Subunits Activated by Serotonin Mediate Presynaptic Inhibition by Regulating Vesicle Fusion Properties

G protein ␤␥-subunits activated by serotonin mediate presynaptic inhibition by regulating vesicle fusion properties

Huzefa Photowala*†‡, Trillium Blackmer*†§, Eric Schwartz*, Heidi E. Hamm¶, and Simon Alford*ʈ

*Department of Biological Sciences, University of Illinois, 840 West Taylor Street, Chicago, IL 60607; and ¶Department of Pharmacology, Vanderbilt University Medical School, 23rd Avenue South at Pierce, Nashville, TN 37232

Communicated by Richard W. Tsien, Stanford University School of Medicine, Stanford, CA, January 20, 2006 (received for review October 25, 2005) Neurotransmitters are thought to be released as quanta, where binding. Modification of synaptotagmin–SNARE complex interac- synaptic vesicles deliver packets of neurotransmitter to the syn- tions can cause incomplete fusion of dense core vesicles (27, 28), aptic cleft by fusion with the plasma membrane. However, synaptic which may modify neurotransmitter release (17). 5-HT-mediated vesicles may undergo incomplete fusion. We provide evidence that inhibition at the lamprey giant synapse is blocked by presynaptic G protein-coupled receptors inhibit release by causing such incom- injection of a G␤␥ scavenger (C terminus of G protein receptor plete fusion. 5-hydroxytryptamine (5-HT) receptor signaling po- kinase 2; ref. 2). Similarly, 5-HT no longer inhibits neurotransmitter tently inhibits excitatory postsynaptic currents (EPSCs) between release after presynaptic injection of botulinum toxin A, which lamprey reticulospinal axons and their postsynaptic targets by a cleaves the nine most C-terminal residues of SNAP-25, nor after direct action on the vesicle fusion machinery. We show that 5-HT presynaptic injection of a peptide corresponding to the C-terminal receptor-mediated presynaptic inhibition, at this synapse, involves 14 amino acids of SNAP-25 (9, 26). We now demonstrate that this a reduction in EPSC quantal size. Quantal size was measured presynaptic 5-HT GPCR reduces the quantal amplitude of excita- directly by comparing unitary quantal amplitudes of paired EPSCs tory postsynaptic currents (EPSCs). Unitary EPSC recording and before and during 5-HT application and indirectly by determining staining of synaptic vesicles with FM dyes demonstrate that this the effect of 5-HT on the relationship between mean-evoked EPSC effect can be explained by a change in vesicle fusion properties. amplitude and variance. Results from FM dye-labeling experiments indicate that 5-HT prevents full fusion of vesicles. 5-HT reduces Results FM1-43 staining of vesicles with a similar efficacy to its effect on 5-HT Reduces Evoked Unitary Quantal Amplitude of EPSCs. 5-HT the EPSC. However, destaining of FM1-43-labeled vesicles is abol- inhibits release by acting through G␤␥ on the SNARE complex (2, ished by lower concentrations of 5-HT that leave a substantial 9). In PC12 cells, G␤␥ also inhibits release at the SNARE complex EPSC. The use of a water-soluble membrane impermeant quench- (26) and modulates quantal size (15). To investigate whether 5-HT ing agent in the extracellular space reduced FM1-43 fluorescence reduces the amplitude or number of quanta underlying EPSCs, we during stimulation in 5-HT. Thus vesicles contact the extracellular chose a dose of 5-HT close to the EC50 (600 nM) (29) that inhibited space during inhibition of synaptic transmission by 5-HT. We EPSCs but left sufficient response to investigate its effect on ␤␥ conclude that 5-HT, via free G , prevents the collapse of synaptic quantal amplitudes. The mean EPSC amplitude was reduced by vesicles into the presynaptic membrane. 5-HT (to 51 Ϯ 9% of control, n ϭ 10 pairs). In five pairs, the release probability was sufficiently low to separate statistically the unitary ͉ ͉ ͉ synaptic transmission kiss and run glutamate release response from recording noise (Fig. 1). At least 200 stimuli were 5-hydroxytrytamine applied before (Fig. 1A) and 200 in 5-HT (Fig. 1B). Whole cell access was simultaneously monitored (Fig. 1C). EPSC amplitude protein-coupled receptor (GPCR)-mediated presynaptic in- plots reveal that separation between failures and the lowest quantal Ghibition is widespread, although the underlying mechanisms amplitude, which remained constant throughout the control re- are not always clearly understood. Activation of a presynaptic cording, was lost in 5-HT (Fig. 1D), even while a significant fraction 5-hydroxytryptamine (5-HT) GPCR nearly abolishes glutamate (Ϸ50%) of the EPSC was retained. Fig. 1 E and F shows amplitude release from lamprey reticulospinal synapses (1) independently of data (Fig. 1D) binned into histograms. The smooth curves centered presynaptic calcium (2, 3). Calcium channel independent modula- at the 0 pA amplitude bin are single Gaussians fitted to the tion by GPCRs, first shown at the neuromuscular junction (4), has recording noise, demonstrating no difference between failures and been described since at other synapses (5–8). At the lamprey noise. In all five recordings for the control amplitude distributions, synapse, 5-HT inhibition is mediated by G␤␥ (2, 3) that binds to a significant separation is seen between failures and unitary events. SNAP-25 in SNARE complexes (9). This separation is also apparent in Fig. 1D, where the region Neurotransmitters are widely thought to be released as vesicular between 0 and –5pA is clear of responses throughout the control quanta (10, 11), whereby all vesicles contain equal amounts of recording. Multiple Gaussian curves were fitted to the evoked transmitter, and their entire contents are liberated into the synapse EPSCs in the control histograms (Fig. 1E; ref. 10; Materials and at fixed rates. However, partial fusion events (kiss and run) that might alter neurotransmitter release have been proposed. In neu- roendocrine cells, large dense-core vesicles can transiently fuse with Conflict of interest statement: No conflicts declared. the membrane (12) to vary release (13, 14) an effect that may be Abbreviations: EPSC, excitatory postsynaptic current; 5-HT, 5-hydroxytryptamine; GPCR, caused by G␤␥ (15). Synaptic vesicles may also partially fuse G protein-coupled receptor. (16–20). Similarly, FM dye fluorescence measurements (21–25) †H.P. and T.B. contributed equally to this work. demonstrate that vesicles fusing with the presynaptic membrane ‡Present address: Department of Physiology, Northwestern University Medical School, may subsequently withdraw without complete loss of fluorophore- Chicago, IL 60611. labeled lipid. §Present address: Oregon Hearing Research Center and Vollum Institute, Oregon Health We have demonstrated that GPCR activation leads to G␤␥ Sciences University, 3181 Southwest Sam Jackson Park Road, Portland, OR 97239. ʈ binding to the C-terminal of SNAP-25 in the SNARE complex (2, To whom correspondence should be addressed. E-mail: [email protected]. NEUROSCIENCE 9, 26) at a site coincident with synaptotagmin–SNARE complex © 2006 by The National Academy of Sciences of the USA

