Clathrin/AP-2 Mediate Synaptic Vesicle Reformation from Endosome-like Vacuoles but Are Not Essential for Membrane Retrieval at Central Synapses

Transcription

1 Neuron Report Clathrin/AP-2 Mediate Synaptic Vesicle Reformation from Endosome-like Vacuoles but Are Not Essential for Membrane Retrieval at Central Synapses Natalia L. Kononenko, 1,2 Dmytro Puchkov, 1 Gala A. Classen, 1,2 Alexander M. Walter, 1,2 Arndt Pechstein, 1 Linda Sawade, 1 Natalie Kaempf, 1 Thorsten Trimbuch, 2 Dorothea Lorenz, 1 Christian Rosenmund, 2 Tanja Maritzen, 1 and Volker Haucke 1,2, * 1 Leibniz Institut für Molekulare Pharmakologie (FMP) & Freie Universität Berlin, Robert-Roessle-Straße 10, Berlin, Germany 2 Charite Universitätsmedizin, NeuroCure Cluster of Excellence, Virchowweg 6, Berlin, Germany *Correspondence: haucke@fmp-berlin.de SUMMARY Neurotransmission depends on presynaptic membrane retrieval and local reformation of synaptic vesicles (SVs) at nerve terminals. The mechanisms involved in these processes are highly controversial with evidence being presented for SV membranes being retrieved exclusively via clathrin-mediated endocytosis (CME) from the plasma membrane or via ultrafast endocytosis independent of clathrin. Here we show that clathrin and its major adaptor protein 2 (AP-2) in addition to the plasma membrane operate at internal endosome-like vacuoles to regenerate SVs but are not essential for membrane retrieval. Depletion of clathrin or conditional knockout of AP-2 result in defects in SV reformation and an accumulation of endosome-like vacuoles generated by clathrin-independent endocytosis (CIE) via dynamin 1/3 and endophilin. These results together with theoretical modeling provide a conceptual framework for how synapses capitalize on clathrinindependent membrane retrieval and clathrin/ AP-2-mediated SV reformation from endosome-like vacuoles to maintain excitability over a broad range of stimulation frequencies. INTRODUCTION Neurotransmission depends on the endocytic recycling of synaptic vesicles (SVs) at nerve terminals. During sustained activity efficient endocytic membrane retrieval is required to keep membrane surface area constant and SVs need to be reformed to replenish the SV pool (Dittman and Ryan, 2009; Saheki and De Camilli, 2012). The mechanisms involved in membrane retrieval and SV reformation have been debated controversially over more than four decades (Ceccarelli et al., 1972; Cheung and Cousin, 2013). Early studies of frog neuromuscular junctions (Heuser and Reese, 1973) and subsequent work on mammalian synapses have suggested that SVs reform by clathrin/ap-2- mediated endocytosis (CME) from the plasma membrane (Dittman and Ryan, 2009; Granseth et al., 2006; Kim and Ryan, 2009; Saheki and De Camilli, 2012), a pathway that crucially depends on clathrin and its major adaptor AP-2 (Dittman and Ryan, 2009; McMahon and Boucrot, 2011). A key role for clathrin/ap-2 in SV recycling is supported by the association of SV proteins with clathrin adaptors such as AP-2 and stonin 2 (Kononenko et al., 2013) and by the fact that clathrin-coated vesicles isolated from nerve terminals are highly enriched in SV proteins (Maycox et al., 1992). Recent data using single optogenetic stimuli in the presence of high calcium concentrations paired with rapid flashand-freeze electron microscopy have challenged the proposal that SV membrane retrieval occurs exclusively via CME (Granseth et al., 2006) and instead suggest that SV membranes are recaptured via ultrafast (Watanabe et al., 2013) clathrin-independent endocytosis (CIE) involving vacuolar membrane invaginations akin to bulk endocytic membrane retrieval (Cheung and Cousin, 2013). Hence, in spite of more than 40 years of research, the role of clathrin and AP-2 in SV recycling has remained enigmatic. Here we show that clathrin and AP-2 mediate SV reformation from internal endosomal structures, whereas presynaptic membrane retrieval largely occurs by clathrin-independent endocytosis via dynamin 1/3 and endophilin. Our results provide a conceptual framework for how synapses capitalize on clathrin-independent endocytosis and clathrin/ap-2- mediated SV reformation from endosome-like vacuoles to maintain excitability. RESULTS Frequency Modulation of SV Membrane Retrieval Mammalian CNS neurons can respond to a broad range of physiological stimulation patterns ranging from a few to tens of hertz (Hz). One possible reason for the apparent discrepancies between various studies on SV recycling, thus, is that the mode of presynaptic membrane retrieval may depend on the frequency of stimulation. To probe how stimulation frequency affects SV membrane retrieval in hippocampal neurons, we employed the ph-sensitive fluorescent protein phluorin fused to the lumenal domain of the SV calcium sensor synaptotagmin 1 (Syt1- phluorin) (Kononenko et al., 2013). PHluorins undergo quenching within the acidic SV lumen and are dequenched upon exocytotic fusion, thereby serving as reporters for SV exo-/endocytosis Neuron 82, , June 4, 2014 ª2014 Elsevier Inc. 981

