The synaptic vesicle protein CSP alpha prevents presynaptic degeneration

Neuron, Vol. 42, 237–251, April 22, 2004, Copyright 2004 by Cell Press The Synaptic Vesicle Protein CSP␣ Prevents Presynaptic Degeneration hanced CSP␣function could attenuate neurodegenerative diseases. Rafael Ferna ´ndez-Chaco ´n, 1,2 Markus Wo ¨lfel, 5 Hiroshi Nishimune, 7 Lucia Tabares, 2 Frank Schmitz, 3,4 Manuel Castellano-Mun ˜oz, 2 Christian Rosenmund, 6 Maria L. Montesinos, 2 Introduction Joshua R. Sanes, 7 Ralf Schneggenburger, 5 and Thomas C. Su ¨dhof 1, *When an action potential invades a presynaptic nerve 1 Center for Basic Neuroscience terminal, voltage-gated Ca 2 ⫹channels open, and Ca 2 ⫹ Department of Molecular Genetics and entering the terminal triggers neurotransmitter release Howard Hughes Medical Institute by synaptic vesicle exocytosis (Katz, 1969). Empty syn- University of Texas Southwestern Medical Center aptic vesicles then undergo endocytosis to recycle for Dallas, Texas 75390 a new round of exocytosis. Synaptic vesicle exo- and 2 Departamento Fisiologı ´aMe ´dica y Biofı ´sica endocytosis are mediated by a sophisticated apparatus Universidad de Sevilla that turns nerve terminals into membrane trafficking Avenida Sa ´nchez-Pizjua ´n4 machines. Presynaptic terminals widely differ in ultra- 41009-Sevilla structural appearance, neurotransmitter type, and use- Spain dependent plasticity, but all terminals share the same 3 Universita ¨t Leipzig functional architecture that depends on the autono- Institut fu ¨r Anatomie mous, Ca 2 ⫹-regulated exo- and endocytosis of synaptic Liebigstrasse 13 vesicles. Since synaptic vesicles are the central organ- 04103 Leipzig elles in neurotransmitter release, the characterization of Germany synaptic vesicle proteins is an important goal in under- 4 Institut fu ¨r Anatomie und Zellbiologie standing synaptic function. This goal has been largely Universita ¨t des Saarlandes Homburg/Saar achieved for cataloging synaptic vesicle proteins but is Germany far from complete for the functional understanding of 5 AG Synaptic Dynamics and Modulation these proteins (reviewed in Ferna ´ndez-Chaco ´n and Su ¨d- 6 AG Synaptic Physiology hof, 1999). Among the functions performed by synaptic Department of Membrane Biophysics vesicle proteins, exo- and endocytosis are relatively MPI fu ¨r biophysikalische Chemie easy to measure, while other stages of the vesicle cycle— 37070 Go ¨ttingen e.g., the biogenesis and recycling or synaptic vesicles or Germany the maintenance of vesicles as separate organelles—are 7 Department of Anatomy and Neurobiology difficult to evaluate. As a result, vesicle proteins that Washington University School of Medicine participate in exo- and endocytosis are better under- 660 South Euclid Avenue stood than proteins involved in other functions. In fact, St Louis, Missouri 63110 some of the proteins that are currently thought to participate in exo- or endocytosis may act at a different stage of the vesicle cycle and influence exo- or endocytosis only indirectly. Summary Cysteine string protein (CSP) is an abundant, evolutionarily conserved synaptic vesicle protein that con- Cysteine string protein ␣(CSP␣)—an abundant synaptains a string of C-terminal cysteine residues (Zinsmaier tic vesicle protein that contains a DNA-J domain char- et al., 1990; Gundersen and Umbach, 1992). In Drosoacteristic of Hsp40 chaperones—is thought to regulate phila mutants that lack CSP, most flies die during devel- Ca 2ⴙ channels and/or synaptic vesicle exocytosis. We opment, but a small percentage survives into adulthood now show that, in young mice, deletion of CSP␣does with a temperature-sensitive paralytic phenotype (Umnot impair survival and causes no significant changes bach et al., 1994; Zinsmaier et al., 1994). Studies of the in presynaptic Ca 2ⴙ currents or synaptic vesicle exoneuromuscular junction of larvae from surviving mutant cytosis as measured in the Calyx of Held synapse. flies revealed that deletion of CSP severely depressed At 2–4 weeks of age, however, CSP␣-deficient mice evoked neurotransmitter release at the permissive temdevelop a progressive, fatal sensorimotor disorder. perature and nearly abolished release at the nonpermis- The neuromuscular junctions and Calyx synapses of sive temperature (Umbach et al., 1994; Zinsmaier et al., CSP␣-deficient mice exhibit increasing neurodegen- 1994). Based on these results and on studies of verteerative changes, synaptic transmission becomes sebrate CSP, three different hypotheses for the function verely impaired, and the mutant mice die at ⵑ2 months of CSP were proposed. of age. Our data suggest that CSP␣is not essential for One hypothesis is that CSP is involved in Ca 2 ⫹channel the normal operation of Ca 2ⴙ channels or exocytosis but function. Gundersen and Umbach (1992) observed that acts as a presynaptic chaperone that maintains contin- CSP antisense RNA inhibited the expression of N-type ued synaptic function, raising the possibility that en- Ca 2 ⫹channels when coinjected with Torpedo electric lobe mRNA into Xenopus oocytes. The connection of CSP to Ca 2 ⫹channels was strengthened by the finding *Correspondence: [email protected] Neuron 238 that voltage-gated Ca 2 ⫹influx was impaired in CSP muis essential for the long-term survival of synapses and tant Drosophila synapses and that neurotransmitter refunctions