Introduction

Chemical synaptic transmission involves a cycle of biogenesis of synaptic vesicles (SVs), their fusion with the plasma membrane (PM), as well as their recycling by endocytosis and de novo formation in the endosome. Coupling of SV exocytosis and endocytosis must be tightly controlled to avoid depletion of the reserve pool of SVs and to enable sustained neurotransmission (Haucke et al., 2011; Lou, 2018). Different hypotheses have been formulated as to how this is achieved within neurons. One hypothesis suggests that an SV-integral transmembrane protein called Flower could form ion channels and gets inserted into the PM during SV fusion (Yao et al., 2009). Subsequently, it may facilitate endocytosis by conducting Ca²⁺ into the cytoplasm, thus contributing to defining endocytic sites (Yao et al., 2017). In Drosophila melanogaster, Flower was proposed to increase phosphatidylinositol-4,5-bisphosphate (PI(4,5)P₂) levels through Ca²⁺ microdomains, which was suggested to drive activity-dependent bulk endocytosis (ADBE) and formation of new SVs after prolonged, intense neurotransmission (Li et al., 2020). The hypothesis that Flower may have Ca²⁺ channel activity was further proposed based on sequence similarities between Flower and the Ca²⁺ selectivity filter of voltage-gated Ca²⁺ channels (VGCCs; Yao et al. (2009)). Additional evidence from Drosophila suggests that Flower may regulate clathrin-mediated endocytosis (CME) in a Ca²⁺ independent fashion (Yao et al., 2017). However, while a facilitatory role of Flower in endocytosis appears to be evolutionarily conserved, and was observed in different organisms and tissues, including non-neuronal cells (Yao et al., 2017; Chang et al., 2018; Rudd et al., 2023), its Ca²⁺ channel activity and its influence on SV recycling is debated (Lou, 2018; Coelho and Moreno, 2020). The kinetics of Ca²⁺ influx mediated by Flower appear to be too slow and the amount conducted too low to have a major impact (Xue et al., 2012). The function of Flower and its dependence on Ca²⁺ moreover seem to be highly dependent on the cell type and the extent of synaptic activity (Yao et al., 2017; Chang et al., 2018). Apart from its role in endocytosis, Flower is involved in cell survival mechanisms during development (Coelho and Moreno, 2020; Costa-Rodrigues et al., 2021). Intriguingly, loss of the mammalian homolog of Flower can reduce tumor growth, suggesting an important function in tumor cell survival and introducing possible options for cancer treatment (Petrova et al., 2012; Madan et al., 2019). However, sufficient research of the exact signaling pathways by means of which Flower mediates cell survival is lacking (Costa-Rodrigues et al., 2021).

Studies which address this mechanism usually propose a two- or three-transmembrane domain organization of the Flower protein structure in which the C-terminus is exposed to the extracellular space and can thus be used for intercellular communication (Rhiner et al., 2010; Costa-Rodrigues et al., 2021). However, more recent research has shown that both N- and C-terminal tails of Flower are likely cytosolic and the longest mammalian isoforms consists of four transmembrane helices which are connected by short loops (Chang et al., 2018; Rudd et al., 2023). The genome of nematodes is predicted to contain only a single isoform of Flower (FLWR-1; wormbase.org). They therefore serve as excellent model organisms to further investigate the evolutionarily conserved role of Flower in neurotransmission and endocytosis. Here, we studied the function of FLWR-1 in the nematode Caenorhabditis elegans. We find that FLWR-1 localizes to SVs and to the PM and is involved in neurotransmission. We further show that FLWR-1 has a facilitatory but not essential role in endocytosis, confirming previous research in Drosophila and mammalian cells. Loss of FLWR-1 surprisingly conveys increased excitability of motor neurons and neurotransmitter release suggesting a deregulation of Ca²⁺ signaling in the presynapse. This is associated with defective SV recycling at the level of the endosome and thus reduced SV numbers upon sustained stimulation. This excitability change is more pronounced in γ-aminobutyric acid (GABA)ergic neurons and leads to an excitation-inhibition (E/I) imbalance at the neuromuscular junction (NMJ) through increased release of the neurotransmitter GABA. A function of FLWR-1 in endocytosis appears to affect also the plasma membrane Ca²⁺-ATPase MCA-3, required for extrusion of Ca²⁺ from the cytosol, and may thus explain the increased excitability observed in flwr-1 mutant neurons.

Materials and methods

Molecular biology

An overview of plasmids used in this paper and how they were generated can be found in Supplementary Table 1. pZIM902 [punc-17b::GCaMP6fOpt] was a gift from Manuel Zimmer. p1676 [punc-17(short)::TagRFP::ELKS-1] was kindly provided by the lab of Zhao-Wen Wang. pJH2523 [punc-25::GCaMP3::UrSL2::wCherry] was a gift from Mei Zhen (Addgene plasmid # 191358; http://n2t.net/addgene:191358; RRID:Addgene_191358) (Lu et al., 2022).

Cultivation of C. elegans

Animals were kept at 20 °C on nematode growth medium (NGM) plates seeded with OP50-1 bacteria (Brenner, 1974). For optogenetic experiments, OP50-1 was supplemented with 200 µM ATR prior to seeding and animals were kept in darkness. Transgenic animals carrying extrachromosomal arrays were generated by microinjection into the gonads (Fire, 1986). L4 staged larvae were picked ∼18 h before experiments and tested on at least three separate days with animals picked from different populations. Transgenic animals were selected by fluorescent markers using a Leica MZ16F dissection stereo microscope. Integration of the sybIs8965 transgenic array (PHX8965 strain) was performed by SunyBiotech. The RB2305 strain containing the flwr-1(ok3128) allele was outcrossed three times with N2 wild type animals. An overview of strains with their respective genotypes and transgenes can be found in Supplementary Table 2.

Counting live progeny

To count living progeny per animal, five L4 larvae were singled onto NGM plates seeded with 50 µl OP50-1. Animals were then picked onto fresh plates after two days and on each of the following two days and removed on day five. Living progeny per animal was counted after reaching adulthood and summed up over the three plates each animal laid eggs on, respectively. Experiments were performed blinded to the genotype.

Pharmacological assays

1.5 mM aldicarb and 0.25 levamisole plates were prepared by adding the compounds to liquid NGM prior to plate pouring (Mahoney et al., 2006). Aldicarb (Sigma Aldrich, USA) was kept as a 100 mM stock solution in 70 % ethanol. Levamisole (Sigma Aldrich, USA) was stored as a 200 mM solution dissolved in ddH₂O. 15 - 20 young adult animals were transferred to each plate and tested every 15 (levamisole) or 30 (aldicarb) minutes. Assays were performed blinded to the genotype and control groups (wild type and flwr-1 mutants) were measured in parallel on the same day. Animals that did not respond after being prodded three times with a hair pick were counted as paralyzed. Worms that crawled of the plate were disregarded from analysis.

