Nanodomain Coupling between Ca 2+ Channels and Ca 2+ Sensors Promotes Fast and Efficient Transmitter Release at a Cortical GABAergic Synapse

Transcription

1 Artcle Nanodoman Couplng between Ca 2+ Channels and Ca 2+ Sensors Promotes Fast and Effcent Transmtter Release at a Cortcal GABAergc Synapse Iancu Bucurencu, 1 Akos Kulk, 2 Beat Schwaller, 3 Mchael Frotscher, 2 and Peter Jonas 1, * 1 Physologcal Insttute I, Unversty of Freburg, Hermann-Herder-Straße 7, D Freburg, Germany 2 Insttute of Anatomy and Cell Bology, Unversty of Freburg, Albertstraße 17, D Freburg, Germany 3 Unt of Anatomy, Department of Medcne, Unversty of Frbourg, 1 Route Albert-Gockel, CH-1700 Frbourg, Swtzerland *Correspondence: peter.jonas@physologe.un-freburg.de DOI /j.neuron SUMMARY It s generally thought that transmtter release at mammalan central synapses s trggered by Ca 2+ mcrodomans, mplyng loose couplng between presynaptc Ca 2+ channels and Ca 2+ sensors of exocytoss. Here we show that Ca 2+ channel subunt mmunoreactvty s hghly concentrated n the actve zone of GABAergc presynaptc termnals of putatve parvalbumn-contanng basket cells n the hppocampus. Pared recordng combned wth presynaptc patch ppette perfuson revealed that GABA release at basket cell-granule cell synapses s senstve to mllmolar concentratons of the fast Ca 2+ chelator BAPTA but nsenstve to the slow Ca 2+ chelator EGTA. These results show that Ca 2+ source and Ca 2+ sensor are tghtly coupled at ths synapse, wth dstances n the range of nm. Models of Ca 2+ nflow-exocytoss couplng further reveal that the tghtness of couplng ncreases effcacy, speed, and temporal precson of transmtter release. Thus, tght couplng contrbutes to fast feedforward and feedback nhbton n the hppocampal network. INTRODUCTION The dstance between presynaptc Ca 2+ channels and the Ca 2+ sensor of exocytoss s a key factor that determnes the effcacy and tmng of synaptc transmsson. Although ths dstance s dffcult to measure drectly, t can be probed by exogenous Ca 2+ chelators (Adler et al., 1991). Ths approach has been appled prevously to a varety of synapses. In the squd gant synapse, transmtter release s suppressed by mllmolar concentratons of the fast Ca 2+ chelator BAPTA but s unaffected by 80 mm of the slow Ca 2+ chelator EGTA (Adler et al., 1991), ndcatng that the dffusonal dstance between Ca 2+ source and Ca 2+ sensor s n the nanometer range (Neher, 1998). In contrast, n the young calyx of Held and n neocortcal glutamatergc synapses, transmtter release s suppressed by both 1 mm BAPTA and 1 mm EGTA (Borst and Sakmann, 1996; Menrenken et al., 2002; Fedchyshyn and Wang, 2005; Ohana and Sakmann, 1998; Rozov et al., 2001), suggestng that the source-sensor dstance must be longer, close to the mcrometer range (Menrenken et al., 2002). These results led to the prevalng vew that transmtter release at the squd gant synapse s trggered by Ca 2+ nanodomans, whereas release at central synapses n the mammalan cortex s ntated by Ca 2+ mcrodomans n presynaptc termnals (Augustne et al., 2003). However, whether these conclusons apply generally to central synapses n the mammalan bran has remaned unclear. As tght couplng between Ca 2+ channels and synaptc vescles s expected to mnmze the dffusonal component of the synaptc delay (Menrenken et al., 2002), t s possble that the dstance between Ca 2+ source and Ca 2+ sensor s partcularly short n neuron types specalzed for rapd sgnalng. To test ths dea, we probed the spatal relaton between Ca 2+ source and Ca 2+ sensor at the output synapses of fast-spkng, parvalbumnexpressng GABAergc basket cells (BCs) n the dentate gyrus of the hppocampus. These neurons operate as fast sgnalng devces, generatng rapd and temporally precse nhbtory synaptc output sgnals onto postsynaptc target cells (Jonas et al., 2004; Bartos et al., 2007; Freund and Katona, 2007). We found that the dstance between Ca 2+ source and Ca 2+ sensor at these synapses s n the nanometer range, smlar to the squd gant synapse (Adler et al., 1991), but very dfferent from any other prevously studed cortcal synapse. RESULTS Ca 2+ Channel Immunoreactvty Is Clustered n Actve Zones of BC Synapses We explored the possblty that tght couplng between Ca 2+ source and Ca 2+ sensor may contrbute to the fast sgnalng at the basket cell (BC)-granule cell (GC) synapse n the dentate gyrus. As prevous studes showed that transmtter release at ths synapse s entrely medated by P/Q-type Ca 2+ channels (Hefft and Jonas, 2005; see Poncer et al., 1997; Wlson et al., 2001), we frst examned the dstrbuton of a 1A subunt mmunoreactvty on putatve BC termnals by hgh-resoluton mmunoelectron mcroscopy (Fgure 1). BC termnals were dentfed by the symmetrcal contacts that they formed on granule cell somata. Pre-embeddng mmunogold labelng revealed that a 1A mmunoreactvty was hghly concentrated n the actve zone of putatve BC termnals. In both adjacent seral sectons (Fgures 1A and 1B) and n three-dmensonally reconstructed boutons 536 Neuron 57, , February 28, 2008 ª2008 Elsever Inc.