to prevent the degeneration of nerve terminals. lease could be rescued in these flies by application of These data are consistent with a role for CSP␣as a aCa 2 ⫹ionophore (Ranjan et al., 1998; Umbach et al., molecular chaperone that makes it possible for syn- 1998). Furthermore, introduction of recombinant CSP apses to keep running for extended time periods. into chick nerve terminals increased Ca 2 ⫹channel activity (Chen et al., 2002), and CSP␣may bind to vertebrate Results Ca 2 ⫹channel proteins (Leveque et al., 1998) or to G protein subunits regulating Ca 2 ⫹channels (Magga et KO of CSP␣Causes Postnatal Lethality al., 2000). The idea that CSP mediates the assembly or Sequence analyses revealed that the mouse genome regulation of Ca 2 ⫹channels at the active zone is attraccontains three CSP genes encoding CSP␣,␤, and ␥ tive in view of the localization of CSP on synaptic vesi- (Supplemental Figure S1 at http://www.neuron.org/cgi/ cles. However, this hypothesis was challenged by fluocontent/full/42/2/237/DC1). Traditionally, only CSP␣ rometric Ca 2 ⫹measurements that demonstrated that (which is widely expressed throughout development in nerve terminals from Drosophila CSP mutants display most tissues Chamberlain and Burgoyne, 1996) was an increase, and not a decrease, in Ca 2 ⫹influx (Dawsonstudied. To determine the expression patterns of CSP␤ Scully et al., 2000) and by measurements of Ca 2 ⫹cur- or ␥that could potentially be functionally redundant with rents in peptidergic terminals from CSP mutant flies CSP␣, we used RNA blotting and quantitative RT-PCR. (Morales et al., 1999). We found that CSP␤and ␥mRNAs are abundant in A second hypothesis for CSP function is that CSP testis but are not detectable in other adult tissues (Supdirectly regulates exocytosis. Consistent with such a plemental Figure S2) or in brain at all developmental role, CSP interacts with syntaxin and synaptotagmin stages tested (P5, P15, and P40; Supplemental Table in pulldown experiments (Nie et al., 1999; Evans and S1). Thus, in mice, brain contains only CSP␣at signifi- Morgan, 2002; however, see Leveque et al., 1998, and cant levels, and CSP␤and CSP␥likely do not have a Matos et al., 2000, for alternative results). Moreover, neuronal function. transmitter release is depressed in Drosophila CSP mu- To generate CSP␣KO mice, we isolated genomic tants despite higher than normal Ca 2 ⫹levels, indicating clones that contained the first coding exon of the mouse a role for CSP in determining the Ca 2 ⫹sensitivity of CSP␣gene and constructed a targeting vector for hothe secretory apparatus (Dawson-Scully et al., 2000). mologous recombination (Figure 1A). We electroporated Furthermore, overexpression of CSP in neuroendocrine the targeting vector into embryonic stem cells, selected cells altered exocytosis but had no effect on Ca 2 ⫹chanhomologous recombinants, and used them to generate nel activity (Brown et al., 1998; Graham and Burgoyne, CSP␣KO mice. Genotyping of newborn offspring from 2000; Chamberlain and Burgoyne, 1998; Zhang et al., heterozygous matings identified apparently normal het- 1998). erozygous and homozygous mutant mice that were born A third hypothesis for CSP function is based on the in the expected Mendelian ratios (data not shown). Thus, observation that CSP contains a DNA-J domain. DNA- the homozygous CSP␣mutation is not pre- or perinatally J domains are characteristic of cochaperones of the lethal. As shown by immunoblotting of brain proteins, Hsp40 family, and they functionally collaborate with the heterozygous mutant mice contained less CSP␣than DNA-K domains of Hsc70 proteins (reviewed in Kelley, wild-type controls, while homozygous mutant mice 1998; Hartl and Hayer-Hartl, 2002). Indeed, CSP actilacked CSP␣(Figure 1B). Control proteins such as vates the ATPase activity of Hsc70 (Braun et al., 1996; SCAMP1, complexins, and synaptobrevin 2 exhibited Chamberlain and Burgoyne, 1997) and forms a trimeric no major change. Quantitative RT-PCR demonstrated complex with Hsc70 and a tetratricopeptide repeat prothat the expression of CSP␤or ␥did not increase in tein called SGT (Tobaben et al., 2001). This complex brains from CSP␣KO mice at P5 or P15, ruling out a catalyzes the ATP-dependent refolding of denatured lucompensatory upregulation of other CSP isoforms in ciferase (Tobaben et al., 2001). A role for CSP as a the KOs (Supplemental Data and Supplemental Table chaperone to renature nerve terminal proteins that mis- S1 at http://www.neuron.org/cgi/content/full/42/2/237/ fold during the continuous operation of the synaptic DC1). In all of the experiments described here, analyses vesicle cycle is also attractive but is not currently supwere carried out with littermates derived from heterozyported by in vivo evidence. gous matings to exclude confounding background ef- We have attempted to test these hypotheses about fects. the function of CSP by analyzing knockout (KO) mice In the first 2 postnatal weeks, the weight and survival lacking CSP. We show that mice contain three CSP of CSP␣KO mice were not significantly different from genes (CSP␣,␤, and ␥), but only CSP␣is expressed in heterozygous and wild-type littermate controls (Figure brain, whereas CSP␤and CSP␥are testis specific. We 1C). However, at 2–3 weeks of age, the KO mice began demonstrate that CSP␣KO mice are relatively normal to suffer from