monolayer on expansion, but are also believed to enhance the monolayer stability on interface compression (20–22).

SP-B and SP-C proteins, although present in small amounts (each <1.5%), thus play an active role. Absence of SP-C leads to altered lung function (23,24), and deficiency of SP-B is lethal (25,26). SP-B is a small (79-residue) hydrophobic protein carrying a total positive charge of +7 (27). SP-B can form a homodimer linked via a cysteine residue (4,28), and six other cysteines form intramolecular disulfide bridges (29). High-resolution structures of synthetic protein fragments (e.g., N-terminal 1–25 (30), or 34-residue mini-B (31)) are available, but the 3D structure of the full-length protein is not known. Given its crucial function, the presence of SP-B is desirable in artificial surfactant replacements. Designing surfactant replacements is mainly targeted at effective treatment of acute respiratory distress syndrome (ARDS) (32). The full-length protein (dimer) is, however, difficult to synthesize, and synthetic peptides based on its motifs are used to avoid animal-derived extracts (12). Synthetic fragments reproduce the fusogenic, lytic, and surface-active properties of SP-B, but are generally less efficient (31,33–39). Design of synthetic preparations is complicated by a lack of understanding of the structural features determining the activity of the protein. A common conclusion of numerous studies (see e.g., Serrano and Pérez-Gil (12) for review) is that SP-B keeps bilayer reservoirs attached to the monolayer, perturbs lipid packing, leading to lipid transfer, and promotes formation of bilayer reservoirs. However, the exact mechanism of its action on the molecular level remains elusive.

In this work, we investigate the possible role of SP-B protein in lipid collective transformations between a monolayer at the air/water interface and bilayers in water. We use the coarse-grained force-field MARTINI (40,41) and perform large-scale molecular dynamics simulations. Previous simulations have examined the interactions between SP-B fragment 1–25 and lipid monolayers in atomic detail (42–45). Monolayer 2D-3D transformations have been simulated in both atomistic and coarse-grained models (46–50). It was shown that in the absence of proteins, lipid monolayers compressed to negative surface tensions undergo collapse from the interface by buckling followed by folding into bilayers (49,50). This collapse pathway is in agreement with theoretical and experimental studies (see, e.g., Lee (51) and Pocivavsek et al. (52) and references therein); however, collapse is expected to occur at much larger and positive surface tensions. In recent simulations with the MARTINI force-field (50), SP-B fragment 1–25 and SP-C protein were shown to act as defects at which bilayer folds were nucleated at near-zero tensions. In this work, we simulate formation of bilayer folds from monolayers promoted by SP-B dimers at positive surface tensions. The proteins attach disconnected bilayers to monolayers, and induce formation of a stalk-like lipid connection, as in the vesicle fusion pathway simulated earlier (53). We reproduce lipid transfer between the interface and the subphase mediated by SP-B and directed in a surface-tension-dependent manner.

Methods

We investigated the role of SP-B protein in the surface activity of lipid monolayers and in transfer of lipids between monolayers and bilayers using molecular dynamics (MD) simulations with the MARTINI force field of a number of different systems (Table 1).

Table 1.

Summary of simulations performed

System setup		Temperature (K)	Surface tension (mN/m)	Simulations
	Mixture			No. of runs	Duration (μs)	Final structure/outcome
Monolayer, vesicle, SP-B	1	310	30	2	5	Attached
			20	1		Attached
			10	2		Transfer from monolayer, collapse
			1	1
	2		30	2		Transfer to monolayer
			20	1		Stalk
			10	2		Transfer from monolayer, collapse
			1	1
Monolayer, vesicle, attached by SP-B	2		35	2		Transfer to monolayer
			30	2		Transfer to monolayer
			20	1		Stalk
			15	1		Transfer from monolayer, collapse
Monolayer, bicelle, attached by SP-B	2	310	35	4	5	Transfer to monolayer
Monolayer, bicelle (lipids only)	2	310	35	2	5	Disconnected
Monolayer, SP-B	1	310	30	2	5	No bending, folds re-spread
			20	2		Minor curvature
			10	2		Local bending
			9	3	10
			7	3	5	Folding, collapse
			5	2
		300	7	1	5	LC + LE, folding, collapse
			5	1
			1	2
		290	7	1	5	LC, no bending
			1	1
	2	310	30	2	5	No bending/ folds re-spread
			20	2		Minor curvature
			12	3	10	Local bending
			10	3	5	Folding, collapse
			9	2
			7	2
			5	1
Monolayer (lipids only)	1	310	1	1	10	LE, stable
		300	7	1	5	LC+ LE, stable
			5	1		LC+LE, stable
			1	1		LC, stable
		290	1	1	5	LC, stable
			10	1
	2	310	1	1	10	LE, stable

