Differential vesicular sorting of AMPA and GABA A receptors

PNAS PLUS Differential vesicular sorting of AMPA and GABAA receptors Yi Gua,1, Shu-Ling Chiua,2, Bian Liua,2, Pei-Hsun Wub,2, Michael Delannoyc, Da-Ting Lina,3, Denis Wirtzb, and Richard L. Huganira,4 a Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, MD 21205; bDepartment of Chemical and Biomolecular Engineering, Johns Hopkins University, Baltimore, MD 21218; and cJohns Hopkins University School of Medicine Microscope Facility, Baltimore, MD 21205 In mature neurons AMPA receptors cluster at excitatory synapses primarily on dendritic spines, whereas GABAA receptors cluster at inhibitory synapses mainly on the soma and dendritic shafts. The molecular mechanisms underlying the precise sorting of these receptors remain unclear. By directly studying the constitutive exocytic vesicles of AMPA and GABAA receptors in vitro and in vivo, we demonstrate that they are initially sorted into different vesicles in the Golgi apparatus and inserted into distinct domains of the plasma membrane. These insertions are dependent on distinct Rab GTPases and SNARE complexes. The insertion of AMPA receptors requires SNAP25–syntaxin1A/B–VAMP2 complexes, whereas insertion of GABAA receptors relies on SNAP23–syntaxin1A/B–VAMP2 complexes. These SNARE complexes affect surface targeting of AMPA or GABAA receptors and synaptic transmission. Our studies reveal vesicular sorting mechanisms controlling the constitutive exocytosis of AMPA and GABAA receptors, which are critical for the regulation of excitatory and inhibitory responses in neurons. AMPA receptor | GABA A receptor | constitutive exocytosis | TIRFM | SNARE I n the mammalian central nervous system, neurons receive excitatory and inhibitory signals at synapses. Specific receptors at postsynaptic membranes are activated by neurotransmitters released by presynaptic terminals. Most fast excitatory neurotransmission is mediated by AMPA receptors, the majority of which are heterotetramers of GluA1/GluA2 or GluA2/GluA3 subunits in the hippocampus (1). Fast synaptic inhibition is largely mediated by GABAA receptors, which are predominantly heteropentamers of two α subunits, two β subunits, and one γ or δ subunit in the hippocampus (2). Numerous studies have demonstrated AMPA receptors are selectively localized at excitatory synapses on dendritic spines, whereas GABAA receptors cluster at inhibitory synapses localized on dendritic shafts and the soma (3). This segregation of excitatory and inhibitory receptors requires highly precise sorting machinery to target receptors to distinct synapses opposing specific presynaptic terminals. However, it is still not clear whether the receptors are sorted before exocytosis into the plasma membrane or are differentially localized only after exocytosis. For example in a “plasma membrane sorting model,” different receptors could be pooled into the same vesicle and inserted along the somatodendritic membrane. The initial sorting would occur on the plasma membrane, where inserted receptors would be segregated by lateral diffusion and stabilization at different postsynaptic zones. Alternatively, in a “vesicle sorting model,” different receptors would first be sorted into different vesicles during intracellular trafficking processes and independently inserted to the plasma membrane, where receptors could be further targeted to specific zones and stabilized by synaptic scaffolds. To date there has been no direct evidence to support either model. However, a large body of literature suggests that the exocytic pathways of AMPA and GABAA receptors have similar but also distinct properties (1, 2). Increasing evidence has suggested roles for the SNARE protein family in vesicular trafficking of AMPA and GABAA receptors (4– 17). SNAREs are a large family of membrane-associated proteins www.pnas.org/cgi/doi/10.1073/pnas.1525726113 critical for many intracellular membrane trafficking events. The family is subdivided into v-SNAREs (synaptobrevin/VAMP, vesicle-associated membrane proteins) and t-SNAREs (syntaxins and SNAP25, synaptosomal-associated protein of 25 kDa) based on their localization on trafficking vesicles or target membranes, respectively. To mediate vesicle fusion with target membranes, SNARE proteins form a four-helix bundle (SNARE complex) consisting of two coiled-coil domains from SNAP25, one coiled-coil domain from syntaxin, and a coiled-coil domain from VAMPs (18). Formation of the helical bundle can be disrupted by neurotoxins, which specifically cleave different SNARE proteins (19). Each SNARE subfamily is composed of genes with high homology but different tissue specificity and subcellular localization. It remains to be determined whether individual SNAREs play specific roles in regulating the membrane trafficking of individual proteins. To address how AMPA and GABAA receptors are sorted in the exocytic pathway and what molecules are involved in regulating exocytosis of these receptors, we specifically studied constitutive exocytosis of AMPA and GABAA receptor subunits using total internal reflection fluorescence microscopy (TIRFM) in combination with immunocytochemistry, electrophysiology, and electron microscopy methods. Together, we revealed that AMPA and GABAA receptors are initially sorted into different vesicles in the Golgi apparatus and delivered to different domains at the plasma membrane and are regulated by specific Rab proteins and Significance In neurons most fast excitatory neurotransmission is mediated by AMPA receptors, which cluster at excitatory synapses primarily on dendritic spines. Fast synaptic inhibition is largely mediated by GABAA receptors, which cluster at inhibitory synapses mainly on the soma and dendritic shafts. It is unclear how these receptors are segregated and delivered to specific locations on the plasma membranes. Here we directly studied the constitutive exocytosis of AMPA and GABAA receptors and demonstrate that they are initially sorted into different vesicles in the Golgi apparatus and inserted into distinct domains of the plasma membrane. Their exocytosis is dependent on distinct Rab GTPases and SNARE complexes. Our results reveal fundamental mechanisms underlying the sorting of excitatory and inhibitory neurotransmitter receptors in neurons. Author contributions: Y.G., D.-T.L., D.W., and R.L.H. designed research; Y.G., S.-L.C., B.L., P.-H.W., and M.D. performed research; Y.G., S.-L.C., B.L., and P.-H.W. analyzed data; and Y.G., S.-L.C., B.L., and R.L.H. wrote the paper. Reviewers: L.M., Medical College of Georgia; and S.J.M., Tufts University. The authors declare no conflict of interest. 1 Present address: Princeton Neuroscience Institute, Princeton University, Princeton, NJ 08544. 2 S.-L.C., B.L., and P.-H.W. contributed equally to this work. 3 Present address: National Institute on Drug Abuse, Baltimore, MD 21224. 4 To whom correspondence should be addressed. Email: rhuganir@jhmi.edu. This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1525726113/-/DCSupplemental. PNAS Early Edition | 1 of 10 NEUROSCIENCE Contributed by Richard L. Huganir, January 6, 2016 (sent for review November 6, 2015; reviewed by Lin Mei and Stephen J. Moss) SNARE complexes. These results reveal fundamental mechanisms underlying the sorting of excitatory and inhibitory neurotransmitter receptors in neurons and uncover the specific trafficking machinery involved in the constitutive exocytosis of each receptor type. Results Dynamic Events of AMPA and GABAA Receptors on the Plasma Membrane of Hippocampal Pyramidal Neurons. To visualize individual exocytosis events of AMPA or GABAA receptors in living hippocampal neurons, we used TIRFM to specifically image trafficking events at or immediately beneath the plasma membrane in contact with the coverslip (100–200 nm) (20). To further ensure the imaging of exocytic events, superecliptic pHluorin (pHluorin or pH) was chosen to tag the extracellular N terminus of AMPA and GABAA receptors. pHluorin is an EGFP variant that fluorescences brightly at pH 7.4 and is fully quenched in the lumen of secretory organelles having a pH <6 (21). Therefore, after exocytosis the fluorescent signal of pHluorin-tagged receptors dramatically increases under the exposure of imaging solution with pH 7.4 (14, 15, 22). pHluorin-tagged GluA2 (pH-GluA2) and γ2S (pH-γ2S) were used for the study, because these subunits are common subunits of AMPA and GABAA receptor complexes in hippocampus, respectively. Previous studies have confirmed that the pHluorin tag does not affect trafficking of these receptor subunits in neurons (15, 23). pHluorin-tagged GluA2 or γ2S was expressed in dissociated hippocampal neurons and directly visualized under TIRFM. Before recording, the entire cell surface in the TIRF field was photobleached to eliminate signals from preexisting surface receptors and isolate new exocytic events (22). We observed robust dynamic events of pH-GluA2 and pH-γ2S throughout the plasma membranes. Most events of GluA2 and γ2S occurred on the extrasynaptic membrane in the cell body and dendritic shafts (Fig. 1A and Movies S1 and S2). We did not observe events of pH-GluA2 on dendritic spines. These dynamic events transiently occurred at high frequency: 95.8% events of pH-GluA2 lasted less than 7 s with the main duration around 2.8 s, whereas 96.7% events of pH-γ2S lasted less than 7 s with main duration around 2.1 s (Fig. 1 B and C). The mean event duration of pH-GluA2 was significantly longer than that of pH-γ2S (Fig. S1A). There are 22 ± 2 events per second per 100 μm2 for pH-γ2S and 15 ± 1 events per second per 100 μm2 for pH-GluA2. These frequencies remained stable under imaging with higher frame rate (Fig. S1B). To confirm that these dynamic events are on the plasma membrane we performed an acidification–neutralization test (24), which included 15–30 s of TIRF imaging in pH 7.4 extracellular solution, then a fast perfusion for 15–30 s with pH 5.5 extracellular solution, followed by a return to pH 7.4 extracellular solution. Most of dynamic events of pH-GluA2 and pH-γ2S quenched upon acidic perfusion and recovered immediately after reneutralization (Fig. 1D and Fig. S1C). In addition, when neurons were perfused with the pH 7.4 solution containing ammonium chloride (NH4Cl), which rapidly alkalinized all of the acidic intracellular pools and revealed intracellular pH receptors (21), the frequency of the dynamic TIRF events remained constant. These results strongly suggest that the dynamic events under TIRFM present on the plasma membrane. Moreover, the frequency of these events was not regulated by neuronal activity, which was acutely suppressed or enhanced by brief application of TTX or KCl, respectively (Fig. S1D), suggesting these are constitutive trafficking events. Overall, these results indicate that the transient extrasynaptic events of GluA2 and γ2S under TIRFM are constitutive dynamics of receptors on the plasma membrane. We noticed that most of the events of GluA2 and γ2S have dim fluorescence intensity, suggesting that each event contains a low number of receptor subunits. To confirm this observation, we measured the number of fluorescent receptors per event (22). Based on the knowledge that the fluorescent intensity of the EGFP monomer is similar to the intensity of pHluorin in the environment of pH 7.4 (24), we compared the fluorescent intensity Fig. 1. Exocytic events of pH-GluA2 and pH-γ2S under TIRFM. (A) Dynamic TIRF events of pH-GluA2 and pH-γ2S are highlighted based on intensity. Typical events are indicated by white arrowheads. Imaging frequency is 1.4 Hz with 500-ms exposure. (Scale bars: 5 μm.) (B and C) Dynamic events of pHGluA2 (B) and pH-γ2S (C) are transient. (Top) Time series of a single event. (Bottom) Distribution of event durations. Arrows indicate the main event duration. n = 120 for both receptors. (Scale bars: 1 μm.) (D) Acidification–neutralization analysis of pH-GluA2 and pH-γ2S events. From left to right: cells in the extracellular solution at pH 7.4, 5.5, and 7.4. (Scale bars: 5 μm.) (E) Dynamics of an exocytic vesicle containing two receptor subunits differentially tagged with pHluorin and tdTomato. (F) Predicted dynamics of tdTomato and pHluorin receptors in the same exocytic vesicle under TIRFM. (G and I) TIRF dynamics of a coinsertion vesicle containing tdt-GluA2 and pH-GluA2 (G) or tdt-γ2S and pH-γ2S (I). (Scale bars: 1 μm.) (H and J) Time course of tdt-GluA2 and pHGluA2 (H) or tdt-γ2S and pH-γ2S (J) from multiple coinsertion events. In H, n = 27; in J, n = 31. (K and L) Dynamic events of GluA2 and γ2S are inhibited by Botox B (K) or Botox C (L). The same control dataset was used. Each data point is one cell. The line in each dataset shows mean frequency. Event frequencies of all cells were normalized by the mean of the control. Asterisks indicate statistical significances. 