Fate of Liposomes in the Presence of Phospholipase C and D: From Atomic to Supramolecular Lipid Arrangement

Understanding the origins of lipid membrane bilayer rearrangement in response to external stimuli is an essential component of cell biology and the bottom-up design of liposomes for biomedical applications. The enzymes phospholipase C and D (PLC and PLD) both cleave the phosphorus–oxygen bonds of phosphate esters in phosphatidylcholine (PC) lipids. The atomic position of this hydrolysis reaction has huge implications for the stability of PC-containing self-assembled structures, such as the cell wall and lipid-based vesicle drug delivery vectors. While PLC converts PC to diacylglycerol (DAG), the interaction of PC with PLD produces phosphatidic acid (PA). Here we present a combination of small-angle scattering data and all-atom molecular dynamics simulations, providing insights into the effects of atomic-scale reorganization on the supramolecular assembly of PC membrane bilayers upon enzyme-mediated incorporation of DAG or PA. We observed that PC liposomes completely disintegrate in the presence of PLC, as conversion of PC to DAG progresses. At lower concentrations, DAG molecules within fluid PC bilayers form hydrogen bonds with backbone carbonyl oxygens in neighboring PC molecules and burrow into the hydrophobic region. This leads initially to membrane thinning followed by a swelling of the lamellar phase with increased DAG. At higher DAG concentrations, localized membrane tension causes a change in lipid phase from lamellar to the hexagonal and micellar cubic phases. Molecular dynamics simulations show that this destabilization is also caused in part by the decreased ability of DAG-containing PC membranes to coordinate sodium ions. Conversely, PLD-treated PC liposomes remain stable up to extremely high conversions to PA. Here, the negatively charged PA headgroup attracts significant amounts of sodium ions from the bulk solution to the membrane surface, leading to a swelling of the coordinated water layer. These findings are a vital step toward a fundamental understanding of the degradation behavior of PC lipid membranes in the presence of these clinically relevant enzymes, and toward the rational design of diagnostic and drug delivery technologies for phospholipase-dysregulation-based diseases.


Materials and Methods
The lipids POPC, POG and POPA were purchased from Avanti Polar Lipids, Alabaster, USA in lyophilized powder form (purity >99%), and used without further purification. Lipids were stored at -20 °C and warmed to room temperature before use. Phosphate buffered saline (140 mM NaCl, 10 mM Na 3 PO 4 , 2.68 mM KCl, pH 7.45) was prepared from commercially available tablets (Life Technologies, Carlsbad, California, USA) and filtered through a 0.2 µm membrane before use. All other chemicals and phospholipases were purchased from Sigma Aldrich and used as supplied.
Liposome formulation: Liposomes were prepared by the freeze-thaw method and extruded to the desired size. Thin films were prepared by evaporating chloroform solutions of POPC. After drying in vacuo overnight, films were incubated for 30 min with 1 mL phosphate-buffered saline solution (PBS) at pH 7.4 (lipid concentration approx. 2.7 mM (SPR experiments) or 27 mM (SANS experiments)).
Subsequently five freeze-thaw cycles were carried out using liquid nitrogen and a 40 ˚C water bath.
The resulting multilamellar vesicles were extruded 15 times using a micro-extruder (Avanti Polar Lipids) and polycarbonate membranes (Whatman Nucleopore) with 200 nm pores. When required they were also further extruded 15 times through 50 nm membranes. For SPR experiments, liposome formulations were diluted to 0.5 mM total lipid concentration in PBS. Typical data collection times on SANS2D were 15 min. The beamline was configured with L 1 = L 2 = 4 m pinhole collimation and sample-detector distances to give a scattering vector Q = (4π/λ)sin(θ/2) range of 0.004 to 0.8 Å -1 , where θ is the scattering angle and neutrons of wavelengths λ of 1.75 to 16.5 Å were used simultaneously by time of flight. Measurements on Larmor were carried out in event mode and the wavelength and Q range for this experiment were 0.9 -12.5 Å and 0.004 -0.8 Å -1 respectively. The instrument was in the 4 m sample-detector configuration, with A1=20mm 2 , S1=14mm 2 , and a sample aperture of 6mm (horizontal) by 8mm (vertical). Data reduction was performed using Mantid 1 and scattering simulations fitted using SasView v3.0. 2 All measurements were fitted using a custom model, taking the sum of two MultiShellVesicle models and fixing the parameters to have one (unilamellar component, see Table 1) and two (bilamellar component) shells respectively (see below for further details).