Lipid monolayers composed of 3:1 DPPC/POPG (mixture 1) and 5:2:2:1 DPPC/POPG/DOPC/PA (mixture 2) were investigated in the specified range of temperatures and surface tensions, in the presence and absence of SP-B proteins. The monolayer-bicelle setup contains a bicelle in the water subphase. The monolayer-vesicle setup includes two small unilamellar vesicles in water. Monolayer and bilayer aggregates are attached by proteins in the starting configuration in selected simulations.

System setup

The system setup included a water slab in vacuum with two symmetric monolayers at the two vacuum/water interfaces; where indicated, two small unilamellar vesicles or a bicelle were placed in water (Fig. 1, a and b). Bicelles, or circular bilayer patches, were obtained by disrupting periodic bilayers and equilibrating in bulk water. The vesicles were formed from bicelles by their spontaneous bending and closing on the timescale of hundreds of nanoseconds. Lipid mixtures of dipalmitoyl phosphatidylcholine (DPPC), dioleoyl phosphatidylcholine (DOPC), palmitoyloleoyl phosphatidylglycerol (POPG), and palmitic acid (PA) were investigated: DPPC/POPG 3:1 (mixture 1); and DPPC/DOPC/POPG/PA 5:2:2:1 (mixture 2). DPPC is a major component of lung surfactant, POPG is a prevalent anionic species, and DOPC (as well as POPG) represents the unsaturated fraction. The surface tension was varied between 0 and 35 mN/m. The temperature was set to 310 K, at which the monolayers of the considered compositions form a liquid-expanded (LE) phase in the given range of surface tensions, whereas bilayer aggregates form a liquid-crystalline phase. Lower temperatures of 300 and 290 K were used to reproduce the liquid-condensed (LC) phase in monolayers in the MARTINI model. The monolayers were equilibrated in a separate protein-free simulation at a surface tension of 35 mN/m. Bicelles with open edges were simulated in a confined space between the two monolayers, preventing their closing. Bicelles were selected as bilayer models to allow water flow as they change their position with respect to the monolayers, which would not be possible for continuous (periodic) bilayers. In the monolayer-vesicle setup, the proteins were placed randomly in water, vacuum, or at interface, and the systems were equilibrated at a constant monolayer area. In the monolayer setup, the proteins were positioned in water in the monolayer vicinity to obtain a symmetric distribution of proteins between the two symmetric monolayers. After equilibration, a range of conditions was applied (Table 1). Both monolayer and monolayer-vesicle setups included 16 SP-B proteins. In additional simulations (mixture 2), two proteins were bound to the surfaces of both monolayer and vesicles or monolayer and bicelle in the starting configuration (called attached setups in the text). Each monolayer consisted of 2304 lipids, resulting in a lateral dimension of ∼40 nm. A vesicle or bicelle contained 1152 lipids, giving a diameter of 14 nm for the former. The simulation box contained ∼200,000 water particles (∼100,000 in the confined case with a bicelle) with Na⁺ ions to neutralize the negative charge of the POPG lipids. The total protein content was ∼4% (wt), slightly higher than in lung surfactant.

Figure 1.

