Introduction

Autophagy is a widespread phenomenon, occurring in virtually all eukaryotic cells. This highly conserved process is aimed at the recycling of cytoplasmic molecules or organelles existing in excess amounts, or containing incorrect structures. The corresponding macromolecules are degraded in the lysosome to their basic building blocks, e.g. amino acids from proteins, or nucleosides from nucleic acids, and the latter are reused in the cellular biosynthetic pathways. The term “autophagy” was invented by Christian de Duve1, the discoverer of lysosomes and peroxisomes. Of the different types of autophagy described in mammalian cells2, macroautophagy is the main autophagic pathway, occurring mainly to eradicate damaged cell organelles or unused proteins. This involves the formation of a double membrane around cytoplasmic substrates resulting in the organelle known as autophagosome (AP). The AP travels through the cytoplasm to a lysosome, and the two organelles fuse. Within the lysosome, the contents of the autophagosome are degraded via acidic lysosomal hydrolases. In the context of this paper, ‘macroautophagy’ will hereafter be designed as ‘autophagy’. This pathway is conserved from yeast to humans. Generation and elongation of the nascent AP constitute the starting point of autophagy, and, as in many other biological events, the first step is critically important for the ensuing developments3,4,5,6,7,8,9.

LC3, a mammalian ortholog of yeast Atg8, is part of an ubiquitin-like system that is crucial to autophagy. Another autophagy protein, Atg3, binds LC3 and directs it to the autophagosomal membrane, in which LC3 is conjugated to phosphatidylethanolamine to produce LC3-PE10,11. LC3-PE is considered as the main autophagy specific marker in cells, and it is also the best known effector in autophagosome elongation8,9,12,13. Thus, LC3 is a major object of research in studies on autophagy and related cellular activities, including apoptosis, and methodological advances such as the one described above might prove very useful in facilitating experimental advances in the field. LC3 is in fact a subfamily of Atg8 orthologs, consisting of three members, namely LC3A, LC3B, and LC3C. In the authors’ laboratory, studies with model membranes, mainly liposomes, have demonstrated the role of certain lipids in modulating LC3 binding to membranes, and the induction of vesicle-vesicle fusion by LC3-PE, liposome fusion being hinted at as a basic mechanism for AP growth. Each of the three members of the LC3 subfamily exhibits characteristic properties in the binding, aggregation and fusion of model membranes, suggesting specific biological roles for each LC3 homolog12,13,14. The current study is devoted to the arguably earliest stage of LC3 interaction with the nascent AP, namely LC3C insertion into the AP bilayer. This occurs in the absence of any ubiquitin-like proteins and does not lead, in itself, to LC3 lipidation3,12.

Fluorescence is the emission of light by a substance that has absorbed light or other electromagnetic radiation. It occurs when a photon from incoming radiation is absorbed by a molecule, exciting it to a higher energy level, followed by the emission of light as the molecule returns to a lower energy state. Fluorescence spectroscopy is extensively used in Membrane Biophysics studies, and particularly in the field of protein-lipid interactions. Fluorescence may arise from native membrane components (intrinsic fluorescence), or from externally added (extrinsic) probes. Both the intrinsic fluorescence of protein residues, mainly tryptophan, and of a myriad extrinsic probes, have been widely applied. Nile red (9-(diethylamino)-5 H-benzo[α]phenoxazin-5-one), a fluorescent hydrophobic probe (Supplementary Chart 1), has lately found a wide range of applications in cell biology15,16. However, its use in the study of lipid bilayers has been limited. In a recent paper, we have shown the use of Nile red as a fluorescent probe in the study of phases and phase transitions in liquid and gel lipid bilayers. A red/orange intensity ratio (ROIR) index was found to be useful in order to normalize the results15. Nile red could be excited without noticeable photolytic effects16.

The present contribution describes a novel application of Nile red in the field of biomembranes, namely the assessment of autophagy protein binding to lipid bilayers. The fluorescent method described here surpasses the traditional density gradient ultracentrifugation measurements in convenience and economy. LC3A, LC3B and LC3C binding to lipid vesicles of various lipid compositions has been comparatively assayed with the ultracentrifugation and the Nile red methods, with closely similar results. The recently described effect of LC3 binding enhancement by ceramide (Cer)17 has been confirmed with the novel fluorescence method.

