The phospho-regulated amphiphysin/endophilin interaction is required for synaptic vesicle endocytosis

The multidomain adaptor protein amphiphysin-1 (Amph1) is an important coordinator of clathrin-mediated endocytosis in non-neuronal cells and synaptic vesicle (SV) endocytosis at central nerve terminals. Amph1 contains a lipid-binding N-BAR (Bin/Amphiphysin/Rvs) domain, central proline-rich (PRD) and clathrin/AP2 (CLAP) domains, and a C-terminal SH3 domain. All domains interact with either lipids or SV endocytosis proteins, with all of these interactions required for SV endocytosis, apart from the Amph1 PRD. In this study, we determined this role and confirmed requirements for established Amph1 interactions in SV endocytosis at typical small central synapses. Domain-specific interactions of Amph1 were validated using in vitro GST pull-down assays, with the role of these interactions in SV endocytosis determined in molecular replacement experiments in primary neuronal culture. Using this approach, we confirmed important roles for CLAP and SH3 domain interactions in the control of SV endocytosis. Furthermore, we identified an interaction site for the endocytosis protein endophilin A1 in the Amph1 PRD and revealed a key role for this interaction in SV endocytosis. Finally, we discovered that the phosphorylation status of Amph1-S293 within the PRD dictates the formation of the Amph1-endophilin A1 complex and is essential for efficient SV regeneration. This work therefore identifies an activity-dependent dephosphorylation-dependent interaction that is key for efficient SV endocytosis.


Introduction 38
Synaptic transmission results from the activity-dependent fusion of neurotransmitter-39 containing synaptic vesicles (SVs) at the presynapse, resulting in the activation of 40 cognate postsynaptic receptors. After neurotransmitter release, presynaptic 41 endocytosis is essential to reform SVs, allowing their refilling and recycling for 42 subsequent rounds of neurotransmission. Early studies identified amphiphysin-1 43 (Amph1), a protein enriched at the presynaptic cytoplasm (Bauerfeind et al. 1997), as 44 a key mediator of SV endocytosis (Shupliakov et al. 1997). This was confirmed when 45 nerve terminals isolated from Amph1 knockout mice displayed reduced formation of 46 both SVs and endosome-like structures, and fewer competent SVs for repeated 47 rounds of exocytosis (Di Paolo et al. 2002). More recently, immunodepletion of Amph1 48 in vivo resulted in various presynaptic deficits after prolonged stimulation, such as SV 49 shortage, and a reduced number of clathrin-coated vesicles and endosome-like 50 structures (Werner et al. 2016). 51 Amph1 is a modular protein, with an N-terminal Bin/Amphiphysin/Rvs (N-BAR) domain 52 characterised by the presence of long stretches of amphiphilic α-helices, and can 53 either homodimerize or form heterodimers with another neuronal isoform, Amph2 54 (Ramjaun et al. 1999; Wigge et al. 1997). The dimeric amphipathic helices of the N-55 BAR domain recruit Amph1 to highly-curved membranes and exhibit fission-inducing 56 activity (Peter et  Amph1 also contains a PRD in its central region, which binds with high affinity to the 68 SH3 domain of endophilin A1 (Micheva et al. 1997b). Similar to Amph1, endophilin A1 69 is essential for optimal SV endocytosis, since either deletion of its gene or its 70 immunodepletion results in inhibition of the process ( targeting control against the sequence 5'-TCGCGATTAGTTCATTAGG-3' 137 (Scrambled) was generated accordingly. The synaptotagmin-1-pHluorin sequence 138 was replaced in the shAmph1-expressing vector by synaptophysin-pHluorin (sypHy) 139 using the primers 5'-GGATCCATGGACGTGGTGAATCAGCTGGTG-3' (sense) and 140 5'-TGGATATCTTTACATCTGATTGGAGAAGGAGGTG-3' (antisense) and enzymes 141 BamHI and EcoRV respectively. The sypHy sequence was subsequently mutated to 142 remove a BglII site. All plasmids were verified by Sanger sequencing. 143

Animals
The wild-type (WT) C57BL/6J mouse colony was utilized for the preparation of primary 145 hippocampal neuronal cultures. Adult female mice were killed by cervical dislocation 146 followed by decapitation, whereas embryos of both sexes were killed by decapitation 147 followed by destruction of the brain at E16-E18. For synaptosomal preparations, adult 148 Sprague Dawley male rats (> 2 months old) were killed by exposure to increasing CO2 149 concentration followed by cervical dislocation. All experimental procedures were 150 performed according to the UK Animal (Scientific Procedures) Act 1986 on the 151 protection of animals used for scientific purposes and were approved by the Animal 152 Welfare and Ethical Review Body at the University of Edinburgh (Home Office project 153 license to M. Cousin -7008878 and PP5745138). All animals were maintained on a 154 12-hour light/dark cycle under constant temperature, with food and water provided ad 155 libitum.  respectively. Images were acquired using a Zeiss AxioCam 506 camera and Zeiss 198 ZEN 2 software. Data analysis was performed using Fiji is just ImageJ. Amph1-199 silenced neurons were visualized at 475 nm excitation wavelength using the anti-GFP 200 antibody as described above. To measure Amph1 expression, regions of interest were 201 drawn around transfected cell bodies (as indicated by GFP staining) and average 202 Amph1 intensity was calculated and normalised to that of untransfected cell bodies. 203 Background was subtracted in all cases. 204

