Distribution of the SNAP25 and SNAP23 synaptosomal-associated protein isoforms in rat cerebellar cortex

Page 1

DA

GGPa

0b

I

A(snSsieatlccopecVSpihntesdrE

Kc

Ss1

*d�EAccfssmg

Neuroscience 164 (2009) 1084–1096

0d

ISTRIBUTION OF THE SNAP25 AND SNAP23 SYNAPTOSOMAL-

SSOCIATED PROTEIN ISOFORMS IN RAT CEREBELLAR CORTEX

ita

fvmcwapvaeoee

aWt

piadeSBgfic

pfchuPos

cC2ira

. MANDOLESI,a1* V. VANNI,a1 R. CESA,b

. GRASSELLI,a F. PUGLISI,a P. CESAREa,b AND. STRATAa,b

EBRI-Santa Lucia Foundation (IRCCS), Via del Fosso di Fiorano 64,0143 Rome, Italy

Department of Neuroscience and National Institute of Neuroscience-taly, University of Turin, Corso Raffello 30, 10125 Turin, Italy

bstract—Synaptosome-associated protein of 25 kDaSNAP25) is a component of the fusion complex that mediatesynaptic vesicle exocytosis, regulates calcium dynamics andeuronal plasticity. Despite its crucial role in vesicle release,NAP25 is not distributed homogenously within the brain. Iteems to be virtually absent in mature inhibitory terminals and

s observed in a subtype of excitatory neurons defined by thexpression of vesicular glutamate transporter 1 (VGluT1). Sincecomplementary distribution of VGluT1 and VGluT2 in excita-

ory synapses is correlated with different probabilities of re-ease (Pr), we evaluated whether SNAP25 localization is asso-iated with specific synaptic properties. In the cerebellum,limbing fiber (CF) and parallel fiber (PF) inputs, which impingento the same Purkinje cell (PC), have very different functionalroperties. In the cerebellum of adult rats, using confocal andlectron microscopy, we observed that VGluT2-positive CFs,haracterized by a high Pr, only weakly express SNAP25, whileGluT1-positive PFs that show a low Pr abundantly expressNAP25. Moreover, SNAP25 was less profuse in the VGluT2-ositive rosettes of mossy fibers (MFs) and was almost absent

n inhibitory terminals. We extended our analysis to the SNAP23omolog; this is expressed at different levels in both �-ami-obutyric acid-containing terminals (GABAergic) and glutama-

ergic terminals of the cerebellar cortex. In conclusion, the pref-rential localization of SNAP25 in specific synaptic boutonsuggests a correlation between SNAP25 and the Pr. This evi-ence supports the hypothesis that SNAP25 has a modulatoryole in shaping synaptic responses. © 2009 IBRO. Published bylsevier Ltd. All rights reserved.

ey words: SNAP-isoform, VGluT1, VGluT2, VGAT, cerebellarortex, probability of release.

ynaptic functional diversification can be explained by thetructural organization of the synapse and by its special-

These two authors contributed equally to this work.Correspondence to: G. Mandolesi, Santa Lucia Foundation (IRCCS), Viael Fosso di Fiorano 64, 00143 Rome, Italy. Tel: �39-06-501703210; fax:39-06-501703327.-mail address: [email protected] (G. Mandolesi).bbreviations: ANOVA, analysis of variance; CF, climbing fiber; CTR,ontrol; MF, mossy fiber; PBS, phosphate buffer solution; PC, Purkinjeell; PF, parallel fiber; PPD, paired pulse depression; PPF, paired pulseacilitation; Pr, release probability; PSD, postsynaptic density; SNAP,ynaptosome-associated protein; SNAREs, soluble N-ethylmaleimideensitive factor attachment protein receptors; VAMP, vesicle-associated

2embrane protein; VGAT, vesicular GABA transporter; VGluT, vesicularlutamate transporter.

306-4522/09 $ - see front matter © 2009 IBRO. Published by Elsevier Ltd. All rightoi:10.1016/j.neuroscience.2009.08.067

1084

zed molecular differentiation. Presynaptic mechanismshat affect neurotransmitter release are important in syn-ptic diversification (Atwood and Karunanithi, 2002).

Among the presynaptic molecules that are candidatesor participating in synaptic performance are synaptobre-in, a vesicle-associated membrane protein (VAMP), plasmaembrane-associated syntaxin 1 and synaptosome-asso-

iated protein of 25 kDa (SNAP25) (Thomson, 2000; At-ood and Karunanithi, 2002). These proteins, also knowns soluble N-ethylmaleimide sensitive factor attachmentrotein receptors (SNAREs), are essential mediators ofesicle fusion and exocytosis (Söllner et al., 1993; Chennd Scheller, 2001; Jahn and Scheller, 2006). Increasingvidence suggests that their spatial and temporal patternsf expression play significant roles in SNARE-mediatedxocytosis and in modulating synaptic transmission (Barkt al., 2004; Matteoli et al., 2009).

In this study, we describe the distribution of SNAP25nd its homolog, SNAP23, in the adult cerebellar cortex.e looked for a correlation between their localization and

he synaptic properties of different neuronal populations.SNAP25 is a membrane-bound protein anchored to the

lasma membrane of neurons via palmitoyl side chains. Its present in two isoforms, “a” and “b”, resulting fromlternative splicing of its pre-mRNA transcript. They areifferentially distributed in the brain; SNAP25a is mostlyxpressed at early stages of development while theNAP25b is highly expressed by adults (Bark et al., 1995;oschert et al., 1996; Oyler et al., 1991). It has been sug-ested that the developmental switch between SNAP25 iso-

orms alters the efficacy of synaptic transmission, which mayn turn contribute to the consolidation of developing neuronalircuitry (Bark et al., 2004).

In addition to its involvement in the formation of theore complex for vesicle exocytosis, SNAP25 has otherunctions. SNAP25 negatively controls the neuronal cal-ium responsiveness to depolarization by specifically in-ibiting neuronal voltage-gated calcium channels (VGCCs)pon phosphorylation of Ser187 (Verderio et al., 2004;ozzi et al., 2008). In summary, the difference in the levelf SNAP25 expression impacts on calcium dynamics andhort-term neuronal plasticity (Pozzi et al., 2008).

SNAP25 is not present in all synaptic terminals of theentral and peripheral nervous system (Oyler et al., 1989;atsicas et al., 1992; Hellstrom et al., 1999; Morris et al.,000; Gibbins et al., 2003). A virtual absence of SNAP25

mmunoreactivity in mature inhibitory synapses has beenecently reported in several brain regions of both rodentsnd humans, including the cerebellum (Verderio et al.,

004; Frassoni et al., 2005; Bragina et al., 2007; Garbelli et

s reserved.

Page 2

aepdmscSt

mSnTSs

m(eea2baw

StdmSs1n1cenM

pe2paSSwn

ailfssal

stcad1mwcttVaGsgaet

A

Wrh(cMa

I

Rp0dscdf�mGmnwdC(umm1mCcsasPlat

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–1096 1085

l., 2008). However, Tafoya et al. (2006) reported a highxpression of SNAP25 in GABAergic terminals of the hip-ocampus and the thalamus. Therefore, there is still aebate about SNAP25 expression in the GABAergic ter-inals of mature brain (Matteoli et al., 2009). Here, we

upport the results of Garbelli et al. (2008) by means ofomplementary experiments. They reported also a lack ofNAP25 immunoreactivity in the Purkinje cell (PC) synap-

ic terminals that contact the deep cerebellar nuclei.On the other hand, different observations have unani-

ously indicated that, at early developmental stages,NAP25 is functionally required at most GABAergic termi-als (Bronk et al., 2007; Delgado-Martinez et al., 2007;afoya et al., 2006) in accordance with the pattern ofNAP25 expression during neuronal differentiation (Fras-oni et al., 2005).

As regards excitatory neurons, SNAP25 has beenainly found in vesicular glutamate transporter-positive

VGluT1-positive) terminals (Verderio et al., 2004; Garbellit al., 2008). SNAP25 has only been characterized in bothxcitatory subtypes in the cerebral cortex, and is virtuallybsent in the VGluT2-positive terminals (Bragina et al.,007). In the cerebellar cortex, its distribution has onlyeen investigated in VGluT1-positive terminals (Garbelli etl., 2008). However, the possible colocalization of SNAP25ith VGluT2 has not been explored.

Among the members of the SNAP family, such asNAP23, SNAP29, SNAP47, we investigated SNAP23. As

he other homologs, SNAP23 is ubiquitously expressed. Itsistribution in the synaptic terminals of the brain has beenore characterized and it seems variably expressed.NAP23 was cloned from human cells and has a strongequence homology to SNAP25 (Ravichandran et al.,996; Wong et al., 1997). It has been identified predomi-antly in non-neuronal, non-endocrine cells (Wong et al.,997; Galli et al., 1998). In vitro, SNAP23 shows bindingharacteristics that are similar to SNAP25 (Ravichandrant al., 1996; Araki et al., 1997). It cannot rescue synchro-ous release in snap-25 null neurons in culture (Delgado-artinez et al., 2007).

