GABARAP and GABARAPL1 showed the highest membrane tethering ability among the six homologs tested
In order to study the effect of Cer on LC3/GABARAP-induced vesicle tethering, the first step was to reconstitute protein lipidation (PE binding) with liposomes either lacking or containing Cer. Egg ceramide (eCer), consisting mainly of C16 Cer, was used as the standard Cer in our experiments. A high (50 mol %) PE concentration was used to favor the covalent binding of LC3/GABARAP to the liposomes. The liposome compositions used in our comparative studies were PC:PE (50:50) and PC:PE:Cer (40:50:10). Lipid compositions are given as mol ratios in all cases. Two different lipidation methods were used in order to test the potential role of Cer in LC3/GABARAP-induced vesicle tethering, namely the chemical and the enzymatic procedures.
Chemical lipidation approach. The chemical lipidation method was used as an initial approach to test the vesicle-tethering capacity of the six homologs. This method had been previously used and its simple design allowed exploring the behavior of the six LC3/GABARAP proteins [8, 11].
The chemical lipidation approach requires the reaction between a maleimide group and a Cys amino acid residue. With this purpose, the LC3/GABARAP proteins were modified so that they exposed a C-terminal Cys, instead of the native Gly. The following mutants were thus obtained: LC3A G120C, LC3B G120C, LC3C G126C, GABARAP G116C, GABARAPL1 G116C and GABARAPL2 G116C. As homologs LC3A and GABARAPL2 possess an additional (native) Cys residue in their sequence, the native Cys in those proteins was mutated to Ser in order to avoid unwanted Cys-maleimide interactions. Thus, LC3A C17S G120C and GABARAPL2 C15S G116C proteins were obtained. In addition, 60 mol% of the PE present in the vesicles was modified through binding a maleimide group (PEmal). This allows the protein C-terminal Cys to react with the maleimide moiety, and anchor the membrane. The compositions tested with the chemical approach were ePC:DOPE:PEmal (50:20:30) and ePC:DOPE:PEmal:eCer (40:20:30:10).
First, LC3/GABARAP protein lipidation was assayed (Fig. 1) to discern the Cer effect on the PE interaction with each of the six homologs. Proteins and liposomes were incubated at 37 °C and aliquots were taken for each protein at different times (0, 5, 10, 40 min).
Fig. 1. Ceramide enhances chemical lipidation of LC3/GABARAP proteins. Protein Cys-terminal lipidation measured at times 0, 5, 10, and 40 min. (a) SDS-PAGE gels representative for each protein and lipid composition. I, non-lipidated protein; II, lipidated protein. (b) Lipidation time courses. (c) Lipidation at time 40 min. 0.4 mM LUV composed of ePC:DOPE:PEmal (40:20:30) (gray) or ePC:DOPE:PEmal:eCer (40:20:30:10) (red) were mixed with 5 μM of the pertinent LC3/GABARAP protein. Liposomes were ≈100 nm in di ameter. Average data ± S.D., n = 3. **p < 0.01, *p < 0.05, ns: non-significant differences.
The percent PE-bound (lipidated) LC3/GABARAP can be estimated by densitometry as it migrates faster than the unbound protein in SDS-PAGE gels. In Fig. 1a,I and II denote respectively the non-lipidated and lipidated forms of the homologs. The percent lipidated protein in Fig. 1b,c is computed from the densitometric intensities of the bands.
All the proteins, except LC3B G120C and GABARAPL2 C15S G120C, showed a Cer-enhanced initial lipidation rate (Fig. 1b). In the absence of Cer, LC3A C17S G120C, GABARAP G116C and GABARAPL1 G116C showed lower lipidation levels than LC3C G126C (Fig. 1c, gray bars). In fact, LC3C G126C was the homolog reaching highest lipidation rates (Fig. 1b, LC3C G126C panel) and top levels after 40 min (Fig. 1c), >80% for both lipid compositions. Note that the basal lipidation levels were above 30% in those four cases. In addition, a significant protein lipidation increase was observed in the presence of Cer (Fig. 1c, red bars), particularly for GABARAP G116C (Fig. 1c). LC3B G120C and GABARAPL2 C15S G116C were the homologs reaching lowest lipidation levels, both below 30%, with either lipid composition (Fig. 1b,c). In summary, protein lipidation was clearly improved by Cer whenever Cer-free lipidation was above ≈30%.
