The wild-type C. elegans Bristol strain N2 was used for some experiments and cultivated as previously described (Brenner, 1974). Also used or generated in this study were the following mutant strains: CB406: unc-57(e406)I, DH1230: chc-1(b1025ts)III, RB700: erp-1(ok462)X, ZX2088: erp-1(ok462)X; unc-57(e406)I. The following transgenic strains were generated and/or used: ZX460: wild type; zxIs6[punc-17::ChR2(H134R)::YFP; lin-15+]V (Liewald et al., 2008), ZX634: unc-57(e406)I; zxIs6[punc-17::ChR2(H134R)::YFP; lin-15+]V, ZX1602: erp-1(ok462)X; zxIs6[punc-17::ChR2(H134R)::YFP; lin-15+]V, ZX1778: chc-1(b1025ts)III; zxIs6[punc-17::ChR2(H134R)::YFP; lin-15+]V, ZX2227: unc-57(e406)I; erp-1(ok462)X; zxIs6[punc-17::ChR2(H134R)::YFP; lin-15+]V, ZX2342: chc-1(b1025)III; unc-57(e406)I; zxIs6[punc-17::ChR2(H134R)::YFP; lin-15+]V, ZX2343: chc-1(b1025)III; erp-1(ok462)X; zxIs6[punc-17::ChR2(H134R)::YFP; lin-15+]V, BC12478: dpy-5(e907)I; sIs10597[rCes F35A5.8a::GFP + pCeh361] (McKay et al., 2003).

Expression of perp-1::GFP (strain BC12478) and punc-17::ChR2(H134R)::YFP (strain ZX460) were analyzed on a Zeiss Axiovert 200 microscope (Zeiss, Germany), with a 40×/0.25 Zeiss oil objective and GFP filter set (Ex 470/40 nm, Em 525/50 nm). Micrographs were taken with an ORCA-Flash4.0 Digital sCMOS Camera (Hamamatsu Photonics, Japan). Animals were transferred onto 10% agarose pads in M9 buffer (K₂PO₄ 20 mM; Na₂HPO₄ 40 mM; NaCl 80 mM; MgSO₄ 1 mM) and immobilized with Polybead polystyrene 0.1 mm microspheres (Polysciences Inc., Warrington, PA, USA).

Transgenic worms were cultivated in the dark at 20°C on nematode growth medium (NGM) dishes with OP50 bacteria (Brenner, 1974) without or with all-trans retinal (ATR; Liewald et al., 2008). Dishes containing ATR were prepared by spreading 320 μl of OP50 culture mixed with 0.64 μl of 100 mM ATR stock (dissolved in ethanol) onto 5.5-cm dishes containing 8.2 ml of NGM. About 18 h before experiments, L4 larvae, grown on ATR plates, were placed on fresh ATR plates. Worms were illuminated with blue light (contraction-assay: 1.4 mW/mm², swimming-assay: 0.53 mW/mm²) from a 50 W mercury lamp, filtered through a GFP excitation filter (450–490 nm), on 5.5 cm diameter dishes, under a 10× objective in a Zeiss Axiovert40 microscope (Zeiss, Germany). Duration of illumination was defined by a computer-controlled shutter (Sutter Instruments). Worms were filmed with a Powershot G9 digital camera (Canon, Japan) at 640 × 480 resolution with 30 fps. Body length was determined as previously described (Erbguth et al., 2012). The length values were normalized to the averaged values measured before illumination (0–4 s), normalization was carried out for each worm. All the values below 80% were excluded and the length-profiles were averaged for each strain. The experiments were repeated on 2–3 different days (worms were picked from different populations on different days); the final graphs show the average of all individual animals. For the heat-shock (hs) experiments worms were cultivated at 15°C. The length of the hs was 14 min at 30°C, worms were picked to pre-warmed (at least 3 h in 30°C) plates. Before recording the contraction assay videos, worms were incubated at RT for 4 min. For further analysis of the loss of contraction during long-term stimulation in SV recycling mutants, the body length was additionally normalized to the time period of 6–10 s during stimulation, and the data was compared during seconds 5–60 of the illumination period. Mean contraction was filtered with a sliding average window of 3 s, and then a linear fit with a forced interception at y = 1 was performed, to allow comparing the slope of the loss of contractions (if any).