2 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1525726113 Gu et al. Dynamic TIRF Events of GluA2 and γ2S Are Exocytic Events. Several lines of evidence suggest that the dynamic surface events of GluA2 and γ2S under TIRFM are exocytic events. First, the exocytic feature is supported by the stereotypic dynamics of these events under TIRFM. As demonstrated in Fig. 1E, when two subunits of the same receptor differentially tagged with pHluorin and a red fluorescent protein (pH-insensitive; for example, tdTomato) and delivered in the same exocytic vesicle (coinsertion), they exhibit different dynamics under TIRFM. The pH-insensitive red fluorescent protein is excited immediately when the exocytic vesicle enters the TIRF field. However, pHluorin remains quenched until it is exposed to the extracellular space (pH 7.4) after the exocytosis. Therefore, the red fluorescence increases in advance of the green fluorescence (Fig. 1F). To investigate whether the events of GluA2 and γ2S also exhibit this stereotypic dynamics of exocytosis under TIRF, we first tagged GluA2 and γ2S with pH-insensitive red fluorescent protein tdTomato and characterized the tagged receptor subunits (Fig. S2). In live hippocampal neurons tdt-GluA2 or tdt-γ2S colocalized with EGFP-GluA2 or EGFP-γ2S, respectively (Fig. S2 A and D). In addition, tdt-GluA2 (Fig. S2 B and C) and tdtγ2S (Fig. S2 E and F) also colocalized well with endogenous GluA1 and β2/3, respectively, indicating that these tdTomatotagged receptors trafficked similarly to endogenous receptors. Moreover, tdt-GluA2 and tdt-γ2S could be stained in live cells with anti-tdTomato antibody, indicating that these receptors were expressed on the surface (Figs. S3 and S4). Total and surface tdtGluA2 colocalized with the excitatory postsynaptic marker PSD95 (postsynaptic density protein 95) (Fig. S3 A and B) and the presynaptic marker VGluT (vesicular glutamate transporter) (Fig. S3 C and D). Similarly, total and surface tdt-γ2S colocalized with the inhibitory postsynaptic marker gephyrin (Fig. S4 A and B) and the presynaptic marker VGAT (vesicular GABA transporter) (Fig. S4 C and D). These data suggest that tdTomato-tagged GluA2 and γ2S are properly trafficked and targeted in hippocampal neurons. We then coexpressed pH receptor and tdt receptor and simultaneously visualized their exocytosis under dual-color TIRFM with 488-nm and 568-nm lasers to excite green and red fluorescent proteins, respectively. Dynamics of events containing both green and red fluorescence signals (coinsertion events) were analyzed. As expected, in the coinsertion events of pH-GluA2 and tdtGluA2, the fluorescence of tdt-GluA2 increased earlier than that of pH-GluA2 (Fig. 1 G and H). Similar dynamics was also observed in the coinsertion events of pH-γ2S and tdt-γ2S (Fig. 1 I and J). The particular dynamics of these coinsertions under TIRF strongly suggest that they are exocytic events. In many events, we also observed that the fluorescence of tdTomato receptor decayed faster than the pHluorin fluorescence. This phenomenon is likely caused by the photoinstability of tdTomato compared to pHluorin (26). Gu et al. Exocytosis of AMPA and GABAA Receptors Is Mediated by Different SNAPs. Although Botox B and C inhibited exocyosis of both GluA2 and γ2S, Botox A, which cleaves SNAP25 (Fig. S5 A and B), showed different effects on exocytosis of pH-GluA2 and pH-γ2S. As shown in Fig. 2A, Botox A inhibited exocytosis of GluA2, but not γ2S. These Botox-treatment results indicate that exocytosis of both AMPA and GABAA receptors require VAMP2 and syntaxins, but the different receptors likely require different SNAPs: SNAP25 mediates AMPA receptor insertion, whereas a Botox A-insensitive SNAP mediates GABAA receptor insertion. Three Botox A-insensitive SNAPs (SNAP23, SNAP29, and SNAP47) are expressed in rat hippocampal neurons (29). All three SNAPs are known to regulate membrane fusion events in neurons (13, 16, 17, 30–32). We examined the effect of shRNAs that specifically knock down each SNAP (Fig. S7 A–E) on the exocytosis frequencies of AMPA and GABAA receptors, compared with a control shRNA with a sequence not targeting any known vertebrate genes. Consistent with Botox A treatment, SNAP25 knockdown reduced the frequency of exocytosis of GluA2, but not γ2S, and this inhibition was fully rescued by shRNA-resistant SNAP25 but not SNAP23 (Fig. 2B). However, knockdown of SNAP23 blocked exocytosis of γ2S, but not GluA2, and this effect was rescued by shRNA-resistant SNAP23 but not SNAP25 (Fig. 2C). The effects of SNAP25 and SNAP23 shRNAs were observed on both somatic and dendritic exocytosis of pH-GluA2 and pHγ2S (Fig. S7 J–M). Knockdown of SNAP29 or SNAP47 did not significantly affect exocytosis of either GluA2 or γ2S (Fig. S7 N and O). These results demonstrated that SNAP25 and SNAP23 specifically mediate the constitutive exocytosis of GluA2- and γ2Scontaining receptors, respectively. If SNAP25 and SNAP23 knockdown reduced constitutive exocytosis of AMPA and GABAA receptors, respectively, we would predict that this would decrease the surface expression of these receptors. To test this possibility, we examined surface levels of endogenous GluA2 and γ2 by immunostaining of surface receptors PNAS Early Edition | 3 of 10 PNAS PLUS The second evidence of exocytosis is based on the results of botulinum toxin (Botox) treatments. Receptor exocytosis occurs when an intracellular vesicle, which carries assembled receptor complexes, fuses to the plasma membrane and the receptor complexes are delivered to the plasma membrane (27). This process highly depends on SNARE proteins, which can be cleaved by different Botoxs (19). Therefore, we tested the effects of Botoxs on the TIRF event frequencies of pH-GluA2 and pH-γ2S. Botox B, which cleaves VAMP2 (Fig. S5 C and D), reduced frequency of both GluA2 and γ2S events (Fig. 1K). Similarly, Botox C, which cleaves rat SNAP25, syntaxin1A, 1B, 2, and 3 (Fig. S5 E–J), inhibited events of GluA2 and γ2S (Fig. 1L). Notably, we detected a low degree of the incomplete blockade of exocytic events, which was commonly reported with botulinum toxin treatments. This is likely due to the inability of Botoxs to proteolyze SNARE proteins in assembled complexes (Fig. S5) (28). Finally, we also observed receptor dispersion to the surrounding regions after exocytosis, which is another stereotypic dynamic of exocytic events (14). As shown in Fig. S6, pH-GluA2 events showed increased fluorescence in the surrounding region while the fluorescence in the insertion spot decayed (Fig. S6 A and B). The appearance of fluorescence peak in the surrounding region was significantly delayed in comparison with the one in the insertion center, strongly indicating the receptor dispersion after the insertion (Fig. S6 C and D). A similar phenomenon was observed for pH-γ2S events (Fig. S6 E–H). In addition, many events showed the separation of receptor subunits during this dispersion process (Fig. S6 A and E), suggesting that each inserted receptor complex can diffuse independently. In summary, the Botox sensitivity and stereotypic dynamics of these events under TIRFM strongly suggest that they are exocytic events of GluA2 and γ2S. NEUROSCIENCE of EGFP monomer to the intensity of single events of pH-GluA2 and pH-γ2S under TIRFM. EGFP monomers were confirmed by their blinking dynamics and single-step photobleaching property (22) (Fig. S1E). The intensity of EGFP monomers, pH-GluA2– containing vesicles, and pH-γ2S–containing vesicles follows Gaussian distributions (Fig. S1 F–H). The peak intensities of fitted Gaussian curves for EGFP monomers, pH-GluA2, and pH-γ2S events indicate that each pH-GluA2 event contains on average two pH-GluA2 subunits (2.2 ± 0.1 subunits per event), whereas each pH-γ2S event contains around four pH-γ2S subunits (3.9 ± 0.2 subunits per event). We and others have previously characterized larger, much less frequent GluA2 and GluA1 insertion events that have slower kinetics and are distinct from these rapid insertion events (14, 15, 22). Because of the much lower frequency of these larger events (two to six insertions per minute) they did not significantly contribute to the quantitation and characterization of the smaller events. Fig. 2. Exocytic events of GluA2 and γ2S are constitutive events mediated by different SNAPs. (A) Insertion of GluA2 and γ2S are differently affected by Botox A treatment. (B) SNAP25 is required for insertions of GluA2, but not γ2S. Scramble shRNA (Scr), SNAP25 shRNA, shRNA-resistant SNAP25 (25), or SNAP23 (23) were coexpressed with pH-receptors for 24 h. (C) SNAP23 is required for insertions of γ2S, but not GluA2. Scramble shRNA (Scr), SNAP23 shRNA, shRNA-resistant SNAP23 (23), or SNAP25 (25) were coexpressed with pH-receptors for 48 h. (D) Knockdown of SNAP25 reduces endogenous surface GluA2. SNAP25 shRNA, shRNA-resistant SNAP25 (SNAP25), or SNAP23 were coexpressed with EGFP to label transfected hippocampal neurons for 72 h. Right panels: high-magnification images of the individual processes boxed in left panels. N, surface GluA2 in nontransfected neurons; T, surface GluA2 in transfected neurons. (Scale bar: 20 μm.) (E) Quantification of surface GluA2 in D. Relative surface GluA2 in transfected neurons is represented by a ratio of surface GluA2 in transfected cells to that of nontransfected cells. The ratios were normalized by the average of the control transfected with the scramble shRNA. Scr: scramble shRNA. 25: shRNA-resistant SNAP25. 23: SNAP23. n = 18–22 neurons for each group. Five processes were selected in each neuron. (F) Knockdown of SNAP23 reduces endogenous surface γ2. SNAP23 shRNA, shRNA-resistant SNAP23 (SNAP23), or SNAP25 were coexpressed with EGFP to label transfected cells for 72 h. (Scale bar: 20 μm.) (G) Quantification of surface γ2 in F was performed similarly as in E. Scr: scramble shRNA. 23: shRNA-resistant SNAP23. 25: SNAP25. n = 25–30 neurons for each group. Five processes were selected in each neuron. Asterisks indicate statistical significances. n.s., no statistical significance. while the SNAPs were specifically knocked down. Indeed, SNAP25 shRNA significantly reduced surface GluA2 levels in hippocampal neurons. This effect was rescued by shRNA-resistant SNAP25 but not SNAP23 (Fig. 2 D and E). Conversely, SNAP23 shRNA drastically reduced surface γ2 levels, which was rescued by shRNAresistant SNAP23 but not SNAP25 (Fig. 2 F and G). However, knockdown of SNAP23 or SNAP25 did not affect surface levels of GluA2 or γ2, respectively (Fig. S8). We then asked whether SNAP25 and SNAP23 regulate surface expression of endogenous GluA2 and γ2 subunits at synapses, respectively. In hippocampal neurons knockdown of SNAP25, rather than SNAP23, reduced synaptic surface GluA2, which colocalized with the excitatory presynaptic marker VGluT (Fig. 3 A and B). However, knockdown of SNAP23, but not SNAP25, significantly reduced synaptic surface levels of γ2, which colocalized with the inhibitory presynaptic marker VGAT (Fig. 3 C and D). Given our observations that most insertions of GluA2 and γ2S occur at extrasynaptic sites (Movies S1 and S2), these reductions in synaptic surface GluA2 and γ2 could result from the depletion of extrasynaptic surface receptors, which supply synaptic receptor pools by lateral diffusion to the postsynaptic membrane. These data suggest that SNAP25 and SNAP23 regulate not only the surface expression of GluA2 and γ2 subunits throughout the entire neuron, but also specifically regulate the synaptic surface expression of GluA2 and γ2, respectively. The reduced surface expressions of GluA2 and γ2 at synapses suggest that SNAP25 and SNAP23 may affect AMPA and GABAA receptor-mediated synaptic transmission. To test this hypothesis, whole-cell patch-clamp recording was used to examine AMPA receptor-mediated mEPSCs (miniature excitatory postsynaptic 4 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1525726113 currents) and GABAA receptor-mediated mIPSCs (miniature inhibitory postsynaptic currents) in hippocampal neurons with SNAP25 or SNAP23 knockdown by specific shRNAs. Strikingly, knockdown of SNAP25, but not SNAP23, preferentially reduced AMPA receptor-mediated mEPSC amplitude (Fig. 3 E and F), whereas knockdown of SNAP23, but not SNAP25, significantly reduced GABAA-receptor mediated mIPSC amplitude (Fig. 3 G and H). These results are consistent with the specific effects of SNAP25 and SNAP23 on constitutive exocytosis and synaptic surface levels of GluA2 and γ2S, respectively. The SNAP25 or SNAP23dependent exocytosis of AMPA or GABAA receptors significantly affects excitatory and inhibitory synaptic transmission in neurons at the basal state. Overall, our results reveal important postsynaptic roles of SNAP25 and SNAP23 on constitutive insertions, surface expression of GluA2 and γ2, and basal synaptic transmission mediated by AMPA and GABAA receptors, respectively. The distinct functions of SNAP25 and SNAP23 support the model that AMPA and GABAA receptors are inserted into the plasma membrane via different vesicles that are under regulation of specific SNAPs. Exocytosis of AMPA and GABAA Receptors Is Mediated by Syntaxin1 and VAMP2. We further investigated other SNARE components necessary for fusion of AMPA or GABAA receptor-containing vesicles to the plasma membrane. In rat hippocampal neurons, five syntaxins are expressed on the plasma membrane: syntaxin1A, syntaxin1B, syntaxin2, syntaxin3, and syntaxin4 (33). All syntaxins, except syntaxin4, can be cleaved by Botox C (Fig. S5 E–J) (34). SNAP25 and SNAP23 have higher affinities to syntaxin1A, 1B, and syntaxin4 than to syntaxin2 and 3 (35–37). We therefore Gu et al. tested whether syntaxin1A, 1B, or syntaxin4 could be the t-SNAREs mediating exocytosis of GluA2 or γ2S. We used specific shRNAs to knock down these three syntaxins (Fig. S7 F–H) and examined the effect on exocytosis of GluA2 or γ2S. Knockdown of syntaxin1A and 1B significantly reduced exocytic frequencies of both GluA2 and γ2S. This effect could be rescued by shRNA-resistant syntaxin1A and 1B, respectively (Fig. 4 A and B). Knockdown of syntaxin4 did not affect exocytosis of either GluA2 or γ2S (Fig. S7P). In conclusion, the constitutive insertion of GluA2- or γ2Scontaining vesicles into the plasma membrane is commonly mediated by two t-SNAREs: syntaxin1A and 1B. There are two VAMPs, VAMP1 and VAMP2, specifically expressed in the rat brain, including hippocampus (38–40). To test which VAMPs are important for pH-GluA2 and pH-γ2S exocytosis, we used shRNAs to specifically knock down VAMP1 and VAMP2 (Fig. S7I) and investigated their effects on the exocytosis of GluA2 and γ2S. Consistent with our previous observation that Botox B, which cleaves VAMP2, reduced the exocytic frequency of both pH-GluA2 and