Measurement of swimming speed using the multi-worm-tracker (MWT)

Swimming speed was measured as described previously (Vettkotter et al., 2022). In short, worms were washed three times with M9 buffer to remove OP50 and transferred onto 3.5 cm NGM plates with 800 µl M9. Animals were visualized using the multiworm tracker (MWT) platform (Swierczek et al., 2011) equipped with a Falcon 4M30 camera (DALSA). Channelrhodopsin-2 (ChR2) was stimulated with 470 nm light at 1 mW/mm² intensity for 90 s. 30 s videos were captured and thrashing was analyzed using the “wrmTrck” plugin for ImageJ (Nussbaum-Krammer et al., 2015). The automatically generated tracks were validated using a custom written Python script (https://github.com/dvettkoe/SwimmingTracksProcessing).

Measurement of body length

Body length was measured as described previously (Liewald et al., 2008; Seidenthal et al., 2022). In short, single animals were transferred onto unseeded NGM plates and illuminated with light from a 50 W HBO lamp which was filtered with a 450 - 490 nm bandpass excitation filter. ChR2 was stimulated with 100 µW/mm² light intensity. A 665 - 715 nm filter was used to avoid unwanted activation by brightfield light. Body length was analyzed using the “WormRuler” software (version 1.3.0) and normalized for the average skeleton length before stimulation (Seidenthal et al., 2022). Values which are more than 20 % higher or lower than the initial body length were discarded as they result from artifacts in the background correction and are biomechanically unlikely. Basal body length was calculated from the skeleton length in pixels which is generated by the WormRuler software. This was converted to mm and averaged for each animal over a duration of five seconds.

Light microscopy and quantification of fluorescence

Animals were placed upon 7 % agarose pads in M9 buffer. Worms were immobilized by either using 20 mM levamisole in M9 or Polybead polystyrene microspheres (Polysciences) for experiments involving GABAergic stimulation. For optogenetic experiments, single animals were placed upon pads to avoid unwanted pre-activation of channelrhodopsins. Imaging was performed using a Axio Observer Z1 microscope (Zeiss, Germany). Proteins were excited using 460 nm and 590 nm LEDs (Lumen 100, Prior Scientific, UK) coupled via a beamsplitter. Background correction was conducted by placing a region of interest (ROI) within the animal yet avoiding autofluorescence.

GCaMP and pHluorin imaging and ChrimsonSA stimulation was performed using a 605 nm beamsplitter (AHF Analysentechnik, Germany) which was combined with a double band pass filter (460 - 500 and 570 - 600 nm). A single bandpass emission filter was used (502.5 - 537.5 nm). pOpsicle assays were performed as described previously using a 100-x objective (Seidenthal et al., 2023). We extended the ChrimsonSA stimulation light pulse to 30 s to be consistent with the stimulus length of electron microscopy (EM) and electrophysiology experiments. Moreover, 2 x 2 binning was applied. Image sequences with 200 ms light exposure (5 frames per second) were acquired with an sCMOS camera (Kinetix 22, Teledyne Photometrics, USA). Video acquisition was controlled using the µManager v.1.4.22 software (Edelstein et al., 2014). The timing of LED activation was managed using a Autohotkey script. GCaMP imaging was conducted similarly with the same setup and light intensity for ChrimsonSA stimulation (40 µW/mm²) yet without binning. Dorsal nerve cord (DNC) fluorescence was quantified with ImageJ by placing a region of interest ROI with the Segmented Line tool. XY-drift was corrected using the Template Matching plugin if necessary. Animals showing excessive z-drift were discarded. A custom written python script was used to summarize background subtraction and normalization to the average fluorescence before pulse start (https://github.com/MariusSeidenthal/pHluorin_Imaging_Analysis). For GCaMP imaging, the bleach correction function of this python script was used which corrects fluorescence traces with strong bleaching. Strong bleaching is defined as the fluorescence intensity of the mean of the last second of measurement being less than 85% of the first second. The script performs linear regression analysis for DNC fluorescent traces and background ROI and corrects values prior to background correction. The increase in GCaMP fluorescence during stimulation was highly dependent on basal fluorescence values before stimulation and transient Ca²⁺ signals. This caused some measurements to show very strong increases in fluorescence. To avoid distortion of statistical analyses, we performed outlier detection for the increase during stimulation on all datasets with the Graphpad Prism 9.4.1 Iterative Grubb’s method. The alpha value (false discovery rate) was set to 0.01. For pOpsicle assays, animals were screened for whether they show a significant increase in fluorescence during stimulation, which is necessary for analysis of fluorescence decay after stimulation. A strong signal was determined as the maximum background corrected fluorescence during the light pulse (moving average of 5 frames) being larger than the average before stimulation + 3 times the standard deviation of the background corrected fluorescence before stimulation. Regression analysis of pHluorin fluorescence decay was performed as described previously by using Graphpad Prism 9.4.1 and fitting a “Plateau followed one-phase exponential decay” fit beginning with the first time point after stimulation to single measurements (Seidenthal et al., 2023). Animals that displayed no increase during stimulation as well as animals showing no decay or spontaneous signals after stimulation were discarded from analysis.

Comparison of dorsal and ventral (VNC) nerve cord fluorescence of GFP::FLWR-1 and mCherry::SNB-1 was conducted by imaging the posterior part of the animal where an abundance of synapses can be found. Images were acquired using the same beamsplitter and excitation filter as used for pOpsicle experiments but equipped with a 500 – 540 nm/600 – 665 nm double bandpass emission filter (AHF Analysentechnik, Germany) and 2 x 2 binning. The Arduino script “AOTFcontroller” was used to control synchronized 2-color illumination (Aoki et al., 2023). VNC and DNC fluorescence were analyzed by choosing single frames from acquired z-stacks in which nerve cords were well focused. Fluorescence intensities were quantified by placing a Segmented Line ROI. Kymographs were generated with the ImageJ Multi Kymograph function.

Confocal laser scanning microscopy was performed on a LSM 780 microscope (Zeiss, Germany) equipped with a Plan-Apochromat 63-x oil objective. The Zeiss Zen (blue edition) Tile Scan and Z-Stack functions were used to generate overview images of animals expressing GFP::FLWR-1. Images were processed and maximum z-projection performed using ImageJ. Colocalization analysis was performed by placing a Segmented Line ROI onto the DNC and plotting the fluorescence profile along the selected ROI for both color channels. Correlation of fluorescence intensities was then conducted using the GraphPad Prism 9.4.1 Pearson correlation function.

Fluorescence of GFP taken up by coelomocytes was visualized with a 40-x objective. An eGFP filter cube (AHF Analysentechnik, Germany) was used and fluorescence was excited with the 460 nm LED. Images were acquired using 50 ms exposure and 2 x 2 binning. Fluorescence was quantified by drawing a ROI around all coelomocytes found within any worm imaged using the Freehand selections tool and performing background correction. Coelomocyte fluorescence was then averaged for each animal.