For analyzing swimming behavior, worms were placed into 96-well plates containing 100 ml NGM and 100 ml of M9 buffer per well. Worms were left in the M9 buffer for 15 min before the first trial of the swimming-assay, then they were filmed for 60 s (swimming-assay before light stimulation). The second round (60 s) of the swimming-assay was filmed after a 90 s light-stimulation and an additional 90 s incubation time. The swimming cycles (the worm’s body bends twice per cycle), were counted manually. The swimming assays were repeated on three different days, with 8–10 worms/group, picked from different populations on different days.

Worms were cultivated at 20°C on NGM dishes with OP50 bacteria. To study aldicarb sensitivity, 1.5 mM aldicarb dishes were prepared (Mahoney et al., 2006). After transferring the animals (15–20 young adults/trial, in total three trials on three different days, with worms picked from different populations) to the dishes, they were scored every 30 min by three gentle touches with a hair pick. The assays were performed blinded and on the same day with the same batch of aldicarb dishes.

For recordings from BWMs animals were immobilized with Histoacryl glue (B. Braun Surgical, Spain) and a lateral incision was made to access neuromuscular junctions (NMJs) along the ventral nerve cord. The basement membrane overlying muscles was enzymatically removed by incubation in 0.5 mg/ml collagenase for 10 s (C5138, Sigma-Aldrich, Germany). Muscles were patch-clamped in whole-cell mode at 22°C using an EPC10 amplifier with head stage connected to a standard HEKA pipette holder for fire-polished borosilicate pipettes (1B100F-4, Worcester Polytechnic Institute, Worcester, MA, USA) of 4–7 MΩ resistance. The bath solution contained: NaCl 150 mM; KCl 5 mM; CaCl₂ 5 mM; MgCl₂ 1 mM; glucose 10 mM; sucrose 5 mM; HEPES 15 mM, pH 7.3 with NaOH, ~330 mOsm. The pipette solution contained potassium-gluconate 115 mM; KCl 25 mM; CaCl₂ 0.1 mM; MgCl₂ 5 mM; BAPTA 1 mM; Na₂ATP 5 mM; Na₂GTP 0.5 mM; cAMP 0.5 mM; cGMP 0.5 mM; HEPES 10 mM, pH 7.2 with 1 M KOH, ~320 mOsm. Recordings were conducted at a holding potential of −60 mV. Light activation was performed using a LED lamp at 470 nm (KSL-70, Rapp OptoElectronic, Germany; 8 mW/mm²) and controlled by Patchmaster software (HEKA, Germany). Data were analyzed by Patchmaster (HEKA, Germany) and MiniAnalysis (Synaptosoft, USA) software.

Transgenic L4 worms were transferred from regular NGM dishes to freshly seeded E. coli OP50 −/+ (0.1 mM) ATR dishes 1 to 2 days before high-pressure freezing (HPF). Young adult animals were used for HPF fixation, based on methods previously described (Rostaing et al., 2004; Weimer, 2006; Kittelmann et al., 2013). Briefly, about 10–40 worms were loaded into a 100 μm deep aluminum planchette (Microscopy Services) filled with E. coli OP50 −/+ ATR, covered with a 0.16 mm sapphire disc and a 0.4 mm spacer ring (Engineering office M. Wohlwend GmbH) for subsequent photostimulation. To prevent preactivation, all manipulations were done under red light. For light stimulation experiments, worms were continuously illuminated with a laser (~20 mW/mm²) for 30 s, followed by HPF at −180°C under 2100 bar pressure in a Bal-Tec HPM010 HPF machine. A ~6 s period of manual insertion of the sample after photostimulation is required with this high pressure freezer. For experiments involving the clathrin mutant, animals were kept at permissive temperature (15°C) at all times. Tools used were equilibrated to the respective temperatures. Worms were either loaded at room temperature to the planchette for freezing directly or incubated at non-permissive temperature at 30°C for 14 min before freezing.

