A Drosophila Model of ALS: Human ALS-Associated Mutation in VAP33A Suggests a Dominant Negative Mechanism

ALS8 is caused by a dominant mutation in an evolutionarily conserved protein, VAPB (vesicle-associated membrane protein (VAMP)-associated membrane protein B)/ALS8). We have established a fly model of ALS8 using the corresponding mutation in Drosophila VAPB (dVAP33A) and examined the effects of this mutation on VAP function using genetic and morphological analyses. By simultaneously assessing the effects of VAPwt and VAPP58S on synaptic morphology and structure, we demonstrate that the phenotypes produced by neuronal expression of VAPP58S resemble VAP loss of function mutants and are opposite those of VAP overexpression, suggesting that VAPP58S may function as a dominant negative. This is brought about by aggregation of VAPP58S and recruitment of wild type VAP into these aggregates. Importantly, we also demonstrate that the ALS8 mutation in dVAP33A interferes with BMP signaling pathways at the neuromuscular junction, identifying a new mechanism underlying pathogenesis of ALS8. Furthermore, we show that mutant dVAP33A can serve as a powerful tool to identify genetic modifiers of VAPB. This new fly model of ALS, with its robust pathological phenotypes, should for the first time allow the power of unbiased screens in Drosophila to be applied to study of motor neuron diseases.


Introduction
Amyotrophic lateral scelerosis (ALS) is a progressive neurodegenerative disease characterized by loss of motor neurons in the spinal cord (''lower'' motor neurons), brainstem, and cortex (''upper'' motor neurons). The combination of muscular atrophy and fasciculations secondary to lower motor neuron loss, as well as spasticity due to upper motor neuron death, gives rise to the characteristic clinical picture of the disorder [1]. Recently, Zatz and coworkers identified a new locus for ALS/motoneuron disease at 20q13.3 (ALS8) in a large Brazilian family [2]. ALS8 is an autosomal dominant slowly progressive disorder characterized by fasciculation, cramps, and postural tremor. Autonomic abnormalities are also seen [3]. ALS8 is caused by a missense mutation in the vesicle-associated membrane protein (VAMP)/ synaptobrevinassociated membrane protein B gene (VAPB) [2].
VAPB is a type II integral membrane protein comprised of an amino terminal major sperm protein (MSP) domain that bears 22% sequence identity to the major sperm protein, a central coiled-coil domain, and a membrane anchored carboxy terminal domain [4,5]. This protein is phylogenetically well conserved; homologs of VAPB are found across species from yeast to mammals [2,6,7]. Several different functions of VAP are known. A conserved function involves recruitment of FFAT motif containing proteins to the ER [6,8]. In addition, VAP has a proposed role in vesicle trafficking [9,10]. In yeast, Scs2, the yeast homolog of VAP, plays a role in the transcriptional regulation of the INO1 gene, which is important for the synthesis of inositol [11]. In Drosophila, dVAP33A, the fly homolog of VAPB (hereafter referred to as VAP), binds to the microtubule network and regulates bouton size at the neuromuscular junction [12].
The ALS8-associated missense mutation occurs in the conserved MSP domain of VAPB. The mutation appears to alter the protein's subcellular distribution in vitro, with the mutant failing to colocalize with ER or Golgi markers and forming intracytoplasmic aggregates [2]. Subsequent biochemical studies have demonstrated a role for VAPB in the unfolded protein response (UPR); the ALS8 mutation impairs the function of the wild type protein in the UPR [13]. More recently, Hoogenraad and colleagues have demonstrated VAPA and VAPB are reduced in ALS patients and in mutant SOD1 transgenic mice, suggesting a wider role for VAPB in sporadic and familial forms of ALS [14].
Here, we have used Drosophila to study the effect of the ALS8associated mutation using fly VAP. We demonstrate that, unlike the wild type protein, misexpression of the ALS-associated VAP mutation has phenotypic effects reminiscent of loss of function mutations in the endogenous gene. The mutant protein aggregates, is ubiquitinated, recruits wild type protein into aggregates, and interferes with its function, consistent with a dominant negative mechanism. Importantly, we find that expression of the mutant protein interferes with BMP signalling at the neuromuscular junction, which not only suggests a possible regulation of this pathway by VAP, but also could have implications in understanding disease pathogenesis. This model thus provides a novel platform for genetic dissection of motor neuron death that may provide powerful insights into both sporadic and familial ALS.

