Hubrecht Institute, Royal Academy of Arts and Sciences and University Medical Center Utrecht, Utrecht, The Netherlands
Corresponding author. Hubrecht Institute, Royal Academy of Arts and Sciences and University Medical Center Utrecht, Upsalalaan 8, Utrecht, CT 3584, The Netherlands. Tel.: +31 30 212 1981; Fax: +31 30 251 6464; E-mail: firstname.lastname@example.org
Wnt proteins are lipid-modified glycoproteins that have important roles in development, adult tissue homeostasis and disease. Secretion of Wnt proteins from producing cells is mediated by the Wnt-binding protein MIG-14/Wls, which binds Wnt in the Golgi network and transports it to the cell surface for release. It has recently been shown that recycling of MIG-14/Wls from the plasma membrane to the trans-Golgi network is required for efficient Wnt secretion, but the mechanism of this retrograde transport pathway is still poorly understood. In this study, we report the identification of MTM-6 and MTM-9 as novel regulators of MIG-14/Wls trafficking in Caenorhabditis elegans. MTM-6 and MTM-9 are myotubularin lipid phosphatases that function as a complex to dephosphorylate phosphatidylinositol-3-phosphate, a central regulator of endosomal trafficking. We show that mutation of mtm-6 or mtm-9 leads to defects in several Wnt-dependent processes and demonstrate that MTM-6 is required in Wnt-producing cells as part of the MIG-14/Wls-recycling pathway. This function is evolutionarily conserved, as the MTM-6 orthologue DMtm6 is required for Wls stability and Wg secretion in Drosophila. We conclude that regulation of endosomal trafficking by the MTM-6/MTM-9 myotubularin complex is required for the retromer-dependent recycling of MIG-14/Wls and Wnt secretion.
During the development of multicellular organisms, cells communicate with each other to coordinate complex processes, such as cell fate determination and morphogenesis. A family of secreted signalling proteins, which is used extensively in development and adult tissue homeostasis, comprises the evolutionarily conserved Wnt proteins (Clevers, 2006). Wnts are characterized by a double lipid modification with palmitic and palmitoleic acid (Willert et al, 2003; Takada et al, 2006), raising questions on how these hydrophobic proteins are secreted and released from Wnt-producing cells and how they disperse in the tissue. Although much attention has been focused on how the spreading of Wnt is mediated and regulated (for review, see Eaton, 2006; Kornberg and Guha, 2007), the mechanism of Wnt secretion is still poorly understood.
Further analysis of the function of MIG-14/Wls and the retromer in Wnt-producing cells provided evidence for a model in which MIG-14/Wls cycles between the Golgi and the plasma membrane to mediate Wnt secretion (Belenkaya et al, 2008; Franch-Marro et al, 2008; Pan et al, 2008; Port et al, 2008; Yang et al, 2008). After transport to the plasma membrane, MIG-14/Wls is endocytosed and recycled back to the Golgi to take part in a new round of Wnt secretion. The endosome to Golgi recycling step is mediated by the retromer complex and mutation of the cargo-selective subunits leads to degradation of MIG-14/Wls in the lysosomal pathway. As a consequence, less MIG-14/Wls is available to mediate Wnt secretion and Wnt signalling is disrupted. The internalization of MIG-14/Wls is dependent on the AP2 endocytotic adaptor complex (Pan et al, 2008; Port et al, 2008; Yang et al, 2008) and when AP2- and clathrin-mediated endocytosis is blocked, MIG-14/Wls accumulates on the cell surface. Also in this case, less MIG-14/Wls is available for Wnt secretion, explaining the Wnt-signalling defect of AP2 subunit mutants (Pan et al, 2008; Port et al, 2008; Yang et al, 2008).
To further examine the regulation of MIG-14/Wls trafficking, we focused on the endocytotic step of the pathway. We analysed an existing panel of C. elegans endocytosis-defective mutants and discovered that two members of the myotubularin-related family of lipid phosphatases, MTM-6 and MTM-9, are required for efficient MIG-14/Wls recycling. We provide evidence that the role of MTM-6 in MIG-14/Wls trafficking is mediated by the sorting nexin family member SNX-3. We further show that the function of MTM-6 in MIG-14/Wls recycling is evolutionarily conserved in Drosophila. Our data extend the current model of regulation of Wnt secretion and provide the first functional link between the conserved MTM-6 and MTM-9 myotubularins and the Wnt signalling pathway.
