1. Neuroscience

A positive feedback loop between Flower and PI(4,5)P2 at periactive zones controls bulk endocytosis in Drosophila

  1. Tsai-Ning Li
  2. Yu-Jung Chen
  3. Ting-Yi Lu
  4. You-Tung Wang
  5. Hsin-Chieh Lin
  6. Chi-Kuang Yao  Is a corresponding author
  1. Institute of Biological Chemistry, Academia Sinica, Taiwan
  2. Neuroscience Program of Academia Sinica, Academia Sinica, Taiwan
  3. Institute of Biochemical Sciences, College of Life Science, National Taiwan University, Taiwan
https://doi.org/10.7554/eLife.60125
Share this article
Cite this article
  1. Tsai-Ning Li
  2. Yu-Jung Chen
  3. Ting-Yi Lu
  4. You-Tung Wang
  5. Hsin-Chieh Lin
  6. Chi-Kuang Yao
(2020)
A positive feedback loop between Flower and PI(4,5)P2 at periactive zones controls bulk endocytosis in Drosophila
eLife 9:e60125.
https://doi.org/10.7554/eLife.60125
  2. Download BibTeX
  3. Download .RIS

Abstract

Synaptic vesicle (SV) endocytosis is coupled to exocytosis to maintain SV pool size and thus neurotransmitter release. Intense stimulation induces activity-dependent bulk endocytosis (ADBE) to recapture large quantities of SV constituents in large endosomes from which SVs reform. How these consecutive processes are spatiotemporally coordinated remains unknown. Here, we show that Flower Ca2+ channel-dependent phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) compartmentalization governs control of these processes in Drosophila. Strong stimuli trigger PI(4,5)P2 microdomain formation at periactive zones. Upon exocytosis, Flower translocates from SVs to periactive zones, where it increases PI(4,5)P2 levels via Ca2+ influxes. Remarkably, PI(4,5)P2 directly enhances Flower channel activity, thereby establishing a positive feedback loop for PI(4,5)P2 microdomain compartmentalization. PI(4,5)P2 microdomains drive ADBE and SV reformation from bulk endosomes. PI(4,5)P2 further retrieves Flower to bulk endosomes, terminating endocytosis. We propose that the interplay between Flower and PI(4,5)P2 is the crucial spatiotemporal cue that couples exocytosis to ADBE and subsequent SV reformation.

Introduction

Proper synaptic vesicle (SV) exocytosis dictates the robustness of brain activity. Coupling SV exocytosis with proper endocytosis is crucial for maintaining a balance of SV proteins at the release site, plasma membrane equilibrium, SV identity, and SV pool size (Chanaday et al., 2019; Haucke et al., 2011; Lou, 2018; Wu et al., 2014a). Currently, four modes of SV endocytosis are proposed. These differ in terms of stimulation intensity for their induction, formation, and molecular components (Haucke et al., 2011; Kononenko and Haucke, 2015; Wu et al., 2014a). Under mild neuronal stimulation, the SV partially fuses with the plasma membrane and reforms at the active zone, the so called ‘kiss and run’ mode. During clathrin-mediated endocytosis (CME), the SV fully collapses into the plasma membrane, followed by reformation of a single SV. Ultrafast endocytosis was also shown to recycle SVs at a sub-second timescale by forming a ~ 80 nm-sized bulk endosome predominantly at the edge of the active zone. SVs subsequently regenerate from this bulk endosome (Granseth et al., 2006; Watanabe et al., 2013a; Watanabe et al., 2013b; Zhu et al., 2009). High frequency stimulation and thus exocytosis could easily surpass the capacity of the three above-described endocytic mechanisms. It has therefore been proposed that activity-dependent bulk endocytosis (ADBE) has the necessary recapture capacity upon intense stimulation (Clayton et al., 2008; Soykan et al., 2017; Wu and Wu, 2007). This retrieval mode is elicited at the periactive zone to recapture large quantities of SV constituents via bulk endosome (~100–500 nm) formation, from which SVs subsequently reform. Hence, specific routes of SV recycling may fit the specific demands of a wide range of neuronal activities at the synapse. However, how stimulation intensity dictates the choice between these different endocytic modes is not well understood.

ADBE has been documented in many different types of neurons in invertebrates and vertebrates (Clayton et al., 2008; Heerssen et al., 2008; Heuser and Reese, 1973; Holt et al., 2003; Kasprowicz et al., 2008; Kittelmann et al., 2013; Miller and Heuser, 1984; Richards et al., 2000; Soykan et al., 2017; Stevens et al., 2012; Vijayakrishnan et al., 2009; Wenzel et al., 2012; Wu and Wu, 2007; Yao et al., 2017). Although the precise mechanism regulating ADBE is not known, multiple lines of evidence suggest that actin polymerization may serve as the membrane invagination force responsible for generating bulk endosomes (Gormal et al., 2015; Holt et al., 2003; Kokotos and Low, 2015; Nguyen et al., 2012; Richards et al., 2004; Soykan et al., 2017; Wu et al., 2016). Furthermore, phosphatidylinositol metabolism has also been implicated in controlling this recycling mode (Gaffield et al., 2011; Holt et al., 2003; Richards et al., 2004; Vijayakrishnan et al., 2009). Importantly, loss of Synaptojanin, the major phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) catalytic enzyme in neurons (Tsujishita et al., 2001), leads to reduced SV endocytosis elicited by intense stimulation, presumably by affecting ADBE (Mani et al., 2007). PI(4,5)P2 is known to promote actin polymerization by activating a number of actin modulators (Janmey et al., 2018). PI(4,5)P2 is clustered to form microdomains upon the demand for diverse cellular functions (Aoyagi et al., 2005; Chen et al., 2015; Honigmann et al., 2013; Kabeche et al., 2015; Mu et al., 2018; Picas et al., 2014; Riggi et al., 2018; van den Bogaart et al., 2011). It has also been reported that formation of PI(4,5)P2 microdomains precedes actin polymerization during a process reminiscent of ADBE in neurosecretory cells (Gormal et al., 2015). Thus, subcellular compartmentalization of PI(4,5)P2 may provide the spatial information dictating where bulk membranes will invaginate. However, the mechanism initiating the formation of the PI(4,5)P2 microdomains is unknown.

Coordinated SV protein and membrane retrieval plays an important role in maintaining the identity of newly formed SVs during SV recycling (Kaempf and Maritzen, 2017; McMahon and Boucrot, 2011; Saheki and De Camilli, 2012; Traub and Bonifacino, 2013). It has been well documented that proper sorting of SV proteins to the nascent SV is achieved in CME by the cooperative action of PI(4,5)P2 and adaptor protein complexes (Saheki and De Camilli, 2012; Traub and Bonifacino, 2013). Recent studies have also revealed a distinct sorting mechanism that retrieves selective SV cargoes to the bulk endosome via ADBE (Kokotos et al., 2018; Nicholson-Fish et al., 2015). Several lines of evidence further suggest that clathrin and adaptor protein complexes are required for reforming SVs from the bulk endosome (Cheung and Cousin, 2012; Glyvuk et al., 2010; Kokotos et al., 2018; Kononenko et al., 2014; Park et al., 2016). A dynaminI/dynaminIII/clathrin-independent mechanism has also been reported as being involved in this process (Wu et al., 2014c). Thus, during SV regeneration via ADBE, multiple protein sorting steps may be required to ensure that the SVs harbor the proper compositions of lipids and proteins, thereby endowing specific release probabilities in relation to other modes of endocytosis (Cheung et al., 2010; Hoopmann et al., 2010; Nicholson-Fish et al., 2015; Silm et al., 2019). The mechanism by which protein sorting and membrane retrieval are coordinated in this process remains to be explored.

Here, we show that, upon intense stimulation, PI(4,5)P2 is compartmentalized into microdomains at periactive zones in the synaptic boutons of Drosophila larval neuromuscular junctions (NMJs). Blockade of PI(4,5)P2 microdomain formation diminishes ADBE and SV reformation from the bulk endosome. Increased intracellular Ca2+ and SV exocytosis are prerequisites for initiating ADBE (Morton et al., 2015; Wu and Wu, 2007). We have previously shown that Flower (Fwe), a SV-associated Ca2+ channel, regulates both CME and ADBE, and that its channel activity is strongly activated upon intense stimulation to elicit ADBE (Yao et al., 2017). We show that Fwe initiates a positive feedback loop upon PI(4,5)P2 increase to ensure the formation of PI(4,5)P2 microdomains and thus trigger ADBE and subsequent SV reformation. Intriguingly, PI(4,5)P2 also participates in retrieval of Fwe to the bulk endosome, thereby stopping membrane recycling. Hence, spatiotemporal interplays between Flower and PI(4,5)P2 coordinate the retrieval of SV cargos and membranes, coupling exocytosis to ADBE and subsequent SV reformation.

Results

Intense neuronal activity induces formation of PI(4,5)P2 microdomains at the presynaptic periactive zone of Drosophila synapses

To investigate the dynamics of PI(4,5)P2 in the presynaptic compartment, we expressed a GFP fusion protein of the pleckstrin homology (PH) domain of PLCδ1 (PLCδ1-PH-EGFP) in synaptic boutons of Drosophila larval NMJs using nSyb-GAL4, a pan-neuronal driver. PLCδ1-PH-EGFP binds to PI(4,5)P2 with high affinity and is widely used to label subcellular compartments in which PI(4,5)P2 is enriched (Chen et al., 2014; Khuong et al., 2010). We delivered 20 Hz stimuli for three minutes to synaptic boutons in a 2 mM extracellular Ca2+ solution and performed live imaging in the third minute. Consecutive snapshot images were taken before stimulation, during the third minute of stimulation, and after stimulation. As shown in Figure 1a–b, we observed a very subtle increase in PLCδ1-PH-EGFP fluorescence in individual boutons (white arrows), similar to findings of a previous study (Verstreken et al., 2009). However, when we raised the stimulus intensity to 40 Hz, we recorded a robust increase in fluorescence relative to a GFP fusion protein of the plasma membrane-integrated mCD8 domain (UAS-mCD8-GFP). Fluorescence signals rapidly returned to basal levels within tens of seconds when the stimuli were removed. We have previously documented that treatment with 40 Hz electric pulses or 90 mM high KCl solution can cause comparable stimulation intensities in Drosophila NMJ boutons (Yao et al., 2017). High K+ treatment also increased the fluorescence signal of PLCδ1-PH-EGFP. No increase in the presynaptic protein level of PLCδ1-PH-EGFP was found under this condition (Figure 1—figure supplement 1a–b), arguing that this stimulation does not induce protein synthesis. These results suggest that, in response to intense stimulation, PLCδ1-PH-EGFP is redistributed and concentrated to PI(4,5)P2-enriched subdomains of the plasma membrane, thereby enhancing the overall fluorescence.

Figure 1 with 3 supplements see all
Download asset Open asset
PI(4,5)P2 forms microdomains at periactive zones under conditions of intense stimulation.

(a–b) Increased fluorescence of PLCδ1-PH-EGFP but not mCD8-GFP in NMJ boutons upon intense stimulation. (a) (Top) Live images of the boutons (arrows) expressing UAS-PLCδ1-PH-EGFP or UAS-mCD8-GFP. The larvae were reared at 25°C. Electrical (20 or 40 Hz) or chemical (90 mM K+) stimulation was conducted in a 2 mM-Ca2+ solution for 3 min (electrical) or 5 min (chemical) and then rested in 0 mM Ca2+ and 5 mM K+. Snapshot images taken before stimulation, at the third (electrical) and fifth (chemical) min of stimulation, and after stimulation. (Bottom) Traces of probe fluorescence for single boutons. The number of boutons imaged (N). (b) Quantification data for EGFP fluorescence change. The resting fluorescence level (F0). Fluorescence change evoked by stimulation (ΔF). (c–f) PLCδ1-PH-EGFP enrichment at periactive zones is dependent on Ca2+ upon intense stimulation. Single-plane confocal (c) or SIM (e) images of the boutons expressing PLCδ1-PH-EGFP. The larvae were reared at 25°C. The boutons subjected to high K+/2 mM Ca2+ (10 min stimulation of 90 mM K+/2 mM Ca2+), high K+/0 mM Ca2+ (10 min stimulation of 90 mM K+/0 mM Ca2+), or high K+/0.5 mM Ca2+ (1 min stimulation of 90 mM K+/0.5 mM Ca2+) treatments were fixed immediately and immunostained for PLCδ1-PH-EGFP (green) and Bruchpilot (Brp) [an active zone scaffold protein; magenta (c), red (e)]. The PLCδ1-PH-EGFP-enriched puncta (Arrows). (d) Quantification data for PLCδ1-PH-EGFP staining intensities, normalized to the value of the resting condition. (f) Average area of PLCδ1-PH-EGFP clusters in individual boutons, as measured by SIM. Individual data values are shown in graphs. p values: ns, not significant; *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. Mean ± SEM. Scale bar: 1 µm (e), 2 µm (a, c). Statistics: one-way ANOVA with Tukey’s post hoc test (b, d).

Figure 1—source data 1

Source data for Figure 1.

https://cdn.elifesciences.org/articles/60125/elife-60125-fig1-data1-v2.xlsx

To characterize the subcellular distribution of the stimulus-dependent PI(4,5)P2 induction, we conducted a chemical fixation protocol whereby the NMJ boutons were fixed at rest or immediately after high K+ stimulation and then immunostained for PLCδ1-PH-EGFP using an α-GFP antibody to enhance the signal for high-resolution microscopic imaging. Confocal images revealed that, for neurons at rest, native PLCδ1-PH-EGFP fluorescence was weakly detected on the presynaptic plasma membrane labeled by α-Hrp staining, with some additional fluorescent signal being dispersed in the cytosol (Figure 1—figure supplement 1c). We then elevated PI(4,5)P2 on the plasma membrane by removing a copy of synaptojanin (synj), which encodes the major neuronal PI(4,5)P2 phosphatase (Tsujishita et al., 2001; Verstreken et al., 2003). Consistent with previous studies (Chen et al., 2014; Verstreken et al., 2009), this reduction in Synj levels enhanced PLCδ1-PH-EGFP fluorescence (Figure 1—figure supplement 1c–d). In this context, we did not observe a significant change in protein expression of PLCδ1-PH-EGFP in the presynaptic compartment (Figure 1—figure supplement 1e–f). By using α-GFP immunostaining, these PLCδ1-PH-EGFP signals could be faithfully amplified (Figure 1—figure supplement 1c–d). Therefore, this immunostaining approach was subsequently used to monitor presynaptic PI(4,5)P2 levels.

Next, we stimulated the boutons expressing PLCδ1-PH-EGFP with 90 mM K+ and 2 mM Ca2+ for 10 min. Similar to our live-imaging results, PLCδ1-PH-EGFP signals were significantly increased on the presynaptic plasma membrane (Figure 1—figure supplement 1g–h). In particular, we observed high-level induction of PLCδ1-PH-EGFP puncta (Figure 1c–d). To assess the possibility that chemical fixation may have altered membrane properties and consequently plasma membrane PLCδ1-PH-EGFP clustering upon intense stimulation, we further examined the distributions of mCD8-GFP and PLCδ1-PHS39R-EGFP, a PI(4,5)P2-binding mutant (Khuong et al., 2010; Verstreken et al., 2009). There was no obvious change in the plasma membrane pattern of mCD8-GFP in fixed boutons stimulated with high K+ (Figure 1—figure supplement 2a–b). Unlike PLCδ1-PH-EGFP, PLCδ1-PHS39R-EGFP was found mainly in the cytosol at rest. After intense stimulation, immunostaining signals of PLCδ1-PHS39R-EGFP did not accumulate on the plasma membrane (Figure 1—figure supplement 2c–d) and were not increased within the boutons (Figure 1—figure supplement 2e). Thus, clustering of PLCδ1-PH-EGFP on the plasma membrane upon stimulation results from an increase in local PI(4,5)P2 concentration, although a potential effect of chemical fixation, if any, on PLCδ1-PH-EGFP clustering cannot be excluded. To assess potential dominant-negative effects of PI(4,5)P2 binding by PLCδ1-PH-EGFP on the stimulation-induced accumulation of PI(4,5)P2, we further examined the recruitment of the AP-2 complex by PI(4,5)P2. Similar to PLCδ1-PH-EGFP, levels of the α subunit of the AP-2 complex (AP-2α) were also increased on the plasma membrane upon high K+ stimulation (Figure 1—figure supplement 3a–b). Together, these data indicate that intense stimulation can promote the formation of PI(4,5)P2 microdomains.

We noted that the induced PI(4,5)P2 microdomains were primarily sited at periactive zones marked by activity-dependent localization of Eps15 (Figure 1c; Figure 1—figure supplement 3c; Koh et al., 2007; Winther et al., 2013). Periactive zones are known hot-spots for ADBE (Chanaday et al., 2019; Kononenko and Haucke, 2015; Wu et al., 2014a). By using structured illumination microscopy (SIM), we estimated that the average size of these PI(4,5)P2 microdomains formed after stimulation is ~300 nm in diameter (Figure 1e–f) (0.07655 ± 0.004 μm2, mean ± S.E.M., n = 37 boutons).

Next, to determine the Ca2+ dependence of the PI(4,5)P2 microdomains, we stimulated the boutons in a solution of 90 mM K+ and 0 mM Ca2+, which resulted in failure to induce PI(4,5)P2 microdomain formation (Figure 1d). We obtained a similar result using 1 min stimulation of 90 mM K+ and 0.5 mM Ca2+ (Figure 1d), which was previously shown to primarily elicit CME but not ADBE (Yao et al., 2017). Hence, these results suggest that intense stimulation can elicit Ca2+-driven compartmentalization of PI(4,5)P2 at the periactive zone in synaptic boutons of Drosophila NMJs.

PI(4,5)P2 microdomains are involved in ADBE initiation and SV reformation from bulk endosomes

Next, we investigated the function of PI(4,5)P2 microdomains in ADBE. High K+ treatment is widely used to trigger ADBE in a broad range of synapses (Akbergenova and Bykhovskaia, 2009; Clayton et al., 2008; Jin et al., 2019; Stevens et al., 2012; Vijayakrishnan et al., 2009; Wu and Wu, 2007; Wu et al., 2014c). To measure ADBE, we induced ADBE with 90 mM K+ and 2 mM Ca2+ for 10 min and then conducted transmission electron microscopy (TEM). ADBE was evoked in wild-type control boutons under these conditions and generated the formation of bulk endosomes (red asterisks, defined as >80 nm in diameter) (Figure 2c–d). To suppress the function of PI(4,5)P2, we expressed PLCδ1-PH-EGFP, anticipating that binding of the PLCδ1-PH domain to PI(4,5)P2 would restrict availability of PI(4,5)P2 to its effectors and metabolic enzymes (Figure 2a; Khuong et al., 2013). By using the GAL4/UAS system, we were able to adjust expression levels of the PLCδ1-PH domain via temperature manipulation (Brand and Perrimon, 1993; D'Avino and Thummel, 1999; Wilder, 2000). When we neuronally expressed PLCδ1-PH-EGFP using nSyb-GAL4 and grew larvae at 25°C, we found that mild expression of PLCδ1-PH-EGFP had a mild inhibitory effect on ADBE induction relative to wild-type control boutons (Figure 2c–d). In contrast, when larvae were grown at 29°C to effect a two-fold increase in PLCδ1-PH-EGFP expression (Figure 2—figure supplement 1a–b), ADBE was almost completely abolished (Figure 2c–d). However, when larvae expressing PLCδ1-PHS39R-EGFP were grown at 29°C, ADBE was not suppressed by the mutant protein (Figure 2c–d).

Figure 2 with 2 supplements see all
Download asset Open asset
PI(4,5)P2 microdomains drive ADBE and SV reformation from bulk endosomes.