Regarding its distribution in the CNS, it has been re-orted that hippocampal and cortical GABAergic neuronsxpress SNAP23 (Verderio et al., 2004; Bragina et al.,007). In cortical neurons, SNAP23 is abundantly ex-ressed in VGluT2-positive terminals, but is poorly detect-ble in VGluT1-positive terminals (Bragina et al., 2007).ince these data suggest a differential expression ofNAP23, which seems to be complementary to SNAP25,e investigated the distribution of SNAP23 in the differenteuronal populations of the cerebellar cortex.

In the cerebellar cortex, there are three main excitatorynd three main inhibitory inputs, which can all be reliably

dentified. Two excitatory synapses are located in the mo-ecular layer and bear remarkably different structural andunctional synaptic properties (Ito, 1984). Both of themupply the same postsynaptic cell, but evoke differentynaptic responses. The parallel fibers (PF) to PC syn-pses are VGluT1-positive. They have one of the lowest

evels of release probability (Pr) in the brain and possess a t

hort-term plasticity characterized by paired pulse facilita-ion (PPF) (Konnerth et al., 1990; Perkel et al., 1990). Inontrast, climbing fiber (CF) synapses are VGluT2-positivend are characterized by the highest Pr and a paired pulseepression (PPD) (Konnerth et al., 1990; Perkel et al.,990). The third excitatory input to the cerebellar cortex isade by the mossy fiber (MF) rosette in the granular layerhere it forms synapses with the dendrites of the granuleells (Eccles et al., 1967). Although structurally similar,hese synapses are actually composed of three distinctypes: those that are positive for VGluT1, those positive forGluT2, and those positive for both. Variation in Pr occurscross MFs (Sargent et al., 2005). Finally, the inhibitoryABAergic synapses are represented by the basket and

tellate cells on the PCs and by the Golgi cells on theranule cells (Eccles et al., 1967). Therefore, by means of

quantitative confocal and ultrastructural analysis, wevaluated the expression of both SNAP25 and SNAP23 byhese different synaptic terminals.

EXPERIMENTAL PROCEDURES

nimals

e used adult Wistar albino rats (Charles River Breeding Labo-atories, Calco, LC, Italy; body weight 150–250 g). Animals wereoused according to the European Community Council Directive86/609/CEE). The experimental protocols were designed in ac-ordance with Italian law DL 116/92 and presented to the Italianinistry of Health. All efforts were made to reduce the number ofnimals used and to minimize their suffering.

mmnohistochemistry

ats were deeply anesthetized with Avertin 0.002 ml/0.01 kg anderfused through the aorta with ice cold 4% paraformaldehyde in.1 M phosphate buffer solution (PBS) at pH 7.4. Tissues wereissected, post fixed for 2 h at 4 °C, and equilibrated with 30%ucrose overnight. Sagittal cerebellar sections (30 �m thick) wereut with a freezing microtome. Sections were blocked with 10%onkey serum solution in PBS containing 0.25% Triton X-100 inor 1 h at room temperature (RT). Then they were incubated at4 °C for 1 day with the following diluted primary antibodies:ouse monoclonal anti-SNAP25 (1:5000; Synaptic SystemmbH, Goettingen, Germany (SYSY) cat. num. 111101) andouse monoclonal anti-SNAP25 from Sternberger (1:1000; cat.um. SMI81; MD, USA; staining pattern identical to that obtainedith the monoclonal antibody from Synaptic Systems (SYSY);ata not shown), rabbit polyclonal anti-SNAP23 (1:100 for 3 days;ovalab, Cambridge, UK, cat. num. pab0057), mouse monoclonal

1:500; SYSY cat. num. 131011) and rabbit polyclonal anti-vesic-lar GABA transporter (VGAT) (1:500; SYSY cat. num. 131002),ouse monoclonal anti- VAMP2 (1:500; SYSY cat. num. 104001),ouse monoclonal anti-syntaxin1 (1:200; SYSY cat. num.

10001), guinea-pig polyclonal anti-VGluT1 (Chemicon, Te-ecula, CA, USA; cat. num. ab5905), mouse monoclonal (1:3000;hemicon cat. num. ab5907) and rabbit polyclonal (1:500; SYSYat. num. 35403) anti-VGluT2. After being washed with PBS, theections were incubated with Cy2-conjugated, Cy-3 conjugatednd Cy-5-conjugated secondary antibodies (Jackson ImmunoRe-earch, West Grove, PA, USA) 1:200 for 2 h at RT and rinsed inBS. Sections were then washed in PBS, mounted on poly-L-

ysine-coated slides and coverslips were mounted with anti-fadinggent (Biomeda, Foster City, CA, USA). To test the specificity ofhe antisera, primary antibodies were substituted with PBS, and

he immunolabeling was absent.

Page 3

W

Ch(M1wna(nfn(scs

C

AsizuwlteFm

acrastre(ATtcmiVhrVfrgt

adWfs

E

Aawi

GtaLotiSntPfsni

obadttaeb

St

TPTPddt(C

mcpls2(tt(at

r3(tPS

S

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–10961086

estern blot

erebellar protein extract was obtained from an adult rat (P32) byomogenizing the cerebellum in a buffer containing 50 mM TrispH 7.5), 300 mM NaCl, 1% Triton X-100, 10% glycerol, 1.5gCl2, 1 mM CaCl2, 1 mM ethylene glycol tetraacetic acid, and% protease inhibitor cocktail (Sigma, Milano, Italy). Crude lysateas centrifuged at 16,000�g for 15 min at 4 °C, and the super-atant was collected. Five �g of cerebellar extract were denatur-ted at 98 °C for 5 min and loaded onto a sodium dodecyl sulfateSDS) polyacrylamide gel (8%). Gels were blotted onto a polyvi-ylidene fluoride (PVDF) membrane. Immunodetection was per-ormed by rabbit polyclonal anti-SNAP23, 1:500 (Covalab, cat.um. pab0057) or mouse monoclonal anti-SNAP25, 1:5000SYSY, cat. num. 111101) and horseradish peroxidize conjugatedecondary antibodies (Chemicon), ECL-Plus reagent (GE Health-are Europe GmbH, Milan, Italy) and the storm 840 acquisitionystem (GE Healthcare Europe GmbH).

onfocal imaging and quantitative analysis

ll images were acquired with a LSM5 Zeiss confocal laser-canning microscope (Zeiss, Göttingen, Germany) using a 63x oilmmersion lens (1.4 numerical aperture) and an additional digitaloom factor of 2x to clearly resolve (0.07�0.07 �m/pixel) individ-al synaptic puncta and dendritic spines. The confocal pinholeas kept at the minimum (1.0), the gain and the offset were

owered to prevent saturation in the brightest signals and sequen-ial scanning for each channel was performed. Three rats fromach experimental group were used for the quantitative analysis.or each cerebellum, we randomly acquired 10–20 images of theolecular and granular layers.

To examine the degree of overlap of the red and green spots,n analysis of pixel colocalization was performed using the colo-alization function of the ImageJ software. The analysis was car-ied out on three animals and z-stacks from each cerebellum werecquired. From the z stacks, the best single section in terms of notaturating the fluorescence intensity was selected. We isolatedhe area of interest (rosettes, CF varicosities and molecular layeregions of 40 �m2) and we performed a colocalization analysis. Tovaluate the colocalization, we used Manders’ overlap coefficientManders et al., 1993): r���i �AixBi��/���i �Ai�2x�i �Bi�2�, where

i and Bi represent signal intensities of pixels for each channel.his coefficient varies from zero to one, the former corresponding

o non-overlapping signals and the latter reflecting 100% of colo-alization between both signals. To assess the reliability of thisethod, we performed a colocalization analysis using the same

mages in the glomeruli localized in the granular layer for: (i)GluT1 and VGluT2 signals as positive control (CTR), which areighly codistributed in the presynaptic terminals of MFs,�0.72�0.02 SE (standard error) (data not shown); (ii) for VGAT-GluT1 as negative CTR, two antigens that are localized in dif-

erent presynaptic terminals (MFs and Golgi synaptic terminals),espectively, but are very close to each other in the cerebellarlomeruli, r�0.114�0.011 SE. An average of coefficients ob-

ained from the examined fields was calculated.A one-way analysis of variance (ANOVA) test was used to

ssess whether Manders’ overlap coefficients were significantlyifferent from those of the corresponding negative CTR group.hen the interaction was significant, a post hoc test was per-

ormed for multiple comparisons. Statistical significance was as-umed when P�0.05.