The myotubularin genes mtm-6 and mtm-9 are required for the EGL-20/Wnt-dependent posterior migration of the QL.d
To understand the mechanism of Wnt secretion better, we focused on the regulation of endocytosis and recycling of MIG-14/Wls. We made use of an existing collection of viable C. elegans mutants with known defect in endocytosis (Fares and Greenwald, 2001). These mutants are defective in the uptake of a secreted form of GFP by specialized endocytotic cells, the coelomocytes (called a Cup phenotype, for coelomocyte uptake defective). We investigated whether these mutants display a defect in Wnt signalling by assessing the position of the Q neuroblast descendants. The C. elegans L1 larva is born with two Q neuroblasts, on the left and right sides (QL and QR), which later generate an identical set of descendants (Q.d). The Q.d on the right side of the larva migrate in a default anterior direction, whereas the Q.d on the left side migrate to the posterior of the animal (Figure 1A). Posterior migration is regulated by the Wnt protein EGL-20 that induces the expression of the homeotic gene mab-5 in QL and its descendants (Salser and Kenyon, 1992). In egl-20/Wnt, mig-14/Wls or vps-35/retromer mutants, mab-5 fails to be expressed and as a consequence, the QL.d migrate towards the anterior (Salser and Kenyon, 1992; Harris et al, 1996; Coudreuse et al, 2006; Prasad and Clark, 2006). Among nine genes tested, we found two Cup mutants with a significant proportion of anteriorly misplaced QL.d, mtm-6 and mtm-9 (Figure 1B). A low but reproducible QL.d migration defect was also seen in cup-4 mutants.
mtm-6 and mtm-9 encode members of the myotubularin family of lipid phosphatases (Xue et al, 2003) and have been shown to form a heterodimer, in which MTM-6 is an active phosphatase, whereas MTM-9 lacks a critical residue in its catalytic site (Dang et al, 2004). We found that mtm-6; mtm-9 double mutants do not show a significantly higher QL.d migration defect than mtm-9 single mutant, which supports their functioning as a complex (Figure 1C). The mtm-9(ar479) allele contains a premature stop (W135STOP) mutation and mtm-6(ok330) is a 1235-base-pair deletion spanning the central region of the protein (Supplementary Figure 1A; Dang et al, 2004). The lower penetrance of the mtm-6(ok330) QL.d migration phenotype in comparison with mtm-9(ar479) might be caused by residual function of the truncated protein. We therefore tested two other mtm-6 alleles: ar513, a stop mutation causing a C-terminal truncation that removes the FYVE domain, and ar515, a missense mutation in the catalytic site (Fares and Greenwald, 2001; Dang et al, 2004). Both alleles show a reproducible QL.d migration phenotype; however, in neither of them the penetrance reaches the levels observed with mtm-9 (15 and 25%, respectively, compared with 62% in mtm-9 mutants; n>100). An alternative explanation may therefore be that mtm-6 acts partially redundantly with other myotubularin-related genes.
There are three other myotubularin genes encoded in the C. elegans genome, mtm-1, mtm-3 and mtm-5 (Xue et al, 2003). To test whether the incomplete penetrance of the QL.d migration defect in mtm-6 mutants is due to functional redundancy among the family members, we scored QL.d migration in mtm-1 and mtm-5 mutants and in animals in which mtm-1, mtm-3 and mtm-5 were knocked down by RNAi (Figure 1C). None of them displayed a defect in QL.d migration. However, in mtm-6; mtm-5 double mutants the QL.d migration defect was similar to that observed in mtm-6; mtm-9 double mutants. A small enhancement was also observed with mtm-5 and mtm-3 RNAi in an mtm-6 mutant background (although only the enhancement by mtm-3 RNAi was statistically significant). A more pronounced enhancement was observed with mtm-1 RNAi in the mtm-6 background but not in mtm-1(op309); mtm-6 double mutants. This may be due to the hypomorphic nature of the op309 allele (M Hengartner, personal communication). Taken together, our data indicate that the myotubularins mtm-6 and mtm-9 are required for proper migration of the QL descendants. Other myotubularins may have a minor role in QL.d migration by acting partially redundantly with mtm-6 and mtm-9. In most of the following study, we focused on mtm-6 as the enzymatically active part of the complex.