pH-γ2S (Fig. 1K), depletion of VAMP2 significantly reduced exocytic frequencies of both pH-GluA2 and pH-γ2S (Fig. 4C). However, knockdown of VAMP1, which is not cleaved by Botox B, had no effect on exocytosis of either pH-GluA2 or pH-γ2S (Fig. S7Q). Together, these results demonstrate that VAMP2, but not VAMP1, serves as a v-SNARE mediating the constitutive exocytosis of both GluA2- and γ2Scontaining AMPA and GABAA receptors. Gu et al. Exocytosis of AMPA and GABAA Receptors Is Differentially Regulated by Specific Rab Proteins. Surface receptors can be delivered to the plasma membrane along different trafficking pathways, such as the de novo exocytic pathway originating from Golgi apparatus and recycling pathways involving early and recycling endosomes. These exocytic pathways are regulated by the small GTPase Rab protein family (41). To investigate the source of the receptorcontaining vesicles, we coexpressed pHluorin-tagged receptors with dominant negative Rab proteins that interfere with specific trafficking pathways. A dominant negative Rab8 [Rab8(T22N)], which blocks vesicle trafficking from the Golgi apparatus to the plasma membrane (de novo exocytosis), reduced the exocytic frequency of both GluA2 and γ2S (Fig. 4D). Dominant negative Rab4, 5, and 11 [Rab4(S22N), Rab5(S34N), and Rab11(S25N)], which block different steps in the vesicle recycling pathway including sorting from early endosomes to the plasma membrane, endocytosis, and trafficking from recycling endosomes to the plasma membrane, only significantly inhibited exocytosis of GluA2, but not γ2S (Fig. 4 E–G). These results suggest that constitutive exocytic events of GluA2 include both de novo exocytic and recycling events, whereas constitutive exocytic events of γ2S are mostly de novo exocytic events. Exocytosis of AMPA and GABAA Receptors Targets Different Zones on the Plasma Membrane. Exocytic events of pH-GluA2 and pH-γ2S not only occurred under different molecular mechanisms, but also show distinct spatial targeting on the plasma membrane. PNAS Early Edition | 5 of 10 PNAS PLUS NEUROSCIENCE Fig. 3. Effects of of SNAP25 and SNAP23 on synaptic surface expression of GluA2 and γ2S and basal synaptic transmission. (A) Knockdown of SNAP25, but not SNAP23, reduces the surface GluA2 at the excitatory postsynaptic membrane. Scramble, SNAP25, or SNAP23 shRNA were coexpressed with EGFP (to label transfected neurons) in hippocampal neurons for 72 h. (Top) Surface GluA2 (s-GluA2). (Middle) VGluT. (Bottom) White puncta showing colocalization of s-GluA2 (magenta) and VGluT (green). Synaptic surface GluA2 in shRNA-transfected cells (shRNA) was compared with nontransfected cells (−) on the same coverslip. Yellow arrows at the corresponding locations in all three panels indicate the s-GluA2, VGluT, and overlaid signals at the same synapse. (Scale bar: 10 μm.) (B) Quantification of the synaptic surface GluA2 in A. In each shRNA-transfected neuron, relative synaptic surface GluA2 was computed as a ratio of the surface synaptic GluA2 of transfected neurons to that of nontransfected neurons. The ratio in each sample was normalized by the average of the scramble shRNA control. Scr: scramble shRNA. 25: SNAP25 shRNA. 23: SNAP23 shRNA. n = 18–22 for shRNAtransfected neurons. Ten synapses were selected in each transfected neuron and nontransfected neurons. (C) Knockdown of SNAP23, but not SNAP25, reduces the surface γ2 at the inhibitory postsynaptic membrane. Scramble, SNAP23, or SNAP25 shRNA were coexpressed with EGFP (to label transfected neurons) for 72 h. (Scale bar: 10 μm.) (D) Quantification of the surface synaptic γ2 in C was performed similarly as in B. n = 28∼30 for shRNA-transfected neurons. (E and G) Whole-cell recordings were performed on hippocampal neurons 3 d after transfection of scramble (Scr), SNAP25 (25), or SNAP23 shRNAs (23). Representative traces of spontaneous AMPA receptor-mediated mEPSCs and GABAA receptor-mediated mIPSCs for each group are shown in E and G, respectively. (F) Quantification of AMPA mEPSC amplitudes in E. SNAP25-knockdown neurons have significantly smaller amplitudes than control (Mann–Whitney test, P < 0.01). Scramble shRNA = −18.2 ± 0.65 pA, SNAP25 shRNA = −15.1 ± 0.69 pA, SNAP23 shRNA = −17.2 ± 1.63 pA. n = 11–14 for each group. (H) Quantification of GABAA mIPSC amplitudes in G. SNAP23 knockdown neurons have significantly smaller amplitudes than control (Mann–Whitney test, P < 0.001). Scramble shRNA = −55.8 ± 4.67 pA, SNAP23 shRNA = −31.8 ± 2.99 pA, SNAP25 shRNA = −51.3 ± 4.55 pA. n = 10–13 for each group. Asterisks indicate statistical significances. n.s., no statistical significance. Fig. 4. Syntaxin1, VAMP2, and specific Rab proteins regulate exocytosis of GluA2 and γ2S. (A–C) Syntaxin1A (A), syntaxin1B (B), and VAMP2 (C) are required for exocytosis of γ2S and GluA2. Scramble shRNA (Scr), syntaxin1A (STX1A), syntaxin1B (STX1B), or VAMP2 shRNA (VAMP2) and shRNA-resistant syntaxin1A (1A), syntaxin1B (1B), or VAMP2 (2) were coexpressed with pH receptors for 48 h. (D–G) Effects of dominant negative Rabs on exocytosis of GluA2 and γ2S. Rab8(T22N) (D), Rab4(S22N) (E), Rab5(S34N) (F), Rab11(S25N) (G), or empty vector was coexpressed with pH receptors for 24 h. The same empty vector control was used in D–G. Asterisks indicate statistical significances. n.s., no statistical significance. Using an intensity-based program, exocytic events of pH-GluA2 and pH-γ2S were automatically isolated (Fig. S9A). Strikingly, we found that the pH-GluA2 exocytic events occur in the central region of the plasma membrane in contact with the coverslip, whereas the pH-γ2S exocytic events distribute in the peripheral region of the plasma membrane (Fig. 5A and Movies S3 and S4). To confirm the spatial segregation of exocytosis of pH-GluA2 and pH-γ2S, we coexpressed GluA2 or γ2S tagged with tdTomato or pHluorin in the same cell and simultaneously visualized their exocytic events using dual-color TIRFM. Consistent with the previous observations, exocytic events of pH-γ2S and tdt-GluA2 have different distributions on the somatic plasma membrane in contact with the coverslip. Vesicles containing pH-γ2S are mainly targeted to the outer peripheral region of the soma, whereas tdt-GluA2–containing vesicles are preferentially targeted to the inner central region of the soma (Movie S5). Quantification of these observations, by counting the number of exocytic events along the long axis of the somatic region, confirmed that exocytic vesicles of pH-γ2S and tdt-GluA2 are spatially segregated on the somatic plasma membrane (Fig. 5 B–D). To rule out any potential artifacts of the fluorescent tags, we swapped the fluorescent tags on the two receptor subunits and imaged exocytosis of pH-GluA2 and tdt-γ2S. Exocytic events of these receptors displayed the same distributions as tdt-GluA2 and pH-γ2S, respectively (Fig. 5E). Exocytic events of pH-GluA2 and tdt-GluA2 occur with a similar inner central-somatic distribution (Fig. 5F), whereas exocytic events of pH-γ2S and tdt-γ2S have the same outer peripheral-somatic distribution (Fig. 5G). In addition, the distribution of pH-GluA2 exocytic events was not affected by the coexpression of other AMPA receptor subunits, such as GluA1 and GluA3 (Fig. S9 B–E). Overall, our observations suggest that the exocytic events of excitatory AMPA receptors and inhibitory GABAA receptors are spatially segregated. The exocytosis of GluA2, a subunit present in most AMPA receptors, occurs at the inner region of the soma in contact with the coverslip, whereas the exocytosis of γ2S, a subunit present in most inhibitory GABAA receptors, occurs at the outer region of the soma in contact with the coverslip. Interestingly, this differential surface targeting of exocytosis of GluA2 and γ2S is regulated by Rab proteins. Whereas dominant negative Rab8 did not change the distribution of GluA2 exocytosis, which mostly occurs at the inner region of the somatic 6 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1525726113 membrane, the residual exocytic events of GluA2 after expression of dominant negative Rab4, Rab5, or Rab11 were distributed more evenly across the somatic membrane (Fig. 5H). The distribution of γ2S exocytosis was not affected by any of the Rab mutants (Fig. 5I). These data suggest that differential targeting of GluA2 and γ2S exocytosis on the plasma membrane potentially reflect specific exocytic pathways for each receptor. Although some GluA2-containing AMPA receptors can be delivered to the plasma membrane through the de novo exocytic pathway, the majority of AMPA receptors are delivered through recycling vesicles and inserted into the inner region of the soma in contact with the coverslip. However, most γ2S-containing GABAA receptors are delivered to the plasma membrane through de novo exocytic vesicles, which specifically insert at the outer regions of the soma. AMPA and GABAA Receptors Exit the Golgi Apparatus As Different Vesicles. Because the constitutive exocytosis of both AMPA and GABAA receptors seems to occur through a de novo exocytic pathway originating from the Golgi apparatus, we further asked whether the two receptor types are trafficked by different vesicles after they exit the Golgi. To image post-Golgi trafficking of receptors, we coexpressed EGFP- or tdTomato-tagged GluA2 and γ2S and then incubated transfected neurons at 20 °C to inhibit vesicle budding from the Golgi apparatus (42). Under this condition, we observed the accumulation of GluA2 and γ2S in the Golgi apparatus (Fig. S10A). After the 20 °C incubation, live neurons were imaged at 32 °C when post-Golgi trafficking is restored (43). We first examined whether the same receptor with different fluorescent tags is cotrafficked in the post-Golgi route by coexpressing EGFP-γ2S and tdt-γ2S, or EGFP-GluA2 and tdtGluA2. We were able to visualize trafficking vesicles containing both EGFP-γ2S and tdt-γ2S (Fig. S10B and Movie S6), or both EGFP-GluA2 and tdt-GluA2 (Fig. S10C and Movie S6), indicating cotrafficking of these differentially tagged receptor subunits. However, we also observed many vesicles containing EGFPor tdTomato-tagged subunits alone. This is likely due to the low number of receptors in each vesicle (Fig. S1 E–H) and the sensitivity of detection. In contrast, when we coexpressed EGFP-GluA2 and tdt-γ2S (Fig. 6A and Movie S7), or EGFP-γ2S and tdt-GluA2 (Fig. 6B and Movie S8), we very rarely observed the cotrafficking of GluA2 and γ2S. The percentage of cotrafficking events of different receptor pairs is significantly lower than that of same Gu et al. PNAS PLUS NEUROSCIENCE Fig. 5. GluA2 and γ2S are inserted into different domains of the somatic plasma membrane. (A) pH-γ2S and pH-GluA2 are inserted into different domains of the somatic plasma membrane. (Left) Neuron morphology under TIRFM. (Middle) Spatial location and intensity (shown in colors) of events for pH-GluA2 (n = 545) and pH-γ2S (n = 1,559) accumulating from 14 s. (Right) Heat map showing spatial density distributions of the events in the middle panel. Event density was calculated as the number of events per second per 100 μm2 within a circular region (diameter: 0.48 μm). (B) Exocytosis of pH-γ2S and tdt-GluA2 in the same cell. Green and magenta rectangular regions represent the same somatic region for pH-γ2S and tdt-GluA2, respectively. (Scale bar: 5 μm.) (C) Quantification of exocytic events of pH-γ2S (green) and tdt-GluA2 (magenta) in the region shown in B. Normalized numbers of events along the long axis of the selected region were plotted against the distance from the center (maximal distances from the center on both directions were normalized as 50% and −50%). (D–G) Averaged distributions of exocytic events of pH-γ2S and tdt-GluA2 (D), pH-GluA2 and tdt-γ2S (E), pH-GluA2 and tdt-GluA2 (F), and pH-γ2S and tdt-γ2S (G). Green and magenta curves represent exocytic event of pH and tdt receptors, respectively. n = 13–17 for each group. (H and I) Effects of dominant negative Rabs on exocytic event distributions of GluA2 (H) and γ2S (I). Empty vector (−), Rab8(T22N), Rab4(S22N), Rab5(S34N), or Rab11(S25N) was coexpressed with pH receptors. n = 16–20 for each group. Asterisks indicate statistical significance compared with empty vector control. receptor pairs (Fig. 6C), suggesting that vesicles exiting the Golgi carry preferentially GluA2 or γ2S alone. These results indicate that GluA2 and γ2S receptors are trafficked in separate vesicles after they exit the Golgi apparatus. Endogenous AMPA and GABAA Receptors Are Sorted into Different Vesicles. Our results in cultured hippocampal neurons strongly suggest that AMPA and GABAA receptors are sorted into different intracellular vesicles before exocytosis. To further investigate whether endogenous receptors were also sorted into separate intracellular vesicular compartments in vivo we performed double-label immunogold EM studies in microsomeenriched fractions (P3) from adult rat brain. Rat brain homogenates were fractionated by differential centrifugation (44) and the fractions were characterized using markers of major intracellular organelles and vesicles (Fig. 7A). The P3 fraction contains membranes from the endoplasmic reticulum (ERP72, endoplasmic reticulum protein 72), lysosomes (LAMP1, lysosomal-associated Gu et al. membrane protein 1), early endosomes (EEA1, early endosome antigen 1), recycling endosomes (synaxin13), and Golgi apparatus (TGN38, trans-Golgi network integral membrane protein 38). Other SNARE proteins, such as SNAP23, SNAP25, and VAMP2, were also present in the P3 fraction. Furthermore, GluA2 and γ2 are enriched in P3 fraction. The EM morphology of P3 fraction showed that the P3 pellet contained intact vesicular structures with different sizes (Fig. 7 B and C). Double-immunogold labeling was performed on thin sections of the P3 pellet after a light