Disruption of synaptic development and ultrastructure by Drosophila NSF2 alleles

Drosophila stocks

The yw strain was used as a control strain throughout. The UAS/GAL4 expression system was used to drive gene expression in the nervous system. To express transgenes specifically in nerve and not in muscle, the GAL4 driver line elav-Gal4 was used. This line expresses GAL4 only in postmitotic neurons. elav^3A-Gal4 is a P-element construct on the third chromosome comprising the elav promoter and Gal4 coding sequence. Construction of the UAS- NSF2^E/Q line was previously described (Stewart et al.,2001). A recombinant third chromosome bearing elav^3A-Gal4 and UAS- NSF2^E/Q was obtained by recombining the UAS transgene onto the same chromosome bearing elav^3A-Gal4. The recombinant chromosome was maintained in a yw background with the third chromosome balancer TM3, Sb Ser y⁺. For the electron microscopic studies herein we used yw; elav^3A-Gal4: UAS- NSF2^E/Q/TM3 Sb,Ser,y+ larvae as our experimental subjects and the elav^3A-Gal4 line and the UAS- NSF2^E/Q line as controls.

For the embryo studies undertaken here, virgin yw females was crossed to yw; elav^3A-Gal4: UAS- NSF2^E/Q/TM3 Sb,Ser,y+ males. The resulting yw; elav^3A-Gal4: UAS- NSF2^E/Q/+ embryos were identified by the reduced pigmentation of the denticle belts and mouth hooks in Stage 16 embryos. The yw; +/TM3 Sb,Ser,y+ siblings served as controls.

The genetic loss of function allele NSF2⁵⁵ was generously supplied by L. Pallanck (Univ. of Washington) and is described in Golby et al. (2000). These alleles were maintained with TM3, KrGFP balancers and homozygous mutant embryos were identified by lack of GFP signal. NSF2⁵⁵/TM3 KrGFP siblings served as controls.

Embryo collection and dissection

For analysis of embryonic NMJs, adult flies of the appropriate genotype were allowed to lay eggs on grape-agar medium. Typically, the adults were kept on medium for 4–6 hours in the afternoon, the medium removed, and the embryos were allowed to develop overnight at room temperature (22–23°C). Embryos were then dechorionated by hand on double-sided tape and Stage 17 embryos were selected for dissection. These embryos were removed from the vitelline membrane under HL3 saline (Stewart et al.,1994) using glass or tungsten dissecting needles. Their heads and tails were then glued to the Sylgard-coated slides, using Nexaband glue, and a dorsal midline incision was made using the needles. The digestive system and fat tissue were removed and the cuticle laid out along the incision and glued to the slide.

Immunofluorescence, image acquisition, and analysis

Following dissection, the preparations were immediately fixed in 3.7% formaldehyde in phosphate-buffered saline (PBS), pH 7.2, for 10–20 minutes. They were then washed on the slide for 3 × 5 minutes in PBT (PBS plus 0.1% Triton X-100) and stained with FITC-conjugated goat anti-HRP, an antibody that labels Drosophila neurons, for 1 hour at room temperature. Following antibody labeling the preparations were washed in PBT, mounted in Vectashield, and coverslipped.

Images were acquired on a Zeiss LSM510 confocal microscope using the 488 line of an argon laser with a 530 nm long-pass filter. A 63×, 1.3 n.a. oil immersion lens was used to acquire Z-sections that were acquired with the pinhole aperture set to acquire images of 1 μm thick. Image sections were then projected onto a single plane and nerve terminal branch lengths were measured using ImageJ software.

Electron microscopy

To process the tissue for ultrastructural analysis, we followed the method of Atwood et al. (1993). Wandering third instar larvae were dissected in HL3 physiological solution and fixed in 0.1 M sodium cacodylate buffer (pH 7.4) containing 1% glutaraldehyde and 4% formaldehyde for 2 hours at room temperature. Following 15 minutes of rinsing in 0.1 M sodium cacodylate, the tissue was postfixed for 30 minutes in 1% osmium tetroxide in 0.1 M sodium cacodylate. The tissue was then dehydrated through an ethanol series and individual abdominal hemisegments 3 or 4 were dissected from the larvae and embedded in Epon-Araldite resin.

The resin blocks were trimmed and thick cross-sections were taken to identify the bodywall muscles. Upon identification of an area of innervation, thin sections (75–100 μm) were cut and collected on Formvar-coated slot grids and doubly stained with uranyl acetate and lead citrate, then examined with a Hitachi H-7500 electron microscope. Images were either collected digitally on a MegaView II camera or by taking photomicrographs.