lectron microscopy

fter perfusion with glutaraldehyde (0.1%), formaldehyde (4%),nd picric acid (0.2%) in 0.12 M PBS, cerebellar slices (500 �m)ere cut by a vibratome (St. Louis, MO, USA) and rapidly frozen

n liquid propane in a cryofixation unit (ASF Auto Leica, Wetzlar, p

ermany), freeze-substituted with methanol in a freeze-substitu-ion apparatus (CS Auto; Wetzlar, Germany) at �80 °C for 36 hnd embedded in methanol/Lowicryl HM20 (Chemische Werkeowi, Waldkraiburg, Germany). Ultrathin sections were collectedn adhesive-coated (Electron Microscopy Sciences, ft. Washing-on, PA, USA) nickel grids (400 mesh) and processed for themmunogold method. Sections were labeled for SNAP25 (1:500,YSY) and for SNAP23 (1:500, Covalab). We observed otheron-labeled structures, such as the PC dendrites, close to in-ensely labeled different structures, such as the cross sectionedFs (Fig. S1A, B). Secondary antibodies (1:20) were goat Fab

ragments coupled to 10 nm colloidal gold particles. To test thepecificity of the antisera, primary antibodies were substituted withormal serum or Tris-buffered saline tween-20 (TBST), and the

mmunolabeling was absent (Fig. S1C, D).Mean densities of SNAP proteins were quantified on the basis

f the number of immunogold particles, representing antiseruminding sites per area of presynaptic terminal and per length of thective zone presynaptic membrane. The presynaptic terminal wasefined as the enlarged portion of the axon which contains clus-ers of synaptic vesicles. The area of PFs, CFs and GABAergicerminals was automatically calculated by the Analysis softwarefter tracing the outline of the presynaptic structure. We consid-red the active zone to be the portion of the presynaptic mem-rane opposite the postsynaptic density (PSD).

RESULTS

NAP25 expression in the glutamatergic synapses ofhe mature cerebellar cortex

he first aim was to analyze the two excitatory inputs to theCs in the molecular layer of the mature cerebellar cortex.he PF and CF inputs contact distinct compartments of theC dendritic arbor, namely the distal and proximal den-ritic regions, respectively. Each input is characterized byistinct electrophysiological properties (Ito, 1984) and byhe complementary expression of the VGluT isoformsFremeau et al., 2001); the PFs express VGluT1 while theFs are VGluT2-positive.

We performed double immunofluorescence experi-ents by incubating cerebellar sagittal sections with spe-

ific antibodies against SNAP25 and VGluT1 or VGluT2roteins. In the rat, the PFs, the axons of the granule cells

ocated in the granular layer, form approximately 200,000ynaptic contacts with the spiny branchlets (Carulli et al.,004). These presynaptic contacts express both VGluT1red, Fig. 1B) and SNAP25 (green, Fig. 1C). In Fig. 1A, Dhe yellow color indicates the strong colocalization of thewo proteins. SNAP25 is also localized at extrasynaptic sitesgreen, Fig. 1D). As we shall demonstrate later, SNAP25 islso present along the axolemma of a high number of PFs in

he molecular layer.The CF is the terminal arbor of the inferior olive neu-

ons. Each PC receives only one CF, which carries about00 varicosities that are specifically labeled by VGluT2red, Fig. 1E, F, H). Each varicosity makes synaptic con-acts with between two and six spines (Palay and Chan-alay, 1974). In these glutamatergic terminals (red),NAP25 (green) was virtually absent (Fig. 1E, G, H).

To determine whether the two other components of theNARE complex were present in the CF terminals, we

erformed a double immunofluorescence study for VGluT2

Page 4

acVSrae

qfoMscmo

Tcc(CSwtasHCSpd

FdPiDir(bw(i C layer.F

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–1096 1087

nd syntaxin1 (Fig. 1I) or VAMP2 (Fig. 1J). We observed alear colocalization between these proteins and theGluT2 marker. The specificity of the antibody for anti-NAP25 was tested by Western blotting analysis on adult

at cerebellum protein extract. As showed in Fig. 1K thentibody recognizes only one intense band with an appar-nt molecular weight of about 25 kDa.

To further support these observations, we performed auantification of pixel colocalization of the immunostainingor SNAP25 and the VGluT-isoforms. We calculated theverlap coefficient (r) in accordance with the study byanders et al. (1993). Complete colocalization corre-

ponds to a value of r equal to one, while zero indicates noolocalization. We used the colocalization index of twoarkers as negative CTR: VGAT and VGluT1 at the level

ig. 1. SNAP25 expression in the glutamatergic synaptic terminals of tistribution (green) in the molecular layer (ml) of a cerebellar sagittalFs (red); the merged image is showed in A. High magnification of the

mmunostaining respectively; D is the merged image. The green labelinouble immunostaining illustrating the virtual absence of SNAP25 (gre

mage is showed in E. A high magnification of the red box in E is shespectively; H is the merged image. I and J are merged images of dogreen, J) in the VGluT2-positive CF varicosity (red). Both componentsy its strong colocalization with VGluT2 (yellow), and also in PF terminith an anti-SNAP25 antibody. (L) Colocalization analysis, expressed b

black column, VGluT2) and PF (white columnn, VGluT1) terminals. Thn the glomerular structure of the granular layer (gray column). Pc, P–J�5 �m.

f the same glomerular structure in the granular layer. i

hey label two different synaptic terminals, which are verylose to each other: the inhibitory terminals of the Golgiells and the excitatory terminals of the MFs respectivelyChaudhry et al., 1998). The r index of such a negativeTR was 0.114 (�0.011 SE, n�55). The r index betweenNAP25 and VGluT2 (S25/V2; 0.173�0.011 SE, n�248)as not significantly different from the negative CTR. On

he contrary, the colocalization index between SNAP25nd VGluT1 (S25/V1; 0.669�0.019 SE; n�40 areas) wasignificantly different (one-way ANOVA, P�0.001; post-hocolm–Sidak test P�0.05 CTR versus VGluT1 and P�0.05TR versus VGluT2) (Fig. 1L). These results show thatNAP25 is differentially expressed in the two excitatory in-uts contacting the PCs. Of note, the presence of SNAP25 isetected in the PF terminals while no detectable expression

ellar molecular layer. (A–D) Double immunostaining showing SNAP25ith a specific localization in VGluT1-positive presynaptic terminals ofA is represented in panels B–D; B and C show VGluT1 and SNAP25

oes not colocalize indicates extrasynaptic SNAP25 localization. (E–H)e CF terminal labeled with an anti-VGluT2 (red) antibody; the mergedpanels (F–H); F and G show VGluT2 and SNAP25 immunostaining,unostaining showing the expression of sintaxin 1 (green, I) or VAMP2NARE complex are abundantly expressed in CF terminals, as shownmarked). (K) Western blotting of adult cerebellar homogenate stainednders’ coefficient, between SNAP25 and the VGluT markers in the CFe CTR is represented by the colocalization index of VGAT and VGluT1*** P�0.001. Error bars indicate SE. Scale bars: A, E 20 �m; B–D,

he cerebsection wred box ing which den) in thowed inuble imm

of the Sals (noty the Ma

e negativ

s observed in the CF terminals.

Page 5

Mtrctdcragbw(orbSCt(2sg00t

dAttw(P

tci

St

Tth(VFtsFdt

Fe(CeawV y columnp Scale ba

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–10961088

We then examined the expression of SNAP25 in theFs. They originate from several brain regions and along

heir extension into the cerebellar granular layer they giveise to a number of “en passant” presynaptic structuresalled rosettes, which provide the major excitatory input tohe cerebellar granule cells. Each rosette contacts theendrites of as many as 50 granule cells. Each granuleell, characterized by an average of four short dendrites,eceives innervations from four different rosettes (Eccles etl., 1967) producing a large glomerular structure. Threeroups of typical rosettes have been characterized on theasis of VGluT expression (Hioki et al., 2003). Accordingly,e identified one group that was only positive for VGluT1

Fig. 2A–E; V1-rosette, red arrowheads in Fig. 2B), a sec-nd group positive only for VGluT2 (V2-rosette, blue ar-owheads in Fig. 2C) and a third fraction that expressedoth markers (V1/V2-rosette, white arrow in Fig. 2B, C).NAP25 was detected in all the rosettes relative to theTR group, but with different levels of expression. In par-

icular, this protein was highly present in the V1-rosettesFig. 2B, D, red arrow) when compared to V2-rosettes (Fig.C, D, blue arrow). Accordingly, the colocalization resultshow a significant difference between the experimentalroups and the CTR group (Fig. 2F), (V2-rosettes.29�0.019 SE; V1-rosettes 0.59�0.024 SE; CTR.114�0.011 SE, one-way ANOVA, P�0.001). In addition,

ig. 2. SNAP25 expression in the glutamatergic synaptic terminals oxpression (green) in the granular layer of a cerebellar sagittal sectionred) and anti-VGluT2 (blue) antibodies; the merged image is showed

and D show VGluT1, VGluT2 and SNAP25 immunostaining, respectixpression in rosettes positive only for VGluT1 (B, E; red arrows), onlrrows). (F) Colocalization analysis, expressed by the Manders’ coeffichich are VGluT2-positive (black column) and/or VGluT1-positive (wGAT and VGluT1 in the glomerular structure of the granular layer (graositive for VGluT2. ** P�0.01; *** P�0.001. Error bars indicate SE.

he r index of the V2-rosettes (n�69) was significantly (

ifferent from the V1-rosettes index (n�41) (one-wayNOVA, P�0.001; post-hoc Holm-–Sidak test P�0.05). In

he V1/V2 rosettes (n�63), the colocalization index be-ween SNAP25 and VGluT1 was 0.52 (�0.013 SE), andas 0.47 between SNAP25 and VGluT2 (�0.046 SE)

one-way ANOVA, P�0.001; post-hoc Holm–Sidak test�0.05).

