The Drosophila stoned locus encodes two novel gene products termed stonedA and stonedB, which possess sequence motifs shared by proteins involved in intracellular vesicle traffic. A specific requirement for stoned in the synaptic vesicle cycle has been suggested by synthetic genetic interactions betweenstoned and shibire, a gene essential for synaptic vesicle recycling (Petrovich et al., 1993). A synaptic role ofstoned gene products also is suggested by altered synaptic transients in electroretinograms recorded from stonedmutant eyes (Petrovich et al., 1993). We show here that the stonedA protein is highly enriched at Drosophila nerve terminals. Mutant alleles that affect stonedA disrupt the normal regulation of synaptic vesicle exocytosis at neuromuscular synapses ofDrosophila. Spontaneous neurotransmitter release is enhanced dramatically, and evoked release is reduced substantially in such stoned mutants. Ultrastructural studies reveal no evidence of major disorganization at stonedmutant nerve terminals. Thus, our data indicate a direct role for stonedA in regulating synaptic vesicle exocytosis. However, genetic and morphological observations suggest additional, subtle effects ofstoned mutations on synaptic vesicle recycling. Remarkably, almost all phenotypes of stoned mutants are similar to those previously described for mutants of synaptotagmin, a protein postulated to regulate both exocytosis and the recycling of synaptic vesicles. We propose a model in which stonedA functions together with synaptotagmin to regulate synaptic vesicle cycling.
- presynaptic function
- neurotransmitter release
- synaptic vesicle fusion
- synaptic vesicle recycling
Chemical synaptic transmission requires the release of neurotransmitter via the fusion of transmitter-filled synaptic vesicles to the presynaptic plasma membrane (for review, see Sudhof, 1995; Matthews, 1996). Although spontaneous synaptic vesicle fusions occur at a low rate in the absence of extracellular Ca2+ (Katz, 1969), stimulus-evoked vesicle fusion depends on the influx of extracellular Ca2+ into the presynaptic terminal (Zucker, 1996). Ca2+ is known to trigger synaptic vesicle fusion by altering dynamic interactions between vesicle-associated proteins and plasma membrane-associated proteins (Sudhof, 1995; Hanson et al., 1997). The synaptic vesicle protein synaptotagmin is believed to act as a Ca2+ sensor that not only inhibits spontaneous Ca2+-independent vesicle fusion but also promotes Ca2+-evoked exocytosis (DiAntonio et al., 1993;Littleton et al., 1993, 1994; DiAntonio and Schwarz, 1994; Geppert et al., 1994; Sudhof and Rizo, 1996).
Subsequent to fusion, synaptic vesicle proteins deposited in plasma membrane are retrieved by endocytosis and recycled into new synaptic vesicles (Kelly, 1993; Cremona and De Camilli, 1997). The budding of endocytic vesicles from presynaptic membrane appears to involve clathrin and several associated proteins, including the adaptor complex AP-2 (De Camilli and Takei, 1996; Robinson, 1997). Fission of budding vesicles from the plasma membrane occurs by a mechanism requiring the GTPase dynamin, which is altered in Drosophila shibire(shi ts) mutants that show a conditional block in synaptic vesicle endocytosis (Kosaka and Ikeda, 1983; van der Bliek and Meyerowitz, 1991; De Camilli and Takei, 1996; Warnock and Schmid, 1996; Grant et al., 1998). Despite this emerging outline of molecular mechanisms that underlie synaptic vesicle cycling (for review, see Cremona and De Camilli, 1997), it is likely that several molecules that participate in the process remain to be identified and studied.
Classical genetic studies provide one avenue for the characterization of such gene products. Previous studies of Drosophila stonedmutants have suggested that stoned (stn) gene products serve important neuronal functions. The mutantsstn ts2 andstn c exhibit behavioral defects; most notably they are sluggish and uncoordinated, suggesting defects of nervous system function (Petrovich et al., 1993). Transients of the electroretinogram, which are believed to represent synaptic transmission between the photoreceptor layer and the lamina within the fly visual system, are altered in stn ts2(Kelly, 1983) and in stn c mutants (Homyk and Pye, 1989), suggesting defects in synapse development or function. Genetic interactions between stoned and shibiremutations may indicate a more specific neuronal function forstoned. The stn ts2 mutation produces synthetic lethality when combined withshi ts1, suggesting a possible role forstoned gene products in synaptic vesicle recycling (Petrovich et al., 1993).
Such a role for stoned in endocytosis is indicated further by a detailed analysis of the sequence of stoned (Andrews et al., 1996) (see Fig. 7). The stoned gene gives rise to a dicistronic mRNA that encodes two structurally unrelated proteins, stonedA and stonedB, from separate open reading frames (ORFs) (Andrews et al., 1996). Both proteins possess sequence motifs found in proteins known or suspected to regulate endocytic vesicle traffic (see Fig. 7). StonedB contains a C-terminal region with 42% identity to a family of adaptor subunit proteins (Andrews et al., 1996) and an N-terminal region with seven repeats of the trimer NPF, recently identified as a consensus binding motif for the clathrin and adaptor-associated protein Eps15 (Benmerah et al., 1996; Salcini et al., 1997). More sequence analysis presented here (see Fig. 7) shows that stonedA also contains sequence motifs characteristic of molecules involved in endocytosis. Taken together with the genetic interaction betweenstn ts2 andshi ts1, the structural features of stonedA and stonedB suggest that both proteins have presynaptic functions, possibly in clathrin-mediated endocytosis (Andrews et al., 1996; Cremona and De Camilli, 1997).