The ultrastructural data from mutant larvae in this report is based on two separate serials totaling 194 sections from one yw; elav^3A-Gal4: UAS- NSF2^E/Q/TM3 Sb, Ser,y+ larvae. We additionally collected a third series of sections from a second mutant larvae that consisted of 370 sections (∼27 μm), that were periodically examined; that examination confirmed our results from the more detailed measurements. For controls, the serial sectioning consisted of one series containing 102 sections from a elav^3A-Gal4 larva. In total, we measured data from 11.7 μm of nerve terminal from the control and 24.5 μm of nerve terminal from the NSF2^E/Q strain.

Data analysis

Data were complied and analyzed in Prism3 software (GraphPad, San Diego, CA). Student's t-test was used to test for statistically significant differences between two genotypes, while one-way ANOVA was used for multiple genotype comparisons, with Dunnett's multiple comparison test used for post-hoc analysis of the difference between experimental and control genotypes.

Early developmental defects at NSF2 mutant NMJs

A previous report on the effects of neural expression of dominant-negative NSF2 in Drosophila neurons showed that there is dramatic overgrowth of the neuromuscular junction in third instar larvae; the total branch length in such larvae is nearly double that of control larvae (Stewart et al.,2002). Examples of larval neuromuscular junctions from three control strains and the NSF2^E/Q strain are shown in Figure 1. In order to distinguish between the possibilities that the overgrowth phenotype is present early in development or acquired later, we examined NMJ morphology at the embryonic NMJ and found that the NMJ show signs of overgrowth and hypersprouting at this early developmental time point (Fig. 2). To quantify the overgrowth we measured the total branch length on muscle 6 and 7 in abdominal segments 3 and 4 for each of the control and NSF2^E/Q strains. In control embryos the total branch length was 29.2 ± 1.3 μm (n = 13) whereas NSF2^E/Q neuromuscular junctions were 42.8 ± 2.0 μm (n = 18). We additionally analyzed a homozygous NSF2 loss of function mutation, dNSF2⁵⁵, in this manner. We found that the total branch length in these mutants was 37.6 ± 3.5 μm (n = 9) in length. Each of the differences between each of the mutants and the control lines was significantly different (ANOVA, P < 0.05). These data therefore show that NMJ overgrowth is present early in development and that the NSF2^E/Q transgene phenocopies a NSF2 loss of function mutation.

Synaptic ultrastructure of NSF2^E/Q larvae

Stewart et al. (2002) previously reported the effects of expressing dominant-negative NSF2 on the morphology and physiology of the third instar NMJ. Among the findings, in addition to the hypersprouting phenotype noted above, was the observation that there is a substantial impairment of synaptic physiology including a reduction in the number of releasable synaptic vesicles. While they provided evidence that there were abundant small clear vesicles at the mutant nerve terminal, a careful examination of synaptic ultrastructure was not preformed at that time. Here we present a more detailed analysis of the nerve terminal ultrastructure derived from serial section electron microscopy.

To accurately quantitate the nerve terminal ultrastructure we collected series of contiguous ultrathin sections in order to reconstruct those areas of the nerve terminal. Table 1 summarizes the nerve terminal, synaptic, and active zones measurements we made. Two motor nerve terminals, called Ib and Is, innervate abdominal muscles 6 and 7 and like previously published reports (Atwood et al.,1993; Stewart et al.,1996) we identified these terminals on the basis of mitochondrial abundance, the relative abundance of dense core vesicles, and the extent of the underlying muscle sarcoplasmic reticulum and collected separate data for each of them. Figure 3 shows sample electron micrographs of the mutant and control NMJs in which two axon profiles are visible in each image. Also labeled on these images are synaptic contacts, synaptic vesicles, and, in the control sample, an active zone T-bar.

Qualitatively, the mutant nerve terminals are characterized by thinner processes and long stretches of nerve terminal devoid of synaptic contacts. We did not take serial section data from such regions, but rather from areas with recognizable synaptic contacts. The data in our table thus likely represent a best-case report of the ultrastructural features in these neurons.

Upon examining and reconstructing the serial section images, we first identified and measured the synaptic contacts and found that there were fewer synapses for a defined nerve terminal length (Table 1). This was also evident from a third serial section that we did not quantitate. The mean size of the synapses, however, was not different, with values near 0.4–0.5 μm² for each of the nerve terminal types in each genotype. Thus, one ultrastructural correlation of the NSF2^E/Q phenotype is a spatial redistribution of synapses along the length of the nerve terminal without an effect on individual synapse size.

