PIP2 Reshapes Membranes through Asymmetric Desorption

Abstract

Phosphatidylinositol-4,5-bisphosphate (PIP2) is an important signaling lipid in eukaryotic cell plasma membranes, playing an essential role in diverse cellular processes. The headgroup of PIP2 is highly negatively charged, and this lipid displays a high critical micellar concentration compared to housekeeping phospholipid analogs. Given the crucial role of PIP2, it is imperative to study its localization, interaction with proteins, and membrane-shaping properties. Biomimetic membranes have served extensively to elucidate structural and functional aspects of cell membranes including protein-lipid and lipid-lipid interactions, as well as membrane mechanics. Incorporation of PIP2 into biomimetic membranes, however, has at times resulted in discrepant findings described in the literature. With the goal to elucidate the mechanical consequences of PIP2 incorporation, we studied the desorption of PIP2 from biomimetic giant unilamellar vesicles by means of a fluorescent marker. A decrease in fluorescence intensity with the age of the vesicles suggested that PIP2 lipids were being desorbed from the outer leaflet of the membrane. To evaluate whether this desorption was asymmetric, the vesicles were systematically diluted. This resulted in an increase in the number of internally tubulated vesicles within minutes after dilution, suggesting that the desorption was asymmetric and also generated membrane curvature. By means of a saturated chain homolog of PIP2, we showed that the fast desorption of PIP2 is facilitated by presence of an arachidonic lipid tail and is possibly due to its oxidation. Through measurements of the pulling force of membrane tethers, we quantified the effect of this asymmetric desorption on the spontaneous membrane curvature. Furthermore, we found that the spontaneous curvature could be modulated by externally increasing the concentration of PIP2 micelles. Given that the local concentration of PIP2 in biological membranes is variable, spontaneous curvature generated by PIP2 may affect the formation of highly curved structures that can serve as initiators for signaling events.

Significance

Phosphatidylinositol-4,5-bisphosphate (PIP2) lipids are implicated in numerous cellular processes, several of which involve recruitment of proteins to reshape membranes. Therefore, the elucidation of PIP2-protein interactions is fundamental to being able to understand these phenomena. Here, we show that PIP2-containing giant unilamellar vesicles can assume significant spontaneous curvature. We demonstrate that this curvature preference arises from the spontaneous development of transbilayer PIP2 asymmetry. We find that PIP2 leaching, amplified by PIP2 oxidation, leads to the formation of inward tubules, whereas outer tubules form upon addition of excess PIP2 micelles. Besides representing a mechanism contributing to the shaping of biological membranes, these findings will allow researchers to better interpret their data from model membrane systems that contain PIP2.

Introduction

Phosphatidylinositol-4,5-biphosphate (PIP2) is an essential lipid in the eukaryotic cell plasma membrane, where its localized generation regulates processes such as endocytosis (1), exocytosis (2), actin cytoskeleton dynamics (3), focal adhesion assembly (4), and ion channel function (5, 6). This lipid is characterized by a highly negative electrostatic charge of ≤ −3 at pH 7 (7, 8, 9, 10), as opposed to other housekeeping phospholipids in the mammalian plasma membrane that are electrically neutral (for example, cholesterol), have zero net charge (for example, phosphatidylcholine), or have a single net negative charge (for example, phosphatidylserine) (11). Consequently, proteins with a cluster of basic residues or cations located at the membrane-solution interface can laterally sequester PIP2 (12). This can then allow the presence of multiple functionally independent pools of PIP2 in the plasma membrane, enabling it to modulate diverse signaling events. However, the high electrostatic charge on PIP2 and high degree of unsaturation in the lipid tail also leads to its higher critical micellar concentration (CMC), which has been estimated to be on the order of μM (13). In addition, the polyunsaturated arachidonic lipid tail on PIP2 is susceptible to oxidation (14), and therefore, introduction of polar moieties in the hydrophobic tail region of PIP2 could further enhance the aqueous solubility of PIP2 lipids and effectively increase their CMC. In contrast, bilayer-forming phospholipids such as dipalmitoyl phosphatidylcholine have a CMC value that is several orders of magnitude lower (nM) (15). The higher electrostatic charge on PIP2 also results in a larger hydrodynamic radius of the headgroup region of PIP2, and therefore, the natural tendency of PIP2 is to form micelles rather than bilayer structures (13, 16, 17). When incorporated into lipid bilayers, PIP2 lipids also tend to have a much shorter residence time than other bilayer-forming phospholipids (18, 19).

Biomimetic membranes have served extensively to elucidate structural and functional aspects of cell membranes including protein-lipid and lipid-lipid interactions, as well as membrane mechanics. Incorporation of PIP2 into biomimetic membranes, however, has at times resulted in discrepant findings described in the literature. For instance, Carvalho et al. have reported an increase in the ζ potential of the PIP2-containing vesicles over time, implying decrease of negative surface charge, which was attributed to PIP2 leaching (20). In another report, Beber et al. claimed irreproducibility of protein-membrane binding studies with PIP2-containing vesicles. By means of a fluorescently tagged septin protein, they also observed that the interaction between PIP2 and the septin protein was substantially suppressed within a few hours after the vesicles were prepared (19). Furthermore, these authors questioned the incorporation of PIP2 lipids into biomimetic membranes at low ionic strength.

Earlier work has put forth several hypotheses for these discrepancies, including desorption of PIP2 from the membrane (20). Even though these reports debated the spontaneous desorption of PIP2 lipids over time, there has been no study to systematically characterize the extent of this desorption and, in particular, its effect on the spontaneous curvature of the membrane. In this contribution, we reduce this gap by quantifying the extent of PIP2 desorption and highlighting the membrane-reshaping property of PIP2 lipids.

In this study, we first verify by means of a fluorescent marker that under our rehydration conditions, PIP2 is being incorporated in the biomimetic membranes. Subsequently, by monitoring the fluorescence intensity of a PIP2 marker over time, we confirm that PIP2 lipids are being desorbed. Next, we aimed to asymmetrically desorb PIP2. Upon sequential dilution of PIP2-containing vesicles, we observed increased internal tubulation within minutes, consistent with asymmetric desorption of PIP2. By performing dilution experiments with saturated chain homolog of PIP2, we showed that the highly polar headgroup of PIP2 might not be sufficient to cause fast desorption of PIP2. Furthermore, we quantified the extent of the generated asymmetry by pulling cylindrical tethers, measuring the resulting pulling forces, and ultimately determining the spontaneous curvature of the membranes. We found that vesicles tend to become more asymmetric with an increase in PIP2 content. Lastly, we show that the PIP2-desorption can be suppressed by increasing the PIP2 content in the solution in the form of micelles.

We conclude that PIP2 lipids asymmetrically desorb from biomimetic membranes over time and that this desorption leads to the generation of an appreciable spontaneous curvature even in the absence of any membrane-reshaping proteins. This finding provides a refined outlook for understanding protein-PIP2 interactions on biomimetic membranes, as well as hinting toward a possible mechanism for membrane remodeling in biological membranes that can be modulated by PIP2 lipids.