Expression of VAP P58S Impairs VAP Activity In Vivo
In order to understand how the ALS8-associated mutation affects VAP function and other cellular processes, we used a gain of function approach. We generated transgenic flies carrying the wild type (VAP wt ) and mutant (VAP P58S ) forms of VAP using the GAL4-responsive UAS expression construct [15]. As a first step, we expressed both constructs in a variety of tissues using different GAL4 driver lines and examined the effect of each on viability of the organism. These included neuronal (elav), glial (repo), muscle/ mesoderm (24B), and ubiquitous drivers (tubulin and actin; Table 1). Overexpression of VAP wt in muscle using 24B-GAL4 [15] was lethal during early larval life (first or second instar). When expression was carried out using a more restricted GAL4 line, G14-GAL4 [16], the animals were able to survive through pupal development but were unable to eclose from their pupal cases and died as pharate adults. In contrast to wild type, overexpression of VAP P58S in muscle was not lethal; viable adult flies were obtained at 25uC (Table 1). However, at 30uC, expression of mutant VAP in muscle was lethal; the animals died during the pupal stage (Table 1). Surprisingly, neuronal expression of neither VAP wt nor VAP P58S had any effect on viability at either temperature.
To ensure that their differential effects were not due to differences in the expression level of VAP wt and VAP P58S transgenes, we performed an immunoblot analysis. Both transgenes were found to have comparable levels of protein expression ( Figure S1). These data suggest that the ALS8 mutation impairs the biological activity of VAP such that higher expression of the mutant protein is required to bring about effects similar to those elicited by the wild type protein.
It has been previously reported that VAP regulates bouton size at the neuromuscular junction in a dose-dependent manner [12]. Increased expression of VAP leads to smaller and more numerous boutons whereas, reduced VAP expression results in fewer but larger boutons. We therefore examined the effect of neuronal overexpression of VAP P58S on bouton morphology at the third instar larval neuromuscular junction. Consistent with the findings of Bellen and coworkers [12], we found that expression of VAP wt using the pan-neuronal elav-GAL4 or the more restricted OK6-GAL4 driver [16] produced smaller boutons ( Figure 1B and E). In contrast, animals overexpressing VAP P58S had larger boutons ( Figure 1C and F).
We measured and compared bouton size of control and animals overexpressing VAP wt and VAP P58S by calculating the average diameter of boutons on muscle 4. Only the largest diameter of each bouton was measured in each case. The average bouton diameter of at least 15 synapses was calculated for each of the genotypes. Panneuronal expression of VAP P58S caused a small but significant increase in bouton size as compared to control animals and wild type overexpressors ( Figure 1G). Similar results were obtained with the OK6-GAL4 ( Figure 1G). As described previously [12], we also observed an increase in bouton number with neuronal expression of wild type VAP. However, overexpression of mutant VAP had no effect on bouton number as compared to control (data not shown). Interestingly, overexpression of VAP in neurons altered the staining pattern of Discs large (DLG), which labels the subsynaptic reticulum (SSR) on the post synaptic surface. Unlike wild type, DLG staining in these animals appeared broad and diffuse (data not shown), indicating that neuronal expression of VAP can influence the organization of post-synaptic elements, consistent with a non cell autonomous action of VAP.
Next, we examined the effects of expression of VAP P58S on bouton size in Drosophila mutants that lack full-length endogenous VAP protein. We expressed VAP wt and VAP P58S in the CNS of the VAP D166 mutant. This mutant is a partial loss-of-function allele of VAP that lacks approximately 100 residues at the amino terminal region of the protein. Hemizygous VAP D166 males survive until the third instar larval stage, but most of them die as pupae and very few ''escaper'' males are observed on eclosion [12]. Similar to VAP null mutants, VAP D166 larvae have enlarged boutons at the larval neuromuscular junction, and this phenotype was rescued by neuronal expression of wild type VAP ( Fig 1I). However, we were unable to rescue this phenotype with VAP P58S (Fig 1J). We measured the average bouton size in all three genotypes: the mean bouton diameter in VAP D166 males was 6.1860.15 mm ( Figure H and K). In the presence of neuronally expressed wild type VAP, however, these boutons were significantly smaller, having an average diameter of 4.8960.13 mm ( Figure 1I and K), similar to the size observed in control animals (Fig 1G, blue bar). In contrast, neuronal expression of VAP P58S in VAP D166 mutants did not cause any significant change in bouton size (5.9060.24 mm; Figure 1J and K). Thus, expression of wild type but not mutant VAP was able to rescue the enlarged bouton phenotype in dVAP D166 mutants.