2 Fgure 1. a 1A Subunt Immunoreactvty Is Concentrated n the Actve Zone of Putatve BC Termnals (A and B) Electron mcrographs showng a presynaptc termnal (T) n consecutve ultrathn sectons that establshes a symmetrcal synaptc contact wth a neuronal soma (S) n the granule cell layer of the dentate gyrus. Pre-embeddng mmunogold labelng revealed that a 1A mmunoreactvty was manly localzed at the presynaptc membrane specalzaton (arrows); mmunogold partcles were only occasonally seen on the extrasynaptc plasma membrane (arrowheads) of boutons. Note that the partcles were present at smlar locatons n adjacent sectons. (C) Three-dmensonal reconstructon of the same axon termnal (arrows pont to the sectons shown n [A] and [B]). Reconstructon was carred out from seral ultrathn sectons (n = 9) obtaned from pre-embeddng materal showng the localzaton of mmunogold partcles for a 1A (whte dots) n relaton to the actve zone (yellow area). (D) Hstogram showng the densty of mmunogold partcles (n = 54) for a 1A on presynaptc termnals (n = 11) n the regon of the actve zone (AZ) and n extrasynaptc membranes. For the extrasynaptc regon, dstances of mmunogold partcles were measured from the closest edge of AZs along the surface of the reconstructed boutons. Densty was determned by countng the number of partcles and dvdng t by the respectve surface area of the presynaptc membrane. Scale bars, 0.2 mm. (Fgure 1C), mmunogold partcles were manly found n the actve zone. Analyss of partcle densty on the plasma membrane n 92 ultrathn sectons from 11 putatve BC termnals revealed that the partcle densty (n = 54) was 9-fold hgher n the actve zone than n the extrasynaptc membrane of the presynaptc termnal (29.2 ± 7.4 mm 2 versus 3.2 ± 0.3 mm 2 ; p < 0.01; Fgure 1D). These results suggest that P/Q-type Ca 2+ channels, the presynaptc Ca 2+ sources for transmtter release at BC output synapses (Hefft and Jonas, 2005), are hghly concentrated n the actve zone, consstent wth a tght couplng of these channels to the exocytotc machnery. Transmtter Release Is Inhbted by the Fast Ca 2+ Chelator BAPTA, but Insenstve to the Slow Chelator EGTA To quantfy the dstance between Ca 2+ source and Ca 2+ sensor, we tested the effect of exogenous chelators on transmtter release at BC-GC synapses (Adler et al., 1991). Untary IPSCs were measured n the pared recordng confguraton, and Ca 2+ chelators were loaded nto presynaptc nterneurons at defned concentratons by patch ppette perfuson (Fgure 2A and Fgure S1). Ppette perfuson allowed us to exchange the soluton at the tp of the patch ppette wthn 2 mn (Fgure 2C). To estmate the tme requred to load chelators nto presynaptc termnals, we smulated chelator dffuson n a detaled model of BC morphology n whch the soma, dendrtes, and proxmal axon were reconstructed and the axon was schematcally extended wth 150 collaterals. Smulaton of dffuson revealed that the chelator concentraton n axon collaterals reached 76% 99% of the somatc concentraton after 100 mn, wth faster exchange n proxmal than n dstal collaterals (Fgure 2D). Thus, chelators can be loaded nto presynaptc termnals wthn typcal recordng tmes f these presynaptc termnals emerge from proxmal axon collaterals. We next examned the effects of Ca 2+ chelators n BC-GC pared recordngs wth <100 mm ntersomatc dstance to mnmze dffuson tmes (Fgure 2A). The effect of Ca 2+ chelators was quantfed as the rato of IPSC chelator /IPSC control, where IPSC control was measured durng a 7.5 mn control perod and IPSC chelator was determned from a 7.5 mn test perod after statonary effects were reached (on average 85 mn after applcaton onset). Loadng the presynaptc BC wth 10 mm of the fast Ca 2+ chelator BAPTA reduced the ampltude of untary BC-GC IPSCs substantally. On average, the peak ampltude of the untary BC-GC IPSCs under steady-state condtons was 16.3% ± 3.6% of the control value (fve pars; Fgures 3A and 3B). Conversely, n cells preloaded wth 10 mm BAPTA, the applcaton of BAPTA-free ntracellular soluton led to a tmedependent ncrease of the IPSC ampltude (Fgures 3C and 3D), suggestng reversblty of the pror effects of BAPTA on transmtter release. On average, the rato of untary BC-GC IPSC ampltude after 90 mn to the mnmal ampltude 20 mn after soluton exchange was 2.5 ± 0.2 (three pars). Collectvely, these results ndcate that 10 mm of the fast Ca 2+ chelator BAPTA consstently and reversbly nhbted synaptc transmsson at the BC-GC synapse. In contrast, loadng the presynaptc BC wth 30 mm of the slow Ca 2+ chelator EGTA reduced the ampltude of the evoked IPSC to only 67.2% ± 4.3%, markedly less than 10 mm BAPTA (fve pars; p < 0.01; Fgures 4A and 4B). We further attempted to determne the half-maxmal nhbtory concentraton of the chelator effects. For BAPTA, the effects were concentraton dependent; 1, 3, 10, and 30 mm BAPTA decreased the IPSC ampltude to 63.7%, 32.3%, 16.3%, and 1.6% of the control value (15 pars total; p < 0.001), yeldng a half-maxmal nhbtory concentraton of 1.6 mm (Fgure 4C). For EGTA, a concentraton-effect relaton could not be obtaned, because even hgh concentratons of Neuron 57, , February 28, 2008 ª2008 Elsever Inc. 537

3 Fgure 2. Loadng of Ca 2+ Chelators nto Presynaptc Termnals of BCs va Somatc Ppette Perfuson (A) Photomcrograph of a BC-GC par, flled wth bocytn, staned usng 3,3 0 -damnobenzdne, and examned lght-mcroscopcally. Patch ppettes and ppette perfuson apparatus are supermposed schematcally. (B) Model of BC morphology used for smulaton of chelator dffuson. Pseudocolor representaton (nset) shows the chelator concentraton after 100 mn. The BC model was taken from Geger et al. (1997). (C) Speed of soluton exchange at the ppette tp, measured as a change n juncton current n an open patch ppette durng a soluton change from 100% Na + -rch to 50% Na + -rch soluton and back va ppette perfuson. (D) Smulaton of the chelator concentraton n 15 axon collaterals as a functon of tme (green curves, collaterals 10 70; black curves, collaterals ; red curve, concentraton at the soma; ndces of collaterals correspond to the numbers n [B]). 100 mn after a somatc concentraton jump, the chelator concentraton n the axon collaterals reaches between 99% and 76% of the somatc concentraton. EGTA had only moderate effects (Fgure 4D). However, assumng that the concentraton dependence follows a Hll equaton wth a Hll coeffcent of 1, the half-maxmal nhbtory concentraton was estmated as 61.5 mm. Thus, the rato of half-maxmal nhbtory concentratons between EGTA and BAPTA (38) s smlar to the nverse rato of the Ca 2+ bndng rates of the chelators (40; Nägerl et al., 2000; Naragh and Neher, 1997; Neher, 1998). To reassure that the effects of Ca 2+ chelators were generated at a presynaptc ste, we tested possble effects on the number of falures and the coeffcent of varaton, ndcators of presynaptc changes (Fgure S2). BAPTA sgnfcantly ncreased the percentage of transmsson falures (p < 0.001; Fgures S2A and S2B) and reduced the coeffcent of varaton of IPSC peak ampltudes rased to the power of 2 (p < 0.001; Fgure S2C), confrmng that the chelator effects were generated at a presynaptc locus. The Exocytotc Ca 2+ Sensor at the BC-GC Synapse Is Not Saturated The relatve nsenstvty of transmtter release at BC termnals to exogenous Ca 2+ chelators, n partcular to the slow Ca 2+ chelator Fgure 3. Transmtter Release at the BC-GC Synapse Is Reversbly Inhbted by 10 mm of the Fast Ca 2+ Chelator BAPTA (A) Presynaptc acton potentals (red) and evoked IPSCs (black) under control condtons (left) and after applcaton of 10 mm BAPTA (rght). Ten consecutve traces are shown supermposed; the average IPSC s shown n green. (B) Plot of IPSC peak ampltude aganst tme durng applcaton of BAPTA va presynaptc patch ppette perfuson. Same par as shown n (A). Tme of applcaton of 10 mm BAPTA s represented by the horzontal bar. Astersks ndcate the tmes at whch traces n (A) were taken. (C and D) Smlar recordng as n (A) and (B), but before and after a change from 10 mm BAPTAcontanng to BAPTA-free ntracellular soluton. Note the ncrease n ampltude of evoked IPSCs, ndcatng reversblty of the pror effect of BAPTA. 538 Neuron 57, , February 28, 2008 ª2008 Elsever Inc.