To investigate directly the potential synaptic functions of thestoned gene products, we examined the effects ofstoned mutations on the function and morphology of synapses on muscles 6 and 7 of the Drosophila third instar larval body wall (Jan and Jan, 1976; Atwood et al., 1993; Kurdyak et al., 1994). Via combined physiological and morphological analyses ofstn ts2 andstn c synapses, we show thatstoned mutations appear to have direct effects on the regulation of neurotransmitter release. Our results also suggest additional subtle effects of stoned mutations on synaptic vesicle recycling and synapse plasticity. During the course of this study it was discovered that stn ts2 andstn c both carry missense mutations within the first ORF of stoned, causing single amino acid substitutions within stonedA (A. M. Phillips and M. Smith, unpublished observations). Thus, our phenotypic characterization ofstn ts2 andstn c mutants provides direct evidence that a stoned gene product, specifically stonedA, acts as a novel regulator of neurotransmitter release.
MATERIALS AND METHODS
Drosophila culture. Flies were raised in vials on instant fly food (Carolina Biological Supply, Burlington, NC) or on medium consisting of instant fly food, agar, and oatmeal (Condie and Brower, 1989) supplemented with yeast. All stocks were maintained at 20–21°C under uncrowded conditions.
Drosophila strains and genetics. Two different wild-type Oregon-R stocks were used, one deriving from Seymour Benzer’s lab (Caltech, Pasadena, CA) and one from Danny Brower’s lab (University of Arizona, Tucson, AZ). stonedmutants stn ts2,stn c,stn 13–120, andstn PH1 were from Len Kelly’s (University of Melbourne, Parkville, Australia) laboratory. The stonedlethal alleles stn 13–120 andstn PH1 are caused by transposon insertions within the stonedA and stonedB ORFs, respectively (Andrews et al., 1996). Whereas stn 13–120 was isolated in a wild-type background, thestn PH1 allele was isolated on a complex chromosome carrying multiple inversions (Zusman et al., 1985); this complex background may cause phenotypes unlinked to stoned. The failure of stn 13–120 andstn PH1 to complement each other suggests that each may impair translation of both stonedA and stonedB proteins via a combination of mRNA truncation and/or polarity effects (Andrews et al., 1996).
To map stn ts2 andstn c physiological phenotypes to thestoned locus, we first mapped these phenotypes to the X chromosome by setting up reciprocal crosses betweenOregon-R and stn ts2 orstn c. Subsequently, we ascertained that the stoned phenotypes were rescued by a duplication ofstoned and uncovered bystn 13–120 andstn PH1 as well as a deficiency forstoned. The duplication Dp(1,Y) y + Ymal +, which consists of a proximal piece of the X chromosome includingstoned attached to the Y chromosome, was from Len Kelly’s laboratory or from the Tata Institute of Fundamental Research, Bombay stock collection. The deficiency stock Df(1)HM430, which covers X chromosome region 20 between uncland l(1)20Cb, was a generous gift of Maurice Kernan (State University of New York, Stony Brook, NY).Df(1)HM430, stn 13–120, andstn PH1 were maintained in stocks over balancers (e.g., FM6) or overy + Ymal +. Because Df(1)HM430 andstn 13–120 could be maintained overy + Ymal +, flies of the genotypesstn c /stn 13–120,stn ts2 /stn 13–120,stn c /Df(stn),stn ts2 /Df(stn),stn c /y + Ymal +, andstn ts2 /y + Ymal +could be generated easily from crosses similar to those shown below:
We controlled extensively for genetic background effects. In larvae of the genotypestoned viable /y + Ymal +, the duplication was derived from one of two different parental stocks (stn 13–120 /y + Ymal +orDf(1)HM430/y + Ymal +). Regardless of parental origin,y + Ymal +always complemented stoned viablephenotypes, except where noted in the text.stn PH1 /y + Ymal +males were usually inviable and, in our hands, survivors were typically sterile. Therefore, stn PH1 was maintained over FM6(T8-lacz), which drives the expression of β-galactosidase from the asense promoter. Female progeny from the crossstoned viable /Y ×stn PH1 /FM6(T8-lacz) were selected for electrophysiology and subsequently were processed for β-galactosidase activity to score the genotype. Briefly, larval CNSs were dissected away and fixed for 3 min in 3% glutaraldehyde in PBS. Fixed CNSs were incubated for 2.5 hr at 37°C in staining solution containing (in mm) 10 PO4 buffer, pH 7.2, 150 NaCl, 1 MgCl2, 3 K4[FeII(Cn)6], and 3 K3[FeIII(Cn)6] plus 0.2% Triton X-100. Just before incubation, a 1:50 volume of freshly prepared 10% X-gal in DMSO was added to staining solution that had been preheated to 65°C. The absence of characteristic blue staining within the CNS was used to identify larvae of the genotypestoned viable /stn PH1.
Larval neuromuscular preparations. For electrophysiology and microscopy of larval neuromuscular junctions (NMJs), larvae were selected from the stocks and dissected to expose the larval body wall muscles, as described previously (Estes et al., 1996). All experiments were performed on muscles 6 and 7 within the third abdominal segment (A3), with the exception of electron microscopy (EM) and stonedA immunohistochemistry studies, in which A2 was included. Muscles 6 and 7 are innervated by a pair of identified motor neurons with well characterized physiological and morphological properties (Jan and Jan, 1976; Atwood et al., 1993; Kurdyak et al., 1994). The motor neurons have distinct firing thresholds and generate a compound response in muscle when they fire simultaneously. During dissections and electrophysiological experiments the larvae were immersed in “HL3” saline (Stewart et al., 1994), which contains (in mm) 70 NaCl, 5 KCl, 1.5 CaCl2, 20 MgCl2, 10 NaHCO3, 5 trehalose, 115 sucrose, and 5 HEPES, pH 7.3. Of those available, the HL3 saline recipe produces an ionic composition and osmolarity most similar to that ofDrosophila hemolymph (Stewart et al., 1994). To prevent muscle contraction that potentially could damage the preparation during dissection, we dissected larvae in low Ca2+ saline or in Ca2+-free saline in which the CaCl2 was replaced by 1.5 mm MgCl2and 0.5 mm EGTA. These treatments did not have any distinguishable effects on the electrophysiological properties of the NMJ.