Electrophysiology

Electrophysiological recordings of body wall muscle cells (BWMs) were done in dissected adult worms as previously described (Liewald et al., 2008). Animals were immobilized with Histoacryl L glue (B. Braun Surgical, Spain) and a lateral incision was made to access NMJs along the anterior VNC. The basement membrane overlying BWMs was enzymatically removed by 0.5 mg/ml collagenase for 10 s (C5138, Sigma-Aldrich, Germany). Integrity of BWMs and nerve cord was visually examined via DIC microscopy.

Recordings from BWMs were acquired in whole-cell patch-clamp mode at 20-22 °C using an EPC-10 amplifier equipped with Patchmaster software (HEKA, Germany). The head stage was connected to a standard HEKA pipette holder for fire-polished borosilicate pipettes (1B100F-4, Worcester Polytechnic Institute, USA) of 4–10 MΩ resistance. The extracellular bath solution consisted of 150 mM NaCl, 5 mM KCl, 5 mM CaCl₂,1 mM MgCl₂,10 mM glucose, 5 mM sucrose, and 15 mM HEPES, pH 7.3, with NaOH, ∼330 mOsm. The internal/patch pipette solution consisted of K-gluconate 115 mM, KCl 25 mM, CaCl₂ 0.1 mM, MgCl₂ 5 mM, BAPTA 1 mM, HEPES 10 mM, Na₂ATP 5 mM, Na₂GTP 0.5 mM, cAMP 0.5 mM, and cGMP 0.5 mM, pH 7.2, with KOH, ∼320 mOsm.

Voltage clamp experiments were conducted at a holding potential of −60 mV. Light activation was performed using an LED lamp (KSL-70, Rapp OptoElectronic, Hamburg, Germany; 470 nm, 8 mW/mm²) and controlled by the Patchmaster software. Subsequent analysis was performed using Patchmaster and Origin (Originlabs). Analysis of mPSCs was conducted with MiniAnalysis (Synaptosoft, Decatur, GA, USA, version 6.0.7) and rate and amplitude of mPSCs was analyzed in 1 s bins. Exponential decay of mPSCs during stimulation was calculated with Graphpad Prism 9.4.1 by fitting a one-phase exponential decay beginning with the first time point during stimulation.

Transmission electron microscopy

Prior to high-pressure freezing (HPF), L4 animals were transferred to freshly seeded E. coli OP50-1 dishes supplemented with or without 0.1 mM ATR. HPF fixation was performed on young adult animals as described previously (Weimer, 2006; Kittelmann et al., 2013b). Briefly, 20 - 40 animals were transferred into a 100 µm deep aluminum planchette (Microscopy Services) filled with E. coli (supplemented with or without ATR, respectively), covered with a sapphire disk (0.16 mm) and a spacer ring (0.4 mm; engineering office M. Wohlwend) for photostimulation. To prevent pre-activation of ChR2, all preparations were carried out under red light. Animals were continuously illuminated for 30 s with a laser (470 nm, ∼20 mW/mm²) followed by HPF at −180 °C under 2100 bar pressure in an HPM100 (Leica Microsystems). Frozen specimens were transferred under liquid nitrogen into a Reichert AFS machine (Leica Microsystems) for freeze substitution. Samples were incubated with tannic acid (0.1% in dry acetone) fixative at −90 °C for 100 h. Afterward, a process of washing was performed for substitution with acetone, followed by incubation of the samples in 2% OsO₄ (in dry acetone) for 39.5 h while the temperature was slowly increased up to room temperature. Subsequently, samples were embedded in epoxy resin (Agar Scientific, AGAR 100 Premix kit – hard) by increasing epoxy resin concentrations from 50% to 90% at room temperature and 100% at 60 °C for 48 h. Electron micrographs of 2 - 5individual animals, and from 12 – 22 synapses per treatment, were acquired by cutting cross sections at a thickness of 40 nm, transferring cross sections to formvar- or pioloform covered copper slot grids. Specimens were counterstained in 2.5% aqueous uranyl acetate for 4 min, followed by washing with distilled water, and incubation in Reynolds lead citrate solution for 2 min in a CO₂-free chamber with subsequent washing steps in distilled water. VNC regions were then imaged with a Zeiss 900 TEM, operated at 80 kV, with a Troendle 2K camera. Images were scored and tagged blind in ImageJ (version 1.53c, National Institute of Health) as described previously, and analyzed using SynapsEM (Watanabe et al., 2020; Vettkotter et al., 2022). Since the number of synaptic organelles varied between synapses of different sizes, their counts were normalized to the average synaptic profile area of 145,253 nm². Endosomes are defined as structures larger than 100 nm and located more than 50 nm away from the dense projection, as described by (Watanabe et al., 2014). Irregular vesicular structures exceeding 100 nm in either height or width were also included.

Protein alignment and structure prediction

Protein alignment was performed with ClustalX (Larkin et al., 2007) and shading of conserved residues created with Boxshade (https://junli.netlify.app/apps/boxshade/). Transmembrane domains as well as membrane orientation were predicted by entering the predicted amino acid sequence to DeepTMHMM (Hallgren et al., 2022). FLWR-1 structure was modeled and visualized with Alphafold3 (AF3) (Abramson et al., 2024). FLWR-1 tetramers have been modeled using AlphafoldMultimer or AF3 (Evans et al., 2022). Protein structures were analyzed and edited using PyMol v2.5.2. Docking of PI(4,5)P₂ to the FLWR-1 AF3 model was conducted in PyMol using the DockingPie plugin running the Autodock/Vina analysis (Eberhardt et al., 2021; Rosignoli and Paiardini, 2022).

Statistical Analysis and generation of graphs

Graphs were created and statistical analyses were performed using GraphPad Prism 9.4.1. Data was displayed as mean ± standard error of the mean (SEM) unless noted otherwise. Unpaired t-tests were used to compare two datasets and one-way ANOVAs or two-way ANOVAs if three or more datasets were assessed, according to the number of tested conditions. Fitting of a mixed-effects model was used instead of two-way ANOVA if the number of data points was not constant between different datasets for example if the number of animals differed between time points. Mann-Whitney tests were performed if two datasets were compared that were not normally distributed and datasets were depicted as median with interquartile range (IQR) in statistical analyses. Kruskal-Wallis tests were conducted if not normally distributed datasets were assessed. Iterative Grubb’s outlier detection was used when required and as indicated. The alpha value (false discovery rate) was set to 0.01. Representations of the exon-intron structures of genes were created using the “Exon-Intron Graphic Maker” (Bhatla, 2012).