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0600509103 PNAS ͉ March 14, 2006 ͉ vol. 103 ͉ no. 11 ͉ 4281–4286 Downloaded by guest on September 28, 2021 Fig. 1. 5-HT inhibits synaptic transmission by reducing quantal amplitude. (A) Paired recordings between axons and postsynaptic neurons. Presynaptic action potentials (Upper) evoke EPSCs (Lower). (B) 5-HT (300 nM) reduced the EPSC amplitude. (C) Electrode access resistance (Ra) remained stable. Means of 20 responses in control and at the experiment end. Stimuli preceded bya5mVvoltage step. Ra estimated by fitting an exponential to the current decay. (D) Amplitude plot of EPSCs. After 200 stimuli, 5-HT (300 nM) inhibited EPSCs and abolished the distinction between failures and the unitary amplitude. (E) Amplitude histograms of data in D. Smooth curves centered at 0 pA are Gaussian distributions fitted to postsynaptic noise, measured before (E) and in 5-HT (F) with no statistical difference. The remaining curves are multiple Gaussians fitted to the data (Materials and Methods). (E) A significant difference is seen between failures and unitary events. No significant difference is seen between failures and noise. (F) In 5-HT, the unitary event amplitude is lost. (G and H) Amplitude histogram from five analyzed pairs before (G) and in 5-HT (H). Amplitudes normalized such that the unitary amplitude was 1 for each pair before averaging. The amplitude bins (gray) between failures and unitary events in control are significantly different in 5-HT.