fixation (Fig. 7D). The morphology of small intracellular trafficking vesicles, which are 50–300 nm in diameter (44), was largely preserved under this condition. GluA2 and γ2 were labeled by specific primary antibodies and secondary antibodies conjugated to 6-nm and 12-nm gold particles, respectively. The average number of 6-nm or 12-nm gold particles on each vesicle is two or three, respectively. The majority of vesicles (88%) contained only a single type of receptor whereas 12% of the vesicles contained both GluA2 and γ2. We observed that 37% of PNAS Early Edition | 7 of 10 Fig. 6. GluA2 and γ2S are trafficked in different vesicles when they exit the Golgi apparatus. (A) Time series of a post-Golgi trafficking vesicle containing only EGFP-GluA2, but not tdt-γ2S, as indicated by arrows at corresponding locations. (Top) EGFP-GluA2. (Middle) tdt-γ2S. (Bottom) Overlay of top and middle panels. (Scale bar: 2.5 μm.) The kymographs show the trafficking of the vesicle along its trajectory for EGFPGluA2, tdt-γ2S and overlaid signal. (B) Time series of a post-Golgi trafficking vesicle containing only EGFPγ2S, but not tdt-GluA2. (C) Quantification of cotrafficking events of EGFP- and tdt-tagged receptors after exit the Golgi apparatus. Asterisks indicate statistical significances. the vesicles contained only GluA2, whereas 51% of the vesicles contained only γ2 (Fig. 7 D and E). Statistical analysis showed that GluA2 and γ2 are independently distributed on these vesicles without a significant colocalization (P > 0.05 compared with the null hypothesis that GluA2-containing vesicles and γ2-containing vesicles are independent vesicle populations; see SI Materials and Methods for details). Taken together, this in vivo result further supported the vesicular sorting model that AMPA and GABAA receptors are sorted into different intracellular trafficking vesicles before exocytosis. Discussion AMPA and GABAA receptors are selectively targeted to excitatory and inhibitory synapses (3), respectively. However, it is not clear when and how AMPA and GABAA receptors are sorted and trafficked into their target zones. To investigate this important question, we performed live TIRF imaging to directly visualize the constitutive exocytic vesicles of AMPA and GABAA receptors. In combination with immunocytochemistry, electrophysiology, and electron microscopy studies, we found that the exocytic sorting of these two receptor types follows the “vesicle sorting model” (Fig. 7F). AMPA and GABAA are initially sorted into different vesicles in the Golgi apparatus. The majority of GABAA receptors are directly delivered to the plasma membrane through the de novo exocytic pathway under the regulation of Rab8. The SNAP23–syntaxin1–VAMP2 complex mediates the fusion of GABAA receptorcontaining vesicle to the plasma membrane. However, exocytosis of AMPA receptors includes not only the Rab8-mediated de novo pathway but also the recycling pathway regulated by Rab4, 5, and 11. The fusion between AMPA receptor-containing vesicle and the plasma membrane is mediated by the SNAP25–syntaxin1–VAMP2 complex. In addition, we observed that vesicles containing AMPA receptors preferentially insert in the central region of the soma, whereas vesicles containing GABAA receptors preferentially insert in the periphery of the soma. This result was surprising and indicated that AMPA and GABAA receptors are not only differentially sorted into distinct vesicles but also targeted to distinct zones of the somatic plasma membrane during exocytosis. This sorting of the major excitatory and inhibitory receptors in the somatodendritic region is reminiscent of the polarized trafficking of apical versus basolateral proteins in epithelial cells (45, 46) and somatodendritic versus axonal proteins in neurons (47, 48), which involves vesicular sorting in TGN and endosomes. Previous studies and our current research suggest a general strategy that proteins that function at different subdomains of the cell are sorted early into separate vesicle populations. This early sorting maximally ensures the independent targeting and regulation of each protein. Fig. 7. Endogenous AMPA and GABAA receptors are sorted into different vesicles. (A) Subcellular fractionation of adult rat brain. Each fraction was normalized based on protein concentration. H, whole brain homogenate; P1, cell debris and nuclei; P2, washed synaptosomal fraction; P3, microsomal pellet; S1, postnuclear supernatant; S2, postsynaptosomal fraction; S3, soluble protein fraction. (B and C) Morphology of vesicles inP3 fraction under EM. (Scale bars: B, 500 nm; C, 100 nm.) (D) Double-immunogold EM of GluA2 and γ2 in P3 sections. GluA2 and γ2 were labeled by 6-nm (arrows) and 12-nm (arrow heads) immunogold beads, respectively. (Scale bar: 100 nm.) (E ) Quantification of vesicles containing GluA2 or γ2 observed under double-immunogold EM. Green: GluA2-only vesicles (37% of all vesicles). Magenta: γ2-only vesicles (51%). Purple: GluA2 and γ2-containing vesicles (12%). n = 73. (F) Vesicular sorting model for constitutive exocytosis of AMPA and GABAA receptors. 8 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1525726113 Gu et al. Gu et al. Methods Animal Use. All animal experiments were performed with approval by the Animal Care and Use Committee at Johns Hopkins University School of Medicine. Fusion Constructs. pHluorin-, EGFP-, and tdTomato-GluA2 were constructed in pcDNA3.1 hygro- vector by inserting the fluorescent proteins between Asn25 and Ser26 amino acids of rat GluA2 (flip). pHluorin-, EGFP-, and tdTomato-γ2S were constructed in pcDNA3.1 hygro- vector by inserting the fluorescent proteins between Asp42 and Asp43 amino acids of mouse γ2S. Dual-TIRFM Imaging. An Olympus IX71 microscope with a plan-Apo objective (100×, N.A. 1.45, oil; Olympus) was used for dual-TIRFM imaging with 488-nm and 568-nm excitation lasers. See SI Materials and Methods for extended details. PNAS Early Edition | 9 of 10 PNAS PLUS did not change in SNAP23+/− neurons (13). However, knockdown of SNAP23 by lentiviral-mediated shRNA expression only induced a modest reduction of surface AMPA receptors and no significant change of surface AMPA receptor levels was detected in SNAP23+/− mice (13), supporting our conclusion that SNAP23 is not required for AMPA receptor exocytosis. However, AMPA receptor surface level was not affected by knockdown of SNAP25 expression with lentiviral-mediated shRNA (13). No postsynaptic defects were detected after application of glutamate agonists in SNAP25-deficient neurons (53). In addition, it has been shown that Botox B rapidly reduced the amplitude of basal AMPA receptor-mediated EPSCs (7) and VAMP2 is required for constitutive delivery of AMPA receptors to the plasma membrane (16), consistent with our observation. In contrast, it has been shown that Botox B application had no effect on basal excitatory synaptic transmission (5). Tetanus toxin, which also cleaves VAMP2 and other Botox B-sensitive VAMPs (19), did not change amplitude of basal AMPA mEPSCs (6). What could be responsible for these contradictory results? First, most previous studies have not directly measured exocytic events. The surface receptor levels or synaptic current amplitude reflect the effects of multiple trafficking steps, including receptor exocytosis, endocytosis, lateral diffusion, and stabilization. So, it is critical to investigate roles of a certain molecule while isolating a particular trafficking event, as we have done here using TIRFM to specifically isolate exocytosis. Second, perturbation of trafficking events by genetic ablation and lentiviral-mediated knockdown of particular genes could induce compensatory expression of other mechanistically related proteins. For this reason, we have used acute neurotoxin treatments and short-term shRNA-mediated knockdown to complement each other. Third, it has been shown that surface levels of postsynaptic receptors, especially AMPA receptors, are regulated by long-lasting homeostatic changes in global neuronal activity, so called “synaptic scaling” (54). SNARE complexes are critical for presynaptic neurotransmitter release (55), and knockdown of particular SNARE components by genetic or virus-based shRNA approaches could possibly modulate neuronal activity in the whole preparation and indirectly affect postsynaptic receptors. Therefore, disruption of SNARE proteins at the single-cell level by sparse transfection of shRNAs, as we have done here, is more reliable when studying SNARE function in postsynaptic receptor trafficking to demonstrate that effects are cell-autonomous and independent of network activity. In summary, by directly studying the constitutive exocytosis of AMPA and GABAA receptors, we found that the segregation of AMPA and GABAA receptors occurs early during intracellular vesicle trafficking. AMPA or GABAA receptor-containing vesicles are sorted in the Golgi and exit via distinct exocytic vesicles. AMPA receptors are highly targeted to recycling pathways, whereas GABAA receptors are not. Moreover, these distinct exocytic events occur in different regions of the cell surface. AMPA and GABAA receptor exocytic events share certain properties but are also distinct in several aspects and are differentially regulated by specific SNARE complexes and Rab proteins. These results demonstrate the neuron’s capacity to elaborately sort different postsynaptic receptors to regulate excitatory and inhibitory transmission. NEUROSCIENCE Moreover, it is surprising that AMPA and GABAA receptors are delivered into distinct domains of the somatic membrane. Our data suggest the vesicles targeted at the central and peripheral regions of the soma originate from endocytic pathways and de novo exocytic pathways, respectively. This phenomenon indicates the presence of specialized zones on the plasma membrane for different exocytic pathways. Why would neurons deliver AMPA and GABAA receptors to different locations and through different pathways on the cell soma? In hippocampal neurons inhibitory synapses are often localized on proximal dendrites and the soma, whereas excitatory synapses are distributed both at proximal and distal dendrites (3, 49). The direct exocytosis of GABAA receptors to the peripheral somatic membrane would place the receptors near the location of inhibitory synapses. However, many AMPA receptors have to travel long distances to reach excitatory synapses on distal dendrites. The high level of constitutive exocytosis of AMPA receptors in the cell soma suggests that lateral diffusion of AMPA receptors from the somatic cell surface to proximal and possibly distal dendrites may play a significant role in maintaining surface and synaptic AMPA receptors. Consistent with this interpretation, a previous study had suggested that endogenous AMPA receptors are mostly exocytosed and recycled at extrasynaptic somatic sites (50). In addition, AMPA receptors may also be delivered through other trafficking pathways. For example, the transport of AMPA receptor-containing vesicles along microtubules certainly delivers AMPA receptors out to distal dendrites for local exocytosis into the extrasynaptic dendritic plasma membrane. It is possible that AMPA receptor containing recycling vesicles preferentially travel along microtubules assisting in the peripheral delivery of the receptors. Moreover, local translation of AMPA receptors subunits in dendrites will also likely play a role in the delivery of AMPA receptors to distal dendrites (1). The constitutive exocytic events characterized here are distinct from previously reported activity-dependent AMPA receptors exocytic events from our laboratory and others (14, 15, 22). Those events for GluA1 and GluA2 homomers and heteromers are brighter and occur much less frequently and have slower kinetics (14, 15). The brighter, long-lasting GluA2 events are moderately regulated by neuronal activity and require the binding of NSF and RNA editing of Q/R site in GluA2 (15). The GluA1 events are significantly regulated by neuronal activity, as well as the binding of the 4.1N protein to GluA1 and by phosphorylation and palmitoylation of GluA1 (14). These brighter and slower events of pH-GluA1 contain around 50 receptor subunits (22). In contrast, we discovered constitutive exocytic events of GluA2 and γ2, which transiently occur at higher frequency and contain fewer than 10 receptor subunits per vesicle. These observations together suggest that activity-dependent and constitutive exocytic events originate from different vesicle populations with distinct properties. However, these two types of exocytosis share a common feature, which is that they both target extrasynaptic sites on the somatic membrane and dendritic shafts. Following the initial extrasynaptic exocytosis, the specialized synaptic clustering of AMPA and GABAA receptors is finally achieved by lateral diffusion of receptors from extrasynaptic pools to the synaptic membrane and stabilization of the receptors on specific postsynaptic membranes by scaffolding proteins (51). The exocytic events of GluA1 have been observed in spines when neuronal activity is stimulated (11, 52) but we and others have rarely observed spine exocytosis even in active neuronal cultures (10, 14, 15, 22). The roles of SNARE complexes on constitutive trafficking or basal surface level of GABAA and AMPA receptors have been reported in many studies. However, the results are not fully consistent. In terms of GABAA receptor, slices from SNAP25 null animals showed an up-regulation of postsynaptic surface GABAA receptors (8), suggesting that SNAP25 is not necessary in GABAA receptor exocytosis and is in agreement with our results. Conversely, the surface and total levels of GABAA receptor α1 subunit ACKNOWLEDGMENTS. We thank Dr. Carolyn Machamer for her valuable advice on experiments and critical reading of the manuscript. We thank members of the R.L.H. laboratory for constructive comments during the execution of this study, Benjamin Lin for developing the ImageSplice software for TIRFM imaging processing, Dr. John Goutsias for advice on TIRFM imaging analysis, Dr. José Esteban for providing cDNAs of Rab dominant negative mutants, Dr. Ann Hubbard for valuable discussion on experiments, Drs. Victor Anggono, Gareth M. Thomas, and Lenora Volk for critical reading of the manuscript, Yingying Wei and Dr. Hongkai Ji for statistical analysis, and Barbara Smith for the EM experiments. TIRF microscope and relevant technical assistance were provided by the Johns Hopkins University School of Medicine Microscope Facility. This research was supported by NIH Grants R01 NS036715 and R01 MH64856. 