Next, we investigated whether other cup mutants act in parallel with mtm-6 in regulating QL.d migration. We therefore constructed double mutants between mtm-6 and other cup mutants and assessed the penetrance of the QL.d migration phenotype. Out of the four mutants tested, only cup-4 enhanced the phenotype of mtm-6 (Figure 1D). This suggests that mtm-6 and cup-4 function synergistically in the regulation of Q cell migration.
mtm-6 mutants display defects in several Wnt-dependent processes
EGL-20/Wnt controls the posterior migration of the QL.d through induction of the homeotic gene mab-5 (Salser and Kenyon, 1992; Harris et al, 1996). To investigate whether the defect in QL.d migration is caused by a loss of mab-5 expression in the QL lineage, we analysed the expression of a mab-5∷lacZ reporter that has been shown to closely mimic the expression of the endogenous mab-5 gene (Cowing and Kenyon, 1992). In wild-type animals, mab-5 was strongly expressed in QL and its descendants, but in a significant fraction of mtm-6(ok330) mutants, no mab-5 expression was observed in the QL lineage (Figure 2A). Importantly, the EGL-20-independent expression of mab-5 in other cells, such as the posterior ventral nerve cord neurons (Salser and Kenyon, 1992), was not affected. These results indicate that mtm-6 is required for the EGL-20/Wnt-dependent induction of mab-5 expression. This conclusion is further supported by the observation that the QL.d migration defect of mtm-6 is rescued by EGL-20/Wnt-independent activation of mab-5. Thus, the QL.d localized at their normal posterior position in double mutants between mtm-6 and the mab-5 gain-of-function mutation e1751 (Figure 2B) (Salser et al, 1993). Furthermore, posterior localization of the QL.d was partially restored in double mutants between mtm-6 and pry-1/Axin (Figure 2B), a negative regulator of the canonical Wnt/β-catenin pathway that is triggered by EGL-20/Wnt in the QL lineage (Maloof et al, 1999; Korswagen et al, 2002).
We next examined whether loss of mtm-6 and mtm-9 also affects other Wnt-dependent processes during C. elegans development, such as the anterior migration of the HSN neurons and the QR descendants and the polarization of the ALM and PLM mechanosensory neurons, which depends on the simultaneous action of several Wnt proteins (Pan et al, 2006; Prasad and Clark, 2006). We found that the HSN neurons and the QR.d were significantly under-migrated in the mtm-6 and mtm-9 single and double mutants (Figure 2C). Furthermore, there was a weak but reproducible defect in the polarity of the ALM neurons (Table I), but no defect was observed in the Wnt-dependent polarization of the seam cell V5 or in the fate specification of the P11/12 ventral epithelial cells. Taken together, these results demonstrate that mtm-6 and mtm-9 are required for several Wnt-dependent processes.
Table 1. ALM and PLM polarity defects in myotubularin and mig-14/Wls mutants
The polarity of the ALM and PLM neurons was assayed using the mec-7∷gfp expressing transgenes muIs32 or muIs35 (Ch'ng et al, 2003). Both reversal and loss of polarity were scored as defective polarization. In each case, n>150.
MTM-6 is required in EGL-20/Wnt-producing cells
It has previously been reported that mtm-6 is predominantly expressed in the intestine (Xue et al, 2003), which is in disagreement with the function of mtm-6 in EGL-20/Wnt signalling. mtm-6 is part of a large operon (Supplementary Figure 1B), which complicates the determination of its expression pattern. The promoter element used by Xue et al (2003) encompasses only part of the operon sequence and may therefore lack essential regulatory regions. We were also unable to generate a reporter construct that includes the complete mtm-6 operon. We therefore determined the expression pattern of mtm-9, as MTM-6 and MTM-9 act as a complex and are thus likely to be co-expressed (Dang et al, 2004). We found that a transcriptional mtm-9∷gfp reporter was expressed in a broad range of tissues, including the muscle, intestine, hypodermis and neurons (Supplementary Figure 2; data not shown). mtm-9 was also expressed in the rectal epithelial cells that are the major source of EGL-20/Wnt (Whangbo and Kenyon, 1999; Figure 3A). The widespread expression of mtm-9 is consistent with the function of the MTM-6–MTM-9 complex in several Wnt-dependent processes.