After freezing, specimens were transferred under liquid nitrogen into a Reichert AFS machine (Leica) for freeze substitution. Tannic acid (0.1% in dry acetone) fixative was used to incubate samples at −90°C for 100 h. Then, a process of washing was performed, followed by an incubation of 2% OsO₄ for 39.5 h (in dry acetone) while slowly increasing the temperature up to room temperature. Afterwards, the process of embedding in Epoxy resin (Agar Scientific, AGAR 100 Premix kit hard) was executed with increasing concentration from 50% to 100% at room temperature and 100% at 60°C over 48 h.

For electron micrographs, cross sections were cut at a thickness of 40 nm, transferred to formvar-covered copper slot grids and counterstained in 2.5% aqueous uranyl acetate for 4 min, followed by washing with distilled water. Then, grids were carried onto Reynolds lead citrate solution for 2 min in a CO₂-free chamber and subsequently washed in distilled water again. Images of regions in the ventral nerve cord were taken with an Erlangshen ES500W CCD camera (Gatan) in a Philips CM12 transmission electron microscope operated at 80 kV. Images were scored blind for each condition and tagged in ImageJ (NIH) for the PM (synapse perimeter). ImageJ ROIs were stored and then quantified based on methods described previously (Steuer Costa et al., 2017). The diameters of synapses from each stimulation condition vary due to the different extent of SV exocytosis or because different synapses were sampled. Thus, each value for the number of docked SVs was normalized and represents the number of docked SVs along a membrane whose perimeter is 1548 nm in a profile; the other organelles are represented as the numbers of SVs or LVs in a typical synaptic profile of 164,100 nm² (Steuer Costa et al., 2017). SV size was scored from 3 to 10 middle images of randomly selected synapses or from randomly selected single images per synapse for each mutant or stimulation protocol. Details (2–3 worms, and typically 10–19 synapses were analyzed per genotype and condition; 1–2 technical replicates were performed) are given in the respective figure legends, and in Supplementary Table S1.

Clustal X2 (Larkin et al., 2007) was used for alignment of endophilin sequences. Alignments were colored using Boxshade¹. The phylogenetic tree was drawn using Phylodendron².

The respective statistics used are indicated for each experiment in the figure legends.

In C. elegans, endophilin A and B are encoded by unc-57 and erp-1, respectively. They show 38.7% (UNC-57 isoform a) and 40.2% (ERP-1 isoform a) identity to their human homologs (Supplementary Figure S1). While the role of UNC-57 in SV recycling was studied earlier and its main sites of action appear to be endocytosis as well as the budding of new SVs from the synaptic endosome (Schuske et al., 2003; Kittelmann et al., 2013), the role of ERP-1 is less clear (Wabnig et al., 2015). For example, could ERP-1 be required for formation of bulk endocytosis structures (Figure 1)? We had previously observed in optogenetic stimulation experiments that erp-1 mutants showed a slight reduction in photoevoked ePSCs (Wabnig et al., 2015). However, no long-term behavioral assays were performed, to assess SV recycling defects under more robust neuronal stimulation conditions. Also, no detailed expression pattern analysis of erp-1 in motor neurons was reported so far, and we wanted to confirm that phenotypes of erp-1 mutants on cholinergic transmission are caused directly in the cholinergic cells and not indirectly in other tissues missing ERP-1.