205
Adult rat brains were homogenized in ice-cold 0.32 M sucrose, 5 mM EDTA (pH 7.4) 206 after removing the cerebellum. The homogenate was centrifuged twice at 950 x g for 207 10 min at 4 o C and the supernatants were combined. After spinning at 20,400 x g for 208 30 min at 4 o C, the pellet (crude synaptosomal fraction) was collected and lysed in the 209 presence of protease inhibitors (Sigma-Aldrich, cat no. P8849). 210

Expression of glutathione S-transferase (GST)-fused Amph1 constructs in Escherichia 212
coli BL21 DE3 cells was induced with 1 mM isopropyl β-D-1-thiogalactopyranoside 213 (Calbiochem, cat no. 420322) at OD600nm 0.6-0.8. After centrifugation at 5000 x g for 214 15 min at 4 o C, the pellets were resuspended in ice-cold buffer containing 10 mM Tris, 215 150 mM NaCl, 1 mM EDTA, pH 8, protease inhibitors, and 1 mM phenylmethylsulfonyl 216 fluoride. Lysozyme (0.0675 μg/μl, Sigma-Aldrich, cat no. L6876), 4 mM dithiothreitol 217 (Sigma-Aldrich, cat no. D0632), and 10 % (v/v) Triton X-100 were also included. The 218 cells were sonicated at 10 kHz and the clear lysates were spun at 17,420 x g for 10 219 min at 4 o C. were examined together as the binding site for CHC ( Figure 1A). WT and mutant 272 versions of GST-Amph1 baits were then bound to glutathione-sepharose and 273 incubated with rat brain synaptosomal lysates. The interactions of Amph1 with CHC 274 and α-AP2 were then assessed by immunoblotting, in addition to Dyn1 and endophilin 275 A1 ( Figure 1B). We confirmed that the Amph1-HSR/SR mutation was sufficient to 276 ablate interactions with CHC ( Figure 1C). Similarly, Amph1-SSR lost its ability to bind 277 to α-AP2 subunits ( Figure 1D). In addition, the Amph1-HSR/SR had a strong negative 278 impact on α-AP2 binding, whereas Amph1-SSR abolished CHC binding. Importantly, 279 the Amph1-SSR and Amph1-HSR/SR mutants bound both Dyn1 and endophilin A1 to 280 a similar extent as Amph1-WT ( Figure 1E,1F would be unable to interact with Dyn1. This proved to be the case, with Amph1 devoid 287 of its SH3 domain incapable of binding to Dyn1 ( Figure 1E). CHC binding to both WT 288 and CLAP domain mutants was unaffected by the absence of the SH3 domain ( Figure  289 1C), whereas deletion of the SH3 domain severely compromised α-AP2 binding to WT 290 Amph1 ( Figure 1D). Finally, removal of the SH3 domain resulted in a significant 291 enhancement of endophilin A1 binding to WT Amph1 and all other CLAP mutants 292 ( Figure 1F). Therefore, intramolecular bonds between the Amph1 PRD and SH3 293 domain may reciprocally control interactions with both α-AP2 and endophilin A1. 294