Methods). In contrast, in 5-HT, no statistically separate failures and some noisy recordings where EPSCs could not be separated from unitary events were distinguished (n ϭ 5). It was not possible to fit recording noise. These exclusions ensured that recording noise did any combination of Gaussians to the histogram data in 5-HT in any not dominate the recorded variance. Thus, in the example (Fig. 2 of the five pairs (Fig. 6, which is published as supporting information A–C), variance of the control recording noise was 4 pA2, and on the PNAS web site). Results from all five pairs were pooled by variance of the control EPSC amplitude was 1,302 pA2. In 5-HT, normalizing the amplitude data to the unitary amplitude. The five variance of the recording noise was 3 pA2, and variance of the EPSC histograms in control and 5-HT were averaged. The trough in amplitude was 399 pA2 (Fig. 2C). The recording noise represented amplitudes between failures and responses was compared to data an insignificant fraction of the recorded EPSC variance. Electrical obtained in 5-HT, where a significantly larger numbers of events components of the EPSCs were unaffected by 5-HT (Fig. 2 A and were recorded (Fig. 1 G and H). These results represent either a C; variance of the peak of the electrical component was not decrease in the postsynaptic response to transmitter release (10) or significantly different from recording noise; P Ͻ 0.05). The elec- a decrease in cleft transmitter concentration. However, the effect of trical component confirmed action potential invasion of the pre- 5-HT is mediated entirely by a presynaptic effect of G␤␥ at the synaptic terminal (32) and acted as a control to ensure no postsyn- SNARE complex (2, 9, 26). aptic dendritic modification of the EPSC occurred. 5-HT (600 nM) This analysis allowed us to directly investigate changes in quantal partially inhibited the EPSC (Fig. 2 A–C) but caused little change size, but was not always possible, because the failure rate was too in M2͞␴2 (Fig. 2Db; mean slope of M2͞␴2 vs. M is 0.52 Ϯ 0.06). The low to observe a unitary quantal amplitude. An alternative analysis result was similar if the slope was measured in three epochs (50 of less ideal data sets was explored (30), in which the mean responses, 75 min; Fig. 2Dc) in 5-HT. In contrast, manipulation of amplitude of the response and its variance was compared (31). The release probability, by alteration of extracellular Mg2ϩ to block relationship mean2͞variance (M2͞␴2) varies linearly or supralin- Ca2ϩ channels or by paired pulse to enhance the Ca2ϩ response early with mean EPSC amplitude when release probability is altered (33), markedly altered M2͞␴2 [Fig. 2Da; mean slope of M2͞␴2 vs. M but is invariant when the amplitude of each release event is altered is 1.21 Ϯ 0.11 (significantly different from 5-HT; Fig. 2Db)]. if recording noise does not significantly impact the recorded M2͞␴2 analysis is invalid if release probability is high, if the EPSCs variance. arise from a pool of synapses with varying properties, or if the We calculated M2͞␴2 for another five pairs in which the failure stimulus paradigm allows failure of the presynaptic action potential rate was insufficient to measure quantal amplitude directly. This to invade the terminal (34). None of these conditions apply in our alternate analysis allows us to independently assess results in Fig. 1. study. All analyses were performed in paired recordings, markedly We excluded synapses with a high failure rate from this analysis (five narrowing the variability of the synaptic pool. Monitoring the EPSC pairs in Fig. 1). We also excluded EPSCs whose amplitude histo- electrical component ensured that presynaptic action potentials grams were skewed to an invariant high release probability and always invaded the terminal (Fig. 2 A and C), and prior data

4282 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0600509103 Photowala et al. Downloaded by guest on September 28, 2021 Fig. 2. Six hundred nanomolar 5-HT does not alter release probability. (Aa) Paired recording with no failures. (Ab) 5-HT (600 nM) reduced EPSC ampli- Fig. 3. 5-HT reduces FM 1-43 staining of the presynaptic terminal. (A) tudes but did not cause failures. NB in control and 5-HT, the electrical com- Schematic of FM1-43 staining protocol. Reticulospinal axon was recorded with ponent, is present and invariant. The amplitude of the chemical component a microelectrode and stimulated (1 Hz; 1,000 action potentials) in FM1-43 (4 varies markedly. (B) From the example in A, EPSC amplitudes plotted before ␮M) perfused over the spinal cord. Excess dye was cleared from the tissue with and in 5-HT. (C) Mean EPSC amplitudes and variances from data in A before Advasep 7, and single confocal sections were imaged. (Ba) Image of an axon and in 5-HT. 5-HT left the electrical component unaltered. (Da) Plot of M2͞␴2 after labeling with FM1-43 in control. (Bb) After 2,000 stimuli to destain the against mean in four pairs. Release probability was altered by raising extra- axon, it was again stimulated with FM 1-43 (1 Hz; 1,000 action potentials) as cellular Mg2ϩ by 1 mM (squares) or paired pulse stimulation (circles). Data in Ba but now in 5-HT (30 ␮M). An image was taken at the same location as Ba. normalized to the events with larger mean amplitude. Decrease in mean No staining was resolved. (Bc) Bar graph comparing the control staining with amplitude is coincident with a decrease in M2͞␴2.(Db) Data are similar to Da the staining in 30 ␮M 5-HT (n ϭ 20 synapses in four preparations). (Ca) Axon but with ﬁve paired recordings before and in 5-HT (600 nM). Data are was labeled with FM1-43 as in Ba and imaged. (Cb) Same axon was destained normalized to control. A decrease in mean amplitude accompanied by only a (1,000 stimuli) and restained with FM1-43 in 5-HT (600 nM; 1 Hz; 1,000 action small decrease in M2͞␴2 represents a change in the unitary amplitude. (Dc) potentials). Staining was apparent at the same locations as Cb, but at reduced Slopes of M2͞␴2 vs. mean with time. Data from three sequential epochs of 50 intensity. (Cc) Bar graph comparing control staining with staining detected in responses in 5-HT. Mean slope is calculated for all ﬁve pairs by using the 600 nM 5-HT (15 synapses in three preparations). control value of M2͞␴2 vs. mean as a reference. Data are errors Ϯ SEM. FM1-43-labeled vesicles. Synaptic vesicle clusters in single giant comparing number of release sites to synaptic responses implies a axons were labeled with FM1-43 by stimulation at 1 Hz (1,000 release probability of Ϸ0.3 (3). Finally, dendritic modification did action potentials) through the recording microelectrode while not contribute because the EPSC electrical component was unaf- FM1-43 was perfused over the spinal cord (Fig. 3A). A small fected by 5-HT or Mg2ϩ. Thus, although not conclusive on their number of stimuli was chosen to allow fusion of approximately the own, these data are supportive of our results in Fig. 1. readily releasable pool (9) to improve chances of completely destaining the labeled pool. FM1-43 perfusion was stopped, and 5-HT Inhibits Stimulus-Evoked FM Dye Uptake Similarly to Its Effect on excess dye was removed with Advasep 7 (1 mM, Ϸ1 min; ref. 37). EPSCs. We conclude that 5-HT-mediated inhibition of EPSCs After loading, vesicle clusters remain labeled (Fig. 3Ba). We have results from a reduction in cleft glutamate concentration. 5-HT has previously shown that this protocol specifically labels synaptic no effect on postsynaptic responses to glutamate (Fig. 7A, which is vesicle clusters and that the fluorescence is significantly higher than published as supporting information on the PNAS web site). background autofluorescence (38). Reduction in cleft glutamate may follow changes in vesicle fusion, Labeled axons were destained by further stimulation. A further vesicle glutamate concentration, or glutamate buffering in the cleft. 2,000 action potentials were applied to the axons to reduce staining However, 5-HT receptor activation and G␤␥ target the vesicle to 0 Ϯ 1% of control. When the axons were relabeled with FM1-43, fusion SNARE complex in Ͻ20 ms (9, 26), and G␤␥ may alter but in a dose of 5-HT sufficient to maximally inhibit EPSCs (30 fusion pore properties in PC12 cells (15). Thus, we hypothesized ␮M), staining was inhibited (Fig. 3 Bb and Bc;18Ϯ 6% of control). that 5-HT, acting through G␤␥ at the SNARE complex, altered The experiment was repeated, but after destaining the axon, only vesicle fusion. Accordingly, we examined synaptic vesicle cycling 600 nM 5-HT was included during restaining. This dose inhibits the properties with FM dyes (35). EPSC to 50% of control (Figs. 1 and 2). In this staining protocol, Synaptic vesicles stained with FM1-43 may remain intact and not labeling is reduced from the control similarly to the effect of the destain after transient fusion (24, 36). Because we hypothesize that same dose of 5-HT on EPSCs (Fig. 3 Cb and Cc;to60Ϯ 12% of