1. Shepherd JD, Huganir RL (2007) The cell biology of synaptic plasticity: AMPA receptor trafficking. Annu Rev Cell Dev Biol 23:613–643. 2. Jacob TC, Moss SJ, Jurd R (2008) GABA(A) receptor trafficking and its role in the dynamic modulation of neuronal inhibition. Nat Rev Neurosci 9(5):331–343. 3. Craig AM, Blackstone CD, Huganir RL, Banker G (1994) Selective clustering of glutamate and gamma-aminobutyric acid receptors opposite terminals releasing the corresponding neurotransmitters. Proc Natl Acad Sci USA 91(26):12373–12377. 4. Maletic-Savatic M, Koothan T, Malinow R (1998) Calcium-evoked dendritic exocytosis in cultured hippocampal neurons. Part II: Mediation by calcium/ calmodulin-dependent protein kinase II. J Neurosci 18(17):6814–6821. 5. Lledo PM, Zhang X, Südhof TC, Malenka RC, Nicoll RA (1998) Postsynaptic membrane fusion and long-term potentiation. Science 279(5349):399–403. 6. Lu W, et al. (2001) Activation of synaptic NMDA receptors induces membrane insertion of new AMPA receptors and LTP in cultured hippocampal neurons. Neuron 29(1):243–254. 7. Lüscher C, et al. (1999) Role of AMPA receptor cycling in synaptic transmission and plasticity. Neuron 24(3):649–658. 8. Tafoya LC, et al. (2006) Expression and function of SNAP-25 as a universal SNARE component in GABAergic neurons. J Neurosci 26(30):7826–7838. 9. Kopec CD, Real E, Kessels HW, Malinow R (2007) GluR1 links structural and functional plasticity at excitatory synapses. J Neurosci 27(50):13706–13718. 10. Makino H, Malinow R (2009) AMPA receptor incorporation into synapses during LTP: The role of lateral movement and exocytosis. Neuron 64(3):381–390. 11. Kennedy MJ, Davison IG, Robinson CG, Ehlers MD (2010) Syntaxin-4 defines a domain for activity-dependent exocytosis in dendritic spines. Cell 141(3):524–535. 12. Park M, Penick EC, Edwards JG, Kauer JA, Ehlers MD (2004) Recycling endosomes supply AMPA receptors for LTP. Science 305(5692):1972–1975. 13. Suh YH, et al. (2010) A neuronal role for SNAP-23 in postsynaptic glutamate receptor trafficking. Nat Neurosci 13(3):338–343. 14. Lin DT, et al. (2009) Regulation of AMPA receptor extrasynaptic insertion by 4.1N, phosphorylation and palmitoylation. Nat Neurosci 12(7):879–887. 15. Araki Y, Lin DT, Huganir RL (2010) Plasma membrane insertion of the AMPA receptor GluA2 subunit is regulated by NSF binding and Q/R editing of the ion pore. Proc Natl Acad Sci USA 107(24):11080–11085. 16. Jurado S, et al. (2013) LTP requires a unique postsynaptic SNARE fusion machinery. Neuron 77(3):542–558. 17. Arendt KL, et al. (2015) Retinoic acid and LTP recruit postsynaptic AMPA receptors using distinct SNARE-dependent mechanisms. Neuron 86(2):442–456. 18. Südhof TC, Rothman JE (2009) Membrane fusion: Grappling with SNARE and SM proteins. Science 323(5913):474–477. 19. Schiavo G, Matteoli M, Montecucco C (2000) Neurotoxins affecting neuroexocytosis. Physiol Rev 80(2):717–766. 20. Axelrod D (2001) Total internal reflection fluorescence microscopy in cell biology. Traffic 2(11):764–774. 21. Miesenböck G, De Angelis DA, Rothman JE (1998) Visualizing secretion and synaptic transmission with pH-sensitive green fluorescent proteins. Nature 394(6689):192–195. 22. Yudowski GA, et al. (2007) Real-time imaging of discrete exocytic events mediating surface delivery of AMPA receptors. J Neurosci 27(41):11112–11121. 23. Jacob TC, et al. (2005) Gephyrin regulates the cell surface dynamics of synaptic GABAA receptors. J Neurosci 25(45):10469–10478. 24. Sankaranarayanan S, De Angelis D, Rothman JE, Ryan TA (2000) The use of pHluorins for optical measurements of presynaptic activity. Biophys J 79(4):2199–2208. 25. Shaner NC, et al. (2004) Improved monomeric red, orange and yellow fluorescent proteins derived from Discosoma sp. red fluorescent protein. Nat Biotechnol 22(12): 1567–1572. 26. Shaner NC, Steinbach PA, Tsien RY (2005) A guide to choosing fluorescent proteins. Nat Methods 2(12):905–909. 27. Wu LG, Hamid E, Shin W, Chiang HC (2014) Exocytosis and endocytosis: Modes, functions, and coupling mechanisms. Annu Rev Physiol 76:301–331. 28. Hayashi T, et al. (1994) Synaptic vesicle membrane fusion complex: Action of clostridial neurotoxins on assembly. EMBO J 13(21):5051–5061. 29. Verderio C, et al. (2006) Entering neurons: Botulinum toxins and synaptic vesicle recycling. EMBO Rep 7(10):995–999. 30. Washbourne P, Liu XB, Jones EG, McAllister AK (2004) Cycling of NMDA receptors during trafficking in neurons before synapse formation. J Neurosci 24(38):8253–8264. 31. Pan PY, et al. (2005) SNAP-29-mediated modulation of synaptic transmission in cultured hippocampal neurons. J Biol Chem 280(27):25769–25779. 32. Su Q, Mochida S, Tian JH, Mehta R, Sheng ZH (2001) SNAP-29: A general SNARE protein that inhibits SNARE disassembly and is implicated in synaptic transmission. Proc Natl Acad Sci USA 98(24):14038–14043. 33. Teng FY, Wang Y, Tang BL (2001) The syntaxins. Genome Biol 2(11)REVIEWS3012. 34. Schiavo G, Shone CC, Bennett MK, Scheller RH, Montecucco C (1995) Botulinum neurotoxin type C cleaves a single Lys-Ala bond within the carboxyl-terminal region of syntaxins. J Biol Chem 270(18):10566–10570. 35. Araki S, et al. (1997) Inhibition of the binding of SNAP-23 to syntaxin 4 by Munc18c. Biochem Biophys Res Commun 234(1):257–262. 36. Chen D, Minger SL, Honer WG, Whiteheart SW (1999) Organization of the secretory machinery in the rodent brain: Distribution of the t-SNAREs, SNAP-25 and SNAP-23. Brain Res 831(1-2):11–24. 37. Pevsner J, et al. (1994) Specificity and regulation of a synaptic vesicle docking complex. Neuron 13(2):353–361. 38. Trimble WS, Cowan DM, Scheller RH (1988) VAMP-1: A synaptic vesicle-associated integral membrane protein. Proc Natl Acad Sci USA 85(12):4538–4542. 39. Baumert M, Maycox PR, Navone F, De Camilli P, Jahn R (1989) Synaptobrevin: An integral membrane protein of 18,000 daltons present in small synaptic vesicles of rat brain. EMBO J 8(2):379–384. 40. Trimble WS, Gray TS, Elferink LA, Wilson MC, Scheller RH (1990) Distinct patterns of expression of two VAMP genes within the rat brain. J Neurosci 10(4):1380–1387. 41. Stenmark H (2009) Rab GTPases as coordinators of vesicle traffic. Nat Rev Mol Cell Biol 10(8):513–525. 42. Matlin KS, Simons K (1983) Reduced temperature prevents transfer of a membrane glycoprotein to the cell surface but does not prevent terminal glycosylation. Cell 34(1):233–243. 43. Yeaman C, et al. (2004) Protein kinase D regulates basolateral membrane protein exit from trans-Golgi network. Nat Cell Biol 6(2):106–112. 44. Lee SH, Valtschanoff JG, Kharazia VN, Weinberg R, Sheng M (2001) Biochemical and morphological characterization of an intracellular membrane compartment containing AMPA receptors. Neuropharmacology 41(6):680–692. 45. Carmosino M, Valenti G, Caplan M, Svelto M (2010) Polarized traffic towards the cell surface: How to find the route. Biol Cell 102(2):75–91. 46. Rodriguez-Boulan E, Kreitzer G, Müsch A (2005) Organization of vesicular trafficking in epithelia. Nat Rev Mol Cell Biol 6(3):233–247. 47. Lai HC, Jan LY (2006) The distribution and targeting of neuronal voltage-gated ion channels. Nat Rev Neurosci 7(7):548–562. 48. Kennedy MJ, Ehlers MD (2006) Organelles and trafficking machinery for postsynaptic plasticity. Annu Rev Neurosci 29:325–362. 49. Megías M, Emri Z, Freund TF, Gulyás AI (2001) Total number and distribution of inhibitory and excitatory synapses on hippocampal CA1 pyramidal cells. Neuroscience 102(3):527–540. 50. Adesnik H, Nicoll RA, England PM (2005) Photoinactivation of native AMPA receptors reveals their real-time trafficking. Neuron 48(6):977–985. 51. Opazo P, Choquet D (2011) A three-step model for the synaptic recruitment of AMPA receptors. Mol Cell Neurosci 46(1):1–8. 52. Patterson MA, Szatmari EM, Yasuda R (2010) AMPA receptors are exocytosed in stimulated spines and adjacent dendrites in a Ras-ERK-dependent manner during long-term potentiation. Proc Natl Acad Sci USA 107(36):15951–15956. 53. Washbourne P, et al. (2002) Genetic ablation of the t-SNARE SNAP-25 distinguishes mechanisms of neuroexocytosis. Nat Neurosci 5(1):19–26. 54. Turrigiano GG (2008) The self-tuning neuron: Synaptic scaling of excitatory synapses. Cell 135(3):422–435. 55. Lin RC, Scheller RH (2000) Mechanisms of synaptic vesicle exocytosis. Annu Rev Cell Dev Biol 16:19–49. 56. Fu J, Naren AP, Gao X, Ahmmed GU, Malik AB (2005) Protease-activated receptor-1 activation of endothelial cells induces protein kinase Calpha-dependent phosphorylation of syntaxin 4 and Munc18c: Role in signaling p-selectin expression. J Biol Chem 280(5):3178–3184. 57. Wu PH, Arce SH, Burney PR, Tseng Y (2009) A novel approach to high accuracy of video-based microrheology. Biophys J 96(12):5103–5111. 58. Wu PH, et al. (2012) High-throughput ballistic injection nanorheology to measure cell mechanics. Nat Protoc 7(1):155–170. 10 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1525726113 Gu et al. Supporting Information Gu et al. 10.1073/pnas.1525726113 SI Materials and Methods Primary Neuron Culture. Rat hippocampi from day-18 embryos were seeded on poly-L-lysine precoated 25-mm (for dual-TIRFM imaging) or 18-mm (for immunocytochemistry) coverslips in Neurobasal media (Invitrogen) containing 50 U/mL penicillin, 50 μg/mL streptomycin, and 2 mM glutamax supplemented with 2% (vol/vol) B27 and 5% (vol/vol) FBS (plating medium). Media was replaced 24 h after plating with feeding medium (plating medium without serum) and neurons were fed twice a week thereafter. Rat cortical neurons were prepared similarly and seeded on poly-L-lysine precoated 12-well plates for electroporation. Neurons were grown at 37 °C and 5% (vol/vol) CO2/95% (vol/vol) air. Transfection and Electroporation. Hippocampal neurons were transfected with Lipofectamine 2000 (Invitrogen) at DIV (day in vitro) 11– 14 and were imaged or fixed 24–72 h posttransfection. Electroporation of cortical neurons was performed at DIV 0 using Amaxa Nucleofector kit (VPG-1003; Lonza) and Nucleofector device (Lonza). Three micrograms of pSuper-shRNAs were electroporated into 3 × 106 cortical neurons. The electroporated neurons were seeded after electroporation and harvested 4 d postelectroporation. shRNAs. To generate shRNAs targeting individual rat SNAPs and syntaxins and VAMPs, pSuper was used as a vector for shRNAs. Oligos were annealed for direct subcloning into pSuper between BglII and HindIII sites (for scramble, SNAP23, SNAP29, and SNAP47 shRNAs) or between BglII and XhoI sites (for SNAP25, syntaxin1A, syntaxin1B, syntaxin4, VAMP1, and VAMP2 shRNAs). The shRNA sequences against SNAP23 and 47 were designed based on siDESGN online tool (Dharmacon). The oligo sequences targeting SNAP23 were 5′-GAT CCC CGG ATA TGG GCA ATG AAA TTT TCA AGA GAA ATT TCA TTG CCC ATA TCC TTT TTA-3′ (sense) and 5′-AGC TTA AAA AGG ATA TGG GCA ATG AAA TTT CTC TTG AAA ATT TCA TTG CCC ATA TCC GGG-3′ (antisense). The oligo sequences targeting SNAP47 were 5′-GAT CCC CAG GAA GAT GTT GAT GAT ATT TCA AGA GAA TAT CAT CAA CAT CTT CCT TTT TTA-3′ (sense) and 5′-AGC TTA AAA AAG GAA GAT GTT GAT GAT ATT CTC TTG AAA TAT CAT CAA CAT CTT CCT GGG-3′ (antisense). The shRNA sequences against SNAP29 were previously published by Pan et al. (31): 5′ GAT CCC CGT GGA CAA GTT AGA TGT CAA TTT CAA GAG AAT TGA CAT CTA ACT TGT CCA CTT TTT A-3′ (sense) and 5′-AGC TTA AAA AGT GGA CAA GTT AGA TGT CAA TTC TCT TGA AAT TGA CAT CTA ACT TGT CCA CGG G-3′ (antisense). The oligo sequences targeting SNAP25, syntaxin1A, and syntaxin1B were obtained from ONTARGET plus siRNAs of Dharmacon. The shRNA sequences against SNAP25 (catalog no. J-093289-11) were 5′-GAT CCC CGG CTT CAT CCG CAG GGT AAT TCA AGA GAT TAC CCT GCG GAT GAA GCC TTT TTC-3′ (sense) and 5′-TCG AGA AAA AGG CTT CAT CCG CAG GGT AAT CTC TTG AAT TAC CCT GCG GAT GAA GCC GGG-3′ (antisense). The shRNA sequences against syntaxin1A (catalog no. J-08917211) were 5′-GAT CCC CCA CCA AAG GTC TCG GTA CAT TCA AGA GAT GTA CCG AGA CCT TTG GTG TTT TTC-3′ (sense) and 5′-TCG AGA AAA ACA CCA AAG GTC TCG GTA CAT CTC TTG AAT GTA CCG AGA CCT TTG GTG GGG-3′ (antisense). The shRNA sequence against syntaxin1B (catalog no. J-090348-12) were 5′-GAT CCC CCG GTC CAA GTT GAA AGC GAT TCA AGA GAT CGC TTT CAA CTT GGA CCG TTT TTC-3′ (sense) and 5′-TCG AGA AAA ACG Gu et al. www.pnas.org/cgi/content/short/1525726113 GTC CAA GTT GAA AGC GAT CTC TTG AAT CGC TTT CAA CTT GGA CCG GGG-3′ (antisense). The shRNA sequences against syntaxin4 were designed based on the previously published shRNA sequences targeting human syntaxin4 (56): 5′-GAT CCC CAA GGA AGA AGC TGA TGA GAA TTT CAA GAG AAT TCT CAT CAG CTT CTT CCT TTT TTT C-3′ (sense) and 5′-TCG AGA AAA AAA GGA AGA AGC TGA TGA GAA TTC TCT TGA AAT TCT CAT CAG CTT CTT CCT TGG G-3′ (antisense). The shRNA sequence against VAMP1 (catalog no. J-091077-09) were 5′ GAT CCC CGA CCA GTA ACA GAC GAT TAT TCA AGA GAT AAT CGT CTG TTA CTG GTC TTT TTC 3′ (sense) and 5′ TCG AGA AAA AGA CCA GTA ACA GAC GAT TAT CTC TTG AAT AAT CGT CTG TTA CTG GTC GGG 3′ (antisense). The shRNA sequences of VAMP2 (catalog no. J-090962-09) were 5′ GAT CCC CAC CAG AAG CTA TCG GAA CTT TCA AGA GAA GTT CCG ATA GCT TCT GGT TTT TTC 3′ (sense) and 5′ TCG AGA AAA AAC CAG AAG CTA TCG GAA CTT CTC TTG AAA GTT CCG ATA GCT TCT GGT GGG 3′ (antisense). The scramble shRNA sequences are 5′-GAT CCC CGC GCG CTT TGT AGG ATT CGT TCA AGA GAC GAA TCC TAC AAA GCG CGC TTT TTA-3′ (sense) and 5′-AGC TTA AAA AGC GCG CTT TGT AGG ATT CGT CTC TTG AAC GAA TCC TAC AAA GCG CGC GGG-3′ (antisense). Pharmacology. Chemicals and botulinum type A toxin were purchased from Sigma-Aldrich. TTX and dynasore were purchased from Tocris Bioscience. Botulinum type B and C toxins were purchased from METAbiologics. For Botox A treatment, Botox A was originally dissolved in a solution containing 20 mM Hepes, 1.25% (wt/vol) lactose, and 1 mg/mL BSA. Prior applying to neurons, Botox A was incubated in ACSF containing 5 mM DTT for 30 min to activate the toxin; 260 nM Botox A was applied to hippocampal neurons in media for 1 h and the cells were then imaged under TIRFM. The solvent was applied to the neurons in the control group. For Botox C treatment, 100 nM Botox B or 100 nM Botox C was applied to hippocampal neurons in media for 6 h and then imaged under TIRFM. The solvent (PBS) was applied in the media of the control group. Dual-TIRFM Imaging. An Olympus IX71 microscope with a planApo objective (100×, N.A. 1.45, oil; Olympus) was used for dualTIRFM imaging. The excitation laser was a 1-W Kr/Ar TIRF laser with launch (SP_2018; Prairie Technologies). The fluorescent signal of pHlourin was imaged with a 488-nm excitation laser (1.7 mW to the back aperture of the objective) and collected through a 525- to 550-nm emission filter, and the fluorescent signal of tdTomato was imaged with a 568-nm laser (2.8 mW to the back aperture of the objective) and a 605- to 655-nm emission filter. Images were captured by a 512 × 512 back-illuminated CCD camera with on-chip multiplication gain and 16- × 16-μm pixels (Cascade 512B; Photometrics) and MetaMorph acquisition software (Molecular Devices) at a rate of 1.4 Hz and 500-ms exposure (except some data in Fig. S1B: 100-ms exposure at 10 Hz). Neurons between the ages of DIV 13 and 16 were used for imaging. All imaging experiments were carried out in artificial cerebrospinal fluid [ACSF, 120 mM NaCl, 5 mM KCl, 2 mM CaCl2, 2 mM MgCl2, 25 mM Hepes (pH 7.4), and 30 mM glucose]. The imaging temperature was maintained at 35–37 °C by a temperature controller (TC-202A; Harvard Apparatus). To increase the signal-to-noise 1 of 21 ratio, we photobleached the fluorescent signals for 20 s with maximal laser power under TIRF mode before data acquisition. TIRF Data Analysis. Insertion kinetics of pHluorin-tagged receptors were analyzed by custom-made automatic software. TIRF images of pH-GluA2 and pH-γ2S from a time-lapsed video were first filtered through a low-pass filter (57, 58) to enhance signal-tonoise ratio for the detection of event objects. Then a 2D Gaussian distribution was used to fit the local intensity distribution of each regional brightest pixel in the filtered image. A threshold value for peak intensity and R-squared value of the fits were set manually. All of the image processing and analysis were performed using custom-made software developed in MATLAB. The distributions of pHluorin- or tdTomato-tagged receptor insertions were analyzed by ImageJ software (National Institutes of Health). For each cell the image stacks from green and red channels were properly aligned with fluorescent beads (TetraSpeck Fluorescent Microspheres Size Kit; Molecular Probes). We cropped a same rectangular region crossing the somatic area in green or red channels. Insertion events in the region were isolated by subtraction of the sequential images 2–31 (acquired in 0.7– 21.7 s) by the series of images from 1 to 30 (acquired in 0–21 s). Insertions were then counted by ImageJ under appropriate thresholds and normalized by the largest number of insertions obtained along the axis. Normalized numbers of events were plotted against the length of the soma. The total length of the soma is normalized to 1. Insertion frequencies of pHluorin-tagged receptors were analyzed by combining DiaTrack 3.03 software (Vallotton Semasopht), ImageJ (National Institutes of Health), and a custom-developed software program, ImageSplice. Insertion events were generated by subtraction of the sequential images 2–11 (acquired in 0.7–7.7 s) by the series of images from 1 to 10 (acquired in 0–7 s) (ImageSplice). The number of insertions of each subtracted image from each channel is detected and counted by DiaTrack. The number of insertions was normalized to the surface area of the cell including somatic membrane and proximal dendritic membrane (ImageJ). Single-Molecule Analysis. The numbers of pH-GluA2 or pH-γ2S in individual exocytic vesicles were estimated as previously described (22). Insertions of pH-GluA2 and pH-γ2S in transfected hippocampal neurons were imaged as described above. Under the same condition, fluorescent signals of purified EGFP monomer absorbed to poly-L-lysine–coated coverslips were imaged under TIRFM. Single molecules of EGFP were identified by the blinking and single step photobleaching characters. The intensity of all of the fluorescent events, including pH-GluA2, pH-γ2S, and EGFP, were background-subtracted, plotted, and fitted by Gaussian curve (OriginPro 8; OriginLab). The intensity corresponding to the peak of the curve was used to estimate the number of receptors in vesicles. Live-Cell Imaging of Fluorescently Tagged Receptor Expression and Post-Golgi Trafficking. Expressions of EGFP-GluA2, tdt-GluA2, EGFP-γ2S, and tdt-γ2S in live neurons were imaged on a Zeiss LSM510 using a 63× objective (N.A. 1.40) (Carl Zeiss MicroImaging Group Inc.). To image post-Golgi trafficking in live neurons, neurons that were seeded on coverslips were transfected with different combinations of EGFP-GluA2, tdt-GluA2, EGFP-γ2S, and tdt-γ2S on DIV 11. The transfected neurons are incubated at 37 °C for 2 h to allow protein expression. Cells were then transferred to 25 mM Hepes (pH 7.4)-buffered Neurobasal medium (Invitrogen) and incubated in 20 °C water bath for 5 h to allow EGFP- or tdt-tagged receptors to traffic to and accumulate in the Golgi apparatus. Cells were imaged and maintained at 32 °C using a temperature controller (TC-202A; Harvard Apparatus) in ACSF; 20 °C incubations and live-cell imaging were performed in the Gu et al. www.pnas.org/cgi/content/short/1525726113 presence of cycloheximide (20 μg/mL; Sigma). Live-cell imaging experiments were performed on an Olympus IX71 microscope with a plan-Apo objective (100×, N.A. 1.45, oil; Olympus). The excitation laser was a 1-W Kr/Ar TIRF laser with launch (SP_2018; Prairie Technologies). The fluorescent signal of GFP was imaged with a 488-nm excitation laser and collected through a 525- to 550-nm emission filter, and the fluorescent signal of tdTomato was imaged with a 568-nm excitation laser and a 605- to 655-nm emission filter. To image intracellular trafficking of EGFP- or tdTomato-tagged receptors, the angle of lasers was adjusted to epi-fluorescent mode. Images were captured by a 512 × 512 back-illuminated CCD camera with on-chip multiplication gain and 16- × 16-μm pixels (Cascade 512B; Photometrics) and MetaMorph acquisition software (Molecular Devices) at a rate of 1.4 Hz and 500-ms exposure. Immunocytochemical Experiments. Surface expression of tdt-GluA2 and tdt-γ2S (Fig. S3) was labeled by incubating live hippocampal neurons in ACSF containing 10% (wt/vol) BSA for 20 min at 10 °C with anti-DsRed antibody (Clontech), which also recognizes tdTomato. Neurons were then washed twice with ice-cold ACSF containing 10% (wt/vol) BSA for washout of the excess antibodies and then fixed with 4% (vol/vol) paraformaldehyde and 4% (wt/vol) sucrose in ice-cold PBS at room temperature for 15 min. Alexa568-conjugated secondary antibodies (Invitrogen) were applied following fixation to visualize the surface staining of DsRed antibody [in PBS containing 10% (wt/vol) BSA, room temperature, 1 h]. Neurons were then permeabilized with PBS containing 0.25% Triton X-100 at room temperature for 20 min. Coverslips were blocked in 10% (wt/vol) BSA in PBS at room temperature for 1 h and primary antibodies for synaptic marker proteins were then applied for 2 h: PSD95 (clone 6G6; Millipore), VGluT (Millipore), gephyrin (Synaptic Systems), or VGAT (Synaptic Systems). Alexa647-conjugated secondary antibodies (Invitrogen) were applied for 1 h to visualize the staining of these synaptic markers. Total expressions of tdTomato-tagged receptors were represented by natural fluorescence signals of the receptors. Endogenous surface γ2 was labeled following the similar protocol using anti-γ2 N terminus antibody (Alomone Labs). Owing to the low affinity of the antibody, endogenous surface GluA2 was labeled by incubating the anti-GluA2 N terminus antibody (MAB397; Millipore) with live neurons at room temperature for 30 min. To detect the expressions of tdt-GluA2, tdt-γ2S, endogenous GluA1, and β2/3 (Fig. S2 B and E), dissociated hippocampal neurons were fixed and permeabilized as previously described. Expressions of tdt-GluA2 and tdt-γ2S were detected by antiDsRed antibody. GluA1 was detected by anti-GluA1 N terminus antibody (clone 007, 4.9D; R.L.H. laboratory). β2/3 was detected by anti-β2/3 N terminus antibody (Millipore). The accumulation of EGFP-GluA2, tdt-GluA2, EGFP-γ2S, or tdt-γ2S in the Golgi apparatus after the 20 °C incubation was detected by staining of EGFP-GluA2 and EGFP-γ2S with anti-GFP antibody (JH4030, made in laboratory), or staining of tdt-GluA2 and tdt-γ2S with antiDsRed antibody. GM130 was stained with anti-GM130 antibody (BD Biosciences). All fluorescence images, including live neurons and fixed immunostaining samples, were acquired on a Zeiss LSM510 using a 63× objective (N.A. 1.40) (Carl Zeiss MicroImaging Group Inc.). Fluorescent intensities were quantified using ImageJ (National Institutes of Health). Electrophysiology. To examine functions of SNAP23 and 25 in syn- aptic transmission, spontaneous AMPA receptor-mediated mEPSC or GABA receptor-mediated mIPSC were measured in cultured hippocampal neurons on DIV 13–17, 3 d after transfection of shRNAs specifically against SNARE23, SNARE25, or scramble control. Neurons were perfused in Hepes-buffered extracellular solution (143 mM NaCl, 5 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 10 mM Hepes, 10 mM glucose, pH 7.2, osmolality 305–310 mOsm) 2 of 21 in the presence of 1 μM TTX and 100 μM DL-2-amino-5-phosphonopentanoic acid. Whole-cell recording pipettes (4–8 MOhm) were filled with internal solution I (115 mM Cs-MeSO4, 0.4 mM EGTA, 5 mM TEA-Cl, 2.8 mM NaCl, 20 mM Hepes, 3 mM Mg-ATP, 0.5 mM Na2-GTP, pH 7.2, osmolality 295–300 mOsm) containing 100 μM picrotoxin for mEPSC recording or internal solution II (120 mM CsCl, 11 mM EGTA, 5 mM TEA-Cl, 0.1 mM CaCl2, 20 mM Hepes, 3 mM Mg-ATP, 0.5 mM Na2-GTP) containing 10 μM NBQX for mIPSC recording. Cells were held at −70 mV holding potential and recording was performed at room temperature. Upon entering whole-cell mode, we allowed 5 min for dialysis of the intracellular solution before collecting data. Signals were measured with a MultiClamp 700B amplifier and digitized using a Digidata 1440A analog-to-digital board. Data acquisition was performed with pClamp 10.2 software and digitized at 10 kHz. mEPSCs or mIPSCs were detected with a template matching algorithm in Clampfit 10.2 software. All equipment and software are from Axon Instruments/Molecular Devices. Averaged mEPSC and mIPSC amplitude and frequency were calculated by collecting 2–4 min of recording for at least 100 events for each cell. All data are presented as mean ± SEM. Statistical significance was calculated using a Mann–Whitney test. Western Blot and Antibodies. Cortical neurons were harvested 4 d after electroporation (DIV 4) in 60 μL SDS lysis buffer (1% SDS, 1 mM EDTA, and protease inhibitor mixture) and further diluted in 300 μL dilution buffer (1% TritonX-100, 1 mM EDTA, 1 mM EGTA, and protease inhibitor mixture). Endogenous expression of SNAP23, SNAP29, SNAP47, syntaxin1A, syntaxin1B, syntaxin4, and VAMP2 was detected by antibodies specific to each protein (Synaptic Systems). Endogenous expression of SNAP25 was detected by anti-SNAP25 antibody (clone SMI-81R; Covance). Endogenous expression of VAMP1 was detected by anti-VAMP1 antibody (Novus). Subcellular Fractionation. Cerebral cortex from one adult male Sprague–Dawley rat (∼P60) was homogenized and fractionated as previously described (44). The brain was weighed and homogenized in 10 volumes of homogenization buffer (0.32 M sucrose, 4 mM Hepes, pH 7.4, and protease inhibitors) by 30 strokes in a Teflon-glass homogenizer (whole-brain homogenate, H). The H fraction was centrifuged for 10 min at 1,100 × g to remove cell debris and nuclei (pellet 1, P1). The supernatant (S1) was centrifuged for 10 min at 9,200 × g and the pellet (P2) was removed. Supernatant from this centrifugation (S2) was further centrifuged for 2 h to obtain the microsomal pellet (P3) and soluble fraction (S3). Each fraction was normalized based on their protein concentrations before being analyzed by Western blot. Electron Microscopy. The morphological characterization of the P3 fraction was done as previously described (6). The P3 pellet was fixed in 5% (vol/vol) paraformaldehyde (freshly prepared from EM-grade prill form) + 0.1% glutaraldehyde (EM grade) + 0.1% tannic acid + 3 mM MgCl2 + 0.1 M phosphate buffer (pH 7.2– 7.4; Sorensen’s) at 4 °C overnight under gentle agitation. Following rinsing with 3% (wt/vol) sucrose + 3 mM MgCl2 + 0.1 M phosphate buffer, samples were postfixed with 1% osmium tetroxide (nonreduced) + 3 mM MgCl2 + 0.1 M phosphate buffer for 1 