Altogether these results show that SNAP25 is differen-ially expressed in the excitatory terminals of the cerebellarortex. In particular, a low or null immunoreactivity appears

n the VGluT2 positive terminals.

NAP23 expression in the glutamatergic synapses ofhe mature cerebellar cortex

he differential expression of SNAP25 in the glutamatergicerminals prompted us to characterize the distribution of itsomolog, SNAP23. As shown in Fig. 3A–D, SNAP23green in Fig. 3C) is expressed in the molecular layer byGluT1-positive PFs (red in Fig. 3B). The protein (green inig. 3E, G) is also present in the VGluT2-positive CF

erminals (red in Fig. 3E, F), and their colocalization ishown in Fig. 3H. A quantitative assessment is shown inig. 3O. Both colocalization coefficients were significantlyifferent from the negative CTR group (0.114�0.011 SE);he r index between SNAP23 and VGluT2 was 0.346

bellar granular layer. (A–E) Triple immunostaining showing SNAP25cular in the presynaptic terminals of MF labeled with the anti-VGluT1igh magnification of the red box in A is represented in panels B-E; B,a merged image of all markers. The arrowheads in D indicate SNAP25uT2 (C, E; blue arrows) and expressing both markers (B, C, E; whiteeen SNAP25 and the VGluT markers in the three groups of rosettes,n). The negative CTR is represented by the colocalization index of). The SNAP25 expression was lower in the single rosettes that were

rs: A�20 �m; B–E�5 �m.

f the cere, in partiin A. A h

vely; E isy for VGlient, betw

hite colum

�0.032 SE, n�150), while the r index between SNAP23

Page 6

aA

ribvV

hf(Pwr

FetlViciVbbcvwV mogenouE �m.

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–1096 1089

nd VGluT1 was 0.601 (�0.039 SE, n�40 regions) (one-wayNOVA, P�0.001; post-hoc Holm-Sidak test P�0.05).

In the granular layer, SNAP23 was expressed by allosettes (Fig. 3I–M). In the single rosette, the colocal-zation index between SNAP23 and VGluT2 (Fig. 3K–M,lue arrowheads) was 0.290 (�0.034 SE, n�34). Thisalue is very similar to that found between SNAP23 and

ig. 3. SNAP23 expression in the glutamatergic synaptic terminalsxpression (green) in the molecular layer (ml) of a cerebellar sagittal sehe merged image is showed in A. A high magnification of the red box inabeling, respectively; D is the merged image. (E–H) Double immunostGluT2 (red); the merged image is shown in E. The F–H panels repre

mmunostaining, respectively; H is the merged image. (I–M) Triple imerebral cortex; I represents the merged image. Panels J–M representmmunostaining, respectively; M is a merged image of all markers. TGluT1 (J, M; red arrowheads), only for VGluT2 (K, M; blue arrowhelotting of adult cerebellar homogenate stained with an anti-SNAP23etween SNAP23 and the VGluT markers in the CF (black column)olocalization index of VGAT and VGluT1 in the rosettes of the granaricosities. (P) Colocalization analysis, expressed by Manders’ coeffichich are VGluT2-positive (black column) and/or VGluT1-positive (wGAT and VGluT1 in the rosettes of the granular layer. SNAP23 is horror bars indicate SE. Scale bars: A, E, J�20 �m; B–D and F–H�5

GluT1 (0.322�0.032 SE, n�56) (Fig. 3I–M, red arrow- l

eads). Both coefficients were significantly differentrom the CTR (0.114�0.011 SE) as shown in Fig. 3Pone way ANOVA, P�0.001; post-hoc Holm–Sidak test�0.05). Similar colocalization indexes (Fig. 3O, n�30)ere found in the V1/V2 rosettes (Fig. 3I–M, white ar-

owheads).We also observed a SNAP23 signal in areas not

cerebellar cortex. (A–D) Double immunostaining showing SNAP23articular in the VGluT1-positive presynaptic terminals of the PFs (red);resented in panels B–D. B, C show VGluT1 (red) and SNAP23 (green)strating SNAP23 expression (green) in the CF varicosities positive forgh magnification of the red box in E. F, G show VGluT2 and SNAP23aining showing SNAP23 expression in the granular layer (gl) of thegnification of the red box in I. J–L show VGluT1, VGluT2 and SNAP23heads in L indicate SNAP23 expression in rosettes positive only forexpressing both markers (J, K, M; white arrowheads). (N) Western

. (O) Colocalization analysis, expressed by the Manders’ coefficient,(white column) terminals. The negative CTR is represented by the(gray column). The SNAP23 expression is less abundant in the CF

ween SNAP23 and the VGluT markers in the three groups of rosettesn). The negative CTR is represented by the colocalization index ofsly expressed in all rosettes (gray column). ** P�0.01; *** P�0.001.

of thection, in p

A is repaining illusent a himunost

a high mahe arrowads) andantibodyand PF

ular layerient, bet

hite colum

abeled for the VGluT markers, indicating expression

Page 7

atbAi2

bat

Ss

Ibc

ddtoaGbu

Sblcmus

FeiraimrtSV

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–10961090

long the axolemma of several fibers. The specificity ofhe antibody for anti-SNAP23 was tested by Westernlotting analysis on adult rat cerebellum protein extract.s showed in Fig. 3N the antibody recognizes only one

ntense band with an apparent molecular weight of about3 kDa.

Altogether, these observations indicate that SNAP23 isroadly distributed in the cerebellar cortex without showingpreferential distribution in specific subsets of excitatory

erminals.

NAP25 and SNAP23 expression in the GABAergicynapses of the mature cerebellar cortex

n the molecular layer, inhibitory neurons are representedy stellate and basket cells. The former neurons makeontacts with the proximal and distal regions of the PC

ig. 4. SNAP25 and SNAP23 expression in GABAergic synaptic termxpression in VGAT-positive presynaptic terminals of stellate and bas

mage is showed (A). A high magnification of the red box in A is represespectively; D is the merged image. SNAP25 immunoreactivity is not dnd VGAT (red) in the granular layer. The GABAergic terminals of Go

mmunostaining of SNAP23 (green) and VGAT (red) in the molecularagnification of the red panel is shown in panel J (merged image); pa

epresentative double immunostaining image of SNAP23 (green) and Verminals of the cerebellar cortex. (M) Colocalization analysis, expres

NAP23 and VGAT (white column) marker in both molecular and granular layGAT and VGluT1 in the glomerular structure of the granular layer (gray colum

endritic tree while the latter ones contact the proximal PCendritic domain and form, in addition, an intricate struc-ure surrounding the soma and the initial axonal segmentf the PCs. GABAergic terminals of PC axon collateralsre also present in this layer. In the granular layer, theolgi cells contact the granule cell dendrites in the cere-ellar glomeruli. To identify the GABAergic terminals, wesed an antibody against VGAT.

Consistent with Garbelli et al. (2008), we did not detectNAP25 (green) labeling in GABAergic terminals (red) ofoth the molecular (Fig. 4A–D) and granular (Fig. 4E, F)

ayers by using a different antibody. Fig. 4M shows theolocalization indexes between SNAP25 and VGAT in theolecular layer (0.214�0.005 SE, n�255) and in the gran-lar layer (0.207�0.031 SE, n�65). Both values were notignificantly different from the negative CTR (0.114�0.011 SE)

e cerebellar cortex. (A–D) Double immunostaining showing SNAP25, inhibitory neurons localized in the molecular layer (ml); the mergedpanels B–D. B and C show VGAT (red) and SNAP25 (green) labeling

in these terminals. (E–F) Double immunostaining of SNAP25 (green)re labeled in red and virtually do not express SNAP25. (G–J) Doublee cerebellar cortex; the merged image is shown in panel (G). A high

I illustrate VGAT and SNAP23 immunostaining, respectively. (K–L) Ad) in the granular layer. SNAP23 is weakly expressed in the inhibitoryanders’ coefficient, between SNAP25 and VGAT (black column) or

inals of thket cellsented inetectablelgi cells alayer of thnels H,GAT (resed by M

ers. The negative CTR is represented by the colocalization index ofn). * P�0.05. Scale bars: A, G�20 �m; B–F and H–L�5 �m.

