Loss of lipid asymmetry facilitates plasma membrane blebbing by decreasing membrane lipid packing

Significance

The lipid asymmetry of the plasma membrane and its regulated release through lipid scrambling are critical determinants of membrane properties and functions. Such scrambling occurs during many biological events, including apoptosis, extracellular vesicle (EV) formation, sperm–egg fertilization, bone mineralization, viral infection, and blood coagulation. However, how lipid scrambling changes membrane properties that facilitate these events is currently unknown. This study reveals that loss of lipid asymmetry lowers membrane lipid packing, thereby facilitating the membrane deformation during blebbing. Our findings suggest that reducing lipid packing by lipid scrambling is an important contributor to the formation of EVs and potentially other cellular processes involving membrane deformation.

Abstract

Membrane blebs have important roles in cell migration, apoptosis, and intercellular communication through extracellular vesicles (EVs). While plasma membranes (PM) typically maintain phosphatidylserine (PS) on their cytoplasmic leaflet, most blebs have PS exposed on their outer leaflet, revealing that loss of steady-state lipid asymmetry often accompanies PM blebbing. How these changes in PM lipid organization regulate membrane properties and affect bleb formation remains unknown. We confirmed that lipid scrambling through the scramblase TMEM16F is essential for chemically induced membrane blebbing across cell types, with the kinetics of PS exposure being tightly coupled to the kinetics of bleb formation. Measurement of lipid packing with environment-sensitive probes revealed that lipid scrambling changes the physical properties of the PM, reducing lipid packing and facilitating the bilayer bending required for bleb formation. Accordingly, reducing lipid packing of the PM through cholesterol extraction, elevated temperature, or treatment with biological amphiphiles promoted blebbing in the absence of TMEM16F. Consistent with these cellular observations, blebbing in Caenorhabditis elegans embryos measured via EV production was significantly reduced by depleting the TMEM16-homolog ANOH-2. Our findings suggest that changing membrane biophysical properties by lipid scrambling is an important contributor to the formation of blebs and EVs and potentially other cellular processes involving PM deformation.

The asymmetric distribution of lipid species between the two leaflets of the bilayer is a fundamental feature of the mammalian plasma membrane (PM) (1–3), which also extends to many other eukaryotes (4, 5). Energy-dependent flippase enzymes translocate essentially all amino-headgroup lipids [phosphatidylethanolamine (PE) and phosphatidylserine (PS)] to the PM cytoplasmic (inner) leaflet, while floppases enrich phosphatidylcholine (PC) and sphingomyelin in the exoplasmic (outer) leaflet (4, 6, 7). Notably, lipid asymmetry is not only evident in the interleaflet distributions of lipid headgroups but also in their hydrophobic acyl chains, with inner leaflets rich in polyunsaturated lipids and outer leaflets rich in saturated lipids (4). These structural features appear to persist throughout eukaryotic PMs (4), implying important functional roles for membrane asymmetry that remain only partially understood (2).

One important aspect of lipid asymmetry is its regulated release, usually achieved by activation of lipid channels called scramblases, which include TMEM16F (8), Xkr8 (9), and others (1, 10, 11). These energy-independent lipid channels facilitate the translocation (scrambling) of lipids between leaflets down their chemical potential gradients, which is typically detected by the exposure of PS and/or PE on the outer leaflet (8). One well-known role of PS exposure is as an “eat me” signal on the surface of dying cells and apoptotic blebs to facilitate their receptor-mediated phagocytosis. PE exposure also occurs on healthy cells during cytokinesis, and PE asymmetry must be reestablished for animal cells to separate through abscission (12–14). Release of lipid asymmetry is also associated with many other biological events, including fertilization, muscle cell fusion, bone mineralization, viral infection, immune cell activation, and blood coagulation (15). In most cases, it is not yet clear how these processes are affected by lipid scrambling; however, one possibility is that lipid asymmetry affects biophysical PM properties (e.g., diffusivity, permeability, stiffness) that regulate these cellular functions (16).

The loss of lipid asymmetry, usually characterized by PS exposure, is particularly characteristic of various types of extracellular vesicles (EVs), including submicron exosomes, PM-derived ectosomes, and large blebs such as apoptotic bodies (17). Similarly, in Caenorhabditis elegans, loss of PE flippase activity results in PE exposure and increased production of EVs from the PM (18, 19). The physiological functions of EVs include intercellular communications (20), apoptotic signaling (21), cancer cell survival (22), and ameboid cell migration (23, 24); however, whether and how the loss of lipid asymmetry contributes to their formation is unknown.

An experimentally tractable version of membrane blebs are Giant PM Vesicles (GPMVs), which have been widely used for studying PM properties (25) like lipid phase separation (26, 27) and protein affinity for lipid raft domains (28, 29). GPMVs are produced by treating cells with sulfhydryl-reactive agents [e.g., formaldehyde, maleimide, iodoacetamide (25, 30, 31)] in Ca²⁺-containing buffers, which results in PS exposure (32). Interestingly, formation of GPMVs from HEK293 cells requires the PS scrambling activity of the scramblase TMEM16F (32), suggesting that release of lipid asymmetry is important for the formation of large blebs. Why loss of lipid asymmetry through TMEM16F is required for giant bleb formation is unknown.