Electrophysiology. All electrophysiological recordings were made from muscle 6, with the larval preparation immersed in a low volume of the HL3 saline described above. Temperature of the saline was 19–21°C, and was checked often with a thermocouple microprobe (type IT-21, Physitemp, Clifton, NJ). Where noted in the text, Ca2+ concentration sometimes was reduced to subphysiological levels by replacing CaCl2 with MgCl2. In all preparations the CNS was gently cut away to prevent endogenous motor activity. Motor nerves were stimulated with glass-tipped suction electrodes. For recordings of excitatory junctional potentials (EJPs) and excitatory junctional currents (EJCs), an isolated pulse stimulator (A-M Systems, Everett, WA) was used to deliver 1 msec pulses at a frequency of 1 Hz and at an intensity ∼1.5× that required to elicit the compound response. All recordings were acquired with an Axoclamp 2B amplifier in conjunction with pClamp 6 software (Axon Instruments, Foster City, CA). Recording electrodes were pulled from thick-walled borosilicate capillary tubes (FHC, Bowdoinham, ME) with a Sutter Instruments (Novato, CA) electrode puller. For intracellular recordings the electrodes were backfilled with 2 m KAc, yielding resistances of 25–40 MΩ. For two-electrode voltage clamp (TEVC) experiments, the recording electrode was backfilled with 3 m KCl; the tip of the current-passing electrode was filled with 2 m K-citrate and the remainder was backfilled with 3 m KCl. Resistances of recording and current-passing electrodes were 10–20 and 15–35 MΩ, respectively. After impalement with both electrodes, the resting membrane potential of muscle 6 was usually at approximately −50 mV but always was clamped to −70 mV. Input resistance was determined by delivering a 20 mV hyperpolarizing pulse from the −70 mV holding potential. The EJC or EJP amplitude for each preparation was determined from an average of 10 consecutive evoked responses. During TEVC, clamp conditions were optimized so that deviation from the holding potential on stimulation was <5 mV. EJCs were low-pass-filtered at 1 kHz. Although the frequencies of mejcs and mejps appeared similar, mejp recordings were much cleaner and therefore were used to quantify mini frequencies. For each preparation, the number of mejps occurring consecutively within 10 sec was used to determine mini frequency. Although the reported mini frequencies were recorded in physiological saline (Ca2+ = 1.5 mm), reducing extracellular Ca2+ concentration did not appear to affect mini frequency significantly (data not shown).
Confocal microscopy. Dissected larvae (see above) were processed by previously described immunohistochemical procedures (Estes et al., 1996). Briefly, each was fixed in 3.5% paraformaldehyde prepared in PBS containing 0.5 mm EGTA and 0.2 mm MgCl2. After washes in PBS the larval preparation was blocked in 2% BSA and 5% goat serum in PBS containing 0.15% Triton X-100 (TBS). Then the preparation was incubated for 2 hr in primary antibody at an appropriate dilution (see below), followed by a 1 hr incubation in fluorescent secondary antibody at a final dilution of 1:200. The preparation was mounted in 0.1% paraphenylene diamine (Sigma, St. Louis, MO) in 95% glycerol and viewed. Rabbit anti-synaptotagmin (DSYT-2) antibody was a generous gift from Troy Littleton and Hugo Bellen (Baylor College of Medicine, Waco, TX) and was used at a final dilution of 1:200. Mouse anti-cysteine string protein (anti-csp) antibody (mAb49), kindly provided by Erich Buchner (Universitat Wurzburg, Germany) and Konrad Zinsmaier (University of Pennsylvania School of Medicine, Philadelphia, PA), was used at a final dilution of 1:50. Rabbit anti-stonedA antiserum was generated against a stonedA–maltose binding protein (MBP) fusion protein and was used at a final dilution of 1:1000. A Texas Red-conjugated goat anti-rabbit antibody (ICN, Costa Mesa, CA) was used to visualize synaptotagmin and stonedA, and an FITC-conjugated goat anti-mouse antibody (ICN) was used to visualize csp. A Bio-Rad 600 laser-scanning confocal microscope and COMOS software (Bio-Rad, Richmond, CA) were used for all image analyses. For synaptotagmin and stonedA single-label experiments, the YHS filter supplied by Bio-Rad was used. For synaptotagmin/csp double-label experiments, the K1/K2 filter was used. For examination of bouton morphology, 1 μm optical sections were collected with the use of a 60× objective. For examination of synaptotagmin and csp distribution within boutons, 1 μm sections were collected by using 60× power at 5× zoom. Images of anti-stonedA staining were taken at 60× power, 1× zoom.
Electron microscopy. Each larval preparation was processed as described previously (Estes et al., 1996). Briefly, the dissected larva was incubated in Trump’s fixative overnight at 4°C. Then the preparation was washed in 100 mm cacodylate buffer containing 264 mm sucrose. The preparation was post-fixed with 1% OsO4 in 100 mm cacodylate buffer for 2 hr, dehydrated in an ethanol series, and embedded in Epon/Araldite (Embed 812, Electron Microscopy Sciences, Fort Washington, PA). Grids were poststained with 2% uranyl acetate and 1% lead citrate and examined under a Jeol 1200EX electron microscope. Only type I boutons on muscle 6 or 7 within A2 and A3 were selected for analysis. Synaptic vesicle content was determined by manual measurements from only those bouton cross sections that contained one or more active zones and in which subsynaptic reticulum was evident.
Statistics. Numerical data reported in the text are mean ± SEM; p values reported in the text were determined by Student’s t test.