295
The Amph1 mutants characterised above placed us in an ideal position to determine 296 the role of these interactions in SV endocytosis. However, we first confirmed that loss 297 of Amph1 function resulted in defects in this process. To determine this, we generated 298 and validated an shRNA vector against mouse Amph1 (shAmph1) that co-expresses 299 the genetically-encoded reporter sypHy (synaptophysin-pHluorin). SypHy comprises 300 the abundant SV protein synaptophysin with an EGFP that has been engineered to be 301 exquisitely pH-sensitive, inserted into an intralumenal loop (Granseth et al. 2006). 302 Since the interior of SVs is acidic, sypHy fluorescence is quenched in resting neurons, 303 however on SV exocytosis the reporter is exposed to the extracellular environment, 304 resulting in its unquenching.  figure 1A,1B). When SV regeneration was examined using 310 sypHy, depletion of Amph1 resulted in slower SV endocytosis following stimulation at 311 10 Hz for 30 s, compared to scrambled controls (Figure 2A,2B). Therefore, loss of 312 Amph1 function results in defects in SV regeneration. 313 The defect in SV endocytosis observed on Amph1 depletion provides an excellent 314 molecular replacement system to determine which Amph1 interactions are required 315 for optimal SV regeneration. All exogenously-expressed Amph1 rescue constructs for 316 these functional experiments were full-length, to mimic physiological context. 317 Furthermore, mutants were tagged with mCer at their C-termini to allow their 318 visualisation during imaging. We first examined the CLAP missense mutants (SSR 319 and HSR/SR) for their ability to restore SV endocytosis in neurons depleted of 320 endogenous Amph1. All mutants were expressed in the shAmph1-silenced 321 background as confirmed by immunolabelling ( Supplementary Figure 2A,2B). Amph1 322 knockdown neurons expressing both sypHy and Amph1-mCer mutants (or an empty 323 mCer vector control) were subjected to stimulation at 10 Hz for 30 s to evoke SV 324 recycling ( Figure 2C). Neurons expressing WT Amph1 had fully restored SV 325 endocytosis kinetics when compared to the empty vector control ( Figure 2D). In 326 contrast, neither of the Amph1 CLAP mutants (HSR/SR or SSR) were able to restore 327 SV retrieval kinetics compared to WT ( Figure 2D). Therefore, the interactions of either 328 or both CHC and α-AP2 with the CLAP domain of Amph1 are essential for optimal SV 329 regeneration at small central nerve terminals. 330 The interaction between Amph1 and Dyn1 is essential for SV endocytosis at a number 331 of large, atypical synapses (Jockusch et al. 2005;Shupliakov et al. 1997). We next 332 determined whether this interaction was also required at small central synapses. We 333 avoided truncating the Amph1 SH3 domain to address this question, since we 334 discovered that its removal affects interactions with multiple partners (Figure 1B-F interaction mutant was expressed in Amph1-depleted neurons, it failed to restore SV 338 endocytosis kinetics following stimulation at 10 Hz for 30 s ( Figure 2E,2F). Therefore, 339 the Amph1-Dyn1 interaction is essential for optimal SV endocytosis at central nerve 340 terminals. 341

342
The Amph1 PRD interacts with the SH3 domain of endophilin A1 (Micheva et al. 343 1997b) however, the specific interaction site on Amph1 remains unknown. Therefore, 344 to determine the role of the Amph1-endophiln A1 complex in SV recycling, it is 345 essential to pinpoint the exact binding site on Amph1. In contrast to Amph1, the binding 346 site of endophilin A1 on Amph2 has been mapped, and consists of a RKGPPVPPLP 347 motif within its PRD domain (Micheva et al. 1997b formation of an intramolecular loop ( Figure 3A). 369 To determine the endophilin A1 interaction site, recombinant GST-tagged Amph1 370 mutants were used as baits to perform pull-down assays from brain synaptosomal 371 lysates. Their potential interactions with endophilin A1 as well as a number of other 372 SH3-containing molecules, such as PSD-95 and syndapin, were then assessed by 373 immunoblotting ( Figure 3B). We revealed that all three PPVPP-related mutants 374 (A1A2VA3A4, A1A2, and A3A4) disrupted the binding to endophilin A1 ( Figure 3C). The 375 AVRA mutant also exhibits compromised binding, however, this did not reach 376 statistical significance ( Figure 3C). These data indicate that endophilin A1 377 preferentially interacts with the motif 301 PPVPP 305 on Amph1 with all the proline 378 residues being variably critical. Importantly, mutation of this site has no impact on 379 either CHC or α-AP2 interactions ( Figure 3B). 380 We next determined the effect of mock phosphorylation of S293 on endophilin A1 381 binding. We found that the phospho-mimetic mutant S293E abolished the Amph1-382 endophilin A1 interaction, in contrast to the phospho-null substitution S293A which 383 was not significantly different to GST-Amph1-WT ( Figure 3C). As above, these 384 phospho-null and -mimetic substitutions did not impact the association of Amph1 with 385 other SH3-containing endocytosis proteins or CHC and α-AP2 ( Figure 3B). Therefore, 386 the phosphorylation status of Amph1 S293 is a key regulator of the Amph1-endophilin 387 A1 complex. 388

