Formation of tubules and helical ribbons by ceramide phosphoethanolamine-containing membranes

Ceramide phosphoethanolamine (CPE), a major sphingolipid in invertebrates, is crucial for axonal ensheathment in Drosophila. Darkfield microscopy revealed that an equimolar mixture of bovine buttermilk CPE (milk CPE) and 1,2-dioleoyl-sn-glycero-3-phosphocholine (diC18:1 PC) tends to form tubules and helical ribbons, while pure milk CPE mainly exhibits amorphous aggregates and, at low frequency, straight needles. Negative staining electron microscopy indicated that helices and tubules were composed of multilayered 5–10 nm thick slab-like structures. Using different molecular species of PC and CPE, we demonstrated that the acyl chain length of CPE but not of PC is crucial for the formation of tubules and helices in equimolar mixtures. Incubation of the lipid suspensions at the respective phase transition temperature of CPE facilitated the formation of both tubules and helices, suggesting a dynamic lipid rearrangement during formation. Substituting diC18:1 PC with diC18:1 PE or diC18:1 PS failed to form tubules and helices. As hydrated galactosylceramide (GalCer), a major lipid in mammalian myelin, has been reported to spontaneously form tubules and helices, it is believed that the ensheathment of axons in mammals and Drosophila is based on similar physical processes with different lipids.


Discussion
CPE is enriched in Drosophila glial cells 6,7 and is required for the development of cortex glia 22 . However, little is known about the physical properties of CPE. The present study shows that hydrated CPE:PC mixtures form vesicular, tubular and helical structures, depending on the N-linked acyl chain length and incubation conditions. The gel to liquid crystalline phase transition temperature of d18:1/N-16:0 CPE (64 °C) has been reported to be 23 degrees higher compared to d18:1/N-16:0 SM (41 °C) 16 and is consistent with our results (63 °C). Similarly, the phase transition temperature of d18:1/N-24:0 CPE (72 °C) and d18:1/N-24:1 Δ15 (c) CPE (50 °C) is more than 20 degrees higher compared to their SM analogues N-24:0 SM (47.5 °C) 23 and N-24:1 Δ15 (c) SM (22.3 °C) 3 . Similar differences in the physical properties of choline-containing lipids compared to ethanolamine-containing lipids were observed in glycerolipids, as the phase transition temperature of saturated PE is 20-30 degrees above the corresponding PC 24 . These differences are attributed to both the stronger headgroup interaction and the tighter packing owing to the small size of the headgroup and the intermolecular hydrogen bonding of PE molecules. By www.nature.com/scientificreports www.nature.com/scientificreports/  www.nature.com/scientificreports www.nature.com/scientificreports/ analogy with PE, the small headgroup size of CPE has been proposed to play a critical role to facilitate the strong CPE-CPE interaction 16 . This is in line with our recent report of stronger amide bond interaction between CPE molecules compared to SM molecules 17 .
Milk CPE exhibits a main phase transition temperature of 56-59 °C and the primary N-acyl lengths are 22:0 (35.8%), 23:0 (29.7%) and 24:0 (18.3%). Even above transition temperature, pure milk CPE films poorly detached from the tube walls and the resulting suspension exhibited amorphous aggregates. Needle and helical structures were observed at low frequency. In contrast, equimolar milk CPE: diC18:1 PC mixture was readily suspended in MilliQ water or PBS and exhibited helical structures. At the observing temperature 21-22 °C, milk CPE is in gel phase while diC18:1 PC (phase transition at −17.3 °C 25 ) is in liquid crystalline phase. Consequently, phase separation of the two lipid phases would be expected. The fluorescent dye DiI C18 labeled helical and tubule structures, but not lipid aggregates formed in equimolar lipid suspensions. This suggests that helical structures are enriched with solid milk CPE whereas aggregates are diC18:1 PC-rich. Systematic variation of the milk CPE:diC18:1 PC ratio revealed that helical structures were dominant in mixtures with high CPE content, above 50%, consistent with the notion that helices are mainly composed of CPE. The homogeneously labeled helical structures by DiI www.nature.com/scientificreports www.nature.com/scientificreports/ C18 suggest that helical structures are not phase separated. However, the presence of phase separated domains below the resolution limit of optical microscopy in the helical structures cannot be excluded.