StonedA immunoreactivity is present at presynaptic terminals and is reduced in stnc mutants
Given the genetic and molecular evidence that stonedA may be involved in synaptic functions, we sought evidence for the presence of stonedA at Drosophila synapses. Using an anti-stonedA antiserum (raised against an MBP fusion to residues 27–350 of stonedA), we observed strong immunoreactivity within motor terminals of wild-type larvae (Fig. 1). The staining was eliminated completely by the preincubation of the antiserum with a stonedA–GST fusion protein, indicating that the antiserum recognizes an epitope present on stonedA (data not shown). StonedA immunoreactivity within boutons is not attributable to cross-reacting epitopes, because it is altered substantially in stonedmutants. Although stonedA immunoreactivity is at nearly wild-type levels in stn ts2 boutons, it is barely detectable in stn c mutant motor terminals (Fig. 1). The altered stonedA immunoreactivity withinstn c boutons probably indicates a reduction of stonedA protein levels at nerve terminals, rather than an altered epitope, because the presumptivestn c mutation lies considerably C-terminal to the region of stonedA used to raise the antiserum. Most importantly, the presence of stonedA at nerve terminals suggests a presynaptic function for the protein.
In addition to stonedA immunoreactivity at presynaptic boutons, we also observed a characteristic striated pattern of staining in body wall muscle. Although this could be competed out by preincubation of the serum with stonedA fusion protein, the muscle staining was not affected significantly by the stn c mutation. Thus, we are unable to determine whether stonedA immunoreactivity in muscles truly represents muscle expression (and potential function) of stonedA, or a conserved epitope in a different protein. However, our data unambiguously show that stonedA protein is concentrated at motor nerve terminals.
stoned mutations alter mejp frequency and EJP amplitude
To investigate the role of stonedA in synaptic transmission, we used electrophysiology to assess synaptic function atstn ts2 andstn c mutant NMJs compared with wild-type controls. From current-clamp recordings of miniature excitatory junctional potentials (mejps) as well as TEVC recordings of currents (mejcs), we found an approximately threefold enhancement in the frequency of spontaneous miniature events (minis) atstn ts2 andstn c NMJs relative to wild-type NMJs (Fig. 2 A, Table1). Mini frequencies were 4.8 ± 0.4/sec in wild-type, 14.1 ± 0.6/sec instn ts2 (p = 1.5 × 10−11), and 14.2 ± 0.8/sec instn c (p = 2.6 × 10−8). The elevated mini frequency ofstn ts2 andstn c mutants indicates a relatively high rate of spontaneous synaptic vesicle fusion events atstn ts2 andstn c motor terminals. This phenotype could result from weakened regulation of synaptic vesicle exocytosis.
In addition to altered spontaneous synaptic activity, postsynaptic responses evoked by motor nerve stimulation were altered substantially in stn c, but notstn ts2, mutants (Fig.2 B, Table 2). In physiological saline (Ca2+ = 1.5 mm), the peak amplitudes of EJPs in wild-type (39.2 ± 2.4 mV) andstn ts2 (41.2 ± 2.1 mV) were similar, whereas stn c EJPs were much smaller (20.1 ± 2.4 mV) than wild-type (p= 3.5 × 10−6). To monitor directly the response of the muscle to neurotransmitter secretion, we eliminated voltage-activated components of the evoked response as well as nonlinear summation of quantal events by clamping wild-type andstn c muscle fibers at −70 mV and recording the EJCs. EJC measurements revealed the true extent of defects in stn c-evoked responses: the EJC amplitude of stn c (17.4 ± 3.6 nA) was only 15% of that of wild-type (132.3 ± 13.8 nA;p = 5.7 × 10−5) (Fig.2 B). Because input resistances ofstn c and wild-type muscles were comparable (wild-type, 3.9 ± 0.5 MΩ;stn c, 5.1 ± 0.75 MΩ), this difference clearly arises from defective synaptic transmission atstn c NMJs rather than more general defects in stn c muscles. Thestn ts2 allele appeared almost wild-type in EJC amplitude, suggesting that thestn ts2 mutation affects a domain of stonedA more essential for regulating spontaneous, rather than evoked, synaptic transmission. At room temperature and in the EJC assay,stn ts2 behaves like a weak allele ofstoned. This is consistent with our observation that stonedA immunoreactivity is clearly reduced instn c, but notstn ts2, mutants. Although originally isolated in a screen for mutants paralyzed at 29°C, in reality the behavioral phenotype of stn ts2 mutants (sluggishness and uncoordination) shows minimal temperature sensitivity. Consistent with observations on the behavior ofstn ts2 mutants, there is little effect of elevated temperature (30°C) on the electrophysiological phenotypes we observe (data not shown). In summary, our physiological analysis indicates that stoned mutations not only elevate the frequency of spontaneous synaptic vesicle fusions but also decrease evoked synaptic transmission, phenotypes consistent with a role for stonedA in the regulation of neurotransmitter release.
We confirmed that the electrophysiological defects instn ts2 andstn c larvae derive from mutations in thestoned locus by ensuring that these phenotypes were uncovered by lethal genetic lesions of stoned and were complemented by duplications of stoned. For uncoveringstoned phenotypes, we used the deficiencyDf(1)HM430, which deletes the region of the X chromosome containing stoned (region 20B) and two homozygous lethal alleles of stoned, stn 13–120 andstn PH1. The duplication Dp(1,Y) y + Ymal + is a modified Y chromosome carrying region 20 of the X chromosome, which includes the stoned locus; for simplicity, we refer to this duplication asy + Ymal +. As shown in Table 1, the electrophysiological phenotypes ofstn ts2 andstn c were uncovered by Df(1)HM430, stn 13–120, andstn PH1 and were complemented byy + Ymal +. All phenotypes we observed were recessive: EJPs and mejps instn ts2 /+, stn c /+ Df(1)HM430/+, stn 13–120 /+, stn PH1 /+, and+/y + Ymal +were similar to wild-type. Thus, allstn ts2 andstn c phenotypes map to thestoned locus (Table 1).