Materials and Methods

Materials

The lipids 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) and L-α-phosphatidylinositol-4,5-bisphosphate ammonium salt (brain, porcine) (PI(4,5)P2) were obtained from Avanti Polar Lipids (Alabaster, AL). Unless stated otherwise, porcine brain PIP2 will be referred to PIP2 throughout this manuscript. When necessary to distinguish different molecular PIP2 species, porcine brain PIP2 will be called brain PIP2. Porcine brain PIP2 is a natural lipid mixture whose major component (∼74 mol%) contains one saturated (C18:0) and one unsaturated (C20:4) fatty acyl tail. PIP2 diC16 and 1-palmitoyl-2-hydroxy-sn-glycero-3-phosphoinositol (16:0 LPI; lyso-PI) were obtained from Echelon Biosciences (Salt Lake City, UT). Texas Red 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine, triethylammonium salt (TR-DHPE) was obtained from Invitrogen/Life Technologies (Grand Island, NY). HEPES, NaCl, and EDTA were obtained from Fisher Scientific (Rochester, NY). 3 μm latex beads, β-casein from bovine milk, and 98% hydrolyzed 146–186 kDa polyvinyl alcohol (PVA) was purchased from Sigma-Aldrich (St. Louis, MO). All commercial reagents were used without further purification.

Protein purification

The Pleckstrin homology (PH) domain (1–170) of human phospholipase C δ 1 (PLCδ1) (21, 22) was cloned into the pET16b vector with an N-terminal glutathione S-transferase (GST) tag and C-terminal EGFP via VectorBuilder (Santa Clara, CA). The plasmid was obtained in Escherichia coli Stbl3 cells and was extracted with a PureYield miniprep kit (Promega, Madison, WI). EGFP-PH-PLCδ1 was expressed in E. coli BL21-(DE3) cells in an luria broth medium supplemented with ampicillin (100 μg/mL) and was induced with 100 μM isopropyl-β-d-thiogalactoside at 18°C for 21 h. After lysis through tip sonication in lysis buffer (50 mM Tris (pH 8.0), 300 mM NaCl, 1 mM EDTA, 2 mM dithiothreitol), the expressed protein was extracted from the supernatant by a GST affinity column, eluted (with 50 mM Tris (pH 8.0), 150 mM NaCl, 20 mM reduced glutathione, 1 mM EDTA, 2 mM dithiothreitol), and then digested with PreScission protease (GE Healthcare, Pittsburgh, PA). The resulting solution was further purified to remove the GST tag through affinity chromatography. Finally, the eluate was loaded on an equilibrated (20 mM HEPES (pH 7.4), 150 mM NaCl) Superdex 200 16/600 column (GE Healthcare, Chicago, IL) for size exclusion chromatography. Protein purity was assessed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), and the concentration of purified protein was calculated by measuring the EGFP absorbance at 488 nm. The purified protein was aliquoted and stored at −80°C.

To exclude potential aggregation and/or degradation of EGFP-PH-PLCδ1 throughout the duration of experiments, we used ultracentrifugation with ultraviolet-visible spectroscopy and SDS-PAGE, respectively, to analyze the state of the biosensor. The stock solution of purified EGFP-PH-PLCδ1 was stored frozen at −80°C. Freezing typically resulted in some aggregation of the protein. Therefore, we used a Sorvall MX 120+ Micro-Ultracentrifuge (Thermo Fisher Scientific, Waltham, MA) to initially prepare the EGFP-PH-PLCδ1 sample by centrifuging (using a S120-AT3 rotor) at 94,000 rotations per minute (rpm) at 4°C for 30 min to obtain aggregate free sample. We then measured the absorbance of this initial sample at 280 and 488 nm (Nanodrop 2000c; Thermo Fisher Scientific). Then, the sample was incubated at room temperature for 1 h, and subsequently, the absorbance was measured, followed by a final 6-min spin at 94,000 rpm. No change in absorbance values for these two samples confirmed the absence of aggregation of EGFP-PH-PLCδ1 during this time period. Similarly, to check for degradation of EGFP-PH-PLCδ1 over time, we ran a 12% SDS-PAGE gel on two samples, one that remained on ice and one that was incubated at room temperature for 1 h, after the EGFP-PH-PLCδ1 sample was ultracentrifuged at 94,000 rpm for 15 min after thawing from −80°C on ice. The SDS-PAGE gel was stained with GelCode Blue Stain Reagent (Thermo Fisher Scientific) and destained with MilliQ water. No significant degradation of the EGFP-PH-PLCδ1 was observed, as shown in Fig. S1.

Circular dichroism

Circular dichroism measurements were performed using AVIV circular dichroism spectrometer model 425 (Aviv BioMedical Inc., Lakewood, NJ) with 0.1 cm pathlength cuvette. EGFP-PH-PLCδ1 samples were measured at room temperature at a concentration of 5 μM in 20 mM HEPES, 150 mM NaCl (buffer pH 7.4) (Fig. S2). All resulting spectra were subjected to subtraction of the background control spectra.

Thin layer chromatography of lipids

Thin layer chromatography (TLC) was used to validate the quality of brain PIP2 lipids. Silica-coated TLC plates (HPTLC Silica Gel 60, 10 × 10 cm; Merck Darmstadt, Darmstadt, Germany) were pretreated with 1% potassium oxalate in methanol:water (2:3 v/v) for 12 h at room temperature and dried at room temperature. 2 μg of PIP2 lipids were spotted on TLC plates and placed in a closed glass chamber with chloroform/acetone/methanol/acetic acid/water (46:17:15:14:8 v/v). Subsequently, the TLC plate was air-dried, and lipids were visualized by spraying the plate with chromo-sulfuric acid (0.6% w/v K₂Cr₂O₇, 30% v/v H₂SO₄) followed by charring at 120°C on a hot plate.

GUV formation methods

Giant unilamellar vesicles (GUVs) were prepared by electroformation in 110 mOsm sucrose solution as described elsewhere (23). Briefly, lipids were mixed in chloroform (or in a 2:1 mixture of chloroform/methanol when PIP2 was present) at a total concentration of 1 mM. PIP2 was used at x (= 0, 0.5, 2, or 5) mol%, POPC at (99.8 − x) mol% and TR-DHPE at 0.2 mol %. PI(4,5)P2 diC16 was used at 5 mol%, whereas lyso-PI (16:0 LPI) was used at 2 mol% for preparing GUVs. Each of these latter compositions contained 0.2 mol% TR-DHPE, with the remaining fraction consisting of POPC. The 40 μL solution of the lipid stock with the desired composition was spread on each of the indium tin oxide (ITO)-coated slides (Delta Technologies, Stillwater, MN) and then put under vacuum for at least 2 h to form a dry lipid film (24). The electroformation chamber was formed by two lipid-coated ITO slides separated by a silicone spacer. The lipid film was hydrated with 450 μL of 110 mOsm sucrose solution and connected to an alternating current field (4 Hz) with a peak-to-peak amplitude of 4 V for 2 h at room temperature to produce the final GUV dispersion.

To prepare GUVs in salt solutions, a PVA-gel hydration-based method was used as in Weinberger et al. (25). A dry PVA film was formed by spreading 25 μL of the 5% PVA solution onto a glass coverslip and dried at 70°C for 30 min. Lipid stock solution was then spread onto the dried PVA patch, and the solvent was evaporated by applying vacuum for at least 1 h. The lipid film was then hydrated with 225 μL of buffer solution and left to swell for 30 min to an hour. Osmolarities were adjusted to 110 or 320 mOsm depending on salt concentration, as detailed below.