To investigate whether MTM-6 is required in Wnt-responding or -producing cells, we expressed MTM-6 tagged with GFP under the control of different tissue-specific promoters in a mtm-6(ok330) mutant background and assayed the EGL-20/Wnt-dependent migration of the QL.d. We found that expression of MTM-6 under the control of the egl-20 promoter (Coudreuse et al, 2006) was sufficient for a complete rescue of the QL.d migration defect (Figure 3B). Expression of MTM-6 from the cdh-3 promoter (Pettitt et al, 1996), which overlaps in expression with the egl-20 promoter in only one cell, resulted in a partial rescue. Importantly, when we made a phosphatase-inactive version of MTM-6 by mutating a conserved cysteine in the catalytic domain (C335S) (Dang et al, 2004) and expressed it from the egl-20 promoter, we did not observe any rescue. These results demonstrate that MTM-6 is required in Wnt-producing cells and that the phosphatase activity is indispensable for its function.
Knockdown of the Vps34 PI3 kinase complex restores EGL-20/Wnt signalling in mtm-6 mutants
Phosphatidylinositol-3-phosphate (PI3P), the substrate of MTM-6, is generated by the phosphatidylinositol-3-kinase VPS-34 (Schu et al, 1993). To further investigate the importance of phosphoinositide turnover in EGL-20/Wnt signalling, we examined the function of vps-34 in QL.d migration. As vps-34 mutants are embryonic lethal (Xue et al, 2003), we used RNAi to partially knock down vps-34 expression. Knockdown of vps-34 produced no defect in the localization of the QL.d in wild-type animals. However, vps-34 RNAi partially rescued the QL.d migration defect of mtm-6 mutants (Figure 3C). A similar effect was achieved with RNAi against ZK930.1 and bec-1, genes encoding two subunits of the Vps34 kinase complex, Vps15 and Vps30/Beclin1, respectively (Figure 3C; Stack et al, 1993; Kihara et al, 2001). These results support the idea that the primary cause of the Wnt-signalling defect in mtm-6 mutants is an excess of PI3P and that a proper balance between PI3P synthesis and degradation is required for efficient Wnt signalling.
MIG-14/Wls protein levels are reduced in mtm-6 mutants
As mtm-6 is required in Wnt-producing cells, we examined whether mtm-6 genetically interacts with mig-14/Wls, the putative sorting receptor for Wnt that is an essential component of the Wnt secretion pathway (Banziger et al, 2006; Bartscherer et al, 2006; Goodman et al, 2006). Animals carrying the mig-14 null allele or78 die during early embryogenesis (Thorpe et al, 1997), so to test for genetic interactions, we made use of two viable reduction-of-function alleles, mu71 and ga62 (Yang et al, 2008). As these mutants show a fully penetrant defect in QL.d migration, we focused on the Wnt-dependent polarization of the ALM and PLM neurons. Whereas mtm-6(ok330) animals displayed only a weak defect in ALM polarity and no defect in PLM polarity (see above), the mtm-6 mutation strongly enhanced both the ALM and PLM polarization defect of mig-14(mu71) and the PLM polarization defect of mig-14(ga62) (Table I). Although these results do not distinguish between a function of mtm-6 and mig-14 in shared or parallel genetic pathways, they are consistent with a functional interaction between MTM-6 and MIG-14 in Wnt signalling.