Next, in order to study whether a Cer effect was also noticeable on the LC3/GABARAP-promoted vesicle tethering, liposomes of the desired lipid compositions were diluted in buffer and suspension turbidity measured as absorbance at 400 nm (A400). Once a stable baseline was observed for 5 min, the Cys-terminal LC3/GABARAP protein was added (Fig. 2). The Gly-terminal counterparts of the proteins were used as controls (Fig. 2, dashed lines).
Fig. 2. Ceramide enhances vesicle tethering promoted by chemically lipidated LC3/GABARAP proteins. Liposome-tethering activity assayed as an increase in absorbance at 400 nm (ΔA400). 0.4 mM LUV composed of ePC:DOPE:PEmal (40:20:30) (gray) or ePC:DOPE:PEmal:eCer (40:20:30:10) (red) were mixed with 5 μM of the pertinent LC3/GABARAP protein. The arrows indicate protein addition. Solid lines, Cys-terminal proteins: LC3A C17S G120C, LC3B G120C, LC3C G126C, GABARAP G116C, GABARAPL1 G116C and GABARAPL2 C15S G116C. Dashed lines, control experiments, Gly-terminal proteins: LC3A C17S, LC3B, LC3C, GABARAP, GABARAPL1 and GABARAPL2 C15S. Liposomes were ≈100 nm in diameter. Average data ± S.D., n = 3, n = 2 for the dashed control curves. The dashed lines are often indistinguishable from the X axis.
Experiments in Fig. 2 show that those proteins that reached lipidation levels below 30% (Fig. 1c), LC3B G120C and GABARAPL2 C15S G116C, were unable to induce vesicle tethering even 35 min after protein addition (Fig. 2, LC3B and GABARAPL2 panels). These results were indistinguishable from those with the corresponding Gly-terminal variants of the latter proteins, which could not be chemically lipidated (Fig. 2, dashed lines).
The other four proteins (LC3A C17S G120C, LC3C G126C, GABARAP G116C and GABARAPL1 G116C) caused liposome tethering only when the Cys-terminal forms were used, underlining the requirement of lipidation for vesicle tethering (Fig. 2, solid lines). GABARAP and GABARAPL1 were by far the two most efficient proteins in terms of vesicle tethering (Fig. 2). Furthermore, those four proteins showed higher tethering activities when Cer was part of the lipid composition (Fig. 2, solid red lines vs. solid gray lines).
From the results in Fig. 2, two parameters were computed, namely the maximum initial slope (Fig. 3a) (corresponding to the vesicle tethering rate) and the lag times (Fig. 3b), measured from the time of protein addition. They were not computed for LC3B G120C or GABARAPL2 C15S G116C, which failed to induce any vesicle tethering. The other four proteins showed increased tethering rates and shorter lag times with Cer-containing vesicles (Fig. 3).
Fig. 3. Ceramide increases rates and shortens lag times of vesicle tethering induced by chemically lipidated LC3/GABARAP. (a) Vesicle tethering rates. (b) Lag times. Data obtained with proteins susceptible to lipidation, as in Fig. 2. 0.4 mM LUV composed of ePC:DOPE:PEmal (40:20:30) (gray bars) or ePC:DOPE:PEmal:eCer (40:20:30:10) (red bars) were mixed with 5 μM protein. Liposomes were ≈100 nm in diameter. Representative curves are shown in Fig. 2. Average data ± S.D., n = 3. ***p< 0.001, **p < 0.01, *p < 0.05, ns: non-significant differences.
The proteins causing higher liposome aggregation, in terms of ΔA400 and tethering rates, were GABARAP G116C and GABARAPL1 G116C (Fig. 2 and Fig. 3a). With GABARAP G116C, in the presence of Cer, the aggregates seemed to exceed 400 nm in size, i.e. surpassing the Rayleigh limit, as the apparent absorbance decreased with time (Fig. 2, GABARAP).
LC3C G126C did not reach high ΔA400 values (Fig. 2), but showed almost no lag time with any of the lipid compositions (Fig. 3b). Correspondingly, it was also the protein showing fastest and most extensive lipidation (Fig. 1b,c).
Lowest vesicle tethering was induced by LC3A C17S G120C (Fig. 2 and Fig. 3a), which showed the longest lag times among the LC3/GABARAP proteins tested. Its low activity made the effect of Cer even more visible as the lag time was shortened from ≈17 to ≈10 min (Fig. 3b).