4 Fgure 4. Transmtter Release at the BC-GC Synapse Is Resstant to the Slow Ca 2+ Chelator EGTA (A) Presynaptc acton potentals and evoked IPSCs under control condtons (left) and after applcaton of 30 mm EGTA to the presynaptc BC by ppette perfuson (rght). Ten consecutve traces are shown supermposed; the average IPSC s shown n green. (B) Plot of IPSC peak ampltude aganst tme durng applcaton of EGTA. Same cell as shown n (A). Tme of applcaton of 30 mm EGTA s represented by the horzontal bar. Astersks ndcate the tmes at whch traces n (A) were taken. (C) Tme course of the onset of the effect of 1 (crcles), 3 (squares), 10 (nverted trangles), and 30 mm BAPTA (trangles). Data from four, four, fve, and two pars were algned to the onset tme of the BAPTA applcaton, normalzed to the control ampltude, condensed by formng means of ten adjacent data ponts, and averaged across pars. (D) Tme course of the onset of the effect of 30 mm EGTA (damonds) and control soluton (mock applcaton, flled crcles). Data from fve and four pars, respectvely. Note that EGTA has small, but detectable effects. The red curves n (C) and (D) represent sldng averages. Error bars ndcate SEM. EGTA, suggests a tght couplng between Ca 2+ source and Ca 2+ sensor. Alternatvely, the relatve nsenstvty may be explaned by saturaton of the sensor durng presynaptc Ca 2+ transents. In ths scenaro, changes n presynaptc Ca 2+ nflow would not lead to correspondng changes n transmtter release. We therefore examned the possblty of saturaton of the presynaptc Ca 2+ sensor by varyng the extracellular Ca 2+ concentraton between 0.5 and 4 mm, whle keepng the extracellular Mg 2+ concentraton constant (Fgure 5). Reducton of the external Ca 2+ concentraton below the control value of 2 mm resulted n a marked decrease n IPSC peak ampltude and a parallel ncrease n the proporton of falures. Conversely, ncrease n the external Ca 2+ concentraton above 2 mm resulted n an ncrease n IPSC peak ampltude beyond the control value. Analyss of the quanttatve relaton between IPSC ampltude and extracellular Ca 2+ concentraton revealed that the half-maxmal effectve concentraton of Ca 2+ was 1.76 mm (Fgure 5C). Thus, the degree of saturaton of the Ca 2+ sensor for 2 mm Ca 2+ was Fgure 5. The Exocytotc Ca 2+ Sensor at BC-GC Synapses Is Not Saturated by Presynaptc Ca 2+ Inflow (A) Presynaptc acton potentals (red) and evoked IPSCs (black) under control condtons (center, 2 mm extracellular Ca 2+ ) and n the presence of dfferent extracellular Ca 2+ concentratons, as ndcated on top. The extracellular Mg 2+ concentraton was held constant at 1 mm. Ten consecutve traces are shown supermposed; the average IPSC s shown n green. (B) Plot of IPSC peak ampltude aganst tme durng applcaton of extracellular solutons wth dfferent Ca 2+ concentratons. Same par as shown n (A). The tme of applcaton of the dfferent extracellular solutons s represented by horzontal bars. (C) Relaton between IPSC ampltude and extracellular Ca 2+ concentraton. IPSC peak ampltudes were normalzed to the value at 2 mm extracellular Ca 2+, averaged across pars, ftted wth the functon f([ca]) = f max [1 + (K/[Ca]) n ] 1, yeldng parameter values of f max = 1.76, K = 1.76 mm, and n = 2.23, and renormalzed to f max. Red curve, normalzed ftted functon; dashed lnes, normalzed IPSC ampltude at 2 mm extracellular Ca 2+, ndcatng that the Ca 2+ sensor showed 57% saturaton. Data from fve to seven pars for each concentraton. Error bars ndcate SEM. Neuron 57, , February 28, 2008 ª2008 Elsever Inc. 539

5 Fgure 6. Concentraton Dependence of BAPTA and EGTA Effects Suggests Nanodoman Couplng between Ca 2+ Source and Ca 2+ Sensor (A and B) Analyss of the chelator effects usng the analytcal steady-state soluton to the lnearzed reacton-dffuson equatons. (A) Plot of rato of IPSC peak ampltude n the presence of chelator to that n control condtons aganst the concentraton of BAPTA (open symbols) or EGTA (flled symbol). Curves represent the predctons of the models ft to the entre data set (contnuous curve, sngle-channel model; dashed curve, cluster model). Schematc llustraton of the models s shown on top (red dots, Ca 2+ channels; green dot, synaptc vescle wth Ca 2+ sensor). The best ft parameters were r = 12 nm (sngle-channel model) and d = 12 nm wth s = 8 nm (cluster model). (B) Both the sngle-channel and the cluster model ft the expermental data substantally better than models wth more homogeneous Ca 2+ sources. Hstograms of the sum of squared dfferences between data and model (c 2 ) for 1000 bootstrap replcatons of the orgnal data ftted wth ether the sngle-channel model (open bars) or the cluster model (flled bars). Arrows ndcate the c 2 values of models n whch the sze of the Ca 2+ channel cluster s corresponds to the average radus of an actve zone (174 nm; AZ ) or n whch the Ca 2+ channels were homogeneously dstrbuted over the entre bouton area ( nfnte ). (C and D) Analyss of the chelator effects usng the tme-dependent soluton to the full reacton-dffuson equatons, consderng moble and fxed buffers as well as buffer saturaton. Data were obtaned as the numercal soluton of Equatons 5 7. (C) Ca 2+ transents at 20 nm dstance from the source under control condtons (upper black lne) and n the presence of dfferent concentratons of EGTA (lower black lnes) or BAPTA (red lnes). Dashed red lne, normalzed presynaptc acton potental; dashed blue lne, normalzed presynaptc Ca 2+ current. (D) Concentraton dependence of the chelator effects, analyzed wth the tme-dependent soluton to the full reacton-dffuson equatons. Curves represent the predctons of the ftted model wth r = 17 nm. Error bars ndcate SEM. 57%, mplyng that the sensor operates n a subsaturatng regme. These results corroborate the concluson that the relatve nsenstvty aganst exogenous Ca 2+ chelators at the BC-GC synapse s caused by tght couplng of source and sensor rather than saturaton of the sensor. The Dstance between Ca 2+ Source and Ca 2+ Sensor Is n the Nanometer Range To quantfy the dstance between Ca 2+ source and Ca 2+ sensor at the BC-GC synapse, we modeled the concentraton dependence of the effects of BAPTA and EGTA on BC-GC IPSCs (Fgure 6). The rato of Ca 2+ concentratons at the sensor n the presence and absence of exogenous chelator was modeled usng the analytcal steady-state soluton to the lnearzed reacton-dffuson equatons (Neher, 1998), and the correspondng rato of IPSC ampltudes was predcted usng the expermentally determned relaton between Ca 2+ concentraton and transmtter release (Fgure 5C). We frst examned the smplest possble model n whch a sngle Ca 2+ channel was coupled to the Ca 2+ sensor (.e., the vescle) at varable dstance r (sngle-channel model; Fgure 6A) (Stanley, 1993, 1997; Brandt et al., 2005; Shahrezae et al., 2006). We further tested a more complex model n whch a cluster of Ca 2+ channels, represented by a normally dstrbuted channel densty wth standard devaton s, was coupled to the Ca 2+ sensor located at varable dstance d from the cluster center (cluster model; Fgure 6A). The dffuson coeffcent of Ca 2+, the physcochemcal propertes of the chelators, the bologcal propertes of the endogenous buffer, and the restng Ca 2+ concentraton n BCs (Y. Aponte et al., 2006, Soc. Neurosc., abstract; Lee et al., 2000) were expermentally constraned (Expermental Procedures), leavng r or d and s as free parameters. The sngle-channel model provded an adequate ft to the expermental observatons; r was estmated as 12 ± 1 nm under these condtons. The cluster model also descrbed the expermental data; d was estmated as 12 ± 4 nm and s as 8 ± 5 nm. The values obtaned for the source-sensor dstance were nsenstve to varatons n the assumed bologcal parameters; a 10-fold decrease or ncrease n the endogenous buffer product changed r by 3 and 10%, and a 3-fold decrease or ncrease n the restng Ca 2+ concentraton changed r by 10 and 25%, respectvely. We also tested whether models n whch Ca 2+ channels were normally dstrbuted over the entre actve zone (radus 174 nm) or homogeneously dstrbuted over the entre bouton area were able to descrbe the chelator effects. However, we found that these models were nconsstent wth the expermental observatons (p < 0.001; Fgure 6B). To test whether smlar estmates of the couplng dstance between Ca 2+ source and Ca 2+ sensor are obtaned n a more detaled model that takes nto account both coexstence of 540 Neuron 57, , February 28, 2008 ª2008 Elsever Inc.