MIG-14/Wls cycles between the Golgi complex, the plasma membrane and endosomes to mediate Wnt secretion (Belenkaya et al, 2008; Franch-Marro et al, 2008; Pan et al, 2008; Port et al, 2008; Yang et al, 2008). To investigate whether mtm-6 affects the endocytosis or trafficking of MIG-14/Wls, we examined the level and subcellular localization of a functional MIG-14∷GFP fusion protein in mtm-6(ok330) mutants. As shown in Figure 4A and B, there was a significant reduction in MIG-14∷GFP protein levels, with most of the remaining MIG-14∷GFP protein localizing to punctate structures, which most likely represent localization of MIG-14∷GFP to the endolysosomal system. A similar effect on MIG-14∷GFP level and subcellular localization is observed in mutants of the cargo-selective retromer sub-complex (Belenkaya et al, 2008; Franch-Marro et al, 2008; Pan et al, 2008; Port et al, 2008; Yang et al, 2008). Comparison between MIG-14∷GFP in mtm-6 and vps-26(tm1523) mutants showed, however, that MIG-14 protein levels were about three-fold lower in vps-26 than in mtm-6 mutants. The stronger reduction in MIG-14 protein levels in vps-26 mutants is in agreement with the observation that vps-26 mutants (Coudreuse et al, 2006) have a more penetrant defect in EGL-20/Wnt signalling than mtm-6 mutants.
To investigate whether MIG-14/Wls becomes limiting for Wnt signalling in mtm-6 mutants, we tested whether mig-14 overexpression can rescue EGL-20/Wnt signalling. To overexpress mig-14, an extrachromosomal transgene was used that expresses mig-14 under the control of the heat-shock promoter (Yang et al, 2008). A short heat shock during the early L1 stage, which is before the EGL-20/Wnt-dependent activation of mab-5 expression, fully rescued QL.d migration in animals carrying the transgene, but not in their non-transgenic siblings (Figure 4C). The rescue of Wnt signalling by mig-14 overexpression is consistent with a function of mtm-6 in MIG-14/Wls trafficking and suggests that the defect in Wnt signalling is caused by a reduction in the pool of MIG-14/Wls that is available to mediate Wnt secretion.
Dmtm6 is required for Wg secretion and Wls stability in Drosophila
To investigate whether the function of mtm-6 in MIG-14/Wls trafficking is evolutionarily conserved, we turned to another well-studied model of Wnt signalling, which is the Drosophila wing imaginal disc. In the wing disc, a stripe of cells at the dorsoventral boundary of the disc expresses the Wnt protein Wingless (Wg; Zecca et al, 1996). When we knocked down the mtm-6 orthologue Dmtm6 in the posterior compartment of the disc by transgene-mediated RNAi (Supplementary Figure 3), we observed the accumulation of Wg protein in the Wg-expressing cells in the posterior compartment, but not in the control anterior compartment, suggesting that Wg is not efficiently secreted upon Dmtm6 knock-down (Figure 4D, Supplementary Figure 4). Importantly, Dmtm6 knockdown did not affect secretion of Hedgehog, another lipid-modified morphogen (Figure 4E, Supplementary Figure 4) showing that loss of Dmtm6 does not lead to a general defect in protein secretion. This conclusion is further supported by the mtm-6 mutant phenotype in C. elegans: apart from a defect in Wnt signalling, mtm-6 mutants have no other obvious developmental defects (data not shown).
To address whether Dmtm6 is also required for the control of Wls levels, we stained endogenous Wls protein in discs in which Dmtm-6 was knocked down. As shown in Figure 4D and Supplementary Figure 4, there was a significant reduction of Wls staining in the Wg-producing cells. Interestingly, Wls levels were also strongly reduced outside of the Wg-producing cells. This phenotype is different from that of Dvps35, where a reduction in Wls levels is only observed in the wg-expressing cells (Belenkaya et al, 2008; Franch-Marro et al, 2008; Port et al, 2008).