h and then rinsed with 3% (wt/vol) sucrose + 0.1 M phosphate buffer. The samples were immersed for 30 min with filtered 2% (wt/vol) uranyl acetate (aqueous) and dehydrated through a graded series of ethanol from 50 to 100% (vol/vol). Dehydrated samples were transferred through propylene oxide and embedded in Eponate 12 (Pella) and cured at 60 °C for 2 d. Thin sections were cut on a Riechert Ultracut E with a Diatome Diamond knife. Eighty-nanometer sections were picked up on formvar-coated 1- × 2-mm copper slot grids (Pella) and stained with uranyl acetate followed by lead citrate. Gu et al. www.pnas.org/cgi/content/short/1525726113 For the postembed immunoelectron microscopy, P3 samples were fixed in 4% (vol/vol) paraformaldehyde + 0.2% glutaraldeyde + 0.1 M Na3PO4 buffer + 3 mM MgCl2 overnight at 4 °C. Fixation and all subsequent steps up to infiltration were carried out at 4 °C followed by curing at 50 °C. After rinsing with 0.1 M phosphate + 3% (wt/vol) sucrose, samples were incubated with 0.12% tannic acid (Malinkrodt) for 1 h. Upon buffer rinse, uncross-linked glutaraldeyde was reduced for 30 min with 50 mM NH4Cl in buffer. After a 45-min rinsing in 0.1 M maleate buffer (pH 6.2), fractions were en bloc-stained in 2% (wt/vol) uranyl acetate (filtered) in maleate for 30 min. After ethanol series dehydration, samples were infiltrated with a 90% (vol/vol) ethanol: LR White (with catalyst) (2:1) mixture for 1 h. Samples were further infiltrated in 90% (vol/vol) ethanol:LR White (1:1 and 1:3) then placed in pure LR White overnight at 4 °C . After three changes in fresh LR White (4 °C), samples were polymerized in tightly sealed gelatin capsules (00) at 50 °C for 24 h. Polymerized blocks were trimmed and sectioned as described above. For immunolabeling grids were treated with 5% (wt/vol) sodium periodate (aqueous; Sigma) followed by 50 mM NH4Cl in TBS (50 mM Trizma base + 150 mM NaCl, pH 7.4). After a block in 10% (vol/vol) normal goat serum (NGS) + 10% (vol/vol) donkey serum (DS) in TBST (10 mM Tris + 500 mM NaCl + 0.05% Tween 20, pH 7.2) for 10 min, grids were incubated with primary antibodies, including antiGluA2 N terminus antibody (Millipore) at 1:5 and anti-γ2 N terminus antibody (Synaptic Systems) at 1:16–24, in 1% NGS + 1% DS in TBST overnight at 4 °C. Samples without the first antibodies served as controls. Grids were then incubated for 10 min in 1% BSA in TBST and sequentially incubated in gold-conjugated secondary antibodies (6 nm Au-goat anti-mouse, 1:40, and 12 nm Au-donkey antirabbit, 1:40; Jackson Laboratories) at room temperature for 1 h for each antibody. After rinsing in TBS then in distilled H2O grids were incubated in 2% (vol/vol) glutaraldehyde for 5 min. After rinsing grids were stained with 2% (wt/vol) filtered uranyl acetate (aqueous) and then allowed to dry. All grids were viewed with a Phillips CM 120 TEM operating at 80 kV and images were captured with an XR 80-8 Megapixel CCD camera by AMT. Quantitation of Immunogold Labeling. The areas of the sections with immunogold labeling were photographed randomly at final magnification of 65,000×. Numbers of 50- to 300-nm diameter vesicles, which were unlabeled, or labeled by 6-nm or 12-nm immunogold particles, were counted manually. Evaluation of statistical significance of colabeling of GluA2 and γ2 was done to test the null hypothesis that GluA2-containing vesicles and γ2-containing vesicles are independent vesicle populations. The observed frequencies (Oi) of four vesicle populations, including unlabeled, labeled for GluA2 only, labeled for γ2 only, and labeled for both GluA2 and γ2, were compared with the frequencies expected (Ei) if GluA2 and γ2 were distributed independently. The deviation of the observed from the expected frequencies was quantified by Pearson’s statistic (χ2 test), χ2 = 4 X ðOi − Ei Þ2 i=1 Ei . The corresponding P value was obtained from a χ2 distribution with one degree of freedom. The distribution of the observed vesicle populations is not significantly different from the null hypothesis when two-sided P > 0.05. Statistics. Student t test, Mann–Whitney test, or one-way ANOVA was used for statistical analysis and was performed using Prism software (GraphPad Software). P < 0.05 was taken as a statistically significant difference, unless otherwise noted. Values are reported as mean ± SEM, unless otherwise noted. 3 of 21 Fig. S1. Characterization of surface dynamic events of pH-GluA2 and pH- γ2S. (A) Duration of pH-GluA2 events is longer than that of pH-γ2S events. Each data point represents one cell. (B) The frequency of dynamic TIRF events of pH-γ2S is significantly higher than that of pH-GluA2 under two different imaging frequencies. The exposure time was 100 ms for 10-Hz imaging. The exposure time was 500 ms for 1.4-Hz imaging. Each data point represents one cell. (C) Frequencies of dynamic TIRF events of pH-GluA2 and pH-γ2S are sensitive to pH of extracellular solution, which is indicated in the figure. Fifty millimolar NH4Cl was used in the extracellular solution to alkalinize all of the intracellular vesicles. (D) Frequency of the dynamic events of GluA2 and γ2S is not affected by neuronal activity. One micromolar TTX or 12 mM KCl were applied to hippocampal neurons in the media for 1 h and the cells were then imaged under TIRFM. Asterisk indicates statistical significance. n.s., no statistical significance. (E, Top) Sequential images showing the kinetics of surface-absorbed EGFP monomer under single-step photobleaching visualized by TIRFM. Image interval: 0.7 s. (Scale bar: 1 μm.) (Bottom) An example of fluorescent intensity trace of EGFP monomer under single-step photobleaching. (F) Binned fluorescence intensity distribution of individual EGFP monomers (gray bars) and fitted Gaussian curve based on the distribution (purple curve). The dashed line shows the peak intensity of the Gaussian curve. n = 561. (G) Binned fluorescence intensity distribution of individual pH-GluA2–containing vesicles (gray bars) and fitted Gaussian curve based on the distribution (purple curve). The dashed line shows the peak intensity of the Gaussian curve. n = 446. (H) Binned fluorescence intensity distribution of individual pH-γ2S–containing vesicles (gray bars) and fitted Gaussian curve based on the distribution (purple curve). The dashed line shows the peak intensity of the Gaussian curve. n = 733. Gu et al. www.pnas.org/cgi/content/short/1525726113 4 of 21 Fig. S2. Characterization of tdTomato-GluA2 and tdTomato-γ2S. (A) Coexpression of EGFP-GluA2 (top left) and tdt-GluA2 (top right) in live hippocampal neurons. (Scale bar: 20 μm.) High-magnification images from the selected region are shown in bottom panels: bottom left, EGFP-GluA2; bottom middle, tdtGluA2; bottom right, overlay left and middle panels. (Scale bar: 5 μm.) (B) Colocalization of tdt-GluA2 and endogenous GluA1 in hippocampal neurons at DIV 19. Top left, total tdt-GluA2 stained with anti-DsRed antibody; top middle, endogenous GluA1 stained with anti-GluA1 antibody; top right, overlay of total tdtGluA2 and endogenous GluA1. (Scale bar: 20 μm.) High-magnification images from the selected region are shown in bottom panels. (Scale bar: 5 μm.) (C) Quantification of colocalization of tdt-GluA2 and endogenous GluA1 in B. Magenta column: 43.4 ± 4.6% of total tdt-GluA2 colocalizes with total GluA1. Green column: 75.5 ± 2.2% of synaptic tdt-GluA2 colocalizes with synaptic GluA1. (D) Similar to A with coexpressions of EGFP-γ2S and tdt-γ2S. (E) Colocalization of tdt-γ2S and endogenous β2/3 in hippocampal neurons at DIV 19. Top left, total tdt-γ2S stained with anti-DsRed antibody; top middle, endogenous β2/3 stained with anti-β2/3 antibody; top right, overlay of total tdt-γ2S and endogenous β2/3. (Scale bar: 20 μm.) High-magnification images from the selected region are shown accordingly in bottom panels. (Scale bar: 5 μm.) (F) Quantification of colocalization of tdt-γ2S and endogenous β2/3 in E. Magenta column: 51.3 ± 4.4% of total tdt-γ2S colocalizes with total β2/3. Green column: 85.4 ± 1.4% of synaptic tdt-γ2S colocalizes with synaptic β2/3. Each group of quantified data is collected from three cells and three different areas were used in each cell. Gu et al. www.pnas.org/cgi/content/short/1525726113 5 of 21 Fig. S3. Colocalizations of tdt-GluA2 with PSD95 and VGluT. (A) Immunostainings of surface tdt-GluA2 and PSD95 in hippocampal neurons at DIV 18. Top left, fluorescent signal of total tdt-GluA2 without staining; top middle, surface tdt-GluA2; top right, endogenous PSD95. (Scale bar: 20 μm.) High-magnification images from the selected region are shown in middle panels. (Scale bar: 5 μm.) Bottom panels: overlaid figures of middle figures as indicated in the figure. (B) Quantification of colocalization of tdt-GluA2 with PSD95 in A. Magenta column: 37.3 ± 3.1% of total tdt-GluA2 colocalizes with total PSD95. Green column: 31.4 ± 2.6% of surface tdt-GluA2 colocalizes with total PSD95. Stripped magenta column: 82.1 ± 2.0% of synaptic total tdt-GluA2 colocalizes with synaptic PSD95. Stripped green column: 69.4 ± 3.0% of synaptic surface tdt-GluA2 colocalizes with synaptic PSD95. (C) Immunostainings of surface tdt-GluA2 and VGluT in hippocampal neurons at DIV 18. Top left, fluorescent signal of total tdt-GluA2 without staining; top middle, surface tdt-GluA2; top right, endogenous VGluT. (Scale bar: 20 μm.) High-magnification images from the selected region are shown in middle panels. (Scale bar: 5 μm.) Bottom panels: overlaid figures of middle figures as indicated in the figure. (D) Quantification of colocalization of tdt-GluA2 with VGluT in C. Magenta column: 18.7 ± 2.8% of total tdt-GluA2 colocalizes with total VGluT. Green column: 18.3 ± 2.0% of surface tdt-GluA2 colocalizes with total VGluT. Stripped magenta column: 46.9 ± 3.1% of synaptic total tdtGluA2 colocalizes with synaptic VGluT. Stripped green column: 42.9 ± 2.0% of synaptic surface tdt-GluA2 colocalizes with synaptic VGluT. Each group of quantified data is collected from three cells and three different areas were used in each cell. Gu et al. www.pnas.org/cgi/content/short/1525726113 6 of 21 Fig. S4. Colocalizations of tdt-γ2S with gephyrin and VGAT. (A) Immunostainings of surface tdt-γ2S and gephyrin in hippocampal neurons at DIV 16. Top left, fluorescent signal of total tdt-γ2S without staining; top middle, surface tdt-γ2S; top right, endogenous gephyrin. (Scale bar: 20 μm.) High-magnification images from the selected region are shown in middle panels. (Scale bar: 5 μm.) Bottom panels: overlaid figures of middle figures as indicated in the figure. (B) Quantification of colocalization of tdt-γ2S with gephyrin in E. Magenta column: 65.0 ± 6.0% of total tdt-γ2S colocalizes with total gephyrin. Green column: 23.8 ± 3.3% of surface tdt-γ2S colocalizes with total gephyrin. Stripped magenta column: 84.6 ± 2.0% of synaptic total tdt-γ2S colocalizes with synaptic gephyrin. Stripped green column: 62.8 ± 3.7% of synaptic surface tdt-γ2S colocalizes with synaptic gephyrin. (C) Immunostainings of surface tdt-γ2S and VGAT in hippocampal neurons at DIV 16. Top left, fluorescent signal of total tdt-γ2S without staining; top middle, surface tdt-γ2S; top right, endogenous VGAT. (Scale bar: 20 μm.) High-magnification images from the selected region are shown in middle panels. (Scale bar: 5 μm.) Bottom panels: overlaid figures of middle figures as indicated in the figure. (D) Quantification of colocalization of tdt-γ2S with VGAT in E. Magenta column: 56.6 ± 7.1% of total tdt-γ2S colocalizes with total VGAT. Green column: 25.9 ± 3.8% of surface tdt-γ2S colocalizes with total VGAT. Stripped magenta column: 77.9 ± 1.8% of synaptic total tdt-γ2S colocalizes with synaptic VGAT. Stripped green column: 63.1 ± 4.2% of synaptic surface tdt-γ2S colocalizes with synaptic VGAT. Each group of quantified data is collected from three cells and three different areas were used in each cell. Gu et al. www.pnas.org/cgi/content/short/1525726113 7 of 21 Fig. S5. Cleavage of SNARE proteins by botulinum toxins. (A) Endogenous SNAP25 was cleaved by Botox A. Botox A (260 nM) was applied to hippocampal neurons in media for 1 h. The cells were then lysed and the endogenous SNAP25 level was analyzed. (B) Quantification of cleavage of SNAP25 by Botox A. Averaged protein levels of SNAP25 were normalized by the control without toxin treatment. (C) Endogenous VAMP2 was cleaved by Botox B. Botox B (100 nM) was applied to hippocampal neurons in media for 6 h. The cells were then lysed and the endogenous VAMP2 level was analyzed. (D) Quantification of cleavage of VAMP2 by Botox B. Averaged protein levels of VAMP2 were normalized by the control without toxin treatment. (E) Endogenous syntaxin1A, 1B, 2, 3, and SNAP25 were cleaved by Botox C. Botox C (100 nM) was applied to hippocampal neurons in media for 6 h. The cells were then lysed and the endogenous SNAP25, syntaxin 1A, 1B, 2, and 3 levels were analyzed. (F–J) Quantification of cleavage of syntaxin1A, 1B, 2, 3, and SNAP25 by Botox C. Averaged protein levels of each SNARE were normalized by the control without toxin treatment. Quantified data shown in B, D, and F–J are from three independent experiments. Asterisk indicates statistical significance. Gu et al. www.pnas.org/cgi/content/short/1525726113 8 of 21 Fig. S6. Lateral diffusion of pH-GluA2 or pH-γ2S following exocytosis. (A) Dynamics of an exocytic event of pH-GluA2. Note that after exocytosis, one exocytic event splits into two events, which separately diffuse away from the insertion spot. (B) Fluorescence time courses on the right panel represent averaged fluorescence changes of the pH-GluA2 event in A in selected regions shown on the left. The green trace represents the fluorescence time course in the center of the exocytic spot (green circle). The magenta trace represents