Materials and methods

Materials

Bioreagent quality Nile red (19123) and sphingomyelinase (EC 3.1.4.12) from Bacillus cereus (SMase, S9396) were purchased from Sigma-Aldrich (St. Louis, MO). Sphingomyelin (egg, chicken) (eSM, 860061); 1,2-dioleoyl-sn-glycero-3-phosphatidylethanolamine (DOPE, 850725); heart bovine cardiolipin, 90% tetralinoleoyl (CL, 840012); ceramide (egg, chicken) (eCer, 860051), were obtained from Avanti Polar Lipids, Inc. (Alabaster, AL). Methanol and chloroform were from Fisher (Suwanee, GA). Buffer solution, unless otherwise stated, was 50 mM Tris-HCl, 150 mM NaCl, pH 7.4. All salts and organic solvents were of analytical grade.

Recombinant protein expression and purification

The pGEX4T-1 plasmids for the expression of glutathione S-transferase (GST)-tagged human orthologs of yeast Atg8, (human LC3A, human LC3B, human LC3C) were kindly provided by Dr. Ivanna Novak (School of Medicine, University of Split, Croatia). LC3 proteins were purified from soluble fractions of bacterial extracts obtained in the absence of detergents, and they were > 90% pure as evaluated by Coomassie Brilliant Blue-stained SDS-PAGE. Escherichia coli BL21 (λDE3) cells were transformed with the appropriate plasmids. After breaking the cells by sonication and removing cellular debris by centrifugation at 30,000 x g for 30 min at 4 °C, the sample supernatant fraction was incubated with 1 mL Glutathione Sepharose 4B beads for 3 h at 4 °C to bind GST-tagged proteins. Bound proteins were cleaved with Thrombin Protease (GE Healthcare, 27–0846-01) overnight at room temperature in Thrombin Cleavage Buffer (10 mM Na2HPO4, 1.8 mM KH2PO4, 140 mM NaCl, 2.7 mM KCl, 1 mM DTT, pH 7.3). Further details on LC3 protein expression and purification can be found in14.

Large unilamellar vesicle (LUV) Preparation

The appropriate lipids were mixed in organic solution and the solvent was evaporated to dryness under a N2 stream. Nile red was added with the lipids at a 250:1 lipid: probe ratio. Then, the sample was kept under high vacuum for 2 h to remove solvent traces. The lipids were swollen in buffer in order to obtain multilamellar lipid vesicles (MLV). When required, hydration was enhanced by stirring with a glass rod, and the vesicles were homogenized by forcing the sample ≈ 80 times between two syringes through a narrow tube (0.5-mm internal diameter, 10-cm long) at 50 °C. MLV were subjected to 10 freeze/thaw cycles, and then extruded using 0.1-µm pore size Nuclepore filters to obtain ≈ 100-nm LUV18. LUV were centrifuged for 30 min at 9,000 x g and 4 °C, and the supernatants were saved for further use. Vesicle size was checked using a Malvern Zeta-Sizer Nano ZS (Malvern, Instruments, UK). Phospholipid concentration was determined with a phosphate assay19. For protein binding assays, an LUV suspension (0.3 mM lipid, containing Nile red) was incubated with the appropriate protein (up to 3 µM) for 45 min before the fluorescence measurements.

Sphingomyelinase (SMase) treatment

SMase (0.05 units) was added to 1.0 ml vesicle suspension, 0.3 mM in lipid, at 37 °C, in the presence of 2 mM o-phenanthroline to inhibit possible traces of phospholipase C activity. SMase activity was assayed by determining phosphorus contents in the aqueous phase of an extraction mixture (chloroform: methanol, 2:1) after addition of aliquots from the reaction mixture at different times. Because of the 1:1 ceramide: phosphate stoichiometry of the enzyme products, enzyme activities could be given either as ceramide production or as phosphate production.