389
The functional role of the Amph1-endophilin A1 interaction in SV recycling has never 390 been determined. Our generation of Amph1 mutants that selectively ablate endophilin 391 A1 binding allow this question to be addressed for the first time. We therefore 392 examined whether these Amph1 PRD mutants were able to restore defects in SV 393 recycling observed when Amph1 is depleted in central neurons. As before, 394 hippocampal neurons were transfected with shAmph1, or full-length Amph1-mCer with 395 one of the PRD mutants (AVRA, A1A2VA3A4, A1A2, A3A4) or an empty mCer vector 396 control. All mutants were expressed in the shAmph1-silenced background as 397 confirmed by immunolabelling (Supplementary figure 3A,3B). Neurons were 398 challenged with 300 APs at 10 Hz as before ( Figure 4A,4C) and the sypHy response 399 quantified. The AVRA mutant (which did not significantly impact endophilin A1 binding) 400 was able to restore SV endocytosis equivalent to Amph1-WT-mCer ( Figure 4A,4B). In 401 contrast, the three other PRD mutants (which did inhibit endophilin A1 binding) could 402 not restore normal SV endocytosis in Amph1-knockdown neurons ( Figure 4C,4D). 403 Intriguingly, the A1A2VA3A4 mutant, together with A3A4 mutant, resulted in slightly more 404 pronounced retardation of SV endocytosis, implying a potential dominant-negative 405 role. Taken together, these data reveal that the association of Amph1 with endophilin 406 A1 is required for efficient SV endocytosis at central nerve terminals. The Amph1 phospho-mutants, S293A and S293E, differ in their ability to bind to 410 endophilin A1, implying that the activity-dependent phosphorylation status of Amph1-411 S293 regulates SV recycling through the assembly/disassembly of the Amph1-412 endophilin A1 complex. Therefore we next determined whether Amph1 S293 413 phosphorylation controlled SV endocytosis. First, to examine any dominant-negative 414 effect of these phospho-mutants, hippocampal neurons were co-transfected with 415 sypHy and either S293A or S239E to assess their impact on SV endocytosis following 416 stimulation at 10 Hz for 30 s ( Figure 5A). In parallel experiments, sypHy was co-417 expressed with either Amph1-WT-mCer or mCer empty vector ( Figure 5A). 418 Overexpression of Amph1-WT-mCer had no dominant-negative effect on SV 419 endocytosis when compared to the empty vector control ( Figure 5B). Similarly, 420 overexpression of either S293A or S293E did not alter SV endocytosis kinetics in WT 421 neurons ( Figure 5B). Therefore the phospho-null and -mimetic S293 mutants are not 422 dominant-negative. 423 Next, we determined whether the S293A or S293E mutants could restore function in 424 neurons depleted of endogenous Amph1. As before, neurons were co-transfected with 425 shAmph1 and mCer-tagged versions of full-length Amph1-mCer or mCer. All mutants 426 were expressed when assessed by immunolabelling in Amph1-silenced neurons 427 (Supplementary figure 3A,3B). Following stimulation with 300APs at 10 Hz ( Figure  428 5C), neither S293A nor S293E was able to restore SV endocytosis in Amph1-silenced 429 neurons in contrast to WT Amph1 ( Figure 5D). Overall, these data reveal that the 430 phosphorylation status of Amph1-S293 is critical for controlling both endophilin A1 431 interactions and SV recycling at central synapses. 432

Discussion 433
Amph1 is established as a key mediator of SV endocytosis at central synapses, via a 434 series of defined interactions with molecules such as CHC, α-AP2, and Dyn1. Using a 435 molecular replacement strategy, we confirmed that these interactions are critical for 436 optimal SV endocytosis at typical small central nerve terminals. Furthermore, we 437 mapped the interaction site for endophilin A1 within the Amph1 PRD and revealed for 438 the first time that this interaction is essential for efficient SV endocytosis. Finally, we 439 determined that the activity-dependent phosphorylation status of S293 within the 440 Amph1 PRD is critical in controlling its interaction with endophilin A1 and SV 441 endocytosis. 442 These findings were revealed via the integration of interaction assays using Amph1 443 fragments that were fused to GST as bait, coupled to molecular replacement 444 experiments where full-length Amph1 mutants were expressed in neurons in which 445 endogenous Amph1 was depleted. with α-AP2 but not CHC. In contrast, Dyn1 is recruited independent to any prior 535 interaction with either α-AP2 or CHC. Therefore, there is potential for the 536 intramolecular interaction between the Amph1 PRD and SH3 domains to generate a 537 binding platform for α-AP2 in addition to previously characterised motifs. The near 538 absence of α-AP2 binding to Amph1 that lacks an SH3 domain suggests that this 539 interaction may only occur when the SH3 domain of Amph1 is unoccupied by external 540 ligands such as Dyn1. This has important implications for the sequence of Amph1-541 dependent interactions during SV endocytosis with α-AP2 recruitment potentially 542 occurring prior to Dyn1 association, after its dissociation, or both. 543 In summary, we have revealed a multifaceted role for Amph1 in SV endocytosis at 544 central nerve terminals. We confirmed key roles for interactions via both its CLAP and 545 SH3 domain and revealed that an activity-and phospho-dependent interaction with 546 endophilin A1 is essential for efficient SV endocytosis. 547