Here, we show that lipid scrambling is required for membrane blebbing across cell lines and that this effect is independent of overall PS levels or the presence of lipids that link the PM to the actin cytoskeleton. In vivo, we find that lipid scrambling affects ectosome production in C. elegans embryos. We show that loss of lipid asymmetry reduces lipid packing, potentially softening the membrane and thus facilitating the membrane bending required for bleb formation. Consistent with this hypothesis, blebbing can be enhanced by manipulating lipid packing in the PM via cholesterol extraction, increased temperature, or biological amphiphiles. Our findings suggest that membrane softening through lipid scrambling is an important aspect of membrane blebbing, potentially relevant in many biological events that involve membrane deformation (15).

Results

Chemically Induced Blebbing Requires Lipid Scrambling.

To study the role of lipid scrambling in membrane blebbing, we evaluated how PS exposure and bleb formation depend on the lipid scramblase TMEM16F. To that end, we induced GPMVs in mouse BaF3 cells (a suspension pro-B cell line) with a widely used blebbing buffer (25) (2 mM Ca²⁺, 2 mM dithiothreitol, 25 mM paraformaldehyde) and measured PS exposure by the binding of fluorescent AnnexinV (AxV). We observed rapid PS exposure (within 5 min of blebbing buffer addition), reaching a plateau within 10 min (Fig. 1 A, Top). In contrast, BaF3 cells in which TMEM16F was genetically deleted (TMEM16F-KO) did not expose PS within the first 5 min and showed ~10-fold lower PS exposure than WT cells 10 min after blebbing buffer treatment (Fig. 1 A, Middle). Even after 1 h, AxV staining was notably lower in TMEM16F-KO cells than in WT cells (Fig. 1 A, Right). The residual PS exposure in TMEM16F-KO cells is likely mediated by other scramblases (9, 11, 33). Reintroduction of TMEM16F to KO cells rescued PS scrambling (Fig. 1 A, Bottom), indicating that TMEM16F is the predominant lipid scramblase under these conditions. As intracellular Ca²⁺ levels are known to increase as a result of these treatments (34), our observations suggest that exposure of PS during giant bleb formation (32, 34, 35) is mediated by Ca²⁺-induced TMEM16F activation, in good agreement with previous reports (32). Consistently, removal of extracellular Ca²⁺ prevented blebbing (SI Appendix, Fig. S1).

Fig. 1.

Bleb formation in BaF3 cells requires TMEM16F-mediated lipid scrambling. (A) Flow cytometry shows reduced PS exposure (revealed by AxV-Alexa568 staining) in TMEM16F-KO BaF3 cells compared to WT after blebbing buffer treatment at 37 °C. Reexpression of TMEM16F-GFP in TMEM16-KOs rescued PS exposure. (B) Epifluorescence microscopy shows that TMEM16F-KO cells are deficient in PS exposure and do not produce blebs. Reexpression of TMEM16F-GFP rescues PS exposure and bleb formation. (Right) Quantification of fraction of blebbing cells. Symbols represent independent experiments; ***P < 0.001 using an unpaired t test; ns = no significant difference from WT. (C) Mixture of TMEM16F-KO cells (arrow) and GFP-tagged TMEM16F-repleted cells (arrowhead) shows that scramblase repletion induces extensive PS exposure and produces blebs. (D) Quantification of AxV intensity (PS exposure) of TMEM16F-KO cells and TMEM16F-GFP cells. Means ± SD of individual cells from one (of three) representative experiment are shown. ***P < 0.001 using an unpaired t test. Scale bars are 5 μm.

These effects of TMEM16F on lipid scrambling were correlated with bleb formation. WT BaF3 cells showed extensive membrane blebbing (Fig. 1B) after 1 h treatment with blebbing buffer [as do most other mammalian cells (31)]. In contrast, TMEM16F-KO cells did not produce giant blebs (Fig. 1B). Again, reintroduction of TMEM16F-GFP rescued blebbing (Fig. 1 B, Right). To further confirm that TMEM16F was required for bleb formation, we mixed nonfluorescent TMEM16F-KO cells with fluorescent cells expressing TMEM16F-GFP and treated with blebbing buffer. Only TMEM16F-GFP⁺ cells showed extensive AxV binding (i.e., PS exposure) (Fig. 1 C and D) and formed large blebs (Fig. 1C, arrowheads), whereas TMEM16F-KO cells in the same preparation bound minimal AxV (Fig. 1 C and D) and did not bleb (Fig. 1C, arrows). To ensure that these effects were not exclusive to giant blebs induced by chemical treatment, we induced blebbing by directly increasing cytosolic Ca²⁺ via the calcium ionophore A23187. While A23187-treated BaF3 cells avidly produced blebs, TMEM16F-KOs did not (SI Appendix, Fig. S2). Collectively, these results reveal that giant bleb formation in BaF3 cells requires Ca²⁺-activated TMEM16F-mediated lipid scrambling.