Reducing PI(4,5)P2 availability suppresses ADBE and subsequent SV reformation. (a) A schematic for PI(4,5)P2 suppression by PLCδ1-PH-EGFP or Synj expression. (b) Expression of Synj reduced presynaptic plasma membrane PI(4,5)P2 upon high K+ treatment. The boutons co-expressing UAS-PLCδ1-PH-EGFP with UAS-RFP (control) or UAS-synj using nSyb-GAL4 were reared at 25°C and subjected to resting condition (10 min incubation of 5 mM K+/0 mM Ca2+) or high K+ stimulation (10 min stimulation of 90 mM K+/2 mM Ca2+), followed by α-GFP immunostaining. Single-plane confocal images of the boutons are shown in Figure 2—figure supplement 1c. Quantification data for PLCδ1-PH-EGFP staining intensity are shown, normalized to the value of the resting condition of controls. (c) TEM images of the boutons of controls (nSyb-GAL4/+ at 29°C), mild PLCδ1-PH-EGFP expression (nSyb-GAL4/UAS-PLCδ1-PH-EGFP at 25°C), high PLCδ1-PH-EGFP expression (nSyb-GAL4/UAS-PLCδ1-PH-EGFP at 29°C), high PLCδ1-PHS39R-EGFP expression (nSyb-GAL4/UAS-PLCδ1-PHS39R-EGFP at 29°C), mild Synj expression (nSyb-GAL4/UAS-synj at 25°C), or high Synj expression (nSyb-GAL4/UAS-synj at 29°C). At rest (10 min incubation of 5 mM K+/0 mM Ca2+). High K+ (10 min stimulation of 90 mM K+/2 mM Ca2+). 10 min recovery (10 min stimulation of 90 mM K+/2 mM Ca2+, followed by 10 min incubation of 5 mM K+/0 mM Ca2+). 20 min recovery (10 min stimulation of 90 mM K+/2 mM Ca2+, followed by 20 min incubation of 5 mM K+/0 mM Ca2+). Bulk endosomes (>80 nm in diameter, red asterisks). Mitochondria (mt). Quantification data for total number of bulk endosomes per bouton area (d). Individual data values are shown in graphs. p values: ns, not significant; *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. Mean ± SEM. Scale bar: 500 nm. Statistics: one-way ANOVA with Tukey’s post hoc test.

Figure 2—source data 1

Source data for Figure 2.

https://cdn.elifesciences.org/articles/60125/elife-60125-fig2-data1-v2.xlsx

Synj comprises a central 5-phosphatase domain that specifically dephosphorylates the 5' position of PI(4,5)P2 to produce PI(4)P (McPherson et al., 1996; Woscholski et al., 1997). In addition, an N-terminal Sac1 domain converts several phosphatidylinositides—including PI(3,5)P2, PI(3)P, and PI(4)P—to PI (Figure 2a; Guo et al., 1999). Indeed, overexpression of Synj reduced the formation of the PI(4,5)P2 microdomains induced by high K+ (Figure 2b; Figure 2—figure supplement 1c). Importantly, overexpression of Synj resulted in dosage-dependent inhibition of ADBE (Figure 2c–d), similar to the effect of increasing levels of the PLCδ1-PH domain. Together, these data suggest that the PI(4,5)P2 microdomains initiate ADBE to control SV membrane retrieval upon intense stimulation.

SVs regenerate from bulk endosomes within minutes of their formation (Cheung and Cousin, 2012; Glyvuk et al., 2010; Kononenko et al., 2014; Stevens et al., 2012; Wu et al., 2014c). Using an approach employed previously (Stevens et al., 2012), we treated the boutons with high K+ followed by an incubation in 5 mM K+ and 0 mM Ca2+ solution for 10 or 20 min, allowing the SVs to reform from the bulk endosomes. In controls (nSyb-GAL4), the number and area of the induced bulk endosomes reverted to almost basal levels within 10 min (Figure 2c–d; Figure 2—figure supplement 2), demonstrating proper SV reformation. Interestingly, upon expression of low levels of PLCδ1-PH-EGFP, the induced bulk endosomes remained after 10 min and 20 min recovery periods, yet ADBE was only partially impaired (Figure 2c–d; Figure 2—figure supplement 2). In contrast, SV reformation actively occurred in the boutons expressing high levels of PLCδ1-PHS39R-EGFP (Figure 2c–d; Figure 2—figure supplement 2). Furthermore, we observed a loss of SV reformation ability upon mild overexpression of Synj (Figure 2c–d; Figure 2—figure supplement 2). Hence, the formation of PI(4,5)P2 microdomains is essential for both ADBE and SV reformation from the bulk endosome, with the latter event being elicited by relatively high levels of PI(4,5)P2. Note that perturbation of PI(4,5)P2 microdomain formation did not cause SV accumulation on the bulk endosome during recovery periods, indicating that PI(4,5)P2 microdomains may play an early role in SV reformation from bulk endosomes.

PI(4,5)P2 microdomains are established via a positive feedback loop of fwe and PI(4,5)P2

We have previously shown that the SV-associated Ca2+ channel Flower (Fwe) elevates presynaptic Ca2+ levels in response to strong stimuli to trigger ADBE (Yao et al., 2009; Yao et al., 2017). Given the Ca2+ dependence of PI(4,5)P2 microdomains (Figure 1d), we hypothesized that exocytosis evoked by intense stimulation promotes Fwe clustering at periactive zones, where it may provide the Ca2+ influx to induce PI(4,5)P2 microdomain formation. To test this hypothesis, we first conducted a proximity ligation assay (PLA) (Söderberg et al., 2008) to investigate if there is a close association between Fwe and PI(4,5)P2 in response to stimulation. In our PLA (Figure 3a), UAS-Flag-Fwe-HA was expressed in a fwe mutant background to replace endogenous Fwe protein with a tagged protein, and expression of UAS-PLCδ1-PH-EGFP reported the localization of PI(4,5)P2. Primary antibodies against the HA tag and EGFP protein were used to detect interactions between Flag-Fwe-HA and PLCδ1-PH-EGFP. The PLA signal was low in resting boutons, whereas high K+ treatment significantly increased PLA signal intensity (Figure 3a–b). Therefore, these results suggest that Fwe and PI(4,5)P2 are closely colocalized when intense stimulation triggers PI(4,5)P2 microdomain formation.

Figure 3
Download asset Open asset
Fwe and PI(4,5)P2 form a positive feedback loop to establish PI(4,5)P2 microdomains.

(a–b) Fwe and PLCδ1-PH-EGFP interact upon intense stimulation. (a) (Left) A schematic for PLA. (Right) Z-projected confocal images of fwe mutant boutons co-expressing Flag-Fwe-HA and PLCδ1-PH-EGFP (nSyb-GAL4/UAS-Flag-fwe-HA/UAS-PLCδ1-PH-EGFP in fweDB25/DB56 at 25°C). After resting condition or high K+ stimulation, the boutons were subjected to PLA. α-GFP and α-HA stained for PLCδ1-PH-EGFP (green) and Flag-Fwe-HA (magenta), respectively. PLA signals (red). (b) Quantification data for PLA signal intensities, normalized to the value of the resting condition. (c–f) Loss of Fwe or its Ca2+ channel activity perturbs PI(4,5)P2 microdomain formation. (c and e). (c) Single-plane confocal images of the boutons neuronally expressing UAS-PLCδ1-PH-EGFP in fweDB25/+ (control) or fweDB25/DB56 at 25°C. The boutons were subjected to resting condition or high K+ stimulation, followed by α-GFP immunostaining. Quantification data for PLCδ1-PH-EGFP staining intensity are shown, normalized to the value of the resting condition of controls (d). (e) Single-plane confocal images of the boutons neuronally expressing lexAop2-PLCδ1-PH-EGFP using vglut-lexA in wild-type Fwe rescue boutons (vglut-lexA/lexAop2-PLCδ1-PH-EGFP, nSyb(w)-GAL4/UAS-flag-fwe-RB-HA in fweDB25/fweDB56 at 25°C), or FweE79Q rescue boutons (vglut-lexA/lexAop2-PLCδ1-PH-EGFP, nSyb(w)-GAL4/UAS-flag-fweE79Q-RB-HA in fweDB25/fweDB5 at 25°C). When the weak nSyb(w)-GAL4 driver drove low expression of UAS-fwe transgenes, this binary system (vglut-lexA/lexAop2-PLCδ1-PH-EGFP) was used to produce detectable levels of PLCδ1-PH-EGFP in boutons. Quantification data for PLCδ1-PH-EGFP staining intensity are shown, normalized to the value of the resting condition of Fwe-rescued boutons (f). (g–h) canA1 RNAi knockdown impairs PI(4,5)P2 microdomain formation. Single-plane confocal images of the boutons co-expressing UAS-PLCδ1-PH-EGFP with UAS-RFP (control), UAS-canA1-RNAi (TRiP.JF01871), or UAS-canA1-RNAi (FB4) using nSyb-GAL4. The larvae were reared at 25°C. After resting condition or high K+ stimulation, the boutons were stained with α-GFP. (h) Quantification data for PLCδ1-PH-EGFP staining intensities, normalized to the value of the resting condition of controls. (i–j) Blockade of the PI(4,5)P2 microdomains attenuates Fwe Ca2+ conductance. (i) Snapshot Ca2+ images of the boutons (arrows) expressing lexAop2-GCaMP6f using vglut-lexA. The larvae of control (w1118), fwe mutant (fweDB25/DB56), canA1 RNAi (JF) (nSyb-GAL4/UAS-canA1-RNAi (TRiP.JF01871)), canA1 RNAi (FB) (nSyb-GAL4/UAS-canA1-RNAi (FB4)), Synj overexpression (nSyb-GAL4/UAS-synj), PLCδ1-PH-APEX2-HA expression (vglut-lexA/LexAop2-PLCδ1-PH-APEX2-HA), or fwe mutant expressing PLCδ1-PH-APEX2-HA (vglut-lexA/LexAop2-PLCδ1-PH-APEX2-HA in fweDB25/DB56) were reared at 25°C. Imaging was taken in the fifth minute for one minute after high K+ (2 mM Ca2+) stimulation. (j) Quantification data for evoked Ca2+ level, normalized to the value of controls. Evoked Ca2+ levels are shown as the increase in GCaMP6f fluorescence under high K+ stimulation. Individual data values are shown in graphs. p values: ns, not significant; *p<0.05; **p<0.01; ****p<0.0001. Mean ± SEM. Scale bar: 2 µm. Statistics: Student t-test (b). One-way ANOVA with Tukey’s post hoc test (d, f, h, j).

Figure 3—source data 1

Source data for Figure 3.

https://cdn.elifesciences.org/articles/60125/elife-60125-fig3-data1-v2.xlsx

Next, we investigated the effect of loss of Fwe on PI(4,5)P2 microdomain formation. Basal levels of PI(4,5)P2 were not affected in the fwe mutant relative to control (Figure 3c–d). However, intense stimulation failed to elicit PI(4,5)P2 microdomain formation in the fwe mutant (Figure 3c–d), revealing a crucial role for Fwe in this process. To determine if the Ca2+ channel activity of Fwe is responsible for this activity, we conducted rescue experiments based on our previous report in which fwe mutant boutons exhibited expression of the wild-type Fwe transgene or the FweE79Q mutant transgene that has reduced Ca2+ conductance (Yao et al., 2017). We found that the wild-type Fwe transgene promoted PI(4,5)P2 microdomain formation, whereas the FweE79Q mutant transgene lost that ability (Figure 3e–f). Thus, Fwe triggers the formation of PI(4,5)P2 microdomains in a Ca2+ channel-dependent manner.

Calmodulin and Calcineurin are thought to be the Ca2+ sensors for ADBE (Evans and Cousin, 2007; Jin et al., 2019; Marks and McMahon, 1998; Sun et al., 2010; Wu et al., 2009; Wu et al., 2014b). The Calcineurin complex dephosphorylates PI(4,5)P2 metabolic enzymes, including Synj and Phosphatidylinositol 4-phosphate 5-kinase Iγ (PIPKIγ) (Cousin and Robinson, 2001; Lee et al., 2005; Lee et al., 2004; van den Bout and Divecha, 2009). Drosophila possesses three isoforms of the catalytic subunit of Calcineurin, i.e., CanA1, CanA-14F, and Pp2B-14D. CanA1 regulates development of Drosophila NMJ boutons (Wong et al., 2014). Therefore, we knocked down canA1 in neurons by using two independent UAS-canA1-RNAi constructs, UAS-canA1-RNAi(JF01871) (Wong et al., 2014) and UAS-canA1-RNAi(FB4) (Dijkers and O'Farrell, 2007). Similar to the effect of loss of Fwe, reducing CanA1 levels via expression of either RNAi construct greatly suppressed the formation of PI(4,5)P2 microdomains compared to the control (Figure 3g–h), indicating that Calcineurin mediates the Ca2+ influx conducted by Fwe to induce PI(4,5)P2 microdomains.

Our previous Ca2+ imaging data showed that intense stimulation activates Fwe to increase presynaptic Ca2+ concentrations (Yao et al., 2017). We measured intracellular Ca2+ levels with the Ca2+ indicator GCaMP6f (Chen et al., 2013), and found that whereas control boutons exhibited a robust increase in intracellular Ca2+ upon high K+ stimulation, fwe mutant boutons exhibit an impaired Ca2+ response (Figure 3i–j). Unexpectedly, canA1 knockdown also elicited the same deficient Ca2+ response (Figure 3i–j). Similar suppressive effects were obtained upon overexpressing Synj or the PLCδ1-PH domain (Figure 3i–j). Hence, Calcineurin activation may increase PI(4,5)P2 activity, which may in turn promote the Ca2+ channel activity of Fwe to further increase intracellular Ca2+ levels. To investigate the potential for such feedback regulation, we expressed the PLCδ1-PH domain in a loss of Fwe background. Expression of the PLCδ1-PH domain did not rescue the low Ca2+ concentration caused by the fwe mutation (Figure 3i–j), showing that the Ca2+ suppression exerted by the PLCδ1-PH domain indeed depends on Fwe. These findings support that a positive feedback loop involving Fwe and PI(4,5)P2 is responsible for the formation of PI(4,5)P2 microdomains.

PI(4,5)P2 gates fwe

A well-known function of PI(4,5)P2 is to modulate ion channel activity through its electrostatic binding to clustered positively-charged amino acids adjacent to the transmembrane domains of ion channels (Hille et al., 2015; Suh and Hille, 2008). Through protein alignment analysis, we found tandem positively-charged amino acids, including lysine (K) and arginine (R), in the intracellular juxta-transmembrane regions of Fwe. These residues are evolutionarily conserved in mice and humans (Figure 4a), whereas other cytosolic residues show poor conservation. This feature inspired us to test the potential impact of PI(4,5)P2 on the channel function of Fwe. To determine direct interaction between Fwe and PI(4,5)P2, we conducted a nanoluciferase (Nluc)-based bioluminance resonance energy transfer (BRET) assay (Cabanos et al., 2017). In our BRET assay (Figure 4a), upon ion channel binding of BODIPY-TMR-conjugated PI(4,5)P2, illumination of Nluc-fused ion channels wrapped in detergent-formed micelles can excite BODIPY-TMR-conjugated PI(4,5)P2 to emit a BRET signal. As shown in Figure 4b, we reconstituted the micelles containing purified Nluc-Fwe-1D4 fusion proteins and, after adding BODIPY-TMR-PI(4,5)P2 and furimazine (a Nluc substrate), we observed a remarkable increase in BRET signal. Excess cold PI(4,5)P2 reduced the signal to ~50% of the BRET signals by competing for the PI(4,5)P2 binding sites in Fwe, suggesting direct PI(4,5)P2 binding to Fwe. To assess the involvement of the positively-charged amino acids of Fwe in PI(4,5)P2 binding, we mutated all of the clustered positively-charged amino acids to non-charged alanine to eliminate the electrostatic interactions. Residue substitution resulted in a significant reduction in BRET signal, comparable to the competitive effect attributable to provision of excess cold PI(4,5)P2. Therefore, the majority of the PI(4,5)P2 binding activity of Fwe is mediated by these positively-charged amino acids. Specifically, alanine substitution of residues in both the middle (K95, K100, R105) and C-terminal (K146, K147, R150) regions of Fwe reduced PI(4,5)P2-specific binding (Figure 4b–c). Moreover, when N-terminal residues (K29, R33) of Fwe were further substituted with alanines, binding of PI(4,5)P2 was also reduced (Figure 4b–c). Hence, these in vitro assays reveal that Fwe directly binds PI(4,5)P2 through multiple regions.

Figure 4 with 2 supplements see all
Download asset Open asset
PI(4,5)P2 binds to Fwe and promotes its Ca2+ channelling activity.

(a–c) PI(4,5)P2 binds Fwe. (a) A schematic of the Fwe structure and a BRET assay. Stars highlight conserved lysine (K) and arginine (R) in juxta-transmembrane regions: N-terminus (N), middle-domain (M), and C-terminus (C). Drosophila melanogaster (Dm). Mus Musculus (Ms). Human Sapiens (Hs). N-terminal fusion of Nluc to Fwe allows detection of PI(4,5)P2 binding. A 1D4 epitope was used for protein purification. (b) Nluc-Fwe-1D4 in micelles excites BODIPY-TMR-conjugated PI(4,5)P2 to emit BRET signal, which is decreased by competitive cold PI(4,5)P2 (1 mM). Alanine substitution of positively-charged residues in all regions (N/M/C > A), both middle-domain and C-terminus (M/C > A), or N-terminus (N > A) reduced BRET signals. (c) Corresponding PI(4,5)P2 binding ability was calculated by subtracting the signal values of N/M/C > A, M/C > A or N > A from that for WT. Quantification data was normalized to the signal value of all mutations (N/M/C). (d) PI(4,5)P2 controls Fwe channel activity. Snapshot Ca2+ images of the boutons (arrows) expressing lexAop2-GCaMP6f using vglut-lexA. fwe mutant (fweDB25/DB56). HA-Fwe[WT]-APEX2 rescue (nSyb-GAL4/UAS-HA-Fwe[WT]-APEX2 in fweDB25/DB56). HA-Fwe[K29A/R33A]-APEX2 rescue (nSyb-GAL4/UAS-HA-Fwe[K29A/R33A]-APEX2 in fweDB25/DB56). The larvae were reared at 25°C. Images were taken in the fifth minute for one minute after high K+ stimulation. (e) Quantification data for evoked Ca2+ level, normalized to the value of HA-Fwe-APEX2 rescue larvae. (f) Single-plane confocal images of the boutons expressing UAS-PLCδ1-PH-EGFP. The larvae were reared at 25°C. After resting conditions or high K+ stimulation, the boutons were stained with α-GFP. (g) Quantification data for the PLCδ1-PH-EGFP staining intensity, normalized to the value of the resting condition of HA-Fwe-APEX2 rescue larvae. Individual data values are shown in graphs. p values: ns, not significant; *p<0.05; ***p<0.001; ****p<0.0001. Mean ± SEM. Scale bar: 2 µm (d, f). Statistics: Student t-test (b). One-way ANOVA with Tukey’s post hoc test (b, c, e, g).

Figure 4—source data 1

Source data for Figure 4.

https://cdn.elifesciences.org/articles/60125/elife-60125-fig4-data1-v2.xlsx

To directly test how PI(4,5)P2 affects Flower Ca2+ channel function, we generated UAS transgenes for the Fwe variants in which all or subsets of positively-charged amino acids were mutated to alanine and performed a mutant rescue experiment using nSyb-GAL4. Mutations of all nine residues or only those in the middle region (K95/K100/R105) led to very low protein expression levels, preventing further study. However, alanine substitution of C-terminal residues K146/K147/R150 did not affect SV localization of Fwe or its ability to regulate presynaptic Ca2+ concentration and induce PI(4,5)P2 microdomain formation (Figure 4—figure supplement 1a–f), suggesting that these residues do not play a regulatory role in Fwe channel activity. We also mutated the N-terminal residues K29/R33. The resulting K29A/R33A variant was still able to properly localize to presynaptic terminals (Figure 4—figure supplement 2a–b). However, upon high K+ stimulation, the K29A/R33A variant lost that ability to maintain proper intracellular Ca2+ levels (Figure 4d–e). Moreover, that variant failed to promote PI(4,5)P2 microdomain formation upon high K+ stimulation (Figure 4f–g). These results reveal that the positive feedback loop involving Fwe and PI(4,5)P2 relies on PI(4,5)P2-dependent gating control of Fwe.

Blockade of the positive feedback loop reduces ADBE and SV reformation from bulk endosomes

Next, we assessed the impact of the Fwe and PI(4,5)P2 regulatory feedback loop on ADBE. Loss of Fwe severely impaired formation of bulk endosomes induced by ADBE under high K+ conditions compared to the fwe mutant rescue control (Figure 5a–b), consistent with our previous findings (Yao et al., 2017). Next, we found that expression of wild-type Fwe protein or the K146A/K147A/R150A mutant variant restored proper ADBE in the fwe mutant background (Figure 5a–b; Figure 4—figure supplement 1g–h), whereas expression of the K29A/R33A variant failed to rescue ADBE deficiency (Figure 5a–b). Consistent with the suppressive effects caused by expression of the PLCδ1-PH domain or Synj (Figure 2), all of the bulk endosomes that remained in boutons lacking fwe could not generate new SVs during a 10 min or even 20 min recovery period (Figure 5a–b; Figure 5—figure supplement 1). Expression of wild-type Fwe protein but not the K29A/R33A mutant variant rescued this defect (Figure 5a–b; Figure 5—figure supplement 1).

Figure 5 with 1 supplement see all
Download asset Open asset
Perturbations of the positive feedback loop involving Fwe and PI(4,5)P2 suppress ADBE and SV reformation from bulk endosomes.