System setup: water slab in vacuum with two monolayers at the two interfaces. Two vesicles (a) or a bicelle (b) are placed in water. Lipid polar groups are shown in green, apolar groups in orange, double bonds in the hydrocarbon chains in white, backbone of SP-B proteins in red (side chains are not shown), and water in cyan.

Force field

We used the MARTINI coarse-grained model for lipids and proteins (40,41). In this model, the molecules are represented by grouping four heavy atoms (two to three in the case of ring structures) into a particle. DPPC, DOPC, and PA are standard components of the force field; in POPG, the glycerol group in the headgroup was represented by a polar particle (P4). In DOPC and POPG, the angle potential in the unsaturated hydrocarbon chains between particles C1, C2, and D3 was modified analogous to polyunsaturated chains in this model. The secondary structure of SP-B protein (Fig. 2 a) was obtained using homology with the saposin family based on the known structures of saposin C and NK-lysin (54–56), as well as its fragment called mini-B (31,37). The protein structure represents a characteristic saposin fold with four α-helices connected by unstructured loops and linked by three intramolecular disulfide bridges (27,57). We have tested several similar homology models of the protein (results not shown) and use here the most active one. The protein topology was built using the scripts downloaded from the MARTINI website (http://md.chem.rug.nl/∼marrink/MARTINI/). The LINCS algorithm (58,59) was used to constrain the bonds in the aromatic side chains and the bonds between hydrophobic side chains and backbone in the proteins. The SP-B dimer was modeled by introducing an additional harmonic bond between C48 residues.

Figure 2.

Structure and orientation of SP-B proteins on the surface of a lipid monolayer. (a) The secondary structure of the protein contains four α-helices (h1–h4) connected by unstructured loops (solid lines) and linked by disulfide bridges (dashed lines) forming a hairpin shape. (b and c) A monomer bound to the monolayer surface in the bent (b) and extended (c) conformations. (d–f) Dimers in antiparallel (d) and parallel (e) orientations formed by aggregation, and a covalently linked dimer (f). View from the water subphase. See Fig. 1 legend for color scheme; water not shown.

Simulation parameters

For nonbonded interactions, the standard cutoffs for the MARTINI force field were used: the Lennard-Jones potential was shifted to zero between 0.9 and 1.2 nm; the Coulomb potential was shifted to zero between 0 and 1.2 nm, with a relative dielectric constant of 15. The time step was 20 fs, with neighbor list updates every 10 steps. Lipids, water, and proteins were coupled separately to a target temperature using the velocity rescaling thermostat (60) with a time constant of 1 ps. For simulations at constant monolayer area, no pressure coupling was used. A surface tension was applied using a surface-tension coupling scheme with the Berendsen barostat and a time constant of 4 ps (61). The compressibility was set to zero in the normal to the interface direction (the z axis) to maintain vacuum. Initial equilibrations were 1 μs long, and production runs were 5 μs long, with selected simulations extended to 10 μs as specified in Table 1.

Results

We explored surface tensions between 0 and 35 mN/m, which covers the physiological range for lung surfactant (62). At 310 K, the considered lipid mixtures were laterally homogeneous (no spatial clustering of lipids), and the monolayers formed a liquid-expanded (LE) phase. The proteins inserted into the headgroup region of lipid monolayers (Fig. 3, a and b) and associated primarily with POPG (anionic) or DOPC (the most unsaturated lipid). Near the equilibrium surface tension (at 20 mN/m), the average numbers of protein-lipid contacts at distances <1 nm were 2.0 and 4.0 for DPPC and POPG, respectively, in mixture 1, and 1.6, 3.7, 3.6, and 1.3 for DPPC, POPG, DOPC, and PA, respectively, in mixture 2 (the numbers are per protein particle, normalized by concentration of each lipid).

Figure 3.

Density profiles along the normal to the interface for SP-B, lipid groups, and solvent at 20 mN/m, 310 K (a); 7 mN/m, 310 K (b), and 7 mN/m, 290 K (c).