2 Neuron SV Reformation from Endosomes via Clathrin/AP-2 Figure 1. Frequency Modulation of Endocytic Membrane Retrieval at Hippocampal Synapses (A) Average traces of hippocampal neurons expressing Syt1-pHluorin in response to 200 APs applied at 5 Hz, followed by 200 APs applied at 40 Hz (normalized to Syt1-pHluorin peak fluorescence at 40 Hz). (B) Decreased relative peak fluorescence of Syt1-pHluorin at 5 Hz (41% ± 5%) versus 40 Hz (100% ± 12%, p = 0.001). (C) Average Syt1-pHluorin traces (from neurons coexpressing scr-shrna) in response to 200 APs at 5 Hz or 40 Hz in the absence or presence of folimycin (normalized to maximum F change after 200 APs). n > 400 boutons. (D) Fraction of duringstimulus endocytosis during sustained synaptic transmission at 5 Hz and 40 Hz. During-stimulus endocytosis in response to 200 APs at 5 Hz was set to 100% (100% ± 5.46%, p < ). Membrane retrieval in response to 200 APs at 40 Hz is mainly supported by poststimulus endocytosis (fraction of during-stimulus endocytosis: 1.08% ± 5.46%). (Ei Eiv and Fi Fiv) Electron micrographs of endocytic intermediates at hippocampal synapses challenged with 200 APs delivered at 5 Hz (Ei Eiv) or at 40 Hz (Fi Fiv). Clathrin-coated pits (CCPs) are frequently observed at synapses stimulated at 5 Hz (Ei Eiv), while endosome-like vacuoles (ELVs) dominate at 40 Hz (Fi Fiv). Scale bar, 100 nm. (G) Average number of CCPs in boutons stimulated at 5 Hz (0.93 ± 0.11) or at 40 Hz (0.21 ± 0.05, p < ). (H) Average number of ELVs in boutons stimulated at 40 Hz (5.68 ± 0.50) or at 5 Hz (2.37 ± 0.26, p < ). n > 200 synapses/condition. All data represent mean ± SEM. See also Figure S1. and subsequent reacidification (Granseth et al., 2006; Kim and Ryan, 2009). Hippocampal neurons were probed with identical numbers of action potentials (APs) applied at either moderate (5 Hz) or high (40 Hz) frequency and SV exoendocytosis was monitored by tracing Syt1-pHluorin fluorescence. The relative peak fluorescence of Syt1-pHluorin at the end of the stimulation period was significantly lower when neurons were stimulated at 5 Hz compared to 40 Hz (Figures 1A and 1B). To probe whether this difference reflects elevated exocytotic fusion or reduced during-stimulus endocytosis at 40 Hz, we applied the H + -ATPase inhibitor folimycin. This drug prevents reacidification of internalized membranes, thereby allowing us to quantitatively determine exocytotic membrane fusion as well as endocytic membrane retrieval during stimulation (Kim and Ryan, 2009). Syt1-pHluorin peak amplitudes (Figure S1A available online) and release rates (Figure S1B) elicited by 200 APs applied at 5 Hz or 40 Hz were identical under endocytosis-blinded conditions in the presence of folimycin. By contrast, prominent during-stimulus endocytosis was observed at 5 Hz but barely detectable at 40 Hz (Figures 1C and 1D). To visualize endocytic intermediates under these conditions, we employed electron microscopy (Figures S1C and S1D). Synapses stimulated with 200 APs applied at 5 Hz accumulated endocytic plasma membrane clathrin-coated pits, intermediates of CME, but also free endosome-like vacuoles (Figures S1C and S1D; Figures 1Ei 1Eiv, 1G, and 1H). During high-frequency stimulation the number of endosome-like vacuoles increased, while plasma membrane clathrin-coated pits were more rarely detected (Figures 1Fi 1Fiv, 1G, and 1H). Less frequently, we also observed plasma 982 Neuron 82, , June 4, 2014 ª2014 Elsevier Inc.

3 Neuron SV Reformation from Endosomes via Clathrin/AP-2 membrane invaginations (Figures 1Fiii and 1Fiv) that sometimes were capped by clathrin-coated buds at their ends (Figure 1Fiv). Clathrin/AP-2 Contribute to Membrane Retrieval at Moderate Frequency but Are Dispensable for Membrane Retrieval under Conditions of High-Frequency Stimulation The distinct morphological intermediates and the different fraction of endocytosis during low- versus high-frequency stimulation could conceivably either reflect a single mode of SV endocytosis running through distinct kinetic bottlenecks or different endocytic modes operating in parallel. It has been proposed that SV membrane retrieval at hippocampal synapses occurs exclusively via CME (Granseth et al., 2006). We put this proposal to the test by depleting neurons of endogenous clathrin using lentivirally delivered shrna. Clathrin knockdown led to a strong reduction of endogenous clathrin levels to about 25% 30% of that observed in controls (Figures S2A and S2B) and a concomitant inhibition of CME of transferrin (Figure S2C), in good agreement with earlier studies in neurons (Granseth et al., 2006). We then monitored exoendocytosis and reacidification of Syt1- phluorin in clathrin-depleted neurons stimulated with 200 APs at 5 Hz or 40 Hz. Clathrin-depleted neurons remained exocytosis competent, although the rate of exocytosis was slightly reduced by about 25% (Figure S2D; consistent with recent findings [Hua et al., 2013]). Surprisingly, clathrin depletion did not result in a block of activity-dependent endocytosis as claimed before (Granseth et al., 2006) (Figures 2A 2D). Quantitative measurement of Syt1-pHluorin peak fluorescence in the absence or presence of folimycin revealed a significant reduction of during-stimulus endocytosis rate (Figure S2E) and extent (Figures 2A and 2B) in clathrin knockdown neurons when compared to scrambled shrna-treated controls. Slowed endocytosis and reacidification at 5 Hz was also overt from the analysis of the poststimulus Syt1-pHluorin fluorescence decay reflected by a doubling of the time-for-decay to reach 1/e of the initial value (t)(figure 2C; t ctrl =24±3s,t CHC = 49 ± 7 s). These data indicate that clathrin and by extension CME contributes to compensatory membrane retrieval at hippocampal synapses stimulated at 5 Hz, though it does not appear to be essential. We reasoned that perhaps CME at synapses operates faster than in nonneuronal cells and therefore the small amounts of clathrin remaining after RNAi would suffice to sustain CME at a lower rate. If that would be the case, then the phenotype of clathrin depletion should become more prominent under conditions of high-frequency stimulation, where massive SV exocytosis occurs during a short time. Strikingly, Syt1-pHluorin exoendocytosis proceeded unaltered in clathrin-depleted neurons stimulated with 200 APs at 40 Hz with a near-identical t value to that measured in control neurons (Figure 2D; t ctrl = 37 ± 3 s, t CHC = 34 ± 2 s). Membrane retrieval also proceeded unperturbed in neurons stimulated with 40 APs at 20 Hz and with only a slight delay if ten APs at 20 Hz were applied (Figures S2F and S2G), contrary to an earlier report (Granseth et al., 2006). Hence, clathrin appears to be dispensable for membrane retrieval under high-frequency (20 40 Hz) stimulation. To challenge these surprising results by independent means, we generated mice in which AP-2(m) was targeted for conditional deletion by appropriately placed loxp sites (Figure S2H). AP-2 is a heterotetrameric complex comprised of a, b, m, and s subunits that links clathrin and other endocytic proteins to sites of CME (Dittman and Ryan, 2009). Knockout of AP-2(m) in mice causes early embryonic lethality (Mitsunari et al., 2005), and depletion of AP-2(m) in mammalian cells results in loss of the entire AP-2 complex and potent inhibition of CME (McMahon and Boucrot, 2011). AP-2(m) lox mice were mated with a tamoxifen-inducible Cre line (B6.Cg-Tg(CAG-cre/Esr1*)5Amc/J) to generate a strain, in which tamoxifen application results in targeted deletion of AP-2(m). Treatment of primary hippocampal neurons prepared from AP-2(m) lox mice with tamoxifen resulted in a substantial depletion of AP-2 by about 80% (Figures S2I and S2J) until barely detectable by immunostaining (Figures S2K and S2L). Further AP-2 depletion below this level caused massive neuronal cell death consistent with the fact that AP-2(m) is an essential gene in mammals (Mitsunari et al., 2005). SV membrane retrieval measured by Syt1-pHluorin in response to 40 Hz stimulation proceeded unperturbed (Figure 2F) and with about 2-fold slower kinetics at 5 Hz in AP-2(m) lox knockout (KO) neurons (Figure 2E), phenocopying clathrin loss of function. Thus, clathrin and AP-2 surprisingly are dispensable for membrane retrieval, at least under conditions of high-frequency stimulation. So, how are SV membranes retrieved during high-frequency stimulation if clathrin and AP-2 are dispensable? Recent data using single optogenetic stimuli in the presence of high calcium concentrations paired with rapid flash-and-freeze electron microscopy have suggested that SV membranes can be recaptured via CIE, a process inhibited by dynasore (Watanabe et al., 2013), a compound that inhibits dynamin among other targets. To probe a possible function of dynamin in clathrin/ap-2-independent membrane retrieval at high frequency, we downregulated expression of the functionally redundant neuron-specific dynamins 1 and 3 (Ferguson et al., 2007) or of the ubiquitously expressed isoform dynamin 2 using specific sirnas (Figures S2M and S2N). Downregulation of dynamin 1/3 profoundly impaired endocytic retrieval of Syt1-pHluorin in neurons stimulated with 200 APs at 40 Hz (Figure 2H; t ctrl = 52 ± 4 s, t Dyn1,3 = 124 ± 13 s; in agreement with Ferguson et al., 2007) but only led to a minor slowing of Syt1-pHluorin endocytosis in neurons stimulated with 200 APs at 5 Hz (Figure 2G; t ctrl = 32 ± 3 s, t Dyn1,3 = 42 ± 6 s). A nearly identical phenotype was seen in neurons overexpressing a dominant-negative mutant version (endophilin1dh0) of the dynamin-binding NBAR protein endophilin (Figures S3A and S3B; in agreement with (Jockusch et al., 2005), while expression of full-length endophilin 1A had no effect (Figure S3C). SiRNA-mediated downregulation of dynamin 2 expression led to a phenotype converse to that of dynamin 1/3 depletion: endocytic retrieval of Syt1-pHluorin was slowed only to a minor degree when 200 APs were applied at 40 Hz, whereas retrieval was significantly slowed if neurons were stimulated at moderate frequency (Figures S3D and S3E), similar to depletion of clathrin or AP-2 (Figure 2), suggesting that clathrin/ap-2 and dynamin 2 operate together as they do in nonneuronal cells (McMahon and Boucrot, 2011). Taken together, these data demonstrate that SV membrane retrieval at hippocampal synapses challenged with high-frequency stimulation occurs largely by dynamin 1/3- and endophilin-mediated Neuron 82, , June 4, 2014 ª2014 Elsevier Inc. 983