Synaptic vesicle release is thought to occur at synaptic active zones. At the ultrastructural level active zones in Drosophila are thought to correspond to densely staining T-bar structures associated with synaptic contacts. We therefore examined the occurrence of active zones in NSF2^E/Q and control samples and found substantial differences between the genotypes. By counting the number of active zones on each synapse we found that, on average, there are about half as many active zones per synapse in the NSF2^E/Q synapses compared to the control (Table 1). This is a statistically significant difference (t-test; P < 0.05). Since the individual synapses size is not different between the genotypes, there is a resulting reduction in the number of active zones for each μm² synaptic contact. For example, we observed about four active zones for each 1 μm² of synaptic contact in the controls but only two active zones per μm² in the NSF2^E/Q synapses.

In many systems there is a correlation between the active zone density and synaptic strength; we therefore examined the frequency distribution of the number of active zones on each synapse. For the controls, many synapses had at least one active zone and a substantial proportion of synapses had more than one active zone; only 10% of Ib synapses and 20% of Is synapses had no active zones in the control. Examples of control boutons showing multiple T-bars on each synapse are shown in Figure 4. The images represent five adjoining sections through the nerve terminal with several T-bars appearing through the sequence. In our estimation, the T-bar marked with the arrowhead appears in the first section, then there are two intervening sections, and then a second T-bar appears marked with a double asterisk. In contrast, the mutant synapses had much higher proportions with 0 or 1 active zone, a small fraction had 2 active zones, and no synapses had >2 active zones. Where we did see T-bars in the mutants they appeared normal; a detail of one is shown in Figure 5. Altogether, these data show that despite the synapses in the two genotypes being of similar size, the number active zones per synapse is greatly reduced in the mutant.

Lastly, we examined the number and distribution of synaptic vesicles near active zones in the nerve terminals from randomly selected sections. In both of the genotypes there were accumulations of synaptic vesicles near synaptic contacts. We examined the number of vesicles within 50 nm of the plasma membrane, the number associated with the active zone T-bar, and the number within a 150-nm arc about the active zone T-bar. None of these measures showed significant differences between the genotypes except for the number of vesicles within the 150-nm arc of the type Ib boutons.

To address whether there is a reduction in the number of synaptic vesicles in the mutant larvae, we counted the total number of vesicles in cross-sections of the nerve terminals. We found on average that there were 225 ± 20 (n = 13 cross sections) vesicles in the mutant and 164 ± 25 (n = 14 cross sections) vesicles in the controls. This difference was not significant. Likewise, when we analyzed a different set of cross sections and normalized for the cross-sectional area of the profile, there were no significant differences. The control showed 201 ± 38 vesicles/μm², while the NSF2^E/Q showed 150 ± 30 vesicles/μm² (n = 7 profiles each). Thus, there were no detectable differences in the abundance of synaptic vesicles within the nerve terminals that we examined. Therefore, the major ultrastructural effects that we detected in the NSF2^E/Q larvae were a reduction in the number of synapses along the nerve terminal and reduction in the number of active zones found on each synapse.

Early developmental phenotype

In Drosophila, neuromuscular synaptogenesis begins about 13–15 hours after egg laying at 25°C (Broadie and Bate,1993) and the embryos we examined here were about 17–19 hours old. Our results undeniably show that the NSF2^E/Q phenotype is present very early in the life of the synapse, and indeed a similar phenotype is present in the NSF2⁵⁵ loss of function allele. The latter point is important since it supports the fact that NSF2^E/Q represents a bona fide NSF2 phenotype. A previous investigation of NSF2 loss of function alleles did not examine embryonic NMJ morphology, so this is the first report of such a phenotype (Golby et al.,2001). It is also worthwhile noting that downregulation of NSF in mammalian hippocampal neurons leads to neurite outgrowth (Yu et al.,2002), signifying that the overgrowth phenotype is not specific to Drosophila but is likely to be a common feature of NSF activity. Interestingly, overexpression of NSF's binding partner α-SNAP in the Drosophila wildtype background produces a phenotype opposite to the NSF2^E/Q phenotype: namely, reduced nerve terminal branches (Babcock et al.,2004). It remains to be determined if α-SNAP loss of function mutants display an NMJ morphological phenotype, but the overexpression result indicates both NSF and α-SNAP may play a role. Additionally, we recently demonstrated (Laviolette et al.,2005) that expression of a UAS-Snap transgene can restore the NSF2^E/Q neuromuscular overgrowth phenotype. Altogether, these data clearly establish a role for NSF/α-SNAP in synaptic development.