It has previously been suggested [9] that each LC3/GABARAP protein may need to reach a specific lipidation threshold value, different for each homolog, in order to induce vesicle tethering. Therefore, the absence of measurable liposome aggregation with LC3B G120C or GABARAPL2 C15S G116C (Fig. 2) could be explained if their lipidation thresholds were above 30%, i.e. above the values reached in the experiments (Fig. 1c).
Furthermore, although LC3A C17S G120C and GABARAPL1 G116C followed similar lipidation trends (Fig. 1), their ability to induce vesicle tethering varied, LC3A C17S G120C showing much less tethering than GABARAPL1 G116C (Fig. 2 and Fig. 3a). This reinforces the hypothesis of a lipidation threshold, which would be different for each of these proteins. According to the lag times (Fig. 3b), GABARAPL1 G116C would reach its threshold during the first minutes after protein addditon, while LC3A C17S G120C would take a longer time.
In turn, the Cer-related lag time shortening (Fig. 3b) could be due to the proteins reaching their lipidation threshold faster in the presence of Cer (Fig. 1b). In fact, there was an increased vesicle tethering rate with Cer (Fig. 3a) that correlated with shorter lag times (Fig. 3b). Note that Cer did not appreciably decrease the lag time of LC3C G126C-induced vesicle aggregation, probably because this time was already very short in the absence of Cer.
Overall, these results imply that, even if individual variations occur among the different homologs, when Cer is part of the lipid composition these proteins do not only need less time to reach the onset of vesicle tethering, but they also do so at a faster rate once vesicle aggregation starts. In summary, Cer appears to potentiate LC3/GABARAP ability to promote vesicle tethering.
Enzymatic lipidation approach. The enzymatic lipidation procedure requires a LC3/GABARAP protein with Gly at its C terminus, ATG7, ATG3, MgCl2 and ATP. The lipid compositions used in this approach were ePC:DOPE (50:50) and ePC:DOPE:eCer (40:50:10).
Fig. 4. Ceramide does not enhance enzymatic lipidation of LC3A, GABARAP or GABARAPL1. Lipidation measured at 0, 10, 30 and 40 min after addition of 0.4 mM LUV, 0.5 μM ATG7, 2 μM ATG3, 1 mM MgCl2, 5 μM LC3/GABARAP, 5 mM ATP. (a) Representative SDS-PAGE gels for each protein and lipid composition. -Cer panels: ePC:DOPE (50:50); +Cer panels: ePC:DOPE:eCer (40:50:10). I, non-lipidated protein; II, lipidated protein. (b) Time course of protein lipidation. Data from experiments in Fig. 4a. (c) Percent protein lipidation at time 40 min. Data from experiments in Fig. 4b. For b andc c, LUV were composed of ePC:DOPE (50:50) (gray) or ePC:DOPE:eCer (40:50:10) (red). Liposomes were ≈80 nm in diameter. Average data ± S.D., n = 3. ns: non-significant differences.
Just three of the six homologs, LC3A, GABARAP and GABARAPL1 were chosen for experiments following the enzymatic lipidation approach. GABARAP and GABARAPL1 were selected as the two proteins showing highest tethering effect with the chemical approach (Fig. 3a). LC3C was not included in this part of the study because previous experiments from this laboratory [11] indicated that its behavior, and probably its in vivo function too, differed from the other homologs, thus making it worthy of a separate investigation [28]. LC3A was chosen as a representative of the LC3 subfamily.
Lipidation levels of the three proteins (LC3A, GABARAP and GABARAPL1) were measured for 40 min. In all cases, lipidation values were much lower (<20%) than those found with the chemical approach (Fig. 1 vs. Fig. 4). A slight tendency towards a higher lipidation with Cer was observed with GABARAP and GABARAPL1, but differences measured at the end of the experiment were not significant (Fig. 4b,c).
Fig. 5. Ceramide increases the rate of liposome tethering induced by enzyme-lipidated GABARAP or GABARAPL1. Liposome tethering was assayed as ΔA400. 0.4 mM LUV composed of ePC:DOPE (50:50) (gray) or ePC:DOPE:eCer (40:50:10) (red) were mixed with 0.5 μM ATG7, 2 μM ATG3, 1 mM MgCl2, 5 μM LC3/GABARAP. Liposomes were ≈80 nm in diameter. (a) Vesicle tethering representative time courses. Arrows indicate 5 mM ATP (solid lines) or buffer (dashed lines) addition. (b) Vesicle tethering rates. Average data ± S.D., n = 3. **p < 0.01, *p < 0.05, ns: non-significant differences.