Results

FLWR-1 is involved in neurotransmission

The amino acid sequence of C. elegans FLWR-1 is conserved to its D. melanogaster (Fwe-Ubi) and Homo sapiens (hFwe4) homologues (Fig. 1A). Fwe-Ubi has been shown to facilitate recovery of neurons following intense synaptic activity (Yao et al., 2009; Yao et al., 2017). To investigate whether this involvement of Flower in neurotransmission is evolutionarily conserved, we studied a mutant lacking most of the flwr-1 coding region (ok3128, Fig. 1B) (Consortium, 2012). Animals lacking FLWR-1 did not display severe phenotypic differences to wild type animals. Basal locomotion in liquid (Fig. 1C, D seconds 0 – 30) and body length of young adults (Fig. S1A) were not significantly different from the respective values of wild type animals. However, fertility appeared slightly reduced (Fig. S1B). Thus, it is likely that FLWR-1 does not have an essential function in neurotransmission but rather a regulatory or facilitatory role. Previous studies have shown that mutations which only weakly affect basal locomotion, can display severely reduced recovery of swimming speed after strong optogenetic stimulation of cholinergic neurons (Yu et al., 2018). We therefore subjected flwr-1(ok3128) mutants expressing channelrhodopsin-2(H134R), which were treated with the chromophore all-trans retinal (ATR), to blue light during swimming (Nagel et al., 2005; Liewald et al., 2008). Photostimulation resulted in a stop of all swimming during the light pulse, followed by a slow recovery in the dark. Indeed, loss of FLWR-1 led to a significantly slowed recovery of swimming locomotion (Fig. 1C, D). This could be fully rescued by transgenic expression of genomic FLWR-1 including a two kilobase sequence upstream of the putative start codon, hereafter called pflwr-1 (as the promoter of flwr-1). To further evaluate the involvement of FLWR-1 in neurotransmission, we exposed flwr-1 mutants to aldicarb. This acetylcholine esterase inhibitor induces paralysis due to accumulation of acetylcholine in the synaptic cleft (Mahoney et al., 2006; Blazie and Jin, 2018). Resistance to aldicarb indicates either reduced release or reception of acetylcholine or increased inhibitory signaling (Vashlishan et al., 2008; Janosi et al., 2024). Loss of FLWR-1 led to a significant delay in paralysis indicating involvement in transmission at the NMJ (Fig. 1E). Full rescue of the flwr-1 mutant phenotypes suggests that expression from the 2 kB fragment of the endogenous promoter fully recapitulates the native expression in the context of NMJ function.

Fig. 1.

FLWR-1 is expressed in excitable cells and localizes to synaptic vesicles

Invertebrate and vertebrate homologues of FLWR-1 are predicted to be membrane proteins containing four transmembrane helices with both C- and N-termini located in the cytosol (Yao et al., 2009; Chang et al., 2018; Rudd et al., 2023). In silico predictions using the FLWR-1 sequence of 166 amino acids suggest evolutionary conservation of the tetraspan structure (Fig. S2A) (Hallgren et al., 2022). Accordingly, Alphafold3 prediction of FLWR-1 protein structure implies a four helical configuration, with helix lengths that could span biological membranes, and three additional, shorter α-helices (Fig. S2B) (Abramson et al., 2024). We sought to determine the cellular as well as subcellular localizations of FLWR-1 by tagging its C-terminus with GFP and expressing the construct using the endogenous promoter (Fig. 2A – C). Green fluorescence could be observed in developing embryos in the uterus, as well as in various tissues including neurons, body wall and pharyngeal muscles (Fig. 2A). FLWR-1 localized to neurites and cell bodies of nerve ring neurons, representing the central nervous system (brain) of the nematode (Fig. 2B) (Ward et al., 1975). These results are in accordance with single-cell RNA-sequencing data obtained from C. elegans which show near ubiquitous expression of flwr-1/F20D1.1 (Fig. S2C, Supplementary Table 3; Taylor et al. (2021)). Within muscle cells, FLWR-1 was primarily targeted to the PM (Fig. 2C). As its D. melanogaster homolog was localized to synaptic vesicles (SVs) (Yao et al., 2009), we wondered whether FLWR-1 would colocalize with known SV markers such as synaptobrevin-1 (SNB-1) (Nonet, 1999; Calahorro and Izquierdo, 2018). Indeed, GFP::FLWR-1 fluorescence largely overlapped with co-expressed mCherry::SNB-1 in the dorsal nerve cord (DNC; Fig. 2D, E). FLWR-1 further seemed to be enriched in fluorescent puncta in DNC and sublateral nerve cords indicating synaptic localization (Fig. 2D - F), however it was not restricted to synaptic regions only, meaning it is likely present also in the PM.

Fig. 2.

Furthermore, we observed moving particles, probably SV precursors, which contain SNB-1 and FLWR-1, travelling along commissures between ventral nerve cord (VNC) and DNC (Fig. 2F, G and Supplementary Movie 1). Anterograde transport of these precursors towards synapses has been shown to be dependent on kinesin-3/UNC-104 (Hall and Hedgecock, 1991; Klopfenstein and Vale, 2004). We used a strong reduction-of-function allele (e1265) affecting interaction of UNC-104 with its cargo to investigate whether FLWR-1 is actively transported towards synapses (Cong et al., 2021). Indeed, animals lacking functional UNC-104 showed a reduced amount of axonal GFP fluorescence in the nerve ring and DNC while cell bodies were clearly visible (Fig. S2D). The ratio of DNC to VNC fluorescence is significantly decreased in unc-104 mutants suggesting defective anterograde transport (Fig. 2H). Distribution of SNB-1 was similarly affected (Gally and Bessereau, 2003; Cuentas-Condori et al., 2023). Together, these results argue that FLWR-1 is expressed in neurons (as well as in muscles and other cell types) and actively transported towards synapses.

GABAergic signaling is increased in flwr-1 knockout mutants

Since FLWR-1 is also expressed in body wall muscles (BWMs), we wondered whether the observed aldicarb resistance originates from reduced acetylcholine reception by acetylcholine receptors (AChR) (Mahoney et al., 2006). To test this, we exposed animals lacking FLWR-1 to levamisole, an AChR agonist which induces paralysis by hypercontraction, similar to aldicarb (Sattelle et al., 2002; Gottschalk et al., 2005). We found no significant differences in the rate of paralysis between wild type and mutant animals (Fig. S3A). However, transgenic expression of FLWR-1 in BWMs did rescue the aldicarb resistance partially (Fig. S3B). Therefore, loss of FLWR-1 seems to decrease excitability of muscle cells in an aldicarb dependent context which may contribute to the aldicarb resistance of flwr-1 mutants.

