MTM-6, a Phosphoinositide Phosphatase, is Required to Promote Synapse Formation in Caenorhabditis elegans

Forming the proper number of synapses is crucial for normal neuronal development. We found that loss of function of the phosphoinositide phosphatase mtm-6 results in a reduction in the number of synaptic puncta. The reduction in synapses is partially the result of MTM-6 regulation of the secretion of the Wnt ligand EGL-20 from cells in the tail and partially the result of neuronal action. MTM-6 shows relative specificity for EGL-20 over the other Wnt ligands. We suggest that the ability of MTM-6 to regulate EGL-20 secretion is a function of its expression pattern. We conclude that regulation of secretion of different Wnt ligands can use different components. Additionally, we present a novel neuronal function for MTM-6.


Introduction
Construction of functional nervous systems depends on the ability of neurons to form precise connections. The process of forming functional synapses begins with proper axon guidance and ends with the formation of synaptic contacts on specific cellular domains. As part of this process, neurons must form and maintain the appropriate number of synapses with their partners. Synapse formation and elimination is dynamic throughout the life of the neuron [1][2][3][4][5]. An excess of connections is formed early in development, and then a subset of synapses is eliminated as the neural circuits matures [1][2][3]. Abnormal synapse number is found in several forms of mental retardation, and a reduced number of synapses is strongly correlated with cognitive impairment in Alzheimer's Disease [5,6].
Much of the research on synapse number has focused on changes in the dendritic arbor and dendritic spines in the mammalian central nervous system using dissociated cultured neurons [1][2][3][5][6][7]. In hippocampal cultures, the calcium influx that occurs during stimulus triggers dephosphorylation of the transcription factor myocyte enhancement factor 2 (MEF2) [8,9]. In turn, MEF2 activity limits the number of dendritic spines that form [8]. Additionally, an RNAi screen in cultured hippocampal neurons gave a number of potential cues for factors affecting synapse number including the cadherins, the semaphorins, and the GTPase REM2 [10]. Yet, the heterogeneous cell population and the issue of separating changes in morphology from changes in number present difficulties in understanding the molecular mechanisms controlling the synapse number. Here, we investigated the molecular mechanisms that regulate synapse number in an in vivo cell system in a neuron that exhibits a stereotyped number of synapses. We demonstrate a role for the myotublarin lipid phosphatase MTM-6 in the maintenance of proper synapse number.

Results
Loss of the myotubularin phosphatase mtm-6 causes a reduction of synapse number The C. elegans motor-neuron DA9 has a stereotyped synapse localization and number. The axon of DA9 extends posteriorly from a ventral cell body and reaches the dorsal nerve cord through a commissure. Within the dorsal nerve cord, the axon extends anteriorly and forms a string of en passant synapses in a specific region along the axon that begins approximately 25 mm anterior to the dorsal commissural turn in adult animals (Fig. 1A). The DA9 synapses can be labeled by cell-specific expression of GFP fused with synaptic vesicle or active zone proteins [56]. During the search for genes required for proper development of the DA9 synapses, we found that a mutant for a myotubularin phosphatase, mtm-6(ok330), shows a reduction in synapse number ( Fig. 1C-F). Using GFP::RAB-3 as a marker for synapses, we observed that in wild-type animals there are an average of 22.4 (+/22.6) puncta within the dorsal cord. Both the number and the subcellular distribution pattern of the GFP::RAB-3 puncta closely agree with what was reported from serial electron microscopy reconstruction studies, suggesting that the RAB-3 marker represents true synapses. We found that the number of DA9 synapses is reduced to 14.4 (+/22.8) puncta in mtm-6 mutant animals (Fig. 1B and Table S1, n.60, SD, p,0.0001). The reduction in synapse number is also present when synapses are visualized with an active zone marker: quantifying GFP::UNC-10, the C.elegan's RIM homolog, we observed that mtm-6 mutants have an average of 14.4 (+/23.3) puncta in comparison to the 21.6 (+/23.4) puncta found in wild-type (Fig. 1B and Table S1, n.60, SD, p,0.0001). The mtm-6(ok330) allele is a 1,235 bp deletion that is predicted to cause a frameshift towards the end of the phosphatase that affects all four isoforms of the gene. Due to the temperature sensitive nature of some mutants used in the paper and the temperature sensitivity of the observed phenotypes (as marked by GFP::RAB-3 or GFP::UNC-10) all animals were maintained at 20˚C for at least two generations prior to quantification.