Page 8

(P

dtSimufh

vaa

Udc

Dndal

wmtiedPlIzpDn

Psca2tpgtna

mT((wS

i

ms(rtaS

1zGdt

Stct

Iisostttdp

S

ItirtTpgutqTct

opPSpewuV(

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–1096 1091

one-way ANOVA, P�0.001; post-hoc Holm–Sidak test�0.05).

No evidence is at present available on the SNAP23istribution in the cerebellar GABAergic terminals. In all ofhese structures, we detected a low immunoreactivity forNAP23 (Fig. 4G–L, green). The SNAP23 colocalization

ndex with VGAT was 0.306 (�0.029 SE, n�197) in theolecular layer and 0.224 (�0.017 SE, n�58) in the gran-lar layer (Fig. 4M). Both values were significantly differentrom the negative CTR (one-way ANOVA, P�0.001; post-oc Holm–Sidak test P�0.05).

Collectively, these observations show that SNAP25 isirtually absent in the inhibitory terminals of basket, stellatend Golgi cells, while SNAP23 is expressed at low levels inll of the GABAergic terminals.

ltrastructural analysis of the SNAP25 and SNAP23istribution in the molecular layer of the cerebellarortex

ue to the limitations of the immunofluorescence tech-ique, we used the freeze substitution and post-embed-ing immunogold technique to evaluate the expressionnd the density of the two SNAP isoforms in the molecular

ayer at the ultrastructural level.The two types of excitatory boutons that form synapses

ith PC dendritic spines were identified by their particularorphology: small clusters of vesicles abutting the synap-

ic contact characterize PF terminals. In contrast, CF var-cosities contain a high density of vesicles that are homog-nously distributed inside the terminal and some of theense core vesicles (Palay and Chan-Palay, 1974). In theF boutons that contact the spines of the PC, we observed

abeling of both SNAP25 (Fig. 5A) and SNAP23 (Fig. 5B).n addition to their localization in the presynaptic activeone, we found both proteins also localized inside theresynaptic structure and along the PF axolemma (Fig. 5B,, arrow), while PC dendrites and spines were clearlyegative (Fig. 5A–D, S1).

We calculated the mean density of gold particles in theF presynaptic terminals in both the entire presynaptictructure and in the active zone. The density has beenalculated as the total number of gold particles per terminalrea (�m2); the values were 43.2 (�4.0 SE, n�42) and5.9 (�2.8 SE, n�47) for SNAP25 and SNAP23, respec-ively (Fig. 5E). We also analyzed the densities of the tworoteins in the active zone, expressed as the number ofold particles per �m of presynaptic membrane juxtaposedo the PSD. We found a value of 5.23 particles (�0.73 SE,�43) and 1.40 particles (�0.36 SE, n�48) for SNAP25nd SNAP23, respectively (Fig. 5F).

CF varicosities also expressed both proteins, but with auch lower density than that observed in PF terminals.he density per area of varicosity for SNAP25 was 9.2�1.1 SE, n�19) (Fig. 5C, E) and for SNAP23 was 8.9�0.6 SE, n�15) (Fig. 5E). In the active zone, the valuesere 1.63 (�0.6 SE, n�18) for SNAP25 and 0.90 (�0.4E, n�10) for SNAP23 (Fig. 5F).

We analyzed the distribution of both proteins in the

nhibitory interneurons of the molecular layer. These ter- p

inals form typical symmetrical synapses onto the smoothurface of the dendritic trunks rather than on the spinesPalay and Chan-Palay, 1974). A low gold particle densityelative to the PF terminals was observed for both proteins:he values for SNAP25 were 10.9 (�1.3 SE, n�13) perrea of presynaptic terminal (Fig. 5D, E), and 1.66 (�0.2E, n�10) per �m of active zone length (Fig. 5D, F).

For SNAP23, the value in the presynaptic area was1.6 (�2.2 SE, n�13) (Fig. 5E) and was 0.83 in the activeone (�0.5 SE, n�13) (Fig. 5F). In the granular layer,arbelli et al. (2008) similarly showed a complementaryistribution of SNAP25 in GABAergic and glutamatergicerminals.

Altogether, our results show a differential expression ofNAP25 and SNAP23 in both the excitatory and inhibitory

erminals of the molecular layer as observed by means ofonfocal imaging. In addition, they show a consistent ex-rasynaptic expression of both proteins.

DISCUSSION

n this study, we examined the expression of SNAP25 andts homolog, SNAP23, in the glutamatergic and GABAergicynaptic terminals of the cerebellar cortex. Two relevantbservations emerged: (i) a remarkable differential expres-ion of SNAP25 in the two excitatory glutamatergic inputso PCs that bear very different morpho-functional proper-ies; and (ii) a poor expression of SNAP25 in adult GABAergicerminals. In addition, we observed that SNAP23 is broadlyistributed in the cerebellar cortex without a preferential ex-ression in specific neuronal subtypes.

NAP25/SNAP23 and glutamatergic terminals

n excitatory synapses, the complementary expression ofhe VGluT isoforms is correlated with distinct electrophys-ological properties. In particular, VGluT1 subsets of neu-ons present a low Pr and PPF while VGluT2-positiveerminals show a high Pr and PPD (Fremeau et al., 2004b;akamori, 2006). CFs and PFs represent a typical exam-le of such a correlation. VGluT isoforms are involved inlutamate uptake, but their mechanism of action is stillnknown. Contradictory data have been reported abouthe involvement of VGluT in the presynaptic regulation ofuantal size (Wojcik et al., 2004; Fremeau et al., 2004a;akamori, 2006). It has been suggested that they mayontrol either the differential recycling or refilling rates ofhe vesicles (Fremeau et al., 2004b).

By means of confocal and electron microscopy, webserved that SNAP25 is poorly expressed in VGluT2-ositive CFs while it is highly present in VGluT1-positiveFs. On the contrary, the other two components of theNARE complex, syntaxin 1 and VAMP2, are highly ex-ressed in both varicosities. We also detected a differentialxpression of SNAP25 in the MF presynaptic structures,hich provides the major excitatory input to cerebellar gran-le cells. These terminals bear one or both VGluT1 andGluT2 proteins and the variation in Pr occurs across MFs

Sargent et al., 2005). SNAP25 is expressed less in VGluT2-

ositive terminals relative to VGluT1-positive terminals.

Page 9

F((catz

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–10961092

ig. 5. Immunogold labeling and quantitative analysis for SNAP25 and SNAP23 proteins in the cerebellum. Synaptic contacts between PF-PC spines) show an intense labeling for SNAP25 (A) and for SNAP23 (B). The two arrowheads in A delimit the active zone. CF varicosity contacting a spineC) and a GABAergic terminal (int) forming a synapse onto the PC dendritic shaft membrane (D) show a lower intensity of SNAP25 labeling. In bothases the labeling is also present in the plasma membrane of cross-sectioned PFs, as indicated by the arrows in B and D, and it is negative in spinesnd dendritic shaft. (E) Density distribution of SNAP25 and SNAP23 is expressed as number of immunogold particles per area of presynapticerminals. (F) Density distribution of SNAP25 and SNAP23 is expressed as the number of immunogold particles per length of membrane in the active

one of the presynaptic terminals. Scale bars: A–C�0.18 �m; D�0.28 �m.

Page 10

aniettr

edistg

Svparetts

S

Waundcaptm

dsoatplbeGfb

G(tltvtt

U

Dndalssfipat

alocSaadbA

aoTittggrzt21

aobG

S

Adetlct

avoc(

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–1096 1093

Interestingly, an analogous correlation between SNAP25nd the VGluT isoforms has been reported in corticaleurons (Bragina et al., 2007). SNAP25 is also detectable

n VGluT1-positive terminals in the hippocampus (Verderiot al., 2004; Frassoni et al., 2005). Therefore, the localiza-ion or absence of SNAP25 in specific subtypes of excita-ory neurons in several brain regions suggests a regulatoryole of SNAP25 in shaping the synaptic response.

Regarding SNAP23, we observed its expression in allxcitatory inputs of the cerebellar cortex. Despite its abun-ant expression in PF terminals, we found that this protein

s not preferentially localized in the active zone. It is pos-ible that it is involved in non-synaptic exocytosis in bothhe regulated and constitutive secretory pathways (Bur-ess and Kelly, 1987).

In endocrine cells, the relative expression levels ofNAP25 and SNAP23 might control the mode (regulateds. basal) of granule release by forming docking com-lexes at different calcium ion thresholds (Chieregatti etl., 2004). In regard to the distribution of SNAP23 in otheregions of the brain, it has been reported that it is prefer-ntially present in the cerebral cortex in VGluT2-positiveerminals (Bragina et al., 2007). Thus, in several regions ofhe brain, a strict correlation between SNAP23 and specificubsets of synapse is not evident.