6 Fgure 7. Tght Couplng Increases Effcacy, Speed, and Temporal Precson of Transmtter Release at BC Output Synapses (A and B) Ca 2+ transents at dfferent dstances from the source ( nm, 20 nm ncrement). Ca 2+ transents (black lnes) were calculated usng the tme-dependent soluton to the full reactondffuson equatons (Equatons 5 7; see nset). In (A), all Ca 2+ transents were normalzed to the same peak value at 20 nm. In (B), all Ca 2+ transents were normalzed to ther ndvdual peak values. (C and D) Release rate for dfferent source-sensor dstances ( nm, 20 nm ncrement). Release rates (black lnes) were calculated from the Ca 2+ transents usng a 12 state transmtter release model (nset). In (C), all release rates were normalzed to the same peak value at 20 nm. In (D), all release rates were normalzed to ther ndvdual peak values. In (A) (D), the dashed red lnes represent the normalzed presynaptc acton potental, and the dashed blue lnes ndcate the normalzed presynaptc Ca 2+ current. (E) Plot of delay from the peak of the Ca 2+ current to the maxmal release rate (crcles, red lne) and half-duraton of the release perod (squares, blue lne) aganst couplng dstance. Tght couplng decreases the synaptc delay and the half-duraton of the release perod. (F) Semlogarthmc plot of peak release rate (crcles, red lne, normalzed to the peak release rate at 20 nm) and rato of release rate n the steady state (at 5 ms)/ release rate at the peak (squares, blue lne) aganst couplng dstance. Tght couplng ncreases the peak release rate and decreases the proporton of asynchronous release. Lnes n (E) and (F) represent fts wth polynomal functons. moble and mmoble buffers and buffer saturaton, we further modeled the concentraton dependence of the chelator effects usng the tme-dependent soluton to the full reacton-dffuson equatons obtaned numercally (Fgures 6C and 6D) (Smth, 2001). In a sngle-channel scenaro, the model adequately descrbed the expermental observatons wth a couplng dstance of 17 nm. Thus, ndependently of the detals of the model, the couplng dstance between Ca 2+ source and Ca 2+ sensor at the BC-GC synapse s nm, comparable to the squd gant synapse, but much tghter than at any other cortcal synapse studed to date. Tght Couplng Enhances Effcacy, Speed, and Temporal Precson of Transmtter Release To address the functonal sgnfcance of tght couplng for fast nhbtory synaptc transmsson, we smulated Ca 2+ transents at varous dstances from the Ca 2+ source, usng the tme-dependent soluton to the full reacton-dffuson equatons (Fgure 7). For a bouton of 0.5 mm radus, the peak Ca 2+ steeply declned as a functon of dstance, as expected for dffuson from a pont source (Fgure 7A). Furthermore, wth ncreasng dstance from the source, the rato of Ca 2+ at the peak to that at later tmes decreased substantally (Fgure 7B). Thus, tght couplng between Ca 2+ source and Ca 2+ sensor ncreases the peak ampltude of the Ca 2+ transent, whle decreasng the relatve ampltude of the Ca 2+ concentraton at later tmes n the presynaptc termnals. To examne how the tghtness of couplng affects transmtter release, we further modeled the correspondng release rates usng a Ca 2+ sensor model wth fve Ca 2+ bndng stes (Fgures 7C and 7D; Lou et al., 2005). Increasng the dstance between source and sensor from 20 to 200 nm ncreased the synaptc delay (measured from the rse of the presynaptc acton potental to the peak release rate) 1.57-fold from 560 to 880 ms and ncreased the half-duraton of the tme course of release fold (Fgures 7D and 7E). Thus, tght couplng between Ca 2+ source and Ca 2+ sensor ncreases both speed and temporal precson of transmtter release at BC output synapses. Furthermore, ncreasng the dstance between source and sensor from 20 to 200 nm substantally reduced synchronous release (measured as the peak release rate) but reduced asynchronous release (measured at t = 5 ms) to a much smaller extent, leadng to a relatve enhancement of asynchronous components (Fgures 7D and 7F). As the couplng dstance was ncreased from 20 to 200 nm, the rato of asynchronous to synchronous release rate ncreased 20,000-fold. Thus, tght couplng between Ca 2+ source and Ca 2+ sensor ncreases the speed and temporal precson of synaptc transmsson, whle reducng the relatve contrbuton of asynchronous release. DISCUSSION We found that P/Q-type Ca 2+ channels are concentrated at the actve zone of putatve BC termnals and that transmtter release at BC-GC synapses s surprsngly resstant to Ca 2+ chelators. The most lkely explanaton of ths result s a tght couplng between the Ca 2+ source and the Ca 2+ sensor of exocytoss. Alternatve explanatons, such as saturaton of the exocytotc Ca 2+ sensor or a hgh endogenous Ca 2+ buffer capacty of BCs, are unlkely. Saturaton of the Ca 2+ sensor durng synaptc transmsson (caused for example by the hgh affnty of the sensor) s ncompatble wth the expermentally determned relaton between IPSC ampltude and extracellular Ca 2+ concentraton, whch further ncreases as the Ca 2+ concentraton s rased Neuron 57, , February 28, 2008 ª2008 Elsever Inc. 541

7 above the control value of 2 mm. Hgh endogenous buffer capacty s also unlkely, snce the somatodendrtc endogenous Ca 2+ bndng rato of fast-spkng BCs n the dentate gyrus s 200 (Y. Aponte et al., 2006, Soc. Neurosc., abstract; Lee et al., 2000; Colln et al., 2005), 20-fold lower than the exogenous Ca 2+ bndng rato of 1 mm BAPTA (k concentraton/dssocaton constant = 1000 mm/0.22 mm 4500) (Neher, 1998). Thus, tght couplng between the presynaptc Ca 2+ channels and the exocytotc Ca 2+ sensors s the only nterpretaton consstent wth all expermental observatons. Wth respect to the tght couplng between Ca 2+ source and Ca 2+ sensor, the BC-GC synapse s smlar to the squd gant synapse (Adler et al., 1991) but dfferent from any other cortcal synapse studed to date (Ohana and Sakmann, 1998; Rozov et al., 2001). The BC-GC synapse also dffers markedly from the young calyx of Held (Borst and Sakmann, 1996; Menrenken et al., 2002), although a developmental ncrease n the tghtness of couplng between Ca 2+ channels and transmtter release was suggested (Fedchyshyn and Wang, 2005; see Iwasak and Takahash, 1998). The estmated dstance between Ca 2+ source and Ca 2+ sensor at the BC-GC synapse s n the nanometer range, suggestng that couplng s medated by proten-proten nteractons (Berkefeld et al., 2006). Our results mply that these nteractons must be hghly specfc, requrng both P/Q-type Ca 2+ channels and specfc reacton partners, presumably components of the release machnery of parvalbumn-expressng GABAergc nterneurons. In contrast, couplng s loose n both cortcal glutamatergc synapses n whch a substantal proporton of release s medated by