To understand how the mtm-6 mutant synapse number phenotype arises during development, we counted the number DA9 synapses during development in wild-type and mtm-6 mutants. For these experiments, we used the mig-13 promoter, which is active in DA9 during early larval stages, to drive expression of SNB-1/synaptobrevin::YFP. The DA9 neuron is born embryonically and forms synapses before hatching. After hatching, as the animals develop through four larval stages to adulthood, the number of synapses gradually increases as the body lengthens (Fig. S1). There is a small, but significant, difference in the number of SNB-1::YFP puncta at the L1 stage between wild-type and mtm-6 mutants ( Fig. S1 and Table S1, 2.35 puncta, n540, SD, p,0.001). However, there is no significant difference between wild-type and mtm-6 animals at either the L2 or L3 stages. After the L3 stage the number of wild-type puncta continues to increase. However, in mtm-6 mutants, the number of puncta appears to stall between L3 and young adulthood. While an average of 2.35 puncta are added between the young adults and adults with one row of eggs in the mtm-6 mutant, this number is lower than the average 4.6 puncta added between wild-type young adults and 1-row adults ( Fig. S1 and Table S1). This developmental study shows that the synapse number phenotype in the mtm-6 mutant is not likely due to synapse elimination or Quantification of the number of puncta formed in wild-type and mtm-6 (ok330) animals using GFP::RAB-3 to mark presynapses and UNC-10::GFP as a marker for active zones (n.60, **** is p,0.0001). (C-F) mtm-6 (ok330) animals show a reduced number of dorsal synaptic puncta when compared to wild-type animals for the GFP::RAB-3 and UNC-10::GFP markers. (C) Wild-type; GFP::RAB-3. (D) Wild-type; UNC-10::GFP. (E) mtm-6(ok330); GFP::RAB-3. (F) mtm-6(ok330); UNC-10::GFP. The UNC-10::GFP puncta on the ventral side of the animal correspond to the VA12 motor neuron. The dashed boxes mark the regions that are enlarged for the insets above each window. Error bars are SD. Scale bar is 10 mm. maintenance defects, but instead represents a compromised ability to add synapses during development.
To test if mtm-6 and mtm-9 function together, we examined the phenotypes in the double mutants. The mtm-9 (ok3523) allele is a large deletion and a null [30]. If mtm-6 and mtm-9 function in the same genetic pathway, the GFP::RAB-3 phenotype in the double mutants should not be more severe when compared to either of the single mutants. Indeed, mtm-6; mtm-9 double mutants show 14.8 (+/22.3) GFP::RAB-3 puncta, which is not significantly different from mtm-6 or mtm-9 alone (Fig. 2C and Table S1, n.60, SD). The double mutant between the two inactive phosphatases shows no significant decrease, which suggests it is unlikely mtm-5 functions with mtm-6 or a different myotubularin to reduce the number of synaptic puncta ( Fig. 2C and Table S1, 14.1+/23.0, n.60, SD). mtm-6 functions both within the DA9 neuron and within cells in the tail to promote DA9 synapse formation mtm-6 is part of a large operon, which has made it challenging to observe its expression pattern directly. Previous studies on mtm-6 used fragments of the operon to drive expression of fluorescent proteins, and they reported expression of the gene in the intestine [23]. Since the inactive myotubularin MTM-9 is known to act as a dimer with MTM-6, Silhankova et al. (2010) observed the mtm-9 expression pattern and inferred that the mtm-6 expression pattern would be similar. They found that mtm-9 is expressed in the hypodermis, neurons, intestine, muscle, and epithelial cells, including the rectal epithelial cells known to secrete EGL-20. We tagged the fosmid containing the complete mtm-6 operon with GFP::SL2::mCherry. This method revealed that mtm-6 is expressed in many neurons throughout the worm, the pre-anal ganglion, hypodermal cells, the anal Loss of a Lipid Phosphatase Reduces Synapse Number in C. elegans depressor muscle, and additional non-neuronal of cells in the tail (Fig. 3). The expression pattern is fairly constant through development.