G␤␥ inhibits release late in vesicle fusion, we wished to determine control). When the FM1-43 is mainly in the aqueous phase (extra- NEUROSCIENCE whether 5-HT could disrupt FM1-43 staining and destaining of cellular solution in the synaptic cleft), its loading into synaptic

Photowala et al. PNAS ͉ March 14, 2006 ͉ vol. 103 ͉ no. 11 ͉ 4283 Downloaded by guest on September 28, 2021 Fig. 4. 5-HT prevents FM1-43 destaining. (A) FM1-43 was loaded into synaptic vesicles in lamprey giant axons. Images show after dye loading (Upper), and after destaining with 5,000 stimuli (Lower). (B)AsinA, another axon was stained with FM1-43 (Upper). The axon was then stimulated 5,000 times in 5-HT (600 nM) (Lower). Destaining was blocked (n ϭ 9 synapses, 3 preparations). (C) Destaining quantiﬁed in control (F), during inhibition of synaptic release with Mg2ϩ (4 mM, E), or in 5-HT (600 nM, ᮀ). (D) 5-HT (600 nM; a) and Mg2ϩ (4 mM, b) inhibition of paired EPSCs. (E) Comparison of the inhibition of EPSCs or FM1-43 destaining by 5-HT (a) and Mg2ϩ (b). 5-HT more profoundly inhibited FM1-43 destaining than the EPSC, whereas Mg2ϩ showed similar inhibition of both.