P/Q-type channels (Red et al., 2003; Ohana and Sakmann, 1998) and n GABAergc synapses of CCK-expressng nterneurons n whch release s exclusvely medated by N-type Ca 2+ channels (Hefft and Jonas, 2005). One possblty s that P/Q-type Ca 2+ channels couple drectly to the Ca 2+ sensor, presumably synaptotagmn (Südhof, 2002; Sheng et al., 1997). In ths framework, synaptc specfcty could be conferred by the selectve expresson of P/Q-type channels and dstnct synaptotagmn soforms. Consstent wth ths hypothess, fast-spkng, parvalbumn-expressng BCs use aca 2+ sensor dfferent from synaptotagmn 1 (presumably synaptotagmn 2; A.M. Kerr et al., 2006, Soc. Neurosc., abstract). Alternatvely, couplng may occur between synprnt stes of Ca 2+ channels and SNARE complex protens (Rettg et al., 1996; Bezprozvanny et al., 2000; Mochda et al., 2003). In ths scenaro, specfcty mght be conferred by the selectve expresson of P/Q-type channels and dstnct syntaxns or soluble NSF attachment protens (SNAPs) n GABAergc nterneurons (Verdero et al., 2004). Tght couplng between Ca 2+ nflow and transmtter release at BC-GC synapses wll contrbute to the fast sgnalng propertes of BCs n multple ways. Frst, tght couplng wll ncrease the effcacy of GABAergc synaptc transmsson. For a gven number of presynaptc Ca 2+ channels, ncreasng the couplng dstance from 20 to 200 nm decreases the peak release rate by orders of magntude (Fgure 7F). Although ths decrease n release probablty could be compensated by an ncrease n the number of presynaptc Ca 2+ channels, the physcal sze of the actve zone places clear lmts to the extent of any compensatory effects. Thus, tght couplng between Ca 2+ source and Ca 2+ sensor may explan the hgh probablty of transmtter release at BC output synapses (Kraushaar and Jonas, 2000). Second, tght couplng wll mnmze the dffusonal component of the synaptc delay (Menrenken et al., 2002). In our model of Ca 2+ - dependent transmtter release, ncreasng the couplng dstance from 20 to 200 nm ncreases the synaptc delay by 300 ms (Fgure 7E). Thus, the couplng dstance s a major determnant of the synaptc delay, whch s <1 ms at room temperature. Thrd, tght couplng wll ncrease the temporal precson of phasc release. Fnally, tght couplng wll ncrease the rato of synchronous to asynchronous or ectopc release (Hefft and Jonas, 2005; Matsu and Jahr, 2004). Smlarly, tght couplng may explan the lack of facltaton and potentaton at the BC-GC synapse (Kraushaar and Jonas, 2000; Tang et al., 2000; Awatraman et al., 2005; Zucker and Regehr, 2002). Thus, tght couplng between Ca 2+ source and Ca 2+ sensor extends the repertore of rapd sgnalng mechansms of fast-spkng, parvalbumnexpressng BCs (Jonas et al., 2004; Freund and Katona, 2007). As BCs are embedded n both feedback and feedforward nhbtory mcrocrcuts, tght couplng wll contrbute to the speed and relablty of nhbton n the hppocampal network (Poulle and Scanzan, 2001, 2004). If tght couplng s a general property of BC output synapses, t wll be also mportant for fast sgnalng at BC-BC synapses and the generaton of fast network oscllatons n nterneuron networks (Bartos et al., 2007). EXPERIMENTAL PROCEDURES Pre-Embeddng Immunoelectron Mcroscopy Sectons of the hppocampus of 2-month-old Wstar rats were processed for mmunogold labelng (Baude et al., 1993; Kulk et al., 2004) wth an affntypurfed polyclonal ant-a 1A antbody (Alomone Laboratores, Jerusalem, Israel). Brefly, anmals were deeply anesthetzed by Narkodorm-n (180 mg/kg,.p.) (Alvetra, Neumünster, Germany), and perfused transcardally wth 0.9% salne followed by fxatve contanng 4% paraformaldehyde, 15% saturated pcrc acd, and 0.05% glutaraldehyde n 0.1 M phosphate buffer (PB). Tssue blocks were washed n 0.1 M PB and 50 mm thck sectons were cut on a vbratome. Sectons were cryoprotected n a soluton contanng 25% sucrose and 10% glycerol n 50 mm PB and processed for pre-embeddng mmunogold labelng as descrbed (Kulk et al., 2004). Seral ultrathn (70 nm) sectons were cut from the surfaces of the samples. Axon termnals of putatve parvalbumnexpressng nterneurons were dentfed by ther symmetrc synaptc contacts wth somata of dentate gyrus granule cells. Densty of mmunopartcles was quantfed as the number of partcles n the actve zone versus the extrasynaptc membrane, dvded by the correspondng surface areas. Pared Recordng Transverse hppocampal slces (300 mm thckness) were cut from brans of 18- to 21-day-old Wstar rats usng a vbratome. Anmals were klled by decaptaton, n agreement wth natonal and nsttutonal gudelnes. Patch ppettes were pulled from thck-walled boroslcate glass tubng (2 mm outer dameter, 1 mm nner dameter); when flled wth ntracellular soluton, the resstance was MU. Smultaneous recordngs from monosynaptcally connected basket cells (BCs) and granule cells (GCs) n the dentate gyrus (Kraushaar and Jonas, 2000) were obtaned under vsual control usng nfrared dfferental nterference contrast vdeomcroscopy. A tght-seal whole-cell recordng was frst establshed n a putatve BC. Selected cells had somata located n the granule cell layer near the hlar border and generated >50 acton potentals durng 1 s depolarzng current pulses (600 pa to 1 na). Subsequently, whole-cell recordngs were made from GCs. To optmze the dffusonal access to presynaptc termnals, pared recordngs were made from adjacent cells wth <100 mm ntersomatc dstance. The recordng temperature was 22 C±2 C. 542 Neuron 57, , February 28, 2008 ª2008 Elsever Inc.

8 Two Axopatch 200A amplfers (Axon Instruments) were used for recordng. The presynaptc neuron was held n the current-clamp mode (I-clamp normal, holdng potental 70 mv). Acton potentals were elcted by sngle or trans of current pulses (duraton 1 ms, ampltude na, repetton tme 6 or 16 s, respectvely). The postsynaptc cell was held n the voltage-clamp mode at a holdng potental of 80 mv. Postsynaptc seres resstance (range: 5 10 MU) was not compensated but was contnuously montored usng a 4 mv test pulse; only pars wth <2 MU change were analyzed. Sgnals were fltered at 5 khz (4-pole low-pass Bessel flter) and dgtzed at 10 khz usng a CED 1401plus nterface (Cambrdge Electronc Desgn). Pulse generaton and data acquston were performed usng FPulse (U. Fröbe, Physologcal Insttute) runnng under Igor (verson 5.03) on a PC. Morphology of pre- and postsynaptc neuron was vsualzed by post hoc stanng, usng 3,3 0 -damnobenzdne as chromogen (Geger et al., 1997). Solutons For storage of slces, a soluton contanng 87 mm NaCl, 25 mm NaHCO 3,10mM glucose, 75 mm sucrose, 2.5 mm KCl, 1.25 mm NaH 2 PO 4, 0.5 mm CaCl 2, and 7 mm MgCl 2 (equlbrated wth 95% O 2 /5% CO 2 gas mxture) was used. For the experments, the slces were superfused wth physologcal salne contanng 125 mm NaCl, 25 mm NaHCO 3, 25 mm glucose, 2.5 mm KCl, 1.25 mm NaH 2 PO 4, 2 mm CaCl 2 (or mm CaCl 2 n a subset of experments wth replacement by sucrose as requred), and 1 mm MgCl 2 (95% O 2 /5% CO 2 ). The ntracellular soluton for the presynaptc neuron contaned 135 mm Kgluconate, 20 mm KCl, 2 mm MgCl 2, 2 or 0.02 mm Na 2 ATP, 0.5 mm NaGTP, 5 mm phosphocreatne, 10 mm HEPES, and 2 mg ml 1 bocytn. The ntracellular soluton for the postsynaptc neuron contaned 110 mm KCl, 35 mm Kgluconate, 10 mm EGTA, 2 mm MgCl 2, 2 mm Na 2 ATP,10mMHEPES,1mMQX- 314, and 2 mg ml 1 bocytn; the ph was adjusted to 7.2 wth KOH n both cases. In solutons n whch chelators were added, the concentraton of Kgluconate was reduced accordngly. 