We used an erp-1 promoter GFP transcriptional fusion and compared the expression pattern to that of punc-17::ChR2::YFP as a marker for cholinergic cells (Figure 2). GFP expression was visible in many tissues, as reported in an earlier high-throughput study (Hunt-Newbury et al., 2007); we focused on the ventral nerve cord motor neuron cell bodies. This part of the nervous system contains cholinergic and GABAergic motor neurons, that can be identified based on anatomical features (position relative to the vulva, left or right commissural processes to the dorsal side, etc.). When compared to the punc-17 reporter, it was obvious that ERP-1 is expressed in many more neurons in the ventral nerve cord than just the cholinergic motor neurons (i.e., in >63 visible neurons, vs. only 43 neurons in the punc-17 reporter strain). This suggests that ERP-1 is present in GABAergic neurons as well (as these are the only other cell type present in this part of the nervous system). We tentatively assigned the individual cell bodies (Figure 2; Supplementary Figure S2), verifying this notion. Thus, the lack of erp-1 should lead to direct effects in cholinergic cells, particularly in single-cell assays, such as in electrophysiological recordings and EM analyses.

To probe for ERP-1 function in cholinergic cells, we first performed a simple pharmacological assay for malfunction of cholinergic transmission. Aldicarb is an ACh esterase inhibitor that leads to accumulation of ACh in the synaptic cleft and to progressive paralysis of the animals. In mutants with reduced cholinergic transmission, paralysis is delayed or does not occur (Miller et al., 1996). Mutants lacking erp-1 showed a mild ric (resistant to inhibitors of choline esterase) phenotype, while mutants in unc-57, and erp-1; unc-57 double mutants showed very strong ric phenotypes and did not paralyze at all during 5 h of exposure to aldicarb (Figure 3A), just as the positive control unc-10 (encoding rab3-interacting molecule—RIM). This demonstrates that erp-1 mutants are somewhat affected for cholinergic transmission, however, much less than unc-57 endophilin A mutants.

Another behavior that can report on cholinergic function at the NMJ is the swimming locomotion of C. elegans. Nematodes exhibit rapid swimming behavior in liquid, and bend their body up to 110 times min⁻¹, while cholinergic mutants show lower swimming rate or are not swimming at all, depending on the severity of the mutation. Under standard conditions, mutants lacking erp-1 showed a slight, but significant reduction of swimming locomotion compared to wild-type animals (Figure 3B). We wondered, if under conditions of increased cholinergic transmission, a possible SV recycling defect may affect swimming more pronouncedly. Thus, we used animals expressing channelrhodopsin-2 (ChR2) in cholinergic cells (transgene zxIs6, Liewald et al., 2008), and assessed them in swimming assays before 90 s continuous photostimulation of the cholinergic neurons, as well as after a recovery period of 90 s in a second swimming assay. While wild-type animals did not show a significant reduction of swimming rate in the second assay, regardless if the animals were cultivated in the presence or absence of the ChR2 co-factor ATR; (i.e., leading to functional or non-functional ChR2, respectively), erp-1 mutants raised with ATR showed significantly less swimming cycles after the light stimulus. Thus, erp-1 mutants show fatigue, likely due to reduced cholinergic transmission, upon light-evoked “exercise”.

To further examine this, we used an assay that allows to (indirectly) measure the rate of cholinergic transmission in live animals while they are undergoing strong neuronal activity. Here, animals expressing transgene zxIs6 are assessed for the extent of the evoked muscle contraction by video microscopy. ChR2 photoactivation leads to spastic paralysis due to simultaneous activation of all body wall muscle cells, and animals show a (longitudinal) contraction of the body (Liewald et al., 2008). Animals with SV recycling defects cannot sustain the contraction for prolonged stimulation periods. When we photostimulated wild-type animals, raised in the presence of ATR, for 2 min, they contracted by about 6% of the body length, until the end of the photostimulus (Figures 4A,B). In contrast, erp-1 mutants initially contracted also by about 6%, but then progressively reduced the contraction to only 3% at the end of the stimulus period (Figures 4A,B). This difference to wild type was statistically significant. Animals raised in the absence of ATR did not show contractions, as expected. Thus, erp-1 mutants exhibit typical phenotypes of a SV recycling defect, even though this was less prominent than for endophilin unc-57 mutants, which lost the contraction almost completely already after 30 s of stimulation (our previous report; Kittelmann et al., 2013). We wondered if erp-1 and unc-57, based on this assay, have redundant functions or could be part of independent pathways, which could indicate different sites of action of the two proteins. Thus, we analyzed erp-1; unc-57 double mutants and compared them to either single mutant (Figure 4C; see Supplementary Figure S3A for statistical analysis, and for analysis of the number of measurable animals during the time course of the assay, when many animals show increased coiling; Liewald et al., 2008). However, the unc-57 mutation dominated the phenotype, and erp-1; unc-57 double mutants did not show an exacerbated phenotype; if at all, they even showed a less prominent reduction of the contraction than unc-57 single mutants (see also Figure 9; Supplementary Figure S6). Possibly, complete absence of endophilin function may have uncovered a “backup” mechanism that allows for SV recycling in the absence of these proteins.