(a) TEM images of the boutons of HA-Fwe-APEX2 rescue (nSyb-GAL4/UAS-HA-Fwe-APEX2 in fweDB25/DB56 at 25°C), fwe mutant (fweDB25/DB56 at 25°C), HA-Fwe[K29A/R33A]-APEX2 rescue (nSyb-GAL4/UAS-HA-Fwe[K29A/R33A]-APEX2 in fweDB25/DB56 at 25°C), control (nSyb-GAL4 at 25°C), canA1 RNAi (JF) knockdown (nSyb-GAL4/UAS-canA1-RNAi (TRiP.JF01871) at 29°C), canA1 RNAi (JF) knockdown plus Fwe overexpression (nSyb-GAL4/UAS-canA1-RNAi (TRiP.JF01871)/UAS-HA-Fwe-APEX2 at 29°C). TEM processing was performed after the following treatments: at rest (10 min incubation of 5 mM K+/0 mM Ca2+); high K+ stimulation (10 min stimulation of 90 mM K+/2 mM Ca2+); 10 min recovery (10 min stimulation of 90 mM K+/2 mM Ca2+, followed by 10 min incubation of 5 mM K+/0 mM Ca2+); or 20 min recovery (10 min stimulation of 90 mM K+/2 mM Ca2+, followed by 20 min incubation of 5 mM K+/0 mM Ca2+). Bulk endosomes (>80 nm in diameter, red asterisks). Mitochondria (mt). (b) Quantification data of total numbers of bulk endosomes per bouton area. Individual data values are shown in graphs. p values: ns, not significant; *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. Mean ± SEM. Scale bar: 500 nm. Statistics: one-way ANOVA with Tukey’s post hoc test.

Figure 5—source data 1

Source data for Figure 5.

https://cdn.elifesciences.org/articles/60125/elife-60125-fig5-data1-v2.xlsx

Neuronal canA1 knockdown by expressing canA1 RNAi construct also impaired ADBE relative to the nSyb-GAL4 control (Figure 5a–b). To verify if CanA1 regulates ADBE in an Fwe-dependent manner, we overexpressed Fwe to increase the intracellular Ca2+ levels under canA1 RNAi knockdown conditions. Fwe overexpression significantly reversed the ADBE defect (Figure 5a–b), suggesting that increasing Fwe-dependent Ca2+ influx can augment activation of the residual CanA1 enzymes and thus normalize downstream ADBE. Taken together with our results reported in previous sections, we propose that the positive feedback loop involving Fwe, CanA1 and PI(4,5)P2 compartmentalizes PI(4,5)P2 microdomains at the periactive zone of boutons to dictate and coordinate ADBE and subsequent SV reformation.

PI(4,5)P2 facilitates retrieval of fwe to bulk endosomes

It was reported recently that a SV protein sorting process occurs during ADBE (Kokotos et al., 2018; Nicholson-Fish et al., 2015). VAMP4, a v-SNARE protein, is essential for ADBE to proceed, and it is selectively retrieved by ADBE (Kokotos et al., 2018; Nicholson-Fish et al., 2015). Given the important role of Fwe in triggering ADBE, we wondered if Fwe is sorted to bulk endosomes during ADBE. To visualize the vesicular localization of Fwe, we rescued the fwe mutant by expressing a UAS transgene of the APEX2 fusion protein of Fwe (UAS-HA-Fwe-APEX2), and then conducted diaminobenzidine (DAB) labeling and TEM. By means of confocal microscopy, we observed that HA-Fwe-APEX2 immunostaining signals were properly present on the SVs marked by Syt and Csp immunostaining (Figure 6—figure supplement 1a). Furthermore, expression of this fusion protein was able to rescue the endocytic defects (Figures 4 and 5) and early animal lethality (not shown) caused by loss of fwe. Therefore, HA-Fwe-APEX2 is functionally equivalent to endogenous Fwe. APEX2 is an engineered peroxidase that is capable of catalyzing DAB polymerization and proximal deposition, with the DAB polymers binding electron-dense osmium to enhance electron microscopy contrast (Lam et al., 2015). Whereas no DAB staining signals were observed in SVs in the Flag-Fwe-HA rescue control boutons stimulated by high K+ (yellow arrows, Figure 6—figure supplement 1b'; Figure 6a), the SV-based localization of HA-Fwe-APEX2 was clearly revealed by DAB staining in HA-Fwe-APEX2-rescued boutons under the resting and stimulation conditions (yellow arrows, Figure 6—figure supplement 1c–d). Moreover, specific DAB labeling also revealed the presynaptic plasma membrane localization of HA-Fwe-APEX2 (white arrows, Figure 6—figure supplement 1b–c), an outcome consistent with previous immunogold staining assays on endogenous Fwe (Yao et al., 2009). Upon high K+ stimulation, SV and plasma membrane localizations were retained (Figure 6—figure supplement 1d'). Remarkably, DAB signals were apparent on all bulk endosomes (Figure 6b–c; Figure 6—figure supplement 1d and d'). To minimize staining variability across boutons, we compared DAB staining intensities on bulk endosomes and SVs in the same boutons. Compared to Flag-Fwe-HA rescue boutons, in HA-Fwe-APEX2 rescue boutons, the staining intensities of bulk endosomes were more abundant compared to those of the surrounding SVs (Figure 6f), revealing a mechanism by which Fwe is recycled to bulk endosomes after it initiates ADBE.

Figure 6 with 1 supplement see all
Download asset Open asset
PI(4,5)P2 facilitates recycling of Fwe to bulk endosomes.

(a–f) PI(4,5)P2 retrieves Fwe to bulk endosomes. (a–e) TEM images of the boutons of Flag-Fwe-HA rescue (nSyb-GAL4/UAS-Flag-Fwe-HA in fweDB25/DB56 at 25°C, a), HA-Fwe-APEX2 rescue (nSyb-GAL4/UAS-HA-Fwe-APEX2 in fweDB25/DB56 at 25°C, b–c), or HA-Fwe-APEX2 rescue coexpressing Synj (nSyb-GAL4/UAS-HA-Fwe-APEX2/UAS-synj in fweDB25/DB56 at 25°C, d–e). After high K+ stimulation, the boutons were subjected to DAB labeling. (a) The Flag-Fwe-HA rescue boutons presented DAB-negative bulk endosomes (red asterisks) and SVs (yellow arrows). (b and c) In the HA-Fwe-APEX2 rescue boutons, DAB signals on bulk endosomes (red asterisks) were higher than those on the SVs (yellow arrows). The views in b and c are from different boutons. (d and e) Under the condition of Synj overexpression, perturbation of PI(4,5)P2 microdomain formation predominantly reduced DAB levels on the bulk endosomes (red asterisks). Views in d and e are from different boutons. (f) Quantification data for the DAB staining intensity ratio of bulk endosomes to surrounding SVs. The number of bulk endosomes, NMJ boutons, and larvae counted (Flag-Fwe-HA rescue control): Bulk endosomes (n = 33) derived from 5 NMJ boutons of two different larvae; HA-Fwe-APEX2 rescue: Bulk endosomes (n = 47) derived from 6 NMJ boutons of two different larvae; HA-Fwe-APEX2 rescue expressing Synj: Bulk endosomes (n = 59) derived from 9 NMJ boutons of two different larvae. Individual values were shown in graphs. p values: ns, not significant; ****p<0.0001. Mean ± SEM. Scale bar: 200 nm (a–e). Statistics: one-way ANOVA with Tukey’s post hoc test.

Figure 6—source data 1

Source data for Figure 6.

https://cdn.elifesciences.org/articles/60125/elife-60125-fig6-data1-v2.xlsx

PI(4,5)P2 is known to recruit adaptor protein complexes to sort SV proteins to the nascent SV during CME (Saheki and De Camilli, 2012). Next, we assessed if PI(4,5)P2 microdomain formation may be involved in sorting of Fwe to bulk endosomes. When PI(4,5)P2 microdomains were perturbed by Synj overexpression (Figure 2b), the bulk endosome localization of HA-Fwe-APEX2 was significantly reduced (Figure 6d–f), whereas there was only a mild reduction in its SV localization (Figure 6d–e). Therefore, in addition to initiating ADBE, PI(4,5)P2 microdomains play a role in facilitating the retrieval of Fwe to the bulk endosome, enabling ADBE to remove its trigger via a negative feedback regulatory mechanism and reducing endocytosis to prevent excess membrane uptake.

Discussion

ADBE occurs immediately after exocytosis to retrieve required SV protein and lipid constituents to further regenerate SVs under conditions of high-frequency stimulations. Here, we show that the Fwe Ca2+ channel-dependent compartmentalization of PI(4,5)P2 orchestrates coupling of exocytosis to ADBE and subsequent SV reformation. Based on our findings, we propose a model for this interplay (depicted in Figure 7). Under conditions of strong stimulation, SV exocytosis transfers Fwe from SVs to the periactive zone, where some of the activated Fwe provides the low Ca2+ levels that initiate Calcineurin activation to upregulate PI(4,5)P2 (Step 1). Increased PI(4,5)P2 enhances Fwe Ca2+ channel activity, thereby establishing a positive feedback loop that induces PI(4,5)P2 microdomain formation (Step 2). High levels of PI(4,5)P2 within these microdomains elicit bulk membrane invagination by triggering actin polymerization (Step 3). In parallel, PI(4,5)P2 facilitates proper retrieval of Fwe to the bulk endosome (Step 4), thereby terminating the ADBE process. Finally, PI(4,5)P2 microdomains dictate SV reformation from the bulk endosomes (Step 5), coordinating ADBE and subsequent SV reformation.

Figure 7
Download asset Open asset
A proposed model for the role of Fwe-dependent PI(4,5)P2 microdomains in coordinating ADBE and SV reformation from bulk endosomes.

Details are described in the Discussion section.

Fwe-dependent PI(4,5)P2 microdomains trigger ADBE

The role of actin polymerization in ADBE has been investigated in mammals (Kononenko et al., 2014; Soykan et al., 2017; Wu et al., 2016), as well as in Drosophila (Akbergenova and Bykhovskaia, 2009). PI(4,5)P2 is known to control a range of actin regulators, thereby modulating the dynamics of actin polymerization and branching (Janmey et al., 2018). It has been shown previously that, in response to nicotine stimulation, PI(4,5)P2 forms clustered microdomains of sub-micrometer scale prior to the appearance of an actin-based ring structure in bovine chromaffin cells (Gormal et al., 2015). In agreement with this observation, we show that intense activity stimulation drives the formation of PI(4,5)P2 microdomains at the periactive zone of Drosophila NMJ synaptic boutons. Perturbations of the formation of these microdomains reduces ADBE activity very significantly, demonstrating that rapid accumulation of PI(4,5)P2 in microdomains is needed to trigger extensive actin polymerization, which likely generates sufficient mechanical force to produce the large endosomes. Furthermore, loss of fwe or RNAi-mediated calcineurin knockdown effectively inhibited PI(4,5)P2 microdomain formation and, as a consequence, ADBE. Those results are consistent with previous data supporting that Ca2+ promotes ADBE by activating its sensor Calcineurin (Cousin and Robinson, 2001; Jin et al., 2019; Sun et al., 2010; Wu et al., 2009; Wu et al., 2014b; Xue et al., 2011).

Our data provide evidence that Fwe-derived Ca2+ regulates PI(4,5)P2 dynamics through activation of Calcineurin. Phosphatidylinositol 4-phosphate 5-kinases (PIP5Ks) are the major kinases that promote PI(4,5)P2 production from precursor phosphoinositides (van den Bout and Divecha, 2009). At mammalian central synapses, PIP5Kγ661, one of the PIP5Kγ isoforms, is the most abundant relative to other isoforms (Wenk et al., 2001). Activation of PIP5Ks can be engaged by binding to its regulators (van den Bout and Divecha, 2009). In response to Ca2+ activation, Calcineurin has been shown to dephosphorylate PIP5Kγ661 to promote its interaction with AP-2 complexes, which augments the enzymatic activity of PIP5Kγ661 (Nakano-Kobayashi et al., 2007). Synj is the major PI(4,5)P2 phosphatase in neurons (Tsujishita et al., 2001) and it is involved in the SV endocytosis prompted by strong stimulation (Mani et al., 2007). Whereas phosphorylation of Synj by CDK5 inhibits its activity, Calcineurin enhances Synj activity via dephosphorylation (Lee et al., 2004). In yeasts, cells undergo bulk membrane invagination, a process reminiscent of neuronal ADBE, under conditions of hyperosmotic stress (Guiney et al., 2015). However, in that process, Calcineurin dephosphorylates Synj to alter its association with other endocytic partners rather than affecting its enzymatic activity to regulate PI(4,5)P2 distribution. Intriguingly, divalent Ca2+ ions are known to control the lateral organization of PI(4,5)P2, further compartmentalizing PI(4,5)P2 into ~70 nm-sized microdomains in monolayers of lipid bilayers via electrostatic interactions between Ca2+ and PI(4,5)P2 (Carvalho et al., 2008; Ellenbroek et al., 2011; Levental et al., 2009; Sarmento et al., 2014; Wang et al., 2012; Wen et al., 2018). Similarly-sized PI(4,5)P2 clusters have been observed in PC12 cells under non-stimulated conditions (van den Bogaart et al., 2011). Therefore, tight spatial and temporal control of the localizations and activities of Fwe, Calcineurin, PIP5K, and Synj may drive Ca2+-mediated PI(4,5)P2 clustering, perhaps accounting for the Fwe-dependent formation of PI(4,5)P2 microdomains at the periactive zone prior to ADBE. Further investigations are needed to characterize the underlying mechanisms.

Our data show that direct binding of PI(4,5)P2 is required for the Ca2+ channel activity of Fwe. Perturbation of PI(4,5)P2-Fwe binding further impaired the formation of PI(4,5)P2 microdomains as well as ADBE initiation. Hence, PI(4,5)P2 controls Fwe gating, so that Fwe can promote PI(4,5)P2 compartmentalization through positive feedback regulation. Furthermore, loss of Fwe impaired the intracellular Ca2+ increase that was evoked upon strong activity stimulation (Figure 3g; Yao et al., 2017). These results support that, in addition to PI(4,5)P2, the channel function of Fwe may be gated by a significant change in membrane potential. Expanding on that notion, it is therefore possible that both factors may gate Fwe, thereby only allowing channel opening when exocytosis directs Fwe to periactive zones. Future studies should explore the details of this channel gating mechanism.

PI(4,5)P2 microdomains coordinate retrieval of SV membranes and proteins for SV reformation via ADBE

Since ADBE is triggered very rapidly by intense stimuli, it was thought that this type of recycling randomly retrieves SV proteins and that the sorting process takes place when SVs regenerate from bulk endosomes. However, recent work has highlighted a distinct retrieval route for SV proteins during ADBE (Kokotos et al., 2018; Nicholson-Fish et al., 2015). Very little is known about the mechanisms underlying that retrieval route. Interestingly, removing VAMP4 or mutating its di-leucine motif was shown to impair ADBE (Nicholson-Fish et al., 2015). The di-leucine motif of transmembrane proteins is known to mediate binding with the AP-2 adaptor complex (Traub and Bonifacino, 2013). Given that the AP-2 adaptor complex works closely with PI(4,5)P2 (McMahon and Boucrot, 2011), these findings imply a role for PI(4,5)P2 and the AP-2 adaptor complex in SV protein sorting via ADBE. Indeed, our data show that bulk endosomes recycle few in a PI(4,5)P2 microdomain-dependent manner. Hence, in addition to initiating ADBE, PI(4,5)P2 may participate in SV protein sorting to bulk endosomes.

SV regeneration occurs following formation of the bulk endosome. Our results also show that either removing Fwe-derived Ca2+ or perturbing PI(4,5)P2 activity impaired the ability of SVs to reform from the bulk endosome, highlighting the essential role of PI(4,5)P2 microdomains in this process. How could PI(4,5)P2 of the plasma membrane regulate subsequent SV reformation? It has been shown that PI(4,5)P2 is rapidly downregulated on bulk endosomes once formed by ADBE (Chang-Ileto et al., 2011; Cremona et al., 1999; Milosevic et al., 2011). It is conceivable that the high concentrations of PI(4,5)P2 in microdomains may compensate for rapid turnover, thereby ensuring appropriate concentrations of PI(4,5)P2 or PI(4)P for further recruitment of clathrin and adaptor protein complexes, such as AP-1 and AP-2 (Blumstein et al., 2001; Cheung and Cousin, 2012; Faúndez et al., 1998; Glyvuk et al., 2010; Kokotos et al., 2018; Kononenko et al., 2014; Park et al., 2016). Alternatively, PI(4,5)P2 may facilitate SV protein sorting prior to ADBE, meaning proper SV protein compositions on bulk endosomes could control recruitment of adaptor protein complexes. Both of these potential mechanisms are not mutually exclusive and may operate in parallel. Therefore, we propose that the Fwe-dependent formation of PI(4,5)P2 microdomains is potentially important in coordinating retrieval of SV membranes and cargos when SVs are recycled via ADBE. Notably, the Fwe channel is evolutionarily conserved from yeast to human (Yao et al., 2009). We have also previously demonstrated conserved functions of Fwe in CME and ADBE at the mammalian central synapse (Yao et al., 2017). A recent study has also identified Fwe as a key protein mediating Ca2+-dependent granule endocytosis in mouse cytotoxic T lymphocytes (Chang et al., 2018). Hence, we hypothesize that the mechanism of ADBE we report here may be generally deployed across synapses and species, even in other non-neuronal cells.

Materials and methods

Drosophila strains and genetics

Request a detailed protocol

fwe mutants and transgenes: fweDB25 and fweDB56 (Yao et al., 2009); UAS-Flag-Fwe-HA (Yao et al., 2009); UAS-Flag-Fwe[E79Q]-HA (Yao et al., 2017); UAS-canA1-RNAi(FB4) (Dijkers and O'Farrell, 2007); UAS-canA1-RNAi(TRiP.JF01871) (Bloomington Drosophila Stock Center, BDSC#25850); UAS-synj (Khuong et al., 2013); UAS-PLCδ1-PH-EGFP (Verstreken et al., 2009) (Bloomington Drosophila Stock Center, BDSC#39693); UAS-PLCδ1-PHS39R-EGFP (Verstreken et al., 2009) (Bloomington Drosophila Stock Center, BDSC#39694); nSyb-GAL4 (Pauli et al., 2008); UAS-mCD8-EGFP (Kyoto Stock Center, DGRC#108068); vglut-lexA (Bloomington Drosophila Stock Center, BDSC #60314); 13XLexAop2-IVS-GCaMP6f (Bloomington Drosophila Stock Center, BDSC #44277)); synj1 (Verstreken et al., 2003) (Bloomington Drosophila Stock Center, BDSC#24883; LexAop2-PLCδ1-PH-APEX2-HA (This paper); and LexAop2-PLCδ1-PH-EGFP (This paper). Fly stocks were reared on regular food at 25°C or as otherwise indicated.

Molecular cloning and transgenesis

Request a detailed protocol

For pUAST-HA-Fwe[WT]-APEX2, the coding sequence of the Fwe B isoform and APEX2 were separately PCR-amplified from pUAST-Flag-Fwe-HA (Yao et al., 2009) or pcDNA3 APEX2-NES (addgene #49386), respectively, and then subcloned into the pUAST vector. pUAST-HA-Fwe[K29A/R33A]-APEX2-HA was generated from pUAST-HA-Fwe[WT]-APEX2 through site-directed mutagenesis. pUAST-Flag-Fwe [K29/R33/K95/K100/R105/K146A/K147A/R150A]-HA, pUAST-Flag-Fwe[K146A/K147A/R150A]-HA and pUAST-Flag-Fwe[K95/K100/R105A]-HA were generated from pUAST-Flag-Fwe-HA through site-directed mutagenesis. For plexAop2-PLCδ1-PH-APEX2-HA, the coding sequence of the PLCδ1-PH domain and APEX2-HA were separately PCR-amplified and then subcloned into pJFRC19-13XLexAop2-IVS-myr:GFP (addgene #26224). For plexAop2-PLCδ1-PH-EGFP, the coding sequence of PLCδ1-PH-EGFP was PCR-amplified from genomic DNA of the UAS-PLCδ1-PHS39R-EGFP fly stock and then subcloned into pJFRC19-13XLexAop2-IVS-myr:GFP. For YeMP-Nluc-Fwe-1D4, the cDNA fragment of Nluc and Fwe-1D4 was PCR-amplified from pNL1.1[Nluc] (a gift from Yi-Shiuan Huang) and YeMP-Fwe-1D4 plasmid (Yao et al., 2009), respectively, and then subcloned into the yeast expression YeMP vector. For YeMP-Nluc-Fwe[K29A/R33A]−1D4, YeMP-Nluc-Fwe[K95/K100/R105/K146/K147/R150A]−1D4, and YeMP-Nluc-Fwe[K29/R33/K95/K100/R105/K146/K147/R150A]−1D4, the DNA fragments of the Fwe variant were PCR-amplified from pUAST plasmids and sublconed into the YeMP-Nluc-Fwe-1D4 vector. PCR primers are indicated in key source table (Appendix 1). Transgenic flies were made by Wellgenetics Inc.