Micelle preparation

PIP2 from porcine brain (Avanti Polar Lipids), dissolved in chloroform/methanol/water (20:9:1, v/v) at a concentration of 1 mg/mL, was dried in a glass flask under nitrogen and then overnight under vacuum. The dried PIP2 lipids were rehydrated in 500 μL solution of 20 mM HEPES and 45 mM NaCl buffer by vortexing for 15 s and then sonicating in a Branson 1510 ultrasonic cleaner (Branson Ultrasonics, Danbury, CT) for 40 min. Micelles were determined to have a mean hydrodynamic radius of 18.4 nm, measured by dynamic light scattering (Zetasizer Nano ZS; Malvern Panalytical, Malvern, UK).

Micropipette aspiration

Micropipettes were fabricated by clipping the tips of pulled capillaries (World Precision Instruments, Sarasota, FL) using a microforge. The inner diameters of the micropipettes were ∼7 μm. To prevent irreversible membrane/pipette adhesion, micropipette tips were incubated with 5 mg/mL pure β-casein for at least 15 min using a MicroFil needle (World Precision Instruments), and the pipettes were then filled with the chamber solution. The lateral membrane tension—which is the sum of a bare tension, Σ_bare (arising because of constraints on the area of the membrane segment) and a contribution due to spontaneous curvature—is equal to the aspiration tension, Σ_asp, which can be related to the pipette aspiration pressure (24, 26) as

$${\Sigma }_{asp}={\Sigma }_{bare}+\frac{1}{2}\kappa C_0^2\approx \frac{R_p{R}_v\mathit{\Delta }P}{2\left(R_v-R_p\right)}.$$ (1)

Here, κ is the local bending rigidity of the membrane, C₀ is the spontaneous curvature of the membrane, R_p is the radius of the cylindrical portion within the micropipette, R_v is the radius of the spherical segment of the vesicle, and ΔP is the suction pressure.

Fluorescence intensity measurements

The PH domain of PLCδ1 (PH-PLCδ1) is known to bind specifically to the headgroup of PI(4,5)P2 because of strong interactions between PH domain residues and the 4-and 5-phosphates on the inositol ring (27). The fluorescence intensity of PH-PLCδ1 fused with enhanced green fluorescent protein (EGFP) has been extensively used to detect PIP2 within the plasma membrane, as well as in model membrane systems (28, 29, 30, 31). The GUVs incubated with the EGFP-PH-PLCδ1 were imaged by confocal fluorescence microscopy. To perform the incubation experiments with EGFP-PH-PLCδ1, the vesicles were first diluted 100-fold in a 1:1:1 solution of sucrose (110 mM)/glucose (110 mM)/(20 mM HEPES + 45 mM NaCl) at pH 7.4. The diluted vesicles were then incubated with 200 nM EGFP-PH-PLCδ1 for 15 min in an Eppendorf tube. The solution was then transferred to an imaging chamber, prepared by squeezing solution between two coverslips that were held together with vacuum grease, and then imaged. The sequential line scan feature of the confocal microscope (IX83; Olympus, Center Valley, PA) was used to avoid any fluorescence bleedthrough between the lipid fluorophore and the protein fluorophore channel. To prevent spontaneous rupture of vesicles on the glass surface, the glass coverslips were pretreated with 5 mg/mL of β-casein solution.

Dilution experiments

The GUV dispersion obtained after electroformation was transferred into an Eppendorf tube and allowed to sit overnight. This was done to dissipate any residual membrane tension that could be present in the vesicles because of electroformation. The vesicles were then diluted either 10-fold or 100-fold in the GUV preparation solution (a 110 mOsm sucrose solution) and imaged after 15 min of dilution. Vesicles without dilution were imaged after a similar time delay. Increasing the dilution factor led to an increase in the proportion of internally tubulated vesicles within 15 min. These dilution experiments were performed for GUVs in presence and in the absence of PIP2 lipids in the membrane. The vesicles were characterized as internally tubulated if they had tube-like structures toward the interior of the vesicle. Additionally, the tubules needed to have dimensions above the diffraction limit and to be visibly connected to the periphery of the GUV (Fig. S5).

Chamber preparation and tether pulling

The GUV dispersions were diluted 100-fold in a 1:1:1 solution of sucrose (110 mM)/glucose (110 mM)/20 mM HEPES + 45 mM NaCl in an Eppendorf tube. For experiments with PIP2-containing vesicles, 0.05 mM EDTA was also added to this solution to chelate any remnant calcium ions. The diluted GUV dispersion (240 μL) was mixed with 0.5 μL of 3 μm latex bead solution and injected into a measurement chamber (25 × 10 × 1 mm) constructed from 150-μm-thick glass coverslips. To facilitate the interaction between the latex bead and the PIP2-containing vesicles to pull out tethers, the latex beads were incubated with 200 nM EGFP-PH-PLCδ1 for 30 min to allow nonspecific adsorption of the protein onto the beads. These beads were subsequently spun down and washed with a buffer containing 20 mM HEPES and 45 mM NaCl three times before mixing them with GUVs. For tether-pulling experiments from vesicles not containing PIP2, latex beads were used directly from their stock solution without incubation with EGFP-PH-PLCδ1 (26, 32).

The glass coverslips used to prepare the chamber were pretreated by immersing them in a solution of 5 mg/mL β-casein and 0.05 mM EDTA and were subsequently rinsed with deionized water. The solution chamber was then mounted on an inverted microscope (IX71; Olympus) equipped with a 60× 0.9 NA water immersion objective (Olympus), a TR-DHPE filter cube (Chroma, Rockingham, VT), and a back-illuminated electron-multiplying charge-coupled-device camera (ImagEM; Hamamatsu, Bridgewater, NJ). A second independently controlled objective (60×, 1.1 NA, water immersion, long working distance; Olympus), oriented opposite to the imaging objective was used for optical trapping using a 1064 nm laser (IPG Photonics, Oxford, MA) (33). A three-dimensional motorized micromanipulator system (Luigs & Neumann, Ratingen, Germany) was used to insert a micropipette into the solution chamber. After insertion, a small fluorescent vesicle was stabilized within the pipette tip to attain a zero-pressure differential. The adjustments in the height of an attached water reservoir, measured with a pressure transducer (Validyne Engineering, Los Angeles, CA), allowed control of the aspiration pressure. The chamber was left undisturbed for ∼15 min to allow settling of GUVs and beads at the bottom of the chamber. Vesicles of roughly 10–20 μm in diameter, which displayed sufficient excess area, were selected such that aspiration at low pressure led to a projection (referred to as the fraction of the vesicle aspirated into the pipette) with a length greater than the pipette radius. A vesicle was pipette-aspirated with an initially small suction pressure (∼10 Pa) while the trapping objective was brought to focus on a membrane-free bead. The trapped bead and the GUV both were moved up by ∼80 μm from the bottom of the chamber to prevent interference from other vesicles and beads. The trap stiffness was obtained at this height (see below), after which the trapped bead was then positioned adjacent and coplanar to the aspirated vesicle. The aspirated vesicle was slowly moved toward the bead until contact and was then moved away to pull a membrane tube of ∼10 μm in length. The suction pressure was changed in increments of ∼10 Pa, starting from ∼10 to 60 Pa and then back to ∼10 Pa while simultaneously recording the pulling force applied on the bead through the tether.