To further investigate this, we used electrophysiology to measure spontaneous and optogenetically evoked transmission at the NMJ (excitatory post-synaptic currents—ePSCs). Also these assays can uncover SV recycling defects (Liewald et al., 2008). Optogenetic stimulation of the cholinergic neurons at 0.5 Hz, which evokes strong photo-ePSCs in wild type, caused significantly smaller currents in erp-1 mutants (ca. 2/3 of wild type; Figures 5A,B). In unc-57 mutants, photo-ePSCs were ~15% of wild-type photo-ePSCs for the first stimulus, and then almost immeasurable for further stimuli. unc-57; erp-1 double mutants had essentially the same phenotype as unc-57 single mutants. However, these strong phenotypes were particularly evident for stimulated transmission, while for spontaneous miniature post-synaptic currents (mPSCs), only unc-57 mutants showed slightly reduced amplitudes. In contrast, both unc-57 and unc-57; erp-1 double mutants showed significantly reduced mPSC frequency, as compared to wild type. This is likely a consequence of reduced SV numbers. Thus, endophilin A/UNC-57 appears to have much more pronounced functions in SV recycling and maintenance of synaptic transmission than endophilin B/ERP-1.

To directly analyze the site of action of ERP-1, and the morphology of the synapse under resting and stimulated conditions, we turned to ultrastructural analyses of thin sections of cholinergic en passant synapses (flash’n’freeze EM; Kittelmann et al., 2013; Watanabe et al., 2013a). Briefly, animals expressing ChR2 in cholinergic neurons were photostimulated for 30 s with blue light, and then transferred into the freezing chamber of a high pressure freezer (this process requires 5–6 s), which can cryo-fix the tissue within ms. This freezes all morphological changes and features in native conditions. Following a freeze-substitution and heavy metal staining protocol, the synaptic ultrastructure is then analyzed in 40 nm thin sections by transmission EM (TEM; Figure 6A). Photostimulation of zxIs6 synapses in animals raised in the absence of ATR did not evoke any obvious effects in the synapse, while the stimulation of functional ChR2 (animals raised with ATR) led to drastic alterations in the synaptic profile: first, the synaptic profile area was significantly increased, which we interpret as a result of excessive SV fusion that adds membrane to the PM of the en passant synaptic bouton (Figure 6B). Consequently, the total number of SVs was reduced for wild-type animals (Figure 6C), as was the number of docked SVs, i.e., SVs that are in zero visible distance to the PM and thought to be ready for fusion (i.e., primed SVs; Figure 6D). In addition, large, clear-core vesicles (LVs), end-products of bulk/ultrafast endocytosis, were observed at elevated numbers in stimulated synapses (Figure 6E), though this did not meet significance under conditions of multiple-comparisons statistics. We compared wild type to erp-1 mutants. These had enlarged synaptic profile areas, even without stimulation, possibly due to inefficient SV recycling; however, they exhibited a reduced total number of SVs even before stimulation. This was no further reduced by light stimulation, indicating that the readily releasable pool (RRP) and/or RP are reduced in these animals, and that a lower turnover of SVs occurs, in line with our behavioral and electrophysiological assays. erp-1 mutants also showed reduced docked SV numbers, that were not further reduced upon photostimulation. LV numbers in erp-1 were increased upon stimulation, just as for wild type. Next, we analyzed endophilin A/UNC-57 mutants: unc-57(e406) synapses had significantly reduced numbers of total and docked SVs (~30% and <15% of wild type, respectively), that were further depleted upon stimulation, indicating a strong defect in SV formation and/or recycling, consistent with the electrophysiological results (Figure 5). Upon stimulation, unc-57 mutants showed a strong increase (~4 times more than wild type) in the number of LVs per synaptic profile (this was unexpected given our previous findings; Kittelmann et al., 2013), where unc-57 mutants had less, but larger LVs than wild type; the difference is due to different analyses: previously, we counted LVs per synapse, not per thin section profile; if we do the same here, we see no major difference to wild type, i.e., 2.2 vs. 2 LVs per synapse, respectively. Consistently, LV size was increased in unc-57 mutants (Figure 6H). Thus, UNC-57/endophilin A is not essential for formation of LV/endosomes upon synaptic stimulation, yet these LVs cannot be broken down into new SVs. ERP-1 also contributes to formation of SVs, as no LVs accumulate in its absence; yet, its site of action remains unclear.