Interestingly, stn ts2 fails to complement fully the reduced EJP phenotype of stn calthough stn ts2 homozygotes do not show detectable reductions in EJP amplitude. This observation further suggests that stn ts2, in terms of the EJP phenotype, is a weak, partially expressive allele of stoned(Table 1). Overall, our complementation studies show that the synaptic transmission defects of stn ts2 andstn c map specifically to thestoned locus. The recessive nature of the phenotypes suggests that the synaptic defects derive from a loss of function or expression of stonedA.
Enhanced transmission variability, increased transmission failure, and decreased quantal content indicate thatstnc impairs evoked neurotransmitter release
Although the enhanced mini frequencies ofstn ts2 andstn c point to a presynaptic function for stonedA, the decrement in evoked response atstn c NMJs theoretically could be a consequence of either presynaptic or postsynaptic defects. We found that EJP amplitudes were twice as variable in individualstn c larvae as compared with individual wild-type larvae (p = 0.002) (Fig.3 A). Because the number of neurotransmitter receptors and ion channels on postsynaptic membrane should not show stochastic variation, the enhanced EJP variability most likely reflects variability in the number of neurotransmitter quanta released on stimulation of stn c motor terminals. This suggests that stn cmutants do not suffer a simple block in postsynaptic responsiveness to neurotransmitter. The enhanced EJP variability ofstn c mutants was not complemented byy + Ymal +. Nonetheless, EJP variability was low instn c /+ heterozygotes and was exacerbated by stn 13–120,stn PH1, and Df(1)HM430 (Fig.3 A). Thus, loss of stonedA function may reduce the tight coupling between neurotransmitter release and voltage-dependent Ca2+ entry at nerve terminals.
Further evidence for defects in evoked neurotransmitter release atstn c NMJs is provided by failure–frequency analysis (Fig. 3 B). We reduced extracellular Ca2+ to 0.3 mm, a level at which quantal variation in EJPs can be observed in wild-type. We then compared how often synaptic transmission failed atstn c and control NMJs subjected to repetitive stimulation. Under these conditions, nerve stimulation failed to evoke a postsynaptic response nearly 50% of the time atstn c NMJs, whereas failure occurred only 16% of the time at wild-type NMJs (p = 0.003) (Fig. 3 B). The failure frequency ofstn c /stn 13–120NMJs was not significantly different from that of stnc NMJs, and the failure frequency ofstn c /y + Ymal +NMJs was not significantly different from that of wild-type NMJs (Fig.3 B). The enhanced rate of transmission failure atstn c NMJs under low Ca2+ conditions strongly suggests an impairment of Ca2+-dependent neurotransmitter release fromstn c presynaptic terminals.
To assess the magnitude to which Ca2+-dependent neurotransmitter release is impaired instn c mutants, we determined quantal content, a direct estimate of the number of synaptic vesicles fusing during a single evoked event. Because EJCs and mejcs represent linear responses of the muscle to neurotransmitter, we calculated mean quantal content (m) for wild-type andstn c from the mean peak amplitudes of EJCs and mejcs (m = EJC/mejc) at 1.5 mmCa2+. Mean quantal content atstn c NMJs (m = 26.6 ± 5.6) was fivefold smaller than that at wild-type NMJs (m = 127.8 ± 11.2; p = 2.2 × 10−5) (Fig. 3 C). An interesting observation is that, when compared with wild-type mejcs (1.0 ± 0.07 nA), stn ts2 mejcs (0.8 ± 0.05 nA; p = 0.03) and stn cmejcs (0.6 ± 0.03 nA; p = 5 × 10−4) were consistently smaller in amplitude. The reduced quantal size in stn ts2 andstn c mutants raises the possibility that muscle-derived stonedA, the presence of which is hinted at by immunocytochemistry (see Fig. 1), in some way regulates postsynaptic sensitivity to neurotransmitter. It is equally plausible that alterations in presynaptic stonedA affect either the amount of neurotransmitter packaged within synaptic vesicles or change postsynaptic responsiveness via a homeostatic mechanism. Although further experiments are required to investigate these possibilities, reduced quantal content at stn c NMJs unequivocally indicates that mutations of stonedA affect neurotransmitter release.
We next sought to determine whether this reduction in Ca2+-dependent neurotransmitter release atstn c NMJs represents a static uncoupling of neurotransmitter release from presynaptic Ca2+entry or a change in the dynamic Ca2+ sensitivity of the release machinery, as observed in some synaptotagminmutants (Littleton et al., 1993). An estimate of Ca2+ sensitivity can be obtained from the slope of a log–log plot of extracellular Ca2+ concentration versus peak EJC amplitude. We thus recorded EJCs at 0.3, 0.5, and 0.7 mm extracellular Ca2+, a concentration range unlikely to saturate the release machinery (Fig.4). We found that the log–log relationships between extracellular [Ca2+] and EJC amplitude were comparable for wild-type,stn c, andstn ts2 (Fig. 4). Thus, although Ca2+-dependent neurotransmitter release is impaired at stn c NMJs, the sensitivity of neurotransmitter release to relative Ca2+ levels is unaltered.
Subtle alterations in presynaptic morphology cannot explain the physiological phenotypes of stoned mutants
The electrophysiological phenotypes ofstn ts2 andstn c mutants demonstrate that stonedA is required for normal levels of spontaneous and evoked synaptic activity at the third instar larval NMJ. To exclude the possibility that these phenotypes arise from gross disorganization of synapses, we examinedstoned nerve terminals by EM. We examined general synapse ultrastructure and quantified the number, density, and size of synaptic vesicles in wild-type, stn ts2, andstn c boutons (Fig.5 A–C). The fractional area occupied by vesicles was not significantly different among wild-type,stn ts2, andstn c boutons, although there was a trend toward wider synaptic vesicle distribution withinstn ts2 andstn c boutons (Fig. 5 D). We found that wild-type and stn ts2 boutons had approximately equivalent synaptic vesicle densities, whereasstn c boutons had a small but significant increase in synaptic vesicle density (p = 0.009) (Fig. 5 D). Synaptic vesicles themselves were indistinguishable between wild-type (diameter, 34.0 ± 0.47 nm) and stoned mutants (32.8 ± 0.51 nm instn ts2; 34.2 ± 0.51 nm instn c). Thus, we observed no major ultrastructural changes in stoned mutant motor terminals; specifically, there was no depletion of morphologically defined synaptic vesicles and no gross abnormalities that could form the basis for the electrophysiological defects of stoned mutants.