Spontaneous curvature determination from plots of pulling force versus tension^0.5

The energy functional of a thin cylindrical tether connected to a large, pipette-aspirated GUV (34) can be written as

$$F=\frac{\kappa }{2}\int dA{\left(C_1+C_2-C_0\right)}^2+\frac{\overline{\kappa }\mathit{\pi }}{2\overline{A}{D}^2}{\left(\text{Δ}A-\overline{\mathit{\Delta }{A}_0}\right)}^2-f_{0}L+\text{Σ}A,$$ (2)

where κ is the bending rigidity, $\overline{\kappa }$ is the nonlocal bending rigidity (35), C₁ and C₂ are the principal curvatures of the tether, C₀ is the spontaneous curvature of the membrane (resulting from asymmetric distribution of lipids with nonzero molecular spontaneous curvature), $\overline{A}$is the total area of the vesicle plus tether, D is the membrane thickness, f₀ is the tether-pulling force, ΔA is the area difference between the two bilayer leaflets, $\overline{\mathit{\Delta }{A}_0}$ is the preferred area difference between the two bilayer leaflets, L is the length of the pulled tether, R is the radius of the tether, Σ = Σ_bare is the bare membrane tension to be distinguished from the aspiration tension, and A is the area of the tether.

The mechanical equilibrium of the tether results from the established balance between the lateral tension and the membrane rigidity of the membrane. The equilibrium radius R₀ and pulling force f₀ can then be calculated by minimizing the free energy of the tube with respect to the tether radius and tether length, resulting in

$$f_0=2\pi \sqrt{2\kappa {\Sigma }_{asp}}-2\pi \kappa C_0+\frac{4{\pi }^3\overline{\kappa }L\left(1-\mathit{\Delta }{a}_0\right)}{\overline{A}}=2\pi \sqrt{2\kappa {\Sigma }_{asp}}-2\pi \kappa C_{eff},$$ (3)

where $\mathit{\Delta }{a}_0=\left(\mathit{\Delta }\overline{A_0}\right)/\left(2\pi LD\right)$ and $C_{eff}=C_0-\left(2{\pi }^2\alpha L\left(1-\mathit{\Delta }{a}_0\right)\right)/\left(\overline{A}\right)$. Here, α is defined as the ratio of nonlocal bending $\left(\overline{\mathrm{\kappa }}\right)$ and local bending rigidity (κ) and is approximately equal to unity (36).

Accordingly, by plotting the equilibrium pulling force f₀ against the square root of membrane tension √Σ_asp, the spontaneous curvature can be deduced from the intercept (37). Note that C_eff has two contributions: the intrinsic curvature of (asymmetrically distributed) constituent lipids (C₀), as well as from the area difference between the two leaflets of the bilayer membrane arising because of lipid number density asymmetry $\left(\left(2{\pi }^2\alpha L\left(1-\mathit{\Delta }{a}_0\right)\right)/\left(\overline{A}\right)\right)$.

Optical trap design and calibration

The optical trap setup was custom built on an inverted microscope (IX71; Olympus) and has been described previously (33). A movable objective lens attached to a stage was used to translate a trapped bead with submicron precision along the axis of tubes pulled from vesicles without measurable change in trap stiffness. The drag-force method (38) was used to measure the stiffness of the trap and was typically found to be around 0.05 pN/nm. The tether-pulling forces were measured from the displacement of the bead from the trap center.

Image analysis

The radius of the aspirated GUVs and the length of pulled tethers were measured using ImageJ software (National Institutes of Health, Bethesda, MD). The fluorescence intensities of the EGFP-PH-PLCδ1 bound to GUVs were obtained by fitting a two-dimensional Gaussian ring and extracting its amplitude in MATLAB (The MathWorks, Natick, MA).

Results and Discussion

PIP2 is incorporated in the membrane

GUVs are most frequently prepared using sucrose solutions at low ionic strength (26, 37, 39) because this improves the yield and quality of vesicles relative to protocols suitable for use at higher, physiological ionic strength. Earlier reports have debated the incorporation of PIP2 lipids in biomimetic membranes under different rehydration conditions. In particular, Beber et al. have claimed that PIP2 cannot be incorporated into biomimetic membranes in the absence of salt in the solution. Opposing this finding, Carvalho et al. have demonstrated that PIP2 can be incorporated in the membrane in the absence of salt. To verify that our preparation protocol leads to incorporation of PIP2 into GUVs in the absence of salt, we formed vesicles in the presence or absence of PIP2. GUVs also contained a trace amount (0.2 mol%) of TR-DHPE to facilitate visualization of the membrane. We used the PIP2 specific marker EGFP-PH-PLCδ1 (40, 41) to monitor incorporation of PIP2 in the membrane. Each vesicle preparation was incubated with 200 nM EGFP-PH-PLCδ1 for 15 min and then imaged. A clear periphery of the membrane can be seen in the TR-DHPE channel for vesicles with and without PIP2 (Fig. 1 a). Vesicles without PIP2 showed no detectable fluorescence in the EGFP channel, whereas those containing 5% PIP2 showed a clear periphery of the PIP2 specific sensor EGFP-PH-PLCδ1 on the membrane (Fig. 1 a). This confirmed that through our preparation protocol, we are able to incorporate PIP2 into the membrane.

Figure 1.

PIP2-containing membranes exhibit internal tubules, hypothesized to be due to asymmetric lipid desorption. The GUVs were prepared using the electroformation protocol described in the Materials and Methods. EGFP-PH-PLCδ1 was used as a marker to detect the incorporation of PIP2 lipids in the membrane. (a) No binding of EGFP-PH-PLCδ1 was observed when PIP2 was absent from the membrane. For the GUVs composed of POPC and TR-DHPE (referred as TR channel in the figure above), no fluorescence was observed in the EGFP channel, whereas in the presence of 5% PIP2, fluorescence intensity from the EGFP channel was clearly detectable. The 100-fold dilution of PIP2-containing vesicles for the incubation experiments with EGFP-PH-PLCδ1 resulted in generation of internal tubules as observed after dilution followed by 15 min of incubation. Experiments +/− PIP2 were carried out under the same illumination/detection conditions and at the same protein concentration. Scale bars, 5 μm. We hypothesize these tubules to result from asymmetric desorption of PIP2 lipids. (b) A cartoon representation shows PIP2 desorption, which occurs because of its higher CMC value compared to POPC. Upon dilution and incubation as mentioned above, the PIP2 lipids desorb asymmetrically in a larger number from the outer leaflet simply because the outer leaflet has a larger accessible solution volume than the inner leaflet. This asymmetry induces tensile and compressive stresses in the outer and the inner leaflet, respectively, that can be released by bending the bilayer membrane toward the interior of the vesicle. To see this figure in color, go online.

Typically, measurements by tether pulling of mechanical membrane properties are carried out with GUVs that are diluted relative to the initially formed dispersion (26, 33, 37). The rationale behind this practice is to improve imaging quality and to reduce premature coating of pulling beads with lipids. Interestingly, upon 100-fold dilution of PIP2-containing vesicles, we observed development of internal tubules in the majority of these vesicles (Fig. 1 a). We hypothesized that the development of internal tubules on dilution of PIP2 vesicles could be a result of asymmetric desorption of PIP2 lipids (Fig. 1 b). We elaborate on experiments performed to evaluate this hypothesis in the subsequent sections.