progressive weakness. After P15, the KO at birth but exhibit a progressive lethal phenotype that mice stopped gaining weight (Figure 1C). At ⵑP30, CSP␣ manifests as muscle weakness and a sensorimotor dis- KO mice entered into a lethargic state but were able to order at 2–4 weeks of age. Analysis of synaptic transmiseat and move when stimulated. CSP␣KO mice started sion in CSP␣KO mice at the Calyx of Held synapse, a to die in the third postnatal week, most KO mice perished preparation that allows electrophysiological measurein the second postnatal month, and none survived bements of unparalleled resolution, documented that yond 3 months (Figure 1D). Internal examination of KO CSP␣was not required for Ca 2 ⫹channel function or Ca 2 ⫹-triggered exocytosis. Instead, we show that CSP␣mice revealed bilateral intraabdominal cryptorchidism CSP␣Function in Neurotransmitter Release 239 Figure 1. Generation and Characterization of CSP␣KO Mice (A) Schematic diagram of the wild-type CSP␣gene, the targeting vector used for homologous recombination, and the KO allele after gene targeting. In the targeting vector, exon 1 encoding residues 1–36 of CSP␣was replaced by a neomycin resistance gene (neo) for positive selection. The short arm of the vector was flanked by a diphtheria toxin gene (DT) for negative selection. The positions of strategic restriction sites and of PCR primers (“A,” “B,” and “C”) are indicated. Note that mice contain two additional CSP genes (CSP␤and CSP␥; Supplemental Figure S1 at http://www.neuron.org/cgi/content/full/42/2/237/DC1) that were detectable only in testis and not in brain (Supplemental Figure S2). (B) Immunoblot analysis of brain proteins from wild-type (⫹/⫹), heterozygous (⫹/⫺), and homozygous CSP␣KO mice (⫺/⫺) with antibodies to CSP␣, SCAMP1, complexins (Cpx), and synaptobrevin/VAMP 2 (Syb2). (C) Body weights of wild-type, heterozygous, and homozygous CSP␣KO mice as a function of age (n ⫽23 CSP ⫹ / ⫹ ,25CSP ⫹ /⫺ , and 19 CSP ⫺/⫺ mice; data shown are mean ⫾SD). (D) Postnatal survival of CSP␣KO mice. The survival of the offspring from heterozygous matings is plotted as percent of viable mice remaining as a function of age (n ⫽23 CSP ⫹ / ⫹ ,25CSP ⫹ /⫺ , and 26 CSP ⫺/⫺ mice). but no other developmental abnormalities (data not ting up after they were placed on their right or left backside, with a severe deterioration in this condition after shown). Heterozygous mutant and wild-type mice exhib- P25 (Figure 2D). ited no differences in body weight or mortality, sug- The poor performance of CSP␣KO mice in locomotor gesting that the decrease in CSP␣levels in heterozytests suggested that synaptic transmission at the neurogotes is not sufficient to impair survival. muscular junction (NMJ) may have been impaired. To examine this, we studied the electromyographic re- Impaired Sensorimotor Performance and sponses in wild-type and CSP␣KO mice at P15, P23, Neuromuscular Function in CSP␣KO Mice and P45 (Figure 2E). In wild-type mice, the mean ampli- CSP␣KO and littermate control mice were subjected to tudes of compound muscle action potentials (CMAPs) a standard neurological examination (modified SHIRPA were small at P13–P15 but increased ⵑ2-fold at P23 protocol; Rogers et al., 1997). In the open field test, and P45–P47 (Figure 2F and Supplemental Table S2 at CSP␣KO mice exhibited a dramatic loss of spontaneous http://www.neuron.org/cgi/content/full/42/2/237/DC1). activity that was already apparent at P15 (Figure 2A). In CSP␣KO mice, the mean CMAP amplitudes were The impairment was also present to a lesser degree in similar to wild-type littermate controls at P13–P15 and heterozygous CSP␣mutant mice that were normal in all P23 but dropped significantly at P43–P47. This result other parameters examined (Figure 2). CSP␣KO mice indicates an initially robust function at the NMJ which manifested muscle weakness as evident from a tenlater deteriorates. The defects in neuromuscular synapdency to clasp the hindlimbs when suspended by the tic transmission in the CSP␣KO mice became even tail (data not shown) and a lack of gripping strength on more apparent when we examined responses to short fore- and hindlimbs that was observed as early as P15 stimulus trains (Figure 2G). In wild-type mice, repetitive (Figure 2B). The KO mice presented a progressively stimulation elicited the same amount of maximal deprespoorer posture, often with the typical leg position of sion at P15, P23, and P45 (to 64%–69% ⫾5% of initial “inverted champagne bottles” and a thin pelvis due to amplitude). In contrast, we observed enhanced synaptic decreased girdle muscle mass. No tremor was evident. depression in CSP␣KO mice already at P15 (52% ⫾ CSP␣KO mice were unable to stay on a turning rod 30 5%; n ⫽4) and even more severe depression at P23 cm above the floor (Figure 2C) and displayed a poor and P45 (to 35% ⫾3%, n ⫽4, and 39% ⫾6%, n ⫽3, acoustic startle reflex (data not shown). In addition, alrespectively; Figure 2G). Together, these data suggest that CSP␣KO mice exhibit a limited functional impair-ready at P15, the CSP␣KO mice had difficulties in get- Neuron 240 Figure 2. Sensorimotor Deficits in CSP␣KO Mice (A) Spontaneous open field activity of CSP ⫹ / ⫹ , heterozygous CSP ⫹ /⫺ , and CSP ⫺/⫺ mice. The number of squares touched by all four paws in a 30 s time period was measured at postnatal days P15–P40. (B) Grip strength