We further counted the number of synaptic contact sites (boutons) over larval muscles 6 and 7. An increased number of synaptic sites could, in principle, produce the enhanced mini frequencies of stonedmutants; likewise, a decreased number of synaptic sites could give rise to the small evoked response of stn cmutants. To gain insight into potential origins of these phenotypes, we used an antibody to synaptotagmin to visualize and examine neuromuscular synapses of wild-type,stn ts2, andstn c NMJs (Fig.6 A–C). We found that the number of boutons per arbor was slightly, but significantly, increased at both stn ts2 andstn c NMJs (83.3 ± 4.6,p = 0.04; 96.2 ± 3.5, p = 9.8 × 10−6) as compared with wild-type (70.6 ± 3.8) (Fig. 6 D). Although these differences raise interesting questions, the increase in bouton number cannot explain the decrease in evoked synaptic transmission atstn c NMJs. Furthermore, these differences represent only 18 and 36% increases in bouton number atstn ts2 andstn c NMJs, respectively, and cannot account for the ∼300% increase in mini frequency atstn ts2 andstn c NMJs. Thus, the physiological phenotypes of stn ts2 andstn c mutants most likely arise from defective regulation of synaptic vesicle fusion within individual boutons.
Unexpectedly, during our light microscopic analyses ofstn ts2 andstn c NMJs we observed distinct alterations in the presynaptic localization of the synaptic vesicle proteins synaptotagmin and cysteine string protein (csp) atstn c terminals. These alterations were difficult to quantify, but in “blind” tests we were able to distinguish between wild-type and stn cpreparations stained with anti-synaptotagmin (or anti-csp) antibodies with close to 100% accuracy. As shown previously (Estes et al., 1996), the distribution of synaptotagmin and csp within wild-type presynaptic boutons reflects the expected locations of synaptic vesicle pools—synaptotagmin and csp both are highly concentrated within boutons but are excluded from intervening axonal regions (Fig.6 E) (data not shown). There also are small regions within boutons that are devoid of synaptotagmin and csp.
In general, stn c mutants appeared to show an increase in synaptotagmin and csp immunoreactivity at the bouton periphery (Fig. 6 G). In addition, we frequently observed synaptotagmin and csp immunoreactivity in axonal regions connecting boutons (Fig. 6 G). Alterations instn ts2 boutons were more subtle, and somestn ts2 preparations were not easily distinguishable from wild-type (Fig. 6 F). These altered patterns of synaptotagmin and csp immunoreactivity instoned boutons could suggest that synaptotagmin and csp are not recycled efficiently from the plasma membrane and consequently diffuse laterally through the membrane. However, no synaptic vesicle depletion or major morphological alterations are apparent instn c nerve terminals viewed under the electron microscope. Possible interpretations of the morphological alterations we observe under the light microscope are considered in Discussion.
The stoned gene first was associated with nervous system function when the mutants stn ts2and stn c were isolated on the basis of obvious behavioral abnormalities (Grigliatti et al., 1973; Homyk, 1977;Homyk and Sheppard, 1977). Subsequent genetic and phenotypic studies provided indirect evidence that stoned gene products might be required for synaptic transmission (Petrovich et al., 1993; Andrews et al., 1996). In this paper we demonstrate that a product of thestoned gene is present at presynaptic terminals and is involved in the regulation of neurotransmitter release. Thus, our analysis provides the first direct evidence that a stonedgene product regulates presynaptic functions.
Genetic and phenotypic analysis of stoned
The Drosophila stoned locus is extremely unusual in its organization: a single dicistronic transcript from the gene encodes two structurally unrelated polypeptides termed stonedA and stonedB (Andrews et al., 1996) (Fig. 7). Thus, unlike most genetic loci, single mutational events could cause phenotypes that result from defects in both molecules. This is likely the case for the lethal alleles of stoned used in this study (Andrews et al., 1996). However, the viable alleles of stoned,stn ts2 andstn c, recently have been shown to carry missense mutations that alter the stonedA product (A. M. Phillips and M. Smith, unpublished observations). Although complex effects of these single base substitutions on transcript stability or on translation rates of the second cistron cannot be excluded completely, it is likely that these missense mutations affect activity of only the stonedA protein. StonedA is enriched at nerve terminals where its levels are reduced substantially in stn cmutants (see Fig. 1). Thus, it is likely that the neurological phenotypes of stn ts2 andstn c mutants result from altered stonedA function.
Our analyses of stn ts2 andstn c mutants and other allelic combinations that survive to the third larval instar show that several properties of neurotransmitter release are altered by a disruption of stonedA. Mutant stn c andstn ts2 larvae exhibit similar rates of spontaneous neurotransmitter release that are elevated approximately threefold over wild-type (see Fig. 2 A, Table 1). The quantal content of the evoked response is reduced significantly instn c mutants (see Fig. 3 C), a phenotype that is not complemented bystn ts2, althoughstn ts2 homozygotes do not themselves show perceptible defects in evoked transmitter release (see Fig.2 B, Table 2). In addition to reduced quantal content,stn c mutations cause reduced fidelity in neurotransmission (see Fig. 3 A,B). All phenotypes ofstn c orstn ts2 homozygotes described above are uncovered by deficiencies or lethal alleles of stoned and complemented by a small chromosomal duplication that includes thestoned locus (Tables 1, 2); thus, these defects are caused by mutations in stoned.