PIP2 desorbs from the outer leaflet of the bilayer membrane

To evaluate PIP2 desorption from the membrane, we used the fluorescence intensity of membrane-adsorbed EGFP-PH-PLCδ1 as a readout for the content of PIP2 on the outer leaflet of the membrane. We first incubated freshly prepared GUVs containing 0.5% PIP2 with EGFP-PH-PLCδ1 and measured the fluorescence intensity. Alternatively, the same vesicle preparation was incubated with the PIP2 sensor protein 12 h after vesicle formation. This revealed that the fluorescence intensity had decreased over time, as shown in Fig. 2 for electroformed vesicles (15 mM NaCl). In both cases, before imaging, the vesicles were diluted 100-fold in a 1:1:1 solution of sucrose (110 mM)/glucose (110 mM)/(20 mM HEPES + 45 mM NaCl) and were then incubated with the protein solution for 15 min. The fluorescence intensity values were normalized with respect to the value obtained for the earlier measurement. We observed a decrease in fluorescence intensity to 72% (12 h after vesicle preparation) compared to the initially measured fluorescence intensity for n = 33 vesicles. This suggested that PIP2 content decreased spontaneously over time on the outer leaflet as a result of 12 h aging.

Figure 2.

PIP2 spontaneously leaches out of the membrane over time. Vesicles were prepared using two different swelling methods, namely electroformation using ITO-coated slides and gel-assisted GUV formation in either 110 or 320 mOsm sucrose solution, depending on the NaCl concentration in the exterior. For the case in which the ionic strength of the outer solution was 15 mM, the lipid composition of POPC (99.3%), PIP2 (0.5%), and TR-DHPE (0.2%) was used. However, when the ionic strength was increased to 150 mM NaCl, the PIP2 content in the membrane was increased to 2% to facilitate binding of EGFP-PH-PLCδ1. Immediately after preparation, GUVs were incubated with 200 nM EGFP tagged with PH-PLCδ1 for 15 min, and then the fluorescence intensity of the EGFP-PH-PLCδ1 bound to the membrane was recorded for each of these GUV formation methods. A different sample of GUVs from the same batch was allowed to sit overnight and was only then incubated with EGFP-PH-PLCδ1, again for 15 min before imaging. The decrease in the fluorescence intensity of the bound protein under different swelling conditions, as well as different ionic strengths, showed that PIP2 leached out of the vesicles over time at both high and low ionic strengths and was not simply an artifact of our GUV preparation protocol. All values were normalized with the fluorescence intensity at time zero for each preparation protocol. A total of 33 vesicles were used for imaging for each of these conditions. Fluorescence images used for analysis were qualitatively equivalent to those shown in Fig. 1a, bottom row. Errors were estimated as SEM of all the values. Data points are mean ± SEM. ^∗p < 0.05.

We considered the following alternative interpretations of our findings above. Steinkühler et al. (42) have shown that electroformation of vesicles on ITO-coated slides (23) may result in an asymmetric distribution of negatively charged lipids. Per se, this might result in an increased concentration of PIP2 lipids on the outer leaflet compared to the inner leaflet right after electroformation. Once this asymmetry would relax through passive lipid flip-flop, the concentration of PIP2 would decrease on the outer leaflet. This could then result in decreased binding of EGFP-PH-PLCδ1. To eliminate this possibility, we prepared vesicles using gel-assisted GUV formation, following the protocol from Weinberger et al (25). Steinkühler et al. demonstrated that gel-assisted GUV formation results in a symmetric distribution of negatively charged lipids in GUVs. Using these GUVs, we performed exactly the same incubation experiments as those for the electroformed GUVs. GUVs prepared by gel-assisted formation also showed a decrease in the EGFP-PH-PLCδ1 fluorescence intensity upon aging of the vesicles. Fig. 2 shows the fluorescence intensity values after normalization, again comparing aged to nonaged vesicles. These results confirmed that PIP2 content of the outer leaflet indeed decreased over time and that the loss in fluorescence intensity was not an artifact of our GUV preparation protocol.

There could be three additional reasons for PIP2 content in the outer leaflet to decrease over time. First, the PIP2 lipids could be selectively flipped toward the interior, and therefore, their concentration would decrease in the outer leaflet. This seems unlikely because there is no driving force to selectively flip PIP2 lipids toward the interior leaflet over a period of 12 h. If anything, we would expect the flipping time of PIP2 lipids to be on the order of days, given that other housekeeping phospholipids such as POPC are known to have a half-life of flip-flop ranging from several hours to days (43, 44). Because the headgroup of PIP2 is highly charged, the energetic cost for the transport of this headgroup through the hydrophobic interior of the bilayer membrane is likely to be significantly higher than that for zwitterionic phospholipids. This would render the flipping timescale of PIP2 lipids significantly longer than the timescale of our experiments.

The second reason for the decrease in outer leaflet PIP2 content could be chemical degradation of PIP2 lipids. In an earlier study, Beber et al. showed, through mass-spectrometry-based lipidomics, absence of PIP2 degradation within 5 h in an aqueous buffer (19). However, those studies did not eliminate the possible presence of oxidized products in the PIP2 lipid stocks.

The third and the most likely reason is that PIP2 lipids desorb over time from the membrane, and this desorption could be enhanced by applying a dilution. PIP2 desorption has been earlier proposed by Carvalho et al., who observed an increase in the ζ potential values on GUVs and LUVs over time (20). This desorption can be explained by the high CMC value of PIP2 lipids as compared to other housekeeping lipids. This relatively higher CMC value means that a larger number of free PIP2 lipids can remain in the solution before forming micellar or bilayer structures. The CMC value of PIP2 lipids could be further raised by their susceptibility to oxidation. PIP2 lipids desorbing into the solution further becomes favorable upon GUV dilution, when PIP2 lipids leach to re-establish partition equilibrium. Finally, the lipids that desorb to reach equilibrium are likely to form micelles because of the conical lipid shape, which is itself caused by a highly charged headgroup and increased hydrodynamic radius. Overall, PIP2’s high CMC, combined with the experimental GUV dilution step, is the most logical explanation for PIP2 desorption over time. Our interpretation of Carvalho et al.’s result is consistent with that of Beber et al., who also interpreted the increase in ζ potential values for PIP2-containing GUVs over 24 h to desorption of PIP2 lipids from the membrane (19). Importantly, both of those studies used brain PIP2, which is prone to oxidation, and this could have contributed to the desorption of PIP2 as observed by these two research groups.

In our experiments, we have used an ionic strength of 15 mM NaCl, which is lower than physiological conditions. This served to minimize the effect of an asymmetric distribution of ions on the spontaneous membrane curvature (45, 46, 47). It has been reported that the CMC of PIP2 lipids decreases with increasing ionic strength (48, 49). To verify that the decrease in fluorescence intensity that we observed at the low ionic strength of 15 mM NaCl was relevant at physiological ionic strength, we increased the concentration of NaCl in the outer solution to 150 mM. We then performed the GUV-incubation experiments with 200 nM EGFP-PH-PLCδ1 immediately after electroformation and then after 12 h as shown in Fig. 2 (150 mM NaCl). Because the interaction between EGFP-PH-PLCδ1 and PIP2 lipids is electrostatic in nature, increasing the ionic strength in the solution shielded the interaction between EGFP-PH-PLCδ1 and PIP2 lipids in the membrane. To compensate for this reduction, we increased the PIP2 content in the membrane from 0.5% to 2%. We still observed the decrease in fluorescence intensity of bound EGFP-PH-PLCδ1 as the vesicles aged (Fig. 2; Fig. S4). Our observations are in line with those of Beber et al. (19), who used high ionic strength (75 mM NaCl) solutions in experiments with PIP2 vesicles and also noticed a decrease in binding of the fluorescently tagged septin protein with PIP2-containing vesicles. These observations support our claim of significant PIP2 desorption from bilayers at both low and high ionic strengths.