Immunohistochemistry

Request a detailed protocol

Third instar larvae were fixed with 4% paraformaldehyde for 20 min. We used 1xPBS buffer containing 0.1% Tween-20 to stain the HA-tagged Fwe proteins. We used 1xPBS buffer containing 0.1% Triton X 100 to stain PLCδ1-PH-EGFP or AP-2α. We used 1xPBS buffer containing 0.2% Triton X 100 to stain GCaMP6f. Primary antibodies were used as follows: guinea pig α-Fwe B isoform (1:400) (Yao et al., 2017), chicken α-GFP (Invitrogen, 1:500); mouse α-HA (Sigma, 1:400), mouse α-Bruchpilot (Developmental Studies Hybridoma Bank nc82, 1:100); rabbit α-AP-2α (1:3000) (González-Gaitán and Jäckle, 1997) guinea pig α-Eps15 (1:3000) (Koh et al., 2007); rabbit α-HRP conjugated with Alexa Fluor 488, Cy3 or Cy5 (Jackson ImmunoResearch Laboratories, 1:250). Secondary antibodies conjugated to Alexa Fluor 488, Alexa Fluor 555, or Alexa Fluor 647 (Invitrogen and Jackson ImmunoResearch) were used at 1:500. The NMJ boutons were derived from muscles 6 and 7 of abdominal segments 2/3. To detect PI(4,5)P2 microdomains, the NMJ boutons were fixed immediately after high K+ stimulation. To quantitatively compare PLCδ1-PH-EGFP or AP-2α immunostaining signal among different sets of experiments, fixed larval fillets derived from different conditions were collected into the same Eppendorf tube. The NMJ boutons were stained for PLCδ1-PH-EGFP using α-GFP and for the neuronal membranes using fluorescein-conjugated α-HRP. Consecutive single-plane images of the boutons of muscles 6 and 7 in abdominal segments 2 or 3 of all different experimental sets were taken using a Zeiss LSM 780 confocal microscope with a Plan-Apochromat 63x/1.4 Oil DIC M27 objective under a 1 μm interval setup and equal laser power and laser exposure time. For data quantification, single-plane images of five different individual boutons from each NMJ bouton image were used. The presynaptic plasma membrane regions of the type Ib boutons defined by α-HRP immunostaining were outlined manually, and native fluorescence or α-GFP immunostaining signal intensities of PLCδ1-PH-EGFP or AP-2α on the plasma membrane were quantified using ImageJ and averaged to serve as one individual data-point. For data quantification of SIM images, we chose single-plane SIM images focused on a central section of an individual bouton and outlined plasma membrane-associated PLCδ1-PH-EGFP-immunostained clusters using ImageJ. The area of these clusters of each individual bouton was measured using ImageJ, and the areas of clusters measuring over 0.032 μm2 (equal to a ~ 200 nm diameter circle, i.e. the resolution limit of SIM) were averaged to serve as one data-point. We assessed 37 boutons derived from five NMJs of three different larvae. Image processing was achieved using LSM Zen.

Western blot

Request a detailed protocol

For western blotting, the brain and ventral nerve chord of larval fillets were removed and subjected to different stimulation conditions. Afterwards, the fillets were crushed in 1xSDS sample buffer and boiled for 5 min. Dilutions for primary antibodies were as follows: mouse anti-α-actin, 1:20000 (Sigma); chicken anti-GFP, 1:5000 (Invitrogen).

PLA

Request a detailed protocol

Third-instar larvae were fixed with 4% paraformaldehyde for 20 min and permeabilized with 1xPBS buffer containing 0.1% Tween-20. Larval fillets were incubated with mouse α-HA (Sigma, 1:200) and rabbit α-GFP (Invitrogen, 1:500) in 1xPBS buffer containing 0.1% Tween-20 at 4°C for 12 hr. Excess antibodies were washed out using 1xPBS buffer containing 0.1% Tween-20. The samples were mixed with the PLA probe (Sigma, 1:5) for 2 hr at 37°C. After washing with 1x buffer A, the samples were incubated with ligation solution (1:40) for 1.5 hr at 37°C. After again washing with 1x buffer A, the samples were incubated with amplification solution (1:80) for 2 hr at 37°C. Next, the samples were washed with 1x buffer B and then 0.01x buffer B. The samples were stained with anti-chicken Alexa Fluor 488-conjugated IgG and anti-mouse Alexa Fluor 647-conjugated IgG, followed by a wash of 1x PBS buffer containing 0.1% Tween-20. To quantitatively compare PLA signal, fixed larval fillets derived from different experimental conditions were collected into the same Eppendorf tube and processed. Consecutive single-plane images of the boutons of muscles 6 and 7 in abdominal segments 2 or 3 of all different experimental sets were taken using a Zeiss LSM 780 confocal microscope with a Plan-Apochromat 63x/1.4 Oil DIC M27 objective under a 1 μm interval setup and equal laser power and laser exposure time. For data quantification, consecutive Z-plane images spanning whole NMJ were projected under maximal fluorescence intensity. All type Ib boutons in individual Z-projection image were outlined according to PLCδ1-PH-EGFP-stained regions. PLA or antibody immunostaining signal intensities within the boutons and background staining signals in surrounding muscles were counted using ImageJ and averaged. One individual data-point was obtained by muscular background signal subtraction. Image processing was achieved using LSM Zen.

Nluc-Fwe-1D4 purification and BRET assay

Request a detailed protocol

The Nluc-Fwe-1D4 fusion protein was purified as described previously (Yao et al., 2009). Briefly, plasma membrane was isolated from a two liter culture of yeast strain BJ5457 expressing Nluc-Fwe-1D4 protein. The plasma membranes were solubilized at 4°C with 10x critical micelle concentration (CMC) DDM (Anatrace) in a solution of 20 mM HEPES (pH8.0), 300 mM NaCl, 10% Glycerol, 2.0 mM DTT, and 1 mM PMSF for 2 hr. Insoluble membranes were spun down by centrifugation at 100,000 ×g for 60 min. The lysates including solubilized Nluc-Fwe-1D4 proteins were cleaned with CNBr sepharose 4B at 4°C for 1 hr. The samples were then mixed with α−1D4-conjugated CNBr-activated Sepharose 4B at 4°C for 8–12 hr. After washing with a solution of 20 mM HEPES (pH8.0), 150 mM NaCl, 10% Glycerol, 2.0 mM DTT, 1 mM PMSF and 2.6xCMC DDM, the protein was eluted with a 1D4 peptide-containing buffer (3 mg of 1D4 peptide in 1 ml of washing buffer). Purified proteins were subjected to SDS-PAGE and detected using Lumitein staining or western blotting with anti-α−1D4 antibody at 1:5000 (Yao et al., 2009). The BRET assay was performed in a 384-well plate, with each well containing 30 μl of reaction solution [0.5 nM purified proteins, 5 μM BODIPY-TMR Phosphatidylinositol 4,5-bisphosphate (C-45M16A, Echelon Bioscience), furimazine (Promega, 1:2000), 20 mM HEPES, 150 mM NaCl, 10% Glycerol, 2 mM DTT, 1 mM PMSF, 4 mM DDM]. Fluorescence signal was detected using a Microplate Reader M1000 pro (Tecan) with two different emission spectrum filters, that is 500–540 nm for Nluc and 550–630 nm for BODIPY-TMR Phosphatidylinositol 4,5-bisphosphate. For competition assay, 30 μl of the reaction solution was included with 1 mM brain phosphatidylinositol 4,5-bisphosphate (Avanti). The BRET signal was calculated according to the following formula:

Live imaging

Request a detailed protocol

For PLCδ1-PH-EGFP imaging, third-instar larvae were dissected in a zero-calcium HL-3 solution at room temperature. For groups stimulated with electric pulses, larval fillets were bathed in a solution of 2 mM Ca2+ (70 mM NaCl, 5 mM KCl, 10 mM MgCl2, 10 mM NaHCO3, 5 mM trehalose, 5 mM HEPES (pH 7.4), 115 mM sucrose, 2 mM CaCl2). High concentrations of glutamate were used to desensitize glutamate receptors, thereby reducing muscle contraction when stimulated. A cut axonal bundle was sucked into the tip of a glass capillary electrode and then stimulated at 20 or 40 Hz for 3 min. Stimulus strength was set at 5 V and 0.5 ms duration by means of pClamp 10.6 software (Axon Instruments Inc). Ten images were taken from larval fillets at rest. After stimulation for 2 min, muscle contraction significantly decelerated. Thus, we captured 60 consecutive snapshot images every second from the third minute. Muscles 6 and 7 of abdominal segment three were imaged. Under the condition of high K+ stimulation, larval fillets were bathed in a solution of 90 mM K+/2 mM Ca2+/7 mM glutamate (25 mM NaCl, 90 mM KCl, 10 mM MgCl2, 10 mM NaHCO3, 5 mM trehalose, 5 mM HEPES (pH 7.4), 30 mM sucrose, 2 mM CaCl2, 7 mM monosodium glutamate) for 5 min. Sixty consecutive snapshot images were captured every second from the fifth minute of stimulation. Images were taken using a long working distance water immersion objective (XLUMPLFLN20XW, Olympus) and EMCCD camera (iXon, Andor) mounted on a SliceScope Pro 6000 (Scientifica) microscope and employing MetaFluor software (Molecular Devices). For GCaMP6f imaging, third instar larvae were dissected in a zero-calcium HL-3 solution at room temperature. Ten images were taken from larval fillets at rest. Subsequently, larval fillets were stimulated with a solution of 90 mM K+/2 mM Ca2+/7 mM glutamate for 5 min. Sixty consecutive snapshot images were captured every second from the fifth minute of stimulation. The lexA/lexAop2 binary system was used to stably express a comparable level of GCaMP6f in the presynaptic compartment of NMJ boutons for all tested genotypes, allowing us to compare Ca2+ imaging results when the resting Ca2+ levels were potentially affected by differences in genetic background. We stained for the GCaMP6f protein using a-GFP antibody and confirmed comparable GCaMP6f levels among the different genotypes tested in each dataset. Evoked Ca2+ levels were calculated by subtracting the resting GCaMP6f fluorescence from the GCaMP6f fluorescence induced by high K+ stimulation. The NMJs were derived from muscles 6 and 7 of abdominal segment 2/3. Images were taken using a water immersion objective (W Plan-Apochromat 40x/1.0 DIC M27, Zeiss). For each imaging experiment, at least three focused images for the same boutons under resting, stimulation, or post-stimulation conditions were used for data quantification. Fluorescence intensities of PLCδ1-PH-EGFP or GCaMP6f within the boutons were quantified using ImageJ and averaged to serve as one individual data-point. Image processing was achieved using LSM Zen.

Transmission electron microscopy

Request a detailed protocol

Third instar larval fillets were prepared in zero-calcium HL-3 solution at room temperature. For the resting conditions, the fillets were bathed in zero-calcium HL-3 solution at room temperature for another 10 min before fixation. For the high K+ stimulation conditions, fillets were bathed in a solution of 90 mM K+ and 2 mM Ca2+ for 10 min. The stimulation was terminated by washing three times with zero-calcium HL-3 solution, followed by fixation. For recovery conditions, following high K+ stimulation, fillets were bathed in zero-calcium HL-3 solution at room temperature for 10 or 20 min before fixation. Larval fillets were fixed for 12 hr at 4°C in 4% paraformaldehyde/1% glutaraldehyde/0.1 M cacodylic acid (pH 7.2), rinsed with 0.1 M cacodylic acid (pH 7.2), and postfixed with 1% OsO4 and 0.1 M cacodylic acid at room temperature for 3 hr. These samples were then subjected to a series of dehydration steps using 30–100% ethanol. After 100% ethanol dehydration, the samples were sequentially incubated with propylene, a mixture of propylene and resin, and pure resin. Finally, the samples were embedded in 100% resin. TEM images were captured using Tecnai G2 Spirit TWIN (FEI Company) and a Gatan CCD Camera (794.10.BP2MultiScanTM). NMJ boutons were captured at high magnifications. For each condition, NMJ bouton images were taken from at least five different NMJs of each third-instar larvae, and three to five larvae were used. Quantifications were performed using ImageJ. For diaminobenzidine (DAB) polymerization, third instar larvae were dissected at room temperature in zero-calcium HL-3 medium, followed by a 10 min incubation in 5 mM K+/0 Ca2+ mM solution or a 10 min stimulation of 90 mM K+/2 mM Ca2+. Next, the samples were subjected to 30 min fixation in ice-cold 4% paraformaldehyde/1% glutaraldehyde/0.1 M cacodylic acid (pH 7.2). Subsequently, the samples were transferred to Eppendorf tubes for 15 min incubation with a solution of 0.5 mg/ml DAB solution, followed by incubation with a solution of 0.5 mg/ml DAB and 0.006% H2O2 for 15 min at room temperature. This latter step was repeated once to ensure DAB polymerization. Samples were washed three times with 1xPBS buffer for 10 min and then fixed with a solution of 4% paraformaldehyde/1% glutaraldehyde/0.1 M cacodylic acid (pH 7.2) for 12 hr at 4°C, followed by fixation with a solution of 1% OsO4/0.1 M cacodylic acid at room temperature for 3 hr. Then, standard dehydration, embedding, and imaging were performed. For data quantifications of DAB intensities, the display color of TEM images was reverted to grayscale using ImageJ. Average DAB staining intensity on each individual bulk endosome was quantified. Then, the average DAB staining intensity on 50–100 surrounding SVs from the same bouton image was used to assess the relative level of HA-Fwe-APEX2 on bulk endosomes vs SVs.

Statistics

All data analyses were conducted using GraphPad Prism 8.0, unless stated otherwise. Paired and multiple datasets were compared by Student t-test or one-way ANOVA with Tukey’s post hoc test, respectively. Individual data values are biological replicates. Samples were randomized during preparation, imaging, and data processing to minimize bias.