Previous examination of NSF2^E/Q showed a dramatic overgrowth of the NMJ in the third larval instar in addition to impaired synaptic physiology (Stewart et al.,2002). The overgrowth phenotype was a surprising one and several potential explanations for it remain open. One possible explanation is that the nerve terminal expands in response to the functional impairment. We think that explanation is unlikely because hypomorphic alleles of the SNARE proteins n-synaptobrevin and syntaxin, which reduce synaptic transmission to levels seen in NSF2^E/Q, do not have a comparable morphological phenotype (Stewart et al.,2000). Likewise, severe alleles of synaptotagmin substantially reduce synaptic output without inducing nerve terminal overgrowth (DiAntonio and Schwarz,1994). Therefore, a reduction in synaptic strength by itself does not appear to be causative factor controlling nerve terminal expansion.

Our data clearly show that nerve terminal overgrowth occurs very early in development of the NSF2^E/Q nerve terminal. It is difficult to rule out the possibility of feedback from the muscle to the nerve as a factor in this phenotype because we simply do not know the period of time such a mechanism would require. However, we observe nerve terminal expansion within a few hours of synaptogenesis in NSF2^E/Q embryos and we think it is unlikely that such retrograde signaling and remodeling could occur in the short period of time between synaptogenesis and when we examine the embryonic NMJ. We therefore favor the idea that the overgrowth phenotype is primarily an autonomous developmental phenotype and that it is not secondary to changes in synaptic physiology.

Nerve terminal ultrastructure

Our results on the changes in nerve terminal ultrastructure help us to more fully understand the effects of NSF2^E/Q expression and delineate the underlying causes of the physiological and developmental phenotypes.

One of the major differences observed in the mutant was a reduction by one-half of the number of synaptic contacts per μm of nerve terminal length. However, considering that the total length of all the nerve terminal branches is about doubled in the mutant strain (Stewart et al.,2002), there appears to be conservation of the total number of synaptic contacts throughout the entire NMJ in the control and mutant strains. Therefore, rather than the absolute number of synaptic contacts being the key factor, the spatial distribution of the synapses may be more important to the NSF2^E/Q physiological phenotype.

The second major difference was in the number of active zones found on the individual synapses. In this case there appears not to be a conservation in the total number across the genotypes, since the number of active zones per μm of terminal length in the mutant is about one-fourth that found in the control, whereas the total length is about doubled. Thus, to a first approximation there would be one-half the total number of active zones in the mutant larval NMJ. Since the active zones are believed to be a site of vesicle release, the decrease in the number of active zones would strongly indicate that a major factor underlying reduced synaptic strength in the NSF2^E/Q larvae is a reduction in release sites.

It is also important to note that not only is the number of active zones reduced, but their distribution upon synapses is also different. Many of the control synapses have more than one active zone per contact but these are rare in the NSF2^E/Q mutant, with many synapses having only one active zone and a sizable fraction having none. Our previous physiological study showed a nearly 5-fold reduction in synaptic output at external calcium concentrations of 1.0 and 1.5 mM, values in the physiological range (Stewart et al.,2002). While the total number of active zones is reduced, they are reduced by one-half and this may not fully account for the physiological impairment. However, if we examine the synapses in which there are two or more active zones we find that there is a 5-fold reduction in the Is terminal and a 4-fold reduction in the Ib terminal. Therefore, a strong correlation can be made between impaired synaptic output and the reduced number of synapses with two or more active zones. This observation is in accord with previous studies on crustacean synapses that showed a strong correlation between the number of active zones on each synapse and synaptic strength (Cooper et al.,1995,1996). If presynaptic calcium channels are associated with the active zones, then this correlation might arise from the intermixing of calcium from multiple sources to effectively increase the intracellular calcium concentration (Llinas et al.,1992; Cooper et al.,1996). The present results follow the same pattern in that the nerve terminal with fewer active zones per synapse is weaker than the nerve terminal with more active zones per synapse.

Finally, our previous study (Stewart et al.,2002) showed that in neurons expressing NSF2^E/Q there are fewer synaptic vesicles in the releasable pool. While we showed that there were abundant vesicles in the nerve terminal, here we further demonstrate that there is no significant reduction of vesicle number in the mutant, and therefore we think that a major change in vesicle population is an unlikely explanation for the 80% reduction in the size of the releasable pool of vesicles. It may appear that our sampling biases us towards this conclusion since we have taken these data from synaptic areas, while a confounding feature of the NSF2^E/Q ultrastructure is fewer synapses per length of nerve terminal. However, since vesicle clusters are associated with synaptic areas and there appears to be an overall conservation of synapse number between the mutant and controls, we likewise think that there is an overall conservation of vesicle number. Therefore, the reduced size of the releasable vesicle pool does not appear to be simply related to an overall reduction in vesicle number. Rather, the inability of the vesicles to enter the releasable pool may be due to an altered biochemical state of the vesicles. We previously found no change in the abundance of SNARE complexes in the NSF2^E/Q adult or larval brain (Stewart et al.,2002), suggesting that either another of NSF's biochemical partners is required for this task or, alternatively, that the vesicles have reduced ability to move within the nerve terminal boutons