of fore- and hindlimbs (see Experimental Procedures). The percentage of mice capable of effectively gripping onto a grid is displayed as a function of postnatal age. (C) Balance test. Data shown are the times (in seconds) that a mouse can stay on a slowly rotating square thin stick (10 s ⫽maximum). (D) Righting times. Mice are placed on the right or left back side, and the time required to become upright is measured. Note that wild-type controls achieve an upright position instantaneously (dotted line). (E) Representative recordings of compound muscle action potentials (CMAPs) from CSP␣ wild-type and KO mice (P23). (F) Absolute CMAP amplitudes in response to supramaximal stimulation at P13–15, P23, and P43–P7. (G) Normalized CMAP amplitudes during supramaximal stimulation (50 Hz for 1 s) at P15, P23, and P45. All data are means ⫾SEMs. ment of the NMJ at P15 that becomes severe at P23 3-week-old mice for electron microscopy. In many re- and P45. spects, synaptic ultrastructure was normal in mutants: vesicle-laden nerve terminals abutted an enfolded, electron-dense postsynaptic membrane; synaptic vesicles Degeneration of the NMJ were clustered at active zones on the presynaptic mem- The weakness of CSP␣KO mice indicates that their brane; and terminals were capped by processes of NMJs might be defective. We used light and electron Schwann cells. Moreover, the numbers of presynaptic microscopy to test this possibility. First, to assess the active zones and postsynaptic folds did not differ signifisize and shape of the NMJ, we double-stained sections cantly between mutants and controls (Table 1). On the with rhodamine-␣-bungarotoxin (to label acetylcholine other hand, mutant terminals were strikingly abnormal receptors in the postsynaptic membrane) and antibodin two ways. First, 42% (19/45) of terminals contained ies to neurofilaments and SV2 (to label motor axons vacuoles and/or multilamellar bodies, both of which are and nerve terminals). In controls, NMJs were elaborately symptomatic of degenerative processes or defective branched by 3 weeks of age, and each receptor-rich membrane trafficking (Figures 3F and 3G). Such inclubranch of the postsynaptic apparatus was directly apsions were seen in only 6% (2/35) of controls. Second, posed by a presynaptic nerve terminal. In mutants, in slender Schwann cell processes invaginated the surface contrast, the postsynaptic apparatus was less mature, of the nerve terminal at 45% of the sites examined (Figoften having a perforated appearance, which correures 3D and 3E). This number is a minimal estimate sponds to a transitional form in the development of the because we examined only a single section (ⵑ0.1 ␮m “pretzel-shaped” aggregate (Sanes and Lichtman, 1999). thick) from each neuromuscular junction (ⵑ30–50 ␮m Moreover, portions of the postsynaptic apparatus were diameter), so it is possible that all terminals were invagiunoccupied by presynaptic nerve branches, and some nated at least once. In contrast, similar but smaller invanerve branches appeared thinner than normal (Figures ginations were seen in only 2/35 control terminals. The 3A and 3B). Because the pre- and postsynaptic apparati aberrant Schwann cell processes found in the CSP␣KO develop in parallel, these results imply that the nerve mice are quite distinct from those observed in mutants terminal arbor formed normally but then began to degenlacking laminin ␤2 (Patton et al., 1998) or following muserate. Next, we prepared sternocleidomastoid muscles from cle damage (e.g., Jirmanova, 1975). In those cases, the CSP␣Function in Neurotransmitter Release 241 Figure 3. Light and Electron Microscopy of Neuromuscular Junctions from Wild-Type and CSP␣KO Mice (A and B) Confocal images of neuromuscular junctions (NMJs) from wild-type (A) and CSP␣KO mice (B) double labeled with rhodaminetagged ␣-bungarotoxin (red) and antibodies to neurofilaments and SV2 (green). (C–G) Electronmicrographs of NMJs from wild-type (C) and CSP KO mice (D–G) to illustrate morphological changes caused by the deletion of CSP␣(see Table 1 for a quantitation of changes). Arrows in panels (D) and (E) identify Schwann cell intrusions, asterisks in panels (F) and (G) indicate multilamellar bodies, and arrowheads in panel (G) indicate vacuoles. (A)–(F) are from P23 animals; (G) is from a P17 animal (calibration bars: [A], 10 ␮m; [C], [E], and [G], 0.25 ␮m; [B], [D], and [F] are the same magnification as [A], [C], and [E], respectively). processes extend between the presynaptic membrane To ask whether the defects observed in CSP␣KO mice were age dependent, we determined the incidenceand the synaptic cleft and ensheath the nerve terminal. In the CSP␣mutants, in contrast, the terminal is indented of abnormalities in the ultrastructure of NMJs at P7, P14, and P35 (Table 1). Defects were evident at all agesrather than wrapped, and it is the “back side” of the terminal rather than the presynaptic face that is affected. examined but were less striking at P7 than at later Table 1. Ultrastructural Analysis of Neuromuscular Junctions in CSP␣KO Mice % with Active % with Schwann % with Multilamellar Age Genotype n Folds/Profile Zones/Profile Cell Protrusions Vacuole Bodies P7 ⫹/⫹26 1.1 1.4 0.0 42.3 7.7 ⫺/⫺40 1.2 1.4 7.5 22.5 17.5 P14 ⫹/⫹44 5.5 1.6 4.5 9.5 4.8 ⫺/⫺38 4.7 1.6 21.1** 28.9** 28.9** P23 ⫹/⫹35 5.0 1.4 5.7 2.9 2.9 ⫺/⫺45 5.7 1.0 44.4** 