Appendix 1

key resources table

Appendix 1—key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Genetic reagent (D. melanogaster)fweDB25Yao et al., 2009fwe mutant allele used in Figures 3–6, Figure 4—figure supplements 1–2, Figure 5—figure supplement 1, and Figure 6—figure supplement 1
Genetic reagent
(D. melanogaster)		fweDB56Yao et al., 2009fwe mutant allele used in Figures 3–6, Figure 4—figure supplements 1–2, Figure 5—figure supplement 1, and Figure 6—figure supplement 1
Genetic reagent (D. melanogaster)UAS-Flag-Fwe-HAYao et al., 2017Wild-type version of fwe transgene used in Figure 3a–b, Figure 6a, Figure 4—figure supplement 1, and Figure 6—figure supplement 1b
Genetic reagent (D. melanogaster)UAS-Flag-Fwe[E79Q]-HAYao et al., 2017Channel deficient version of fwe transgene used in Figure 3e–f
Genetic reagent (D. melanogaster)UAS-canA1-RNAi (FB4)Dijkers and O'Farrell, 2007RNAi line of canA1 used in Figure 3g–j
Genetic reagent (D. melanogaster)UAS-canA1-RNAi(TRiP.JF01871)Bloomington Drosophila Stock Center
Dijkers and O'Farrell, 2007		RRID:BDSC_25850RNAi line of canA1 used in Figure 3g–h, 5
Genetic reagent (D. melanogaster)UAS-synjKhuong et al., 2013Wild-type version of synj transgene used in Figure 2b–d, Figure 3i–j, Figure 6d–f, and Figure 2—figure supplement 1c.
Genetic reagent (D. melanogaster)UAS-PLCδ1-PH-EGFPBloomington Drosophila Stock Center
Verstreken et al., 2009		RRID:BDSC_39693PI(4,5)P2 reporter transgene used in Figures 1–4, Figure 1—figure supplement 1, Figure 1—figure supplement 3c, Figure 2—figure supplements 1–2, Figure 4—figure supplement 1e–f.
Genetic reagent (D. melanogaster)UAS-PLCδ1-PHS39R-EGFPBloomington Drosophila Stock Center
Verstreken et al., 2009		RRID:BDSC_39694PI(4,5)P2-binding mutant transgene used in Figure 2c–d, Figure 1—figure supplement 2c–e
Genetic reagent (D. melanogaster)nSyb-GAL4Bloomington Drosophila Stock Center
Pauli et al., 2008		RRID:BDSC_51635Neuronal GAL4 driver used in all figures
Genetic reagent (D. melanogaster)vglut-lexABloomington Drosophila Stock CenterRRID:BDSC_60314Motor neuron lexA driver used in Figure 3e–f, Figure 3i–j, Figure 4d–e, and Figure 4—figure supplement 1c–d
Genetic reagent (D. melanogaster)13XLexAop2-IVS-GCaMP6fBloomington Drosophila Stock CenterRRID:BDSC_44277Ca2+ indicator transgene used in Figure 3i–j, Figure 4d–e, and Figure 4—figure supplement 1c–d
Genetic reagent (D. melanogaster)synj1Bloomington Drosophila Stock Center
Verstreken et al., 2003		RRID:BDSC_24883synj mutant allele used in Figure 1—figure supplement 1c–d
Genetic reagent (D. melanogaster)UAS-mCD8-EGFPKyoto Stock CenterRRID:DGRC_108068mCD8-EGFP transgene used in Figure 1—figure supplement 2a–b
Genetic reagent (D. melanogaster)UAS-HA-Fwe[WT]-APEX2This paperAPEX2 fusion version of wild-type fwe transgene used in Figure 4d–g, Figure 5, Figure 6b–f, Figure 4—figure supplement 2, and Figure 6—figure supplement 1c–d
Genetic reagent (D. melanogaster)UAS-HA-Fwe[K29A/R33A]-APEX2This paperAPEX2 fusion version of fwe[K29A/R33A] transgene used in Figure 4d–g, Figure 5, Figure 4—figure supplement 2
Genetic reagent (D. melanogaster)UAS-Flag-Fwe[K29A/R33A/K95A/K100A/R105A/K146A/K147A/R150A]-HAThis paperfwe[K29A/R33A/K95A/K100A/R105A/K146A/K147A/R150A] transgene was not expressed stably
Genetic reagent (D. melanogaster)UAS-Flag-Fwe[K146A/K147A/R150A]-HAThis paperfwe[K146A/K147A/R150A] transgene used in Figure 4—figure supplement 1
Genetic reagent (D. melanogaster)UAS-Flag-Fwe[K95A/K100A/R105A]-HAThis paperfwe[K95A/K100A/R105A] transgene was not expressed stably
Genetic reagent (D. melanogaster)LexAop2-PLCδ1-PH-APEX2-HAThis paperAPEX2 fusion version of PLCδ1-PH transgene used in Figure 3i–j
Genetic reagent (D. melanogaster)LexAop2-PLCδ1-PH-EGFPThis paperlexAop2 version of PLCδ1-PH-EGFP transgene used in Figure 3e–f
Antibodyα-GFP
(Chicken polyclonal)		InvitrogenCat #A10262, RRID:AB_2534023IF: 1:500
WB: 1:5000
Antibodyα-HA
(Mouse monoclonal)		SigmaCat # H3663, RRID:AB_262051IF: 1:400
PLA: 1:200
Antibodyα-Bruchpilot
(Mouse monoclonal)		DSHB
Wagh et al., 2006		Cat# nc82, RRID:AB_2314866IF: 1:100
Antibodyα-AP-2α
(Rabbit polyclonal)		González-Gaitán and Jäckle, 1997IF: 1:3000
Antibodyα-Eps15
(Guinea pig polyclonal)		Koh et al., 2007IF: 1:3000
Antibodyα-Fwe B isoform
(Guinea pig polyclonal)		Yao et al., 2017IF: 1:400
AntibodyCy3 AffiniPure Rabbit Anti-Horseradish PeroxidaseJackson ImmunoResearch LabsCat# 323-165-021, RRID:AB_2340262IF: 1:250
AntibodyAlexa Fluor 488 AffiniPure Rabbit Anti-Horseradish PeroxidaseJackson ImmunoResearch LabsCat# 323-545-021, RRID:AB_2340264IF: 1:250
Antibodyα-GFP
(Rabbit polyclonal)		Thermo Fisher ScientificCat# A-6455, RRID:AB_221570PLA: 1:500
Antibodyα-actin
(Mouse monoclonal)		SigmaCat# A7732, RRID:AB_2221571WB:1:20000
Antibodyα−1D4
(Mouse monoclonal)		Yao et al., 2009WB: 1:2000
Strain, strain background (Escherichia coli)DH5α
Strain, strain background (Escherichia coli)yeast strain BJ5457Yao et al., 2009
Peptide, recombinant proteinNluc-Fwe-1D4 fusion proteinThis paperUsed in Figure 4b–c
Peptide, recombinant proteinNluc-Fwe[K29A/R33A]−1D4This paperUsed in Figure 4b–c
Peptide, recombinant proteinNluc-Fwe[K95A/K100A/R105A/K146A/K147/R150A]−1D4This paperUsed in Figure 4b–c
Peptide, recombinant proteinNluc-Fwe[K29A/R33A/K95A/K100A/R105A/K146A/K147A/R150A]−1D4This paperUsed in Figure 4b–c
Chemical compound, drugn-Dodecyl-β-D-MaltopyranosideAnatraceCat# D310Used in Figure 4b–c
Chemical compound, drugBODIPY-TMR Phosphatidylinositol 4,5-bisphosphateEchelon BioscienceCat#C-45M16AUsed in Figure 4b–c
Chemical compound, drugFurimazinePromegaCat# N1110Used in Figure 4b–c
Chemical compound, drugBrain Phosphatidylinositol 4,5-bisphosphateAvanti Polar LipidsAV-840046PUsed in Figure 4b–c
Commercial assay or kitDuolink In Situ Red Starter Kit Mouse/RabbitSigmaCat# DUO92101Used in Figure 3a–b
Commercial assay or kitNEBuilder HiFi DNA Assembly Master MixNEBCat# E2621
Recombinant DNA reagentYeMPYao et al., 2009
Recombinant DNA reagentpcDNA3 APEX2-NESAddgeneRRID:Addgene_49386
Recombinant DNA reagentpJFRC19-13XLexAop2-IVS-myr:GFPAddgeneRRID:Addgene_26224
Recombinant DNA reagentpNL1.1[Nluc]PromegaCat# #N1001
Recombinant DNA reagentpUAST-HA-Fwe[WT]-APEX2This paperDNA construct of APEX2 fusion version of wild-type fwe transgene used in Figure 4d–g, Figure 5, Figure 6b–f, Figure 4—figure supplement 2, and Figure 6—figure supplement 1c–d
Recombinant DNA reagentpUAST-HA-Fwe[K29A/R33A]-APEX2This paperDNA construct of APEX2 fusion version of fwe[K29A/R33A] transgene used in Figure 4d–g, Figure 5, Figure 4—figure supplement 2
Recombinant DNA reagentpUAST-Flag-Fwe-HAYao et al., 2017
Recombinant DNA reagentpUAST-Flag-Fwe[K29/R33/K95/K100/R105/K146A/K147A/R150A]-HAThis paperDNA construct of fwe[K29A/R33A/K95A/K100A/R105A/K146A/K147A/R150A] transgene
Recombinant DNA reagentpUAST-Flag-Fwe[K146A/K147A/R150A]-HAThis paperDNA construct of fwe[K146A/K147A/R150A] transgene used in Figure 4—figure supplement 1
Recombinant DNA reagentpUAST-Flag-Fwe[K95/K100/R105A]-HAThis paperDNA construct of fwe[K95A/K100A/R105A] transgene
Recombinant DNA reagentplexAop2-PLCδ1-PH-APEX2-HAThis paperDNA construct of APEX2 fusion version of PLCδ1-PH transgene used in Figure 3i–j
Recombinant DNA reagentplexAop2-PLCδ1-PH-EGFPThis paperDNA construct of lexAop2 version of PLCδ1-PH-EGFP transgene used in Figure 3e–f
Recombinant DNA reagentYeMP-Nluc-Fwe-1D4This paperDNA construct of expression of Nluc-Fwe-1D4 recombinant protein used in Figure 4b–c
Recombinant DNA reagentYeMP-Nluc-Fwe[K29A/R33A]−1D4This paperDNA construct of expression of Nluc-Fwe[K29A/R33A]−1D4 recombinant protein used in Figure 4b–c
Recombinant DNA reagentYeMP-Nluc-Fwe[K95A/K100A/R105A/K146A/K147/R150A]−1D4This paperDNA construct of expression of Nluc-Fwe[K95A/K100A/R105A/K146A/K147/R150A]−1D4 recombinant protein used in Figure 4b–c
Recombinant DNA reagentYeMP-Nluc-Fwe[K29A/R33A/K95A/K100A/R105A/K146A/K147A/R150A]−1D4This paperDNA construct of expression of Nluc-Fwe[K29A/R33A/K95A/K100A/R105A/K146A/K147A/R150A]−1D4-1D4 recombinant protein used in Figure 4b–c
Sequence-based reagentHA-Fwe[WT]-APEX2 forward-1This paper5'- tcgttaacagatcttGCGGCCGCATGTACCCATACGATGTTCCAGATTACG-3'
Sequence-based reagentHA-Fwe[WT]-APEX2 Reverse-1This paper5'- tgaacctcctccgccTGTGGGGCGCCAGACATC-3'
Sequence-based reagentHA-Fwe[WT]-APEX2 forward-2This paper5'- gccccacaggcggaggaggttcaggaggcggaggttcgGGTACCGGCGGAAAGAGTTACCCGAC-3'
Sequence-based reagentHA-Fwe[WT]-APEX2 Reverse-2This paper5'- tccttcacaaagatccTCTAGAGGCGTCGGCGAATCCCAG −3'
Sequence-based reagentHA-Fwe[K29A/R33A]-APEX2-HA forwardThis paper5'-CAGCCGTGGTATCTCGCATATGGAAGTGCATTGCTGGGCATTG-3'
Sequence-based reagentFlag-Fwe[K95/K100/R105A]-HA forwardThis paper5'-GGAATCTGCGCCCTTGTACTTCGCTGCCGGGCTCTACATTGCCATGGCCATTCCGCCCATTATT-3'
Sequence-based reagentFlag-Fwe[K146A/K147A/R150A]-HA forwardThis paper5'-CAGAGGACATGGCCGCCGCTGCCACATCGCCCACACAGATGGCCGGCAGTCAGGCGGGCGG-3'
Sequence-based reagentPLCδ1-PH-APEX2-HA forward-1This paper5'-tcttatcctttacttcagGCGGCCGCATGGACTCGGGCCGGGAC-3'
Sequence-based reagentPLCδ1-PH-APEX2-HA forward-2This paper5'- GCTGTACAAGggcggaggaggttcaggaggcggaggttcgGGCGGAAAGAGTTACCCGAC-3'
Sequence-based reagentPLCδ1-PH-APEX2-HA reverse-1This paper5'- cctccgccggtaccAAGATCTTCCGGGCATAGCTGTCG-3'
Sequence-based reagentPLCδ1-PH-APEX2-HA reverse-2This paper5'- ggttccttcacaaagatcctctagattaagcgtaatctggaacatcgtatgggtaGGCGTCGGCGAATCCCAG −3'
Sequence-based reagentPLCδ1-PH-EGFP forwardThis paper5'- tcttatcctttacttcagGCGGCCGCATGGACTCGGGCCGGGAC-3'
Sequence-based reagentPLCδ1-PH-EGFP reverseThis paper5'- TTTCTAGATTACTTGTACAGCTCGTCCAT-3'
Sequence-based reagentYeMP-Nluc-Fwe-1D4 forward-1This paper5'- CCACTAGTATGGTCTTCACACTCGAAG-3'
Sequence-based reagentYeMP-Nluc-Fwe-1D4 forward-2This paper5'- AAAACTAGTGTCGACTCGTTTGCGGAAAAGATAACG
−3'
Sequence-based reagentYeMP-Nluc-Fwe-1D4 Reverse-1This paper5'-AAAGTCGACCGAACCTCCGCCTCC-3'
Sequence-based reagentYeMP-Nluc-Fwe-1D4 Reverse-2This paper5'- AAAACGCGTTTACGCAGGCGCGACTTGG-3'
Sequence-based reagentYeMP-Nluc-Fwe[K29A/R33A]−1D4, YeMP-Nluc-Fwe[K95A/K100A/R105A/K146A/K147/R150A]−1D4, YeMP-Nluc-Fwe[K29A/R33A/K95A/K100A/R105A/K146A/K147A/R150A]−1D4 forwardThis paper5'- AAAACTAGTGTCGACTCGTTTGCGGAAAAGATAACG-3'
Sequence-based reagentYeMP-Nluc-Fwe[K29A/R33A]−1D4, YeMP-Nluc-Fwe[K95A/K100A/R105A/K146A/K147/R150A]−1D4, YeMP-Nluc-Fwe[K29A/R33A/K95A/K100A/R105A/K146A/K147A/R150A]−1D4 ReverseThis paper5'- AAAACGCGTTTACGCAGGCGCGACTTGGCTGGTCTCTGTTGTGGGGCGCC-3'
Software, algorithmLSM ZenZeisshttps://www.zeiss.com/microscopy/us/products/microscope-software/zen-lite.html
Software, algorithmImage Jhttps://imagej.net/contributorshttps://imagej.nih.gov/ij/
Software, algorithmpClamp 10.6Moleculardeviceshttps://www.moleculardevices.com/products/cellular-imaging-systems/acquisition-and-analysis-software/metamorph-microscopy#gref
Software, algorithmMetaMorphMoleculardeviceshttps://www.moleculardevices.com/products/cellular-imaging-systems/acquisition-and-analysis-software/metamorph-microscopy#gref
Software, algorithmGraphPad Prism 8.0Prismhttps://www.graphpad.com/

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided.

References

    1. Brand AH
    2. Perrimon N
    (1993)
    Targeted gene expression as a means of altering cell fates and generating dominant phenotypes
    Development 118:401–415.
    1. Hille B
    2. Dickson EJ
    3. Kruse M
    4. Vivas O
    5. Suh B-C
    (2015) Phosphoinositides regulate ion channels
    Biochimica Et Biophysica Acta (BBA) - Molecular and Cell Biology of Lipids 1851:844–856.
    https://doi.org/10.1016/j.bbalip.2014.09.010

Decision letter

  1. Hugo J Bellen
    Reviewing Editor; Baylor College of Medicine, United States
  2. Richard W Aldrich
    Senior Editor; The University of Texas at Austin, United States
  3. Patrik Verstreken
    Reviewer; KU Leuven, Belgium

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Your work nicely reveals new aspects of the function of Flower in bulk endocytosis by regulating the levels of PI(4,5)P2. The observation that Flower recruits PI(4,5)P2 which directly enhances Flower channel activity is elegantly demonstrated. Moreover, the fact that PI(4,5)P2 drives ADBE which in turn removes Flower, nicely explains how the process is terminated.

Decision letter after peer review:

Thank you for submitting your article "Flower initiates a positive feedback loop upon PIP2 enrichment at periactive zones to control bulk endocytosis" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by Hugo Bellen, Reviewing Editor and Richard Aldrich, Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Patrik Verstreken (Reviewer #1).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission. As you will notice the reviewers find your data interesting and they have a positive tone. However, they raise a number of issues. I have read their reviews and your manuscript. Many of the issues that are raised relate to better controls, textual changes, clarifications, and quantification of data. We think you can address most of these in a timely manner.

Reviewer #1:

In this paper Li et al. show how the calcium channel and PI(4,5)P2 regulate bulk endocytosis at synapses. Their work brings exciting new mechanistic insight into the regulation of this process, and what I find of particular interest is the notion that Fwe might be gated by PI(4,5)P2 to control the retrieval of synaptic membrane. I think this is an exciting paper and I have mostly minor comments on interpretation and one suggestion for a set of genetic interactions to cement the role for calcineurin in this pathway.

1) The authors elegantly confirm and demonstrate further that PI(4,5)P2levels increase during stimulation. They conclude this is in microdomains based on several pieces of data. This indeed seems to be the case but I think it is important to note that chemical fixation may alter distribution. I would either note this in the text or try to include single particle tracking experiments (eg sptPALM).

2) In Figure 2 they investigate the role for PI(4,5)P2 in ADBE using PLCdPH over expression and Synj over expression. It might be interesting to look at the role for Tweek in this process as well, giving the opposing effect of Tweek in comparison to Synj1.

3) I wonder if in the reformation assays in Figure 2 the authors can observe the formation of small SVs at the surface of bulk endosomes?

4) I think the conclusion that the PLA signal indicates a physical interaction between Fwe and PI(4,5)P2 is overstated: to conclude that Fwe moves in to PI(4,5)P2-enriched domains would be more correct. The demonstration of a direct interaction is given later in the paper in Figure 4.

5) The authors write "Similar to the effect of loss of Fwe, reduced CanA1 levels greatly suppressed the formation of PIP2 microdomains compared to the control (Figure 3E-F), indicating that Calcineurin mediates the Fwe Ca2+ channel to induce PIP2 microdomains." This is not shown: the authors show that both Fwe and CanA1 reduce the formation of PI(4,5)P2 microdomains during stimulation, but not that CanA1 mediates this effect in an Fwe-dependent manner. This should be shown using genetic interactions: eg test if double mutants show a similar (non-additive) effect compared to either single mutant, or overexpress Fwe in CanA1 mutants (and vice versa) to determine if this rescues the defect.

6) Figure 6G-H seem superfluous to me and not central to the story. I would remove this from the manuscript.

7) I also suggest to write PI(4,5)P2 or PhIns(4,5)P2 instead of PIP2 because the latter refers to all phosphoinositides that have 2 phosphate groups, and not specifically to PI(4,5)P2.

Reviewer #2:

This paper describes an intriguing model for a feedback pathway by which the calcium channel Flower, Ca++ signaling, and PIP2 together regulate activity dependent bulk endocytosis and synaptic vesicle regeneration at a neuronal synapse. This is an interesting and novel idea, which could illuminate key molecular dynamics downstream of synaptic activity that regulate synaptic vesicle recycling. The strengths of the paper are the in vitro-in vivo analysis of fwe-PIP2 binding and function, and in the measurement of each of PIP2, Calcium signaling, and membrane traffic for each component.

1) The authors claim the increase in fluorescence of PHPLC-GFP as evidence of clustering in microdomains. However, this is a circular argument, as the increase in fluorescence is proposed as the evidence that PIP2 microdomains have formed. This claim is an important premise of the paper, as PHPLC fluorescence intensity is the primary quantified metric for a central conclusion of the paper (that PIP2 microdomains form upon stimulation, and require fwe).

At a technical level, it is difficult to understand why clustering should increase fluorescence, since the total amount of protein is not changed by Western blotting. One possibility is that diffuse fluorescence is below the detection limit of the microscope, while clustered fluorescence is above this limit; however, comparing measurements in this way from inside and outside the linear range of detection is misleading. Another possibility is that GFP fluorescence emission (but not levels) is increased in the microenvironment of the puncta, but this is not evidence for clustering, but instead for a changed microenvironment. The authors could quantify this by post-staining after fixation with anti-GFP antibodies in a second channel (they would expect to see no clustering if the effect was purely on GFP fluorescence). Either way, the whole-bouton quantification presented here is not sufficient, and more rigorous quantification of clustering is required (see point 4).

Finally, there are concerns about potential effects of PH-PLCdelta (which has been used here and elsewhere both as a reporter and manipulation for PIP2). What level of expression were used in Figure 1 compared to 2, and could the "clusters" coincide with malfunctioning endocytosis? Levels should be compared directly (rather than inferred from temperature).

Due to these questions, some form of independent evidence (published or novel) for PIP2 clustering must be shown or cited.

2) The consistency of phenotypes among experimental manipulations on Calcium, PIP2, and vesicle dynamics is compelling evidence for a feedback loop. However, not all functional links in the loop are completely tested; for instance, whether CanA1 loss of function leads to calcium signaling defects via synj/PIP2, or through its myriad other targets. While it might be possible to “normalize” synj phosphorylation using minibrain heterozygotes or mutants, this is beyond the scope of suggested revisions. However, where a link in the chain has not been as rigorously tested (as for CanA1), the authors should discuss other possible interpretations of these data (as they do for other possible means of gating Flower).

3) Figure 2—figure supplement 2- why does FM get taken up to similar extent with high PLC or synj if EM data suggests ABDE is not happening? And why is unloading not impaired if vesicle recycling is diminished?

4) There are several instances where images are not quantified, or where the method of quantification is insufficiently described:

a) Currently Materials and methods indicate only that “image processing and quantification were achieved using LSM Zen and ImageJ”. The authors must describe in detail how fluorescence imaging was processed and quantified. What processing steps were performed? In 2D or 3D? Projections or slices? How were NMJ masks or ROIs created? Materials and methods do not describe what kind of microscope and objective were used for live imaging (important for understanding resolution). All this information is required to correctly interpret the imaging data.

b) Microdomains are poorly described: how is their size quantified? 200nm is very close to the resolution limit of SIM, suggesting many measurements may have been at the floor of the resolution of the system and that this measurement likely represents an overestimate of the size of the domain. It would be informative to see a histogram of values to determine to what degree this is the case, and discuss if appropriate.

c) How was DAB staining (Figure 6) quantified? What was the criterion used to discriminate between no labeling and “similar to SV” (eg in d, where the overall staining appears to be more similar to a than to b). In the text, the data are presented as both SV and bulk endosome levels decreasing, though these data are not quantified or presented in the figure.

5) If fwe initiates a positive feedback loop, why do only high levels of activity stimulate clustering and ADBE?

6) The claim that Fwe is “selectively sorted” by PIP2 seems overstated, especially given the companion idea that Fwe sorts PIP2 into microdomains. In general, the data in Figure 6 are a bit unconvincing or overinterpreted:

a) AP2 complexes strongly associated with PIP2 microdomains- this is completely unsupported by the data, which show an exceedingly low correlation.

b) How were “colocalized” pixels determined in Figure 6H (not described in legend or Materials and methods)? Further, the figure legend for 6I indicates that “Quantification data for Pearson correlation coefficients” is the overlapping area divided by the overall area. This is not a Pearson's correlation coefficient, which rather depends on the linear relationship between channel intensities of all (or a subset) of pixels.

c) “synj altered the localization of the AP2 complex”- this is not quantified and is very difficult to determine from the images presented. This is also a clear example where it is not obvious that images are being taken from the same Z-plane, which further confounds interpretation of the data.

d) AP2-fwe IPs are not very convincing. The difference between negative control and IP is subtle, only a very small fraction of input, and neither repeated or quantified. These data should be strengthened or removed.

e) Is synj expressed at room temperature or 29 degrees?

Reviewer #3:

Previous work by the Yao lab showed that the SV-associated Ca2+ channel Flower (Fwe) is required for both clathrin-mediated endocytosis (CME) of synaptic vesicles (SVs) from the plasma membrane (PM) and activity-dependent bulk endocytosis (ADBE). Importantly, Fwe has likely a distinct role for each endocytotic route since only ADBE requires Ca2+ channel function. This follow-up study pursues the mechanisms underlying Fwe's role for ABDE. The study provides evidence for an interesting model of Fwe's role in ADBE that may significantly advance our understanding of mechanisms underlying ADBE. Specifically, it suggests that freshly exocytosed Fwe on the PM forms a positive Ca2+-dependent feedback loop that facilitates the generation and accumulation of phosphatidylinositol 4,5-bisphosphate (PIP2) in microdomains on the PM. In turn, accumulating PIP2 may facilitate compartmentalization of Fwe to bulk endosomes, thereby terminating ADBE. I have a number of concerns.

1) To visualize PIP2 at fly NMJs, the entire study relies only on a single PIP2 marker, the Gal4-driven expression of PLCδ1-PH-EGFP. Unfortunately, this marker not only reduces "free" PIP2 levels but also has dominant-negative effects on synaptic growth and SV cycling (Khuong et al., 2010). To properly support the notion that intense stimulation causes accumulation of PIP2 in microdomains on the PM, a number of corrections are required:

a) the PLCδ1-PH-EGFP imaging experiments require a proper control, like the expression of PLCδ1-PHS39R-GFP, which does not bind PIP2. Measuring mCD-EGFP expression levels is not a proper control.

b) the stimulation-induced accumulation of PIP2 should be verified by a second independent marker like quantitative immunostainings for edogenous α-adaptin or other PIP2-binding proteins. This is especially important to rule out unaccounted dominant-negative effects.

c) It is unclear how changes in PLCδ1-PH-EGFP fluorescence (Figure 1A-B, D, plus all subsequent experiments using PLCδ1-PH-EGFP including Figure 3C-F, 4F-G) are measured at larval NMJs since no information about respective ROIs is provided. To make a strong statement about a PIP2 accumulation on the PM, PLCδ1-PH-EGFP fluorescence measured for an entire bouton (from one or a few slices) should be subtracted by the PLCδ1-PH-EGFP fluorescence obtained from the cytosol of the bouton. Measurements using ROIs for the entire bouton do not allow specific conclusions about PM levels.