To investigate this, we also analyzed unc-57; erp-1 double mutants. These animals showed significantly increased synapse cross-sections after 30 s photostimulation (Figure 6B), indicating that some SV fusion took place, despite the very low number of fusion events evident from electrophysiological analysis of these mutants (Figure 5), and that an impairment in endocytosis may occur. Just as unc-57 mutants, unc-57; erp-1 double mutants had extremely reduced total SV and docked SV numbers (~20% and ~10% of wild type, respectively). In contrast to wild type and also to unc-57 single mutants, the double mutants showed highly increased numbers of LVs (~350% of wild type), already before stimulation. This indicates major difficulties of these synapses to break down the end-products of bulk endocytosis; yet, endocytosis per se does occur. LVs in erp-1; unc-57 double mutants were significantly larger than in wild type or erp-1 single mutants (Figure 6H). After stimulation, the numbers of LVs, surprisingly, were reduced compared to the non-stimulated condition, and were comparable to stimulated wild type or erp-1 synapses. This was in strong contrast to unc-57 mutants. This indicates that in the absence of all endophilins, (activity-, possibly Ca²⁺-dependent) mechanisms exist that enable some formation and recycling of LVs, such as to sustain a low level of SV exocytosis and recycling.

We also analyzed the diameter of SVs, to assess whether the regulation of the organelle characteristics is affected. SV diameters did not exhibit any obvious difference in the mean, or in the diameter distribution in wild type, erp-1 and unc-57; erp-1 double mutants, neither under resting or stimulation conditions (Figures 6F,G; Supplementary Figure S4).

We next asked at which stage during SV recycling clathrin function is required. Clathrin was shown to affect SV size in C. elegans (SVs became smaller in the absence of functional clathrin heavy chain CHC-1; Sato et al., 2009), and in mammalian synapses, clathrin was demonstrated to act at the synaptic endosome (Watanabe et al., 2014). As clathrin heavy chain (encoded by chc-1) is essential for survival, we could assess only a chc-1(b1025ts) temperature sensitive allele: these animals are comparably unaffected at the permissive temperature (15°C), but are deprived of CHC-1 function at elevated temperatures (30°C). First, we performed a time-dependent optogenetic contraction assay, comparing wild type to chc-1 mutants at elevated temperature. After the shift to 30°C, the animals showed a (mild) decline in contractions over the 1 min photostimulation period, that was significantly different from wild type and from the initial contraction, while wild-type animals showed comparable and consistent contractions both at 15°C (no hs) and at 30°C (with hs; Figures 7A,B, 7A,B, 9A; Supplementary Figure S5). Thus, chc-1(b1025ts) mutants showed a mild SV recycling defect at non-permissive temperature.