VAP P58S May Function as a Dominant Negative
The inability of VAP P58S to rescue the bouton phenotype associated with VAP D166 shows that the ALS8-associated missense mutation impairs VAP function. This, together with the observation that neuronal expression of VAP P58S increases bouton size similar to VAP loss of function, suggests that the mutation might function as a dominant negative. To explore this possibility, we compared the ability of VAP wt and VAP P58S to rescue the lethality associated with VAP loss of function. Since neuronal expression of VAP is sufficient to rescue lethality [12], we used elav-GAL4 to express VAP wt and VAP P58S transgenes in genetic backgrounds containing two independent VAP mutations: VAP D20 (a null allele) and VAP D166 . Males hemizygous for these mutations do not survive to eclosion as adults. In both cases, the percentage of mutant VAP hemizygous males that eclosed as adults was higher with VAP wt as compared to VAP P58S (Figure 2A). The rescue by VAP wt was approximately 63% and 33% greater as compared to VAP P58S in VAP D20 and VAP D166 backgrounds, respectively. Of note, when VAP wt and VAP P58S were coexpressed, the percentage of rescued males that eclosed was very similar to that observed with VAP P58S alone, suggesting that the ALS8-associated mutation may impair the function of the wild type protein, consistent with a dominant negative mode of action ( Figure 2A). We were intrigued by the apparent inhibition of normal neuronal VAP function by the ALS8 mutation and sought to examine the consequences of this putative dominant negative interaction in a different neuronal context. Overexpression of neurotoxic genes including mutant ataxin-1 in sensory precursors using the scabrous-GAL4 driver leads to loss of bristles in the adult [17]. We found that overexpression of VAP wt using scabrous-GAL4 resulted in failure of thoracic macrochaetae to develop ( Figure 2C). This bristle loss was not observed in sca-GAL4/UAS-VAP P58S animals ( Figure 2D). However, coexpression of VAP P58S with VAP wt suppressed the bristle phenotype associated with the latter (Figure 2E), again supporting a dominant negative effect of VAP P58S .

VAP P58S Appears to Aggregate and Recruit Normal VAP into Aggregates In Vivo
Abnormal aggregation of cellular proteins is a common feature of many neurodegenerative diseases, including Alzheimer's, Huntington's, and Parkinson's diseases, prion disorders, and ALS [18][19][20][21][22]. Biochemical studies have demonstrated that the ALS8 mutation results in insolubility and aggregate formation of VAPB in non-ER fractions [2,13,14]. To test if VAP P58S formed aggregates similar to the human mutant VAPB protein, we compared the cellular distribution of wild type and ALS8 mutant VAP in the muscle of third instar larvae because of the ease with which this tissue can be visualized. Expression of VAP wt resulted in a uniform increase in the intensity of VAP staining throughout the muscle ( Figure 3B); on the other hand, expression of VAP P58S resulted in a more punctate appearance in muscle ( Figure 3C). To  . VAP P58S appears to function as a dominant negative by inhibiting the activity of wild type VAP. (A) Histogram comparing the rescue from lethality of VAP D20 (null mutant) and VAP D166 (partial deletion), using neuronal expression of VAP wt , VAP P58S , or a combination of both. The ordinate represents the percentage of mutant hemizygous males obtained (i.e., VAP mutant hemizygous males/VAP mutant hemizygous males+balancer chromosome males). Each bar represents the mean obtained from at least 2-3 independent experiments. Both VAP P58S and VAP wt are able to rescue the larval lethality observed in VAP D20 and VAP D166 mutants. However, the percentage of mutant males obtained with VAP wt is higher (72.967.3 for VAP D20 and 61.561.2 for VAP D166 ) as compared with VAP P58S (45.063.1 for VAP D20 and 46.660.06 for VAP D166 ). For each mutant, the value for wild type rescue differed significantly from mutant (p,0.05 for. D20 and 0.01 for D166; ANOVA with Sidak-Holm test). Rescue of lethality is also obtained by co-expressing VAP P58S and VAP wt . In this case, the percentage of mutant males obtained was similar to that observed with VAP P58S alone (45.763.3 for VAP D20 and 40.362.7 for VAP D166 ). However, for each mutant, rescue using the recombined VAP P58S , VAP wt chromosome differed significantly from that obtained using VAP wt alone (p,0.05 for D20 and 0.01 for D166, ANOVA with Sidak-Holm test). n.s., not significant. (B-E) VAP P58S suppresses the thoracic bristle phenotype produced by expression of VAP wt using Sca-GAL4. (B) Shown in this panel is the thorax of a control (sca-GAL4/+) adult fly with a stereotyped pattern and number of dorsal macrochaetae (arrows). (C) In Sca-GAL4/UAS-VAP wt animals, the dorsal macrochaetae either fail to develop or are dramatically reduced in number. (D) Expression of VAP P58S has no effect on the development or number of bristles. (E) However, co-expression of VAP P58S with VAP wt results in a suppression of the bristle loss phenotype caused due to overexpression of VAP wt . doi:10.1371/journal.pone.0002334.g002 ensure that this apparent aggregate formation was independent of cellular context, we also examined the cellular distribution of VAP wt and VAP P58S in neurons of third instar larvae. As in muscle, neurons overexpressing VAP P58S showed punctate cytoplasmic staining with reduced staining at the cell membrane, whereas more evenly distributed cytoplasmic staining was observed in neurons overexpressing VAP wt ( Figure S2).