2) To significantly strengthen the conclusion that PIP2 accumulation drives ADBE initiation and SV reformation (Figure 2) a proper control should be used for the postulated PLCδ1-PH-EGFP-mediating reduction of PIP2 levels. Ideally, one may use expression of the PLCδ1-PHS39R-GFP mutant. Overexpression of synaptojanin alone does not make the point that the reduced ADBE levels are due to PIP2 depletion since synaptojanin also affects several other phosphatidylinositides.

3) The proximity ligation experiments of Figure 3A require a proper control like PLCδ1-PHS39R-GFP co-expression with Flag-Fwe-HA. Otherwise, the observed proximity could be due to other factors.

4) The images of Figure 3C, e show little evidence of PLCδ1-PH-EGFP microdomains at the PM, much in contrast to Figure 1. Does this mean that the quantification associated with these figures mostly reflects PIP2 on cytosolic bulk endosomes?

5) The shown knockdown data of CanA1 in Figure 3E-H require a control for specificity, best co-expression of an RNAi-resistant WT transgene.

6) The study suggests that PIP2 binding to freshly exocytosed Fwe gates Fwe's Ca2+ channel activity. In turn, the resulting Ca2+ influx activates calcineurin and the postulated feedback loop accumulating PIP2 at the PM. If correct, one would expect that a mutation that several affects Fwe' Ca2+ channel activity (like FweE79Q) should significantly reduce the accumulation of PIP2. Is this indeed the case? If not, the Fwe-dependent PIP2 accumulation may occur through a different mechanism.

7) The HA-Fwe-APEX2 experiments of Figure 6 require further controls including documentation that the fusion protein is properly localized to SVs and not excessively retained on the PM during ADBE. The EM alone is not convincing.

8) The co-immunoprecipitation of Fwe with AP2 shown in Figure 6G requires quantification.

9) The evidence for the conclusion that accumulating PIP2 facilitates "sorting" of Fwe for ADBE is weak because it essentially rests on the HA-Fwe-APEX2 experiment. Since reducing PIP2 levels by synaptojanin expression reduced the bulk endosome localization of Fwe as judged by DAB staining levels (Figure 6D-F), the authors conclude that "PIP2 microdomains can facilitate the retrieval of Fwe to the bulk endosome". However, this appears to be somewhat a catch22 since reduced PIP2 levels decrease the overall amount of bulk endocytosis. Whether the potentially reduced levels of internalized Fwe are disproportionally reduced has not been addressed. In addition, DAB staining intensity should not be used to quantify protein levels as the enzymatic reaction may have different constrains on SVs versus endosomes.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "A positive feedback loop between Flower and PI(4,5)P2 at periactive zones controls bulk endocytosis in Drosophila" for further consideration by eLife. Your revised article has been evaluated by Richard Aldrich (Senior Editor) and Hugo Bellen (Reviewing Editor).

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

The authors have provided a meaningfully improved manuscript that largely addresses the major concerns of the paper. In particular, the authors provide several new lines of evidence that levels or organization of PI(4,5)P2 are likely changing in response to stimulation: There is no change in the new negative control PH S39 mutant, AP2a levels increase at the membrane, and the levels of native and anti-GFP fluorescence of PH-PLC increase at the membrane. Especially given the strong correspondence between in vitro and in vivo data and the consistent phenotypes among the perturbations in the proposed pathway, these new data further strengthen the main claim that Fwe acts in a positive feedback loop with Ca++ and PI(4,5)P2.

It is worth noting that the precise nature of the microdomains remains a bit unclear. It is difficult to discern whether PI(4,5)P2 microdomains are forming de novo following stimulation, or whether PI(4,5)P2 ¬signal increases in existing microdomains (for example, AP2a recruitment to the membrane is quite high even under resting conditions). And as noted by the authors, while the PH domain is the best available tool to answer the questions posed, it remains possible that properties of the microdomains partially reflect the function-perturbing aspects of the probe. However, these caveats do not significantly detract from the core conclusions of the paper.

Figure 3H. Something is wrong with the shown KD data by UAS-canA1-RNAi(FB5) or the drawn conclusion that "reducing CanA1 levels via expression of either RNAi construct greatly suppressed the formation of PI(4,5)P2 microdomains compared to the control (Figure 3G-H)". The resting levels for the canA1 KD are only 50% of control. The K+-induced increase PLCδ1-PH-EGFP in the KD is as large as in control. Accordingly, the KD seems to have no effect, which would suggest that one of the two KDs is not specific for canA1.

If the FB5 KD is not working, the one needs a wild type CanA1 co-expression for the KD with the canA1-RNAi(JF01871) transgene to demonstrate specificity.

Please clarify.

Figure 3I. The canA1 KD uses the controversial UAS-canA1-RNAi(FB5) transgene, which does not appear to block the formation of PI(4,5)P2 microdomains (Figure 3H). The experiment should be done with the UAS-canA1-RNAi(JF01871) transgene as only this transgene reliable affects PI(4,5)P2 microdomains (assuming the KD can be rescued). Otherwise, one cannot conclude that "Calcineurin mediates the Ca2+ influx conducted by Fwe to induce PI(4,5)P2 microdomains". Resolving this issue is critical.

"Together, these data suggest that the PI(4,5)P2 microdomains initiate ADBE to control SV membrane retrieval upon intense stimulation." This sentence should be rephrased. The shown FM1-43 data (Figure 2—figure supplement 2B) only suggest that overexpression of PLCδ1-PH-EGFP or Synj impairs SV endocytosis independent of the mode of endocytosis. The shown FM1-43 data do not exclude a role of PI(4,5)P2 for other modes of endocytosis.

Moreover, the FM1-43 data use a much shorter (30 seconds) and milder stimulation than the solid EM data visualizing ADBE (10 minutes; Figure 2). Therefore, the FM1-43 and EM data are not directly comparable.

The same concern applies to the FM1-43-visulaized SV exocytosis data, which indicate that SV exocytosis is not impaired by overexpression of PLCδ1-PH-EGFP or Synj if endocytosis is stimulated for 1 minute. Again, the data are not comparable to the ADBE defect shown in Figure 2. I suggest either to omit the data of Figure 2—figure supplement 2 or use the same stimulation paradigm for the FM1-43 and EM data.

https://doi.org/10.7554/eLife.60125.sa1

Author response

Reviewer #1:

In this paper Li et al. show how the calcium channel and PI(4,5)P2 regulate bulk endocytosis at synapses. Their work brings exciting new mechanistic insight into the regulation of this process, and what I find of particular interest is the notion that Fwe might be gated by PI(4,5)P2 to control the retrieval of synaptic membrane. I think this is an exciting paper and I have mostly minor comments on interpretation and one suggestion for a set of genetic interactions to cement the role for calcineurin in this pathway.

1) The authors elegantly confirm and demonstrate further that PI(4,5)P2levels increase during stimulation. They conclude this is in microdomains based on several pieces of data. This indeed seems to be the case but I think it is important to note that chemical fixation may alter distribution. I would either note this in the text or try to include single particle tracking experiments (eg sptPALM).

For our live-imaging experiments (also see Materials and methods), we stimulated the NMJs with 40-Hz electric stimulation or a 90-mM K+ chemical stimulation under conditions of 2 mM extracellular Ca2+. However, these stimuli caused pronounced muscle contraction, which prevented us from precisely focusing the objective lens on a single NMJ bouton over the imaging timespan for confocal microscopy. Attempts to slow down contractions by desensitizing the glutamate receptors with 7 mM glutamate treatment were unsuccessful. Therefore, in our experiments, we manually adjusted the focus plane and captured snapshot images of the same single boutons each second for one minute during stimulation by using a long working distance water immersion objective (XLUMPLFLN20XW water lens) and an EMCCD camera (iXon, Andor) mounted on a SliceScope Pro 6000 (Scientifica) microscope. For each imaging experiment, at least three focused images of the same boutons under resting, stimulation, or post-stimulation conditions were used for data quantification. Fluorescence intensities of PLCδ1-PH-EGFP within the boutons were assessed using ImageJ and averaged to serve as one individual data point. Due to these technical issues, it is difficult to perform single-particle tracking.

We have assessed the possibility that chemical fixation may alter membrane properties and thereby account for plasma membrane clustering of PLCδ1-PH-EGFP upon intense stimulation. We have now examined the distribution of mCD8-GFP and PLCδ1-PHS39R-EGFP (as also suggested by reviewer #3), a PI(4,5)P2-binding mutant Khuong et al., 2010Verstreken et al., 2009(; ). We found that there was no obvious change in the plasma membrane pattern of mCD8-GFP in fixed boutons stimulated with high K+ (Figure 1—figure supplement 2A-B). Unlike PLCδ1-PH-EGFP, PLCδ1-PHS39R-EGFP was observed mainly in the cytosol at rest. After intense stimulation, immunostaining signals of PLCδ1-PHS39R-EGFP did not accumulate on the plasma membrane (Figure 1—figure supplement 2C-D) and did not increase within the boutons (Figure 1—figure supplement 2E). In addition, we have further assessed potential dominant-negative effects of PI(4,5)P2 binding by PLCδ1-PH-EGFP on the stimulation-induced accumulation of PI(4,5)P2 by examining recruitment of the AP-2 complex by PI(4,5)P2. Similar to PLCδ1-PH-EGFP, levels of the α subunit of the AP-2 complex (AP-2α) were also increased on the plasma membrane upon high K+ stimulation (Figure 1—figure supplement 3A-B). Taken together, these results suggest that PLCδ1-PH-EGFP clustering on the plasma membrane upon stimulation arises from an increase in local PI(4,5)P2 concentration, although a potential effect of chemical fixation, if any, on PLCδ1-PH-EGFP clustering cannot be excluded. The new data from PLCδ1-PHS39R-EGFP and AP-2α experiments have been added to the figures indicated in our explanation here, and a respective statement has been added to the main text.

2) In Figure 2 they investigate the role for PI(4,5)P2 in ADBE using PLCdPH over expression and Synj over expression. It might be interesting to look at the role for Tweek in this process as well, giving the opposing effect of Tweek in comparison to Synj1.

We agree it would be interesting to test for a potential role of Tweek in ADBE, although we also feel it may be beyond the scope of this paper. We did attempt to obtain tweek mutant stocks (tweek1 and tweek2) from the Bloomington Drosophila Stock Center and other sources. Unfortunately, after multiple requests, due to the hot summer weather, the mutant stocks died during delivery. Therefore, we could not perform the suggested TEM analysis within the limited revision period.

3) I wonder if in the reformation assays in Figure 2 the authors can observe the formation of small SVs at the surface of bulk endosomes?

We thank the reviewer for raising this interesting point. We have carefully re-examined our TEM images collected during recovery periods. However, we did not observe any accumulation of SVs budding from the bulk endosome in control boutons or boutons in which the PI(4,5)P2 microdomains were perturbed. This observation further indicates that PI(4,5)P2 microdomain formation likely plays an early role in promoting SV reformation from the bulk endosome. We now state this observation in the revised manuscript.

4) I think the conclusion that the PLA signal indicates a physical interaction between Fwe and PI(4,5)P2 is overstated: to conclude that Fwe moves in to PI(4,5)P2-enriched domains would be more correct. The demonstration of a direct interaction is given later in the paper in Figure 4.

As suggested by this reviewer and also by reviewer #3, we have now rephrased our conclusion to: “The PLA signal was low in resting boutons, whereas high K+ treatment significantly increased PLA signal intensity (Figure 3A-B). Therefore, these results suggest that Fwe and PI(4,5)P2 are closely colocalized when intense stimulation triggers PI(4,5)P2 microdomain formation.”

5) The authors write "Similar to the effect of loss of Fwe, reduced CanA1 levels greatly suppressed the formation of PIP2 microdomains compared to the control (Figure 3E-F), indicating that Calcineurin mediates the Fwe Ca2+ channel to induce PIP2 microdomains." This is not shown: the authors show that both Fwe and CanA1 reduce the formation of PI(4,5)P2 microdomains during stimulation, but not that CanA1 mediates this effect in an Fwe-dependent manner. This should be shown using genetic interactions: eg test if double mutants show a similar (non-additive) effect compared to either single mutant, or overexpress Fwe in CanA1 mutants (and vice versa) to determine if this rescues the defect.

As suggested, we have now assessed if Fwe overexpression can rescue the ADBE defect caused by canA1 RNAi knockdown. Neuronal canA1 was knocked down by expressing two independent canA1 RNAi lines, which impaired ADBE relative to the nsyb-GAL4 control (Figure 5A-B)[To address the similar concern raised by reviewer #3, we have conducted an additional canA1 RNAi knockdown experiment using UAS-canA1-RNAi (JF01871)]. We found that Fwe overexpression was able to significantly reverse the ADBE defect under both of these canA1 RNAi knockdown conditions (Figure 5A-B). Therefore, increasing Fwe-dependent Ca2+ influx can augment activation of the residual CanA1 enzymes and thus normalize downstream ADBE. These new data support that CanA1 mediates this effect in an Fwe-dependent manner. The new data have been added to Figure 5A-B, and a respective statement has been added to the main text.

We have also attempted to express a constitutive form of CanA1 in the fwe mutant background using two different neuronal GAL4 drivers, nsyb-GAL4 and elav-GAL4. Whereas single expression of a constitutive form of CanA1 did not affect larval viability, this expression caused precocious lethality of fwe mutants at very early larval stages. Unfortunately, the tiny size of the resulting larvae prevented us from further NMJ characterization.

6) Figure 6G-H seem superfluous to me and not central to the story. I would remove this from the manuscript.

We thank the reviewer for this constructive suggestion. We have removed these data from Figure 6.

7) I also suggest to write PI(4,5)P2 or PhIns(4,5)P2 instead of PIP2 because the latter refers to all phosphoinositides that have 2 phosphate groups, and not specifically to PI(4,5)P2.

Acknowledged. We have now changed PIP2 to PI(4,5)P2 in the revised manuscript.

Reviewer #2:

This paper describes an intriguing model for a feedback pathway by which the calcium channel Flower, Ca++ signaling, and PIP2 together regulate activity dependent bulk endocytosis and synaptic vesicle regeneration at a neuronal synapse. This is an interesting and novel idea, which could illuminate key molecular dynamics downstream of synaptic activity that regulate synaptic vesicle recycling. The strengths of the paper are the in vitro-in vivo analysis of fwe-PIP2 binding and function, and in the measurement of each of PIP2, Calcium signaling, and membrane traffic for each component.

1) The authors claim the increase in fluorescence of PHPLC-GFP as evidence of clustering in microdomains. However, this is a circular argument, as the increase in fluorescence is proposed as the evidence that PIP2 microdomains have formed. This claim is an important premise of the paper, as PHPLC fluorescence intensity is the primary quantified metric for a central conclusion of the paper (that PIP2 microdomains form upon stimulation, and require fwe).

At a technical level, it is difficult to understand why clustering should increase fluorescence, since the total amount of protein is not changed by Western blotting. One possibility is that diffuse fluorescence is below the detection limit of the microscope, while clustered fluorescence is above this limit; however, comparing measurements in this way from inside and outside the linear range of detection is misleading. Another possibility is that GFP fluorescence emission (but not levels) is increased in the microenvironment of the puncta, but this is not evidence for clustering, but instead for a changed microenvironment. The authors could quantify this by post-staining after fixation with anti-GFP antibodies in a second channel (they would expect to see no clustering if the effect was purely on GFP fluorescence).

We thank the reviewer for raising this concern. We have now performed the suggested experiments. Our confocal images revealed that, for neurons at rest, there was weak fluorescence signal of native PLCδ1-PH-EGFP on the presynaptic plasma membrane labelled by α-Hrp staining, with some additional signal dispersed in the cytosol (Figure 1—figure supplement 1C). First, we elevated PI(4,5)P2 levels on the plasma membrane by removing a copy of synaptojanin (synj), which encodes the major neuronal PI(4,5)P2 phosphatase (Tsujishita et al., 2001; Verstreken et al., 2003). The resulting reduced level of Synj enhanced PLCδ1-PH-EGFP fluorescence (green channel, Figure 1—figure supplement 1C-D), consistent with previous studies (Chen et al., 2014; Verstreken et al., 2009). In this context, we did not observe a significant change in protein expression of PLCδ1-PH-EGFP in the presynaptic compartment (Figure 1—figure supplement 1E-F). By using α-GFP immunostaining (red channel), these PLCδ1-PH-EGFP signals could be faithfully amplified (Figure 1—figure supplement 1C-D).

Next, we stimulated the boutons expressing PLCδ1-PH-EGFP and treated with 90 mM K+ and 2 mM Ca2+ for 10 min. Similar to our live-imaging results, both native fluorescence and α-GFP immunostaining signals of PLCδ1-PH-EGFP were significantly increased on the presynaptic plasma membrane (Figure 1—figure supplement 1G-H) In particular, we observed high-level induction of PLCδ1-PH-EGFP puncta (Figure 1C-D). As a control experiment, we examined the distribution of mCD8-GFP and PLCδ1-PHS39R-EGFP, a PI(4,5)P2-binding mutant (Khuong et al., 2010; Verstreken et al., 2009) (as suggested by reviewer #3), which revealed no obvious change in the plasma membrane pattern of mCD8-GFP in fixed boutons stimulated with high K+ (Figure 1—figure supplement 2A-B). Unlike PLCδ1-PH-EGFP, PLCδ1-PHS39R-EGFP was found mainly in the cytosol at rest. After intense stimulation, immunostaining signals of PLCδ1-PHS39R-EGFP did not accumulate on the plasma membrane (Figure 1—figure supplement 2C-D) and were not increased (Figure 1—figure supplement 2E). As suggested by reviewer #3, we have now assessed recruitment of the AP-2 complex upon PI(4,5)P2 increase. Similar to PLCδ1-PH-EGFP, levels of the α subunit of the AP-2 complex (AP-2α) were also increased on the plasma membrane upon high K+ stimulation (Figure 1—figure supplement 3A-B). Thus, the clustering of PLCδ1-PH-EGFP on the plasma membrane upon stimulation results from an increase in local PI(4,5)P2 concentration. As pointed out by this reviewer, we think that the signal of the diffuse PLCδ1-PH-EGFP probe in Drosophila NMJ boutons is below the detection limit of confocal microscopy, whereas the signal from the clustered probe on the plasma membrane exceeds this limit and thus is much more detectable.

These new data have been added to the above-mentioned figures, and a respective statement has been added to the revised manuscript.

Either way, the whole-bouton quantification presented here is not sufficient, and more rigorous quantification of clustering is required (see point 4).

Since the signals of PLCδ1-PH-EGFP are mainly associated with the plasma membrane in boutons (Figure 1—figure supplement 1), we previously quantified the immunostaining signals of PLCδ1-PH-EGFP from the entire bouton outlined by α-Hrp staining to indicate any change in the plasma membrane association of PLCδ1-PH-EGFP. As suggested, we have now re-quantified PLCδ1-PH-EGFP signal on the presynaptic plasma membrane defined by α-Hrp staining.

For details, please also see our response to point 4.

Finally, there are concerns about potential effects of PH-PLCdelta (which has been used here and elsewhere both as a reporter and manipulation for PIP2). What level of expression were used in Figure 1 compared to 2, and could the "clusters" coincide with malfunctioning endocytosis? Levels should be compared directly (rather than inferred from temperature).

We apologize for the confusion. In Figure 1, we expressed PLCδ1-PH-EGFP using nsyb-GAL4 at 25 °C. We also quantified the level of PLCδ1-PH-EGFP expressed at 25 °C vs 29 °C. As shown in Figure 2—figure supplement 1A-B, compared to its expression at 25 °C, elevating the temperature to 29 °C led to a two-fold increase in probe expression. This information has now been added to the revised manuscript. We also now state the experimental temperatures in all figure legends.

It is conceivable that PLCδ1-PH-EGFP can suppress the binding of PI(4,5)P2 effectors and PI(4,5)P2 catalytic enzymes to PI(4,5)P2 when this reporter is highly expressed. For all figures, we measured the level of PI(4,5)P2 under conditions of mild PLCδ1-PH-EGFP expression (reared at 25 °C). This expression level only mildly affects normal ADBE (Figure 2C-D). Therefore, such a mild ADBE defect should not account for PI(4,5)P2 clustering upon intense stimulation. In support of this notion, similar to PLCδ1-PH-EGFP, the AP-2 complex was recruited to the plasma membrane upon high K+ stimulation (Figure 1—figure supplement 3A-C). However, we cannot exclude the possibility that PLCδ1-PH-EGFP expression may suppress the binding of PI(4,5)P2 catalytic enzymes to slow rapid PI(4,5)P2 turnover, thereby enabling us to capture PI(4,5)P2 clustering. Nevertheless, our data suggest that PI(4,5)P2 microdomains are formed upon intense stimulation.

Due to these questions, some form of independent evidence (published or novel) for PIP2 clustering must be shown or cited.