Lipid Scrambling, But Not PS Abundance, Affects Bleb formation.

We further validated that lipid scrambling is associated with bleb formation in other cell lines with varying scrambling and blebbing kinetics. To that end, we used live cell imaging to quantify AxV binding and the onset of bleb formation in two adherent cell lines: Rat Basophilic Leukemia cells [RBL, a mast cell model widely used for blebbing studies (35–37)] and HEK293 cells. RBL cells scrambled PS rapidly after treatment with blebbing buffer, reaching maximal staining within 30 min (Fig. 2A). The first blebs were observed within 20 min in this cell type (Fig. 2A, red dots). In contrast, HEK293 cells scrambled PS slowly, requiring >3 h to reach maximal AxV staining. These HEK293 cells also required much longer time for initial bleb formation, typically ~2 h (Fig. 2A). The slow kinetics of lipid scrambling of HEK293 cells is likely related to their low expression of TMEM16F (32). These results demonstrate that the kinetics of bleb formation correlate with the kinetics of lipid scrambling. Curiously, in both RBL and HEK293 cells, bleb formation initiated when PS scrambling reached approximately half of maximal AxV staining (Fig. 2A, pink band and black dotted line). These findings imply that progressive loss of lipid asymmetry leads to progressive changes in membrane properties, which allow membrane blebbing only beyond a certain threshold.

Fig. 2.

Lipid scrambling, not PS abundance, determines bleb formation. (A) Kinetics of PS scrambling and bleb formation of RBL and HEK293 cells. Red dots indicate the time point of initial bleb formation in individual cells. The mean normalized AxV intensity when blebs were first observed across all cells measured is shown as the gray dotted line (pink band = SD). (B) PS-deficient PSA-3 cells do not constitutively bleb, but expose PS and produce blebs similar to PS-rescued PSA-3 cells supplemented with ethanolamine for PS synthesis. (C) Quantification of AxV intensity (PS exposure) and bleb formation reveals no significant difference (ns, unpaired t test) between PS-deficient and PS-rescued cells. Mean ± SD of individual cells from one (of three) representative experiment are shown. (D) PE externalization induced by blebbing buffer detected by duramycin-PEG-GFP. In contrast to evenly distributed AxV staining, duramycin stains only the cell body (arrow) of the scrambled cells, but not the blebs (arrowhead). Scale bars are 5 μm (Top) and 10 μm (Bottom).

As lipid scrambling results in both increased PS exposure in the outer leaflet and a decreased level of PS in the inner leaflet, we next tested whether bleb formation may be induced by a reduced abundance of PS in the inner leaflet. To that end, we used PSA-3 cells, which are CHO cells deficient in PS synthetase 1 (PSS1), resulting in ~30% lower PS levels in the PM (38, 39). Importantly, these cells do not bleb constitutively (Fig. 2 B, Left), which suggests that lower inner leaflet PS levels are not sufficient to induce blebbing. PS levels in these cells can be rescued by feeding with ethanolamine, which bypasses PSS1 deficiency by enhancing PSS2-catalyzed PS synthesis from PE (38, 39). While PS-deficient PSA-3 cells were morphologically different from PS-rescued (ethanolamine fed) PSA-3 cells (elongated versus polygonal, Fig. 2B), both exposed PS and produced blebs at similar levels after treatment with blebbing buffer (Fig. 2 B and C). Thus, manipulating overall PS abundance did not affect blebbing, suggesting that lipid scrambling, not inner leaflet PS levels, affects blebbing.

Consistent with a general loss of lipid asymmetry during chemically induced blebbing, we also detected PE exposure on the outer leaflet of the PM using the PE-binding probe duramycin-polyethylene glycol (PEG)-green fluorescence protein (GFP), an effective and well-tolerated PE reporter (40). Treatment of both BaF3 cells and RBL cells with blebbing buffer led to duramycin binding (i.e., PE exposure) (Fig. 2D), consistent with previous reports that TMEM16F is a nonselective phospholipid channel (41). Unexpectedly, while the PE binding probe brightly labeled cell bodies after treatment with blebbing buffer, it was largely absent from the surface of membrane blebs, in striking contrast to AxV (PS-binding probe), which was distributed uniformly throughout the cell and bleb surface (Fig. 2D). This effect was not due to cross-reaction between the probes, as it was also observable without AxV. This staining pattern either reflects that PE lipids are segregated from PS lipids on the surface of membrane blebs or that duramycin fails to bind PE in membranes under conditions present in blebs; these effects require future study.

Lipid Scrambling Decreases Lipid Packing.