vesicles is reduced by 5-HT, but not abolished. This result does not vesicle release probability (high Mg2ϩ) reduced EPSC amplitudes distinguish between possible effects of 5-HT on synaptic transmis- and FM1-43 destaining similarly. Because individual vesicle fusion sion. The reduction in labeling might be due to a reduction in vesicle events are unaltered in high Mg2ϩ, there are simply fewer events, fusion probability or in labeling of fusing vesicles through an both the mean EPSC amplitude and destaining should be reduced endocytic pore. To investigate this phenomenon more directly, we proportionately. In 5-HT, a very different effect on destaining is determined the effect of 5-HT on destaining of labeled vesicles. seen. Although a significant EPSC is recorded, destaining is prevented. This destaining is not proportional with the effect of 5-HT Low-Dose 5-HT Profoundly Inhibits Destaining of FM1-43. Synaptic on EPSC amplitudes. vesicle clusters were labeled with FM1-43 (2,000 stimuli; ref. 37). We propose that 5-HT liberates G␤␥ to interact with the After clearing the tissue of extracellular dye, 5,000 stimuli reduced SNARE complex (2, 9, 26) to cause transient fusion. During this fluorescence to 9.6 Ϯ 4.1% of prestimulus intensity (n ϭ 9; Fig. 4 incomplete fusion, glutamate can escape from vesicles, because A and C; we define this reduction of fluorescence as control glutamate dissolved in the vesicle interior can escape a transiently destaining). However, destaining was nearly abolished by using the formed pore, but the lipophilic FM1-43 cannot (28). If our hypoth- same protocol in 600 nM 5-HT (Fig. 4 B and C; after 5,000 stimuli esis is true, then the vesicle interior must contact the extracellular fluorescence remained at 89 Ϯ 10% of prestimulus intensity; n ϭ space to allow glutamate release. 5). Thus, in 600 nM 5-HT, we recorded only 15 Ϯ 10% of destaining Continuing this line of reasoning, water-soluble molecules in the in controls. extracellular space (e.g., sulforhodamine, Fig. 5A) should permeate We included a control to monitor the effect of reducing vesicle the fusion pore. Sulforhodamine quenches FM1-43 by fluorescence fusion probability on FM1-43 destaining. Increasing Mg2ϩ concen- resonance energy transfer and has been used to reduce nonsynaptic trations immediately reduces vesicle fusion probability by inhibiting fluorescence in FM1-43-labeled hippocampal cultures (39). Exci- Ca2ϩ entry. This reduction of probability reduced the mean EPSC tation with a 488-nm laser will not directly excite sulforhodamine amplitude (Fig. 4Db) but did not change the vesicle fusion mode. (Fig. 5C). The dye is a small molecule (molecular mass 560 Da), Destaining of FM1-43 in 4 mM Mg2ϩ after 5,000 stimuli was therefore, it may have access to synaptic vesicle pores even if vesicles inhibited to 46.5 Ϯ 5.4% of the prestimulus intensity (Fig. 4C). undergo incomplete fusion in 5-HT. Sulforhodamine is highly We compared the effects of 5-HT and Mg2ϩ on EPSCs (Fig. 4D) water-soluble and should not have access to axon interiors when and on destaining. Bar graphs comparing the effect of 5-HT on applied extracellularly. To confirm this inaccessibility, we super- FM1-43 destaining and on inhibition of the mean EPSC amplitude fused the tissue with sulforhodamine (Fig. 5A;25␮M; two spinal are shown (Fig. 4E). 5-HT (600 nM) reduced the mean EPSC cords), imaged the tissue confocally, and reconstructed the vol- amplitude to 45 Ϯ 7% (n ϭ 22) of control and reduced FM1-43 umes. An axon was filled with fluorescently labeled phalloidin destaining to 15 Ϯ 10% of control destaining (Fig. 4Ea; n ϭ 5). (phalloidin Alexa 488) to mark its synapses. Clearly, the dye was Mg2ϩ (4 mM) reduced the EPSC to 65 Ϯ 7% of control (n ϭ 3; Fig. restricted to the extracellular space. Background fluorescence in the 4Eb) and FM1-43 destaining to 56 Ϯ 7% of control (n ϭ 8; Fig. 4 axon interiors was unaltered by sulforhodamine. We confirmed that C and Eb). sulforhodamine neither altered EPSCs, nor prevented 5-HT- Each destaining data set was compared to the others (two-factor mediated inhibition. In paired recordings (n ϭ 3; Fig. 5B), sulfor- ANOVA), comparing the fluorescence at each stimulus number hodamine (25 ␮M) caused no change in EPSCs (102.3 Ϯ 7.9% of with the equivalent in control conditions. For FM1-43, the effects control) and 5-HT (10 ␮M) inhibited EPSCs similarly to controls of Mg2ϩ and 5-HT were significantly different from each other and (15.8 Ϯ 5.1% of control). control (P Ͻ 0.05). 5-HT (600 nM; reduced EPSCs to Ϸ50% of We tested whether vesicles stained with FM1-43 showed a control) largely blocked vesicle destaining. In contrast, reduced stimulus-evoked reduction in FM1-43 fluorescence in 5-HT and