As also suggested by reviewer #3, we have now investigated recruitment of the AP-2 complex upon PI(4,5)P2 increase. Similar to PLCδ1-PH-EGFP, levels of the α subunit of the AP-2 complex (AP-2α) were also increased and clustered on the plasma membrane upon high K+ stimulation (Figure 1—figure supplement 3A-B). This finding represents further evidence of PI(4,5)P2 clustering on the plasma membrane upon stimulation.

Moreover, we now cite additional studies showing PI(4,5)P2 clustering in different cellular contexts.

2) The consistency of phenotypes among experimental manipulations on Calcium, PIP2, and vesicle dynamics is compelling evidence for a feedback loop. However, not all functional links in the loop are completely tested; for instance, whether CanA1 loss of function leads to calcium signaling defects via synj/PIP2, or through its myriad other targets. While it might be possible to “normalize” synj phosphorylation using minibrain heterozygotes or mutants, this is beyond the scope of suggested revisions. However, where a link in the chain has not been as rigorously tested (as for CanA1), the authors should discuss other possible interpretations of these data (as they do for other possible means of gating Flower).

We have now added further discussion on this topic in the revised manuscript.

3) Figure 2—figure supplement 2- why does FM get taken up to similar extent with high PLC or synj if EM data suggests ABDE is not happening? And why is unloading not impaired if vesicle recycling is diminished?

For our loading/unloading FM4-64 dye uptake assay, we stimulated the NMJ boutons with high K+ to take up the dye for a long period of time, e.g. 5 minutes, so that a majority of both the exo-endo cycling pool (ECP) and reserved pool (RP) were loaded with FM dye. Then we stimulated them with high K+ for 1 minute to release the loaded dye. By calculating the ratio of unloaded dye to loaded dye, we could rule out the effect of SV endocytosis and measure the efficacy of SV exocytosis. Our results revealed no effect on the rate of SV exocytosis when the PLCδ1-PH domain or Synj was expressed. Since we did not image the fluorescence intensities of FM dye under the same imaging detection conditions, the signal intensities of loaded dye by endocytosis cannot be compared directly.

To address the reviewer’s question, we have conducted a FM1-43 dye uptake assay. We stimulated the boutons with a 30-second pulse under conditions of 90 mM K+ and 2 mM Ca2+ to trigger SV exocytosis and prompt the recycling SVs to take up FM dye (Figure 2—figure supplement 2A). We imaged the boutons under different genotypes with same imaging setup. Whereas control boutons exhibited appropriate dye uptake, overexpression of either PLCδ1-PH-EGFP or Synj impaired dye uptake (Figure 2—figure supplement 2B-C). These findings are consistent with ADBE being defective. This new data has been added to Figure 2—figure supplement 2A-C and a respective statement has been added to the revised manuscript.

4) There are several instances where images are not quantified, or where the method of quantification is insufficiently described:

a) Currently Materials and methods indicate only that “image processing and quantification were achieved using LSM Zen and ImageJ”. The authors must describe in detail how fluorescence imaging was processed and quantified. What processing steps were performed? In 2D or 3D? Projections or slices? How were NMJ masks or ROIs created? Materials and methods do not describe what kind of microscope and objective were used for live imaging (important for understanding resolution). All this information is required to correctly interpret the imaging data.

We apologize for this oversight. We have now ensured that quantification data now accompanies all images. Moreover, detailed procedures for our imaging experiments are now provided in the revised Materials and methods (as follows):

For live-imaging of PLCδ1-PH-EGFP:

“Ten images were taken from larval fillets at rest. Muscle contraction significantly decelerated after stimulation for 2 min. Thus, we captured 60 consecutive snapshot images every second from the third minute. Muscles 6 and 7 of abdominal segment 3 were imaged. Under the condition of high K+ stimulation, larval fillets were bathed in a solution of 90 mM K+/2 mM Ca2+/7 mM glutamate (25 mM NaCl, 90 mM KCl, 10 mM MgCl2, 10 mM NaHCO3, 5 mM trehalose, 5 mM HEPES (pH 7.4), 30 mM sucrose, 2 mM CaCl2, 7 mM monosodium glutamate) for 5 min. Sixty consecutive snapshot images were captured every second from the fifth minute of stimulation. Images were taken using a long working distance water immersion objective (XLUMPLFLN20XW, Olympus) and EMCCD camera (iXon, Andor) mounted on a SliceScope Pro 6000 (Scientifica) microscope and employing MetaFluor software (Molecular Devices). For each imaging experiment, at least three focused images for the same boutons under resting, stimulation, or post-stimulation conditions were used for data quantification. Fluorescence intensities of PLCδ1-PH-EGFP within the boutons were quantified using ImageJ and averaged to serve as one individual data-point. Image processing was achieved using LSM Zen.”

For live-imaging of GCaMP6f :

“Ten images were taken from larval fillets at rest. Subsequently, larval fillets were stimulated with a solution of 90 mM K+/2 mM Ca2+/7 mM glutamate for 5 min. Sixty consecutive snapshot images were captured every second from the fifth minute of stimulation. The NMJs were derived from muscles 6 and 7 of abdominal segment 2/3. Images were taken using a water immersion objective (W Plan-Apochromat 40x/1.0 DIC M27, Zeiss). For each imaging experiment, at least three focused images of the same boutons under resting, stimulation, or post-stimulation conditions were used for data quantification. Fluorescence intensities of GCaMP6f within the boutons were quantified using ImageJ and averaged to serve as one individual data-point. Image processing was achieved using LSM Zen.”

For PLCδ1-PH-EGFP or AP-2α immunostaining signal:

“Consecutive single-plane images of the boutons of muscles 6 and 7 in abdominal segments 2 or 3 of all different experimental sets were taken using a Zeiss LSM 780 confocal microscope with a 63X 1.6 numerical aperture oil-immersion objective under a 1-μm interval setup and equal laser power and laser exposure time. The zoom value was adjusted to 2 to obtain maximal imaging resolution. For data quantification, single-plane images of five different individual boutons from each NMJ bouton image were used. The presynaptic plasma membrane regions of the type Ib boutons defined by α-HRP immunostaining were outlined manually, and native fluorescence or α-GFP immunostaining signal intensities of PLCδ1-PH-EGFP or AP-2α on the plasma membrane were quantified using ImageJ and averaged to serve as one individual data-point. Image processing was achieved using LSM Zen.”

For PLA:

“To quantitatively compare PLA signal, fixed larval fillets derived from different experimental conditions were collected into the same Eppendorf tube and processed. Consecutive single-plane images of the boutons of muscles 6 and 7 in abdominal segments 2 or 3 of all different experimental sets were taken using a Zeiss LSM 780 confocal microscope with a 63X 1.6 numerical aperture oil-immersion objective under a 1-μm interval setup and equal laser power and laser exposure time. The zoom value was adjusted to 2 to obtain maximal imaging resolution. For data quantification, consecutive single-plane images were projected, and the type Ib boutons were outlined based on the region of PLCδ1-PH-EGFP immunostaining. PLA or antibody immunostaining signal intensities were quantified using ImageJ and averaged to serve as one individual data-point. Image processing was achieved using LSM Zen.”

For FM1-43 dye assay:

“Consecutive bouton images of different fly genotypes were captured with a Zeiss 780 confocal microscope with a 63 x 1.6 numerical aperture oil-immersion objective under a 1-μm interval setup and equal laser power and laser exposure time. For data quantification, consecutive single-plane images were projected, and the type Ib boutons were outlined according to the region of FM dye labeling. FM dye signal intensities were quantified using ImageJ and averaged to serve as one individual data-point. Image processing was achieved using LSM Zen.”

For FM4-64 dye loading/unloading assay:

“Bouton snapshot images were captured using a water immersion objective (W Plan-Apochromat 40x/1.0 DIC M27, Zeiss) and EMCCD camera (iXon, Andor) mounted on a SliceScope Pro 6000 (Scientifica) microscope and employing MetaFluor software (Molecular Devices). FM dye fluorescence intensities in type Ib boutons from each image were quantified using ImageJ and averaged to serve as one individual data-point. Image processing was achieved using LSM Zen.”

If the reviewer still requests more details, we will add accordingly.

b) Microdomains are poorly described: how is their size quantified? 200nm is very close to the resolution limit of SIM, suggesting many measurements may have been at the floor of the resolution of the system and that this measurement likely represents an overestimate of the size of the domain. It would be informative to see a histogram of values to determine to what degree this is the case, and discuss if appropriate.

We thank the reviewer for raising this issue. To quantify SIM imaging data, we chose single-plane SIM images focused on the central section of individual boutons and outlined the immunostained and plasma membrane-associated PLCδ1-PH-EGFP clusters using ImageJ. The area of these clusters for each individual bouton was measured using ImageJ and then averaged to serve as one data-point. We assessed 37 boutons derived from five NMJs of three different larvae. The area of the clusters in each individual bouton was measured using Image J, and the areas of the clusters over 0.032 μm2 (equal to a ~200-nm diameter circle, i.e., the resolution limit of SIM) were averaged to serve as one data-point. We calculated an average cluster area of 0.07655 μm2 (equal to a ~300-nm diameter circle). We have now revised our statement in the main text. The requalified data have been added to Figure 1F, and respective details have been added to the Materials and methods.

c) How was DAB staining (Figure 6) quantified? What was the criterion used to discriminate between no labeling and “similar to SV” (eg in d, where the overall staining appears to be more similar to a than to b). In the text, the data are presented as both SV and bulk endosome levels decreasing, though these data are not quantified or presented in the figure.

To also ease the reviewer #3’s concern that “DAB staining intensity should not be used to quantify protein levels as the enzymatic reaction may have different constrains on SVs versus endosomes”, we have now re-quantified the DAB signal intensity ratio of bulk endosome to SV to indicate the protein levels of HA-Fwe-APEX2 on bulk endosomes, relative to those on the SVs. With this normalization, we can verify the role of PI(4,5)P2 microdomains in Fwe localization on bulk endosomes. We provide details of description and data quantification in the revised main text, Materials and methods, and figure legends as follows:

“To minimize staining variability from bouton to bouton, we compared DAB staining intensities on bulk endosomes and SVs from the same boutons. As shown in Figure 6B-F, compared to Flag-Fwe-HA rescue boutons, in HA-Fwe-APEX2 rescue boutons, the staining intensities of bulk endosomes were more abundant compared to those of the surrounding SVs, revealing a mechanism by which Fwe is recycled to bulk endosomes after it initiates ADBE.”

When we perturbed PI(4,5)P2 microdomains by Synj overexpression (Figure 2B), the bulk endosome localization of HA-Fwe-APEX2 was significantly reduced (Figure 6D-F), but there was only a mild reduction in SV localization (Figure 6D-E). Therefore, in addition to initiating ADBE, PI(4,5)P2 microdomains play a role in facilitating the retrieval of Fwe to the bulk endosome, enabling ADBE to remove its trigger via a negative feedback regulatory mechanism, thereby reducing endocytosis to prevent excess membrane uptake. We now provide this explanation in the revised manuscript

For data quantification of DAB intensities, the display color of TEM images was reverted to grayscale using ImageJ. Average DAB staining intensity for each individual bulk endosome was assessed. Then the average DAB staining intensities of 50-100 surrounding SVs from the same bouton image were used to assess the relative level of HA-Fwe-APEX2 on bulk endosomes vs SVs.

We add quantification data for the DAB staining intensity ratio of bulk endosomes to SVs the surrounding SVs in Figure 6F. The number of bulk endosomes, NMJ boutons, and larvae counted (Flag-Fwe-HA rescue control: Bulk endosomes (n=33) derived from 5 NMJ boutons of two different larvae; HA-Fwe-APEX2 rescue: Bulk endosomes (n=47) derived from 6 NMJ boutons of two different larvae; HA-Fwe-APEX2 rescue expressing Synj: Bulk endosomes (n=59) derived from 9 NMJ boutons of two different larvae. Individual values were shown in graphs. This description is now stated in figure legends.

5) If fwe initiates a positive feedback loop, why do only high levels of activity stimulate clustering and ADBE?

Our data show that direct binding of PI(4,5)P2 is required for the Ca2+ channel activity of Fwe. Perturbation of PI(4,5)P2-Fwe binding further impaired the formation of PI(4,5)P2 microdomains, as well as ADBE initiation. Hence, PI(4,5)P2 controls Fwe gating, so that Fwe can promote PI(4,5)P2 compartmentalization through positive feedback regulation. Furthermore, loss of Fwe impaired the intracellular Ca2+ increase evoked upon strong activity stimulation (Figure 3I-J) (Yao et al., 2017). These results indicate that, in addition to PI(4,5)P2, the channel function of Fwe may be gated by a significant change in membrane potential. Therefore, it is possible that both factors may gate Fwe, thereby only allowing channel opening when exocytosis targets Fwe to periactive zones. Future studies should explore the details of this channel gating mechanism. These points have now been added to our revised Discussion.

6) The claim that Fwe is “selectively sorted” by PIP2 seems overstated, especially given the companion idea that Fwe sorts PIP2 into microdomains. In general, the data in Figure 6 are a bit unconvincing or overinterpreted:

Acknowledged. We have now toned down this conclusion and revised the statement in our Introduction to: “PI(4,5)P2 also participates in retrieval of Fwe to the bulk endosome, thereby stopping membrane recycling.”. We have also revised our Abstract and Results.

a) AP2 complexes strongly associated with PIP2 microdomains- this is completely unsupported by the data, which show an exceedingly low correlation.

b) How were “colocalized” pixels determined in Figure 6H (not described in legend or Materials and methods)? Further, the figure legend for 6I indicates that “Quantification data for Pearson correlation coefficients” is the overlapping area divided by the overall area. This is not a Pearson's correlation coefficient, which rather depends on the linear relationship between channel intensities of all (or a subset) of pixels.

c) “synj altered the localization of the AP2 complex”- this is not quantified and is very difficult to determine from the images presented. This is also a clear example where it is not obvious that images are being taken from the same Z-plane, which further confounds interpretation of the data.

d) AP2-fwe IPs are not very convincing. The difference between negative control and IP is subtle, only a very small fraction of input, and neither repeated or quantified. These data should be strengthened or removed.

e) Is synj expressed at room temperature or 29 degrees?

As suggested by reviewer #1, since our data on Fwe and AP-2 complex association (original Figure 6G-H and text) were deemed superfluous, we have removed this topic from the revised manuscript.

Reviewer #3:

Previous work by the Yao lab showed that the SV-associated Ca2+ channel Flower (Fwe) is required for both clathrin-mediated endocytosis (CME) of synaptic vesicles (SVs) from the plasma membrane (PM) and activity-dependent bulk endocytosis (ADBE). Importantly, Fwe has likely a distinct role for each endocytotic route since only ADBE requires Ca2+ channel function. This follow-up study pursues the mechanisms underlying Fwe's role for ABDE. The study provides evidence for an interesting model of Fwe's role in ADBE that may significantly advance our understanding of mechanisms underlying ADBE. Specifically, it suggests that freshly exocytosed Fwe on the PM forms a positive Ca2+-dependent feedback loop that facilitates the generation and accumulation of phosphatidylinositol 4,5-bisphosphate (PIP2) in microdomains on the PM. In turn, accumulating PIP2 may facilitate compartmentalization of Fwe to bulk endosomes, thereby terminating ADBE. I have a number of concerns.

1) To visualize PIP2 at fly NMJs, the entire study relies only on a single PIP2 marker, the Gal4-driven expression of PLCδ1-PH-EGFP. Unfortunately, this marker not only reduces "free" PIP2 levels but also has dominant-negative effects on synaptic growth and SV cycling (Khuong et al., 2010). To properly support the notion that intense stimulation causes accumulation of PIP2 in microdomains on the PM, a number of corrections are required:

a) the PLCδ1-PH-EGFP imaging experiments require a proper control, like the expression of PLCδ1-PHS39R-GFP, which does not bind PIP2. Measuring mCD-EGFP expression levels is not a proper control.

As suggested, we have now examined the expression and distribution of PLCδ1-PHS39R-EGFP at rest and after high K+ stimulation. We found that, unlike PLCδ1-PH-EGFP, PLCδ1-PHS39R-EGFP mainly localized in the cytosol of cells at rest. After intense stimulation, immunostaining signals of PLCδ1-PHS39R-EGFP did not accumulate on the plasma membrane (Figure 1—figure supplement 2C-D) and were not increased within the boutons (Figure 1—figure supplement 2E). Thus, the clustering of PLCδ1-PH-EGFP on the plasma membrane upon stimulation results from an increase in local PI(4,5)P2 concentration. These new data have been added to Figure 1—figure supplement 2C-E, and a respective statement has been added to the revised manuscript.

b) the stimulation-induced accumulation of PIP2 should be verified by a second independent marker like quantitative immunostainings for edogenous α-adaptin or other PIP2-binding proteins. This is especially important to rule out unaccounted dominant-negative effects.

As requested, we have now assessed recruitment of endogenous α-adaptin upon high K+ stimulation. We found that, similar to PLCδ1-PH-EGFP, levels of the α subunit of the AP-2 complex (AP-2α) were also increased on the plasma membrane upon high K+ stimulation (Figure 1—figure supplement 3A-B). The new data have been added to Figure 1—figure supplement 3A-B and a respective statement has been added to the revised manuscript.

c) It is unclear how changes in PLCδ1-PH-EGFP fluorescence (Figure 1A-B, D, plus all subsequent experiments using PLCδ1-PH-EGFP including Figure 3C-F, 4F-G) are measured at larval NMJs since no information about respective ROIs is provided. To make a strong statement about a PIP2 accumulation on the PM, PLCδ1-PH-EGFP fluorescence measured for an entire bouton (from one or a few slices) should be subtracted by the PLCδ1-PH-EGFP fluorescence obtained from the cytosol of the bouton. Measurements using ROIs for the entire bouton do not allow specific conclusions about PM levels.

We thank the reviewer for raising our oversight. As also suggested by reviewer #2, we have now re-quantified the signals of PLCδ1-PH-EGFP on the presynaptic plasma membranes, which were defined by α-Hrp staining.

For details of our revision, please see our response to point 4 in the comments of reviewer #2.

2) To significantly strengthen the conclusion that PIP2 accumulation drives ADBE initiation and SV reformation (Figure 2) a proper control should be used for the postulated PLCδ1-PH-EGFP-mediating reduction of PIP2 levels. Ideally, one may use expression of the PLCδ1-PHS39R-GFP mutant. Overexpression of synaptojanin alone does not make the point that the reduced ADBE levels are due to PIP2 depletion since synaptojanin also affects several other phosphatidylinositides.

We appreciate this suggestion. We conducted the recommended experiments and found that when larvae expressing PLCδ1-PHS39R-EGFP were grown at 29 ºC, the mutant protein failed to suppress ADBE (Figure 2C-D). The new data have been added to Figure 2C-D, and a respective statement has been added to the revised manuscript.

3) The proximity ligation experiments of Figure 3A require a proper control like PLCδ1-PHS39R-GFP co-expression with Flag-Fwe-HA. Otherwise, the observed proximity could be due to other factors.

As suggested, we have now performed a PLA assay to assess the association of Fwe with PLCδ1-PHS39R-EGFP. As shown in Figure 1—figure supplement 1 and 2, whereas PLCδ1-PH-EGFP mainly localized on the plasma membrane, PLCδ1-PHS39R-EGFP localized in the cytosol, both at rest and even upon high K+ stimulation. We further found that PLCδ1-PHS39R-EGFP indeed locates to the SVs marked by the SV proteins, Csp and Fwe (see Author response image 1). Consequently, when we did the PLA assay on the Flag-Fwe-HA rescue boutons expressing PLCδ1-PHS39R-EGFP, the close proximity between Fwe and SV-localized PLCδ1-PHS39R-EGFP due to 30-50 nm-sized SV already generated very strong PLA signals when the boutons were at rest and even after high K+ stimulation (see Author response image 2). Therefore, since PLCδ1-PH-EGFP and PLCδ1-PHS39R-EGFP have totally different subcellular localizations, we are not comfortable using PLCδ1-PHS39R-EGFP as a control protein in our PLA assay.

Author response image 1
Download asset Open asset
PLCδ1-PHS39R-EGFP locates to SVs marked by Csp or Fwe.
Author response image 2
Download asset Open asset
A close association between PLCδ1-PHS39R-EGFP and Flag-Fwe-HA on the SVs generates PLA signals at rest or after high K+ stimulation.

4) The images of Figure 3C, E show little evidence of PLCδ1-PH-EGFP microdomains at the PM, much in contrast to Figure 1. Does this mean that the quantification associated with these figures mostly reflects PIP2 on cytosolic bulk endosomes?

We apologize for the confusion. As also suggested by reviewer #2, we have replaced Z-projected images with single-plane images that focus on the central section of the boutons to represent protein distribution. PLCδ1-PH-EGFP microdomains were found mostly on the plasma membrane.