Fluorescence spectroscopy

1 mM LUV, containing 4 µM Nile red (250:1 lipid: probe mole ratio), were prepared. Then, the samples were diluted to a 0.3 mM lipid concentration and the fluorescence spectra of the Nile Red probe were retrieved in a QuantaMaster 40 spectrofluorometer (Photon Technology International, Lawrenceville, NJ). Emission spectra were collected between 570 and 800 nm, exciting at 540 nm. A thermal TC125 controller (Quantum Northwest, Liberty Lake, USA) was used to stabilize the sample temperature at 37 °C. Once the emission spectra were obtained, in order to calculate the Red/Orange Intensity Ratios (ROIR), the integral areas under the orange (573–613 nm) and red (650–800 nm) regions were computed with the PTI FelixGX software (Photon Technology International, Lawrenceville, NJ)15 using this simple equation:

\({\text{ROIR}}\,=\,{{\text{A}}_{{\text{65}}0\, - \,{\text{8}}00}}/{{\text{A}}_{{\text{573}}\, - \,{\text{613}}}}\)

where A650 − 800 and A573 − 613 correspond, respectively, to the integral areas of the red and orange regions in the Nile red emission spectra.

Vesicle flotation

The standard assay of protein interaction with membranes uses flotation in sucrose gradients. Liposomes were incubated with the indicated proteins for 1 h at 37 °C in buffer. The protein/lipid mix was adjusted to 1.4 M sucrose concentration in 300 µl and transferred to a centrifuge tube. This first (bottom) layer was overlaid with successive solutions containing 0.8 M (400 µl) and 0.5 M (300 µl) sucrose. The three-layer gradients were centrifuged in a TLA-120.2 rotor (Beckman Coulter, Brea, CA, US) at 355,040 × g for 50 min at 4 °C. After centrifugation, four 250-µl fractions were collected, starting from the bottom. Proteins were detected in SDS-PAGE gels using Coomassie Brilliant Blue staining. Densitometry of the protein bands was performed using ImageJ software, and the percent liposome-bound protein was estimated from the band intensities measured in the third + fourth fractions (floating vesicle fractions), relative to the total sum of intensities measured in all fractions14.

Cryo-EM sample Preparation and image collection

LUVs with and without protein were incubated at 37 °C for 1 h with continuous stirring and then loaded on freshly glow-discharged 300- mesh R2/2 Quantifoil holey carbon grids (Quantifoil Micro Tools GmbH). Vitrification was performed on LEICA GP2 automatic plunge freezer (LEICA microsystems) maintained at 8 °C at a relative humidity close to saturation (90% rH). Grids were loaded with 4 µL sample solutions for 30 s, blotted with absorbent standard filter paper, and plunged into a liquid ethane bath. The vitrified grids were removed from the plunger and stored under liquid nitrogen. Imaging of cryo-EM samples was made on a JEM-2200FS/CR (JEOL Europe, CIC bioGUNE, Spain) transmission electron microscope operated at 200 kV and images were recorded under low-dose conditions, with a total dose of the order of 30–40 electrons/Å2 per exposure, at defocus values ranging from − 1.5 to -4.0 μm. The in-column Omega energy filter of the microscope helps to record images with improved signal-to-noise ratio (SNR) by zero-loss filtering, using an energy selecting slit width of 20 eV centered at the zero-loss peak of the energy spectra. Digital images were recorded on a GATAN K2 summit direct detection camera 4 K × 4 K (5 μm pixels) (Gatan Inc., Pleasanton, CA) using DigitalMicrograph (Gatan Inc.) software, at a nominal magnification of 30,000×, resulting in final sampling of 1.3 Å/pixel20.

Statistics

Unless otherwise stated, all data are given as average values ± SD of three measurements obtained with different vesicle preparations. When required, statistical significance was estimated with the Student’s t-test.

Results

Representative examples of Nile red fluorescence emission spectra in lipid bilayers are shown in Fig. 1. Bilayers of four different lipid compositions are depicted, in all cases in the presence of 3 µM LC3C, at a 250:1 lipid: protein mole ratio. The differences were small but highly reproducible, and data analysis was made easier by the use of ROIR. Nile red shows absorption maxima at 520–550 nm and emission maxima at 580–640 nm, single photon excitation not being damaging to the system. A red/orange integral intensity ratio (ROIR) index has been used to normalize the emission spectra results15. The larger the 650–800 nm (red) area with respect to the 573–613 nm (orange) area, the larger the ROIR, and the more fluid/disordered the bilayer.