4284 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0600509103 Photowala et al. Downloaded by guest on September 28, 2021 destaining similar to that seen with no 5-HT present implies that 5-HT (600 nM) does not alter vesicle fusion probability because an equal number of vesicles contact the membrane in 5-HT as in controls. This contact allows sulforhodamine to enter the same number of vesicles from which FM1-43 would have escaped in controls and to effect a similar reduction in fluorescence. Discussion Transient vesicle fusion may represent an important component of synaptic regulation, but a controlled change of fusion mode requires a molecular mechanism. GPCRs can inhibit synaptic transmission by a direct action of G␤␥ at the SNARE complex (2, 9, 26). We now conclude that this pathway causes transient fusion of synaptic vesicles, resulting in reduced EPSC amplitudes. Cleavage of the presynaptic G␤␥ target blocks the effect of 5-HT entirely (2, 9). This result, together with data showing no postsynaptic effect of 5-HT, leaves a presynaptic locus by which 5-HT alters the EPSC quantal amplitude. 5-HT, at a dose that reduces quantal size (Figs. 1 and 2), mediates a profoundly different effect on FM1-43 destaining. This difference is explained if 5-HT causes the vesicle to fuse transiently and then withdraw from the membrane without full collapse (23, 24, 28, 40). Thus, although 5-HT inhibits neurotransmitter release, it inhibits only to the extent that a fusion pore restricts glutamate access to the synaptic cleft. Destaining, during which FM1-43 is mostly lipid bound in the synaptic vesicle, is, in contrast, prevented by the same dose of 5-HT. If the vesicle forms a pore and then withdraws from the terminal membrane, the FM dye will remain largely trapped and will not fully destain (28). If our hypothesis is correct, then to allow the release of some glutamate, vesicles must transiently fuse with the presynaptic membrane, which will expose the vesicle interior to the extracellular space. The stimulation-dependent quenching of FM1-43-stained Fig. 5. Sulforhodamine quenches FM1-43 trapped in vesicles by 5-HT. (A) vesicles with extracellular sulforhodamine strongly supports this Extracellular sulforhodamine does not enter axons. Orthogonal views of the hypothesis. ventro-medial spinal cord: sagittal volume (Left) and volume from dorsal ␤␥ surface (Right). Sulforhodamine (25 ␮M) was superfused over the spinal cord We conclude that 5-HT, acting via G (2) at the SNARE (red). An axon was labeled with Alexa phalloidin 488 (green) to mark presyn- complex (9, 26), alters synaptic transmission to a mode in which aptic terminals and to demonstrate how it penetrates the intracellular void vesicles transiently meet the plasma membrane but allow neuro- unstained by sulforhodamine. (B) Sulforhodamine (25 ␮M) had no effect on transmitters to exit or sulforhodamine to enter. G␤␥ binds the EPSC. 5-HT (10 ␮M) in sulforhodamine inhibited the EPSC as in control. SNAP-25 at its C terminus in the SNARE complex and can Traces are the means of 10 sequential responses. (C) Sulforhodamine absorp- compete with synaptotagmin for binding (9, 26). Manipulations of tion overlaps FM1-43 emission. We excited FM1-43 at 488 nm and detected synaptotagmin also alter fusion properties (27, 41, 42). It is tempting ﬂuorescence with a bandpass ﬁlter (505–560 nm). Sulforhodamine is not to speculate that fusion pore stability might be modulated by excited directly at these wavelengths and does not emit at the detected signaling at the synaptotagmin–SNARE complex interface. This wavelengths. (D) Stained axon destained in 600 nM 5-HT and sulforhodamine after 2,000 and 3,000 stimuli. (E) 5-HT (600 nM) prevents destaining. In hypothesis is supported by our data with FM dyes to analyze sulforhodamine (25 ␮M) and 5-HT, destaining was close to complete. (F) synaptic vesicle fusion in 5-HT. During lipid–lipid interactions Model showing loss of FM1-43 (green) from vesicles during control stimula- between the vesicle and the plasma membrane, which has been tion. (G) Model showing sulforhodamine (red) quenching FM1-43 trapped in thought to underlie kiss-and-run fusion, FM dyes rapidly diffuse in the vesicle during stimulation in 5-HT. the membrane (43). Recording EPSCs while destaining is blocked implies that this fusion does not involve continuity between the vesicle lipid and plasma membrane. A more likely scenario is that sulforhodamine. Vesicles were labeled with FM1-43. Superfusate the SNARE complex forms a channel-like protein pore (42) in containing 600 nM 5-HT and 25 ␮M sulforhodamine was added. which alterations in synaptotagmin interaction (9, 26) favors tem- This reduced background fluorescence such that only the labeled porary pore formation (27, 41, 42). vesicle clusters were visible above the noise of the confocal pho- We propose that G␤␥ changes the transmitter cleft concentration tomultiplier tube detectors. Destaining was then monitored during by altering vesicle fusion (Fig. 5 F and G). However, this mechanism stimulation. During stimulation, FM1-43 fluorescence of labeled is not the only one by which G proteins alter release. Regulation of puncta (505–560 nm) was significantly reduced (Fig. 5 D and E)in Ca2ϩ channels is also important at the presynaptic terminal (44, 45). contrast with the effect of 5-HT destaining with no sulforhodamine Perhaps G␤␥ acts through parallel pathways, causing reduced Ca2ϩ (Fig. 4; also for comparison in Fig. 5E). Destaining of FM1-43 in 600 influx and transmitter release from the exocytosing vesicle. Rapid nM 5-HT alone was blocked (89 Ϯ 10% of fluorescence remaining recycling associated with partial vesicle fusion may also conserve after 5,000 stimuli compared to 20 Ϯ 12% in 600 nM 5-HT and 25 vesicles from depletion during sustained stimulation. Furthermore, ␮M sulforhodamine, n ϭ 9; P Ͻ 0.01). We conclude that FM1-43 if the quantity of transmitter released is varied presynaptically, remains largely trapped in synaptic vesicles undergoing transient mechanisms of synaptic plasticity are available that have not been fusion (Fig. 4). Sulforhodamine, however, gains access to the widely discussed: for example, differential control of postsynaptic vesicles by diffusion through an aqueous fusion pore, leading to receptors with differing distributions or with different affinities to

almost complete quenching of vesicle fluorescence of Ͼ5,000 the released transmitter (e.g., N-methyl-D-asparate vs. ␣-amino-3- NEUROSCIENCE stimuli (Fig. 5 C and E). That sulforhodamine causes apparent hydroxy-5-methyl-4-isoxazolepropionic acid receptors) (25, 46).