Yet, we also observed enrichment of FLWR-1 in fluorescent puncta indicating presynaptic localization (Fig. 3A, B). Moreover, FLWR-1 partially colocalized with the dense projection marker ELKS-1 in cholinergic as well as in GABAergic neurons (Fig. 3A, B; Fig. S3C-E; Dai et al. (2006); Kittelmann et al. (2013a)). Unlike ELKS-1, FLWR-1 could also be found to a lesser extent in intersynaptic regions. This indicates that FLWR-1 might be involved in neurotransmission at both cholinergic and GABAergic NMJs yet is not exclusively localized to active zones. To determine whether the site of action of FLWR-1 in NMJ signaling is evenly located to cholinergic and GABAergic neurons, we rescued FLWR-1 in each subset of these neurons. To achieve the intended expression, the promoters of the vesicular transporters of acetylcholine (UNC-17) and GABA (UNC-47) were inserted in front of the FLWR-1 genomic sequence. Surprisingly, expression of FLWR-1 in GABAergic, but not in cholinergic, neurons rescued aldicarb resistance (Fig. 3C). This was unexpected since mutations affecting SV recycling commonly lead to a slowed replenishment of SVs and thus reduced acetylcholine release (Salcini et al., 2001; Schuske et al., 2003; Yu et al., 2018). However, our results indicate that loss of FLWR-1 leads to increased release of GABA which counteracts the aldicarb induced paralysis (Camara et al., 2019). To confirm this, we crossed flwr-1 mutants into animals which lack the vesicular GABA transporter UNC-47. This abolishes GABA release and induces hypersensitivity to aldicarb (Vashlishan et al., 2008). We observed that the additional mutation of unc-47 led to a complete loss of aldicarb resistance in flwr-1 mutants (Fig. 3D). This supports the notion that increased GABA signaling, not decreased acetylcholine release, is the main cause of aldicarb resistance in flwr-1 animals.

Fig. 3.

Neuronal excitability is increased in flwr-1 mutants

Since the release of GABA appeared to be increased in flwr-1 mutant animals, we wondered whether the excitability of GABAergic neurons is increased. To test this, the fluorescent Ca²⁺ indicator GCaMP was expressed in GABAergic neurons, allowing us to estimate relative Ca²⁺ levels at presynaptic sites (Nakai et al., 2001; Lu et al., 2022). We co-expressed the red-shifted ChrimsonSA channelrhodopsin variant to control depolarization of neurons, independent of GCaMP excitation light (Oda et al., 2018; Seidenthal et al., 2022). As expected, optogenetic stimulation caused an increase in GCaMP fluorescence at NMJs in the DNC (Fig. 4A). Comparing wild type and flwr-1 mutants, we found that the gain in fluorescence intensity was significantly higher in animals lacking FLWR-1 (Fig. 4A, B), supporting our previous finding of increased neurotransmission in GABAergic neurons. To verify this at the behavioral level, we used a strain expressing ChR2(H134R) in GABAergic neurons, as it can be used to assess GABA release through measurement of body length (Liewald et al., 2008). Since GABA receptors hyperpolarize muscle cells, optogenetic stimulation of GABAergic motor neurons causes relaxation and thus increased body length (Schultheis et al., 2011; Seidenthal et al., 2022). To augment the effect of ChR2 stimulation on body length and to observe the effect of GABA release independent of cholinergic neurotransmission, the assay was performed in a mutant background lacking a functional subunit of the levamisole receptor (unc-29; Fleming et al. (1997); Richmond and Jorgensen (1999)). As expected, stimulation with blue light led to increased body length (Fig. 4C). In accordance with Ca²⁺ imaging results, flwr-1; unc-29 double mutants showed a significantly stronger elongation during stimulation compared to the unc-29 single mutant (Fig. 4C, D). These results indicate that Ca²⁺ influx into the synapse is increased, likely causing more GABA to be released in animals lacking FLWR-1. The previously observed aldicarb resistance indicates that this change in GABAergic transmission outweighs possible changes in cholinergic transmission (Fig. 3). To investigate whether neuronal excitability may be generally upregulated in flwr-1 mutants, we assessed whether Ca²⁺ influx is affected in cholinergic motor neurons as well (Fig. 4E, F). Again, we observed increased GCaMP fluorescence levels during optogenetic stimulation. Together, these results indicate that loss of FLWR-1 conveys an upregulation of neuronal excitability during continuous stimulation in both classes of motor neurons (Fig. 4G). However, the overall E/I balance appears to be shifted towards stronger inhibition of BWMs.

Fig. 4.

Endocytosis is slowed in flwr-1 mutants in non-neuronal cells and neurons

Homologues of FLWR-1 were found to be involved in endocytosis in neuronal as well as non-neuronal cells (Yao et al., 2009; Chang et al., 2018; Rudd et al., 2023). We thus wondered whether this function is evolutionarily conserved in nematodes. One possibility to assess this in C. elegans is to observe the efficiency of endocytosis by coelomocytes (Fares and Greenwald, 2001). These scavenger cells continuously endocytose fluid from the body cavity and loss of endocytosis-associated factors leads to a defective uptake of proteins secreted from other tissues (Fares and Grant, 2002). By expressing GFP fused to a secretory signal sequence which is discharged from BWMs, endocytic uptake by coelomocytes can be quantified by fluorescence microscopy (Fares and Greenwald, 2001; Bednarek et al., 2007). According to single-cell RNAseq data, FLWR-1 is expressed in coelomocytes (Supplementary Table 3; Taylor et al. (2021)). Indeed, mutation of flwr-1 led to strongly reduced GFP fluorescence levels within coelomocytes indicating a reduced uptake by endocytosis (Fig. 5A, B).

Fig. 5.

We further used the recently established pOpsicle (pH-sensitive optogenetic reporter of synaptic vesicle recycling) assay to estimate the amount of SV fusion and the rate of recycling of SV components in cholinergic neurons (Seidenthal et al., 2023). This assay combines a pHluorin based probe fused to an SV-associated protein (SNG-1) and optogenetic stimulation of neurotransmitter release; this way, pHluorin fluorescence is unquenched during stimulated exocytosis and quenched during SV endocytosis and recycling (Sankaranarayanan et al., 2000). Loss of FLWR-1 led to significantly increased fluorescence signals during stimulation which indicates more SV fusion compared to wild type (Fig. 5C, E). This is in accordance with the increased excitability of cholinergic neurons we observed earlier (Fig. 4E, F). Moreover, the rate of fluorescence decay after stimulation was significantly delayed in flwr-1 mutants suggesting slower recycling of SVs, or reduced acidification of endosomal structures or SVs after recycling (Fig. 5D, F).