Fig. 1
figure 1

Fluorescence emission spectra of Nile red in vesicle lipid bilayers, in the presence of 3 µM LC3C. Bilayer lipid compositions, in mole ratios, were SM: DOPE (50:50), SM: DOPE: eCer (40:50:10), SM: DOPE: CL (33:33:33), and SM: DOPE: CL: eCer (23:33:33:10). In all cases, lipid: probe mole ratio was 250:1.

Full size image

ROIR decreased with LC3C concentration, and the protein effect was much more marked when CL was present in the bilayers (Fig. 2). Fluorescence spectra were retrieved 45 min after protein addition. Preliminary experiments had shown that 45 min were enough for the system to reach equilibrium (supplementary figure S1). Supplementary figure S2 shows an overlay of Nile red spectra in the presence and absence of protein. In the presence of protein, the left-hand part of the spectrum is shifted to lower wavelengths, and this leads to the decreased ROIR. The effect is apparently small, but highly reproducible, judging from the very good fitting (high correlation coefficients) of the experimental points to the computed hyperbolas. Parallel experiments with cryo-EM had shown that LC3C failed to induce observable changes in the LUV morphology upon bilayer insertion (supplementary figure S3). LC3C concentrations were up to 3 µM, due to difficulties in obtaining larger amounts of protein. However, the effect of 10 µM LC3C was tested with the SM: DOPE: CL: eCer lipid composition, and the data fitted very well the computed hyperbola (data not shown).

Fig. 2
figure 2

% ROIR decrease of Nile red fluorescence emission as a function of LC3C concentration, in bilayers of various lipid compositions. (A) Binding profile. The data have been fitted to hyperbolic equations, of the general form y = y0 + a*x/(b + x). Correlation coefficients R for the various lipid compositions are SM: DOPE (R = 0.9899), SM: DOPE: eCer (R = 0.8919), SM: DOPE: CL (R = 0.9995), and SM: DOPE: CL: eCer (R = 0.9998). (B) Detailed view of binding at 0–3 µM LC3C concentrations. The straight lines in (B) correspond to the initial (maximum) slopes of the % ROIR decrease vs. LC3C concentration curves. Data obtained from experiments as shown in Fig. 1. ROIR were computed as indicated under Methods. Quantitative lipid compositions were SM: DOPE (50:50), SM: DOPE: eCer (40:50:10), SM: DOPE: CL (33:33:33), and SM: DOPE: CL: eCer (23:33:33:10).

Full size image

In CL-containing bilayers, with LC3C, the additional presence of Cer further decreased ROIR. ROIR decreased with LC3C concentrations following a hyperbolic curve (Fig. 2A) and this allowed an estimation of the maximum decrease for each lipid composition, as shown in Fig. 3A. The slopes of the ‘ROIR vs. LC3C concentration’ curves are an additional parameter that could be used in the fluorescence data analysis (Fig. 3B). Figure 3 shows very clearly the effects of CL and Cer: when added to SM: DOPE (50:50) CL causes a significantly larger decrease of ROIR (Fig. 3A), and the decrease is much faster (Fig. 3B). In turn the presence of Cer, when added at 10 mol% to a eSM: DOPE: CL (33:33:33) mixture, giving rise to a eSM: DOPE: CL: eCer (23:33:33:10) composition, induces a larger decrease of ROIR (Fig. 3A) and a larger slope (faster change in ROIR) (Fig. 3B).

Similar effects of LC3C on ROIR, although quantitatively less extensive, were found when experiments were performed starting with a more complex mixture, mimicking autophagosomes21 (supplementary figure S4). Again, the larger LC3C binding occurred in the presence of CL, and/or of CL + Cer.