PIP2 asymmetrically desorbs from the bilayer membrane

A recent report by Dasgupta et al. (26) has shown that bulky monosialotetrahexosylganglioside (GM1) lipids can asymmetrically desorb from bilayer membranes upon dilution. Therefore, we sequentially diluted vesicles using the GUV preparation solution (a 110 mOsm sucrose solution) to enhance asymmetric desorption of PIP2. We observed an increase in the number of vesicles with internal tubulation (Fig. 3 a) with the increase in the dilution factor for vesicles containing PIP2. As a control, we performed a similar dilution for vesicles containing only POPC and TR-DHPE (Fig. 3 b). For these vesicles, there was no increase in the number of internally tubulated vesicles, thus suggesting that in the presence of PIP2, negative curvature developed because of its asymmetric desorption.

Figure 3.

Sequential dilution of PIP2-containing vesicles leads to an increase in the number of internally tubulated vesicles. Vesicles were prepared with POPC (with or without PIP2) and TR-DHPE using electroformation in 110 mOsm sucrose solution. (a) shows a sequential increase in the percentage of internally tubulated vesicles on diluting the GUV dispersion solution by 10-fold and 100-fold, respectively, in 110 mOsm sucrose solution for vesicles containing PIP2. (b) A similar sequential dilution of vesicles without PIP2 showed no statistically significant change in the number of internally tubulated vesicles. Before performing dilution experiments, the vesicles were allowed to age for 12 h after preparation (without changing solution conditions) to relax any residual tension that might be present because of the electroformation process. The vesicles were imaged for tubules, 15 min after dilution. A total of 30 vesicles were imaged for both these lipid compositions. The data are represented as mean ± SEM. ^∗p < 0.05.

Diluting the vesicles creates a new partition equilibrium between the membrane and the solution. Because the interior volume of the vesicle is much smaller compared to the outside chamber solution (Fig. 1 b), more lipids need to be desorbed from the outer leaflet to reach a new partition equilibrium. Since the CMC values of zwitterionic phospholipids are on the order of nM (compared to the μM range for PIP2), they tend to remain in the bilayer even on dilution, with a residence time of days (44). However, PIP2 lipids will leach out of the membrane, as confirmed by the decrease in fluorescence intensity values obtained earlier using EGFP-PH-PLCδ1. Therefore, upon dilution of PIP2-containing vesicles, more PIP2 lipids will be desorbed from the outer leaflet to reach the new partition equilibrium between the membrane and the solution as compared to the inner leaflet. This lipid number asymmetry will induce tensile and compressive stresses in the outer and the inner leaflet, respectively. These stresses can be relaxed such that the inner, compressed leaflet expands while the outer, dilated leaflet shrinks. This is achieved by bending the bilayer membrane toward the interior of the vesicle, resulting in internal tubule formation. It should be noted that the PIP2 desorption occurs spontaneously and is attributable to its conical shape and relatively high CMC. Dilution is likely to accelerate this spontaneous desorption, consistent with the observation of internal tubules within 15 min of dilution.

However, for double-chained phospholipids, it is known that the intervesicular transfer rate of lipids is on the order of hours to days (50, 51, 52), whereas in our experiments, we observe internal tubule formation within 15 min of dilution. A similar observation has been made for GM1 lipids by Dasgupta et al., who showed internal tubules emanating from the vesicles within a few minutes after dilution (26). This prompted us to explore the possibility that molecularly modified PIP2 caused this fast desorption.

One of the possibilities, in case of PIP2, could be the presence of four double bonds in one of its lipid tails. The presence of an increasing number of double bonds is known to decrease the desorption time of lipids by roughly one order of magnitude each (53). Therefore, in addition to the presence of a highly polar headgroup, the presence of four double bonds could substantially reduce the desorption time of PIP2 lipids compared to other housekeeping phospholipids. To test this, we performed dilution experiments with saturated chain PIP2 lipids (diC16 PIP2) under exactly the same conditions as was done with brain PIP2 lipids. Interestingly, we did not observe the generation of internal tubules within 1 h of 100-fold dilution of GUVs containing 5% saturated chain PIP2 lipids (Fig. 4 a). This showed that the presence of unsaturated lipid tail of PIP2 is contributing to the fast desorption of PIP2 from GUVs, as evidenced by our experiments. We confirmed that the saturated chain PIP2 had been incorporated in the membrane by checking the binding of EGFP-PH-PLCδ1 to GUVs containing 5% diC16 PIP2. Clear fluorescence from the periphery of these GUVs confirmed that saturated PIP2 had been incorporated into the membrane (data not shown).

Figure 4.

Evaluating the role of polyunsaturated lipid tail on fast desorption of PIP2. (a) No increase in the percentage of internally tubulated vesicles is observed on diluting the GUV dispersion 100-fold in 110 mOsm sucrose solution for vesicles containing 5 mol% of saturated chain homolog of PIP2 (diC16 PIP2) within 1 h of dilution. (b) On the other hand, the presence of a single saturated chain in lyso-PI (16:0 LPI) appears to lead to desorption from the membrane within 15 min of dilution. This was shown by observing an increase in the proportion of internally tubulated vesicles upon dilution for membranes containing 2% lyso-PI. (c and d) GUVs prepared from freshly opened brain PIP2 lipid stock do not show an increase in the proportion of internally tubulated vesicles upon 100-fold dilution, whereas the GUVs prepared from a couple months old brain PIP2 do result in desorption of lipid within 15 min of dilution. A total of 33 vesicles were imaged for each trial of these lipid compositions. The data are represented as mean ± SEM. ^∗p < 0.05.

Another possibility for phospholipids containing polyunsaturated fatty acids, such as porcine brain PIP2, is their susceptibility to oxidative modification by reactive oxygen species (14, 54, 55, 56, 57, 58, 59). Polyunsaturated fatty acids readily undergo peroxidation under various conditions to afford a plethora of different reaction products, including truncated phospholipids and different types of low molecular weight aldehydes (60). The process of oxidation of the unsaturated acyl chains of the phospholipids introduces polar moieties to the originally hydrophobic parts of lipids, which dramatically changes their biophysical parameters and, consequently, the properties of membranes containing them (59). Therefore, it is also possible that the arachidonic lipid tail of brain PIP2 is modified to produce truncated PIP2 that desorbs within 15 min of dilution.

Earlier studies on desorption kinetics of double-chained phospholipids and lysolipids have shown a reduction in desorption time from hours to minutes for lysolipids compared to double-chained phospholipids (50, 61, 62). Therefore, to test whether the absence of one lipid tail can significantly affect the desorption time of saturated chain PIP2 lipids, we used lyso-PI (16:0 LPI), which has a relatively less polar headgroup compared to PIP2 but, more importantly, has only one hydrophobic lipid tail. Lyso-PI was chosen because of the unavailability of commercial lyso-PIP2. In our dilution experiments with GUVs containing 2% lyso-PI, we observed internal tubules within 15 min of 100-fold dilution (Fig. 4 b). This observation suggests that if one of the products of lipid oxidation resulted from truncation of a PIP2 fatty acyl tail, this oxidation product could desorb within minutes.