The phenotype in the DA9 motor neuron and the expression pattern of mtm-6 suggest the hypothesis that gene could act within neurons. Though some myotubularin family members affect muscle structure in humans, when we compared mtm-6 mutant animals to wild-type we found that they exhibit normal muscle morphology (Fig. S2). As a result, the hypothesis that the neuronal phenotype of mtm-6 could arise from a defect in its post-synaptic partner is unlikely. An alternate hypothesis arises from the fact that MTM-6 is also reported to regulate Wnt secretion by influencing the recycling of the Wnt ligand carrier protein MIG-14/WIs [30]. Loss of mtm-6 causes Wnt-like neuronal migration phenotypes in C.elegans [30]. The loss of mtm-6 causes a visible reduction in the levels of MIG-14 in both C.elegans and D. melanogaster [30]. It is plausible that the reduction of GFP::RAB-3 puncta in the mtm-6 mutant could be due to a defect in Wnt secretion.
To test where MTM-6 functions, we expressed mtm-6 in various cell types in the mtm-6 mutant animals. We observed significant rescue of the GFP::RAB-3 puncta number when we expressed genomic DNA containing mtm-6 and its operon ( Fig. 4G and Table S1, 19.7+/24.3, p,0.0001, n.50, SD). We also created transgenic lines that express the cDNA of mtm-6a, one isoform of mtm-6, in the DA9 neuron (mig-13 promoter), in LIN-44 secreting cells (lin-44 promoter) and in cells expressing egl-20 (egl-20 promoter) (Fig. 4A-F). The mig-13 promoter rescues the number of puncta in mtm-6 mutants so that they are indistinguishable from wild-type levels ( Fig. 4B-D, G and Table S1, 21.7+/24.2, n.60, SD). This result suggests that mtm-6 functions cell-autonomously in the DA9 neuron to regulate synapse number. Since the mig-13 promoter is expressed in the VA12 neuron and a couple of tail cells in addition to DA9, we wanted to more stringently observe whether or not mtm-6 functions in DA9. We placed a loxp::stop::loxp cassette between the mig-13 promoter and mtm-6a so that expression of the cDNA would be suppressed in the absence of the cre recombinase. We expressed this construct in the background of a strain that contained a single copy insertion of cre driven by a promoter that is only expressed in the DA neurons. DA9 is the only cell that both the mig-13 promoter and the DA specific promoter share, so the suppression of the expression of the mtm-6a cDNA is only removed within DA9. When we used this system to limit the expression of mtm-6a to DA9, we found that there is still significant rescue in animals containing the rescue construct with respect to the mtm-6 mutant ( Fig. 4G and Table S1, 17.7+/23.1, p,0.0001, n.40, SD). The rescue is also significantly different from wild-type (p,0.0001), which implies that MTM-6 functions partially within the neuron, but might also have non-cell autonomous roles.
If the mtm-6 phenotype is the result of a combined effect of mtm-6 action within in the neuron and in the EGL-20 secreting cells, then coexpression of the mtm-6a cDNA using the egl-20 promoter and the mig-13/cre-lox system should demonstrate improved rescue as compared with individual expression under the promoters. When we coexpressed mtm-6a from the two promoters we found that we achieved a rescue of the phenotype that is indistinguishable from wild-type ( Fig. 4G and Table S1, 22.4+/23.3).
Finally, we tested whether or not the phosphatase activity of MTM-6 is required to form the proper number of synapses using a previously characterized point mutation [24]. The introduction of the C335S mutation into the phosphatase domain of the mtm-6a construct prevents the rescue of the mtm-6 mutant phenotype ( Fig. S3 and Table S1, 16.95+/23.1, n.40, SD). Furthermore, the C335S mutation causes no dominant negative effect in the wild-type background ( Fig. S3 and Table S1). This suggests the phosphatase domain is required to maintain the proper number of synapses in DA9.

MTM-6 functions partially through EGL-20 secretion to promote DA9 synapse formation
Our previous works show that two wnt gradients formed by LIN-44 and EGL-20 in the tail region of the worm inhibit synapse formation in the most posterior segment of DA9 axon [56,60]. The synapse-free segments of the axon include the commissure and an ''asynaptic'' region in the dorsal DA9 axon [56,60]. Crucially, Silhankova et al. (2010) observed that multiple Wnt processes are impacted by loss of mtm-6. Consistent with the hypothesis that the DA9 phenotype in mtm-6 mutants is partially due to impaired Wnt secretion, we observed a subtle shift of synaptic material into the asynaptic region of the neuron with respect to wild-type ( Fig. 4B-C). Therefore, we compared the phenotypes of the mtm-6 mutant with those of different Wnt mutants.