the fluorescence change over time in the annulus surrounding the exocytic spot (the area between green and magenta circles). The baseline was subtracted in each trace. The maximal fluorescence was normalized as 1. (C) Averaged fluorescence time course of multiple pH-GluA2 insertion events on the exocytic spot (green) and surrounding region (magenta). Note that the magenta trace peaks when the green trace starts decaying. n = 30. (D) Peak time of fluorescence of pH-GluA2 events on the exocytic spot and surrounding region. n = 30. (E) Dynamics of an exocytic event of pH-γ2S. (F) Fluorescence time courses on the right panel represent averaged fluorescence changes of the pH-γ2S event in E in selected regions, as indicated on the left. (G) Averaged fluorescence time course of multiple pH-γ2S insertion events on the exocytic spot (green) and surrounding region (magenta). n = 30. (H) Peak time of fluorescence of pH-γ2S events on the exocytic spot and surrounding region. n = 30. Asterisk indicates statistical significance. Gu et al. www.pnas.org/cgi/content/short/1525726113 9 of 21 Fig. S7. shRNA knockdown of SNARE proteins, and effects of shRNAs of SNARE proteins on exocytosis of GluA2 and γ2S. (A, B, and D–I) shRNA knockdown of SNARE proteins. Three micrograms of shRNA specifically targeting each SNARE protein was electroporated into dissociated rat cortical neurons at DIV 0 immediately before plating. The cells were lysed 4 days after electroporation and expressions of endogenous SNARE proteins were analyzed. Tubulin was used as an internal control in all experiments. (A) SNAP23; (B) SNAP25; (D) SNAP29; (E) SNAP47; (F) syntaxin1A; (G) syntaxin1B; (H) syntaxin4; (I) VAMP1 and VAMP2. (C) Specificity of SNAP23 and SNAP25 shRNAs. shRNA specificity was tested in HEK293T cells. Myc-SNAP23 and GFP-SNAP25 were cotransfected with either SNAP23 or SNAP25 shRNA. Two days after transfection cells were lysed and expression of myc-SNAP23 and GFP-SNAP25 was determined by blotting with antimyc and anti-GFP antibodies, respectively. (J and K) SNAP25 is required for both somatic (J) and dendritic (K) exocytosis of GluA2, but not γ2S. Scramble shRNA (Scr), SNAP25 shRNA, shRNA-resistant SNAP25 (25), or SNAP23 (23) were coexpressed with pH-GluA2 or pH-γ2S in hippocampal neurons. One day after transfection, exocytosis of pH-GluA2 or pH-γ2S were imaged under TIRFM. The somatic and dendritic exocytic events were then analyzed separately. (L and M) SNAP23 is required for both somatic (L) and dendritic (M) exocytosis of γ2S, but not GluA2. Scramble shRNA (Scr), SNAP23 shRNA, shRNA-resistant SNAP23 (23), or SNAP25 (25) were coexpressed with pH-GluA2 or pH-γ2S in hippocampal neurons. Two days after transfection, exocytosis of pH-GluA2 or pH-γ2S were imaged under TIRFM. The somatic and dendritic exocytic events were then analyzed separately. (N) SNAP29 is not required for exocytosis of GluA2 and γ2S. In dissociated hippocampal neurons, scramble shRNA (Scr) or SNAP29 shRNA (29) was coexpressed with pH-GluA2 or pH-γ2S. Two days after transfection exocytosis of pH-GluA2 or pH-γ2S were imaged under TIRFM. (O) SNAP47 is not required for exocytosis of GluA2 and γ2S. In dissociated hippocampal neurons, scramble shRNA (Scr) or SNAP47 shRNA (47) was coexpressed with pH-GluA2 or pH-γ2S. Two days after transfection exocytosis of pH-GluA2 or pH-γ2S were imaged under TIRFM. (P) Syntaxin4 is not required for exocytosis of GluA2 and γ2S. In dissociated hippocampal neurons, scramble shRNA (Scr) or syntaxin4 shRNA (STX4) was coexpressed with pH-GluA2 or pH-γ2S. Two days after transfection exocytosis of pH-GluA2 or pH-γ2S were imaged under TIRFM. (Q) VAMP1 is not required for exocytosis of GluA2 and γ2S. In dissociated hippocampal neurons, scramble shRNA (Scr) or VAMP1 shRNA was coexpressed with pH-GluA2 or pH-γ2S. Three days after transfection exocytosis of pH-GluA2 or pH-γ2S were imaged under TIRFM. Asterisks indicate statistical significances. n.s., no statistical significance. Gu et al. www.pnas.org/cgi/content/short/1525726113 10 of 21 Fig. S8. Knockdown of SNAP23 and SNAP25 does not affect surface expression of endogenous GluA2 and γ2, respectively. (A) Knockdown of SNAP23 did not affect surface expression of endogenous GluA2. In dissociated hippocampal neurons, scramble shRNA, SNAP25 shRNA, or SNAP23 shRNA were coexpressed with EGFP to label transfected cells. Surface levels of endogenous GluA2 were stained 72 h after transfection. Right panels show higher-magnification images of the individual processes boxed in left panels. N, surface GluA2 in nontransfected neurons; T, surface GluA2 in transfected neurons. (Scale bar: 20 μm.) (B) Quantification of endogenous surface GluA2 levels in A. Relative GluA2 surface level of transfected neurons is represented by a ratio of surface GluA2 intensity in transfected cells to that of surrounding nontransfected cells. The ratios were further normalized by the average of the control sample transfected with the scramble shRNA. Scr: scramble shRNA. 23: shRNA-resistant SNAP23. 25: SNAP25. n = 18–22 neurons for each group. Five processes were selected from each neuron. (C) Knockdown of SNAP25 did not affect surface expression of GABAA receptor γ2 subunit. In dissociated hippocampal neurons at DIV 11, scramble shRNA, SNAP23 shRNA, or SNAP25 shRNA were coexpressed with EGFP to label transfected cells. Seventy-two hours after the transfection, endogenous γ2 on the surface was stained. Right panels show higher-magnification images of individual processes boxed in left panels. N, surface γ2 in nontransfected cells; T, surface γ2 in cells transfected with shRNAs. (D) Quantification of the surface γ2 levels in C. Relative γ2 surface level is represented by a ratio of surface γ2 intensity in transfected cells to that of surrounding nontransfected cells. The ratios were further normalized by the average of the control sample transfected with the scramble shRNA. Scr: scramble shRNA. 25: SNAP25 shRNA. 23: SNAP23 shRNA. n = 25–30 neurons for each group. Five processes were selected from each neuron. Asterisks indicate statistical significances. n.s., no statistical significance. Gu et al. www.pnas.org/cgi/content/short/1525726113 11 of 21 Fig. S9. Distributions of exocytosis of AMPA and GABAA receptors on the plasma membrane under coexpression of GluA1 or GluA3. (A) Demonstration of event detection by automated program. A raw image of pH-GluA2 from TIRF imaging was first obtained (Left). Event enhancement filtering was applied to reduce the noise (Middle). Objects with high intensity were then identified as events and their locations in the image were calculated (Right). (B) Averaged distribution of insertion events of pH-γ2S and tdt-GluA2 with coexpression of GluA1. The green and magenta curves represent the exocytic event distributions of pH-γ2S and tdt-GluA2, respectively. n = 17. (C) Averaged distribution of insertion events of pH-GluA2 and tdt-γ2S with coexpression of GluA1. The green and magenta curves represent the exocytic event distributions of pH-GluA2 and tdt-γ2S, respectively. n = 16. (D) Averaged distribution of insertion events of pH-γ2S and tdt-GluA2 with coexpression of GluA3. The green and magenta curves represent the exocytic event distributions of pH-γ2S and tdt-GluA2, respectively. n = 12. (E) Averaged distribution of the insertion events of pH-GluA2 and tdt-γ2S with coexpression of GluA3. The green and magenta curves represent the exocytic event distributions of pH-GluA2 and tdt-γ2S, respectively. n = 13. Asterisk indicates statistical significance. Gu et al. www.pnas.org/cgi/content/short/1525726113 12 of 21 Fig. S10. The trafficking of EGFP- or tdTomato-tagged GluA2 and γ2S was blocked in Golgi apparatus after 20 °C incubation, and cotrafficking of GluA2 or γ2S with different fluorescent tags when they exit the Golgi apparatus. (A) EGFP-γ2S, tdt-γ2S, EGFP-GluA2, or tdt-GluA2 were expressed in hippocampal neurons at DIV 11. After the incubation at 20 °C, EGFP-γ2S, tdt-γ2S, EGFP-GluA2, or tdt-GluA2 were stained by anti-GFP or tdTomato antibodies with a costaining of Golgi marker GM130. In the overlaid images, magenta shows the GM130 staining and green shows the staining of EGFP- or tdTomato-tagged receptors. (Scale bar: 20 μm.) (B) Time series of an example of a post-Golgi trafficking vesicle containing EGFP-γ2S and tdt-γ2S, as indicated by arrows (green, magenta, and white) at corresponding locations. For each time point, top: time-series images of EGFP-γ2S; middle: time-series images of tdt-γ2S; bottom: overlaid time-series images of top and middle panels. (Scale bar: 2.5 μm.) See Movie S6 for complete time lapse. The kymographs show the trafficking of the vesicle along its trajectory for EGFP-γ2S, tdt-γ2S, and overlaid signal. (C) Similar to B, time series of an example of a post-Golgi trafficking vesicle containing EGFP-GluA2 and tdt- GluA2, as indicated by arrows (green, magenta, and white) at corresponding locations. See Movie S6 for complete time lapse. Gu et al. www.pnas.org/cgi/content/short/1525726113 13 of 21 Movie S1. Plasma membrane insertions of pH-GluA2. pH-GluA2 was expressed in hippocampal neurons. Two days after transfection, plasma membrane insertions of pH-GluA2 under TIRFM with an excitation laser at 488 nm. The movie was taken at 500-ms exposure and 1.4-Hz acquisition frequency. Movie S1 Gu et al. www.pnas.org/cgi/content/short/1525726113 14 of 21 Movie S2. Plasma membrane insertions of pH-γ2S. pH-γ2S was expressed in hippocampal neurons. Two days after transfection, plasma membrane insertions of pH-GluA2 under TIRFM with an excitation laser at 488 nm. The movie was taken at 500-ms exposure and 1.4-Hz acquisition frequency. Movie S2 Gu et al. www.pnas.org/cgi/content/short/1525726113 15 of 21 Movie S3. Identification of insertion events of pH-GluA2 using automatic program. Insertion events of pH-GluA2 were automatically identified by a custommade program under a proper threshold. The events were marked as magenta dots. Movie S3 Gu et al. www.pnas.org/cgi/content/short/1525726113 16 of 21 Movie S4. Identification of insertion events of pH-γ2S using automatic program. Insertion events of pH-γ2S were automatically identified by a custom-made program under a proper threshold. The events were marked as magenta dots. Movie S4 Gu et al. www.pnas.org/cgi/content/short/1525726113 17 of 21 Movie S5. Plasma membrane insertions of tdt-GluA2 and pH-γ2S have different distributions. tdt-GluA2 and pH-γ2S were coexpressed in hippocampal neurons. Two days after transfection, plasma membrane insertions of tdt-GluA2 and pH-γ2S were simultaneously imaged under dual TIRFM. Left channel represents insertions of pH-γ2S, which excited by a laser at 488 nm. Right channel represents insertions of tdt-GluA2, which excited by a laser at 568 nm. The movie was taken at 500-ms exposure and 1.4-Hz acquisition frequency. Movie S5 Gu et al. www.pnas.org/cgi/content/short/1525726113 18 of 21 Movie S6. Cotrafficking of EGFP-γ2S/tdt-γ2S and EGFP-GluA2/tdt-GluA2 after exiting the Golgi apparatus. Trafficking vesicles were imaged at 32 °C at 0.7-s interval after a preincubation at 20 °C. Top panel: cotrafficking of EGFP-γ2S and tdt-γ2S. Left channel represents a trafficking of EGFP-γ2S-containing vesicle. Middle channel represents the trafficking of the same vesicle containing tdt-γ2S. Right channel shows an overlaid movie of left and middle channels, representing a cotrafficking vesicle containing both EGFP-γ2S (green) and tdt-γ2S (magenta). Arrows point out the position of EGFP-γ2S/tdt-γ2S vesicle in each channel. Bottom panel: cotrafficking of EGFP-GluA2 and tdt-GluA2. Left channel represents the trafficking of an EGFP-GluA2-containg vesicle. Middle channel represents the trafficking of the same vesicle containing tdt-GluA2. Right channel shows an overlaid movie of left and middle channels, representing a cotrafficking vesicle containing both EGFP-GluA2 (green) and tdt-GluA2 (magenta). Arrows point out the position of EGFP-GluA2/tdt-GluA2 vesicle in each channel. (Scale bar: 2.5 μm.) Movie S6 Gu et al. www.pnas.org/cgi/content/short/1525726113 19 of 21 Movie S7. Trafficking of EGFP-GluA2 and tdt-γ2S after exiting the Golgi apparatus. Trafficking vesicles containing EGFP-GluA2 or tdt-γ2S were simultaneously imaged at 32 °C at 0.7-s interval after a preincubation at 20 °C. Left channel represents the EGFP-GluA2 channel. Middle channel represents the tdt-γ2S channel. Right channel shows an overlaid movie of left and middle channels. EGFP-GluA2 is shown in green and tdt-γ2S is shown in magenta. Top panel: an example of a trafficking vesicle containing only tdt-γ2S, but not EGFP-GluA2. Arrows point out the position of tdt-γ2S-vesicle in each channel. Bottom panel: an example of a trafficking vesicle containing only EGFP-GluA2, but not tdt-γ2S. Arrows point out the position of EGFP-GluA2-vesicle in each channel. (Scale bar: 2.5 μm.) Movie S7 Gu et al. www.pnas.org/cgi/content/short/1525726113 20 of 21 Movie S8. Trafficking of EGFP-γ2S and tdt-GluA2 after exiting the Golgi apparatus. Trafficking vesicles containing EGFP-γ2S or tdt-GluA2 were simultaneously imaged at 32 °C at 0.7-s interval after a preincubation at 20 °C. Left channel represents the EGFP-γ2S channel. Middle channel represents the tdt-GluA2 channel. Right channel shows an overlaid movie of left and middle channels. EGFP-γ2S is shown in green and tdt-GluA2 is shown in magenta. Top panel: an example of a trafficking vesicle containing only EGFP-γ2S, but not tdt-GluA2. Arrows point out the position of EGFP-γ2S-vesicle in each channel. Bottom panel: an example of a trafficking vesicle containing only tdt-GluA2, but not EGFP-γ2S. Arrows point out the position of tdt-GluA2-vesicle in each channel. (Scale bar: 2.5 μm.) Movie S8 Gu et al. www.pnas.org/cgi/content/short/1525726113 21 of 21