NAP25/SNAP23 and GABAergic terminals

hether SNAP25 is present in adult GABAergic terminals ismatter of debate. In fact, SNAP25 immunoreactivity is

ndetectable in inhibitory synapses of mature hippocampaleurons (Verderio et al., 2004; Frassoni et al., 2005) of ro-ents, in human brain (Garbelli et al., 2008) and in the raterebral cortex (Bragina et al., 2007). In contrast, Tafoya etl. (2006) reported the presence of high SNAP25 levels inerisomatic basket cell terminals on CA1 pyramidal cells of

he adult mouse hippocampus and in the thalamus of adultice.

In the cerebellar cortex, Garbelli et al. (2008) did notetect SNAP25 immunoreactivity in GABAergic terminals oftellate, basket and Golgi cells. Our results support theirbservations by means of a complementary colocalizationnalysis and electron microscopy studies. At the ultrastruc-ural level, we found a very low density of SNAP25 goldarticles in the inhibitory terminals located in the molecular

ayer. In the granular layer, Garbelli et al. (2008) have com-ined pre-embedding immunolabeling for SNAP25 with post-mbedding immunogold localization for the neurotransmitterABA and described GABAergic terminals that are negative

or SNAP25. These observations indicate that mature cere-ellar GABAergic neurons weakly express SNAP25.

There is evidence that SNAP23 is expressed inABAergic terminals of hippocampal and cortical neurons

Verderio et al., 2004; Bragina et al., 2007). On the con-rary, we observed by means of a quantitative analysis aow but detectable expression of SNAP23 in the inhibitoryerminals of the cerebellar cortex. Altogether, these obser-ations indicate that both SNAP25 and SNAP23 are nothe predominant SNAP isoforms in the mature inhibitory

erminals of the cerebellar cortex. m

ltrastructural distribution

ue to the limitations of the immunofluorescence tech-ique, we used the freeze substitution and post-embed-ing immunogold technique to evaluate the expressionnd density of the two SNAP homologs in the molecular

ayer at the ultrastructural level. The procedure of freezeubstitution and low-temperature embedding of neuralamples in Lowicryl HM20 resin is not able to preserve thene structural characteristics of tissues, but overcomes theroblems of the penetration of immunoreagents in tissuesnd offers the high-resolution localization of neuronal an-igens (Baude et al., 1993; Nusser et al., 1994).

Besides their localization in the active zone of presyn-ptic terminals, both SNAP25 and SNAP23 were also

ocalized inside the presynaptic structure. Accordingly, inther brain regions, SNAP25 was found on synaptic vesi-les together with syntaxin 1 (Schulze et al., 1995; Walch-olimena et al., 1995; Kretzschmar et al., 1996; Morel etl., 1998; Zhang et al., 2000; Tao-Cheng et al., 2000). Inddition, SNAP25 was identified in PC12 cells on largeense core vesicles (Marxen et al., 1997) and it seems toe involved in endosome fusion (Tao-Cheng et al., 2000;ikawa et al., 2006).

We also detected both SNAP homologs in the PFxolemma in accordance with observations reported inther brain regions (Garcia et al., 1995; Zhang et al., 2000;ao-Cheng et al., 2000). The localization of both proteins,

n particular of SNAP25, along the axolemma suggestshat they are involved in non-synaptic exocytosis, in bothhe regulated and constitutive secretory pathways (Bur-ess and Kelly, 1987; Lysakowski et al., 1999). Secretoryranules containing neuropeptides are most likely to beeleased at non-synaptic locations away from the activeones (Lysakowski et al., 1999). CFs release the neuropep-ide corticotrophin-releasing factor (CRF) (Schmolesky et al.,007) and some release neuropeptide-Y (Ueyama et al.,994; Morara et al., 1997).

Alternatively, they may be involved in the constitutiveddition of membrane proteins into the axolemma. More-ver, SNAP25 may be a target receptor of other mem-rane-bound proteins that are polarized to axons, such asAP 43 (Skene, 1989).

NAP25 as a modulator of synaptic transmission

ltogether, these observations suggest that synapses withifferent release properties are characterized by differentxpression levels of SNAP25. Based on data reported inhe literature, we can speculate how SNAP25 may modu-ate the Pr and short-term plasticity. In particular, we fo-used our attention on the CF and PF inputs whose elec-rophysiological properties have been well characterized.

Notably, the large difference in the Pr of these syn-pses cannot be accounted for by the number of dockedesicles. In fact, despite their distinct overall ultrastructuralrganization, they have similar numbers of docked vesi-les, a similar size of PSD and postsynaptic spine volumeXu-Friedman et al., 2001). It has been proposed that

olecular differences in release machinery or calcium sig-

Page 11

nc2

c2bck2rtmsts(tcSccost

Stht(rhn

iSih(Sipma

SSacciiscmaa

da

ttcpdF

tdfHlgec

AwuehtaMSc(

SaehSbipi

AdttAM5F

A

A

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–10961094

aling, superimposed onto the synaptic morphology, mayontribute to the regulation of Pr (Xu-Friedman et al.,001).

Recent observations point to a role for SNAREs inontrolling the Pr (Finley et al., 2002; Hu and Davletov,003). In particular, mutations that alter the interactionetween the SNAP25 C-terminal coil and the other SNAREoils dramatically reduce the transmitter Pr, but leave theinetics of synaptic responses unaltered (Finley et al.,002). Moreover, a stochastic model of neurotransmitterelease has been proposed in which the availability of-SNARE heterodimers (SNAP25-syntaxin) in the targetembrane may be a novel regulator of the probability of

ynaptic vesicle fusion (Hu and Davletov, 2003). In turn,he existence of multiple isoforms or phosphorylationites of such proteins could differentially affect the PrSüdhof, 1995). Interestingly, there is increasing evidencehat SNAP25 also modulates various voltage–gated ionhannels (Catterall, 1999; Zamponi, 2003). In particular,NAP25 interaction causes a reduction of N-type calciumhannel currents (Wiser et al., 1996), inhibition of L-typealcium channel currents (Ji et al., 2002), and a reductionf the activity of P/Q-type calcium channels by negativelyhifting the steady state voltage dependence of inactiva-ion (Zhong et al., 1999).

The negative modulation of cytosolic calcium byNAP25 has been associated with the alteration of short-

erm plasticity events. Hippocampal slices from SNAP25eterozygous mice display significantly larger PPF relativeo wild-type mice, specifically at short inter-pulse intervalsPozzi et al., 2008). Furthermore, higher calcium peaks areeached after depolarization by SNAP25 null and SNAP25eterozygous cultured neurons compared with wild-typeeurons (Pozzi et al., 2008).

Similar effects on PPF were previously shown to occurn SNAP25 mouse mutants that overexpressed theNAP25a isoform, where the shift between the “a” and “b”

soforms is impaired (Bark et al., 2004), and in culturedippocampal neurons treated with BoNT/A or BoNT/EYoung, 2005). The inhibition of VGCCs is mediated byNAP25 phosphorylation at Ser187. Since neuronal activ-

ty transiently induces Ser187 phosphorylation, it has beenroposed that SNAP25 may provide a negative feedbackechanism for controlling neuronal excitability (Pozzi etl., 2008).

The negative modulation of cytosolic calcium byNAP25 fits very well with the differential expression ofNAP25 between CFs and PFs. We suggest that a highermount of SNAP25 in the PF terminals causes a loweralcium influx, which correlates with a low Pr. On theontrary, a low amount of SNAP25 in the CF terminal ismportant to maximize the presynaptic calcium entry induc-ng a multivesicular neurotransmitter release, typical of thisort of synapse (Foster et al., 2002). Both mechanisms,ombined to the saturation of postsynaptic receptors,ake the CF synapse both powerful and reliable (Foster etl., 2002). The high Pr combined with the synchronous

ctivation of a high number of presynpatic terminals pro-

uces a large synaptic response that makes such a syn-pse unique in the CNS.

The amount of SNAP25 may influence short-term plas-icity. It has been suggested that PPD and PPF depend onhe Pr. When the Pr is high the vesicle pools are signifi-antly depleted following an action potential so that de-ression dominates in CF, whereas when Pr is low, vesicleepletion is minimal and facilitation dominates in PF (Xu-riedman and Regehr, 2004).

In the MFs, the Pr varies between them. Such a struc-ure, with its high density of release sites, would allowesensitization to play a role in short-term plasticity evenor a low Pr synapse (Xu-Friedman and Regehr, 2003).owever, Pr is relatively uniform across the 200–400 re-

ease sites on an individual MF presynaptic structure (Sar-ent et al., 2005). Accordingly, we found that a variablexpression of SNAP25 in these terminals correlated withomplementary expression of the VGluT isoforms.

CONCLUSION

ltogether, these observations leave open the question ofhether a low amount of SNAP25 is responsible for reg-lated exocytosis at the CF-PC synapses and at the cer-bellar GABAergic terminals. Alternatively, other SNAPomologs could replace SNAP25 in the role that it ishought to play. SNAP23 is not a good candidate for such

function due to its low expression in these terminals.oreover, it has been shown that lentiviral expression ofNAP23 in SNAP25 null neurons in culture does not res-ue the fast release component of the evoked releaseDelgado-Martinez et al., 2007).