There are five Wnt ligand genes in the C. elegans genome: lin-44, egl-20, cwn-1, cwn-2 and mom-2. We focused our attention on the first four as mom-2 predominantly functions in endoderm specification [61]. To better understand the mtm-6 phenotypes, we counted the total number of puncta in the dorsal region, the number of puncta in the dorsal asynaptic region and the number of puncta in the commissure (Fig. 5A). The quantification of these parameters allows for characterization of two different synaptic phenotypes. The first phenotype is the number of synapses. The second phenotype is the position of the synapses. The loss of inhibitory Wnt signaling causes a shift of synaptic material into the dorsal asynaptic region. The most severe Wnt phenotypes show an additional shift of material into the commissure.
If MTM-6 primarily regulates EGL-20 secretion, then there should be no enhancement between the egl-20 and mtm-6 mutants. Indeed, we found that the mtm-6; egl-20 double mutant shows no significant enhancement of the synaptic number phenotype (Fig. 5C-E, H and Table S1, 16.1+/22.3, n540, SD, p.0.9998). In contrast, the mtm-6; lin-44 double mutant shows enhanced phenotypes with puncta appearing in the commissure (Fig. 6C-E and Table S1, 3.6+/23.1, n540, SD). From prior reports, the appearance of puncta in the commissure is typical of loss of the frizzled receptor lin-17 or loss of the combination of egl-20 and lin-44 (Fig. 6B-D and Table S1) [56,60]. As previously mentioned, it represents a more severe loss of inhibitory positional cues. The   Table S1). There is no enhancement of the lin-44 mutant by either the cwn-1 or cwn-2 mutants, suggesting that cwn-1 and cwn-2 are not involved ( Fig. S5 and Table S1). The fact that the mtm-6 and egl-20 mutants enhance the lin-44 mutant in a similar manner supports the hypothesis that MTM-6 regulates EGL-20 secretion.
If mtm-6 and egl-20 function in a linear pathway, then the mtm-6; egl-20; lin-44 triple mutant should not be enhanced as compared with the lin-44; egl-20 or the lin-44; mtm-6 double mutants. Consistent with this hypothesis, there is no significant increase in either the dorsal puncta number or commissural puncta number in the triple mutant (Fig. 5H, Fig. 6E, and Table S1). There is an increase in the incidence of defective axon guidance (data not shown). In the lin-44; egl-20 or the lin-44; mtm-6 double mutants, the incidence of defective axon guidance is around 9-10%. The mtm-6; egl-20; lin-44 triple mutant has an incidence of 28.6% for defective axon guidance. We have excluded the animals with axon guidance defects when scoring synapse phenotypes. Together, these genetic interaction results regarding the synapse phenotypes are consistent with the notion that mtm-6 and egl-20 function in the same genetic pathway to prevent DA9 synapses from forming in the asynaptic region. Yet, these results do not explain the neuronal function of the gene or exclude the possibility that mtm-6 could function in a parallel and non-linear pathway.
To further understand the cellular action of egl-20 and mtm-6, we tested whether mtm-6 is required in EGL-20 secreting cells with additional cellautonomous rescue experiments. The initial rescue experiments of the mtm-6 single mutant suggest that the mtm-6 synapse number phenotype is partially dependent on egl-20, since both the mig-13/cre-lox system and the egl-20 promoter partially rescue the phenotype, but can fully rescue the phenotype in combination (Fig. 4G). If this hypothesis is true, then we should see partial rescue of the synapse number phenotype in the egl-20; mtm-6 and the lin-44; egl-20; mtm-6 mutants when mtm-6a is expressed under the mig-13 promoter. Indeed, we observed partial rescue of the reduced synaptic number in the absence of egl-20 ( Fig. 5E-G, I-J, and Table S1). Not only is rescue possible in the mtm-6; egl-20 double mutant, but it is possible to rescue the lin-44; egl-20; mtm-6 mutant by the mig-13 promoter driving mtm-6a (Fig. 5I and Table S1, 18.7+/23.6, n540, p,0.0001, not significantly different from wild-type, SD). In contrast, no rescue is seen in the background of the triple mutant when mtm-6a is expressed under the egl-20 promoter. This result is consistent with two separate functions of MTM-6 in regulating the synapse number of DA9: an EGL-20 independent, cellautonomous function of mtm-6 in DA9, and an EGL-20 dependent function in the EGL-20 secreting cells. While these results elucidate the relationship of mtm-6 and egl-20 in the context of the synapse number phenotype, they do not indicate their interaction with respect to the positional phenotype.