Cellular deposits containing ubiquitin are characteristic of a number of neurodegenerative disorders, including ALS [23,24]. We therefore examined ubiquitin staining in animals expressing VAP wt and VAP P58S . In VAP wt overexpressors, ubiquitin staining was similar to controls (data not shown). On the other hand, in animals overexpressing VAP P58S , punctate ubiquitin immunoreactivity was observed throughout the muscle. These puncta colocalized with VAP P58S aggregates, suggesting that mutant VAP aggregates are ubiquitinated to be targeted for proteosomal degradation ( Figure 3D).
One potential pathogenic mechanism underlying the ALS8 mutation would be recruitment of normal endogenous VAP into aggregates, thereby depleting the cell of wild type protein. To test this possibility, we generated transgenic flies carrying amino terminal hemagglutinin (HA) and Myc-tagged versions of VAP P58S and VAP wt , respectively. These constructs were tested individually by expressing transgenes in muscle and staining with the appropriate epitope tag antibodies. Similar to the untagged form, Myc-VAP wt staining was localized to the muscle membrane, whereas the HA-VAP 58S immunoreactivity appeared punctate throughout the muscle, suggesting that the tagged versions behave similar to the untagged forms ( Figure 3E and F). However, when the two transgenes were expressed together, the Myc-VAP wt staining appeared more punctate and largely colocalized with the HA immunoreactive puncta representing the ALS8 mutant VAP ( Figure 3G). These data suggest that VAP P58S may recruit the wild type protein into aggregates, consistent with a dominant negative effect whereby the normal protein thus recruited is functionally inhibited.

VAP P58S Impairs Microtubule Organization
Impaired microtubule organization and microtubule dependent transport are believed to be a common underlying feature in many neurodegenerative diseases [25]. Microtubule-based transport plays a crucial role in survival of motor neurons in particular. Aberrant axonal transport contributes toward pathogenesis in sporadic ALS and in mice expressing mutant SOD1 [26][27][28]. Investigators previously have shown that VAP associates with microtubules [10,12]. In the absence of VAP, microtubule organization at the neuromuscular junction is abnormal in Drosophila [12]. We were therefore interested in examining the effect of VAP P58S on microtubule organization. We examined the neuromuscular junction in larvae expressing wild type and mutant VAP by immunohistochemistry using the monoclonal antibody 22C10, which recognizes Futsch, a presynaptic microtubuleassociated protein related to vertebrate MAP1B [29]. The pattern of Futsch staining at the neuromuscular junction has been extensively characterized [12,[30][31][32]. In wild type larvae, Futsch staining ( Figure 4A-D, green) in most boutons appears thread-like, running along the axis of each synaptic branch and continuous with the axonal cytoskeleton ( Figure 4A, arrow; also see inset). At distal boutons, Futsch staining appears punctate or diffuse ( Figure 4A, arrowhead; also see inset ). Occasionally, Futsch staining is found to fill the entire presynaptic space. These two forms of staining, i..e, ''diffuse'' and ''filled'', have been described as two variant forms of disorganized microtubules found in mutants affecting synaptic architecture [30,32]. In dVAP null mutants, approximately 70% of the boutons have been reported to show a filled ( Figure 4B, asterisk) or diffuse pattern of Futsch immunoreactivity, consistent with reported microtubule disorganization [12]. To determine if expression of VAP P58S also leads to disorganization of microtubules, we compared the pattern of Futsch staining at the neuromuscular junction of larvae expressing wild type and ALS8 mutant VAP in neurons. Similar to controls, in animals overexpressing VAP wt , the microtubules were well organized; 22C10 staining in most boutons was continuous, with few boutons displaying a diffuse or filled pattern of Futsch staining as compared to controls ( Figure 4C and E; 33% in controls and 37% in VAP wt overexpressors). In contrast, expression of VAP P58S increased microtubule disorganization: 53% of boutons displayed either the filled or diffuse pattern of Futsch staining ( Figure 4D and E). This finding was comparable to that observed in VAP D166 mutants, in which approximately 63% of the boutons showed disorganized microtubules ( Figure 4E). Thus, microtubule organization is impaired by expression of ALS associated mutant VAP.
Overexpression of VAP wt But Not VAP P58S Leads to a Decrease in the Number of Active Zones at the Neuromuscular Junction Having established a putative dominant negative role for VAP P58S , we were interested in determining whether its expression in neurons altered the expression or localization of other synaptic proteins at the neuromuscular junction. Therefore, larvae expressing wild type and ALS8 mutant VAP were stained with antibodies against a variety of synaptic proteins such as synaptotagmin, syntaxin, and CSP. No obvious defects in the expression or localization of any of these markers were apparent (data not shown). However, some differences were observed when these animals were stained with nc82, a monoclonal antibody which recognizes Bruchpilot (brp), a crucial component found in active zones at the synapse [33,34]. Despite an increase in the total number of boutons as a result of overexpression of VAP wt , there were fewer nc82 immunoreactive puncta per synapse in these animals ( Figure 5B and D). Quantification of the total number of nc82 puncta per synapse at muscle 4 for each of the genotypes revealed a significant decrease in wild type overexpressors as compared to mutant and control larvae. No significant differences in the number of nc82 immunoreactive puncta were observed between controls and larvae expressing VAP P58S (Figure 5C and D). To determine if the decrease in the total number of nc82 puncta was an indirect consequence of reduced bouton size caused by overexpression of VAP wt , we quantitated the number of the nc82 positive puncta relative to bouton cross sectional area in all three genotypes. In this case however, we did not observe any significant differences between the three genotypes (Table S1). This analysis suggests that the decrease in the total number of active zones per synapse might indeed be an effect of reduced bouton size rather than loss of active zones per se.