Slowed replenishment of SV pools caused by defective recycling is known to cause synaptic depression, as assessed by electrophysiological measurements of post-synaptic currents (Wu and Betz, 1998; Kittelmann et al., 2013b; Krick et al., 2021). This is also the case in Drosophila for animals lacking Flower (Yao et al., 2009). To address this in C. elegans, we measured miniature post-synaptic currents (mPSCs) in BWMs, which reflect the post-synaptic effects of presynaptic neurotransmitter release (Liewald et al., 2008; Weissenberger et al., 2011). Animals lacking FLWR-1 showed no significant difference in unstimulated mPSC frequency or amplitude (Fig. S4A, B) which is in accordance with FLWR-1 being dispensable in basal swimming locomotion (Fig. 1C, D). Similarly, pulsed optogenetic stimulation of cholinergic neurons at different frequencies did not reveal a difference in the measured currents between wild type and mutants (Fig. S4C, D) suggesting that FLWR-1 might be needed only during continuous stimulation. Indeed, flwr-1 mutants showed an accelerated rundown of the mPSC frequency in BWMs during constant illumination (Fig. 5G, H). This is in accordance with a role of FLWR-1 in SV recycling. Surprisingly, mPSC amplitudes in flwr-1 mutants are reduced during 30 s hyperstimulation (Fig. 5I, J). This suggests that either the amount of neurotransmitter released from a single SV or the number of multivesicular fusion events is decreased (Liu et al., 2005). The absolute mPSC frequency, which represents the number of SVs fusing per time, was also smaller during continuous stimulation in flwr-1 animals (Fig. S4E, F, G). These results contrast the increased excitability of cholinergic neurons we observed (Fig. 4A, E). Since cholinergic neurons also stimulate GABAergic neurons, and since GABAergic minis are also evaluated here, the dissection of animals, required for electrophysiological recordings, could damage neuronal commissures, causing an interruption of physiological signal transmission, which is otherwise observed in intact animals. We further analyzed whether flwr-1 mutants can recover from strong optogenetic 30 s stimulation by applying a short stimulus after a recovery period (inter stimulus interval, ISI) of 15 s, and found no significant difference to wildtype (Fig. S4E, H). This suggests that a 15 s ISI is sufficient for flwr-1 mutants to recover to the same extent as wild type synapses. In sum, our results support a facilitatory role of FLWR-1 in SV recycling specifically during continuous stimulation.

Loss of FLWR-1 leads to depleted SV pools and accumulation of endosomal structures after optogenetic stimulation

Mutations which affect endocytosis commonly have fewer SVs because of defective recovery of SV components which could also be shown for Flower defective Drosophila boutons (Schuske et al., 2003; Yao et al., 2009). Such a defect is even more pronounced when samples are conserved immediately following optogenetic stimulation by high-pressure freezing (Kittelmann et al., 2013b; Yu et al., 2018). In nematodes, this optogenetic stimulation can be controlled by supplementing all-trans retinal (ATR), the chromophore of ChR2 (Nagel et al., 2005). Ultrastructural analysis using transmission electron microscopy (TEM) indeed revealed fewer SVs in stimulated flwr-1 mutant synapses compared to wild type (Fig. 6A, B). This indicates that flwr-1 mutants, unlike wild type, are unable to refill SV pools sufficiently fast. At the same time, loss of Drosophila Flower has been shown to cause defective formation of bulk endosomal structures after strong stimulation (Yao et al., 2017). In contrast to this, we observed an increased, rather than decreased, number of endosomes in stimulated flwr-1 mutant synapses (Fig. 6C). This may either be caused by increased SV fusion which triggers bulk endosomal formation (Clayton et al., 2008; Wu et al., 2014) or defective breakdown of recycling endosomes (Watanabe et al., 2013; Gan and Watanabe, 2018; Yu et al., 2018). Our pHluorin imaging data (Fig. 5C - F), would support the notion that both, increased SV fusion as well as defective recovery and subsequent acidification of SVs, may cause the higher number of endosomes in flwr-1 mutants. However, we note that our stimulation regime likely resembles a more physiological activation of neurotransmission compared to the intense stimuli previously used (Yao et al., 2017), as wild type synapses only rarely contained endosomal structures. Moreover, we observed fewer docked vesicles in flwr-1 animals which have been treated with ATR, compared to those without (Fig. 6D). This might be caused by increased SV fusion and is in line with the increased excitability we observed before. The number of large vesicles and neuropeptide-containing dense core vesicles (DCVs) is unchanged in animals lacking FLWR-1 (Fig. S5).

Fig. 6.

The increased excitability of flwr-1 neurons may be caused by deregulation of MCA-3

While a facilitatory role of Flower in endocytosis appears to be conserved in C. elegans, we found no evidence that FLWR-1 conducts Ca²⁺ upon insertion into the PM (Yao et al., 2009). On the contrary, Ca²⁺ levels are increased during strong stimulation of neurons in animals lacking FLWR-1. We thus wondered whether clearance of Ca²⁺ which has entered the synapse via VGCCs, might be defective in flwr-1 mutants. In neurons and muscles, the plasma membrane Ca²⁺ ATPase (PMCA) is involved in extruding Ca²⁺ from the cell (Boczek et al., 2019; Krick et al., 2021; Krebs, 2022). The C. elegans homolog MCA-3 (also known as CUP-7) is expressed in neurons, muscle cells and coelomocytes (Fig. S6, Supplementary Table 3; Bednarek et al. (2007); Taylor et al. (2021)). Interestingly, reducing the function of MCA-3 by mutation was shown to cause a similar defect in endocytosis of secreted GFP in coelomocytes as the loss of FLWR-1 does (Fig. 5A, B; Bednarek et al. (2007)). We therefore wondered whether MCA-3 might be negatively affected in flwr-1 mutants. To assess this, we used a mutant (ok2048) lacking part of the C-terminus of MCA-3 including part of the regulatory calmodulin-binding domain (Kraev et al., 1999; Consortium, 2012; Mantilla et al., 2023) (Fig. 7A).

Fig. 7.

Since full knockouts of mca-3 are lethal, it is likely that this represents a reduction-of-function mutation (Bednarek et al., 2007). As expected, mca-3 mutants displayed increased Ca²⁺ influx (or net Ca²⁺ levels, representing the summed VGCC-mediated entry and MCA-3-mediated efflux) upon optogenetic stimulation of cholinergic motor neurons (Fig. 7B, C). Elevated Ca²⁺ levels appeared to be further enhanced by additionally mutating flwr-1, however this did not meet significance. This was likely due to the large variance in the measurements, which was due to some cells showing very low basal Ca²⁺ levels (i.e. GCaMP fluorescence) yet reaching similarly high levels during stimulation. As is, our data suggest that the two genes are not in a common pathway, as the effects appear to be additive. However, the mca-3 mutant used here is only a reduction-of-function mutant, thus additional deregulation by loss of FLWR-1 may explain the possible additive effect. A different picture was observed in aldicarb assays: Reduction of MCA-3 function conveyed aldicarb resistance that was, however, less pronounced than for flwr-1 mutants (Fig. 7D). Here, the double mutant did not further augment the mutant phenotype of flwr-1, which may be interpreted as that both proteins function in the same pathway, where flwr-1 is upstream of mca-3.