Although stonedA may have additional, as yet unknown, functions in other tissues, our electrophysiological studies suggest that an important function of stonedA is to regulate the presynaptic release of neurotransmitter. However, these studies alone do not exclude the possibility that functional defects in stoned mutants arise from a primary defect in, for instance, synapse development or organization. To address this issue, we examined stonedsynapses by EM. Our analysis did not reveal any ultrastructural defects that could account for the physiological defects we observed atstoned synapses. Thus, the data suggest that altered regulation of synaptic vesicle fusion is the primary defect instoned mutants. An alternative that we have not excluded in our studies is that a subset of synaptic vesicle membrane proteins, required for function but not for assembly of synaptic vesicles, is not recycled correctly in stoned mutants. In this scenario the altered regulation of transmitter release that is observed instoned mutants could arise from a primary defect in an unprecedented, novel pathway for the recycling of specific membrane component(s) of synaptic vesicles. Below, we discuss and try to reconcile these two alternative roles that stonedA may play in synaptic vesicle cycling.
StonedA as part of the Ca2+-sensitive fusion clamp
The elevated rate of spontaneous neurotransmitter release from stn ts2 andstn c motor terminals suggests that in wild-type synapses stonedA function limits the rate of Ca2+-independent synaptic vesicle fusions. Furthermore, the reduction in evoked neurotransmitter release fromstn c motor terminals suggests that stonedA normally promotes Ca2+-dependent synaptic vesicle fusions. Such a dual role in regulating synaptic vesicle exocytosis has been postulated previously for synaptotagmin, a synaptic vesicle membrane protein with two calcium-binding C2domains (Popov and Poo, 1993). Mutations of synaptotagmin inDrosophila give rise to phenotypes strikingly similar to those of stoned mutants (Littleton et al., 1993, 1994;DiAntonio and Schwarz, 1994). Like stn cmutants, Drosophila synaptotagmin (syt) mutants exhibit an approximately threefold enhanced rate of spontaneous neurotransmitter release and a severe reduction in evoked neurotransmitter release. These observations led to the proposal that synaptotagmin acts as a Ca2+-sensitive fusion clamp that inhibits synaptic vesicle fusion in the absence of Ca2+ but facilitates synaptic vesicle fusion in response to presynaptic Ca2+ entry. Such a role for synaptotagmin is supported by a number of biochemical studies (for review, see Sudhof and Rizo, 1996; Bennett, 1997).
The biochemical properties of stonedA are not as well characterized as those of synaptotagmin, and so the detailed mechanisms by which it may regulate transmitter release remain to be discovered. Unlike the case for synaptotagmin, there is no evidence that stonedA is capable of sensing and responding to cytosolic Ca2+ levels. First, although stonedA contains clusters of acidic residues (see Fig.7), it does not contain C2 domains or any other consensus Ca2+-binding domains. Second, although synaptic vesicle fusion in stn c mutants is uncoupled partially from Ca2+, evoked neurotransmitter release in the mutant remains sensitive to Ca2+ (see Fig. 4). A recent observation that recombinant stonedA can bind to recombinant synaptotagmin in vitro (A. M. Phillips, unpublished results) opens the possibility that stonedA interacts with synaptotagmin to regulate synaptic vesicle fusion in a Ca2+-dependent manner.
A possible role for stonedA in synaptic vesicle recycling
Genetic interactions between stn ts2and shi ts1 led to the suggestion that astoned product might regulate synaptic vesicle recycling (Petrovich et al., 1993). Our recent analysis of the stonedA sequence shows that its C-terminal region contains four DPF repeats (see Fig.7), which define a C-terminal domain of Eps15, a protein believed to be involved in membrane internalization (Benmerah et al., 1996). A functional role for the DPF motif is suggested by a recent study that demonstrates that the binding of human Eps15 to the α-adaptin subunit of AP-2 is mediated by a portion of Eps15 that contains four of these DPF repeats (Benmerah et al., 1996). It may be significant that stonedB contains Eps15-binding domains that potentially could interact with the Eps15-like sequences of stonedA.
When we screened protein databases for other polypeptides with DPF repeats, we discovered a rather selective enrichment of two or more closely spaced DPFs in several presynaptic proteins, including Eps15, the clathrin-uncoating cofactor auxilin (Ungewickell et al., 1995) (Entrez accession number S68983; three DPFs), and the clathrin assembly protein AP180 (Ahle and Ungewickell, 1986) (accession number S36327; two DPFs). Other proteins containing DPF repeats are heat–shock proteins, the FMRFamide-related neuropeptides (accession numbersS38816, Q07981, P42565; 7–14 DPFs), phospholipase A-2-activating protein (accession numbers P27612, P54319; three DPFs), and phospholipase C (accession numbers A31225, P13217; two DPFs). Although we can only speculate about the function of DPF repeats in stonedA, the presence of multiple DPFs in Eps15, auxilin, and AP180 indicates a common structural feature among stonedA and proteins thought to be involved in clathrin-mediated endocytosis.