To further understand the interaction between mtm-6 and egl-20 in regulating the location of synapses, we examined the number of ectopic puncta in the commissure in mtm-6; lin-44 as well as lin-44; egl-20; mtm-6 mutants. The commissural puncta phenotype in the mtm-6; lin-44 double mutant is rescued when mtm-6a is expressed from the egl-20 promoter, suggesting that mtm-6 is required for the secretion of EGL-20 for the purpose of specifying the asynaptic region (Fig. 6F). There is also a reduction in the number of commissural puncta when mtm-6a expression is driven by the mig-13 promoter or the mig-13/cresystem ( Fig. 6E and Table S1). However, the ability to rescue the enhanced Wnt phenotype by driving mtm-6a expression with all promoters tested is lost in the lin-44; egl-20; mtm-6 mutant ( Fig. 6E and Table S1). The data suggests that expressing mtm-6a from either neurons or the EGL-20-expressing cells is sufficient to create enough EGL-20 to rescue the enhanced Wnt phenotype in the double. Yet, the failure to rescue the triple mutant ensures that this process is through egl-20 and not a parallel pathway. The relation of mtm-6 and egl-20 with respect to the phenotype of commissural puncta GFP::RAB-3 is linear, which is contrast to the non-linear relation of mtm-6 and egl-20 for the number of GFP::RAB-3 puncta.

The Relative Specificity of mtm-6 for EGL-20 secretion
The genetic data and prior research by Silhankova et al. (2010) support the hypothesis that mtm-6 acts predominantly on egl-20 and not the other Wnt ligands. We hypothesized that the relative specificity of MTM-6 for EGL-20 secretion could result from the expression pattern of mtm-6. We observed whether or not there was overlap between the mtm-6 tagged-fosmid expression pattern and the expression pattern of different Wnts. We found that there is overlap between the mtm-6 and the egl-20 expression patterns in the anal depressor muscle, but there is little to no overlap of the mtm-6 expression pattern with the lin-44 expression pattern (Fig. 7A-F). This implies that different Wnt ligands may rely on distinct components to their secretory paths based on where they are expressed.

Discussion
Initial research on the myotubularin family focused on mutations in the myotubularin genes underlying X-linked myotubular myopathy and two Charcot-Marie-Tooth neuropathies. In the former case myotubularins act in muscle and in the second the myotubularins are required for myelination of neurons by Schwann cells [16,21]. Yet, a growing body of research implies that myotubularins and the phosphoinositides they regulate also function in neurons. MTMR2 is expressed in peripheral nerves and is capable of affecting the number of dendritic spines in excitatory neurons [36,62]. It is plausible that the neuropathies are caused in part by defects in neurons. Here we demonstrate that loss of mtm-6 affects both the number of synaptic puncta and their location within the motor neuron DA9. We propose that mtm-6 impacts synapses in part through its regulation of Wnt secretion and in part through a Wnt independent function within the neuron.