SP-B facilitates lipid transfer between monolayers and bilayers

We first investigated the possible role of SP-B proteins in the transfer of lipid molecules between a monolayer at the air/water interface and a bilayer in water. To this end, two vesicles were added in water, and surface tensions above and below the equilibrium value (∼22 mN/m) were simulated (see Table 1).

SP-Bs diffused towards the membrane/water interface and partitioned into the lipid headgroup region, whether the proteins were positioned initially in water or in a vacuum, outside the monolayer. Protein aggregates also formed in water and on the membrane surface. All proteins and their aggregates became membrane-bound on the simulation timescale. As reported previously (53), the proteins adopted two conformations on the surface of the membrane: bent or extended (Fig. 2, b and c).

SP-Bs in the bent conformation, resembling the structure of NK-lysin (55), attached monolayers to vesicles by binding to the surfaces of both membranes (Fig. 4 a). Once attached, the connection was stable and did not disaggregate on the simulation timescale of 5–10 μs. Some proteins also attached two vesicles together. Single or multiple proteins or their aggregates (of two to four) maintained the connection (Fig. 4, b and c). If the proteins aggregated in water, they formed a globule that could still attach two membranes but provided a farther separation between them, with no significant evolution of the connection.

Figure 4.

SP-B connects bilayer reservoirs to the monolayer. (a–c) An SP-B monomer (a) and aggregates of two (b) and three (c) proteins attach a small vesicle to a monolayer. (d and e) A lipid bridge (d) and a stalk-like connection (e) between the monolayer and the vesicle are established. (f) Proteins line the positively curved side of the stalk. Images correspond to mixture 2 at 310 K and 10 mN/m at simulation times of 1 μs (a), 1.1 μs (d), and 1.2 μs (e); first and second monolayers with vesicles at 30 mN/m at 1.8 μs (b and c), and a monolayer with fused vesicles at 30 mN/m at 4.9 μs (f). Color scheme is as in Fig. 1.

The protein mediated formation of a lipid bridge between the attached monolayer and vesicle. The characteristic times for establishment of a lipid-lined connection varied between tens of nanoseconds and microseconds, increasing for higher surface tensions. The number of lipids protruding out of the membranes grew rapidly, and the lipid bridge transformed into a stalk-like structure (Fig. 4, d and f) on a timescale of tens of nanoseconds. Formation of the stalk followed the same pathway as fusion of two vesicles mediated by SP-B, described earlier (53).

A lipid flow initiated between the monolayer and vesicle via the stalk. The direction of the flow depended on the surface tension in the monolayer. Below the equilibrium surface tension (∼22 mN/m), lipids were transferred from the monolayer to the vesicle (Fig. 4 e, Movie S1 in the Supporting Material). If the surface tension was maintained at the same value, the monolayer collapsed. Above the equilibrium tension, lipids were transferred from the subphase to the monolayer (Movie S2). Near the equilibrium surface tension, the stalk-like lipid connection remained on the simulation timescale, and only minor lipid flow was observed.

Simulation results were similar for the two lipid compositions. The simulation outcomes at different conditions are reported in Table 1. The proteins attached the monolayer and vesicles in all simulations. The stalk-like lipid connection was established more often in mixture 2 due to a higher concentration of unsaturated lipids making the membranes softer. Lower elastic moduli in this mixture reduce the energy cost for highly curved lipid-based connections. To obtain more information on lipid transfer, vesicles and bicelles attached to monolayers by SP-Bs in the starting configuration (for mixture 2, the softer lipid mixture) were simulated. Lipid transfer from the vesicles was limited in each case, and the inner leaflet did not merge with the monolayer. A small vesicle radius leading to stretching of its outer leaflet likely makes the loss of lipids unfavorable. The lipids transferred to the stalk continued to travel to the expanded monolayer, thus disrupting the connection. In contrast, lipid transfer from bicelles was not restricted and continued until the surface tension in the monolayer returned to the equilibrium value (see Fig. 5, a–d). The characteristic times to establish a lipid bridge with an expanded monolayer were also much faster (tens of nanoseconds). In the absence of SP-B, the bicelles remained disconnected from monolayers on the simulation timescale.