New members of the SNARE family, SNAP29 andNAP47, have been characterized (Su et al., 2001; Holt etl., 2006). They have a ubiquitous distribution, but theirxpression in different subsets of neurons is not known. Itas been suggested that they may act as “general utility”NAREs that are capable of providing a SNAP25 homologinding domain to a variety of SNARE complexes operat-

ng inside the cells (Su et al., 2001; Holt et al., 2006). Theossible involvement of other homologs, which are not yet

dentified, should also be taken into consideration.

cknowledgments—We wish to thank Prof. M. Matteoli for heriscussion and comments regarding the manuscript. We alsohank, V. Batocchi and A. Renna for her helpful technical assis-ance. This work was supported by grants from the Italian Spacegency, the Italian Ministry of Universities and Research, theinistry of Health, the European Community (contract number12039), Regione Piemonte, and the Compagnia San Paolooundation.

REFERENCES

ikawa Y, Lynch KL, Boswell KL, Martin TF (2006) A second SNARErole for exocytic SNAP25 in endosome fusion. Mol Biol Cell17:2113–2124.

raki S, Tamori Y, Kawanishi M, Shinoda H, Masugi J, Mori H, Niki T,Okazawa H, Kubota T, Kasuga M (1997) Inhibition of the binding ofSNAP-23 to syntaxin 4 by Munc18c. Biochem Biophys Res Com-

mun 234:257–262.

Page 12

A

B

B

B

B

B

B

B

C

C

C

C

C

C

D

E

F

F

F

F

F

F

G

G

G

G

H

H

H

H

I

J

J

K

K

L

M

M

M

M

M

M

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–1096 1095

twood HL, Karunanithi S (2002) Diversification of synaptic strength:presynaptic elements. Nat Rev Neurosci 3:497–516.

ark C, Bellinger FP, Kaushal A, Mathews JR, Partridge LD, WilsonMC (2004) Developmentally regulated switch in alternativelyspliced SNAP-25 isoforms alters facilitation of synaptic transmis-sion. J Neurosci 24:8796–8805.

ark IC, Hahn KM, Ryabinin AE, Wilson MC (1995) Differential ex-pression of SNAP-25 protein isoforms during divergent vesiclefusion events of neural development. Proc Natl Acad Sci U S A92:1510–1514.

aude A, Nusser Z, Roberts JD, Mulvihill E, McIlhinney RA, SomogyiP (1993) The metabotropic glutamate receptor (mGluR1 alpha) isconcentrated at perisynaptic membrane of neuronal subpopula-tions as detected by immunogold reaction. Neuron 11:771–787.

oschert U, O’Shaughnessy C, Dickinson R, Tessari M, Bendotti C,Catsicas S, Pich EM (1996) Developmental and plasticity-relateddifferential expression of two SNAP-25 isoforms in the rat brain.J Comp Neurol 367:177–193.

ragina L, Candiracci C, Barbaresi P, Giovedì S, Benfenati F, Conti F(2007) Heterogeneity of glutamatergic and GABAergic releasemachineries in cerebral cortex. Neuroscience 146:1829–1840.

ronk P, Deak F, Wilson MC, Liu X, Sudhof TC, Kavalali ET (2007)Differential effects of SNAP-25 deletion on Ca2�-dependent andCa2�-independent neurotransmission. J Neurophysiol 98:794–801.

urgess TL, Kelly RB (1987) Constitutive and regulated secretion ofproteins. Annu Rev Cell Biol 3:243–294.

arulli D, Buffo A, Strata P (2004) Reparative mechanisms in thecerebellar cortex. Prog Neurobiol 72:373–398.

atsicas S, Catsicas M, Keyers KT, Karten HJ, Wilson MC, Milner RJ(1992) Differential expression of the presynaptic protein SNAP-25in mammalian retina. J Neurosci Res 33:1–9.

atterall WA (1999) Interactions of presynaptic Ca2� channels andsnare proteins in neurotransmitter release. Ann N Y Acad Sci868:144–159.

haudhry FA, Reimer RJ, Bellocchio EE, Danbolt NC, Osen KK,Edwards RH, Storm-Mathisen J (1998) The vesicular GABA trans-porter, VGAT, localizes to synaptic vesicles in sets of glycinergic aswell as GABAergic neurons. J Neurosci 18:9733–9750.

hen YA, Scheller RH (2001) SNARE-mediated membrane fusion.Nat Rev Mol Cell Biol 2:98–106.

hieregatti E, Chicka MC, Chapman ER, Baldini G (2004) SNAP-23functions in docking/fusion of granules at low Ca2�. Mol Biol Cell15:1918–1930.

elgado-Martinez I, Nehring RB, Sorensen JB (2007) Differential abil-ities of SNAP-25 homologs to support neuronal function. J Neuro-sci 27:9380–9391.

ccles JC, Ito M, Szentágothai J (1967) The cerebellum as a neuronalmachine. Berlin: Springer Verlag.

inley MF, Patel SM, Madison DV, Scheller RH (2002) The coremembrane fusion complex governs the probability of synaptic ves-icle fusion but not transmitter release kinetics. J Neurosci 22:1266–1272.

oster KA, Kreitzer AC, Regehr WG (2002) Interaction of postsynapticreceptor saturation with presynaptic mechanisms produces a reli-able synapse. Neuron 36:1115–1126.

rassoni C, Inverardi F, Coco S, Ortino B, Grumelli C, Pozzi D,Verderio C, Matteoli M (2005) Analysis of synaptosomal-associ-ated protein of 25 kDa (SNAP25) immunoreactivity in hippocampalinhibitory neurons during development in culture and in situ. Neu-roscience 131:813–823.

remeau RTJ, Kam K, Qureshi T, Johnson J, Copenhagen DR, Storm-Mathisen J, Chaudhry FA, Nicoll RA, Edwards RH (2004a) Vesic-ular glutamate transporters 1 and 2 target to functionally distinctsynaptic release sites. Science 304:1815–1819.

remeau RTJ, Troyer MD, Pahner I, Nygaard GO, Tran CH et al(2001) The expression of vesicular glutamate transporters defines

two classes of excitatory synapse. Neuron 31:247–260.

remeau RTJ, Voglmaier S, Seal RP, Edwards RH (2004b) VGLUTsdefine subset of excitatory neurons and suggest novel roles forglutamate. Trends Neurosci 27:98–103.

alli T, Zahraoui A, Vaidyanathan VV, Raposo G, Tian JM, Karin M,Niemann H, Louvard D (1998) A novel tetanus neurotoxin-insen-sitive vesicle-associated membrane protein in SNARE complexesof the apical plasma membrane of epithelial. Mol Biol Cell 9:1437–1448.

arbelli R, Inverardi F, Medici V, Amadeo A, Verderio C, Matteoli M,Frassoni C (2008) Heterogeneous expression of SNAP-25 in ratand human brain. J Comp Neurol 506:373–386.

arcia EP, McPherson PS, Chilcote TJ, Takei K, De Camilli P (1995)rbSec1A and B colocalize with syntaxin 1 and SNAP-25 throughoutthe axon, but are not in a stable complex with syntaxin. J Cell Biol129:105–120.

ibbins IL, Jobling P, Teo EH, Matthew SE, Morris JL (2003) Heter-ogeneous expression of SNAP-25 and synaptic vesicle proteins bycentral and peripheral inputs to sympathetic neurons. J CompNeurol 459:25–43.

ellstrom J, Arvidsson U, Elde R, Cullheim S, Meister B (1999) Dif-ferential expression of nerve terminal protein isoforms in VAChT-containing varicosities of the spinal cord ventral horn. J CompNeurol 411:578–590.

ioki H, Fujiyama F, Taki K, Tomioka R, Furuta T, Tamamaki N,Kaneko T (2003) Differential distribution of vesicular glutamatetransporters in the rat cerebellar cortex. Neuroscience 117:1–6.

olt M, Varoqueaux F, Wiederhold K, Takamori S, Urlaub H,Fasshauer D, Jahn R (2006) Identification of SNAP-47, a novelQbc-SNARE with ubiquitous expression. J Biol Chem 281:17076–17083.

u K, Davletov B (2003) SNAREs and control of synaptic releaseprobabilities. FASEB J 17:130–135.

to M (1984) The cerebellum and neural control. New York: RavenPress.

ahn R, Scheller RH (2006) SNAREs-engines for membrane fusion.Nat Rev Mol Cell Biol 7:631–643.

i J, Yank SN, Huang X, Li X, Sheu L, Diamant N, Berggren PO,Gaisano HY (2002) Modulation of L-type Ca(2�) channels bydistinct domains within SNAP-25. Diabetes 51:1425–1436.

onnerth A, Llano I, Armstrong CM (1990) Synaptic currents in cere-bellar Purkinje cells. Proc Natl Acad Sci U S A 87:2662–2665.