Fig. 3
figure 3

LC3C-induced % ROIR decrease of Nile red fluorescence emission in bilayers of various lipid compositions. (A) Maximum observed values of % ROIR decrease. (B) Slopes of the % ROIR vs. LC3C concentration curves. Quantitative lipid compositions as in Fig. 1. Average values ± S.D., n = 3. Statistical significance was estimated according to the Student’s t-test, p values are indicated for each pair of significantly different data.

Full size image

A comparison of the fluorescence and ultracentrifugation data for LC3 binding to bilayers can be seen in Fig. 4. The three subfamily members LC3A, LC3B and LC3C were comparatively studied. Maximum decrease of Nile red fluorescence, expressed as percent ROIR, slope of ROIR decrease with LC3, and protein binding measured with ultracentrifugation varied in parallel for all three proteins, following the trend found in Fig. 3 for LC3C. At variance with ultracentrifugation, Nile red fluorescence does not provide absolute values of protein binding. However, the strictly parallel changes of ROIR and centrifugation data in Fig. 4 allowed us to estimate a conversion coefficient, of % ROIR decrease per 1% bound protein, of respectively 0.13 (LC3A), 0.12 (LC3B), and 0.23 (LC3C). Note that, for each new protein to be studied, a calibration of fluorescence vs. ultracentrifugation measurements would be required. Specifically, our method could be used in a rapid initial scanning of different proteins binding to different membranes. Once this semi-qualitative assessment had been performed would ultracentrifugation be used whenever a precise quantification was required.

Fig. 4
figure 4

A comparison of the Nile red fluorescence and the ultracentrifugation methods to estimate LC3 protein binding to lipid bilayers. Effects of CL and Cer. Data for LC3A, LC3B, and LC3 proteins are given. Top row, maximum observed values of % ROIR decrease. Middle row, slopes of the % ROIR vs. LC3 concentration curves. Bottom row, % bound protein, measured with an ultracentrifugation-flotation method10. Bilayer lipid compositions were: 1, eSM: DOPE (50:50); 2, eSM: DOPE: eCer (40:50:10); 3, eSM: DOPE: CL (33:33:33); 4, eSM: DOPE: CL: eCer (23:33:33:10). Average values ± S.D., n = 3. Statistical significance was estimated according to the Student’s t-test, p values are indicated for each pair of significantly different data: ** p < 0.01, *** p < 0.001.

Full size image

As shown above, the presence of Cer in CL-containing bilayers increased LC3 binding measured either by Nile red fluorescence, or by ultracentrifugation. In the above experiments, Cer was included in the lipid mixture in organic solvent, prior to LUV preparation. Alternatively, and considering that our bilayers contain eSM, Cer could be generated in situ through the action of a sphingomyelinase (SMase). This was achieved through addition of bacterial SMase to preformed LUV (devoid of Cer) ± LC3. The time courses of SMase-dependent Cer generation in LUV composed either of eSM: DOPE (50:50 mol ratio) or of eSM: DOPE: CL (at equimolar ratios), in the absence of LC3, are shown in Fig. 5. In the LC3 binding assays with SMase-generated Cer, the enzyme was allowed to act for 1 h before LC3 addition.

Fig. 5
figure 5

Time course of SM hydrolysis by SMase in bilayers (LUV). SMase activity is measured as Cer release, considering a 1:1:1 stoichiometry for SM, Cer and phosphorylcholine. Bilayer lipid compositions were: (A) eSM: DOPE (50:50); (B) eSM: DOPE: CL (33:33:33). Average values ± S.D., n = 3.

Full size image

Binding of LC3A, LC3B, and LC3C, with or without CL, containing Cer either added during bilayer preparation or in situ generated, and assayed either with the fluorescence or the ultracentrifugation method, is shown in Fig. 6. The results obtained with SMase were essentially identical to those in which Cer was added at the LUV preparation stage. This confirms that: (i) Nile red fluorescence parallels the binding results obtained with ultracentrifugation; (ii) Cer increases binding of LC3 proteins to lipid bilayers containing CL, and (iii) the same results are obtained with in situ generated Cer and with externally added lipid.