Figure 5.

SP-B connects bilayer reservoirs to the monolayer. The protein attaches a bicelle to a monolayer (a), promotes formation of a lipid bridge (b), which transforms into a stalk (c), and lipid flow continues to the expanded monolayer (d). Images correspond to a monolayer of mixture 2 at 35 mN/m and 310 K, at 24 ns (a), 113 ns (b), 126 ns (c), and 137 ns (d) of simulation. Color scheme is as in Fig. 1.

SP-B induces bilayer folds in lipid monolayers

SP-B had a strong effect on the properties of lipid monolayers forming a LE phase. The protein in the extended conformation (Fig. 2 c), resembling the open structure of saposin C (63), induced local bending of the monolayers (Fig. 6 a) at surface tensions below the equilibrium value (∼22 mN/m). The bent conformation (Fig. 2 b) did not lead to similar distortion; interconversion between these conformations occurred on the microsecond timescale. On lowering the surface tension, the amplitude of the bending deformation increased (Fig. 6 b), and the protein center of mass shifted out of the monolayer (compare Fig. 3, a and b). Below a certain threshold tension, a bilayer fold was formed from the monolayer (Fig. 6 c). The threshold values for folding depended on lipid composition, and were 9 ± 1 and 12 ± 1 mN/m for mixtures 1 and 2, respectively. At near-zero surface tensions (1 mN/m), a single protein could initiate monolayer folding. At higher surface tensions, protein aggregation was required to produce bilayer folds, and a dimer was sufficient. In the dimer, the proteins adopted parallel or antiparallel orientations (Fig. 2, d and e). Once a bilayer fold was formed, the proteins lined its perimeter characterized by positive curvature, supporting an earlier hypothesis (64). When surface tension was maintained below the equilibrium value, the folds grew in size, eventually leading to monolayer collapse. When surface tension was increased above equilibrium value, the bilayer folds respreaded into the monolayer. In the absence of proteins, the monolayers were stable at the same conditions (Table 1) and collapsed only at zero or negative surface tensions by buckling and folding (results not shown), as characterized in detail in previous works (49,50,52).

Figure 6.

(a) SP-B induces local curvature in lipid monolayers at surface tensions below the equilibrium spreading value. (b and c) Protein aggregates (dimers) initiate formation of bilayer folds, side view. Images correspond to a monolayer of mixture 1 at 7 mN/m and 310 K at simulation times of 1 μs (a), 3.7 μs (b), and 3.76 μs (c). Color scheme is as in Fig. 1.

In contrast, SP-Bs had no effect on monolayers forming a liquid-condensed (LC) phase, similar to SP-B fragments (50). Since bending and folding of monolayers were observed only at low surface tensions at 310 K, we considered the same range of tensions at lower temperatures. At 290 K, mixture 1 formed a LC phase between 1 and 10 mN/m in the absence of proteins. Under these conditions, proteins induced only minor local distortion (perturbation of translational order without bending; see Fig. 3 c) of the monolayer while the monolayers remained in the LC phase. In the coexistence region of the LC and LE phases (5–7 mN/m, 300 K), the proteins partitioned into the LE phase, in agreement with experimental observations (20,65,66), and initiated monolayer folding and collapse. At conditions close to the phase transition (1 mN/m, 300 K), the proteins disordered the monolayer locally and induced formation of LE domains from the LC phase, with consequent monolayer collapse by folding. Reducing the protein concentration twice had the same effect.