Overexpression of VAP wt and VAP P58S Causes Ultrastructural Abnormalities at the Synapse
Studies conducted in mutant SOD1 transgenic mouse models indicate that loss of synaptic connections, leading to nerve retraction, are early events in development of pathological phenotypes [35][36][37]. To evaluate whether expression of VAP P58S causes structural abnormalities that might lead to synaptic instability, we conducted an ultrastructural analysis of the neuromuscular junction in VAP wt and VAP P58S . At least five animals of each genotype were used for the analysis. The ultrastructure of a wild type bouton is shown in Figure 6A; notable structures include the subsynaptic reticulum (SSR), as well as the ''T'' bars associated with active zones (asterisk).
Overexpression of the wild type protein, as previously described (Figure 1), resulted in smaller boutons and, as previously reported by Pennetta and coworkers, these boutons also contained significantly fewer neurotransmitter vesicles [38] (Figure 6B). We observed a 54% decrease in the number of vesicles per bouton cross sectional area in these animals (p,0.001). In contrast, boutons from synapses overexpressing VAP P58S contained abundant neurotransmitter vesicles, similar to controls ( Figure 6C and D, see Table S2). A subset of these vesicles were abnormally large ( Figure 6C and D, dark arrow) and sometimes filled with electron dense material ( Figure 6C, dark arrowhead). We also occasionally observed abnormal electron dense structures in the SSR of these animals ( Figure 6C and D, white arrowhead). We frequently observed electron dense structures attached to neurotransmitter vesicles floating within the cytoplasmic space of boutons and not associated with the presynaptic membrane ( Figure 6D, white arrow). This phenotype, referred to as ''floating T-bars'' [39], was more frequently observed in animals expressing VAP P58S (44% boutons in elav-GAL4/UAS-VAP P58S versus 12.5 % in elav-GAL4/+ animals). In addition, there was a significant increase in the number of these floating T-bar structures in animals expressing VAP P58S as compared to the controls (0.8160.23 as expressed per bouton in elav-GAL4/UAS-VAP P58S versus 0.2560.17 in elav-GAL4/+ animals; p,0.05; see Table S3). We did not observe this phenotype in animals overexpressing VAP wt .

Interaction between VAP and BMP signaling
The phenotype of floating T-bars that was sometimes observed in boutons of animals expressing VAP P58S is reminiscent of the phenotype associated with mutations in components of the bone morphogenetic protein (BMP) signaling pathway [16,[39][40][41]. Given the resemblance between this phenotype and mutants of the BMP pathway, we sought to identify potential perturbations of BMP signaling in animals expressing the ALS8 mutation. In Drosophila, BMP signaling at the neuromuscular junction takes place in an anterograde as well as a retrograde manner and loss of BMP signaling leads to defects in synaptic growth and function [16,[39][40][41]. The signaling pathway involves activation of a heteromeric complex comprised of two type I and two type II receptor molecules. Upon ligand binding, the type II receptor phosphorylates the type I receptors, which transduce the signal by phosphorylating receptorspecific transcription factors called Smads. The phosphorylated Smad (pMAD in Drosophila) complex with Smad4 (Medea in  Drosophila) translocates to the nucleus and activates transcription of target genes [42]. The level of activity of this pathway directly correlates with the level of phosphorylated Smad/pMAD; therefore, changes in levels of phosphorylated Smad or pMAD are used assess the activity of this signaling pathway [43,44]. At the neuromuscular junction, pMAD accumulates at postsynaptic densities opposite the presynaptic active zone marker, bruchpilot (nc82) [45]. In motor neurons, pMAD accumulation takes place in response to retrograde signaling by the ligand glass bottom boat (gbb), mediated by the type II receptor wishful thinking (wit) [16].