As lipid scrambling appears to be critical for formation of giant blebs, we examined which membrane properties that regulate membrane blebbing may be affected by scrambling. It was previously reported that depletion of PI(4, 5)P₂ (PIP₂) can lead to detachment of PM from the cytoskeleton, facilitating bleb formation (34). We hypothesized that PIP₂ may be depleted from the PM inner leaflet in WT (e.g., by direct scrambling or scrambling-associated hydrolysis) but not TMEM16F-KO cells, potentially explaining their differences in blebbing. To test this possibility, we imaged an exogenously expressed PIP₂ sensor (the PH domain of PLCδ) during blebbing buffer treatment of BaF3 cells. We confirmed previous reports that PIP₂ appears to be lost from the cytoplasmic PM leaflet during induction of blebbing, evidenced by relocalization of PH-PLCδ from the PM to the cytosol (Fig. 3 A, Top). However, we also observed similar movement of PH-PLCδ from the PM to the cytosol in TMEM16F-KO cells (Fig. 3 A, Bottom), indicating that TMEM16F and its lipid scrambling are not required for PIP₂ depletion. These findings imply that PIP₂ is not lost from the inner leaflet of the PM by TMEM16F-mediated scrambling; a likely alternative is Ca²⁺-induced activation of cytosolic PLC (42). Importantly, despite depletion of PIP₂, TMEM16F-KO cells did not produce blebs (Figs. 1 and 3A). Thus, while PIP₂ depletion and the resulting release of the PM from the cytoskeleton may be important for bleb formation, this effect does not explain why TMEM16F-mediated lipid scrambling is required for blebbing.

Fig. 3.

Lipid scrambling reduces lipid packing and facilitates blebbing. (A) The presence of PIP₂ lipids in the PM inner leaflet was probed by expression of PH-PLCδ-GFP. Treatment with blebbing buffer relocated PH-PLCδ from the PM to the cytosol in both BaF3 WT and TMEM16F-KO cells, suggesting depletion of PM PIP₂ in both. The scale bar is 5 μm. (B) Calibration of measured Laurdan Generalized Polarization (GP) in synthetic lipid vesicles to membrane bending rigidity calculated from molecular dynamic simulations of the same compositions. Laurdan GP and membrane stiffness are correlated across the compositions investigated: Green symbols represent membranes of POPC+cholesterol (from bottom left to top right: 0, 10%, 40%, and 50%); orange symbols represent DOPC+cholesterol (from bottom left to top right: 0, 10%, 40%, and 50%). Error bars are mean ± SD of >5 giant vesicles per condition. (C) Laurdan GP images of BaF3-WT and TMEM16F-KO cells before and after blebbing buffer treatment. Redder colors represent higher GP (i.e., lipid packing). Insets show magnifications of PMs. The scale bar is 5 μm. (D) Laurdan GP of BaF3-WT and TMEM16F-KO cells before and after blebbing buffer treatment. Blebbing buffer treatment reduces Laurdan GP in BaF3-WT cells, and this effect is attenuated in TMEM16F-KO. Means ± SD of individual cells from one (of three) representative experiment are shown. ***P < 0.001 from the unpaired t test between indicated groups. (E) GP changes due to blebbing buffer treatment (ΔGP) of BaF3-WT and TMEM16F-KO cells across three independent repeats. *P < 0.05 from the paired t test between groups.

Blebbing requires major membrane deformation via outward bending, which is energetically costly due to the stiffness of lipid bilayers. Membrane bending rigidity (i.e., stiffness) has been previously implicated in the bilayer rearrangements required for membrane traffic (43), suggesting that stiffness may be an important regulator of blebbing. A striking recent report showed that synthetic asymmetric membranes are up to twofold stiffer than symmetric vesicles, even when their overall lipid composition is identical (44). This anomalously high stiffness of synthetic lipid bilayers was recently confirmed by neutron spin echo measurements (45). Theoretical and computational studies rationalized this finding by invoking a stress differential between the two asymmetric leaflets resulting from their suboptimal lipid packing (46, 47). Thus, since asymmetric bilayers are stiffer than symmetric, we hypothesized that releasing lipid asymmetry by scrambling may reduce membrane stiffness, thus facilitating membrane deformation and blebbing.

The bending rigidity of the lipid bilayer is difficult to measure directly in living cells due to the tight biochemical and mechanical coupling between the PM and other cellular elements, most notably the cytoskeleton (48). Since membrane bending rigidity is highly correlated with lipid packing (i.e., area per lipid, APL) (49, 50), we estimated stiffness changes due to scrambling by measuring changes in lipid packing via a solvatochromic fluorescent sensor called Laurdan. This sensor reports changes in local lipid packing through a shift in Laurdan’s fluorescence emission spectrum (51). Specifically, Laurdan’s maximum emission in tightly packed membranes is ~440 nm, whereas in loosely packed it is ~490 nm; the normalized ratio between these two maxima is known as generalized polarization (GP), with higher values reflecting more tightly packed membranes. To evaluate the validity of using Laurdan to estimate changes in membrane stiffness, we compared membrane stiffness to Laurdan GP across a range of defined lipid compositions. Laurdan GP was experimentally measured by spectral confocal imaging of giant unilamellar vesicles (GUVs) while bending rigidity was computationally calculated from molecular dynamics (MD) simulations of the same lipid compositions (50). There was a strong correlation between Laurdan GP and membrane stiffness across the dataset spanning ~fivefold difference in stiffness (Fig. 3B), suggesting that Laurdan GP can be used to infer changes in membrane stiffness.