24.4** 22.2** P35 ⫹/⫹33 6.6 2.2 6.1 3.0 9.1 ⫺/⫺29 6.0 1.2* 31.0** 27.6** 24.1** NMJs were examined by electronmicroscopy as described (Patton et al., 2001; see also Figure 3). “N” is the number of NMJ profiles counted, and “folds” refers to “junctional folds,” membrane invaginations into muscle fibers directly across from the presynaptic active zone. Data are from two to three mice per age per genotype. Asterisks indicate significant differences (p ⬍0.05) between mutants and age-matched littermate controls (*, unpaired t test; **, contingency table analysis). Neuron 242 stages. For example, the fraction of synapses with invaginated Schwann cell processes in the nerve terminal increased 300% between P7 and P14. In addition, the fraction of nerve terminals in CSP␣KO mice with multilamellar bodies increased ⵑ50% between P7 and P14, despite the fact that more control terminals bore such bodies (and vacuoles) at P7 than at P14 (Table 1). Thus, the synaptic defects resulting from lack of CSP␣are progressive during early postnatal life, suggesting that these defects are responsible, at least in part, for the sensorimotor impairments (Figure 2). Ca 2ⴙ Channel Function in CSP␣KO Mice The initially normal function but subsequent progressive impairment of NMJs in CSP␣KO mice are most consistent with the chaperone hypothesis of CSP function (see Introduction). However, these findings could potentially also be explained by a secondary effect of a role of CSP␣in controlling presynaptic voltage-gated Ca 2 ⫹ channels or in mediating a step in Ca 2 ⫹-triggered exocytosis. To distinguish more definitively between the three hypotheses of CSP function (Ca 2 ⫹channel activity, Ca 2 ⫹-triggered exocytosis, or chaperone; see Introduction), we examined Ca 2 ⫹channel function and synaptic transmission at the Calyx of Held synapse in the medial nucleus of the trapezoid body (MNTB). This synapse is uniquely suitable for precise measurements of Ca 2 ⫹ currents and neurotransmitter release (Borst and Sakmann, 1996) and thus allows a direct test of the role of CSP␣in these processes. We first investigated presynaptic Ca 2 ⫹channel function in CSP␣KO mice but failed to detect significant changes. We observed a similar activation time course and voltage dependence of Ca 2 ⫹currents in CSP␣KO and control mice (Figure 4A) and similar current/voltage relationships of Ca 2 ⫹currents (Figure 4B). At P9–P12, when these measurements were performed, ⬎85% of the presynaptic Ca 2 ⫹current at the Calyx synapse is mediated by Aga-IVa-sensitive, P/Q-type Ca 2 ⫹channels (Takahashi et al., 1996). Consistent with this finding, Aga-IVa toxin (0.2 ␮M) inhibited the majority of the Ca 2 ⫹ currents in wild-type and KO synapses (Figure 4A). The degree of inhibition was indistinguishable between con- Figure 4. Presynaptic Ca 2 ⫹ Currents at the Calyx of Held from Wildtrol and CSP-deficient synapses (wild-type, 93.4% ⫾ Type and CSP␣KO Mice 6.6%; KO, 92.1% ⫾4.2% inhibition; n ⫽3 cells each). (A) Voltage-gated Ca 2 ⫹ currents in Calyx terminals from wild-type It was proposed that CSP␣regulates G protein modu- (top) and CSP␣KO mice (bottom) at P9–P13. Ca 2 ⫹ currents were lation of Ca 2 ⫹channels (Magga et al., 2000). We therefore measured in response to depolarizations between ⫺70 and ⫹60 mV examined whether the coupling of a G protein-coupled in the absence (left traces) or presence (right traces) of agatoxinreceptor to voltage-gated Ca 2 ⫹channels might be im- IVa (Aga-IVa; 0.2 ␮M), a specific P/Q-type Ca 2 ⫹ channel blocker. (B) Current-voltage (I/V) relations obtained from the experiments paired in CSP␣KO mice. The Calyx of Held expresses shown in (A). Black and gray symbols represent values recorded in presynaptic GABA B receptors whose activation causes the absence and presence of Aga-IVa, respectively. a characteristic slowing of Ca 2 ⫹channel opening and a (C) Voltage-gated Ca 2 ⫹ currents recorded in response to three conreduction of Ca 2 ⫹current amplitude (Takahashi et al., secutive step depolarizations to 0 mV in the absence (black traces) 1998; Isaacson, 1998). Figure 4C illustrates the effect or presence of the GABA B receptor agonist baclofen (50 ␮M; gray traces), which slows Ca 2 ⫹ current activation and reduces peak Ca 2 ⫹ current amplitudes in wild-type (left panel) and CSP␣KO mice (right panel). (D) Whole-cell recordings of presynaptic Ca 2 ⫹ currents in response (F) Plot of the maximal Ca 2 ⫹ current amplitude as a function of to step depolarizations to ⫺10 mV, measured in Calyces of Held at postnatal development between P5 and P13. Each data point repre- P5–P6. Recordings were obtained before and after addition of the sents the Ca 2 ⫹ current measured at the peak of the I/V relation for N-type-specific Ca 2 ⫹ channel blocker ␻-conotoxin GVIA (3 ␮M). a single cell (gray symbols, wild-type; black symbols, KO cells). Both (E) Average blocking efficiency of 3 ␮M␻-conotoxin GVIA, exdata sets were fitted by linear regression (indicated by straight lines), pressed in percent of remaining Ca 2 ⫹ current (n ⫽5 wild-type and revealing a slight developmental increase of Ca 2 ⫹ current amplin⫽3 cells CSP␣KO mice). tudes. CSP␣Function in Neurotransmitter Release 243 of 50 ␮M baclofen, a saturating concentration of this selective GABA B receptor agonist, on presynaptic voltage-gated Ca 2 ⫹currents recorded during a 15 ms step to 0 mV. We