Next, we analyzed the synaptic ultrastructure of chc-1 mutants, at permissive or non-permissive temperatures (Figure 8A). The synaptic profile area of the analyzed synapses, regardless of wild type or chc-1 mutation, and regardless of permissive or non-permissive temperature, increased significantly during photostimulation (Figure 8B). This indicates that CHC-1 is not required for SV fusion. When we analyzed LV numbers, chc-1 mutant synapses showed significantly more LVs than wild-type synapses at the non-permissive temperature (Figure 8C). This was already the case without photostimulation. Thus, CHC-1 is required for SV recycling at the stage of the synaptic endosome, and not at the stage of PM-localized bulk endocytosis. Also, normal synaptic activity is sufficient under the non-permissive conditions to cause accumulation of LVs, which obviously cannot be degraded in the absence of functional CHC-1. Last, we also analyzed the LV diameter and found that chc-1 mutants had significantly larger LVs at the non-permissive temperature (Figure 8D), in line with a function of CHC-1 in the breakdown of LVs to form new SVs. SV diameters were reduced in chc-1 mutants at non-permissive temperature, without ATR (and thus without functional ChR2), as observed previously (Sato et al., 2009). However, photostimulation induced a shift to larger (wild-type-like) SV diameters in the chc-1 mutant, both at permissive and non-permissive temperatures. Possibly, under conditions of evoked SV turnover even the compromised CHC-1 can form normal cages and thus generate normal size SVs; alternatively, other proteins like ERP-1 or UNC-57 can compensate for the lack of clathrin activity.

Despite the lack of endophilins in the unc-57; erp-1 double mutant, SVs were still present and some endocytosis and SV recycling persisted. As clathrin was shown in rodent neurons to act at the PM at non-physiological temperatures, and as a back-up mechanism when ultrafast endocytosis is non-functional (e.g., in the absence of actin polymerization; Watanabe et al., 2014), we wondered if clathrin may act in a similar manner in C. elegans. To address this, we analyzed erp-1; chc-1 and unc-57; chc-1 double mutants. First, we assessed the performance of the NMJ in long term stimulation assays and the resulting body contraction. Since the absolute degree of contraction was very different in different mutants and experimental conditions, and the distinctive phenotype uncovering a defect in SV recycling is the gradual loss of the contraction during the stimulation, we turned to comparing the temporal change (i.e., slope) in (normalized) contractions over the 1 min stimulation period (Figure 9; Supplementary Figure S6). The strongest phenotype (highest slope) was apparent in the unc-57 single mutant. The unc-57; erp-1 or unc-57; chc-1(b1025ts) double mutants (under non-permissive temperature, affecting CHC-1 function), however, somewhat reduced the extent of this phenotype (Figures 9B,C). This could indicate that when all endophilins, or both endophilin A and clathrin are compromised, a compensation by another mechanism (CHC-1/clathrin, UNC-57/endophilin A, or ERP-1/endophilin B, respectively), takes over to enable some endocytosis and SV recycling to occur. In erp-1; chc-1 double mutants, only comparably mild phenotypes were apparent (Figures 9E,F), emphasizing that UNC-57 is the “main” player in both endocytosis and endosomal SV recycling.

To analyze the synaptic correlate of these findings, we analyzed the ultrastructures of the erp-1; chc-1 and unc-57; chc-1 double mutants, at permissive and non-permissive temperatures, without and with stimulation (illumination in absence/presence of ATR; Figure 10). In electron micrographs, particularly large and numerous LVs could be observed in unc-57; chc-1 mutants under hs conditions (Figures 10B,D,E). This was similar for chc-1 single mutants at the non-permissive temperature (though not as pronounced), as well as for unc-57 single mutants. However, it was not the case in erp-1; chc-1 mutants (compare Figures 6E,H, 10A,D,E). Thus, endophilin A, B and clathrin appear to act as mutual back-up mechanisms for SV endocytosis and recycling (with strongest contribution by endophilin A) when the two other players are missing.