4 Neuron SV Reformation from Endosomes via Clathrin/AP-2 Figure 2. Depletion of Clathrin or AP-2 Slows Endocytosis and Reacidification at 5 Hz but Not at 40 Hz (A) Reduced during-stimulus endocytosis in clathrin-depleted neurons. Average fluorescence traces recorded from hippocampal neurons transduced with lentivirus coexpressing Syt1-pHluorin together with anti-clathrin heavy chain shrna (shchc, red) or an inactive scrambled version (scr, black). Neurons were stimulated with 200 APs at 5 Hz in the absence or presence of folimycin. Values were normalized to the maximum fluorescence change at the end of 200 APs in the presence of folimycin. n > 400 boutons/condition. Data for scr-shrna-treated controls are replotted from Figure 1C. (B) Reduced fraction of during-stimulus endocytosis in clathrin-depleted neurons (61.4% ± 12.0% of scr-controls, p = 0.015). (C and D) Depletion of clathrin delays membrane retrieval induced by lowfrequency stimulation (C) (t shchc = 49.0 ± 7.0 s versus t scr = 24.0 ± 2.6 s, p = 0.001). Endocytosis induced by high-frequency stimulation is unaffected (D) (t shchc = 33.7 ± 1.8 s versus t scr = 37.4 ± 3.0 s, p = 0.281). n > 500 boutons/condition. (E and F) Deletion of AP-2(m) (mimics clathrin loss of function). Membrane retrieval induced by low-frequency stimulation is significantly slowed in the absence of AP-2 (E) (t AP-2loxP/Cre+EtOH = 24.6 ± 2.5 s versus t AP-2loxP/Cre+Tmx = 63.3 ± 12.7 s, p = 0.007), while endocytosis during high-frequency stimulation is largely unaffected (F) (t AP-2loxP/Cre+EtOH = 41.7 ± 2.9 s versus t AP-2loxP/Cre+Tmx = 47.0 ± 3.3 s, p = 0.232). n > 500 boutons/condition. (G and H) Depletion of dynamins 1 and 3 (see Figure S2M) significantly delays membrane retrieval induced by high-frequency stimulation with 200 APs at 40 Hz (H) (t sidyn1,3 = ± 13 s versus t scr = 52.1 ± 4.3 s, p < ), while endocytosis induced by low-frequency stimulation with 200 APs at 5 Hz (G) is affected only to a minor degree (t sidyn1,3 = 42.7 ± 5.9 s versus t scr = 32.4 ± 3.3 s, p = 0.120). Note the increased t in calcium phosphatetransfected scrambled sirna-treated compared to nontransfected controls. n > 500 boutons/condition. (I and J) A theoretical model based on CME and CIE operating in parallel agrees well with experimental data. Data points are replotted from (A) for 5 Hz (I) or from Figure 1C for 40 Hz (J) with identical color coding. Solid lines represent model predictions with one set of parameter values determined by fitting all six data sets simultaneously. Note that the slight difference between model prediction and experiment in folimycin-treated cells at 40 Hz is likely due to bleaching that cannot be accounted for by the model. Under conditions of clathrin knockdown, it was assumed that the rate constant for CME was reduced to 30% (k CME_knockdown = k CME, consistent with the level of clathrin knockdown and inhibition of CME of transferrin; Figures S2A S2C), while all other parameters were kept constant (see Table S1 for all parameter values). (K) Illustrated model: after exocytosis, the lumenal face of the SV switches from acidic ph (marked in dark) to the more alkaline ph of the extracellular environment, causing Syt1-pHluorin dequenching. Subsequent membrane retrieval and reacidification via parallel CME and CIE lead to fluorescence requenching. In the model, CME and CIE, and their subsequent acidification occur with different rate constants (k CME, and k CIE, respectively), are driven by fused vesicles (FVs), yet differ with respect to the stoichiometry of membrane internalization: CME internalizes single and CIE multiple (m) vesicle equivalents at once (see Figures S3G S3I for details). All data represent mean ± SEM. See also Figures S2 and S3. CIE, while both CIE and CME contribute to membrane retrieval at moderate stimulation frequencies. A Mathematical Kinetic Model of SV Membrane Retrieval via CIE and CME The experimental findings described thus far indicate that hippocampal synapses have evolved mechanisms to ensure efficient membrane retrieval over a range of stimulation frequencies involving CIE and CME. What determines the choice between CIE and CME and how does the frequency of stimulation affect this choice? To provide rational answers to these questions, we turned to mathematical kinetic modeling of exoendocytosis. We considered a simple scenario in which CIE and CME operate in parallel (consistent with Virmani et al., 2003), are governed by 984 Neuron 82, , June 4, 2014 ª2014 Elsevier Inc.