SAXS measurements:
Mixtures of POPC:POG and POPC:POPA at 100:0, 90:10, 50:50 and 10:90 mol% were co-dissolved in chloroform which was evaporated under a stream of nitrogen, then lyophilized overnight for a minimum of 12 h to remove any residual solvent. Mixtures were hydrated in: a) POPC:POG mixtures, PBS w/w to 60 wt% and b) POPC:POPA mixtures, PBS or pure water (Gibco) w/w to 80 wt%. Samples were sealed and heat cycled 20 times between -196°C and 60 °C.
SAXS data was obtained at 25 °C at Diamond Light Source, UK, using beamline I22, with samples mounted in glass capillaries (Capillary Tube Supplies Ltd SGCT 1.5 mm). The beamline was configured at an X-ray energy of 18 keV and 3.7 m sample to detector distance to give a scattering vector S = (2/λ)sin(θ/2) = Q/(2π) range of 6.5 x 10 -5 to 1 x 10 -1 Å -1 , where θ is the scattering angle and the X-ray wavelength λ is 0.6902 Å. Images were analyzed using the AXcess software package. 3 Briefly, the two-dimensional SAXS images were radially integrated to give one-dimensional diffraction patterns. The Bragg peaks were then fitted using Gaussian functions and indexed by comparison to characteristic peak spacings from known lipid structures.

All-Atom Simulations:
The lipid bilayer was initially created with equilibrated POPC lipids using CHARMM-GUI. 4 A membrane bilayer with 70 POPC lipids in each leaflet was solvated with water molecules, after which 150 mM NaCl was added and the system was equilibrated for 100 ns. The equilibrated POPC lipid bilayer was converted to 10% and 50% POG or POPA systems with unit cell sizes around 65.5 × 65.5 × 89.5 Å 3 comprising ~40000 atoms. All defect systems were run for 200 ns to equilibrate and the last 100 ns were used for analysis.
MD simulations were performed using the NAMD code (version 2.9) 5 with the CHARMM36 lipid, 6 and TIP3P water and NBFIX ion parameters as defined by the CHARMM36 FF. The POG head group (OH) parameters were taken from CHARMM General FF and charges in the OH group were optimized. For POPA, the topology and parameters for the PA head group were taken from the CHARMM36 lipid. 6 Simulations were performed in the NPT ensemble with the temperature and pressure maintained at 300 K and 1 atm, respectively, via Langevin coupling with damping coefficient of 5 ps -1 . All bonds to hydrogen atoms were maintained using the SHAKE algorithm. 7 Periodic boundary conditions were employed with the particle-mesh Ewald algorithm 8 to compute the longrange electrostatic interactions. Lennard-Jones (LJ) potential was switched off within 10-12 Å using a force-switching function. A non-bonded pair list cutoff of 16 Å was used. A time step of 2 fs was maintained throughout the simulations.
Surface Plasmon Resonance: SPR measurements were carried out using a Biacore 3000 and an L1 chip (Sensor Chip L1, GE Healthcare), at 25 ˚C. All solutions were degassed overnight prior to use.

Raman spectroscopy of liposomes incubated with PLC and PLD:
200 nm POPC liposomes (2.5 mg/mL lipid concentration) were incubated for 1 h at 37 ˚C with PLC from Clostridium perfringens (C. welchii), or PLD from Arachis hypegaea (both 50 mU/mL), in PBS buffer containing 0.5 mM CaCl 2 . Subsequently 5 µL of each suspension was pipetted onto a 1 cm 2 MgF 2 slide and evaporated to dryness. A control sample of POPC liposomes incubated without enzyme was also prepared. Raman spectroscopy was carried out on a Witec Alpha 300+ (532 nm laser, 1800 g/mm grating) with laser power set to 1.5 mW and 5 sec integration time. Three to four spectra from different sample positions were collected, the background was corrected using a polynomial subtraction and the background corrected spectra were averaged.