MTM-6 regulates binding of SNX-3 to MIG-14/ Wls-containing endosomes
The substrate of the enzymatic activity of MTM-6, PI3P, is a phosphoinositide species that is highly enriched on endosomal membranes and has an important role in many aspects of endosomal biology. One of its functions is to serve as a docking mark for proteins that regulate endosomal trafficking (for review, see Di Paolo and De Camilli, 2006). Among these are the sorting nexins (SNX), a family of proteins containing a PI3P-binding Phox-homology (PX) domain (Cullen, 2008), which are required for the recruitment of the cargo-selective retromer sub-complex to cargo-containing endosomes (Griffin et al, 2005; Rojas et al, 2007; Wassmer et al, 2007). We hypothesized that by regulating PI3P levels, MTM-6 may affect membrane association of the retromer and consequently MIG-14/Wls recycling. We have found that the sorting nexin SNX-3 functions together with the retromer complex in MIG-14/Wls recycling and Wnt secretion (M Harterink et al, in preparation) and therefore examined the sub-cellular localization of SNX-3 tagged with GFP in the egl-20-expressing cells of wild-type and mtm-6 mutant animals. Localization of SNX-3 in wild-type animals was mostly cytoplasmic with only a few punctate structures, whereas in mtm-6 mutants the punctate localization became much more prominent (Figure 5A and B). A similar phenotype was observed in the Drosophila wing imaginal disc, in which DSnx3 localized much more frequently to punctate structures in posterior cells expressing a Dmtm6 RNAi transgene compared with cells in the anterior where the RNAi transgene was not expressed (Figure 5C). This is consistent with the hypothesis that a defect in mtm-6 leads to increased PI3P levels and therefore to increased recruitment of PI3P-binding proteins to the membrane. To confirm that the observed SNX-3 accumulation is due to binding to endosomal PI3P, we mutated a single amino acid in the SNX-3 PX domain that abrogates PI3P binding (Pons et al, 2008). We again expressed this SNX-3(R70A)∷GFP mutant protein from the egl-20 promoter and compared its localization in wild-type and mtm-6 mutant animals. In this case, we did not observe any punctate localization in either wild-type or mtm-6 mutants (Figure 5A), confirming that the SNX-3 accumulation in mtm-6 mutants is indeed dependent on PI3P binding.
To investigate whether the SNX-3-positive vesicles contain MIG-14, we co-expressed MIG-14∷GFP and SNX-3∷mCherry in EGL-20/Wnt-producing cells. In wild-type animals, MIG-14 showed a more widespread sub-cellular distribution than SNX-3, but there was clear co-localization between the two on intracellular punctae, with most of the SNX-3-positive vesicles also containing MIG-14∷GFP (Figure 5D). In mtm-6 mutants, the remaining MIG-14∷GFP co-localized with SNX-3. Interestingly, the MIG-14-positive vesicles in mtm-6 mutants appeared larger and stained more intensely for MIG-14∷GFP in the SNX-3∷mCherry-expressing transgenic animals than in animals not carrying the snx-3 transgene (Figures 4A and 5D), suggesting that overexpression of SNX-3 in a mtm-6 mutant background may partially block MIG-14 trafficking. Consistently, we found that the mtm-6-induced defect in the EGL-20/Wnt-dependent migration of the QL.d was strongly enhanced by SNX-3 overexpression (18% in mtm-6(ok330) versus 92% in mtm-6(ok330); Pegl-20∷snx-3∷mcherry, whereas overexpression of SNX-3∷mCherry had no effect in wild-type animals, n>200). To investigate whether the enlarged SNX-3-decorated vesicles in mtm-6 mutants are endosomes, we co-expressed SNX-3∷mCherry with the endosomal marker 2 × FYVE∷GFP (Roggo et al, 2002). As shown in Figure 5E, SNX-3 and 2 × FYVE∷GFP localized to the same vesicles, confirming that these structures are endosomes. Taken together, these results demonstrate that mtm-6 regulates the PI3P-dependent association of SNX-3 to MIG-14-containing endosomes (Supplementary Figure 5). Disruption of this regulation induces a defect in MIG-14/Wls trafficking, which interferes with its recycling and ultimately leads to its degradation. We propose that a proper balance of SNX-3 recruitment and release at the endosomal membrane is critical for the efficient retromer-dependent recycling of MIG-14/Wls and the secretion of Wnt.