We next applied Laurdan to BaF3 cells to measure how lipid scrambling changes membrane properties. The GP of the PM decreased significantly after treatment with blebbing buffer (Fig. 3 C–E), suggesting a significant decrease in lipid packing and, by extension, membrane stiffness. This reduction of PM GP was significantly smaller in TMEM16F-KO cells (Fig. 3 C–E), consistent with their reduced lipid scrambling (Fig. 1). Since various membrane probes report on aspects of the membrane with varying sensitivity and specificity (52), we also measured the effects of scrambling on the fluorescent lifetime of Di4-ANNEPDHQ (Di4) (4), another probe sensitive to lipid packing and whose readout also correlates with membrane stiffness (SI Appendix, Fig. S3). As for Laurdan GP, we observed significantly decreased Di4 lifetime in BaF3 cells treated with blebbing buffer, but a much smaller effect in TMEM16F-KO cells (SI Appendix, Fig. S4). These results indicate that lipid scrambling by TMEM16F reduces lipid packing, which implies that membrane stiffness may also be reduced by scrambling, thus facilitating membrane bending and bleb formation. We hypothesize that in the absence of rapid, large-scale lipid scrambling in TMEM16F-KO cells, membrane packing and stiffness remain high, which impedes membrane bending and thus blebbing.

Reducing Membrane Stiffness Promotes Bleb Formation.

To support the notion that changes in membrane stiffness affect blebbing, we relied on the composition and temperature dependence of membrane bending rigidity. Namely, cholesterol stiffens membranes made of biologically relevant lipids (53) while higher temperature softens them (54). We extracted cholesterol using 5 mM MβCD (55) and observed the expected decrease in lipid packing (reported by Di4 lifetime) of HEK293 cell PMs (Fig. 4A). Blebbing buffer treatment induced bleb formation within 1 h in MβCD-treated HEK293 cells (Fig. 4B), in striking contrast to untreated HEK293 cells (Figs. 2 A and 4B). Increasing incubation temperature (from 37 °C to 45 °C) also induced bleb formation (SI Appendix, Fig. S5). Similarly, decreasing membrane stiffness in TMEM16F-KO BaF3 cells by either MβCD, increased incubation temperature, or a biological amphiphile [sodium deoxycholate, DCA, which reduces lipid packing (56)] also rescued bleb formation (Fig. 4C), suggesting that decreasing membrane stiffness promotes membrane blebbing. It is important to note that the perturbations (cholesterol extraction and increased temperature) that promoted membrane blebbing in TMEM16F-KO cells eventually led to lipid scrambling (evidenced by AxV binding after 1 h treatment, SI Appendix, Fig. S6), the mechanism of which is yet unclear.

Fig. 4.

Decreasing membrane packing facilitates blebbing. (A) Treatment of HEK293 cells with MβCD reduced lifetime of Di4 in the PM, indicating decreased PM lipid packing and thus membrane softening. Means ± SD of three independent replicates, ***P < 0.001 from the unpaired t test between groups. (B) Treatment with MβCD (5 mM for 1 h) induces bleb formation in HEK293 cells. The scale bar is 20 μm. (C) Softening the PM of BaF3 TMEM16F-KO cells by extracting cholesterol (MβCD), lowing lipid packing (DCA), or increasing temperature (from 37 °C to 45 °C) leads to bleb formation. The scale bar is 10 μm.

Lipid Scrambling Promotes Membrane Blebbing In Vivo.

We next tested whether lipid scrambling influences PM blebbing in vivo by measuring the production of EV in live C. elegans embryos (18, 19). To visualize EV production, we used a degron-tagged PIP₂ reporter (57) to label EVs on the surface of embryos: In cells, the reporter is degraded, while in EVs that detached from cells, it is protected from degradation and becomes visible as bright puncta (Fig. 5A). Since relatively few EVs are released in wild-type embryos, this assay relies on a partial loss-of-function mutation caused by tagging the trafficking protein PAD-1, which results in a 20-fold increase in EV release through PM blebbing (58). PAD-1 acts in concert with the major PE flippase TAT-5 to maintain PE asymmetry in the PM (18, 19), and disruption of either protein enhances EV production (18, 19), consistent with our observations that loss of lipid asymmetry facilitates blebbing. ANOH-1 and ANOH-2 are the two TMEM16 family proteins in C. elegans (59, 60), with ANOH-1 mediating PS exposure in adult neurons (61). Only ANOH-2 is expressed in embryos, thus we examined its role in embryonic EV formation. Knocking down the putative scramblase ANOH-2 significantly reduced EV production from the PM in GFP::PAD-1 mutant embryos (Fig. 5 A and B), in line with our observations that its homolog scramblase TMEM16F is important for PM blebbing in cultured mammalian cells (Fig. 1).