Unlike milkCPE:PC (1:1) suspensions, equimolar suspensions of milk CPE:diC18:1 PE, milk CPE:diC16:0 PE and milk CPE:cholesterol exhibited amorphous aggregates. It is speculated that the small headgroups of both CPE and PE favor tight interaction, impeding hydration of these mixtures (Fig. 12A,B). In case of CPE:cholesterol, the hydrated membranes likely coexist with cholesterol precipitates due to the low solubility of cholesterol in CPE 16,17 (Fig. 12C). While CPE and PS can engage in intermolecular hydrogen bonding, the comparatively larger headgroup of PS likely facilitates hydration of the membrane. Consequently, it can be speculated that in the presence of PS the formation of CPE-rich lipid domains is prevented, precluding the formation of helical structures (Fig. 12D).   3 . It has been proposed that intermolecular hydrogen bonds play an important role in tubules and helices formation 3 . However, there are several differences between CPE and GalCer. First, GalCer alone forms helices and tubules whereas CPE forms these structures only www.nature.com/scientificreports www.nature.com/scientificreports/ when mixed with PC. Second, N-24:1 GalCer forms tubules whereas N-24:0 GalCer forms helices. In contrast, both d18:1/N-24:0 CPE:diC18:1 PC (1:1) and d18:1/N-24:1 Δ15 (c) CPE:diC18:1 PC (1:1) display helical structures. These results suggest that in addition to the strong hydrogen bonding between the lipid headgroups, other headgroup specific features play an additional role in the formation of tubules and helices.
Helices and tubules formed efficiently upon hydration at the phase transition temperature of the respective CPE. This suggests that dynamic lipid re-organization is required to form helical structures, as the transbilayer lipid movement is accelerated at the phase transition temperature 28 . We speculate that the hydrophobic mismatch between very long chain CPE and diC18:1 PC induces tilting of CPE rich membrane domains (Fig. 12E). Restraining the chiral CPE molecules to a bilayer phase while maintaining a tilt with respect to the local layer normal together with the favored twist of neighbor to neighbor interaction, guides the whole assembly to twist into a helical ribbon 3,29 . Within the plane of the monolayer, the strong linear lipid association necessary to stabilize the formed tubular and helical structures are likely supported by hydrogen bonding 3,29 . Interdigitated interaction between the very long chain CPEs in PC membrane may also stabilize the tilted CPE domains (Fig. 12E). Moreover, the thin water layer on CPE may facilitate the piling up of bilayer as observed by electron microscopy (Figs 5 and 12F).
The EGFP-tagged CPE-binding protein PlyA2 was unable to bind to preformed helical structures composed of milk CPE:diC18:1 PC. Since tight lipid packing was reported to prevent binding of lipid-binding peptides 30 , we speculate that CPE is tightly packed in helical structures. In contrast, PlyA2-EGFP could bind to helical structures when the protein was added during their formation. This indicates that PlyA2 does not inhibit helix formation. As it was shown previously, lipid binding proteins can only bind a few % of target lipids due to steric hindrance 31 , it is thus envisaged that only a limited amount of PlyA2 binds to the forming helices, without significantly affecting the overall lipid assembly.
Although GalCer is enriched in myelin, inhibition of GalCer biosynthesis did not inhibit myelin formation 32,33 . Nevertheless, myelin abnormalities such as reduced myelin thickness, redundant myelin outfoldings and vacuole formation were detected in mutant mice 32,33 . It is interesting to note that mice with double deficiency of GalCer synthase and fatty acid 2-hydroxylase form myelin enriched with SM. Currently it is proposed that GalCer is required for the stability and maintenance of myelin during ageing 34 .