Despite the genetic interactions betweenstn ts2 andshi ts1, and the sequence of stonedA, there is little direct evidence that stoned mutations affect synaptic vesicle recycling. Our EM data show that synaptic vesicles are not depleted in stn c andstn ts2 nerve terminals, and these vesicles are not morphologically distinguishable from wild-type vesicles. However, as shown in Figure 6, under the fluorescence microscope we observe an altered distribution of the synaptic vesicle membrane proteins synaptotagmin and csp instn c mutants. This redistribution is consistent with an accumulation of synaptotagmin and csp throughout the presynaptic plasma membrane of stn ts2 andstn c mutants (Estes et al., 1996). If synaptotagmin and csp are not retrieved efficiently from the plasma membrane in stoned boutons, it seems puzzling thatstoned boutons show no alterations in synaptic vesicle number, distribution, or size. It is possible that a subset of synaptic vesicle proteins, including csp and synaptotagmin, recycle via a stonedA-dependent pathway, whereas others, required for the formation of morphologically identifiable synaptic vesicles, recycle in a stonedA-independent manner. If this were the case, then altered neurotransmitter release in stoned mutants may be a secondary consequence of ineffective synaptotagmin recycling into synaptic vesicles. A second, more conservative alternative is that the apparent redistribution of synaptotagmin and csp, although unequivocal under the fluorescence microscope, derive from a relatively minor fraction of synaptic vesicle proteins that are localized aberrantly. In this model, stoned mutants may have a minor defect in recycling of all synaptic vesicle membrane proteins; this defect is visible in our immunofluorescence studies but is so subtle as to be invisible under the electron microscope. Although additional experiments are required to assess the endocytosis of different synaptic vesicle membrane proteins instn ts2 andstn c mutants, our working hypothesis is that stonedA regulates both synaptic vesicle fusion and endocytosis.
StonedA is not the first protein for which dual roles in synaptic vesicle exocytosis and endocytosis have been proposed. Both biochemical and genetic studies suggest that, in addition to its role in regulating synaptic vesicle fusion, synaptotagmin may regulate synaptic vesicle recycling (Zhang et al., 1994; Jorgensen et al., 1995; Sudhof and Rizo, 1996). Although the possible role of synaptotagmin in synaptic vesicle recycling has not been addressed in studies of Drosophila syt mutants, Caenorhabditis elegans synaptotagmin(snt-1) mutants exhibit phenotypes consistent with impaired synaptic vesicle recycling (Jorgensen et al., 1995). For example, the synaptic vesicle protein synaptobrevin exhibits an abnormally diffuse distribution in the snt-1 nerve cord, suggesting an accumulation and lateral spreading of this protein within neuronal plasma membrane (Jorgensen et al., 1995). This altered distribution of synaptobrevin in snt-1 nerve cord is qualitatively similar to the altered distributions of synaptotagmin and csp instoned boutons. Thus, parallels between stonedmutants and synaptotagmin mutants suggest that stonedA and synaptotagmin may share functions in synaptic vesicle recycling as well as in synaptic vesicle fusion.
If stonedA does in fact regulate general synaptic vesicle endocytosis, could defects in transmitter release atstn ts2 andstn c synapses be secondary to a primary defect in synaptic vesicle recycling? We believe this to be unlikely for the following reasons. First, our EM studies show that any recycling defects at stn ts2 andstn c mutant synapses must be very subtle. Second, partial depletion of synaptic vesicles, which may be achieved by stimulating shi ts1 mutants at nonpermissive temperature, does not result in the specific phenotypes we observe at stoned mutant synapses (Koenig et al., 1983). The elevated mini frequency and the reduced evoked release are unique phenotypes of stoned mutants and thus probably reflect a specific function of stonedA in regulating Ca2+-dependent neurotransmitter release. The simplest interpretation of our data is thatstn ts2 andstn c mutants have independent defects in synaptic vesicle fusion and synaptic vesicle recycling.
Other possible functions of stonedA
In addition to regulating basal presynaptic functions, stonedA may be involved in a cAMP-dependent pathway that regulates synapse plasticity. Genetic studies have shown synthetic lethality betweenstn ts2 anddnc M14, a mutation that affects a cAMP phosphodiesterase (Petrovich et al., 1993). The dunce gene product has been implicated in learning and memory and appears to be an important component of cAMP-responsive pathways that regulate gene expression as well as structural and functional remodeling of synapses (Davis, 1996). It is interesting thatstn ts2 andstn c mutants cause small but significant increases in bouton number at the larval NMJ, a phenotype that has been described for dnc M14 (Zhong et al., 1992; Schuster et al., 1996). This somewhat tenuous association between a mutation in the cAMP pathway and stoned could indicate that stonedA participates in the regulation of synapse plasticity and perhaps plays some role in learning and memory (Petrovich et al., 1993).
Identification of other gene products that interact with stonedA, either directly or indirectly, will facilitate further characterization of stonedA function. In addition, identification of stonedA homologs in other organisms will allow for the development of further tools to examine possible functions of stonedA. We have used a classical genetic approach to manipulate stonedA activity at a relatively simple, experimentally accessible synapse in Drosophila. Our results establish that stonedA is an influential regulator of synaptic vesicle exocytosis and suggest that stonedA may have additional functions in synaptic vesicle recycling and synaptic remodeling.
This work was funded by National Institutes of Health Grant NS34889 (to M.R.) as well as by the McKnight and Alfred P. Sloan foundations. D.S. acknowledges support from Developmental Neuroscience Research Training Grants at the University of Arizona, funded by the Flinn Foundation and National Institutes of Health. We thank A. Marie Phillips for permission to cite her unpublished results and K. S. Krishnan, Sujata Rao, and A. Marie Phillips for useful discussions. We are grateful to Rick Levine and Andrea Yool as well as to Christos Consoulas and Rebecca Johnston of the Levine lab for help with electrophysiology. We acknowledge Patty Jansma for assistance with confocal and electron microscopy. We also thank John Calley for his help in searching protein databases. The microscopy was performed with a Bio-Rad 600 confocal microscope and a Jeol 200EX electron microscope belonging to the Arizona Research Laboratories Division of Neurobiology. We thank Troy Littleton and Hugo Bellen for anti-syt antibodies and Erich Buchner and Konrad Zinsmaier for anti-csp antibodies. This manuscript was improved substantially by comments from Rick Levine and Dave Sandstrom.
Correspondence should be addressed to Daniel T. Stimson, Department of Molecular and Cellular Biology, Life Sciences South, Room 444, University of Arizona, P.O. Box 210106, Tucson, AZ 85721.