In order to assess whether VAP P58S affects BMP signaling, we expressed wild type and mutant transgenes in larval muscle using G14-GAL4 [16] and examined pMAD staining at the neuromuscular junction. As reported previously [45], in control animals, punctate pMAD staining was observed opposite presynaptic active zones immunostained with nc82 ( Figure 7A). In VAP wt transgenic animals, pMAD staining was more intense ( Figure 7B) and a few nc82 immunoreactive puncta isolated from pMAD staining were also observed. In contrast, in animals expressing VAP P58S , the intensity of pMAD puncta was decreased ( Figure 7C). We measured the intensity of pMAD puncta in all three genotypes. As compared to controls, in animals overexpressing VAP wt we observed a 40% increase in intensity of pMAD staining, whereas expression of VAP P58S resulted in a 42% decrease in intensity of pMAD staining. In animals expressing VAP wt , the average relative pixel intensity was 140.61611.11 (mean6SEM). In VAP P58S expressing animals, pixel intensity was 58.3966.60 (mean6SEM). These values differed significantly (p,0.001 by ANOVA with Student-Neuman-Keuls test) from control pixel intensities, which were normalized to 100 (see Materials and Methods). We also examined pMAD expression in nuclei of motor neurons in the larval CNS. As at the neuromuscular junction, expression of VAP P58S lead to a decrease in the intensity of pMAD staining ( Figure S3), suggesting that both anterograde and retrograde BMP signaling may be impaired.
These data indicate that increased VAP wt expression potentiates BMP signaling, suggesting an interaction between VAP and BMP signaling pathways. In order to explore this possibility further, we conducted an epistasis analysis using the bristle phenotype observed in animals overexpressing VAP wt . Overexpression of VAP wt using sca-GAL4 resulted in loss of thoracic bristles in adult flies ( Figure 8B). To test if VAP and BMP signaling pathways interact, we co-expressed a dominant negative form of the type I BMP receptor-thickvein (DN-tkv). We found that co-expression of DN-tkv suppressed the bristle loss phenotype associated with overexpression of VAP wt ( Figure 8C). We quantitated this by counting the number of bristles present within a defined region of the thorax of adult flies. In control animals, the ten dorsal macrochaetae present in this region (circle in Figure 8A) were invariant in number and position in all adult flies examined ( Figure 8D). However, in flies overexpressing VAP wt , an average of only 2-3 bristles were observed. Co-expression of DN-tkv using 2 independently derived transgenic lines resulted in significant suppression of the bristle loss phenotype: on average, 5-6 bristles were observed in these flies ( Figure 8D). Taken together, these results are consistent with the assertion that VAP P58S interferes with BMP signaling at the neuromuscular junction.

Discussion
We have used Drosophila as a model system to gain insight into the mechanism by which the ALS8 associated mutation gives rise to disease. The evolutionary conservation of VAP and the availability of a vast array of genetic tools make the fly a particularly attractive model system with which to dissect pathophysiological mechanisms underlying ALS8. We have compared and assessed the cellular consequences of expression of wild type and ALS8-associated mutant VAP on the development and function of the synapse at the neuromuscular junction. In Drosophila, the synapses at the neuromuscular junction are glutamatergic, providing some similarities to the spinal cord synapse which are affected in ALS patients. Furthermore, the presynaptic neurons are obviously the functional equivalent to the lower motor neurons also affected in ALS patients.
The relationship between mutation and pathogenesis in inherited neurodegenerative disorders is obviously of great interest as it could point towards a mechanistic understanding of the disease and indicate treatment strategies. For genetically dominant disorders, potential mechanisms include toxic gain of function, haploinsufficiency, and dominant negative effects. Pennetta and coworkers concluded that the ALS8 mutation is a toxic gain of function rather than a dominant negative [38]. This conclusion was based on the toxic properties of the mutant protein in conjunction with its retention of activity. However, our data and that of others [13,14] argue against retention of VAP activity by the ALS8 mutant, but rather suggest that that the ALS8 mutation in VAMP-associated-protein (VAP) functions through a dominant negative mechanism. Dominant negative mutations are those in which the product of the mutant gene adversely affects function of the normal gene product within the same cell, for example by forming non-functional dimers. Interestingly, members of the VAP family are able to form dimers [6,8]. The following experiments reported here support the notion that the formation of mixed dimers or aggregates between wild type and ALS8 mutant proteins is at least in part responsible for the dominant phenotype. The VAP loss of function phenotype (VAP D20 and VAP D166 ) is characterized by an increase in bouton size and increased microtubule disorganization. Both these phenotypes are observed when VAP P58S is overexpressed in neurons, indicating that expression of the mutant protein is dominant over the wild type protein. This is not simply due to an excess of mutant protein since VAP P58S can also suppress the phenotypes generated by overexpression of VAP wt (Figure 4). Finally, VAP P58S forms ubiquitinated aggregates that can also recruit and induce aggregation of wild type protein (Figure 3), thereby rendering it functionally inactive. The ability of the mutant protein to recruit and induce aggregation of the wild type protein has also been reported for mammalian VAP in vitro [13,14]. The finding that down regulation of mammalian VAP using RNAi techniques is sufficient to induce cell death in motor neurons [14] also argues that it is loss of VAP function, rather than the formation of aggregates, that is pathogenic.