It is interesting to note that formation of bilayer folds from monolayer was observed only below a certain threshold surface tension. The threshold values depended on the lipid composition, were positive for both mixtures, and were below the equilibrium value (∼22 mN/m). The times required for monolayer folding decreased from the microsecond scale to tens of nanoseconds as the surface tension was reduced from the threshold to a near-zero value. Since protein aggregation was necessary to produce monolayer folding, the observed intervals include the time required to form a dimer from randomly distributed monomers, as well as the time to bend a monolayer into a bilayer. One factor controlling the former is protein diffusion; however, the coefficient of long-time lateral diffusion of protein in the LE phase did not depend noticeably on the surface tension in the considered range (and equaled (4.5 ± 0.6) × 10⁻⁷ cm²/s in mixture 1 and (4 ± 1) × 10⁻⁷ cm²/s in mixture 2). Moreover, aggregation of proteins bound to the monolayer surface was not observed at higher surface tensions (of 20 and 30 mN/m). Establishment of intermolecular contacts, which depends on protein interactions with the monolayer and changes with surface tension (compare the partitioning in Fig. 3, a–c), is likely the limiting factor for dimer formation. SP-Bs are known to dimerize by forming a disulfide bridge between cysteines (C48-C48) (27,28). Introducing a harmonic bond between these residues in the antiparallel dimer (Fig. 2 f) led to faster kinetics for monolayer folding (tens of nanoseconds) but neither changed the threshold surface tension nor had other noticeable effects.

Discussion

We simulated formation of bilayer folds from a monolayer induced by SP-B. The ability of SP-B to promote 2D-to-3D transformations in monolayers became evident in earlier experimental studies (15–17). Monolayer-bilayer transformations were also observed in previous simulations, but only at negative or near-zero surface tensions. Unlike the transformations in the absence of proteins, which start with monolayer buckling/wrinkling (49,52), SP-B fragments were shown to act as defects at which the folds nucleated (67). In this work, the full-length proteins not only form nucleation sites but bend the monolayer locally and line the highly curved perimeter of the folds. This significantly reduces the activation barrier and shifts the monolayer-bilayer transformation to much higher surface tensions. Covalently linked dimers appear to bend the monolayer more efficiently and thus speed up folding.

Bilayer fold formation is followed by monolayer collapse, which restores the surface tension to the equilibrium value. This collapse is characteristic for lipid monolayers/lung surfactant extracts containing hydrophobic surfactant proteins (see, e.g., Rugonyi et al. (11) and Possmayer et al. (68) and references therein). It manifests itself as a plateau on the surface pressure (tension)-area isotherms. In experiments, collapse occurs at surface tensions close to the equilibrium value (16,17,69). In our simulations, a monolayer disconnected from any bilayer structures initially can sustain somewhat lower tensions. Once a bilayer fold is formed, the surface tension is readily restored to the equilibrium value on timescales much faster (tens of nanoseconds) than can be resolved experimentally. This hysteresis is characteristic for first-order phase transitions and is associated with creating a fold nucleus. The energy for fold formation includes the work of increasing the monolayer area against the surface tension. This work vanishes at low tensions but is finite near equilibrium values. The activation barrier for folding while lowered by SP-B increases with increasing surface tension. On the other hand, the presence of an already formed bilayer fold or a vesicle attached by the protein provides lipid flow/monolayer collapse at much higher tensions approaching equilibrium.

Our simulations show that SP-B can attach bilayer reservoirs to the monolayer. Such close proximity or direct association provides an immediate source of surfactant for the interface instead of relying on diffusion of material in the water subphase. Bilayer reservoirs underlying the monolayer are also believed to play an additional role of stabilizing the interface against collapse at low surface tensions (20–22). Our simulations do not address this question, but they suggest that attachment of disconnected bilayers by proteins constitutes the first step for formation of a lipid-lined connection between the interface and the subphase.