found no difference between wild-type and CSP␣KO mice (n ⫽3 cells each) in the responses of Ca 2 ⫹currents to baclofen (Figure 4C), indicating that the G protein-mediated coupling of GABA B receptors to P/Q-type Ca 2 ⫹channels, the predominant Ca 2 ⫹channel subtype at this developmental stage, is unchanged. In calyces of the chick ciliary ganglion, whole-cell perfusion with recombinant CSP␣increased the amplitude of presynaptic Ca 2 ⫹currents, suggesting that CSP␣may recruit previously inactive Ca 2 ⫹channels (Chen et al., 2002). However, the presynaptic Ca 2 ⫹currents in the calyces of the ciliary ganglion are predominantly N-type (Stanley, 1991; Yawo and Chuhma, 1994), while N-type Ca 2 ⫹channels account only for a minority of Ca 2 ⫹channels in the Calyx of Held (Wu et al., 1999). To specifically monitor presynaptic N-type Ca 2 ⫹currents in CSP␣KO mice, we examined calyces from younger mice in which the contribution of N-type Ca 2 ⫹channels to the overall Ca 2 ⫹current is larger (ⵑ30%; Iwasaki et al., 2000; Wu et al., 1999). We asked whether the ␻-conotoxin-sensitive component of the presynaptic Ca 2 ⫹current was reduced in young (P5–P6) CSP␣KO mice, as would be expected if functional N-type Ca 2 ⫹channels were recruited by CSP␣(Chen et al., 2002). We applied 3 ␮M␻-conotoxin to slices and monitored presynaptic Ca 2 ⫹channels by voltage steps (to ⫺10 mV, 15 ms) that were applied every 15 s (Figure 4D). Although the magnitude of the ␻-conotoxin block varied between terminals, we again detected no significant difference between wild-type and CSP␣KO mice in the average blocking effect (Figure 4E). Finally, we searched for alterations in the peak Ca 2 ⫹ currents in calyces of Held from CSP␣KO versus wildtype littermate mice (Figure 4F) at ages between P5 and P13. In Figure 4F, each data point represents the maximal Ca 2 ⫹current amplitude for a given cell, taken from I-V relations as shown in Figure 4B. For both genotypes, we observed a slight developmental increase in peak Ca 2 ⫹current amplitude but detected no consistent difference between wild-type and KO cells. Two calyces (D) Confocal image of a brainstem section from a wild-type mouse at P12 labeled with anti-CSP␣antibodies. Ring-like Calyx synapses adjacent to the somata of the MNTB principal cells and smaller, noncalyceal synapses are clearly visible (calibration bar, 20 ␮m). Parallel studies of CSP␣KO slices showed that the staining was abolished (data not shown). Figure 5. Synaptic Transmission at the Calyx of Held in Wild-Type (E and F) Representative traces of EPSCs during a 100 Hz stimu- and CSP␣KO mice at P9–P11 lus train. (A) Representative traces from whole-cell recordings of excitatory (G) Time course of the depression of EPSCs during 100 Hz stimulus postsynaptic currents (EPSCs) in MNTB principal neurons from wildtrains. EPSC amplitudes during five to eight successive stimulus type or CSP␣KO mice elicited by afferent fiber stimulation. trains were normalized to the first response and averaged for all (B) Average EPSC amplitudes in cells from wild-type and CSP␣KO cells (means ⫾SDs; n ⫽5 wild-type/7 KO cells). mice (n ⫽5 and 7, respectively). Each point represents the average (H) Average mEPSC traces for a wild-type cell (left panel, average first EPSC amplitude of individual neurons recorded during four to of n ⫽634 mEPSCs) and for a CSP␣KO cell (right panel, average eight successive 100 Hz stimulus trains (0.2 s) separated by 40–50 s. ofn⫽28 mEPSCs). Note the similar mEPSC amplitudes and kinetics Bars show means ⫾SDs of the averages for each group. for both genotypes. (C) Mean 20%–80% rise times and mean decay time constants of (I) mEPSC amplitudes in wild-type and CSP␣KO mice (n ⫽5 and EPSCs. The EPSC decay could be fitted with a double-exponential 7, respectively). Each point represents the average mEPSC amplifunction; the plot shows the fast decay time constant that accounted tude for a given cell, taken as the mean of the mEPSC amplitude disfor ⵑ80% of the amplitude in wild-type and CSP␣KO cells. tribution. Neuron 244 from CSP␣KO animals exhibited small peak Ca 2 ⫹cur- Kinetics of Release in CSP␣KO Neurons rents that were not found in the littermate wild-type To search for more subtle changes in exocytosis in the controls (Figure 4F). However, since the remaining 23 KO CSP␣KO mice, we next analyzed the slow and the fast calyces were indistinguishable from wild-type controls, components of transmitter release using paired pre- and the average peak Ca 2 ⫹currents were not significantly postsynaptic voltage-clamp recordings. Recordings different between wild-type (1.14 ⫾0.38 nA; n ⫽18) and were performed in the presence of extracellular cyclothi- CSP␣KO mice (1.05 ⫾0.38 nA; n ⫽25). Together, these azide (100 ␮M) and kynurenic acid (1 mM) to reduce data show that CSP␣is not essential for presynaptic desensitization and saturation of postsynaptic AMPA Ca 2 ⫹channel function. receptor responses. Release was triggered by a presynaptic voltage-clamp protocol in which steps to ⫹80 mV Synaptic Transmission in CSP␣KO Mice are interrupted by brief excursions of 0.8 and 2 ms to We next tested the second hypothesis, namely, that 0 mV (Neher and Sakaba, 2001). This protocol measures CSP is directly involved in exocytosis, by measuring the effectiveness of a defined period of Ca 2 ⫹influx in synaptic transmission in the Calyx of