Lipid accumulation promotes scission of caveolae

Caveolae, bulb-shaped invaginations of the plasma membrane (PM), show distinct behaviors of scission and fusion at the cell surface. Although it is known that caveolae are enriched in cholesterol and sphingolipids, exactly how lipid composition influences caveolae surface stability has not yet been elucidated. Accordingly, we inserted specific lipids into the PM of cells via membrane fusion and studied acute effects on caveolae dynamics. We demonstrate that cholesterol and glycosphingolipids specifically accumulate in caveolae, which decreases their neck diameter and drives their scission from the cell surface. The lipid-induced scission was counteracted by the ATPase EHD2. We propose that lipid accumulation in caveolae generates an intrinsically unstable domain prone to scission if not balanced by the restraining force of EHD2 at the neck. Our work advances the understanding of how lipids contribute to caveolae dynamics, providing a mechanistic link between caveolae and their ability to sense the PM lipid composition. SUMMARY Caveolae serve as mechanoprotectors and membrane buffers but their specific role in sensing plasma membrane lipid composition remains unclear. Hubert et al. show that cholesterol and glycosphingolipids accumulate in caveolae and drive subsequent scission from the cell surface. These results provide new insight into how lipids contribute to budding and scission of membrane domains in cells.

stability has not yet been elucidated. Accordingly, we inserted specific lipids into the PM of 23 cells via membrane fusion and studied acute effects on caveolae dynamics. We demonstrate 24 that cholesterol and glycosphingolipids specifically accumulate in caveolae, which decreases 25 their neck diameter and drives their scission from the cell surface. The lipid-induced scission 26 was counteracted by the ATPase EHD2. We propose that lipid accumulation in caveolae 27 generates an intrinsically unstable domain prone to scission if not balanced by the restraining 28 force of EHD2 at the neck. Our work advances the understanding of how lipids contribute to 29 caveolae dynamics, providing a mechanistic link between caveolae and their ability to sense 30 the PM lipid composition. were rapidly distributed throughout the basal membrane, as observed using live-cell TIRF 130 microscopy (Figs. 1B and S1E, exemplified by LacCer). The total fluorescence attributed to 131 the Bodipy motif increased uniformly in various regions of interest (ROIs) (Fig. 1B, bar plot). 132 To determine the amounts of lipids that were incorporated in the membrane through liposome 133 fusion, we used quantitative mass spectrometry on whole cells as 90% of these lipids are 134 located in the PM (Lorizate et al., 2013). The method was verified by altering the lipid 135 composition using myriocin (24h treatment) or sphingomyelinase (SMase, 2h treatment), 136 which are known to lower the levels of sphingomyelin (Gulshan et al., 2013). Analysis 137 showed that these treatments drastically decreased SM(d18:1/16:0) levels, the major 138 endogenous species of SM (Fig. 1C). Next, we incubated cells with fusogenic liposomes 139 containing Bodipy-labeled LacCer or SM C 12 , and analyzed the lipid composition by LC- Bodipy-LacCer and Bodipy-SM C 12 per 400 000 cells were measured to be 4.2 pmol and 2.7 145 pmol, respectively, (i.e., 6.3 x 10 6 and 4.0 x 10 6 lipids/cell) (Fig. 1D). To assess the 146 incorporation efficiency of Chol, deuterium labeled Chol, d7-Chol, was included into 147 fusogenic liposomes. GC-MS/MS analysis revealed that d7-Chol was incorporated to similar 148 levels as Bodipy-labeled LacCer and SM C 12 . Given that the PM of these HeLa cells harbor 149 around 7 x 10 9 lipids/cell (see Methods section for details), the levels of Bodipy-and d7-150 labeled lipids detected by mass spectrometry led to a 0.02-0.09% increase in specific labeled 151 lipids and a 0.4-1.6% of total lipids. 152 To determine the rate of incorporation of the different Bodipy-labeled lipid species, we used 153 spinning disk microscopy in a central confocal plane of the cell (Fig. 1E). Quantitative 154 analysis of lipid incorporation into the PM over time revealed similar levels for most lipids 155 ranging from 1900 to 4800 arbitrary units at 10 min (Fig. 1E). The variation in incorporation 156 rates between the different lipids may be due to marginal differences in the fusogenicity of 9 (ROI) after bleaching was monitored. Lateral diffusion within the PM was similar between 161 the different Bodipy-lipids employed (Fig. 1F). 162 To conclude, the use of fusogenic liposomes enabled rapid incorporation of approximately 4 163 x 10 6 specific lipids into the PM per living cell over 10 minutes. Based on our tracking data, 164 we estimated that each cell contains around 300 caveolae, comprising around 0.1% of the 165 surface area. Caveolae are approximately 60 nm in diameter, and each lipid occupies 0.62 Å. 166 Therefore ca. 10 x 10 6 lipids are contained within the caveolae, of which 50% is Chol. The 167 amount of specific incorporated lipids in our system is therefore about half of the total 168 amount of lipids contained within caveolae. The immediate addition of extra lipids to the PM 169 did not result in a detectable effect on the cell volume (Fig. S1H).