5 Neuron SV Reformation from Endosomes via Clathrin/AP-2 Figure 3. Reduced SV Density and Accumulation of Endosome-like Vacuoles at Clathrin- or AP-2-Deficient Hippocampal Synapses (Ai, Aii, Bi, and Bii) Representative electron micrographs of SVs and endosome-like vacuoles (ELVs, arrows) in synaptic terminals from cultured hippocampal neurons transduced with lentivirus expressing anti-clathrin heavy chain shrna (shchc; Aii and Bii) or an inactive scrambled version (scr; Ai and Bi) challenged with 200 APs at 5 Hz (Ai and Aii) or 40 Hz (Bi and Bii) and fixed immediately thereafter. Scale bar, 250 nm. (Ci, Cii, Di, and Dii) Representative electron micrographs of SVs and ELVs (arrows) in synaptic terminals from hippocampal neurons of AP-2(m) lox /Cre mice treated either with tamoxifen (Tmx; Cii and Dii) or with ethanol (EtOH; Ci and Di) and challenged with 200 APs at 5 Hz (Ci and Cii) or 40 Hz (Di and Dii) and fixed immediately thereafter. Scale bar, 250 nm. (E J) Average numbers of CCPs (E and H), SVs (F and I), and ELVs (G and J) in control synapses (black) and in clathrin (E G)- or AP-2 (H J)-deficient synapses (red). Depletion of clathrin leads to significant frequency-independent reduction in SV density (scr 5 Hz: ± 4.92, shchc 5 Hz: ± 6.70, p < , scr 40 Hz: ± 4.93, shchc 40 Hz: ± 9.55, p = 0.042) and the accumulation of ELVs (scr 5 Hz: 3.00 ± 0.28 shchc, 5 Hz: 8.55 ± 0.68, p < , scr 40 Hz: 5.42 ± 0.38, shchc 40 Hz: 9.63 ± 1.26, p < ). Note the dramatic reduction of CCPs in clathrindepleted terminals stimulated with 200 APs at 5 Hz (scr 5Hz: 1.52 ± 0.27, shchc 5 Hz: 0.23 ± 0.09, p = 0.001). n > 100 synapses/condition. Similarly loss of AP-2 leads to significant frequency-independent reduction in SV density (EtOH 5 Hz: 133 ± 5.12, Tmx 5 Hz: 95 ± 6.76, p = 0.003, EtOH 40 Hz: ± 5.81, Tmx 40 Hz: ± 9.37, p < ) and the accumulation of ELVs (EtOH 5 Hz: 2.98 ± 0.31, Tmx 5 Hz: ± 1.18, p < , EtOH 40 Hz: 6.31 ± 0.48, Tmx 40 Hz: 9.44 ± 1.45, p = 0.011). Note the dramatic reduction of CCPs in AP-2-depleted terminals stimulated with 200 APs at 5 Hz (EtOH 5 Hz: 1.61 ± 0.29 Tmx 5 Hz: 0.07 ± 0.06, p < ). n > 100 synapses/condition. All data represent mean ± SEM. See also Figures S2 and S3 and Table S2. individual rate constants for membrane retrieval and acidification (k CIE, k CME ), but differ with respect to the stoichiometry of membrane retrieval: the size of clathrin-coated pits (Heuser and Reese, 1973) and vesicles (Maycox et al., 1992) at nerve terminals indicates that CME retrieves single vesicles. By contrast, CIE may occur via large invaginations that internalize multiple vesicle equivalents (Schikorski, 2014; Watanabe et al., 2013) and, thus, is cooperative (Figure 2K; Figures S3G S3I). Fitting the mathematical model to the experimental data obtained from phluorin imaging allowed us to accurately reproduce stimulus-dependent changes in fluorescence decay as well as the different fractions of during-stimulus endocytosis elicited by 5 Hz or 40 Hz stimulation trains (Figures 2I and 2J). Reducing the rate of CME by clathrin knockdown to 30% of that in controls (k CME, shchc = 0.3x k CME, Ctrl ; Figure 2K) predominantly reduced endocytosis after 5 Hz stimulation (Figure 2I) but hardly affected membrane retrieval in response to 40 Hz stimulation (Figure 2J), in agreement with experimental results. The rapid accumulation of fused vesicle membranes (FV; Figure 2K) induced by highfrequency stimulation, hence, favors CIE as a direct consequence of its cooperative nature. The parameters determined by the best fit of our model to the experimental data suggest that the rate constant for CIE is approximately twice as large as the one for CME. Clathrin/AP-2 Mediate SV Reformation from Endosomelike Vacuoles While these experimental and theoretical results indicate that activity-dependent membrane retrieval at synapses can occur via CIE, they do not rule out an important function for clathrin/ AP-2 in subsequent steps of SV cycling such as SV reformation (Heerssen et al., 2008; Kasprowicz et al., 2008), a reaction not probed by phluorin measurements. To analyze such function, we stimulated clathrin-depleted hippocampal neurons with 200 APs at either 5 Hz or 40 Hz followed by rapid fixation and analysis by electron microscopy. Clathrin-coated pits were detected in control synapses stimulated at 5 Hz but were nearly absent in clathrin-depleted synapses (Figure 3E), in agreement with the Neuron 82, , June 4, 2014 ª2014 Elsevier Inc. 985