Held synapse from inducing transmitter release, monitored via the EPSC wild-type and KO mice at P9–P11. Using afferent fiber (Figure 6A). stimulation in brainstem slices, we applied 20 stimuli at The kinetic parameters of the EPSCs were similar be- 100 Hz and employed postsynaptic whole-cell retween CSP␣KO mice and wild-type littermates at P9– cordings to monitor AMPA receptor-mediated excit- P11 (Figure 6A). EPSCs monitored in these recordings atory postsynaptic currents (EPSCs). decayed slightly slower than EPSCs recorded after fiber We first examined the initial synaptic responses in the stimulation (Figure 5), presumably because the cyclothistimulus trains (Figures 5A). The shapes and amplitudes azide delayed receptor desensitization and deactiva- of the EPSCs were similar in wild-type and KO littermate tion. The average amplitudes of the EPSCs during the mice (CSP⫹ / ⫹, 8.6 ⫾2.5 nA; CSP ⫺/⫺ , 7.88 ⫾2.05 nA; first and second stimulus and of the summed EPSCs Figure 5B), and the 20%–80% rise times of the EPSCs were not significantly different between wild-type and were nearly identical (Figure 5C). To assess the decay KO cells (Figure 6D). This indicates that transmitter re- of the EPSCs, we fitted the responses with a doublelease in response to brief Ca 2 ⫹influx was not altered in exponential function. The fast component accounted CSP␣KO mice. for ⵑ80% of the EPSC amplitude and again was indistin- We then analyzed the kinetics of the rapidly and slowly guishable between wild-type and KO cells (Figure 5C). releasable pools of vesicles after prolonged (ⵑ50 ms) These data suggest that CSP␣is not essential for release presynaptic voltage-clamp depolarization to 0 mV (Saat P9–P11. To confirm that CSP␣is in fact expressed kaba and Neher, 2001). We monitored EPSCs in Calyx at this stage in the Calyx synapse, we performed immuterminals that were depolarized to ⫹80 mV for 4 ms and nofluorescence analyses, which demonstrated that then held at 0 mV for either 4 ms or 50 ms (Figure CSP␣could be reliably detected in Calyx terminals in 6B). The short 4 ms stimulus induced a fast burst of wild-type but not KO mice (Figure 5D and data not transmitter release, while the longer 50 ms stimulus inshown). duced a second slower phase in addition to the initial We then asked whether deletion of CSP␣alters synapfast burst (Sakaba and Neher, 2001). We derived the tic responses during stimulus trains (Figures 5E–5G). cumulative transmitter release rates for both stimuli by We observed a marked short-term depression during deconvolution of the EPSCs (Figure 6C) and fitted the 100 Hz stimulation (Borst et al., 1995; von Gersdorff et time course of cumulative release with double-exponen- al., 1997; Wang and Kaczmarek, 1998) without signifitial functions. We detected no consistent difference in cant differences between wild-type and CSP␣KO cells. the time constants of both phases of transmitter release To analyze this quantitatively, we constructed for each between CSP␣KO and wild-type mice (Figure 6E). Thus, cell a mean depression curve from four to eight succesthe kinetics of transmitter release were unchanged in sive 100 Hz stimulus trains. The depression curves were CSP␣KO mice, suggesting that the coupling of Ca 2 ⫹ normalized and averaged for all wild-type and KO cells channels to vesicle fusion and the Ca 2 ⫹sensitivity of (Figure 5G). The relative amplitude of the second comvesicle fusion were not significantly altered. pared to the first EPSC (CSP⫹ / ⫹, 59% ⫾13%; CSP ⫺/⫺ , 62% ⫾34%) and the near steady-state depression at Progressive Loss of Synaptic Function the end of the 100 Hz trains (CSP⫹ / ⫹, 4.7% ⫾1.5%; at the Calyx of Held CSP ⫺/⫺ , 5.6% ⫾3.7% of initial EPSC amplitude) were The Ca 2 ⫹current and EPSC recordings at the Calyx similar. However, synaptic responses in the CSP␣KO synapse effectively rule out the first two hypotheses were more heterogeneous than in wild-type mice, as about CSP function—an essential role in controlling reflected by their larger error bars (Figure 5G). Ca 2 ⫹channel activity or exocytosis—but raise the ques- Finally, to analyze the quantal properties of transmitter tion whether CSP␣has any function in the Calyx. Since release, we sampled spontaneously occurring miniature at the NMJ the CSP␣KO mice exhibited an age-depen- EPSCs (mEPSCs) in the recording intervals between dent progressive neurodegenerative phenotype that evoked EPSCs. We detected no significant difference was indicative of a chaperone function (see Figures 1–3), between wild-type and CSP␣KO cells in the average we asked whether synaptic transmission at the Calyx shape (Figures 5H) or mean amplitude of the mEPSC of Held also becomes abnormal later in development, (Figure 5I). Together with the unchanged EPSC parameconsistent with such a function. To test this, we examters (Figures 5A–5C), these results indicate that, at P9– ined synaptic transmission in Calyx synapses using af- P11, the amount and kinetics of quantal release in referent fiber stimulation at P20–P23 (Figure 7) with exactly sponse to afferent fiber stimulation are not altered in CSP␣KO mice. the same protocol as described above for mice at P9– CSP␣Function in Neurotransmitter Release 245 5.43 nA [n ⫽5]; Figure 7B) than at P9–P11 (7.88 ⫾2.05 nA). Thus, at P20–P23, the average EPSC amplitude in KO cells is ⵑ2-fold lower than in wild-type cells (p ⫽ 0.016 in a one-tailed, two-sample unequal variance t test), whereas at P9–P11 there was no statistically significant difference between KO and wild-type cells. In these experiments, the average EPSC amplitudes were calculated from the highly reproducible mean EPSCs of individually analyzed cells. These mean ESPCs were variable between mutant cells, as reflected in the large standard deviation of the average shown above, whereas the mean EPSCs were much less variable between wild-type cells (Figure 7B). Although we occasionally observed EPSCs in the range of 5–14 nA in KO cells, some cells produced an initial EPSC amplitude of only ⵑ1 nA (1.03 and 0.98 nA; n ⫽2 cells). In fact, we probably overestimated the average EPSC amplitude in the CSP␣ KO mice at P20–P23, because we used extracellular recordings of pre- and postsynaptic action potentials to preselect neurons in which the Calyx input could be stimulated successfully by afferent fiber stimulation (Borst et al., 1995). Normally, our success rate in finding a neuron with pre- and postsynaptic action potentials is 10%–20% (Meyer et al., 2001). However, in CSP␣KO mice at P20–P23, we recorded action potentials only in ⬍5% of postsynaptic cells, indicating that many calyceal EPSCs were too small to reliably induce a postsynaptic action potential, a result that is consistent with the loss of presynaptic nerve terminals seen morphologi- Figure 6. Paired Recordings of Synaptic Responses at the Calyx of cally (see below). Held from Mice at P9–P12 In wild-type mice, the rise times were shorter, and (A) EPSCs elicited by short periods of presynaptic Ca 2 ⫹ influx. The the fast decay component of the EPSC—which now presynaptic voltage-clamp protocol used (schematically diaaccounted for ⬎95% of the total EPSC—was accelergrammed above the electrophysiological traces) depolarizes the cells to ⫹80 mV (at which point the driving force for Ca 2 ⫹ influx ated at P20–P23 compared to P9–P11 (Figures 7A–7C). is strongly reduced). Ca 2 ⫹ influx is then induced by returning the In contrast, the EPSCs in CSP␣KO mice did not rise membrane potential from ⫹80 mV to 0 mV for 0.8 ms (first stimulus) and decay as smoothly but often exhibited irregularities and 2 ms (second stimulus). (Figure 7A, right panel). The average EPSC rise and (B) EPSCs elicited with longer periods of presynaptic Ca 2 ⫹ influx. In decay times were approximately two times slower in this protocol (shown schematically above the traces), the membrane CSP␣KO mice than in wild-type littermate controls (Figpotential is raised to ⫹80 mV for 4 ms, Ca 2 ⫹ influx is then induced ure 7C). This together with the fact that the kinetics of by repolarization to 0 mV for 4 ms or for 50 ms, and finally Ca 2 ⫹ channels are closed by repolarization to ⫺80 mV. mEPSCs were unchanged (see below; Figure 7H) indi- (C) Integrated transmitter release rates obtained by deconvolution cates that transmitter release was more asynchronous of the EPSCs shown in panel (B) (straight and dashed lines ⫽4 and in the CSP␣KO mice. 50 ms stimulations, respectively). Release rates in response to 50 Furthermore, at P20–P23, CSP␣KO mice exhibited ms depolarizations were fitted with a double-exponential function, dramatic changes in short-term plasticity. Wild-type resulting in estimated time constants of a fast and a slower resynapses experienced less depression of EPSCs during lease component. 100 Hz stimulus trains at P20–P23 than at P9–P12 (e.g., (D) Average EPSC amplitudes measured with the protocol shown in panel (A). Bars represent EPSC amplitudes in response to the compare Figure 5E with Figure 7D). In CSP␣KO synfirst (0.8 ms) and second (2 ms) repolarizations to 0 mV and the apses, however, we often observed facilitation instead summed EPSC amplitudes in response to both stimuli. of depression (Figures 7E and 7F). Again, responses (E) Average time constants of pool depletion estimated in the experiwere heterogeneous (see large error bars in the summents shown in (B) (for panels [D] and [E], CSP ⫹ / ⫹ ,n⫽4; CSP ⫺/⫺ , mary graph of Figure 7G) and occasionally obscured by n⫽7 cell pairs). noncalyceal inputs (Figure 7E). The principal neurons of the MNTB receive a second synaptic glutamatergic in- P11 (Figure 5). As summarized below, we observed draput in addition to the dominant Calyx input. This second matic changes in CSP␣KO mice at P20–P23 that are input generates a delayed EPSC of a smaller amplitude consistent with a functional deterioration of synaptic (Forsythe and Barnes-Davies, 1993; see Figure 7D, transmission. arrows). In some KO cells, the small initial EPSC ampli- In wild-type mice, the initial EPSC amplitude was tude made it difficult to distinguish the calyceal input larger at P20–P23 (13.3 ⫾3.1 nA; [n ⫽5]) than at P9–P11 from noncalyceal excitatory inputs. In order to differenti- (8.6 ⫾2.5 nA). This increase reflects the normal maturaate between calyceal and noncalyceal EPSCs, we theretion of the Calyx synapse (Taschenberger and von Gersfore recorded EPSCs at varying stimulation intensities. dorff, 2000; Iwasaki and Takahashi 2001; Joshi and In all cells included in the final data set, the calyceal Wang, 2002). In CSP␣KO mice, however, the average input was clearly separated from a smaller, presumably noncalyceal input on the basis of their different stimula-initial EPSC amplitude was smaller at P20–P23 (5.62 ⫾