Fig. 5.

EV blebbing from the C. elegans PM depends on the presence of TMEM16 homolog ANOH-2 and temperature. (A) Representative fluorescent images of EVs above the surface of GFP::PAD-1 mutant C. elegans embryos visualized with a degron-tagged PIP2 reporter (mCh::PH::CTPD). Fewer EVs are observable in anoh-2 knockdowns (KD) compared to empty vector control. The scale bar is 10 μm. (B) Significant reduction in EV puncta in anoh-2 KD embryos. Small symbols represent EV counts from individual embryos (colors represent samples from different independent experiments); large symbols represent means of the experiments. *P < 0.05 from the unpaired t test of the mean values from the three repeats. (C) EV production in GFP::PAD-1 mutant embryos is temperature dependent. Means ± SD of individual embryos from three experiments are shown., ***P < 0.001 from the unpaired t test.

To connect this effect on embryonic EV release to membrane packing and stiffness, we tested whether temperature impacts PM blebbing (i.e., EV formation) in vivo. Decreasing the temperature of GFP::PAD-1 worms by ~3 °C from the regular culture temperature (22 to 23 °C) significantly decreased EV production (Fig. 5C). Correspondingly, increasing temperature by ~3 °C significantly increased EV formation (Fig. 5C). Although the changes in EV production could also be explained by other effects of temperature, lipid scrambling- and temperature-dependent EV production are broadly in line with our cell culture experiments (2, 3), supporting the notion that lipid scrambling reduces membrane stiffness which in turn facilitates membrane deformation for bleb formation in vivo.

Discussion

Lipid asymmetry between the two leaflets of the PM and its regulated scrambling are important determinants of PM properties and functions (1–4). Here, we studied the role of membrane asymmetry in membrane blebbing using both chemically induced, micrometer-sized PM blebs in cultured cells and submicron EVs in C. elegans embryos. We confirmed that lipid scrambling through scramblase TMEM16F is necessary for formation of chemically induced giant blebs and that the kinetics of blebbing correlate strongly with the kinetics of lipid scrambling (Figs. 1 and 2). Since membrane deformation is an important aspect of many biological processes, we hypothesized that lipid scrambling may reduce membrane stiffness and thus facilitate blebbing. Consistently, we observed that scrambling reduces lipid packing density of the PM, which correlates with decreased PM stiffness, potentially explaining why scrambling facilitates blebbing (Fig. 3). Further, softening the PM through cholesterol extraction (MβCD), decreasing lipid packing with a biological detergent (DCA), or increasing temperature (from 37 °C to 45 °C) all promoted blebbing in the absence of TMEM16F (Fig. 4). Finally, in accordance with cell culture observations, in vivo production of EVs in C. elegans embryos was significantly reduced by knocking down the TMEM16 family homolog ANOH-2, while elevated temperature significantly increased EV release from the PM (Fig. 5).

Our findings suggest that lipid asymmetry is partially responsible for maintaining sufficiently high membrane stiffness to inhibit membrane blebbing at steady state. This mechanism may be important for resisting deformation caused by intracellular osmotic pressure in living cells (62, 63). Conversely, physiological processes where PM deformation is desired (e.g., blebbing, EV production) may rely on activation of scramblases to release asymmetry, decreasing membrane stiffness and thereby facilitating budding of the membrane. Since lipid scrambling is observed during many biological events that involve membrane deformation (15), membrane softening by scrambling may be relevant in many such cellular events.

Loss of lipid asymmetry often involves both scramblase activation and flippase inactivation (1, 2). We demonstrated that the scramblase TMEM16F is responsible for lipid scrambling after BaF3 exposure to blebbing buffer, but it is unknown whether flippase proteins are also inhibited by blebbing buffer. Caspase-mediated cleavage of PS flippases ATP11A and ATP11C promotes PS exposure during apoptosis (64, 65), but it is yet unknown whether these flippases influence apoptotic bleb formation. In C. elegans embryos, disrupting the major PS flippase TAT-1 causes PS exposure (66), but does not increase EV blebbing from the PM (19). However, disrupting either PM localization or lipid flipping activity of the flippase TAT-5 resulted in PE externalization and increased EV production (18, 19). Similarly, knocking down a mammalian homolog of TAT-5, ATP9A, increased EV release in cultured cancer cell lines (67), suggesting that this subfamily of flippases control lipid asymmetry and the associated membrane blebbing.