VAP was originally identified as a ''VAMP-associate protein'' in Aplysia, and antibodies against VAP decrease synaptic transmission [10]. Subsequently, VAP was demonstrated to alter the size and number of boutons at the Drosophila neuromuscular junction [12]. Overexpression of VAP wt leads to a significant decrease in the total number of active zone at the synapse despite an increase in bouton number; fewer nc82 immunoreactive puncta per synapse are present ( Figure 5) even though the number of such puncta per unit cross sectional area of the bouton (mm 2 ) is not significantly different from control and animals expressing mutant VAP. Further research will be necessary in order to determine whether this difference in the total number of active zones might lead to impaired synaptic transmission.
In a recent report, by conducting morophometric analysis on electron micrographs of individual boutons, Pennetta and colleagues show that there is no change in the number of active zones in animals overexpressing VAP wt in neurons [38]. These authors also report locomotory defects, synaptic degeneration and muscle atrophy in larvae expressing VAP P58S in neurons. We did not observe any of these defects in our experiments. Differences in expression levels of the mutant protein could account for these discrepancies.
Extensive crosstalk between motor neurons and muscle is a conspicuous feature of neuromuscular junction development in Drosophila, where synapses must accommodate the increase in size of muscles during larval development by increasing bouton number and active zones per bouton. Components of the TGFb/ BMP signaling pathway are known to regulate neuronal survival, synaptic development and function in vertebrates as well as invertebrates. Recently, reduced TGF-b signaling has been associated with Alzheimers disease and the level of one of the receptors (TGFbRII) is reduced in brain tissue of Alzheimer's disease patients [46].
In Drosophila, retrograde BMP signaling is required for synaptic growth at the neuromuscular junction [39]. At the same time, anterograde BMP signaling in neurons appears to be essential to restore neurotransmission which is significantly lowered in mutants of the BMP pathway. Here, we show using epistasis analysis, the presence of a genetic interaction between VAP and BMP signaling pathways (Figure 8). We also show that overexpression of VAP wt potentiates BMP signaling whereas overexpression of VAP P58S impairs BMP signaling as evidenced by the decrease in intensity of pMAD puncta at the neuromuscular junction ( Figure 7) and nuclei of motor neurons ( Figure S3). Expression of VAP P58S also causes ultrastructural phenotypes reminiscent of mutants in the BMP pathway, e.g., floating T-bars and occasional detached postsynaptic membranes. However, it should be noted that despite the similarities, these phenotypes are by no means identical; BMP mutants show smaller synapses (i.e., fewer boutons) with no significant change in bouton size-a phenotype unlike that observed in animals with neuronal expression of VAP P58S . This suggests that the aberrations associated with ALS8 are likely to be more complex than simply loss of receptors. Recent observations that VAP is reduced in ALS patients and mutant SOD transgenic mice [14] suggest a potential role for VAP in many forms of ALS. Further genetic analysis using this model as a paradigm for ALS is likely to provide insights into pathways that contribute to motor neuron dysfunction and death in ALS.

Drosophila Genetics and Phenotypic Analysis
The VAP D166 and VAP D20 mutant lines and VAP cDNA were kindly provided by Hugo Bellen. G14-GAL4 and the second chromosome elav-GAL4 transgenic line were obtained from Kai Zinn. David Krantz provided the OK6-GAL4 line. The 24B-GAL4, tubulin-GAL4, actin-GAL4, sca-GAL4, and repo-GAL4 drivers were obtained from the Bloomington stock center. UAStkv D GSK (UAS-DNtkv) stocks were obtained from Michael O'Connor; both lines carried two insertions of the transgene [47].
Site-directed mutagenesis using the Stratagene Quick-Change II kit (Stratagene) was used to generate the P58S mutant using the wild type cDNA. This manipulation and placement of epitope tags were performed using the pGEM-7 cloning vector (Promega). Epitope tags (3X Myc or HA) were engineered at the amino terminus of wild type and P58S VAP using PCR. Identity of all constructs was verified by sequencing. Constructs were subcloned into the UAS expression vector and transgenic flies generated using standard techniques. Certain lines were generated using facilities of the Duke Model Systems Genomics Group. Multiple transgenic lines were generated and tested for each construct. At least two independently derived lines on the autosomes were tested for each construct. In no case were insertions selected on the basis of phenotypes or lack thereof. For all untagged experiments shown here, we used UAS-VAP wt-A1. 6 and UAS-VAP P58S-A2.1 , which showed similar elav-GAL4-driven expression levels as assessed using immunblots of head extracts in combination with guinea pig anti-VAP ( Figure S1). For each construct, each of the lines used showed the same patterns of lethality under control of various drivers. Neuronal overexpression with elav-GAL4 was carried out at 30uC. Experiments involving overexpression in the muscle were conducted at 25uC. Bristle phenotype and all rescue experiments were carried out at 23uC.