We observe the formation of a stalk connecting bilayer reservoirs to a monolayer mediated by SP-B. This structure likely represents the rate-limiting step in surfactant collective adsorption/transfer to the interface (11,12), as it is characterized by high curvature and a high energy barrier. The ability of hydrophobic surfactant proteins (SP-B and SP-C) to promote the stalk has been related to their effect on the curvature of lipid phases (13). Inducing a negative spontaneous curvature in a lipid-protein mixture would decrease the energy barrier for forming the negatively curved stalk. Here, we show that the transition to stalk proceeds along a fusion pathway (70) from establishment of close contact by protein binding to both membranes, to lipid protrusions and formation of a lipid bridge along the protein wedge. Once the stalk is formed, we observe lipid transfer to/from the monolayer. The flow is directed to the interface if the surface tension increases above the equilibrium. This prevents further increase of surface tension on expansion of the gas exchange interface at inhalation. On interface compression, lipid flow from the interface could provide refinement of the monolayer composition. Although there is a contradiction to whether the squeeze-out of material from the interface is selective to certain components (unsaturated lipid fraction; see, e.g., (9,71) and (72–74)), experimental studies generally agree that bilayer reservoirs appear and grow from monolayers below the equilibrium tension.

The activities of full-length SP-B observed in this study rely on the ability of the protein to adopt bent and extended conformations upon binding to a membrane surface, and to interconvert between the conformations. Protein fragments (e.g., SP-B 1–25 or mini-B) could achieve similar effects by mimicking these conformations upon aggregation. The activity of SP-B fragments via aggregation would probably have slower kinetics due to times for diffusion and establishment of intermolecular contacts, and could require a higher local concentration. In addition, cooperative rearrangements of protein domains to distort lipid membranes could be less effective for disconnected protein fragments. Earlier experimental studies confirmed that SP-B fragments can reproduce the properties of the full-length protein but are less efficient (31,35,37,75,76). In our simulations, mini-B aggregates attached vesicles and monolayers and induced monolayer folding (results not shown). However, folding occurred over longer times and at lower surface tensions compared to the case with full-length SP-B, whereas attachment did not lead to lipid transfer. In the same manner, covalently linked SP-B homodimer provided faster kinetics for monolayer folding than did monomer aggregates.

The limitations of our simulations are the short timescale (1–10 μs) and the small lengthscale (∼40 nm) compared to those in experiments. In a macroscopically large monolayer versus a few tens of nanometers in simulations, there is a higher probability that monolayer folding will occur. Simulation times used may not be sufficient to observe the system overcoming higher activation barriers, such as for monolayer folding close to the equilibrium tension. In addition, experimentally observed monolayer collapse is only partial near equilibrium values and allows further reduction of surface tension. The properties promoting partial monolayer collapse are currently being investigated (S. Baoukina and D. P. Tieleman, unpublished work). In this work, we considered monomeric proteins, and a covalently linked dimer was used only in the case of monolayer folding. Dimer models tested so far did not lead to noticeable improvements in lipid transfer and remain a subject for further study. The model for the protein monomer was used in an earlier work (53); its secondary structure is fixed by the parameters of the force field, but spatial rearrangements are unrestricted (41). Atomistic simulations could provide more details on the protein's conformational space and interactions with lipids. At the other extreme, less detailed models or much larger simulation models at the current resolution would be needed to study longer-range interactions, e.g., between multiple bilayer reservoirs or networks of proteins and their effects on monolayer-bilayer transformations.

Despite these limitations, our simulations are in good agreement with the hypotheses on SP-B activity in lung surfactant derived from a large number of experimental and theoretical works. This study reveals the molecular mechanisms for connecting bilayer reservoirs to a monolayer, promoting their close proximity; formation of bilayer reservoirs from a monolayer; and surface tension-dependent lipid transfer between the interface and the sub-phase mediated by SP-B proteins.

Conclusions

We have investigated by detailed computer simulations the role of the full-length SP-B protein in a characteristic saposin-fold structure that interconverts between extended and bent conformations. We have shown that SP-Bs 1), induce formation of bilayer reservoirs by monolayer folding below the equilibrium surface tension; 2), attach disconnected bilayer aggregates in water to a monolayer at the interface; 3), facilitate lipid transfer between these structures in a surface-tension-dependent manner. These activities underlie the crucial role of SP-B in the function of lung surfactant, namely, maintaining the surface tension at the gas-exchange interface at low values, which is necessary for breathing.