Safety Statement
No unexpected or unusually high safety hazards were encountered. SANS data were fitted using a multilayer vesicle model in the SasView v3.0. 2 program.
The absolute scattering intensity I(Q) was calculated as the sum of terms for volume fractions φ 1 and φ 2 of uni-lamellar and bi-lamellar vesicles, together with a flat background to allow for residual incoherent and/or inelastic scatter. The code fits the total volume fractions of polydisperse particles, but these can be converted to number densities (particles per unit volume) or volume fractions of material in the shells.
Where V i is an appropriate normalization factor for polydispersity function N(r). For spherical shells of thickness T, with gaps G, (thus a "d-spacing" of T+G): Where the shells have neutron scattering length density of ρ s and the bulk solvent core ρ c The uniand bi-lamellar vesicles were allowed to have different core radii, but their Gaussian polydispersity functions were kept at standard deviations of 12%. For a simple spherical particle system, discounting any polydispersity, the volume fraction φ i of a particle population becomes: where N i is the number of particles and V i is the volume per particle/shell. Given particle core radius (r c ), and bilayer thickness (T), one can calculate V p,i and V s,i , respectively the total volume per particle and shell volume per particle. Thus, the proportion of bilamellar sub-population can be calculated as a function of total particle volume (see Table 1), particle number and lipid volume (see Table S1).  Since many conventional phospholipase assays measure enzyme activity on free lipids in solution, and were therefore not appropriate in this study, we used SPR to determine enzyme-induced cleavage of the phosphorous-oxygen bonds of POPC vesicles ( Figure S2) as a qualitative means of comparing enzyme activity. Specifically, we used a Biacore L1 chip, which is a commercially available dextrancoated gold chip modified with alkyl chains, designed for label-free attachment of liposomes. While it is possible for vesicles to fuse and form a homogeneous membrane bilayer on the chip surface, in general the formation of monolayers of intact liposomes is observed on L1 chips. 9 Here  After binding liposomes to the chip, buffer containing 10 mU/mL of either PLC or PLD was passed over the chip surface. On addition of enzyme, we did not observe an increase in RU corresponding to enzyme-substrate binding (K on ). The PLC and PLD activities used in these SPR experiments correspond to 5-50 pgP/mm 3 , which is close to the limit of detection of enzyme binding to the SPR chip. Given this and the increased distance from the gold surface of the chip compared to standard protein binding affinity measurements (caused by the presence of liposomes), it is not surprising that the enzyme-liposome binding events are not observed here.
In all experiments at 10 mU/mL PL activity, a lag time of around 2 min was observed after enzyme addition before a decrease in RU. Similar lag times are reported in the literature for several phospholipases. [10][11][12] During this time it is likely that structural rearrangement is ongoing with increasing concentration of DAG/PA, until a critical rearrangement at which time there is either a burst in enzyme activity or a large-scale change in lipid packing, and the vesicles detach from the chip.
We observed the latter in SAXS data (see Figure S7) for the PC In all cases, the rate of cleavage (determined by the rate of decrease in RU) was greater for PLC than PLD. This could be perhaps justified by the fact that the mass of the DAG cleavage product from PC reaction with PLC is smaller than that of the PA reaction product from PC reaction with PLD.
However, it is more likely due to differences in activity of the enzymes in these specific conditions. This is corroborated by our observations that, although it followed the same trend, the same experiment with PLC from Bacillus cereus led to a slower rate of detachment than either of the two sensorgrams shown in Figure S2 (raw data available). Reducing enzyme activity ( Figure S2C) leads to a less pronounced overall change in RU, but not an appreciable decrease in the lag phase time before decrease in RU. Due to the almost stoichiometric ratio of enzyme molecules to liposomes, it is possible that there are two vesicle sub-populations representing one population with bound enzyme and one without. Similar systems have previously been studied by total internal reflection fluorescence microscopy and deconvoluted to characterize single enzyme activity on single liposomes. 14 This explains the unchanged lag phase between different enzyme concentrations, because it represents the time before detachment of vesicles having been acted on by a single enzyme.    Movie S1. POG head group -OH forms a hydrogen bond with the carbonyl oxygen of POPC in POPC:POG (9:1) bilayers. Only heavy atoms of a POG and a POPC molecule are shown.