Photowala et al. PNAS ͉ March 14, 2006 ͉ vol. 103 ͉ no. 11 ͉ 4285 Downloaded by guest on September 28, 2021 Methods width, A, B, C...,andF, G, H,.... Experiments were performed on larval lampreys (Petromyzon ma- rinus) anesthetized with tricaine methanesulfonate (MS-222; 100 All histograms from the five pairs were well fit with peak mg͞liter; Sigma), decapitated, and dissected in cold saline: 100 mM amplitudes that were multiples of the unitary amplitude. The mean peak amplitude of the smallest of these Gaussians and the mean NaCl͞2.1 mM KCl͞2.6 mM CaCl2͞1.8 mM MgCl2͞4mMglucose͞5 mM Hepes, adjusted to pH 7.6 with NaOH. width were significantly different from the mean amplitude of recording noise and its fitted width. In all cases, the peak current of each subsequent Gaussian (parameters F, G, H,. . . ) never deviated Electrophysiology. Axons were recorded with microelectrodes (1–3 Ͼ M KCl; 20–50 M⍀), neurons with patch electrodes containing 102.5 10% from a multiple of the minimum peak current for the first mM Cs methane sulfonate, 1 mM NaCl, 1 mM MgCl ,5mM three curves. This uniformity would be expected from a distribution 2 of events from sums of 0, 1, and 2 quanta for each stimulus. EGTA, and 5 mM Hepes, pH adjusted to 7.2 with CsOH. EPSC Goodness of fit were confirmed with Kolmogoroff-Smirnoff tests amplitudes measured were the difference between the current for each pair (Supporting Text, which is published as supporting immediately before the stimulus and at the EPSC peak. Noise was information on the PNAS web site) the amplitude difference between data points with the same time difference but no stimulus. Whole cell access was monitored by Analysis of Quantal Noise. If transmitter release is represented by a voltage step application (Fig. 1C). binomial distribution, where M equals mean synaptic current and ␴2 equals variance of that current over a finite stimulus number, then Imaging. FM1-43 (4 ␮M) was applied over the spinal cord while an axon was intracellularly stimulated. Postsynaptic activity was M2͞␴2 ϭ (Npvz)2͞Np(1 Ϫ p)(vz)2 ϭ Np͞(1 Ϫ p), [2] blocked with glutamate receptor antagonists CNQX and AP5 (5 and 100 ␮M, respectively). Excess dye was removed with Advasep where p equals release probability of N quanta available, v equals 7 (1 mM; ref. 37 and 38; 1 min) to reveal stimulus-dependent content of each vesicle, and z the response to a unit release of 2͞␴2 staining. Dye-stained axons were imaged confocally (modified transmitter. M is independent of z if z or the unit release Bio-Rad MRC 600) or conventionally (Hamamatsu charge- remains unchanged and increases proportionately to N and linearly coupled device). Staining is Ca2ϩ-dependent (37) and synapse- to p. This prediction has been experimentally confirmed at the specific (38). Stimulation through the recording electrode was used neuromuscular junction where the quantal coefficient of variation to probe for exocytosis-dependent destaining. (CV) is low. It is less clear in the CNS, where CV is likely to be high (ref. 34; CV at the lamprey synapse is high, 0.5 Ϯ 0.1; n ϭ 4). Nevertheless, this measure predicts a presynaptic locus with Quantal Analysis. Multiple Gaussian curves were fitted to ampli- ϩ changes in presynaptic Ca2 entry, and a postsynaptic locus after tudes in control histograms (Fig. 1D) as a sequence of Gaussians postsynaptic receptor blockade. with increasing widths corresponding to summed normally distrib- Significance was determined by using Student’s paired t test or uted quantal amplitudes such that ANOVA. P values Ͼ0.05 were determined to be not significant.

xϪF 2 xϪG 2 xϪH 2 Ϫͩ ͪ Ϫͩ ͪ Ϫͩ ͪ f͑x͒ϭAe width ϩ Be 2.width ϩ Ce 3.width ... We thank T. Geraschshenko, J. Buchanan, H. von Gersdorff, and G. Viana di Prisco for critical reading of this and earlier versions of the up to five Gaussians, [1] manuscript and invaluable discussions. This work was supported by grants from the National Institute of Mental Health (to S.A.) and the National Eye where the fit parameters were: Institute (to H.E.H.).