Fig. 6
figure 6

Effect of SM hydrolysis by SMase on LC3 protein binding to lipid bilayers. Data for LC3A, LC3B, and LC3 proteins are given. Top row, maximum observed values of % ROIR decrease. Middle row, slopes of the % ROIR vs. LC3 concentration curves. Bottom row, % bound protein, measured with an ultracentrifugation-flotation method10. Bilayer lipid compositions were: 1, eSM: DOPE (50:50); 2, eSM: DOPE: Cer (40:50:10); 3, eSM: DOPE (50:50) + SMase; 4, eSM: DOPE: CL (33:33:33); 5; eSM: DOPE: CL: Cer (23:33:33:10); 6, eSM: DOPE: CL (33:33:33) + SMase. SMase was allowed to act for 1 h, a time period in which released Cer was about 10% of the total lipid. Average values ± S.D., n = 3. Statistical significance was estimated according to the Student’s t-test, p values are indicated for each pair of significantly different data: ** p < 0.01, *** p < 0.001.

Full size image

Discussion

The present study proposes a simple and relatively low-cost method to assay the binding of amphipathic proteins to lipid bilayers. The results elicit two main questions, namely why/how Nile red ROIR responds to the presence of proteins in contact with the bilayer, and why/how CL is required for Cer to exert its enhancing activity in LC3 binding.

The data in Figs. 2, 3, 4 and 6 show a protein-dependent decrease in Nile red ROIR. Studies in the absence of protein15 have indicated that ROIR varies with the lipid phase, increasing from liquid-ordered to gel, to rippled, and to liquid-crystalline, i.e. increasing with decreasing bilayer molecular order. Proteins are considered to decrease lipid order in bilayers22,23, thus LC3 binding should lead to an increased ROIR, contrary to the experimental observations. However, Sot et al.15 had observed a non-linear correlation between Nile red ROIR and bilayer fluidity in experiments with cholesterol. ROIR values, presumed to decrease with decreasing bilayer fluidities, become actually lower in the presence of cholesterol. This would constitute an anomaly, because cholesterol, at the 30 mol% ratio used in15, is known to decrease the fluidity of bilayers in the Lα phase. Conversely, cholesterol increases the fluidity of bilayers in the Lβ or gel phase, inducing formation of the liquid-ordered Lo phase24,25. The cholesterol results in15 were interpreted as suggesting that the probe was actually reporting on its capacity to diffuse in the membrane matrix. In general, an increased fluidity will facilitate probe diffusion, and ROIR will also increase. However, the probe motion is restricted by the protein in the hydrophobic bilayer matrix, and ROIR decreases. L3C, when inserted into the bilayer, would increase the lateral surface pressure26. Note that Nile red is a very hydrophobic molecule (see Supplementary Chart 1), at variance with phospholipids or Cer, and it is unlikely to be oriented along the bilayer polarity axis. It would rather be found in the nonpolar membrane core, thus any protein- (or cholesterol-) caused local increase in molecular order would not be compensated by an increased lateral diffusion.

The concerted effect of CL and Cer in increasing LC3 binding to bilayers had been observed previously17. This situation might have a parallel in the outer mitochondrial membrane, in which CL externalization acts as an autophagic stimulus to promote damaged mitochondria recognition by LC3/GABARAP, thus reinforcing the potential role of ceramide in mitochondrial autophagy (mitophagy). Varela et al.17, using model lipid membranes, proposed an explanation for the positive cooperativity of both lipids in LC3 binding. Cer is known to cause a reorganization of the membrane lipids, due to its tendency to lateral segregation into Cer-enriched domains27. This causes in turn, due to the high affinity of SM for Cer28,29, the displacement of SM to rigid domains together with Cer. Varela et al.17,30 had previously observed that, even if Cer segregates laterally into very rigid domains, GABARAP (an LC3 ortholog) binds only the more fluid regions, suggesting that the enhancing role of ceramide is exerted by the minority of ceramide molecules dispersed in the fluid phase. Our hypothesis is that Cer would generate nanodomains dispersed within CL-enriched fluid areas, and the rigid-fluid interfaces would be preferential areas for LC3/GABARAP protein binding. The structural defects caused in bilayers by the coexistence of rigid and fluid domains are often preferred sites for protein insertion31,32. Such rigid nanodomains would also contribute to hindering Nile red motion, and decreasing ROIR.