retzschmar S, Volknandt W, Zimmermann H (1996) Colocalizationon the same synaptic vesicles of syntaxin and SNAP-25 withsynaptic vesicle proteins: a re-evaluation of functional models re-quired? Neurosci Res 26:141–148.

ysakowski A, Figueras H, Price SD, Peng YY (1999) Dense-coredvesicles, smooth endoplasmic reticulum, and mitochondria areclosely associated with non-specialized parts of plasma membraneof nerve terminals: implications for exocytosis and calcium buffer-ing by intraterminal organelles. J Comp Neurol 403:378–390.

anders EMM, Verbeek FJ, Aten JA (1993) Measurement of colocal-ization of objects in dual-colour confocal images. J Microsc169:375–382.

arxen M, Maienschein V, Volknandt W, Zimmermann H (1997) Im-munocytochemical localization of synaptic proteins at vesicularorganelles in PC12 cells. Neurochem Res 22:941–950.

atteoli M, Pozzi D, Grumelli C, Condliffe SB, Frassoni C, Harkany T,Verderio C (2009) The synaptic split of SNAP-25: different roles inglutamatergic and GABAergic neurons? Neuroscience 158:223–230.

orara S, Marcotti W, Provini L, Rosina A (1997) Neuropeptide Y(NPY) expression is up-regulated in the rat inferior olive duringdevelopment. Neuroreport 8:3743–3747.

orel N, Taubenblatt P, Synguelakis M, Shiff G (1998) A syntaxin-SNAP 25-VAMP complex is formed without docking of synapticvesicles. J Physiol Paris 92:389–392.

orris JL, Lindberg CE, Gibbins IL (2000) Different levels of immuno-

reactivity for synaptosomal-associated protein of 25kDa in vaso-

Page 13

N

O

O

P

P

P

R

S

S

S

S

S

S

S

T

T

T

T

U

V

W

W

W

W

X

X

X

Y

Z

Z

Z

S

S

G. Mandolesi et al. / Neuroscience 164 (2009) 1084–10961096

constrictor and vasodilator axons of guinea-pigs. Neurosci Lett294:167–170.

usser Z, Mulvihill E, Streit P, Somogyi P (1994) Subsynaptic segre-gation of metabotropic and ionotropic glutamate receptors as re-vealed by immunogold localization. Neuroscience 61:421–427.

yler GA, Higgins GA, Hart RA, Battenberg E, Billingsley M, BloomFE, Wilson MC (1989) The identification of a novel synaptosomal-associated protein, SNAP-25, differentially expressed by neuronalsubpopulations. J Cell Biol 109:3039–3052.

yler GA, Polli JW, Wilson MC, Billingsley ML (1991) Developmentalexpression of the 25-kDa synaptosomal-associated protein(SNAP-25) in rat brain. Proc Natl Acad Sci U S A 88:5247–5251.

alay SL, Chan-Palay V (1974) Cerebellar cortex: cytology and or-ganisation. Berlin: Springer Press.

erkel DJ, Hestrin S, Sah P, Nicoll RA (1990) Excitatory synapticcurrents in Purkinje cells. Proc Biol Sci 241:116–121.

ozzi D, Condliffe S, Bozzi Y, Chikhladze M, Grumelli C, Proux-Gillardeaux V, Takahashi M, Franceschetti S, Verderio C, MatteoliM (2008) Activity-dependent phosphorylation of Ser187 is requiredfor SNAP-25-negative modulation of neuronal voltage-gated cal-cium channel. Proc Natl Acad Sci U S A 105:323–328.

avichandran V, Chawlat A, Roche PA (1996) Identification of a novelsyntaxin and synaptobrevin/VAMP-binding protein, SNAP-23, ex-pressed in nonneuronal tissue. J Biol Chem 271:13300–13303.

argent PB, Saviane C, Nielsen TA, DiGregorio DA, Silver RA (2005)Rapid vesicular release, quantal variability, and spillover contributeto the precision and reliability of transmission at a glomerularsynapse. J Neurosci 25:8173–8187.

chmolesky MT, De Ruiter MM, De Zeeuw CI, Hansel C (2007) Theneuropeptide corticotropin-releasing factor regulates excitatorytransmission and plasticity at the climbing fibre-Purkinje cell syn-apse. Eur J Neurosci 25:1460–1466.

chulze KL, Broadie K, Perin MS, Bellen HJ (1995) Genetic andelectrophysiological studies of Drosophila syntaxin-1A demon-strate its role in non-neuronal secretion and neurotransmission.Cell 80:311–320.

kene JH (1989) Axonal growth-associated proteins. Annu Rev Neu-rosci 12:127–156.

öllner T, Whiteheart SW, Brunner M, Erdjument-Bromage H, Gero-manos S, Tempst P, Rothman JE (1993) SNAP receptors impli-cated in vesicle targeting and fusion. Nature 362:318–324.

u Q, Mochida S, Tian JH, Mehta R, Sheng ZH (2001) SNAP-29: ageneral SNARE protein that inhibits SNARE disassembly and isimplicated in synaptic transmission. Proc Natl Acad Sci U S A98:14038–14043.

üdhof TC (1995) The synaptic vesicle cycle: a cascade of protein–protein interactions. Nature 375:645–653.

afoya LC, Mameli M, Miyashita T, Guzowsky JF, Valenzuela CF,Wilson MC (2006) Expression and function of SNAP-25 as auniversal SNARE component in GABAergic neurons. J Neurosci26:7826–7838.

akamori S (2006) VGLUTs: “exciting” times for glutamatergic re-

search? Neurosci Res 55:343–351. t

ao-Cheng JH, Du J, McBain CJ (2000) Snap-25 is polarized to axonsand abundant along the axolemma: an immunogold study of intactneurons. J Neurocytol 29:67–77.

homson AM (2000) Molecular frequency filters at central synapses.Prog Neurobiol 62:159–196.

eyama T, Houtani T, Nakagawa H, Baba K, Ikeda M, Yamashita T,Sugimoto T (1994) A subpopulation of olivocerebellar projectionneurons express neuropeptide Y. Brain Res 634:353–357.

erderio C, Pozzi D, Pravettoni E, Inverardi F, Schenk U, Coco S,Proux-Gillardeaux V, Galli T, Rossetto O, Frassoni C, Matteoli M(2004) SNAP-25 modulation of calcium dynamics underlies differ-ence in GABAergic and glutamatergic responsiveness to depolar-ization. Neuron 41:599–610.

alch-Solimena C, Blasi J, Edelmann L, Chapman ER, Von MollardGF, Jahn R (1995) The t-SNAREs syntaxin 1 and SNAP-25 arepresent on organelles that participate in synaptic vesicle recycling.J Cell Biol 128:637–645.

iser O, Bennett MK, Atlas D (1996) Functional interaction of syntaxinand SNAP-25 with voltage-sensitive L- and N-type Ca2� channels.EMBO J 15:4100–4110.

ojcik SM, Rhee JS, Herzog E, Sigler A, Jahn R, Takamori S, BroseN, Rosenmund C (2004) An essential role for vesicular glutamatetransporter 1 (VGLUT1) in postnatal development and control ofquantal size. Proc Natl Acad Sci U S A 101:7158–7163.

ong PP, Daneman N, Volchuk A, Lassam N, Wilson MC, Klip A,Trimble WS (1997) Tissue distribution of SNAP-23 and its subcel-lular localization in 3T3-L1 cells. Biochem Biophys Res Commun230:64–68.

u-Friedman MA, Harris KM, Regher WG (2001) Three-dimensionalcomparison of ultrastructural characteristics at depressing andfacilitating synapses onto cerebellar Purkinje cells. J Neurosci21:6666–6672.

u-Friedman MA, Regehr WG (2003) Ultrastructural contributions todesensitization at cerebellar mossy fiber to granule cell synapses.J Neurosci 23:2182–2192.

u-Friedman MA, Regehr WG (2004) Structural contributions to short-term synaptic plasticity. Physiol Rev 84:69–85.

oung SM Jr. (2005) Proteolysis of SNARE proteins alters facilitationand depression in a specific way. Proc Natl Acad Sci U S A102:2614–2619.

amponi GW (2003) Regulation of presynaptic calcium channels bysynaptic proteins. J Pharmacol Sci 92:79–83.

hang L, Volknandt W, Gundelfinger ED, Zimmermann H (2000) Acomparison of synaptic protein localization in hippocampal mossyfiber terminals and neurosecretory endings of the neurohypophysisusing the cryo-immunogold technique. J Neurocytol 29:19–30.

hong H, Yokoyama CT, Scheuer T, Catterall WA (1999) Reciprocalregulation of P/Q-type Ca2� channels by SNAP-25, syntaxin andsynaptotagmin. Nat Neurosci 2:939–941.

APPENDIX

upplementary data

upplementary data associated with this article can be found in

he online version at: doi:10.1016/j.neuroscience.2009.08.067.

(Accepted 7 August 2009)(Available online 6 September 2009)