5) The shown knockdown data of CanA1 in Figure 3E-H require a control for specificity, best co-expression of an RNAi-resistant WT transgene.

The UAS-canA1-RNAi (FB5) used in our work was generated by cloning two copies of a region of CanA1 (320–1160 bp) in reverse orientation spaced by an 800 bp intron from CG7464 (Dijkers and O’Farrell, et al., 2007). It is therefore difficult to predict the functional small RNA fragments generated by this construct and thus to design an RNAi-resistant canA1 transgene. As an alternative, we expressed another UAS-canA1 RNAi (JF01871) (Bloomington Drosophila Stock Center, BDSC#25850), which was generated by using different regions of the CanA1 coding region, the funcionality of which had been tested previously (Wong et al., 2014). We found that expression of UAS-canA1 RNAi (JF01871) using nsyb-GAL4 suppressed PI(4,5)P2 microdomain formation (Figure 3H) and ADBE initiation (Figure 5A-B). These new data strengthen the role for CanA1 in PI(4,5)P2 and ADBE induction. These new data have been added in the figures indicated above, and respective statements have been added to the revised manuscript.

6) The study suggests that PIP2 binding to freshly exocytosed Fwe gates Fwe's Ca2+ channel activity. In turn, the resulting Ca2+ influx activates calcineurin and the postulated feedback loop accumulating PIP2 at the PM. If correct, one would expect that a mutation that several affects Fwe' Ca2+ channel activity (like FweE79Q) should significantly reduce the accumulation of PIP2. Is this indeed the case? If not, the Fwe-dependent PIP2 accumulation may occur through a different mechanism.

As suggested, we have now conducted experiments to determine if the Ca2+ channel activity of Fwe is involved in this process. We undertook rescue experiments based on our previous report in which fwe mutant boutons exhibited expression of the wild-type Fwe transgene or the FweE79Q mutant transgene that exhibits reduced Ca2+ conductance (Yao et al., 2017). We found that the wild-type Fwe transgene promoted PI(4,5)P2 microdomain formation, whereas the FweE79Q mutant transgene lost that ability (Figure 3E-F). Thus, Fwe triggers the formation of PI(4,5)P2 microdomains in a Ca2+-dependent manner. The new data have been added to Figure 3E-F, the detailed genotypes have been added to the respective figure legend, and an explanatory statement has been added to the revised manuscript.

7) The HA-Fwe-APEX2 experiments of Figure 6 require further controls including documentation that the fusion protein is properly localized to SVs and not excessively retained on the PM during ADBE. The EM alone is not convincing.

As suggested, we have now performed confocal imaging analysis to verify the SV localization of HA-Fwe-APEX2. As shown in Figure 6—figure supplement 1A, HA-Fwe-APEX2 (stained with α-Fwe) was highly colocalized with the specific SV markers, Syt and Csp. Furthermore, expression of this fusion protein was able to rescue the endocytic defects (Figures 4 and 5) and early animal lethality caused by loss of fwe. Therefore, HA-Fwe-APEX2 is functionally equivalent to endogenous Fwe. In our TEM images (Figure 6—figure supplement 1C-D), DAB signals generated by HA-Fwe-APEX2 were found on SVs (yellow arrows) and on the plasma membrane (white arrows) under both resting and stimulation conditions. These new data have been added to Figure 6—figure supplement 1, and a respective statement has been added to the revised manuscript.

8) The co-immunoprecipitation of Fwe with AP2 shown in Figure 6G requires quantification.

As suggested by reviewer #1, since the data on Fwe and AP-2 complex association (previous Figure 6G-H and text) was deemed superfluous, we have now removed this data from the revised manuscript.

9) The evidence for the conclusion that accumulating PIP2 facilitates "sorting" of Fwe for ADBE is weak because it essentially rests on the HA-Fwe-APEX2 experiment. Since reducing PIP2 levels by synaptojanin expression reduced the bulk endosome localization of Fwe as judged by DAB staining levels (Figure 6D-F), the authors conclude that "PIP2 microdomains can facilitate the retrieval of Fwe to the bulk endosome". However, this appears to be somewhat a catch22 since reduced PIP2 levels decrease the overall amount of bulk endocytosis. Whether the potentially reduced levels of internalized Fwe are disproportionally reduced has not been addressed.

We thank the reviewer for raising this point. In our data (Figure 6C-E), we controlled animal growth temperature at 25 °C to express a mild level of Synj. Under this condition, the suppressive effect on the formation of PI(4,5)P2 microdomains (Figure 2B) is much stronger than that on ADBE initiation (Figure 2C-D), and increasing Synj expression further abolished ADBE (Figure 2C-D). Therefore, our data argue that reduced Fwe localization in the bulk endosome is unlikely to be due to a reduction in the overall amount of bulk endocytosis. Instead, PI(4,5)P2 plays a role in retrieval of Fwe to the bulk endosomes.

In addition, DAB staining intensity should not be used to quantify protein levels as the enzymatic reaction may have different constrains on SVs versus endosomes.

Upon H2O2 activation, APEX2 catalyzes DAB polymerization, and DAB polymers proximally label the phospholipids where the APEX2-fused proteins are closely localized. Synaptic vesicles and bulk endosomes are generated freshly from the plasma membrane upon exocytosis. Therefore, both vesicular compartments are expected to hold very similar compositions of phospholipids, glycolipids, and cholesterols, even though they also contain distinct types of phosphoinositides as minor components. We certainly agree that the enzymatic reaction may have different constraints on SVs versus endosomes. It has not been demonstrated that the lipid and protein composition of the membrane affects DAB labeling. Therefore, we assume that the DAB staining intensity is positively related to the membrane density of the APEX2-fused Fwe protein on both bulk endosomes and SVs. Even so, to ease the reviewer’s concern, we have now re-quantified the DAB signal intensity ratio of bulk endosome to SV to indicate the protein levels of HA-Fwe-APEX2 on bulk endosomes, relative to those on the SVs. With this normalization, we can verify the role of PI(4,5)P2 microdomains in Fwe localization on bulk endosomes. We provide details of description and data quantification in the revised main text, Materials and methods, and figure legends as follows:

“To minimize staining variability from bouton to bouton, we compared DAB staining intensities on bulk endosomes and SVs from the same boutons. As shown in Figure 6B-F, compared to Flag-Fwe-HA rescue boutons, in HA-Fwe-APEX2 rescue boutons, the staining intensities of bulk endosomes were more abundant compared to those of the surrounding SVs, revealing a mechanism by which Fwe is recycled to bulk endosomes after it initiates ADBE.”

When we perturbed PI(4,5)P2 microdomains by Synj overexpression (Figure 2B), the bulk endosome localization of HA-Fwe-APEX2 was significantly reduced (Figure 6D-F), but there was only a mild reduction in SV localization (Figure 6D-E). Therefore, in addition to initiating ADBE, PI(4,5)P2 microdomains play a role in facilitating the retrieval of Fwe to the bulk endosome, enabling ADBE to remove its trigger via a negative feedback regulatory mechanism, thereby reducing endocytosis to prevent excess membrane uptake. We now provide this explanation in the revised manuscript.

For data quantification of DAB intensities, the display color of TEM images was reverted to grayscale using ImageJ. Average DAB staining intensity for each individual bulk endosome was assessed. Then the average DAB staining intensities of 50-100 surrounding SVs from the same bouton image were used to assess the relative level of HA-Fwe-APEX2 on bulk endosomes vs SVs.

We add quantification data for the DAB staining intensity ratio of bulk endosomes to SVs the surrounding SVs in Figure 6F. The number of bulk endosomes, NMJ boutons, and larvae counted (Flag-Fwe-HA rescue control: Bulk endosomes (n=33) derived from 5 NMJ boutons of two different larvae; HA-Fwe-APEX2 rescue: Bulk endosomes (n=47) derived from 6 NMJ boutons of two different larvae; HA-Fwe-APEX2 rescue expressing Synj: Bulk endosomes (n=59) derived from 9 NMJ boutons of two different larvae. Individual values were shown in graphs. This description is now stated in figure legends.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The authors have provided a meaningfully improved manuscript that largely addresses the major concerns of the paper. In particular, the authors provide several new lines of evidence that levels or organization of PI(4,5)P2 are likely changing in response to stimulation: There is no change in the new negative control PH S39 mutant, AP2a levels increase at the membrane, and the levels of native and anti-GFP fluorescence of PH-PLC increase at the membrane. Especially given the strong correspondence between in vitro and in vivo data and the consistent phenotypes among the perturbations in the proposed pathway, these new data further strengthen the main claim that Fwe acts in a positive feedback loop with Ca++ and PI(4,5)P2.

It is worth noting that the precise nature of the microdomains remains a bit unclear. It is difficult to discern whether PI(4,5)P2 microdomains are forming de novo following stimulation, or whether PI(4,5)P2 ¬signal increases in existing microdomains (for example, AP2a recruitment to the membrane is quite high even under resting conditions). And as noted by the authors, while the PH domain is the best available tool to answer the questions posed, it remains possible that properties of the microdomains partially reflect the function-perturbing aspects of the probe. However, these caveats do not significantly detract from the core conclusions of the paper.

Figure 3H. Something is wrong with the shown KD data by UAS-canA1-RNAi(FB5) or the drawn conclusion that "reducing CanA1 levels via expression of either RNAi construct greatly suppressed the formation of PI(4,5)P2 microdomains compared to the control (Figure 3G-H)". The resting levels for the canA1 KD are only 50% of control. The K+-induced increase PLCδ1-PH-EGFP in the KD is as large as in control. Accordingly, the KD seems to have no effect, which would suggest that one of the two KDs is not specific for canA1.

If the FB5 KD is not working, the one needs a wild type CanA1 co-expression for the KD with the canA1-RNAi(JF01871) transgene to demonstrate specificity.

Please clarify.

Figure 3I. The canA1 KD uses the controversial UAS-canA1-RNAi(FB5) transgene, which does not appear to block the formation of PI(4,5)P2 microdomains (Figure 3H). The experiment should be done with the UAS-canA1-RNAi(JF01871) transgene as only this transgene reliable affects PI(4,5)P2 microdomains (assuming the KD can be rescued). Otherwise, one cannot conclude that "Calcineurin mediates the Ca2+ influx conducted by Fwe to induce PI(4,5)P2 microdomains". Resolving this issue is critical.

For suggested rescue experiment, we have carefully gone through public resource indicated in Flybase. Unfortunately, a wild-type canA1 transgene is not available. If we would begin the experiment by establishing a wild-type canA1 transgene, it will take at least 3-4 months to be able to get the result. Alternatively, we have used a UAS transgene of a constitutively active form of CanA1 (UAS-CanA1.δ.HA(4B)), which was published by Dijkers and O’Farrell, 2007, and is available in Bloomington Drosophila Stock Center (#83343). However, we observed that expression of UAS-CanA1.δ.HA(4B) alone already inhibited the formation of PI(4,5)P2 microdomains in high K+ stimulation and expression of PI(4,5)P2 at rest (see Author response image 3). Therefore, sustained high CanA1 activity can also perturb the PI(4,5)P2 increase after strong stimulation and even lower the resting PI(4,5)P2 concentrations. One would expect that expression of a wild-type canA1 transgene likely causes similar effects. Therefore, we are not comfortable to assess the rescue effect by constitutively increasing CanA1 enzymatic activity if dynamic CanA1 activity is needed to promote PI(4,5)P2 microdomain formation.

Author response image 3
Download asset Open asset

To address the reviewer’s concern, we have tested another independent line of UAS-canA1-RNAi (FB)(Dijkers and O’Farrell, 2007), UAS-canA1-RNAi (FB4). FB4 and FB5 lines carries identical UAS-canA1-RNAi transgene, but their canA1-RNAi transgenes are inserted on different chromosomes, e.g. second chromosome for FB4 and third chromosome for FB5. Unlike UAS-canA1-RNAi (FB5), we found that expression of UAS-canA1-RNAi (FB4) significantly impaired the formation of PI(4,5)P2 microdomains but did not affect the resting level of PI(4,5)P2 (Figure 3H), which is close to the effects of UAS-canA1-RNAi (JF). Thus, the FB5 line may have an unclean genetic background which influences the resting levels of PI(4,5)P2. In addition, we have examined the effect of expression of both UAS-CanA1-RNAi (JF01871) and UAS-CanA1-RNAi (FB4) on intracellular Ca2+ increase upon high K+ stimulation. As shown in Figure 3I-J, expression of UAS-CanA1-RNAi (JF01871) or UAS-CanA1-RNAi (FB4) was able to suppress intracellular Ca2+ increase upon high K+ stimulation. We hope that the new data of two independent RNAi lines we added can support the role of CanA1 in the PI(4,5)P2 microdomain formation. We have done the best we can do during limited revision time. The new data have been added to Figure 3H-J, and a respective statement has been added to the revised manuscript. Accordingly, we have removed the results of UAS-canA1-RNAi (FB5) from Figures 3 and 5.

"Together, these data suggest that the PI(4,5)P2 microdomains initiate ADBE to control SV membrane retrieval upon intense stimulation." This sentence should be rephrased. The shown FM1-43 data (Figure 2—figure supplement 2B) only suggest that overexpression of PLCδ1-PH-EGFP or Synj impairs SV endocytosis independent of the mode of endocytosis. The shown FM1-43 data do not exclude a role of PI(4,5)P2 for other modes of endocytosis.

Moreover, the FM1-43 data use a much shorter (30 seconds) and milder stimulation than the solid EM data visualizing ADBE (10 minutes; Figure 2). Therefore, the FM1-43 and EM data are not directly comparable.

The same concern applies to the FM1-43-visulaized SV exocytosis data, which indicate that SV exocytosis is not impaired by overexpression of PLCδ1-PH-EGFP or Synj if endocytosis is stimulated for 1 minute. Again, the data are not comparable to the ADBE defect shown in Figure 2. I suggest either to omit the data of Figure 2—figure supplement 2 or use the same stimulation paradigm for the FM1-43 and EM data.

As suggested, we have removed Figure 2—figure supplement 2 and the statement.

https://doi.org/10.7554/eLife.60125.sa2

Article and author information

Author details

  1. Tsai-Ning Li

    Institute of Biological Chemistry, Academia Sinica, Taipei, Taiwan
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0195-145X

  2. Yu-Jung Chen

    Institute of Biological Chemistry, Academia Sinica, Taipei, Taiwan
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared

  3. Ting-Yi Lu

    Institute of Biological Chemistry, Academia Sinica, Taipei, Taiwan
    Contribution
    Data curation, Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared

  4. You-Tung Wang

    Institute of Biological Chemistry, Academia Sinica, Taipei, Taiwan
    Contribution
    Resources, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology
    Competing interests
    No competing interests declared

  5. Hsin-Chieh Lin

    Institute of Biological Chemistry, Academia Sinica, Taipei, Taiwan
    Contribution
    Resources, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology
    Competing interests
    No competing interests declared

  6. Chi-Kuang Yao

    1. Institute of Biological Chemistry, Academia Sinica, Taipei, Taiwan
    2. Neuroscience Program of Academia Sinica, Academia Sinica, Taipei, Taiwan
    3. Institute of Biochemical Sciences, College of Life Science, National Taiwan University, Taipei, Taiwan
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    ckyao@gate.sinica.edu.tw
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0977-4347

Funding

Ministry of Science and Technology, Taiwan (107-2311-B-001-003-MY3)

  • Chi-Kuang Yao

Ministry of Science and Technology, Taiwan (106-0210-01-15-02)

  • Chi-Kuang Yao

Ministry of Science and Technology, Taiwan (107-0210-01-19-01)

  • Chi-Kuang Yao

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Hugo Bellen, Patrik Verstreken, Kartik Venkatachalam, the Bloomington Drosophila Stock Center, and the Developmental Studies Hybridoma Bank for stocks and reagents. We thank Hugo Bellen, Ruey-Hwa Chen, and Y Henry Sun for the critical comments. We thank Yi-Shuian Huang for providing the Nluc plasmid. We thank DNA Sequencing Core Facility (AS-CFII-108–115) for sequencing DNA constructs. We thank Wellgenetics for making transgenic lines. We thank the IMB imaging core for helping with TEM and SIM imaging. This work was supported by grants from the Ministry of Science and Technology of Taiwan (107–2311-B-001–003-MY3, 106-0210-01-15-02, and 107-0210-01-19-01).

Senior Editor

  1. Richard W Aldrich, The University of Texas at Austin, United States

Reviewing Editor

  1. Hugo J Bellen, Baylor College of Medicine, United States

Reviewer

  1. Patrik Verstreken, KU Leuven, Belgium

Version history

  1. Received: June 17, 2020
  2. Accepted: December 9, 2020
  3. Accepted Manuscript published: December 10, 2020 (version 1)
  4. Version of Record published: December 18, 2020 (version 2)

Copyright

© 2020, Li et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,848
    Page views
  • 286
    Downloads
  • 4
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Tsai-Ning Li
  2. Yu-Jung Chen
  3. Ting-Yi Lu
  4. You-Tung Wang
  5. Hsin-Chieh Lin
  6. Chi-Kuang Yao
(2020)
A positive feedback loop between Flower and PI(4,5)P2 at periactive zones controls bulk endocytosis in Drosophila
eLife 9:e60125.
https://doi.org/10.7554/eLife.60125

Categories and tags

Research organism

Further reading

    1. Genetics and Genomics
    2. Neuroscience

    Translation of dipeptide repeat proteins in C9ORF72 ALS/FTD through unique and redundant AUG initiation codons

    Yoshifumi Sonobe, Soojin Lee ... Paschalis Kratsios
    Research Article Updated

    A hexanucleotide repeat expansion in C9ORF72 is the most common genetic cause of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). A hallmark of ALS/FTD pathology is the presence of dipeptide repeat (DPR) proteins, produced from both sense GGGGCC (poly-GA, poly-GP, poly-GR) and antisense CCCCGG (poly-PR, poly-PG, poly-PA) transcripts. Translation of sense DPRs, such as poly-GA and poly-GR, depends on non-canonical (non-AUG) initiation codons. Here, we provide evidence for canonical AUG-dependent translation of two antisense DPRs, poly-PR and poly-PG. A single AUG is required for synthesis of poly-PR, one of the most toxic DPRs. Unexpectedly, we found redundancy between three AUG codons necessary for poly-PG translation. Further, the eukaryotic translation initiation factor 2D (EIF2D), which was previously implicated in sense DPR synthesis, is not required for AUG-dependent poly-PR or poly-PG translation, suggesting that distinct translation initiation factors control DPR synthesis from sense and antisense transcripts. Our findings on DPR synthesis from the C9ORF72 locus may be broadly applicable to many other nucleotide repeat expansion disorders.

    1. Cell Biology
    2. Neuroscience

    Removal of extracellular human amyloid beta aggregates by extracellular proteases in C. elegans

    Elisabeth Jongsma, Anita Goyala ... Collin Yvès Ewald
    Research Article Updated

    The amyloid beta (Aβ) plaques found in Alzheimer’s disease (AD) patients’ brains contain collagens and are embedded extracellularly. Several collagens have been proposed to influence Aβ aggregate formation, yet their role in clearance is unknown. To investigate the potential role of collagens in forming and clearance of extracellular aggregates in vivo, we created a transgenic Caenorhabditis elegans strain that expresses and secretes human Aβ1-42. This secreted Aβ forms aggregates in two distinct places within the extracellular matrix. In a screen for extracellular human Aβ aggregation regulators, we identified different collagens to ameliorate or potentiate Aβ aggregation. We show that a disintegrin and metalloprotease a disintegrin and metalloprotease 2 (ADM-2), an ortholog of ADAM9, reduces the load of extracellular Aβ aggregates. ADM-2 is required and sufficient to remove the extracellular Aβ aggregates. Thus, we provide in vivo evidence of collagens essential for aggregate formation and metalloprotease participating in extracellular Aβ aggregate removal.

    1. Computational and Systems Biology
    2. Neuroscience

    Task-dependent optimal representations for cerebellar learning

    Marjorie Xie, Samuel P Muscinelli ... Ashok Litwin-Kumar
    Research Article Updated

    The cerebellar granule cell layer has inspired numerous theoretical models of neural representations that support learned behaviors, beginning with the work of Marr and Albus. In these models, granule cells form a sparse, combinatorial encoding of diverse sensorimotor inputs. Such sparse representations are optimal for learning to discriminate random stimuli. However, recent observations of dense, low-dimensional activity across granule cells have called into question the role of sparse coding in these neurons. Here, we generalize theories of cerebellar learning to determine the optimal granule cell representation for tasks beyond random stimulus discrimination, including continuous input-output transformations as required for smooth motor control. We show that for such tasks, the optimal granule cell representation is substantially denser than predicted by classical theories. Our results provide a general theory of learning in cerebellum-like systems and suggest that optimal cerebellar representations are task-dependent.