170
GSLs and Chol decrease the surface stability of caveolae 171 We next aimed to elucidate if lipids are involved in controlling the balance between stable 172 and dynamic caveolae at the PM and if effects could be attributed to individual lipid species. 173 To visualize caveolae, we generated a stable mammalian Flp-In T-Rex HeLa cell line 174 expressing Cav1-mCherry, hereafter named Cav1-mCh HeLa cells. Protein expression was 175 induced by doxycycline (Dox) to achieve expression of Cav1-mCherry at similar level as 176 endogenous Cav1 (Fig. S1I). Using TIRF microscopy and single-particle tracking, we and fusion ("Surface adjacent"), but remain close to the surface will have a high speed and 186 short duration as they are not stably fused with the PM and short duration ( Fig. 2A) GSLs were due to their PM release, as characterized by loss of the stabilizing protein EHD2. 230 Therefore, we treated Cav1-mCh HeLa cells with the different fusogenic liposomes and 231 visualized endogenous EHD2 using indirect immunofluorescent labeling (Fig. 3A). These 232 experiments revealed that incorporation of GM1, LacCer or Chol into the PM led to a 233 significantly lower amount of EHD2 localized to Cav1 (Fig. 3A, scatter plot). This data 234 suggests that the caveolae release induced by increased PM levels of LacCer and Chol is due 235 to a loss of EHD2-mediated stabilization. Conversely, Cer and SM C 12 , as well as its short 236 chain analogue SM C 5 , did not appear to have any significant effect on the association of 237 EHD2 with Cav1 (Fig. 3A, scatter plot). Further experiments showed that, following lipid 238 treatment, the majority of caveolae remained associated with cavin1, revealing that no 239 disruption of the caveolae coat, and subsequent release of cavin1, occurred (Fig. 3B).  (Fig. S3G). This verified 271 that stable assembly, but not disassembly of EHD2, is necessary to stabilize caveolae. 272 To clarify whether, in order to have a stabilizing role, EHD2 had to be caveolae-associated 273 prior to lipid addition, fluorescently labeled, purified EHD2 (EHD2-647) was microinjected 274 into Cav1-mCh HeLa cells (Fig. 3G). Within 20 min, EHD2-647 colocalized with Cav1, 275 confirming that the microinjected protein was indeed recruited to caveolae (Figs. S3H-I). 276 Next, we tested if an acute injection of EHD2-647 could rescue the effect on caveolae 277 dynamics caused by LacCer. Strikingly, we found that exogenously added EHD2 stabilized 278 the caveolae to the same extent as the overexpressed EHD2, demonstrating that increased 279 levels of EHD2 can acutely reverse the increased mobility of caveolae induced by lipids (Fig. 280 3H).

281
LacCer and Chol accumulate in caveolae and Chol is sequestered within these domains 282 As GSLs and Chol increased the surface release of caveolae, we aimed to determine whether 283 there was a differential accumulation of lipids within caveolae at the PM. We treated Cav1-284 mCh HeLa cells with fusogenic liposomes and followed the distribution of Bodipy-labeled 285 LacCer or Chol using live-cell TIRF microscopy. After 15 min, both lipids were found to These rapid fusion events enabled us to study, for the first time, how caveolae respond to an 377 acute change in PM lipid composition and to observe lipid exchange in the caveolae bulb. 378 Furthermore, the use of labelled lipids allowed us to measure the levels of incorporation in 379 relation to endogenous levels. Our results demonstrate the power of this approach for 380 studying caveolae dynamics and we foresee that our methodology will also be a useful tool 381 outside of this framework. 382 Our work shows that the surface association of caveolae is highly sensitive to changes in the 383 PM lipid composition. An acute increase in the levels of Chol and GSLs, which were found 384 to specifically accumulate in caveolae, dramatically increased caveolae mobility. These 385 caveolae traveled at higher speeds, their PM duration was shorter and they also displayed 386 reduced levels of EHD2, a protein indicative of PM-associated caveolae. Therefore, we 387 conclude that accumulation of Chol and GSLs in caveolae trigger surface release of caveolae. 388 In agreement with this, analysis by EM revealed that the caveolae neck diameter was reduced study, increased caveolae mobility is a direct result of lipid accumulation in these structures. 393 As our methodology allowed us to determine the levels of specifically incorporated lipids, we 394 found that rapid, yet relatively small increases in specific lipids can affect caveolae dynamics. 395 Because caveolae immediately responded to these changes in bilayer composition, we 396 propose that they serve as PM sensors, not only for membrane tension, but also for lipid 397 composition. 398 Previous studies have suggested that a threshold concentration of Chol is required to maintain 399 caveolae integrity and proposes that assembly and disassembly is in a dynamic equilibrium 400 dependent on Chol levels (Hailstones et al., 1998). This is also in line with our experiments 401 showing that excess Chol drives caveolae assembly towards scission and that Chol was 402 indeed found to accumulate in caveolae when these structures were restrained to the surface 403 by EHD2 overexpression.
Furthermore, our methodology enabled, for the first time,

562
One day prior to experiment, cells were seeded in a 6-well plate. Cells were left untreated or 563 treated with 11.7 nmol/ml of the different fusogenic liposomes for 10 min at 37C, 5% CO 2 . 564 The cells were washed three times with PBS and harvested in 500 μl MeOH by scraping. 565 Counting revealed that approximately 4 × 10 5 cells were obtained per sample. For myriocin 566 (2.5 μM) and SMase (0.01 U) treatment, cells were incubated for 24 h or 2 h, respectively. 567 Extraction was performed using a mixer mill set to a frequency 30 Hz for 2 min, with 1 568 tungsten carbide bead added to each tube. Thereafter the samples were centrifuged at 4ºC,    Calculations of the number of PM lipids.