We generated animals with both HA and myc tags for both the ALS8 and wild type transgenes in order to exclude the possibility that aggregation was due to the epitope tag alone (data not shown). For co-expression studies, mitotic recombination was used to place both UAS-VAP wt and UAS-VAP P58S transgenes on the second chromosome. For studies of transgene expression in genetic backgrounds lacking endogenous VAP, VAP D166 /FM7Kr-GFP; elav-GAL4 females were crossed to w/Y; UAS-VAP males (either wild type, P58S, or the recombined chromosome containing both transgenes).

Immunoblotting
Immunoblot analysis of adult heads was carried out using published methods [50] in conjunction with guinea pig anti-VAP (1:10,000) and mouse anti-b-tubulin (1:1000; Accurate Chemical and Scientific).

Image Acquisition and Processing
All confocal images were obtained using a Zeiss Pascal laser scanning confocal microscope in conjunction with a 63X NA 1.4 Zeiss Apochromat or 40X NA 0.75 Axoplan Neofluar lens at room temperature using Vectashield (Vector labs) as an imaging medium and the Zeiss Pascal software. For experiments in which direct comparisons were made based on fluorescence intensity (i.e., Figure 7 and Figure S3) images for different genotypes were acquired under precisely identical conditions and in succession. Images were processed in Adobe Photoshop. Films from immunoblots were scanned with a UMAX Powerlook 1000 transmissive flatbed scanner. Graphical and statistical analyses were performed using SigmaPlot 9.0 and Systat 3.1.
For quantitation of the relative fluorescence intensity levels of pMAD, the pixel intensity of pMAD puncta for all three genotypes (control, G14-GAL4.VAP wt and G14-GAL4.VAP P58S ) was measured using Image J software. Three independent sets of experiments were conducted, and in each case all three genotypes were imaged simultaneously using identical imaging conditions. The change in intensity stated in the text represents the average change observed in these three independent experiments. For each set, 5-8 synapses of each genotype were imaged. The percentage relative fluorescence intensity of in G14-GAL4.VAP wt and G14-GAL4.VAP P58S was calculated by normalizing the control intensity to 100 percent.

Electron Microscopy
Third instar larvae were filleted and pinned first in normal saline, then glutaraldehyde (2% in 0.12 M sodium cacodylate buffer, pH 7.4) for 2 hr with rotation. Samples then were rinsed thrice in 0.12 M sodium cacodylate buffer and postfixed with 1% osmium tetroxide in the same buffer with rotation. After postfixation, samples were washed and stained with 1% aqueous uranyl acetate for 1 hr. Specimens then were washed, dehydrated, and embedded in Eponate 12 resin. Thin sections were stained with uranyl acetate and Sato's lead. Analysis was carried out using a JEOL 100CX transmission electron microscope at an operating voltage of 80 kV. Figure S1 Comparable expression of UAS-VAP wt-A1.6 and UAS-VAP P58S-A2.1 . Immunoblot using head extracts from elav-GAL4.VAP wt and elav-GAL4.VAP P58S animals. The same blot was stripped and reprobed with anti-b-tubulin. Found at: doi:10.1371/journal.pone.0002334.s001 (0.69 MB TIF) Figure S2 A comparison of the cellular staining pattern of VAP wt and VAP P58S in neurons. Shown are images of the 3rd instar larval brain from elav-GAL4/VAP wt (A) and elav-GAL4/VAP P58S animals (B). Green, anti-HRP staining; red, anti-VAP. Overexpression of VAP wt leads to robust intracellular staining (A). In contrast, expression of VAP P58S leads to formation of aggregates (B). Scale bar, 10 mm. Table S1 Analysis of the number of the nc82 immunoreactive puncta relative to bouton cross sectional area. Values shown are mean6SEM, N = 8 for each genotype; one-way Kruskal-Wallis ANOVA. There were no significant differences observed, suggesting that the apparent decrease in the total number of active zones per synapse might indeed be an effect of reduced bouton size rather than loss of active zones.  for Driver alone and VAPwt and 10 for VAPmut; one-way ANOVA with Student-Newman-Keuls comparison; each bouton analyzed was considered as an independent sample for this analysis irrespective of the animal from which it was obtained. Neuronal expression of wild type VAP significantly reduced vesicle numbers as compared to the driver alone or mutant VAP. Found at: doi:10.1371/journal.pone.0002334.s005 (0.04 MB DOC) Table S3 Quantitation of floating T bars per bouton. Values indicated are mean6SEM, N = 16 for Driver alone, 17 for VAPwt, and 27 for VAPmut; one-way ANOVA with Student-Newman-Keuls comparison; each bouton analyzed was considered as an independent sample for this analysis irrespective of the animal from which it was obtained. Neuronal expression of mutant VAP significantly increased the number of floating T bars as compared to the driver alone or mutant VAP. Found at: doi:10.1371/journal.pone.0002334.s006 (0.04 MB DOC)