Next, the vesicle tethering ability of each protein was tested (Fig. 5). In all cases, ATP addition had a positive effect: all three proteins were able to induce some vesicle aggregation (Fig. 5a, solid vs. dashed lines). GABARAP and GABARAPL1 exhibited a similar behavior: Cer almost doubled their vesicle tethering rates, GABARAPL1 reaching a higher rate (Fig. 5b). As expected, LC3A showed the smallest effect, with lower tethering rates and no Cer effects (Fig. 5b). This was expected as almost no lipidation was detected for this protein (Fig. 4c) and, according to the results obtained with the chemical approach, the lipidation threshold seemed to be higher for this protein.
In summary, even if the enzymatic lipidation approach allowed a lower Cer effect on lipidation (Fig. 4c), the increasing effect on tethering rates was noticeable and significant for GABARAP and GABARAPL1 (Fig. 5b), prompting the study of Cer effects on their lipid mixing ability.
Ceramide increased intervesicular lipid mixing induced by GABARAP and GABARAPL1
Liposome tethering is the initial step in the hemifusion or fusion between two vesicles. Any fusion event involves also lipid mixing between the fusing vesicles. Therefore, to test whether LC3/GABARAP were able to induce any kind of vesicle fusion, total lipid mixing (TLM) assays were performed (Fig. 6). For this approach, a vesicle population doped with the fluorophores NBD-PE and Rho-PE was mixed to a 1:9 ratio with unlabeled liposomes of the same lipid composition. NBD-to-Rho Förster energy transfer was measured.
Fig. 6. Both GABARAP and GABARAPL1 induce extensive lipid mixing of ceramide-containing vesicles. Total lipid mixing (TLM) induced by the lipidated proteins, monitored with the NBD-PE/Rho-PE lipid dilution assay. 0.4 mM LUV composed of ePC:DOPE (50:50) (gray) or ePC:DOPE:eCer (40:50:10) (red) were mixed with 0.5 μM ATG7, 2 μM ATG3, 1 mM MgCl2, 5 μM LC3/GABARAP protein. Liposomes were ≈80 nm in diameter. (a) Total lipid mixing, representative time courses. The arrows indicate 5 mM ATP (solid lines) or buffer (dashed lines) addition. (b) Total lipid mixing rates: -ATP (striped bars) or +ATP (solid bars). (c) Total lipid mixing lag times after ATP addition. (b,c) Data from experiments as in (a). Average data ± S.D., n = 3. ***p < 0.001, **p < 0.01, *p < 0.05, ns: non-significant differences.
The results showed that GABARAP and GABARAPL1 caused ATP-dependent, extensive lipid mixing only when the lipid composition included Cer (Fig. 6a, solid gray vs. red lines). Cer effect could be seen also in terms of a higher maximum initial slope (Fig. 6b, solid red bars) and of shorter lag times (Fig. 6c, solid red bars).
Even if some vesicle tethering was observed with LC3A (Fig. 5a), it was not associated to TLM under the assay conditions (Fig. 6a). No lag time could be detected either. With either lipid composition, the time courses after adding ATP overlapped with their respective ATP-free (non-lipidated) controls (Fig. 6a, dashed vs. solid red lines and Fig. 6b, stripped vs. solid red bars). This could indicate either that higher levels of lipidation are needed for LC3A or that this protein is unable to promote intervesicular lipid mixing.
Taken together, these results implied that the main Cer effect was to enhance lipid mixing, even without large increases in protein lipidation or vesicle tethering. Only when Cer was present, and the proteins could be lipidated (+ATP), were they able to induce vesicle lipid mixing (Fig. 6).
Ceramide enhanced GABARAP- and GABARAPL1-induced leakage-free inner-monolayer lipid mixing
In full vesicle fusion, lipids from both monolayers of each membrane become mixed; in hemifusion, only the outer monolayer lipids exchange, while the inner monolayers do not come into contact. Since total lipid mixing assays (TLM) cannot differentiate between those two events, specific inner-monolayer lipid mixing assays (ILM) were performed, in combination with the above-mentioned TLM (Fig. 6), to distinguish which of the two events were taking place (Fig. 7).