FLWR-1 structure prediction does not imply Ca²⁺ conducting ability but PI(4,5)P₂ binding

We observed increased rather than decreased Ca²⁺ influx during stimulation, which contrasts findings from D. melanogaster Flower (Yao et al., 2017). Previously, an evolutionarily conserved glutamate residue within the transmembrane domain has been proposed to be a Ca²⁺ selectivity filter because of similarities to Ca_v1.2 and TRP channels (Yao et al., 2009). This group therefore proposed that Flower, like TRP channels, can form homo-tetramers to conduct Ca²⁺ (Hoenderop et al., 2003; Yao et al., 2009; Zhang et al., 2023). We therefore wondered whether loss of the conserved glutamate residue impacts FLWR-1 function in C. elegans (Fig. 8A). Surprisingly, expressing a mutant variant of FLWR-1 in which glutamate 74 is exchanged to glutamine significantly decreased the aldicarb resistance of the flwr-1 mutant (Fig. 8B). This suggests that the respective glutamate may not be essential for FLWR-1 function but may have an impact on protein stability. We thus performed structure and oligomerization prediction of FLWR-1 using Alphafold 3 (AF3) to estimate the approximate location of the conserved glutamate residue within a putative FLWR-1 homotetramer (Evans et al., 2022; Abramson et al., 2024). One of the predicted complexes indeed revealed a pore-like structure (Fig. S7A). However, while conventional Ca²⁺ conducting channels contain a glutamate residue within their pore domain (Chen et al., 2023) (Fig. S7B), E74 of FLWR-1 is not part of pore-lining residues in the predicted tetramer.

Fig. 8.

FLWR-1, like MCA-3, has further been suggested to bind to phosphatidylinositol-4,5-bisphosphate (PI(4,5)P₂), an important regulator of SV recycling, and affect its levels within the PM (Bednarek et al., 2007; Lopreiato et al., 2014; Li et al., 2020). Binding of transmembrane proteins (including TRP channels) to PI(4,5)P₂ is generally mediated by positively charged amino acid residues (lysine and arginine) close to the intracellular side of the PM, which bind to the negatively charged phosphate residues (Fig. S7C) (Ribalet et al., 2005; Hansen et al., 2011; Duncan et al., 2020; Rohacs, 2024). The predicted FLWR-1 structure similarly contains many basic amino acids which are likely to be close to the intracellular PM leaflet (Fig. S7D). In silico docking prediction was then used to estimate how PI(4,5)P₂ may bind to FLWR-1. In the resulting structure, both phosphate residues are close to lysine 31 and arginine 27 (Fig. S7D). A positive charge at this position is evolutionarily conserved, suggesting an important function (Fig. 8C). Moreover, these amino acids have been suggested to mediate PI(4,5)P₂ binding in D. melanogaster Flower (Li et al., 2020). Expression of a mutated version of FLWR-1 in which both residues were replaced by alanines only partially rescued aldicarb resistance (Fig. 8D), implying that FLWR-1 function may be facilitated by these residues and their potential interaction with PI(4,5)P₂.

These findings implicate that the conserved glutamate residue 74 is not essential, yet still important for FLWR-1 activity. In addition, an interaction with PI(4,5)P₂ via basic amino acids also seems to be an important factor. However, we do note that FLWR-1 structural models as depicted here rely on in silico predictions. The actual biological structure of FLWR-1, or its homologues, has yet to be determined using X-ray crystallography or cryo-EM.

Discussion

Here, we found that loss of FLWR-1 conveyed increased excitability of motor neurons, particularly GABAergic neurons, thus leading to more GABA release. This gives rise to an E/I imbalance at the NMJ. We showed that FLWR-1 is localized to SVs in neuronal cells, and we found an accumulation of endosomal structures as well as slowed acidification kinetics in flwr-1 mutant animals, suggesting defective recovery of SVs from recycling endosomes. We further found indications that deregulation of PMCAs may be the cause of the increased excitability of flwr-1 neurons (see model, Fig. 8E).

MCA-3 has previously been suggested to facilitate endocytosis in other cells, namely coelomocytes (Bednarek et al., 2007). This indicates that both, increased excitability as well as defective endocytosis in coelomocytes, may be influenced by deregulation of MCA-3 in flwr-1 mutants. Whether FLWR-1 directly activates expression or function of MCA-3, or whether this is caused by a homeostatic change because of the loss of FLWR-1, is yet unclear. However, it is important to note that Drosophila Flower was shown to increase neuronal PI(4,5)P₂ levels, which drives ADBE during sustained neurotransmission (Li et al., 2020). We could show that an involvement with PI(4,5)P₂ is likely conserved in nematodes. Similarly, MCA-3 was important to maintain PI(4,5)P₂ in coelomocytes (Bednarek et al., 2007). Thus, FLWR-1 and MCA-3 may have a shared role in the regulation of PI(4,5)P₂ levels. The question thus arises, whether FLWR-1 acts upstream of MCA-3 or vice versa or whether they act in parallel. Our results indicate that MCA-3 may be functionally affected in flwr-1 mutants explaining increased Ca²⁺ levels during stimulation (Brini and Carafoli, 2011; Chamberland et al., 2019). Alternatively, MCA-3 may undergo endo-exocytosis and this could be affected by effects of the flwr-1 mutation on endocytosis and recycling (Ono et al., 2019). The increased Ca²⁺ influx we observed was at first surprising, since work in Drosophila has yielded opposite results during 40 Hz electrical stimulation in neuromuscular boutons of Flower mutants (Yao et al., 2017). At stimulation with 10 Hz, however, these animals showed an increased GCaMP signal compared to wild type, in line with our observations. Potentially, increased excitability is only observed with lower, more physiological stimulation of neurons.

Even though we observed increased, instead of decreased, Ca²⁺ influx into the pre-synapse in flwr-1 mutants, we cannot conclude about the putative Ca²⁺-conducting activity of FLWR-1. However, it was discussed that the kinetics of Ca²⁺ influx induced by Flower in Drosophila are too slow and the extent too little as to directly trigger known modes of endocytosis (Brose and Neher, 2009; Xue et al., 2012; Leitz and Kavalali, 2016). Therefore, rather than having a function in directly coupling exo- and endocytosis, we suggest that FLWR-1 has a regulatory role in facilitating SV recycling, potentially by affecting MCA-3 activity. Increased Ca²⁺ at endocytic sites could, for example, lead to activation of the Ca²⁺ sensor calmodulin which has been shown to promote endocytosis in nerve terminals (Wu et al., 2009). However, whether the small and slow Ca²⁺ influx observed for Drosophila Flower is sufficient to activate calmodulin is unclear (Wu et al., 2009; Yao et al., 2009). To assess the possibility of FLWR-1 forming an ion channel, we used artificial intelligence-based protein structure prediction. Assuming similarity to Ca²⁺ channels, we asked AF3 to model FLWR-1 as a tetramer. The putative pore lining residues were mostly hydrophobic or apolar, thus not well in agreement with a function as an ion channel. The suggested similarity to the selectivity filter of VGCCs, including the sequence EGW or EAW (in FLWR-1 this is EAP; Yao et al. (2009)) was not confirmed in the AF3 models: While glutamate faces the extracellular end of the VGCC pore in all four channel modules, E74_FLWR-1 pointed away from the pore. We thus think that FLWR-1 is very unlikely to form an ion channel.