6 Neuron SV Reformation from Endosomes via Clathrin/AP-2 Figure 4. Reduced SV Density and Accumulation of Endosome-like Vacuoles at Cortical Synapses of AP-2(m) Conditional Knockout Mice (A) Loss of AP-2(a) in the somatosensory cortex of AP-2(m) knockout mice. Immunoblot of lysates from the somatosensory cortex of AP-2(m) KO mice and their wild-type (WT) littermates probed with antibodies against AP-2(a) clathrin heavy chain (CHC) and actin. The remaining AP-2(a) may result from astrocytic pools that are not targeted by Cre-mediated deletion. (B and C) Representative electron micrographs of synaptic terminals from the somatosensory cortex of wild-type and AP-2(m) KO mice. Note the depletion of SVs and the corresponding accumulation of endosome-like vacuoles (ELVs). Scale bar, 500 nm. (D) Severely reduced mean number of synaptic vesicles (SVs) in the somatosensory cortex of AP-2(m) knockout mice (113 ± 14.61) compared to WT littermates (193 ± 7.30, p = 0.006). (E) Average number of ELVs in synaptic terminals from the somatosensory cortex of wild-type (back) and AP-2(m) KO (red) mice (WT: 1.54 ± 0.47, KO: ± 1.23, p = ). (F and G) Representative electron tomograms and 3D reconstructions of presynaptic terminals from the somatosensory cortex of wild-type (F) and AP-2(m) KO mice (G). Tomograms of 200-nm-thick samples were reconstructed and overlaid onto a 2D projection of a single tomogram slice using Amira 3D segmentation editor. AP-2(m) KO synapses display a drastic reduction in SV number and an accumulation of ELVs, most of which were disconnected from the plasma membrane (PM). Scale bar, 200 nm. (H and I) Representative examples of ELVs connected to the PM (H) and of free ELVs that had undergone complete fission (I). PSD, postsynaptic density. All data represent mean ± SEM. See also Figure S4. essential role for clathrin in coated pit formation and with the observed clathrin-sensitive endocytosis component at this frequency (Figures 2C and 2E). Moreover, clathrin-depleted neurons stimulated at either frequency displayed a significant reduction in the number of SVs (Figure 3F) and a concomitant accumulation of internal endosome-like vacuoles of different sizes (Figures 3Ai, 3Aii, 3Bi, 3Bii, and 3G), similar to the vacuoles seen at D. melanogaster larval neuromuscular junctions following acute clathrin inactivation (Heerssen et al., 2008; Kasprowicz et al., 2008). A phenotype closely resembling that elicited by clathrin depletion was observed in tamoxifen-treated AP-2(m) lox KO neurons: stimulation at 5 Hz resulted in the accumulation of clathrin-coated pits in vehicle-treated control synapses, which were virtually absent from AP-2(m) lox KO synapses (Figure 3H). Furthermore, AP-2(m) lox KO synapses stimulated at either 5 Hz or 40 Hz suffered from a partial depletion of SVs (Figure 3I) and a concomitant accumulation of endosome-like vacuoles (Figures 3C, 3D, and 3J), most of which were disconnected from the plasma membrane, consistent with their proper acidification measured by phluorin imaging (compare Figure 2F) and with electron tomography analysis (see below, Figure 4). These ultrastructural data indicate that clathrin/ap-2 play an important role in the maintenance of SV pools and in the processing of endosome-like vacuoles that probably are derived from CIE, consistent with the appearance of clathrin-coated endosome-like vacuoles and a concomitant depletion of SVs in hippocampal neurons treated with the clathrin inhibitor Pitstop 2(von Kleist et al., 2011) (Figure S4). To challenge this hypothesis, we finally sought to analyze the functional effects of AP-2(m) loss of function under physiological activity levels in vivo. To this aim, we crossed AP-2(m) lox mice with a line expressing Cre recombinase under the control of the neuron-specific T1a-tubulin promoter (Cre-tub1a) (Coppola et al., 2004). Homozygous AP- 2(m) lox/lox /Cre+ mice displayed normal AP-2(m) levels at birth but lost neuronal AP-2(m) expression in the entorhinal and somatosensory cortex within the first 3 weeks of their postnatal life (Figure 4A) and died between postnatal days 21 and 23 (data not shown). Electron microscopic analysis of nerve terminals from the somatosensory cortex of these conditional AP-2(m) lox KO mice revealed a profound reduction in SV number (Figures 4B 4D) and a corresponding accumulation of endosome-like vacuoles compared to wild-type littermates (Figures 4B, 4C, and 4E), while no apparent changes in the morphologies of the postsynaptic compartment or of the active zone were detected. Electron tomography analysis revealed that most of these endosome-like vacuoles had undergone complete fission from the plasma membrane, indicating that AP-2 is largely dispensable for endocytic plasma membrane retrieval in vivo 986 Neuron 82, , June 4, 2014 ª2014 Elsevier Inc.