We also compared freshly opened brain PIP2 lipid stock and lipid stock that was a few months old (which was used in most experiments). Consistent with our brain PIP2 oxidation hypothesis, we did not observe internal tubules within 15 min of 100-fold dilution for the GUVs prepared from freshly purchased brain PIP2 stock (Fig. 4 c), whereas GUVs prepared from a few months old stock of brain PIP2 again resulted in internal tubules within 15 min of dilution (Fig. 4 d). Comparison between the TLC of new and old stock of brain PIP2 showed smeared spots for the old brain PIP2 that could be due to oxidation of brain PIP2 lipids (Fig. S3).

In the experiments done by Dasgupta et al. on desorption of GM1 lipids, they observed the appearance of internal tubules within minutes, which they interpret as fast desorption of GM1 lipids from the membrane. However, the measured spontaneous desorption time of tritiated GM1 has been shown to be on the order of days (63, 64). Therefore, the fast desorption of GM1 that has been observed by Dasgupta et al. might also be due to the presence of oxidation products resulting from the unsaturation present in its tail region. We also cannot eliminate the possibility of copurified lysolipids and other kinds of fast-leaching amphiphiles in the ovine brain GM1 that was used in their study (personal communication). From a biological perspective, lipids such as PIP2 that define the nature of an organelle membrane might not benefit from desorbing from that membrane within minutes. The same might apply to gangliosides that define the cell surface for identification through extracellular molecules.

Quantifying the spontaneous curvature generated because of PIP2 desorption

The dilution experiments leading to inward tubulation within 15 min of dilution showed that PIP2 desorption results in negative membrane curvature. To quantify the spontaneous curvature that is being generated by this desorption, we carried out membrane-tether-pulling experiments at 100-fold dilution of vesicles. A single vesicle was aspirated using a micropipette that was connected to a water reservoir. The membrane tension of the aspirated vesicle was controlled by adjusting the height of this water reservoir using a method described by Tian et al. (24). Next, a latex bead was optically trapped and manipulated to pull a thin narrow tether from the aspirated vesicle. The latex beads used to pull tethers from PIP2-containing vesicles were coated with EGFP-PH-PLCδ1, as described in the Materials and Methods. The membrane tension was systematically varied from 10 to 60 Pa in steps of 10 Pa each. Simultaneously, the pulling force needed to keep the bead stable at each tension value was recorded. The pulling force was plotted against the square root of membrane tension and fitted to a straight line. The slope of this linear fit allowed us to obtain the bending rigidity of a vesicle, whereas the y-intercept allowed us to deduce the spontaneous curvature of the membrane using Eq. 3.

Fig. 5 a shows pulling forces at different membrane tensions for three different lipid compositions: POPC only, 2% PIP2, and 5% PIP2. Here, “POPC only” refers to POPC + 0.2% TR-DHPE vesicles, and data points corresponding to that composition are represented by stars. The slope of the linear fit of these data points resulted in a bending rigidity of 1.17 ± 0.09 × 10⁻¹⁹ J, which is in agreement with the literature values for POPC vesicles (65). The y-intercept of this fit was 0.32 ± 0.55 pN (standard error of the fit), which corresponds to a negligible spontaneous curvature as shown in Fig. 5 b. There was a minor solution asymmetry between the interior and the exterior solution for all of our experiments. The solution in the interior of the GUVs consisted of sucrose only, whereas the outer solution consisted of sucrose (110 mM), glucose (110 mM), and HEPES (20 mM) + NaCl (45 mM) buffer in a 1:1:1 ratio. The osmolarities of the interior and the exterior solutions were adjusted such that there was no osmotic difference between the two compartments. Based on the intercept value obtained for POPC-only vesicles, it is evident that this minor solution asymmetry did not significantly affect the spontaneous curvature of these vesicles as shown in Fig. 5 b.

Figure 5.

Tether-pulling experiments confirm the generation of negative spontaneous curvature in the presence of PIP2. (a) shows plots of force versus √Σ_asp for vesicles with varying PIP2 concentrations for outward membrane tethers. The vesicle membrane tension Σ_asp is set by the suction pressure of the micropipette, and corresponding pulling force is measured. The data points are shown for POPC-only (six-pointed star), 2% PIP2- (▿), and 5% PIP2- (⋄) containing vesicles. All three vesicle compositions included 0.2% TR-DHPE to allow visualization of the membrane. The lines connecting the data points are fits from Eq. 3 and are extended to show the y-intercept. The y-intercept value is close to zero for POPC-only vesicles, meaning that there is negligible asymmetry, whereas 2 and 5% PIP2 have a positive intercept value, confirming the presence of asymmetry in these vesicles. The intercept values were converted to spontaneous curvature using Eq. 3 and are shown as bar plots in (b). The POPC-only vesicles resulted in a negligible spontaneous curvature. Incorporation of 2 and 5% PIP2 in membranes upon 100-fold dilution resulted in a spontaneous curvature of −20.7 ± 2.4 and −36.2 ± 2.2 μm⁻¹, respectively. The intercept values reported here are the average of the intercept values obtained between 30 and 90 min after dilution for vesicles that were 4–12 h old. For each membrane composition, the tether-pulling experiments were performed for at least five vesicles, and the data are represented as mean ± SEM. ^∗p < 0.05.

Next, we performed the tether-pulling experiments on vesicles containing 2 or 5% PIP2 (in addition to POPC and 0.2% TR-DHPE) under the same solution conditions as was done with POPC-only vesicles. The force values at different membrane tensions for 2 and 5% PIP2 are shown as inverted triangles and diamonds, respectively, in Fig. 5 a. A linear fit of the data points for these vesicles resulted in a positive intercept value (Fig. 5 a), confirming that PIP2-containing vesicles result in negative spontaneous curvature, as shown in Fig. 5 b. The bending rigidity of the vesicles with 2 and 5% PIP2 was similar to that for POPC-only vesicles. These values are within the error range of bending rigidity values obtained using tether-pulling experiments (65).

We converted the intercept values obtained for the three compositions into corresponding spontaneous curvature values using Eq. 3 (Fig. 5 b). The inclusion of PIP2 in the vesicles resulted in negative spontaneous curvature, which is in line with the observation of internal tubules in PIP2-containing vesicles (Fig. 3). The incorporation of 2% PIP2 in the vesicles resulted in a spontaneous curvature of −20.7 ± 2.4 μm⁻¹. On increasing the PIP2 content to 5% in the membranes, even larger spontaneous curvature of −36.2 ± 2.2 μm⁻¹ was obtained as compared to 2% PIP2-containing vesicles. The increase in the negative value of spontaneous curvature upon increasing PIP2 content indicates that pulling an outward tether from a lipid bilayer at a given membrane tension becomes more difficult when the PIP2 content increases. In other words, the higher the PIP2 content in the vesicles, the higher its asymmetric desorption from the vesicles upon dilution. This asymmetric desorption causes number asymmetry, which is compensated for by inner tubulation. The tendency of the membrane to tubulate internally is consistent with the observation of a higher force required to pull an outward tether compared to a symmetric bilayer. The formation of internal tubules in vesicles containing PIP2 appears to be visible in images of vesicles shown in earlier reports, as well; however, this had not been characterized (13).

The values of the spontaneous curvature due to the asymmetric desorption of PIP2 can be compared to similar observations recently reported for the lipid GM1 (26). These two lipids can be compared considering micelle sizes and CMC values. The reported average value for the radius of PIP2 micelles is ∼30 Å (16, 17). This value is roughly half of that reported for the radius of GM1 micelles (66). A smaller value of the micelle radius for PIP2 suggests that the molecular spontaneous curvature of PIP2 lipids is larger than that of GM1 lipids, and thus, an asymmetric distribution of PIP2 will result in a spontaneous curvature larger than GM1. This would imply a larger value of C₀ for PIP2 in Eq. 3 compared to GM1 if the composition difference between leaflets of these two lipids were equal.