That loss of mtm-6 seems to have a greater impact on egl-20 rather than the other Wnt ligands is consistent with a previous study by Silhankova et al. (2010). In their report, several Wnt-mediated processes are observed to have varying dependencies on the different Wnt ligands. The impact of mtm-6 on the different Wnt mediated processes is also variable. The mtm-6 mutation has a dramatic effect on the QL cell migration, which is a process that is largely dependent on EGL-20 secretion [52,55,63]. Similarly, loss of mtm-6 has a noticeable affect on HSN migration, a process that is known to be affected by loss of egl-20 [55]. However, the mtm-6 mutation has no affect on PLM polarization and only a small but reproducible affect on ALM polarization. PLM polarity is dependent on lin-44, cwn-1 and egl-20, while ALM is dependent on egl-20, cwn-1 and cwn-2. If mtm-6 affected all Wnt secretion equally, then one would expect there to be a more marked affect on ALM and PLM polarization, as is seen by the more generally required Wnt secretion factor mig-14/wntless [55]. However, we acknowledge that interpreting the genetic interaction results both in our works and prior works is complicated with respect to the intricate series of secreted Wnt ligands in this system and that alternate models are possible. How mtm-6 is able to modulate synapse number within the neuron is unclear and could be of interest for future research. It is also unclear at this point how or if the synapse number phenotype of the mtm-6 mutant relates to inhibitory Wnt positional cues. It is of interest is that mtm-6 is expressed in many neurons. We observed a subtle backing defect in mtm-6 worms, which assisted in genotyping mtm-6 homozygous animals. When we observed other neurons we observed minor effects in AIY and PLM. The other active myotubularins, mtm-1 and mtm-3, may act redundantly with mtm-6 to prevent more widespread effects in the nervous system or to prevent more distinct movement defects. Myotubularin expression is also seen in the central nervous system and in peripheral nerves in mammalian systems. Myotubularins have been linked to Creutzfeldt-Jakob disease and epilepsy in genomic studies [64,65]. C. elegans may provide a more genetically tractable system for understanding neuronal myotubularin function and shedding light on their roles in neuronal disease.
The reduced number of synaptic puncta in the mtm-6 mutant led to the realization that egl-20 mutants also have a reduced number of synaptic puncta. This novel observation was made possible by stringently maintaining consistent growth conditions across the different strains observed since the observed phenotypes were temperature sensitive as visualized by GFP::RAB-3 or GFP::UNC-10. That EGL-20 can modulate the number of puncta in DA9 highlights the Wnt family's diverse portfolio of functions in nervous system. Wnts are involved at nearly every step of neuronal circuit development, from the initial precursor migration to the fine-tuning of acetylcholine receptors at a synapse [42][43][44][45][46][47][48][49][50][51][52][53][54][55][56][57][58]. egl-20 is known to have a strong effect in controlling the synaptic pattern of the DA9's neighbor, DA8 [60]. DA8 and DA9 form their synapses in adjacent but non-overlapping regions leading to tiled synaptic innervations [66]. egl-20 inhibits synapse formation in DA8's asynaptic region in much the same way lin-44 inhibits DA9's synapses to assist in creating the tiled synaptic domains [60]. egl-20 may act directly on DA9, or it may be that DA9 reacts to the effect of egl-20 on DA8.
egl-20 and mtm-6 appear to act in parallel paths with respect to controlling synapse number. EGL-20 secretion can influence the number of synapses in the mtm-6 mutant, hinting that there may be overlap between the two genes further downstream of each pathway (Fig. 4G). The developmental data suggests that the mtm-6 phenotype could be the result of inefficient addition of synapses in the process of animal growth (Fig. S1). The addition of synapses requires the biosynthesis of presynaptic material in the cell body, long-range transport of this material along the axon, and local aggregation of synaptic vesicle with active zone proteins. It seems plausible that the MTM-6 could influence vesicle trafficking through phospholipid modulation since phospholipids are implicated in several aspects of intracellular membrane trafficking. Future experiments will be required to determine the site of action for MTM-6 in synapse formation.

Fluorescence Microscopy and Quantification
Images were taken using a Plan-Apochromat 4061.3 objective on a Zeiss LSM710 confocal microscope. Worms were immobilized using a mixture of 5 mM Levamisole (Sigma-Aldrich) and 500 mM 2,3-butanedione monoxime (Sigma-Aldrich) on a 2% agarose pad. Image montages were created and analyzed following the methods previously described with one modification [56]. The images were aligned along anteroposterior axis using the region where the axon passes above the cell body and the gut rather than using DA9 commissure as previously reported.
Muscle arms were imaged by soaking worms in 3 mM muscimol for a 5 minutes and then transferring the worms on to a 5% agarose pad. The worms were rolled using the coverslip so that some of worms were dorsal side up.
Synapses were counted using a Zeiss AxioSkop or AxioVivison epifluorescent microscope with a mercury or X-cite light source. Worms were immobilized with 1 mM Levamisole on a 2% agarose pad. For each strain, n.40. Error bars represent Standard Deviation. Statistical analyses were performed with a one-way ANOVA test followed by Tukey's multiple comparis on test.