It is important to acknowledge that in addition to the decreased membrane packing and its hypothesized effect on membrane stiffness, release of lipid asymmetry likely changes other membrane properties that could potentially contribute to bleb formation and growth. For example, PE lipids have an intrinsic conical shape that promotes membrane curvature (68), thus translocation of PE to the outer leaflet by scrambling may stabilize the highly curved region at the neck of nascent blebs. This curvature sensitivity of PE may be related to the surprising absence of the PE probe from the bleb bodies (Fig. 2D). Further, bleb formation and expansion requires substantial membrane area, which can be supplied by membrane reservoirs at the cell surface (69). Interestingly, lipid scrambling by TMEM16F has been reported to release such membrane reservoirs (70). These factors may each contribute concomitantly to scrambling-dependent bleb formation (71) and their distinct effects require further studies.

Materials and Methods

Cell Culture.

BaF3 cell lines (kind gift of Shigekazu Nagata’s lab, Osaka University) (8) were cultured in RPMI 1640 medium supplemented with 50 μM β-mercaptoethanol and 45 U/mL recombinant mouse IL-3 (R&D Systems). RBL-2H3 cells were grown in Eagle's Minimum Essential Medium. HEK293 cells were cultured in Dulbecco's Modified Eagle Medium. PS-deficient PSA-3 cells were cultured in F-12K medium. PS was rescued in PSA-3 cells by addition of 10 μM ethanolamine, which enhances PSS2-catalyzed PS synthesis from PE (38, 39). All cell culture media were supplemented with 10% fetal bovine serum, 100 units/mL penicillin, and 100 μg/mL streptomycin.

Worm Culture.

C. elegans worms were maintained on Nematode Growth Media plates seeded with OP50 bacteria at room temperature (22 °C to 23 °C), according to standard protocols (72). The WEH642 strain has the genotype pad-1(babIs1[GFP]) I; wurIs155[pAZ132-coPH-oma-1(219-378): pie-1::mCh::coPH::CTPD; unc-119(+)]. For temperature shift experiments, worms were incubated at the indicated temperatures for 3.5 to 5 h before imaging at room temperature (RT).

RNA Interference.

RNA Interference (RNAi) was performed by feeding L1 larvae through adulthood at 25 °C, as described (73). Control WEH642 worms were fed with the empty RNAi vector pPD129.36 in HT115 bacteria. The anoh-2 RNAi clone was provided by Keith Choe (74).

Blebbing Buffer Treatment.

Cultured cells at 80% confluency were treated with blebbing buffer (150 mM NaCl, 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 2 mM CaCl₂, 25 mM paraformaldehyde, and 2 mM dithiothreitol, pH 7.4) at 37 °C for 1 h. Other blebbing conditions including MβCD (5 mM, Sigma) and DCA (0.25 mM, Sigma) treatment were performed as indicated in the manuscript.

Lipid Scrambling Detection.

For microscopic imaging, PS exposure was detected by incubating cells with AxV-AlexaFluor 568/647 (Invitrogen, 1:100 dilution) in blebbing buffer for 5 min at 23 °C. PE exposure was detected by incubating cells with Duramycin-PEG-GFP (10 μg/mL, gift from Dr. Ming Zhao from the Northwestern University) in blebbing buffer for 10 min at 23 °C. In PSA-3 cell experiments, the cells were treated with blebbing buffer for 20 min at 37 °C, and then, AxV568 was added to detect PS exposure. For live-cell experiments of PS exposure kinetics in RBLs and HEK293 cells, AxV647 were added to the blebbing buffer, and the cells were imaged at 23 °C every 30 s. The time-dependent intensity of AxV signal on a given cell was normalized to the maximum value of the same cell. For flow cytometry analysis, the cells were washed with PBS and resuspended in blebbing buffer at 37 °C for 5, 10, and 60 min. Then, AxV568 (1:100 dilution) was added to cells for 5 min at 23 °C and the cells were analyzed on a flow cytometer (BD LSRFortessa).

Observation of PIP₂ Depletion.

BaF3 TMEM16F-WT and -KO cells were transfected with PH-PLCδ-GFP plasmid (PIP₂ probe) through electroporation on a Nucleofector device. 24 h after transfection, the cells were washed with PBS and imaged under a Leica SP8 confocal microscope at 23 °C in imaging buffer (150 mM NaCl, 10 mM HEPES, and 2 mM CaCl₂, pH 7.4) with added PS probe (AxV647, 1:200 dilution). Then, 2× concentrated blebbing buffer was added to the cells 1:1 in volume to the imaging buffer, and PIP₂ depletion from the PM and PS exposure were monitored every 30 s. Images in Fig. 3A were taken 20 min after blebbing buffer addition.

GUV Preparation.

GUVs used for the calibrations in Fig. 3B were prepared by electroformation, as previously described (26). Briefly, lipid solutions with various lipid compositions [1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), 1,2-Dioleoyl-sn-glycero-3-phosphocholine (DOPC), and cholesterol, purchased from Avanti Polar Lipids] in chloroform (1.5 mg/mL, 1.5 μL) were applied to two electrodes of a custom-built chamber and dried under vacuum for 15 min. The electrodes were then submerged in a chamber preloaded with 300 μL of sucrose (0.1 M). Alternating current (2 V, 10 Hz, sine wave) was applied to the electrodes for 1.5 h at 60 °C. The GUVs were detached from the electrodes with a square wave current of 2 Hz, 2 V for 15 min. The GUV suspension was then cooled to RT and added to 1.7 mL of isotonic glucose (0.1 M) for settlement.