1. Buchanan, J. T. & Grillner, S. (1991) Neurosci. Lett. 122, 71–74. 25. Chen, G., Harata, N. C. & Tsien, R. W. (2004) Proc. Natl. Acad. Sci. USA 101, 2. Blackmer, T., Larsen, E. C., Takahashi, M., Martin, T. F., Alford, S. & Hamm, 1063–1068. H. E. (2001) Science 292, 293–297. 26. Blackmer, T., Larsen, E. C., Bartleson, C., Kowalchyk, J. A., Yoon, E. J., 3. Takahashi, M., Freed, R., Blackmer, T. & Alford, S. (2001) J. Physiol. 532, 323–336. Preininger, A. M., Alford, S., Hamm, H. E. & Martin, T. F. (2005) Nat. 4. Silinsky, E. M. (1984) J. Physiol. 346, 243–256. Neurosci. 8, 421–425. 5. Capogna, M., Gahwiler, B. H. & Thompson, S. M. (1996) J. Neurophysiol. 75, 27. Bai, J., Wang, C. T., Richards, D. A., Jackson, M. B. & Chapman, E. R. (2004) 2017–2028. Neuron 41, 929–942. 6. Nurrish, S., Segalat, L. & Kaplan, J. M. (1999) Neuron 24, 231–242. 28. Richards, D. A., Bai, J. & Chapman, E. R. (2005) J. Cell Biol. 168, 929–939. 7. Sakaba, T. & Neher, E. (2003) Nature 424, 775–778. 29. Schwartz, E. J., Gerachshenko, T. & Alford, S. (2005) J. Neurophysiol. 93, 8. Harvey, V. L. & Stephens, G. J. (2004) Eur. J. Neurosci. 20, 684–700. 980–988. 9. Gerachshenko, T., Blackmer, T., Yoon, E. J., Bartleson, C., Hamm, H. E. & 30. Malinow, R. & Tsien, R. W. (1990) Nature 346, 177–180. Alford, S. (2005) Nat. Neurosci. 8, 597–605. 31. Tsien, R. W. & Malinow, R. (1991) Ann. N.Y. Acad. Sci. 635, 208–220. 10. Del Castillo, J. & Katz, B. (1954) J. Physiol. (London) 124, 560–573. 32. Cochilla, A. J. & Alford, S. (1997) J. Physiol. (London) 499, 443–457. 11. Heuser, J. E., Reese, T. S., Dennis, M. J., Jan, Y., Jan, L. & Evans, L. (1979) 33. Cochilla, A. J. & Alford, S. (1998) Neuron 20, 1007–1016. J. Cell Biol. 81, 275–300. 34. Korn, H. & Faber, D. S. (1991) Trends Neurosci. 14, 439–445. 12. Artalejo, C. R., Elhamdani, A. & Palfrey, H. C. (1998) Curr. Biol. 8, R62–R65. 35. Betz, W. J. & Bewick, G. S. (1992) Science 255, 200–203. 13. Graham, M. E., Fisher, R. J. & Burgoyne, R. D. (2000) Biochimie 82, 469–479. 36. Henkel, A. W. & Betz, W. J. (1995) J. Neurosci. 15, 8246–8258. 14. Elhamdani, A., Palfrey, H. C. & Artalejo, C. R. (2001) Neuron 31, 819–830. 37. Kay, A. R., Alfonso, A., Alford, S., Cline, H. T., Holgado, A. M., Sakmann, B., 15. Chen, X.-K., Wang, L.-C., Zhou, Y., Cai, Q., Prakriya, M., Duan, K.-L., Sheng, Snitsarev, V. A., Stricker, T. P., Takahashi, M. & Wu, L. G. (1999) Neuron 24, Z.-H., Lingle, C. & Zhou, Z. (2005) Nat. Neurosci. 8, 1160–1168. 809–817. 16. Ceccarelli, B., Hurlbut, W. P. & Mauro, A. (1973) J. Cell Biol. 57, 499–524. 38. Photowala, H., Freed, R. & Alford, S. (2005) J. Physiol. 569, 119–135. 17. Choi, S., Klingauf, J. & Tsien, R. W. (2003) Philos. Trans. R. Soc. London B 358, 39. Pyle, J. L., Kavalali, E. T., Choi, S. & Tsien, R. W. (1999) Neuron 24, 803–808. 695–705. 40. Stevens, C. F. & Williams, J. H. (2000) Proc. Natl. Acad. Sci. USA 97, 18. Pawlu, C., DiAntonio, A. & Heckmann, M. (2004) Neuron 42, 607–618. 12828–12833. 19. Klyachko, V. A. & Jackson, M. B. (2002) Nature 418, 89–92. 41. Wang, C. T., Lu, J. C., Bai, J., Chang, P. Y., Martin, T. F., Chapman, E. R. & 20. Gandhi, S. P. & Stevens, C. F. (2003) Nature 423, 607–613. Jackson, M. B. (2003) Nature 424, 943–947. 21. Kavalali, E. T., Klingauf, J. & Tsien, R. W. (1999) Philos. Trans. R. Soc. London 42. Han, X., Wang, C. T., Bai, J., Chapman, E. R. & Jackson, M. B. (2004) Science B 354, 337–346. 304, 289–292. 22. Pyle, J. L., Kavalali, E. T., Piedras-Renteria, E. S. & Tsien, R. W. (2000) Neuron 43. Taraska, J. W. & Almers, W. (2004) Proc. Natl. Acad. Sci. USA 101, 8780–8785. 28, 221–231. 44. Catterall, W. A. (1999) Ann. N.Y. Acad. Sci. 868, 144–159. 23. Zakharenko, S. S., Zablow, L. & Siegelbaum, S. A. (2002) Neuron 35, 1099–1110. 45. Mirotznik, R. R., Zheng, X. & Stanley, E. F. (2000) J. Neurosci. 20, 7614–7621. 24. Aravanis, A. M., Pyle, J. L. & Tsien, R. W. (2003) Nature 423, 643–647. 46. Liu, G., Choi, S. & Tsien, R. W. (1999) Neuron 22, 395–409.

4286 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0600509103 Photowala et al. Downloaded by guest on September 28, 2021