Should FLWR-1 exert an influence on PI(4,5)P₂ levels, this may activate PMCAs (Berrocal et al., 2017), which would remove Ca²⁺ and prevent sustained phospholipase C activity, thus protecting PI(4,5)P₂ at the PM from hydrolysis (Lopreiato et al., 2014; Penniston et al., 2014). Increased PI(4,5)P₂ levels could then activate different forms of endocytosis (Posor et al., 2015; Blumrich et al., 2023). Such interactions could further trigger a positive feedback loop leading to invagination of the PM (Yao et al., 2017). Both, PMCAs and Flower, were suggested to directly bind PI(4,5)P₂ with positively charged amino acid residues in areas close to their respective transmembrane regions (Filoteo et al., 1992; Li et al., 2020). In line with this, the FLWR-1 AF3 model exhibits several amino acids in positions that could interact with PI(4,5)P₂ headgroups in the membrane, similar to the interaction of potassium channels with PI(4,5)P₂. This could lead to a close arrangement of both proteins in PI(4,5)P₂ lipid microdomains (Katan and Cockcroft, 2020). Flower could also activate PMCAs in intracellular compartments after endocytosis. One PMCA variant was shown to be mainly localized to recycling SVs and endosomes and Ca²⁺ influx into these is an important factor in clearance from the presynapse (Ono et al., 2019). Overall, this may explain the increased number of recycling endosomes found in flwr-1 synapses. For example, effects of flwr-1 mutations on PI(4,5)P₂ at the PM may negatively affect bulk endocytosis. Yet effects on PI(4,5)P₂ at the endosome may have a stronger impact in slowing down breakdown of these endosomes (Fig. 8E). Still, our ultrastructural analysis contrasts EM analyses in Drosophila which yielded fewer, rather than more, bulk endosomal structures in animals lacking Flower (Yao et al., 2017; Li et al., 2020). However, this group used an unusually long stimulation period of ten minutes which may result in entirely different outcomes compared to our 30 s optogenetic stimulus. Alternatively, the subcellular localization of Flower function may be different between organisms, since, in C. elegans, loss of FLWR-1 seems to delay reformation of SVs from bulk endosomes.

Another possible explanation of how Flower could facilitate endocytosis was suggested by Yao et al. (2017), proposing that calcineurin may function downstream of Flower to stimulate bulk endosome formation. Calcineurin is a Ca²⁺- and calmodulin-dependent phosphatase that activates endocytic proteins by dephosphorylation, thus accelerating recycling of SVs (Cousin and Robinson, 2001; Kumashiro et al., 2005; Yamashita, 2012). The Ca²⁺ increase, driving calcineurin, would result from effects of FLWR-1 on PMCAs in our model. But how do the increased Ca²⁺ levels and the decreased rate of pHluorin decay we observe fit to this idea? One explanation could be that more Ca²⁺ and thus more activation of calcineurin induces a shift of the main endocytic mode to ADBE (Yamashita, 2012). Structures formed by bulk endocytosis are acidified more slowly than small SVs generated by CME, which could explain the delayed decay of pHluorin fluorescence (Gross and von Gersdorff, 2016; Okamoto et al., 2016). These observations are in line with the higher number of endosomal structures in flwr-1 mutants that we detected by EM. More pronounced neurotransmission, due to increased SV fusion, causes an increase in bulk invaginations (Clayton et al., 2008), yet the endosomes formed this way are not efficiently disassembled. The C. elegans homolog of calcineurin, TAX-6, is also involved in activity-dependent bulk endocytic processes, however in a different context (Miller-Fleming et al., 2016; Cuentas-Condori et al., 2023). Here, this protein is rather required for the regeneration of GABAergic synapses during maturation. Thus, the remote possibility that FLWR-1 also takes part in this process needs to be considered, implying that developmental aspects of FLWR-1 function may influence E/I balance even in adult NMJs. Jointly, these ideas suggest that calcineurin/TAX-6 could be upregulated, rather than downregulated, by the increased Ca²⁺ levels during neurotransmission in flwr-1 mutants.

In sum, our work confirms a conserved role of FLWR-1 in neurotransmission and SV recycling even though we observed differences to previous data from Drosophila NMJs. We found an unexpected upregulation of neurotransmission in flwr-1 mutants which has not yet been observed in other animal models. Further investigation will be required to confirm the involvement of MCA-3 and the nature of the relationship between these proteins.

Additional information

Author Contributions

MS created plasmids. MS and NS generated strains. MS performed light microscopy, swimming assays, pharmacological assays and body length measurements. SS conducted aldicarb assays. JR and DR prepared EM samples and JR conducted EM experiments. JL performed electrophysiology. MS and AG designed and coordinated the study. MS and AG wrote the manuscript. AG and SE supervised the work and acquired funding. All authors read and approved the final manuscript.

Funding

This project was funded by Deutsche Forschungsgemeinschaft, Collaborative Research Centre 1080 project B02 (grant DFG CRC1080/B2 to AG), as well as by core funding from Goethe University.

Acknowledgements

We are indebted to Katharina Kuhlmeier for her expert technical assistance, and members of the Gottschalk group for critical comments regarding the manuscript. We further thank Martina Rudgalvyte from the Glauser Lab (Université de Fribourg) for helpful comments. We would also like to show our gratitude to members of the Zimmer Lab (University of Vienna) for providing advice and plasmids for GCaMP imaging, and to the Wang Lab (University of Connecticut) for providing plasmids. Some deletion mutations used in this work were provided by the International C. elegans Gene Knockout Consortium (C. elegans Gene Knockout Facility at the Oklahoma Medical Research Foundation, which is funded by the National Institutes of Health, and the C. elegans Reverse Genetics Core Facility at the University of British Columbia, which is funded by the Canadian Institute for Health Research, Genome Canada, Genome B.C., the Michael Smith Foundation, and the National Institutes of Health). Finally, we thank the Caenorhabditis Genetics Center (CGC), which is funded by the NIH Office of Research Infrastructure Programs (P40 OD010440), for providing strains.

Supplemental figures

Fig. S1.

Fig. S2.

Fig. S3.

Fig. S4.

Fig. S5

Fig. S6.

Fig. S7.