7 Neuron SV Reformation from Endosomes via Clathrin/AP-2 (Figures 4F 4I). The incomplete depletion of SVs at AP-2(m) lox KO synapses probably reflects the fact that many synapses in these young animals do not yet have an extensive stimulation history and have suffered from complete loss of AP-2 only for a few days, during which the remaining amounts of AP-2 may suffice to partially replenish the SV pool. It is also possible that some SV reformation can occur in the absence of AP-2 (i.e., via AP-3). We conclude that AP-2 serves a key function in SV reformation from endosome-like vacuoles in mammalian CNS neurons in vivo. DISCUSSION We show here that clathrin/ap-2 and, thus, CME are not essential for retrieval of fused SV membranes in hippocampal neurons, in particular under conditions of high-frequency stimulation. Instead, mammalian neurons appear to have evolved a dynamin 1/3- and endophilin-dependent CIE mechanism that allows them to uncouple membrane retrieval from vesicle reformation once the rate of exocytotic membrane addition exceeds the capacity and slow speed of CME from the plasma membrane, consistent with our theoretical modeling data. This stimulation-dependent modulation of endocytic mechanisms at synapses may enable mammalian neurons to cope with stimulation-induced membrane expansion over a wide range of frequencies. We further show that SV reformation from endosome-like vacuoles (first observed by Heuser and Reese to contain clathrin coats more than 40 years ago [Heuser and Reese, 1973]) largely depends on clathrin/ap-2, in agreement with a plethora of studies that had suggested a key role for the clathrin pathway in SV cycling (Dittman and Ryan, 2009; Heerssen et al., 2008; Heuser and Reese, 1973; Jockusch et al., 2005; Kasprowicz et al., 2008; McMahon and Boucrot, 2011; Saheki and De Camilli, 2012) and with the fact that clathrin-coated vesicles isolated from nerve terminals are highly enriched in SV proteins (Maycox et al., 1992). These results further suggest that endosome-like vacuoles must preserve their plasma membrane identity to allow for clathrin/ap-2 coats to assemble (McMahon and Boucrot, 2011) in order to replenish the SV pool (in agreement with Hoopmann et al., 2010). We therefore propose a model according to which the primary function of clathrin/ap-2-mediated CME is to regenerate SVs of proper size and composition irrespective of whether they bud from the plasma membrane proper (i.e., during sustained stimulation at low-to-medium frequencies) or from plasma membranederived endosome-like vacuoles (Hoopmann et al., 2010) generated by dynamin-dependent CIE. This model is consistent with prior observations that downregulation of AP-2(m) (Kim and Ryan, 2009) or knockout of the AP-2-associated endocytic adaptor stonin 2 (Kononenko et al., 2013) have surprisingly mild effects on the kinetics of membrane retrieval but cause a partial redistribution of synaptotagmin 1 to the neuronal surface, probably as a result of defective sorting over many SV cycles. Such surface-stranded SV proteins have been proposed to constitute a readily retrievable SV pool (Hua et al., 2011) that might conceivably be internalized by either CIE or CME. The results reported here are difficult to reconcile with earlier findings by Granseth et al. (2006), who, using weak stimuli paired with bleach correction, found that SV membrane retrieval was completely inhibited in hippocampal neurons depleted of clathrin by sirna. While the precise reason for the observed discrepancies is unknown, we note that neither clathrin knockdown nor conditional knockout of AP-2 in our hands led to a block in membrane retrieval under any stimulation condition. We can only speculate at this time about the precise mechanisms mediating CIE. Morphologically, endosome-like vacuoles that accumulate in the absence of clathrin/ap-2 resemble bulk endosomes observed in a variety of systems (Cheung and Cousin, 2013). Similar structures have also been reported by Watanabe et al. (2013), following single optogenetic stimuli, though the time course of membrane retrieval is clearly distinct from that reported here. Our initial molecular analysis suggests that CIE induced by high-frequency stimulation depends on dynamin 1/3 and endophilin 1 as well as on actin polymerization (Figure S3F). Clearly, endophilin 1 fulfils an additional role at later stages of SV recycling in recruitment of synaptojanin that triggers uncoating during clathrin/ap-2-mediated SV reformation (Milosevic et al., 2011). How precisely stimulation frequency modulates the choice between CME and CIE is unclear but may relate to the rate or amount of exocytotic membrane addition and/or the regulation of presynaptic calcium levels (Wu and Wu, 2014). For example, fast membrane fission during CIE may be triggered by calcineurin-mediated dephosphorylation and activation of dynamin 1 (Armbruster et al., 2013), resulting in the insertion of its PH domain into the membrane. We hypothesize that other types of mammalian neurons as well as other organisms capitalize on similar mechanisms of membrane retrieval at synapses or possibly even other cell types. Our work should thus pave the way to dissect these mechanistically and functionally in detail. EXPERIMENTAL PROCEDURES Electron Microscopy Cultured hippocampal neurons were grown and stimulated (at 27 C) as described for phluorin imaging. Cultures were fixed within 10 s after the end of stimulation protocol by immersion into PBS-buffered 2% glutaraldehyde. Details on routine embedding and analysis can be found in Supplemental Experimental Procedures. Statistical Analysis The statistical significance for all data was evaluated with a two-tailed paired Student s t test. Significant differences were accepted at p < See Supplemental Experimental Procedures for additional experimental procedures. SUPPLEMENTAL INFORMATION Supplemental Information includes Supplemental Experimental Procedures, four figures, and four tables and can be found with this article online at ACKNOWLEDGMENTS This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB 958/A01 (to V.H.); Exc-257/ NeuroCure (to V.H. and C.R.); GK1123 (to V.H.), and the Schram Foundation. We thank Sabine Hahn for expert technical support. Neuron 82, , June 4, 2014 ª2014 Elsevier Inc. 987

8 Neuron SV Reformation from Endosomes via Clathrin/AP-2 Accepted: April 22, 2014 Published: June 4, 2014 REFERENCES Armbruster, M., Messa, M., Ferguson, S.M., De Camilli, P., and Ryan, T.A. (2013). Dynamin phosphorylation controls optimization of endocytosis for brief action potential bursts. elife 2, e Ceccarelli, B., Hurlbut, W.P., and Mauro, A. (1972). Depletion of vesicles from frog neuromuscular junctions by prolonged tetanic stimulation. J. Cell Biol. 54, Cheung, G., and Cousin, M.A. (2013). Synaptic vesicle generation from activity-dependent bulk endosomes requires calcium and calcineurin. J. Neurosci. 33, Coppola, V., Barrick, C.A., Southon, E.A., Celeste, A., Wang, K., Chen, B., Haddad, B., Yin, J., Nussenzweig, A., Subramaniam, A., and Tessarollo, L. (2004). Ablation of TrkA function in the immune system causes B cell abnormalities. Development 131, Dittman, J., and Ryan, T.A. (2009). Molecular circuitry of endocytosis at nerve terminals. Annu. Rev. Cell Dev. Biol. 25, Ferguson, S.M., Brasnjo, G., Hayashi, M., Wölfel, M., Collesi, C., Giovedi, S., Raimondi, A., Gong, L.W., Ariel, P., Paradise, S., et al. (2007). A selective activity-dependent requirement for dynamin 1 in synaptic vesicle endocytosis. Science 316, Granseth, B., Odermatt, B., Royle, S.J., and Lagnado, L. (2006). Clathrinmediated endocytosis is the dominant mechanism of vesicle retrieval at hippocampal synapses. Neuron 51, Heerssen, H., Fetter, R.D., and Davis, G.W. (2008). Clathrin dependence of synaptic-vesicle formation at the Drosophila neuromuscular junction. Curr. Biol. 18, Heuser, J.E., and Reese, T.S. (1973). Evidence for recycling of synaptic vesicle membrane during transmitter release at the frog neuromuscular junction. J. Cell Biol. 57, Hoopmann, P., Punge, A., Barysch, S.V., Westphal, V., Bückers, J., Opazo, F., Bethani, I., Lauterbach, M.A., Hell, S.W., and Rizzoli, S.O. (2010). Endosomal sorting of readily releasable synaptic vesicles. Proc. Natl. Acad. Sci. USA 107, Hua, Y., Sinha, R., Thiel, C.S., Schmidt, R., Hüve, J., Martens, H., Hell, S.W., Egner, A., and Klingauf, J. (2011). A readily retrievable pool of synaptic vesicles. Nat. Neurosci. 14, Hua, Y., Woehler, A., Kahms, M., Haucke, V., Neher, E., and Klingauf, J. (2013). Blocking endocytosis enhances short-term synaptic depression under conditions of normal availability of vesicles. Neuron 80, Jockusch, W.J., Praefcke, G.J., McMahon, H.T., and Lagnado, L. (2005). Clathrin-dependent and clathrin-independent retrieval of synaptic vesicles in retinal bipolar cells. Neuron 46, Kasprowicz, J., Kuenen, S., Miskiewicz, K., Habets, R.L., Smitz, L., and Verstreken, P. (2008). Inactivation of clathrin heavy chain inhibits synaptic recycling but allows bulk membrane uptake. J. Cell Biol. 182, Kim, S.H., and Ryan, T.A. (2009). Synaptic vesicle recycling at CNS snapses without AP-2. J. Neurosci. 29, Kononenko, N.L., Diril, M.K., Puchkov, D., Kintscher, M., Koo, S.J., Pfuhl, G., Winter, Y., Wienisch, M., Klingauf, J., Breustedt, J., et al. (2013). Compromised fidelity of endocytic synaptic vesicle protein sorting in the absence of stonin 2. Proc. Natl. Acad. Sci. USA 110, E526 E535. Maycox, P.R., Link, E., Reetz, A., Morris, S.A., and Jahn, R. (1992). Clathrincoated vesicles in nervous tissue are involved primarily in synaptic vesicle recycling. J. Cell Biol. 118, McMahon, H.T., and Boucrot, E. (2011). Molecular mechanism and physiological functions of clathrin-mediated endocytosis. Nat. Rev. Mol. Cell Biol. 12, Milosevic, I., Giovedi, S., Lou, X., Raimondi, A., Collesi, C., Shen, H., Paradise, S., O Toole, E., Ferguson, S., Cremona, O., and De Camilli, P. (2011). Recruitment of endophilin to clathrin-coated pit necks is required for efficient vesicle uncoating after fission. Neuron 72, Mitsunari, T., Nakatsu, F., Shioda, N., Love, P.E., Grinberg, A., Bonifacino, J.S., and Ohno, H. (2005). Clathrin adaptor AP-2 is essential for early embryonal development. Mol. Cell. Biol. 25, Saheki, Y., and De Camilli, P. (2012). Synaptic vesicle endocytosis. Cold Spring Harb. Perspect. Biol. 4, a Schikorski, T. (2014). Readily releasable vesicles recycle at the active zone of hippocampal synapses. Proc. Natl. Acad. Sci. USA 111, Virmani, T., Han, W., Liu, X., Südhof, T.C., and Kavalali, E.T. (2003). Synaptotagmin 7 splice variants differentially regulate synaptic vesicle recycling. EMBO J. 22, von Kleist, L., Stahlschmidt, W., Bulut, H., Gromova, K., Puchkov, D., Robertson, M.J., MacGregor, K.A., Tomilin, N., Pechstein, A., Chau, N., et al. (2011). Role of the clathrin terminal domain in regulating coated pit dynamics revealed by small molecule inhibition. Cell 146, Watanabe, S., Rost, B.R., Camacho-Pérez, M., Davis, M.W., Söhl-Kielczynski, B., Rosenmund, C., and Jorgensen, E.M. (2013). Ultrafast endocytosis at mouse hippocampal synapses. Nature 504, Wu, X.S., and Wu, L.G. (2014). The yin and yang of calcium effects on synaptic vesicle endocytosis. J. Neurosci. 34, Neuron 82, , June 4, 2014 ª2014 Elsevier Inc.