Fig. 7. Both GABARAP and GABARAPL1 induce fusion in a fraction of the vesicle population. Total (TLM) and inner (ILM) lipid mixing induced by the lipidated proteins, monitored with NBD-PE/Rho-PE lipid dilution assays. 0.4 mM LUV composed of ePC:DOPE (50:50) (gray) or ePC:DOPE:eCer (40:50:10) (red) were mixed with 0.5 μM ATG7, 2 μM ATG3, 1 mM MgCl2, 5 μM LC3/GABARAP protein. Liposomes were ≈80 nm in diameter. Light colors ILM, dark colors TLM. (a) Representative time courses. The arrows indicate 5 mM ATP addition. (b) Total and inner lipid mixing rates. (c) Percent total and inner lipid mixing at 40 min. (b,c) Data from experiments as in (a). Average data ± S.D., n = 3, **p < 0.01, *p < 0.05.
With ePC:DOPE vesicles, neither GABARAP nor GABARAPL1 induced any marked ILM, the effect of GABARAP being virtually undetectable, and GABARAPL1 reaching only ≈10% after 35 min (Fig. 7a, light gray lines and bars). Both proteins were able to induce some ILM when Cer was part of the lipid mixture, although mixing never went beyond 30% (Fig. 7b,c). Therefore, GABARAP and GABARAPL1 seemed to have the capacity of inducing full fusion, but not of all the liposomes present. GABARAPL1 promoted faster and more extensive fusion than GABARAP.
The low fusogenic activity of LC3A, already hinted at in the previous sections (Fig. 5 and Fig. 6), was confirmed by the results in Fig. 7: with ePC:DOPE (50:50) no ILM was detected, and it barely reached 5% with ePC:DOPE:eCer after 35 min.
An additional parameter to characterize the fusogenic effects of LC3/GABARAP is the mixing of vesicle aqueous contents (ACM). The assay involves testing whether vesicles containing a water-soluble fluorophore (ANTS) fuse with liposomes containing its quencher (DPX). Previously, it is essential to check that the decrease in ANTS fluorescence is not due to an increased membrane permeability (and to the subsequent contents leakage). Fig. 8a shows that, neither in the absence nor in the presence of Cer, did any of the three LC3/GABARAP proteins induce a sizable vesicle permeabilization.
The lack of aqueous contents leakage allowed for reliable ACM assays in which LUV, containing ANTS or DPX, were mixed at a 1:1 ratio. Results in Fig. 8b-d supported what was previously observed in ILM assays (Fig. 7). With GABARAP or GABARAPL1 and ePC:DOPE vesicles, no ACM differences were observed in the absence or presence of lipidated proteins (Fig. 8b, dashed vs. solid gray lines; Fig. 8c,d, striped vs. solid gray bars). With liposomes composed of ePC:DOPE:eCer, ATP addition to GABARAP and GABARAPL1 induced a measurable ACM, significantly different from the controls (i.e. non-lipidated proteins) (Fig. 8b, dashed vs. solid red lines; Fig. 8c,d, stripped vs. solid red bars). In the case of LC3A, no significant differences were seen in any case. Therefore, these results reinforce the role of both Cer and the lipidation of GABARAP and GABARAPL1 (+ATP) to enhance vesicle fusion (Fig. 8).
Fig. 8. Both GABARAP and GABARAPL1 induce leakage-free intervesicular aqueous contents mixing. (a) Vesicle contents leakage induced by the lipidated proteins, monitored with an ANTS/DPX mixing assay. (b) Aqueous contents mixing (ACM) monitored with an ANTS/DPX mixing assay. (a,b) Representative time courses. The arrows indicate 5 mM ATP (solid lines) or buffer (dashed lines) addition. (c) Aqueous contents mixing rates. (d) Percent aqueous contents mixing at 40 min. (c,d) Data from experiments in (b). -ATP (striped bars) or +ATP (solid bars). Average data ± S.D., n = 3. ***p < 0.001, **p < 0.01, *p < 0.05, ns: non-significant differences. 0.4 mM LUV composed of ePC:DOPE (50:50) (gray) or ePC:DOPE:eCer (40:50:10) (red) were mixed with 0.5 μM ATG7, 2 μM ATG3, 1 mM MgCl2, 5 μM LC3/GABARAP protein. Liposomes were ≈80 nm in diameter. Contents leakage and mixing were assayed as detailed under Methods.
Taken together, these results support the notion that Cer promotes the fusion-inducing capacity of LC3/GABARAP proteins. Both the ILM and the ACM assays showed that GABARAP and GABARAPL1 proteins seem to have the capacity to induce vesicle hemifusion and that full fusion only happens in a fraction of the vesicles present in the assay mixture.