In developmental contexts, NSF-dependent trafficking has been studied at the developing Drosophila wing margin (Stewart et al.,2001), but NSF has not been previously assigned a role in synaptic development. In some cases SNARE proteins have been associated with developmental function and, by association, NSF may be implicated (Hepp and Langley,2001; Kimura et al.,2003). However, the overall role of SNARE-mediated transport in the developing growth cone and in synapse formation is not clear; in some systems SNARE-dependent vesicle transport seems to be an important component of growth cone function but in others it is not so. In particular, SNAP-25 (Osen-Sand et al.,1993) and a tetanus toxin-insensitive form of VAMP (Martinez-Arca et al.,2000) have been shown to contribute to neurite extension in experiments using cortical and amacrine neural cell cultures and PC12 cell assays. However, genetic deletion of syntaxin, n-synaptobrevin, or SNAP-25 does not apparently affect in vivo neural synapse development in Drosophila (Schulze et al.,1995; Deitcher et al.,1998; Vilinsky et al.,2002). Identification of developmental defects associated with NSF may point to new roles for this molecule and help to resolve the role of SNAREs in synapse formation.

How could modulation of NSF lead to the results we found? Recent genetic studies of NSF2^E/Q from our laboratory (Laviolette et al.,2005) provide a framework for understanding the novel phenotypes reported here. In a screen to identify genes that interact with the NSF2^E/Q overgrowth phenotype we found that upregulation of several groups of genes can suppress the NSF2^E/Q phenotype (Laviolette et al.,2005). Importantly, upregulating components of the actin cytoskeleton potently suppress the phenotype. Three lines of evidence lead us to speculate that disruption of actin biochemistry is the underlying cause of the NSF2^E/Q phenotype. First, there is likely a vesicular trafficking role for actin within the presynaptic nerve terminal, although the precise function remains controversial (Doussau and Augustine,2000). While there is evidence that actin is an important component of the endocytic pathway (Shupliakov et al.,2002), its role in the exocytic pathway is unclear. Some studies report a restraining role for actin (Morales et al.,2000), some reporting a supportive role (Sankaranarayanan et al.,2003), and others support a positive function (Cole et al.,2000; Zhang and Benson,2001; Sakaba and Neher,2003). In any case, since some of the vesicles at the NSF2^E/Q nerve terminals seem not to be able to be released, a defect in intra-bouton mobility is possible. Second, actin is well known to be important for growth cone movement, and regulation of actin is essential to synaptogenesis (Zhang and Benson,2001). Misregulation of actin might explain the early nerve overgrowth phenotype we see. Third, we have previously shown that there is a considerable fraction of Drosophila NSF1 normally found in noncytosolic cell fractions (Mohtashami et al.,2001) and (Phillips et al.,2001) have demonstrated that actin and NSF are components of the presynaptic particle isolated from rat synaptosomes. To the extent that the mammalian presynaptic particle is equivalent to the Drosophila T-bar active zone, it is reasonable to think that disruption of NSF activity may lead to perturbation of presynaptic particle integrity and the alterations in active zone number we report here. Thus, the disparate phenotypes we observe with respect to physiology, morphology, and ultrastructure in the Drosophila NSF2^E/Q allele may have a common root in perturbation of the actin cytoskeleton. Our future studies will be aimed at testing this hypothesis. Interestingly, it was recently shown that mutation of the Drosophila gene nervous wreck (nwk) causes phenotypes remarkably similar to the NSF2^E/Q phenotype (Coyle et al.,2004). nwk is an SH3 adaptor protein that interacts with WASP, a well-known regulator of actin dynamics.

In summary, here we present evidence to show that the NSF2^E/Q mutant phenotype is present early in neuromuscular development, that it phenocopies an NSF2 loss of function allele, and is associated with substantial ultrastructural rearrangements. These data lead us to the conclusion that the defects associated with NSF2^E/Q are developmental. Future studies that investigate the relationship between the developmental and physiological defects of NSF2^E/Q, and whether they are separable, will prove invaluable to our understanding of the relationship between synapse structure and function.