1,2-Bs(2-amnophenoxy)ethane-N,N,N 0,N 0 -tetraacetc acd (BAPTA) and ethyleneglycol-bs(2-amnoethylether)-n,n,n 0,N 0 -tetraacetc acd (EGTA) were from Sgma, other chemcals were from Merck, Sgma, Redel-de Haën, or Gerbu. Ppette Perfuson Ca 2+ chelators were loaded nto presynaptc BCs by ppette perfuson, usng a custom-made two-port ppette holder and parts of the 2PK+ ppette perfuson kt (ALA Scentfc Instruments, Westbury, NY). Presynaptc ppettes were flled wth a small volume (2 4 ml) of ntracellular soluton. Chelators were appled through one port va a flexble quartz tubng (100 mm outer dameter) coated wth polyamde. To mnmze exchange tmes, the end of the tubng was postoned closely to the ppette tp. The other end was connected to an 0.6 ml reservor wth chelator-contanng soluton. To delver the chelators to the ppette, postve pressure was appled to the reservor and a compensatory negatve pressure was appled at the second port of the ppette holder (whch was also used for sucton durng seal formaton and transton from cell-attached nto whole-cell mode). Both postve and negatve pressure were generated by a pressure/vacuum pump system and controlled precsely usng two ndependent regulators (2 5 mm Hg). Changes n electrode potental durng perfuson were <2 mv for 30 mm of chelator. Recordng tme after the soluton exchange was, on average, 85 ± 5 mn (range: mn). Control experments wth mock applcaton of the orgnal ntracellular soluton revealed that the rundown of evoked IPSCs durng long-lastng recordngs was small (to 80.4% of control value; Fgure 4D and Fgure S1D). Gven the small extent of rundown, no correcton was made. Control experments were also performed to test whether presynaptc chelator concentratons mght be affected by organc anon transporters (OATs). In two BC-GC pars, we tested the effects of ntracellular applcaton of 30 mm EGTA after addng 1 mm of the OAT blocker probenecde to the bath soluton. As the results were smlar to those obtaned n the absence of the drug, data were pooled. Fnally, we tested whether the effects of exogenous chelators could be nfluenced by the expresson of the endogenous Ca 2+ -bndng proten parvalbumn n BCs. In two BC-GC pars from parvalbumn knockout ( / ) mce (Schwaller et al., 1999), we examned the effects of 30 mm EGTA appled va presynaptc patch ppette perfuson. The effects were smlar to those obtaned n wld-type rats (reducton of IPSC peak ampltude to 72.1% of control value). As the results were obtaned from a dfferent speces, data were not ncluded n the mean values. To estmate the concentraton of chelators reached n presynaptc termnals, dffuson from the soma nto the BC axon was modeled wth Neuron 5.9, usng a mechansm based on COMPARTMENT and LONGITUDINAL_DIFFUSION (Carnevale and Hnes, 2006). A prevously descrbed cable model of a dentate gyrus BC wth fully reconstructed somatodendrtc doman and man axon and schematcally extended axonal arborzaton (150 collaterals) was used (Geger et al., 1997). The dffuson coeffcent for BAPTA and EGTA was assumed as 220 mm 2 s 1 (Naragh and Neher, 1997) and the tme step was 100 ms. These smulatons ndcated that the concentraton at presynaptc termnals was close to that at the soma after mn recordng tmes for proxmal collaterals (Fgure 2D). Data Analyss Data analyss was performed usng Mathematca 4.1.2, 5.0, or (Wolfram Research). Untary IPSC peak ampltude was measured as the dfference between baselne and the peak current <8 ms after the frst presynaptc acton potental. Statonarty of IPSC ampltudes was tested by Spearman rank correlaton analyss wth p > 0.1; only pars fulfllng the stablty crteron were used for subsequent analyss. Events were classfed as successes when the ampltude was >3 standard devatons of the baselne nose and as falures otherwse. Coeffcents of varaton (CV, standard devaton/mean) of IPSC peak ampltudes were calculated from 30 traces durng statonary control and test perods (Kraushaar and Jonas, 2000). Modelng Chelator Effects wth the Steady-State Soluton to the Lnearzed Reacton-Dffuson Equatons Effects of Ca 2+ chelators on synaptc transmsson were modeled usng the steady-state soluton to the lnearzed reacton-dffuson problem obtaned analytcally (Neher, 1998). Ths approach assumed that buffer saturaton s neglgble. A vescle carryng the Ca 2+ sensor was ether coupled to a sngle Ca 2+ channel at varable dstance r (sngle-channel model) or to a cluster of Ca 2+ channels at varable dstance d (cluster model). In the cluster model, d was measured from the center of the cluster, and the wdth of the cluster was quantfed as the standard devaton s of the normally dstrbuted channel densty. The rato of Ca 2+ transents n the presence and absence of chelators (R Ca ) was calculated as R Ca = ð1=r expð r=l b ÞÞ=ð1=r expð r=l o ÞÞ ðsngle channelþ (1) ð ð R Ca = n Ca ðrþ expð r=l b Þdr n Ca ðrþ expð r=l o Þdr ðclusterþ wth (2) l o = OD Ca = k o on ½BŠo and (3) h l b = O D Ca = k oon½bšo + k bon BbT KbD = Ca K b D ; (4) r where r s radal dstance, n Ca (r) s the Ca 2+ channel densty dstrbuton,! s a defnte ntegral (ntegraton range nm), D Ca s the dffuson coeffcent of free Ca 2+,k o on [B] o s the buffer product of endogenous buffer, k b on s the rate constant of Ca 2+ bndng to exogenous buffer, [B] b T s the total concentraton of exogenous buffer, K b D s the dssocaton constant of exogenous buffer, and [Ca 2+ ] r s the restng Ca 2+ concentraton. Physcochemcal propertes of Ca 2+ and chelators were assumed as follows (Naragh and Neher, 1997; Neher, 1998; Nägerl et al., 2000): D Ca = 220 mm 2 s 1,k b on =410 8 M 1 s 1 for BAPTA and M 1 s 1 for EGTA, and K b D = 0.22 mm for BAPTA and 0.07 mm for EGTA. Endogenous bufferng propertes of BCs were constraned based on measurement of Ca 2+ transents n proxmal dendrtes of BCs (Y. Aponte et al., 2006, Soc. Neurosc., abstract). [Ca 2+ ] r was taken as the value prevously measured wth 100 mm fura-2 (71 nm), and k o on [B] o was determned as k S /t (see Naragh and Neher, 1997) = 202/0.2 s 1 = 1010 s 1, where k S s the endogenous Ca 2+ bndng rato of BCs (202) and t s the fast decay tme constant of the Ca 2+ transents recorded wth low concentratons of fura-2 or fura- FF (200 ms; Y. Aponte, J. Bschofberger, and P.J., unpublshed data; see Müller et al., 2007). The rato of IPSCs n the presence and absence of Neuron 57, , February 28, 2008 ª2008 Elsever Inc. 543