CPE is required for the ensheathment of peripheral nerves in Drosophila 6,22 . While it cannot be excluded that CPE is involved in ensheathment via direct interaction with regulating proteins, both the multilayer and tight packing nature of CPE:PC membranes suggest that CPE directly supports and stabilizes the glial membranes wrapped around neurons in Drosophila. It is interesting that similar to GalCer in mammalian myelin, replacement of CPE with SM rescued cortex glial abnormalities in Drosophila 22 . This suggests that similar to GalCer, CPE may be involved in the long term stability of glial membranes or in pathological conditions.
preparation of lipid suspensions. Lipid suspensions were prepared by gentle hydration of lipid films. To this end, CPE stock solutions were prepared in chloroform:methanol (2:1) while all other lipid stock solutions were in chloroform only. The stock solutions were mixed in a glass test tube at the desired molar ratio (total 150 n moles), dried under N 2 flow and kept in vacuo at least for 2 h. Removal of trace organic solvent from lipid film is crucial to obtain helical structures. 500 μL MilliQ water (Millipore, Japan) or PBS (1.058 mM KH 2 PO 4 , 2.96 mM Na 2 PO 4 , 155 mM NaCl, pH 7.2) was added to the lipid film at room temperature (21-22 °C) and the resulting mixture was incubated at the indicated temperatures for 3 h to prepare the lipid suspension. The lipid suspensions were observed by darkfield or fluorescence microscopy at room temperature. Alternatively, the lipid suspensions were further processed for electron microscope observation.
Gas chromatography. CPE was transmethylated with boronfluoride-methanol reagent 14% at 100 °C for 90 min as previously described 36 . The resulting fatty acid methyl esters (FAME) were extracted with hexane, and analyzed on a Shimadzu GC-14A gas chromatograph using an Omegawax 320 (30 m × 0.32 mm × 0.25 μm) capillary column (Supelco, Bellefonte, PA, USA) and an oven temperature programmed from 100 °C to 250 °C with helium as a carrier gas 36 . FAME peaks were identified by their retention times compared to known standards and the results expressed in mol %. (2019) 9:5812 | https://doi.org/10.1038/s41598-019-42247-1 www.nature.com/scientificreports www.nature.com/scientificreports/ Differential scanning calorimetry (DSC). The lipid films were prepared as outlined above and hydrated with HEPES buffer (20 mM HEPES at pH 7.0, 100 mM NaCl, 100 mM EDTA) at a final concentration of 0.5 mM. After three freeze-thaw cycles (each cycle: −80 °C, 15 min; 60 °C, 15 min; 30 s vortex mixing), the hydrated lipid samples and reference buffer solutions were degassed in vacuo for at least 5 min (Microcal ThermoVac vacuum pump) prior to loading into the GE Healthcare MicroCal VP-DSC Microcalorimeter. Samples were scanned at a rate of 10 degrees/h over a range of 20-80 °C with 10 repeat cycles. The data of the 10 th heating curve was processed with Origin 7 software. Anisotropy measurement. DPH anisotropy measurements were performed on a Fluorolog spectrofluorometer (Horiba, Kyoto, Japan), operating in the T format as reported previously 37 with slight modifications. The lipids were mixed with DPH at a molar ratio of 1/200 and dried under N 2 gas. The dry samples were kept in high vacuum for at least 1 h. Lipid suspensions were prepared by hydrating 100 or 200 mmol/L lipids in water at 60 °C. The suspensions were sonicated using a bath sonicator (US-1A, As One, Osaka, Japan) for 5 min. Samples were repeatedly scanned three times within the indicated temperature ranges at a rate of 1 degree/min. Excitation and emission wavelengths were 357 and 451 nm, respectively. The steady-state anisotropy, r, was determined from the 4 th scan as described 38 .
Darkfield microscopy and fluorescence microscopy. The shape of lipid assemblies in suspension was observed by darkfield microscopy 39 . The darkfield microscope (Eclipse E600, Nikon, Japan) was equipped with two objectives (100x, N.A. = 0.5-1.3 and 40x, N.A. = 0.75) and a darkfield condenser (N.A. = 1.0-1.4). Images were captured with a sCMOS camera (ORCA-Flash4.0, Hamamatsu photonics, Japan) and processed using imageJ (NIH, USA). Simultaneous observation of darkfield and epifluorescence images was performed with the same objective and appropriate filter units (B3A for EGFP and G2A for DiI C18 and mCherry). electron microscopy. Negative staining electron microscopy was performed as reported previously 30 with some modifications. Lipid suspensions were adsorbed onto poly-D-lysine-treated pioloform-coated specimen grids and negatively stained with 2% uranyl acetate. The samples were examined by transmission electron microscope (JEM1230, JEOL, Japan) with the help of the Materials Characterization Team in RIKEN Advanced Technology Support Division. Electron micrographs were recorded with a CCD camera (Veleta, Olympus-SIS, Germany).