Laurdan GP Imaging.

For GUVs, after 1 h settling at 23 °C, 200 μL GUVs from the tube bottom was stained with Laurdan (1 μg/mL) at RT for 10 min. For GP measurement, the stained GUVs were imaged at 23 °C with a Leica SP8 confocal microscope. The samples were excited at 405 nm, and the emission around 440 nm (420 to 460 nm) and 490 nm (470 to 510 nm) was collected in two separate channels. For cell imaging, BaF3 WT or TMEM16F-KO cells in blebbing buffer with Laurdan (1 μg/mL, from Avanti) were incubated at 23 °C for 30 min before collecting Laurdan images.

GP Analysis of Laurdan Images.

GP analysis was performed as previously described (75). Briefly, cells were washed with PBS and stained with 10 µg/mL C-Laurdan for 10 min on ice and then imaged using confocal microscopy on a Leica SP8 with spectral imaging at 60× (water immersion, NA = X) and excitation at 405 nm. The emission was collected as two images: 420 to 460 nm and 470 to 510 nm. MATLAB (MathWorks, Natick, MA) was used to calculate the two-dimensional GP map, where GP for each pixel was calculated as previously described (76). Briefly, each image was background subtracted and thresholded to keep only pixels with intensities greater than three SD of the background value in both channels. The GP image was calculated for each pixel using Eq. 1, where g is the g-factor. The g-factor was calculated for each independent experiment. GP maps (pixels represented by GP value rather than intensity) were imported into ImageJ. To calculate the average PM GP, the average GP value was calculated after drawing a region of interest (ROI) on the PM of a cell. The GUV membrane was directly used as the ROI to calculate the average GP value of a given GUV.

$$\mathrm{GP}=\frac{{\sum }_{420}^{460}{I}_x-g{\sum }_{470}^{510}{I}_x}{{\sum }_{420}^{460}{I}_x+g{\sum }_{470}^{510}{I}_x}.$$ [1]

FLIM Imaging and Analysis.

Cells were washed with PBS and imaged with Di4 (1.5 μg/mL) in either imaging buffer or blebbing buffer at 23 °C using a Leica SP8 confocal. GUVs prepared in isotonic glucose solution were imaged with Di4 (1.5 μg/mL) at 23 °C. Di4 was excited with a pulsed laser (20 MHz) at 488 nm, and the emission was collected at 560 to 800 nm. During imaging, >100 photons/pixel were counted with no more than 0.5 photons per laser pulse per pixel. Image analysis was performed using Leica software. The pixels representing the PM of the cells was taken as the ROI and the resulting fluorescence decay curves from these pixels were fitted to a biexponential reconvolution function adjusted to the instrument response function. Di4 lifetime was reported using intensity-weighted mean value from the biexponential fitting.

EV Imaging.

mCh::PH::CTPD fluorescence on the surface of C. elegans embryos was observed on a Zeiss Axio Observer 7 microscope with a Plan-Apo 40× 1.4 NA oil objective and Excelitas Technologies X-Cite 120LED Boost illumination. Images were collected with a Hamamatsu ORCA-Fusion sCMOS camera controlled by 3i SlideBook6 software. mCh::PH::CTPD puncta with a signal-to-noise ratio minimum of 4 were detected from images of the top surface of 6- to 15-cell embryos using Matlab code based on CMEAnalysis (77).

MD Simulations.

To examine the relationship between area per lipid and bending rigidity, we used MD simulations. We analyzed simulation trajectories of lipid bilayers composed of either POPC or DOPC with 0, 10, 40, and 50 mol% cholesterol (Chol). The trajectories were taken from refs. 78 (for POPC), (for POPC/Chol) and 49 (for all DOPC bilayers). Briefly, all systems contained 100 lipids per leaflet (200 lipids total) and were constructed and equilibrated with the CHARMM-GUI web server (79–82) in all-atom representation. Subsequent production runs were conducted with NAnoscale Molecular Dynamics (83) using the CHARMM36 force field (84–86) at a fixed temperature of 25 °C (298.15 K) and pressure of 1 atm. Detailed simulation parameters are described in the original studies. Average area per lipid was calculated by dividing the average lateral area of the simulation box by the number of lipids in one leaflet. Bending rigidity was obtained from real-space analysis of lipid splay fluctuations, as described in ref. 50. Briefly, each lipid is defined by a director vector connecting the centers of mass of its headgroup and tail atoms. The distribution of angles between director vectors of neighboring lipids is then used to compute a potential of mean force (PMF). A small region of the PMF is fit to a quadratic function and the bending (splay) modulus is obtained from the coefficients of the best fit. This calculation is done separately for each leaflet and the bilayer bending rigidity is the sum of the splay moduli of the two leaflets.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.