9 chelators (R IPSC ) was then calculated as R IPSC = f(r Ca ), where f was the doubly normalzed verson of the power relaton between transmtter release and extracellular [Ca 2+ ] at the BC-GC synapse (concentraton normalzed to 2 mm and release normalzed to that at 2 mm; Fgure 5C). Confdence ntervals of r, d, and s were obtaned by bootstrap procedures. One thousand artfcal data sets were generated from the means and SEMs of the orgnal data set and were analyzed as the orgnal (Efron and Tbshran, 1998). Bootstrap procedures were also used to compare sngle-channel and cluster model wth models n whch Ca 2+ channels were dstrbuted more homogeneously (Fgure 6B). Tme-Dependent Soluton to the Full Reacton-Dffuson Equatons and Smulaton of Transmtter Release To account for both possble devatons from the steady-state assumpton and saturaton of buffers, the tme-dependent soluton to the full reacton-dffuson equatons was obtaned numercally. Ths approach consdered moble and fxed buffers as well as buffer saturaton. Assumng Ca 2+ nflow at a pont source and radal symmetry, the Ca 2+ concentraton as a functon of tme and radal dstance can be obtaned from a set of partal dfferental equatons (Smth, 2001):! v hca v hca = D Ca r 2v 2 + X k vt r 2 on; hca ½B 2 + Š k off; ½B Š vr vr tot ½B Š (5a) and v½b Š 1 v = D B; r 2v½B Š k vt r 2 on; hca 2 + ½B Š k off; ½B Š vr vr tot ½B Š ; (5b) where r s radal dstance, t s tme, [Ca 2+ ] s the free Ca 2+ concentraton, and D Ca s the dffuson coeffcent for Ca 2+. Furthermore, [B ] s the free buffer concentraton, k on, the bndng rate, k off, the unbndng rate, [B ] tot the total buffer concentraton, and D B, the dffuson coeffcent of the th buffer. The boundary condtons near the source (correspondng to the regon of the actve zone) were gven as! hca lm 4pD Ca r 2v 2 + = 2sðtÞ for r/r mn and (6a) vr lm 4pD B; r 2v½B Š = 0 for r/r mn ; (6b) vr where s(t) s the flux through the Ca 2+ channels as a functon of tme, whch was specfed accordng to prevous models of P/Q-type Ca 2+ channels (Borst and Sakmann, 1998). The flux was multpled by 2 to convert sphercal nto hemsphercal symmetry. The peak ampltude was set to 1 pa, correspondng to <10 P/Q-type Ca 2+ channels (L et al., 2007). Ths ampltude value was chosen to approxmately reproduce the expermentally observed relaton between transmtter release and external Ca 2+ concentraton (Fgure 5C) for a couplng dstance of 20 nm. The boundary condtons remote from the source (correspondng to the outer plasma membrane of the presynaptc termnal) were specfed as! v hca 2 + lm = 0 for r/r max and (7a) vr v½b Š lm = 0 for r/r max : (7b) vr As the gradent s 0, no dffuson s possble beyond the outer boundary. Intal condtons were gven as [Ca 2+ ] = 50 nm, wth [B ] beng set to the correspondng equlbrum values. r mn and r max were chosen as 1 nm and 500 nm, respectvely. The partal dfferental equaton was solved usng NDSolve of Mathematca 6.0.1, wth a spatal grd resoluton of <0.1 nm. Increasng the grd resoluton by a factor of 3 had only mnmal effects on the results. Dffuson coeffcents were chosen D Ca =D B = 220 mm 2 s 1. The affntes of moble buffers (ATP) and endogenous fxed buffers were assumed as 200 and 2 mm, respectvely, and the bndng rates for Ca 2+ were chosen as M 1 s 1 n both cases (Menrenken et al., 2002). The concentraton of moble buffers was assumed as 290 mm (correspondng to 2 mm Mg ATP n the ppette soluton; Menrenken et al., 2002). The concentraton of the fxed buffers was assumed as 160 mm, to account for the relatvely hgh endogenous Ca 2+ bndng rato of BCs (Y. Aponte et al., 2006, Soc. Neurosc., abstract). The rato of IPSCs was calculated as R IPSC = f(r Ca ), where R Ca s the rato of peak ampltudes of Ca 2+ transents. To smulate transmtter release for Ca 2+ transents at varous dstances from the source, we used the prevously publshed allosterc model of Ca 2+ -dependent vescle fuson (Lou et al., 2005). The model conssted of 12 states, 6 before and 6 after fuson. The Ca 2+ bndng rate was assumed as k on = M 1 s 1, the unbndng rate as k off = 4000 s 1, and the basal fuson rate as l + =210 4 s 1. On and off rates were multpled by nteger numbers between 1 and 5 to mplement ndependent Ca 2+ bndng at dfferent stes. The cooperatvty factors were chosen as b = 0.5 and f = 31.3 (Lou et al., 2005). The occupances for the dfferent states of the model were obtaned by solvng the correspondng frst-order ordnary dfferental equaton wth a Q matrx approach. Fnally, the release rate was obtaned by dfferentaton of the sum of occupances of all fused states. Statstcal Analyss Values are gven as means ± SEMs. Error bars n fgures also ndcate SEMs whenever they exceed the sze of the symbols. Sgnfcance of dfferences was assessed by a nonparametrc two-sded Wlcoxon sgned rank test at the sgnfcance level (p) ndcated. SUPPLEMENTAL DATA The Supplemental Data for ths artcle can be found onlne at neuron.org/cg/content/full/57/4/536/dc1/. ACKNOWLEDGMENTS We thank Dr. J. Bschofberger for crtcally readng an earler verson of the manuscrpt; Dr. Arnd Roth for useful dscussons; and S. Becherer, M. Northemann, and K. Wnterhalter for techncal assstance. Supported by the Deutsche Forschungsgemenschaft (SFB 505, SFB 780, and Lebnz program). Receved: May 9, 2007 Revsed: November 16, 2007 Accepted: December 26, 2007 Publshed: February 27, 2008 REFERENCES Adler, E.M., Augustne, G.J., Duffy, S.N., and Charlton, M.P. (1991). Alen ntracellular calcum chelators attenuate neurotransmtter release at the squd gant synapse. J. Neurosc. 11, Augustne, G.J., Santamara, F., and Tanaka, K. (2003). Local calcum sgnalng n neurons. Neuron 40, Awatraman, G.B., Prce, G.D., and Trussell, L.O. (2005). Modulaton of transmtter release by presynaptc restng potental and background calcum levels. Neuron 48, Bartos, M., Vda, I., and Jonas, P. (2007). Synaptc mechansms of synchronzed gamma oscllatons n nhbtory nterneuron networks. Nat. Rev. Neurosc. 8, Baude, A., Nusser, Z., Roberts, J.D., Mulvhll, E., McIlhnney, R.A., and Somogy, P. (1993). The metabotropc glutamate receptor (mglur1a) s concentrated at persynaptc membrane of neuronal subpopulatons as detected by mmunogold reacton. Neuron 11, Berkefeld, H., Saler, C.A., Bldl, W., Rohde, V., Thumfart, J.O., Eble, S., Klugbauer, N., Resnger, E., Bschofberger, J., Olver, D., et al. (2006). 544 Neuron 57, , February 28, 2008 ª2008 Elsever Inc.