9 Neuron, Volume 82 Supplemental Information Clathrin/AP-2 Mediate Synaptic Vesicle Reformation from Endosome-like Vacuoles but Are Not Essential for Membrane Retrieval at Central Synapses Natalia L. Kononenko, Dmytro Puchkov, Gala A. Classen, Alexander M. Walter, Arndt Pechstein, Linda Sawade, Natalie Kaempf, Thorsten Trimbuch, Dorothea Lorenz, Christian Rosenmund, Tanja Maritzen, and Volker Haucke

10 Inventory of Supplemental information for Clathrin/ AP-2 mediate synaptic vesicle reformation from endosome-like vacuoles but are not essential for membrane retrieval at central synapses Natalia L. Kononenko, Dmytro Puchkov, Gala A. Classen, Alexander M. Walter, Arndt Pechstein, Linda Sawade, Natalie Kaempf, Thorsten Trimbuch, Dorothea Lorenz, Christian Rosenmund, Tanja Maritzen, Volker Haucke Supplemental Information contains: 4 supplementary figures and legends: Figure S1 (related to fig. 1): Syt1-pHluorin release rate is insensitive to stimulation frequency Figure S2 (related to figs. 2 and 3): Depletion of clathrin or dynamin by RNA interference or conditional knockout of AP-2(µ) in hippocampal neurons Figure S3 (related to figs. 2 and 3): Differential requirements for dynamins, endophilin, and actin in membrane retrieval at hippocampal synapses Figure S4 (related to fig. 4): Acute pharmacological inhibition of clathrin leads to accumulation of clathrin-coated vesicles and clathrin-coated ELVs 4 supplementary tables, supplementary experimental procedures, and supplementary references. 1

11 SUPPLEMENTARY FIGURES Figure S1 (related to fig. 1). Syt1-pHluorin release rate is insensitive to stimulation frequency. A) Average Syt1pHluorin responses to 40 APs at 20 Hz followed by 200 APs at 5 Hz (black) and, finally, 200 APs at 40 Hz (gray) in the presence of folimycin (>400 boutons). Fluorescence transients at 5 Hz and 40 Hz were normalized to the initial stimulus of 40 APs at 20 Hz. The signal amplitude was insensitive to stimulus frequency. B) Relative release rates calculated from A). Differences in peak fluorescence depicted in A) are not due to differences in release rates (40 Hz=100±6%, 5 Hz=101±9%). C,D) Electron micrographs of synapses used for illustrations of endocytic intermediates in Figs.1 Ei, EII (C) and Fi, Fii (D). Scale bars, 500 nm. All data represent mean ± SEM. 2

12 Figure S2 (related to figs. 2 and 3). Depletion of clathrin or dynamin by RNA interference or conditional knockout of AP-2(µ) in hippocampal neurons. A) Immunoblots of lysates from primary cultures of hippocampal neurons transduced with lentivirus expressing anti-clathrin heavy chain shrna (shchc) or an inactive 3