Furthermore, PIP2 lipids have CMC values (13, 48, 67, 68) that are at least two orders of magnitude higher than those of GM1 lipids (66, 69). The CMC of PIP2 lipids can further be increased if its lipid tail gets oxidized, rendering its hydrophobic tail more hydrophilic. This suggests that PIP2 lipids remain in monomeric form at concentrations roughly two orders of magnitude higher than those for GM1 lipids, leading to a higher tendency for PIP2 lipids to leach out of the membrane and thus resulting in higher asymmetries in lipid number densities. This would then lead to a larger magnitude of the area difference term in Eq. 3. These two effects combined have a reinforcing effect on the magnitude of the effective spontaneous curvature because of asymmetric desorption of PIP2 lipids. Therefore, the effective spontaneous curvature values obtained for PIP2 lipids are expected to be larger than those reported for GM1 lipids, which is consistent with our observation of larger intercepts for PIP2 membranes compared to GM1 (26).

Because plasma membranes contain significant amounts of cholesterol, its potential effects on PIP2 desorption warrant discussion. In an earlier study, through ζ potential measurements, Carvalho et al. (20) showed that inclusion of cholesterol in PIP2-containing membranes decreased the leaching of PIP2. They hypothesized that this might be a result of the membrane-stabilizing role of cholesterol (70, 71, 72). However, in a more recent contribution, Beber et al. (19) incorporated 15% cholesterol in the membranes and still observed a decrease in the fluorescence intensity of septin protein bound to the membrane. Therefore, it seems that cholesterol does not ubiquitously prevent PIP2 desorption.

Asymmetry due to desorption of PIP2 is suppressed by external PIP2 micelle addition

Next, to further support our leaching hypothesis, we asked whether spontaneous curvature generation due to asymmetric desorption of PIP2 can be prevented as an alternative to using saturated chain PIP2 lipids. To suppress the leaching of PIP2 lipids from the membrane, we increased the PIP2 content in the solution. To do this, we introduced a 100-fold higher PIP2 concentration in the micellar form (8.8 μM) compared to the PIP2 content in 100-fold diluted GUV dispersion solution. The GUV-micelle solution was allowed to sit for 15 min before imaging. The presence of excess of PIP2 in the outer solution suppressed the leaching of PIP2 lipids from the outer leaflet, resulting in a significant decrease of internally tubulated vesicles, as shown in Fig. 6 a. In fact, the majority of vesicles developed external tubules because of the insertion of PIP2 lipids selectively in the outer leaflet (Fig. 6 b; Fig. S6). The generation of external tubules within 15 min of incubation with PIP2 micelles agrees with the timescale for formation of inward tubules in our dilution experiments. We suggest that in future experiments testing the lipid-protein interactions involving PIP2, one might externally add an optimized amount of PIP2 to achieve a reference state of GUVs with symmetric PIP2 distribution.

Figure 6.

External PIP2 addition suppresses internal tubulation. The vesicles composed of POPC (94.8%), PIP2 (5%), and TR-DHPE (0.2%) were rehydrated in 110 mOsm sucrose and diluted 100-fold in a 1:1:1 solution of sucrose (110 mM)/glucose (110 mM)/20 mM HEPES + 45 mM NaCl. (a) shows the proportion of externally tubulated, nontubulated, and internally tubulated vesicles in the presence and absence of externally added micelles. In the absence of micelles, the majority of vesicles showed internal tubules. On addition to a separate GUV sample of 100-fold higher concentration of PIP2 in the form of micelles, the number of internally tubulated vesicles decreased substantially within 15 min. In addition to the decrease in number of internally tubulated vesicles, several vesicles showed outer tubulation, suggesting that the PIP2 lipids were incorporated into the outer leaflet of the GUVs. The data are represented as mean ± SEM. (b) shows one of the externally tubulated vesicles as a result of micelle addition in the solution. Scale bars, 5 μm.

Conclusions

In this report, we showed that spontaneous desorption of PIP2 can result in remodeling of membranes. The spontaneous curvature values obtained because of asymmetric desorption of PIP2 lipids in the membrane can result in the generation of tubules 60–100 nm in diameter for 2 and 5% PIP2, respectively. The values of spontaneous curvature resulting from desorption of PIP2 lipids are larger than those obtained for the asymmetric trans-membrane distribution of small molecules such as sugars and ions (45, 47, 73, 74). However, the values are comparable to the magnitude of spontaneous curvature generated by typical curvature-generating proteins, such as BAR proteins (75, 76, 77, 78).

The findings in this report serve to suggest that conclusions that have been derived about PIP2-protein interactions in biomimetic membranes, especially relating to curvature-generating and sensing proteins, will be influenced by the leaching of PIP2 from the membranes, and therefore, results will depend largely on the age and the dilution factor of the vesicles at the time of the experiment. Most of these experiments rely on the generation of highly curved structures such as tubules to elucidate the effect of curvature-generating proteins in the presence of PIP2 lipids. Because the asymmetric desorption of PIP2 in itself will result in the generation of these curved structures, knowing the extent of this spontaneous curvature will allow precise determination of the curvature-generating and sensing properties of tested proteins. This aspect needs to be accounted for when interpreting such results.

We have also shown that the desorption of PIP2 from the membranes could be accelerated by the oxidation of its arachidonic lipid tail. Lipid oxidation products have been implicated in the development of several chronic diseases, including atherosclerosis, cancer, type 2 diabetes, and Alzheimer’s disease (79, 80, 81, 82). In addition, specific oxidized phospholipids have been shown to trigger immune responses, inflammation, and apoptosis (60). Oxidation products of phospholipids thus can play an important role in controlling biological processes, including the remodeling of membranes.

PIP2 lipids are present in small amounts in the mammalian plasma membrane. However, their local concentration in the biological membranes is variable (83, 84). Our findings suggest that generation of PIP2 from PIP and the resulting increase in spontaneous curvature, as well as leaching of PIP2, can give rise to differential stresses between the two leaflets of the plasma membrane. This differential stress can lead to changes in the membrane morphology that can influence signaling events at the plasma membrane. For instance, the processes of clathrin-mediated endocytosis (85, 86), and fast endophilin-mediated endocytosis (87) involve temporal generation of PIP2 lipids, which then control the sequential recruitment of various proteins at different stages of these processes that are accompanied by the generation of highly curved structures. Therefore, remodeling of the membrane due to spontaneous bending caused by PIP2 can influence the recruitment of endocytic proteins during the endocytic vesicle formation process. In addition, differential stresses generated by PIP2 might also play a role in modulating the function of mechanosensitive ion channels. Indeed, PIP2 lipids could play an essential role in controlling the function of several ion channels (88, 89, 90). Differential tension in the two leaflets of the membrane due to changes in local PIP2 density might act as a switch to modulate the conformational states of these ion channels in the plasma membrane. Further studies of the complex and multifaceted role of PIP2 will be enhanced with consideration of its desorption behavior.

Author Contributions

S.S., R.J., and T.B. designed research. S.S., R.J., J.R., and Z.E.Z. performed research. S.S., R.J., and J.R. contributed analytic tools. S.S. and J.R. analyzed data. S.S. and T.B. wrote the manuscript.

Editor: Ilya Levental.