10 BK Ca -Cav channel complexes medate rapd and localzed Ca 2+ -actvated K + sgnalng. Scence 314, Bezprozvanny, I., Zhong, P., Scheller, R.H., and Tsen, R.W. (2000). Molecular determnants of the functonal nteracton between syntaxn and N-type Ca 2+ channel gatng. Proc. Natl. Acad. Sc. USA 97, Borst, J.G.G., and Sakmann, B. (1996). Calcum nflux and transmtter release n a fast CNS synapse. Nature 383, Borst, J.G.G., and Sakmann, B. (1998). Calcum current durng a sngle acton potental n a large presynaptc termnal of the rat branstem. J. Physol. 506, Brandt, A., Khmch, D., and Moser, T. (2005). Few Ca V 1.3 channels regulate the exocytoss of a synaptc vescle at the har cell rbbon synapse. J. Neurosc. 25, Carnevale, N.T., and Hnes, M.L. (2006). The Neuron Book (Cambrdge: Cambrdge Unversty Press). Colln, T., Chat, M., Lucas, M.G., Moreno, H., Racay, P., Schwaller, B., Marty, A., and Llano, I. (2005). Developmental changes n parvalbumn regulate presynaptc Ca 2+ sgnalng. J. Neurosc. 25, Efron, B., and Tbshran, R.J. (1998). An Introducton to the Bootstrap (London: Chapman and Hall/CRC). Fedchyshyn, M.J., and Wang, L.Y. (2005). Developmental transformaton of the release modalty at the calyx of Held synapse. J. Neurosc. 25, Freund, T.F., and Katona, I. (2007). Persomatc nhbton. Neuron 56, Geger, J.R.P., Lübke, J., Roth, A., Frotscher, M., and Jonas, P. (1997). Submllsecond AMPA receptor-medated sgnalng at a prncpal neuron-nterneuron synapse. Neuron 18, Hefft, S., and Jonas, P. (2005). Asynchronous GABA release generates longlastng nhbton at a hppocampal nterneuron prncpal neuron synapse. Nat. Neurosc. 8, Iwasak, S., and Takahash, T. (1998). Developmental changes n calcum channel types medatng synaptc transmsson n rat audtory branstem. J. Physol. 509, Jonas, P., Bschofberger, J., Frcker, D., and Mles, R. (2004). Interneuron dversty seres: Fast n, fast out - temporal and spatal sgnal processng n hppocampal nterneurons. Trends Neurosc. 27, Kraushaar, U., and Jonas, P. (2000). Effcacy and stablty of quantal GABA release at a hppocampal nterneuron-prncpal neuron synapse. J. Neurosc. 20, Kulk, A., Nakadate, K., Hagwara, A., Fukazawa, Y., Luján, R., Sato, H., Suzuk, N., Futatsug, A., Mkoshba, K., Frotscher, M., and Shgemoto, R. (2004). Immunocytochemcal localzaton of the a 1A subunt of the P/Q-type calcum channel n the rat cerebellum. Eur. J. Neurosc. 19, Lee, S.-H., Rosenmund, C., Schwaller, B., and Neher, E. (2000). Dfferences n Ca 2+ bufferng propertes between exctatory and nhbtory hppocampal neurons from the rat. J. Physol. 525, L, L., Bschofberger, J., and Jonas, P. (2007). Dfferental gatng and recrutment of P/Q-, N-, and R-type Ca 2+ channels n hppocampal mossy fber boutons. J. Neurosc. 27, Lou, X., Scheuss, V., and Schneggenburger, R. (2005). Allosterc modulaton of the presynaptc Ca 2+ sensor for vescle fuson. Nature 435, Matsu, K., and Jahr, C.E. (2004). Dfferental control of synaptc and ectopc vescular release of glutamate. J. Neurosc. 24, Menrenken, C.J., Borst, J.G.G., and Sakmann, B. (2002). Calcum secreton couplng at calyx of Held governed by nonunform channel-vescle topography. J. Neurosc. 22, Mochda, S., Westenbroek, R.E., Yokoyama, C.T., Zhong, H., Myers, S.J., Scheuer, T., Itoh, K., and Catterall, W.A. (2003). Requrement for the synaptc proten nteracton ste for reconsttuton of synaptc transmsson by P/Q-type calcum channels. Proc. Natl. Acad. Sc. USA 100, Müller, M., Felmy, F., Schwaller, B., and Schneggenburger, R. (2007). Parvalbumn s a moble presynaptc Ca 2+ buffer n the calyx of Held that accelerates the decay of Ca 2+ and short-term facltaton. J. Neurosc. 27, Nägerl, U.V., Novo, D., Mody, I., and Vergara, J.L. (2000). Bndng knetcs of calbndn-d 28k determned by flash photolyss of caged Ca 2+. Bophys. J. 79, Naragh, M., and Neher, E. (1997). Lnearzed buffered Ca 2+ dffuson n mcrodomans and ts mplcatons for calculaton of [Ca 2+ ] at the mouth of a calcum channel. J. Neurosc. 17, Neher, E. (1998). Usefulness and lmtatons of lnear approxmatons to the understandng of Ca ++ sgnals. Cell Calcum 24, Ohana, O., and Sakmann, B. (1998). Transmtter release modulaton n nerve termnals of rat neocortcal pyramdal cells by ntracellular calcum buffers. J. Physol. 513, Poncer, J.C., McKnney, R.A., Gähwler, B.H., and Thompson, S.M. (1997). Ether N- or P-type calcum channels medate GABA release at dstnct hppocampal nhbtory synapses. Neuron 18, Poulle, F., and Scanzan, M. (2001). Enforcement of temporal fdelty n pyramdal cells by somatc feed-forward nhbton. Scence 293, Poulle, F., and Scanzan, M. (2004). Routng of spke seres by dynamc crcuts n the hppocampus. Nature 429, Red, C.A., Bekkers, J.M., and Clements, J.D. (2003). Presynaptc Ca 2+ channels: a functonal patchwork. Trends Neurosc. 26, Rettg, J., Sheng, Z.H., Km, D.K., Hodson, C.D., Snutch, T.P., and Catterall, W.A. (1996). Isoform-specfc nteracton of the a 1A subunts of bran Ca 2+ channels wth the presynaptc protens syntaxn and SNAP-25. Proc. Natl. Acad. Sc. USA 93, Rozov, A., Burnashev, N., Sakmann, B., and Neher, E. (2001). Transmtter release modulaton by ntracellular Ca 2+ buffers n facltatng and depressng nerve termnals of pyramdal cells n layer 2/3 of the rat neocortex ndcates a target cell-specfc dfference n presynaptc calcum dynamcs. J. Physol. 531, Schwaller, B., Dck, J., Dhoot, G., Carroll, S., Vrbova, G., Ncotera, P., Pette, D., Wyss, A., Bluethmann, H., Hunzker, W., and Celo, M.R. (1999). Prolonged contracton-relaxaton cycle of fast-twtch muscles n parvalbumn knockout mce. Am. J. Physol. 276, C395 C403. Shahrezae, V., Cao, A., and Delaney, K.R. (2006). Ca 2+ from one or two channels controls fuson of a sngle vescle at the frog neuromuscular juncton. J. Neurosc. 26, Sheng, Z.H., Yokoyama, C.T., and Catterall, W.A. (1997). Interacton of the synprnt ste of N-type Ca 2+ channels wth the C2B doman of synaptotagmn I. Proc. Natl. Acad. Sc. USA 94, Smth, G.D. (2001). Modellng local and global calcum sgnals usng reactondffuson equatons. In Computatonal Neuroscence, E. de Schutter, ed. (Boca Raton, FL: CRC Press), pp Stanley, E.F. (1993). Sngle calcum channels and acetylcholne release at a presynaptc nerve termnal. Neuron 11, Stanley, E.F. (1997). The calcum channel and the organzaton of the presynaptc transmtter release face. Trends Neurosc. 20, Südhof, T.C. (2002). Synaptotagmns: why so many? J. Bol. Chem. 277, Tang, Y., Schlumpberger, T., Km, T., Lueker, M., and Zucker, R.S. (2000). Effects of moble buffers on facltaton: expermental and computatonal studes. Bophys. J. 78, Verdero, C., Pozz, D., Pravetton, E., Inverard, F., Schenk, U., Coco, S., Proux-Gllardeaux, V., Gall, T., Rossetto, O., Frasson, C., and Matteol, M. (2004). SNAP-25 modulaton of calcum dynamcs underles dfferences n GABAergc and glutamatergc responsveness to depolarzaton. Neuron 41, Wlson, R.I., Kunos, G., and Ncoll, R.A. (2001). Presynaptc specfcty of endocannabnod sgnalng n the hppocampus. Neuron 31, Zucker, R.S., and Regehr, W.G. (2002). Short-term synaptc plastcty. Annu. Rev. Physol. 64, Neuron 57, , February 28, 2008 ª2008 Elsever Inc. 545