13 scrambled version thereof (scr) probed with antibodies against clathrin heavy chain (CHC) or synapsin 1. B) Relative levels of clathrin heavy chain (CHC) in control neurons (scr) and neurons expressing shrna-targeting clathrin heavy chain (CHC). In neurons transfected with shchc clathrin levels were severely depleted (32.95±4.83%) compared with controls (100±6.80%, p<0.0001). Data represent four independent experiments with 58 images per condition in total. C) Effect of shrna-targeting CHC on CME assayed by transferrin Alexa488 uptake. In neurons transfected with shrna CME of transferrin Alexa488 was significantly inhibited (39.93±3.61%) compared with controls (100±25.41%, p=0.031). Data represent 17 images per condition in total. D) Relative rates of exocytosis in the presence of folimycin is significantly reduced in clathrindepleted neurons (75.5±7.2%) when compared to controls (100±4.4%, p=0.003; synaptic boutons >400). E) Relative rates of during-stimulus endocytosis in neurons expressing shrna-targeting CHC (shchc, red) was significantly slower (62.72±7.33%) when compared to controls (100±9.16%, p=0.006; synaptic boutons >400). F, G) Depletion of clathrin slightly delays membrane retrieval induced by stimulation with 10 APs at 20 Hz (F) (τ chc =44.9±4.7s versus τ scr =30±2.5s, p=0.01), while endocytosis induced by stimulation with 40 APs at 20 Hz (G) is unaffected (τ chc =26.7±1.5s versus τ scr =35.2±4.72s, p=0.148). H) Targeting strategy used to generate conditional AP-2(µ) KO mice. I) Immunoblots of lysates from primary cultures of AP-2floxP/Cre neurons treated either with ethanol (EtOH) or Tamoxifen (Tmx) probed with the indicated antibodies. J) Levels of AP-2 quantified from I). Application of tamoxifen leads to depletion of AP-2(µ) (22.75±3.65%). K) Relative levels of AP-2 in control neurons (EtOH) or neurons treated with tamoxifen (Tmx) as shown in L). In cells transfected with tamoxifen AP-2 was severely depleted (23.66±1.75%) compared with controls (100±6.62%, p<0.0001). Data represent 45 images per condition in total. L) Primary hippocampal neurons from AP-2floxP/Cre mice treated either with ethanol (EtOH) or tamoxifen (Tmx) were immunostained with MAP-2 (green) and AP-2(α) antibodies (red). Scale bar, 50µm. M) Relative levels of dynamin 1 and 3 in control neurons (scr; 100±8.68%, p<0.0001) and neurons transfected with Dyn1,3 sirna (11.17±3.25%). N) Relative levels of dynamin 2 in control neurons (scr; 100±10.6%, p=<0.0001) or neurons transfected with Dyn2 sirna (40.5±6.3%). All data represent mean ± SEM. 4

14 5

15 Figure S3 (related to figs. 2 and 3): Differential requirements for dynamins, endophilin, and actin in membrane retrieval at hippocampal synapses. A,B) Overexpression of endophilin 1A lacking its amphipathic H0 helix (ΔH0) delays membrane retrieval induced by high-frequency stimulation with 200 APs at 40 Hz (A) (τ EndophilinΔHO =86.1±12.8s versus τ mcherry =37.4±2.4s, p=0.003), while endocytosis induced by low-frequency stimulation with 200 APs at 5 Hz (B) is largely unaffected (τ EndophilinΔHO =28.7±3.8s versus τ mcherry =24.0±3.8s, p=0.433). Data are from >500 boutons for each condition. C) Overexpression of full-length endophilin A1 (Endophilin FL, red) does not affect membrane retrieval induced by high-frequency stimulation compared to mcherry expressing controls (τ mcherry =51.92±6.64s versus τ EndophilinFL =45.69±9.66s, p=0.594; synaptic boutons >300). D,E) Dynamin 2 regulates membrane retrieval induced by lowfrequency stimulation. Depletion of dynamin 2 significantly delays membrane retrieval induced by low-frequency stimulation with 200 APs at 5 Hz (E) (τ sidyn2 =64.4±10.7s versus τ scr =29.7±3.3s, p=0.010), while endocytosis induced by high-frequency stimulation with 200 APs at 40 Hz (D) is largely unaffected (τ sidyn2 =77.04±15s versus τ scr =52.2±4.8s, p=0.171). Data are from >500 boutons for each condition. G) Kinetic model used for simulations of exo-endocytosis. H) Simulated Ca 2+ -waves for 5 Hz (left) and 40 Hz (right) stimulation (note the build-up of Ca 2+ between stimuli at 40 Hz). I) Temporal development of fused vesicles (FV) as well as the magnitude of CME and CIE are shown at 5 Hz (left) and 40 Hz (right) in control (black) and clathrin-depleted (clathrin KD, red) synapses. All data represent mean ± SEM. 6

16 Figure S4 (related to fig. 4). Acute pharmacological inhibition of clathrin leads to accumulation of clathrin-coated vesicles and clathrin-coated ELVs. A) Representative electron micrographs of stimulated synaptic terminals from cultured hippocampal neurons in the presence of 30 µm of the clathrin inhibitor Pitstop 2 and fixed immediately after. Arrowheads indicate clathrin-coated structures. Scale bar, 100 nm. B-D) Average number of SVs (B, Ctrl: 51.78±5.68, Pitstop2: 33.14±5.56, p=0.026), clathrin-coated vesicles (CCVs) (C, Ctrl: 1.35±0.35, Pitstop2: 2.30±0.30, p=0.040) and clathrin coated endosomelike vesicles (ELV, Ctrl: 2.30±0.54, Pitstop2: 4.06±0.54, p=0.028) (D) in control (DMSOtreated, black) and Pitstop2-treated (red) presynaptic terminals from 23 control and 30 Pitstop 2-treated synapses. 7

17 SUPPLEMENTARY TABLES Supplementary table 1. Model parameters. The following parameters were either taken from the literature or determined by the simplex: model parameters WT unit comment V tot 120 vesicles Hua et al. (2010) exocytosis k p = k!!"# [Ca!! ] Ca!! + K! 1/s Voets et al. (2000) k 1Max /s best fit K M 3.26 µm best fit k -p /s best fit k /s Wolfel et al (2007) k /s Wolfel et al (2007) b 0.5 Wolfel et al (2007) k /s Lou et al (2005) Endocytosis/ reacidification k CME /s best fit k CIE /s best fit m 1.53 best fit 8

18 Supplementary table 2. Full quantitative analysis of ultrastructural components of clathrinand AP-2(µ)-depleted cultured hippocampal neurons and their corresponding controls. SV, synaptic vesicles; ELVs, endosome-like vacuoles; CCPs, clathrin-coated pits; CCVs, clathrincoated vesicles; eccps, endosomal clathrin-coated pits; uncoated, uncoated plasma membrane invaginations; MVBs, multivesicular bodies. SV ELVs CCPs CCVs eccps uncoated MVBs scrambled shrna, 5 Hz 146±5 3.01± ± ± ± ± ±0.05 scrambled shrna, 40 Hz 138±5 5.42± ± ± ± ± ±0.05 CHC shrna, 5 Hz 92±7 8.56± ± ± ± ± ±0.07 CHC shrna, 40 Hz 107± ± ± ± ± ± ±0.08 AP-2loxP/Cre +EtOH, 5 Hz AP-2loxP/Cre +EtOH, 40 Hz AP-2loxP/Cre +Tmx, 5 Hz AP-2loxP/Cre +Tmx, 40 Hz 134±5 2.99± ± ± ± ± ± ±6 6.31± ± ± ± ± ± ± ± ± ± ± ± ± ±9 9.45± ± ± ± ± ±0.18 9