The discovery of the putative Wnt-sorting receptor, MIG-14/Wls, has shown that secretion of Wnt proteins requires a dedicated pathway that provides a previously unanticipated level of regulation to the Wnt signalling machinery (Banziger et al, 2006; Bartscherer et al, 2006; Goodman et al, 2006). An important aspect of this secretion pathway is that MIG-14/Wls cycles between the Golgi and the plasma membrane to mediate Wnt secretion. In the absence of plasma membrane-to-Golgi transport, Wnt secretion is diminished, as is the case in mutants of the retromer complex and mutants of the AP2 endocytic adaptor complex (Belenkaya et al, 2008; Franch-Marro et al, 2008; Pan et al, 2008; Port et al, 2008; Yang et al, 2008). Proteins directing Wls recycling are therefore important regulators of Wnt secretion.
In this study, we show that the myotubularin lipid phosphatase family members, MTM-6 and MTM-9, regulate MIG-14/Wls recycling in C. elegans and Drosophila. Myotubularin phosphatases have been shown to specifically dephosphorylate PI3P or PI(3,5)P2 (Taylor et al, 2000; Walker et al, 2001; Schaletzky et al, 2003), two important phosphoinositide species involved in regulation of membrane trafficking (for review, see Vicinanza et al, 2008). As several myotubularins act on the same substrate, they may either act redundantly or may dephosphorylate specific sub-pools of PI3P. We show that of the five C. elegans myotubularins genes, only mtm-6 and mtm-9 significantly affect Wnt signalling, suggesting that individual myotubularins have distinct roles. Indeed, minor functions of the other myotubularins in Wnt signalling are only observed in the absence of mtm-6 and mtm-9, indicating that they can only partially substitute for MTM-6/MTM-9 function. The specificity of myotubularins can be achieved either by tissue-specific expression, differential localization within the cell or by specific protein–protein interactions. Although initial studies hinted at the first possibility, showing non-overlapping patterns of mtm-1, mtm-3 and mtm-6 expression (Xue et al, 2003), later reports demonstrated broader and overlapping expression domains for mtm-1 and mtm-3 (Ma et al, 2008; Zou et al, 2009). Our own data also show a broad expression pattern of mtm-9, suggesting that like in mammals (Laporte et al, 2002), C. elegans myotubularins are expressed ubiquitously and their specificity is achieved differently. Localization of myotubularins to distinct sub-cellular compartments has been demonstrated by some in mammalian cells (Lorenzo et al, 2006; Cao et al, 2007), whereas others have reported diffuse cytoplasmic localization (Blondeau et al, 2000; Taylor et al, 2000; Laporte et al, 2002). An intriguing example of differential localization has been described for human MTM1 and MTMR2, in which MTM1 localizes to Rab5-positive early endosomes and only partially to Rab7-positive late endosomes (Cao et al, 2007), whereas MTMR2 localizes predominantly to Rab7-positive endosomes (Cao et al, 2008). Furthermore, corresponding changes in early or late endosomal PI3P pools have been achieved by knockdown of MTM1 or MTMR2, respectively (Cao et al, 2008). C. elegans MTM-6 and MTM-1 have been shown to localize to the plasma membrane in various tissues (Xue et al, 2003; Zou et al, 2009). In our hands, GFP-tagged MTM-6 showed a predominantly diffuse cytoplasmic localization, arguing against localization-driven specificity, although we cannot rule out the possibility that localization to specific sub-cellular compartments is masked by protein overexpression.
Apart from the finding that MTM-6 and MTM-9 (MTMR6 and MTMR9 in mammals) form a heterodimer (Mochizuki and Majerus, 2003; Dang et al, 2004), little is known about their protein interaction partners. MTMR6 can interact with and modulate the activity of the calcium-activated potassium channel KCa3.1 (Srivastava et al, 2005). This interaction is important for the regulation of Ca2+ signalling during T cell activation (Srivastava et al, 2006). Although C. elegans MTM-6 interacts with the KCa3.1 orthologue KCNL-2 in a yeast two-hybrid assay (Srivastava et al, 2005), the physiological relevance of this interaction in C. elegans is not clear. Knockdown of kcnl-2 did not produce any defect in QL.d migration in wild-type animals, nor did it enhance the penetrance of the phenotype in mtm-6 mutants (data not shown), suggesting that this interaction is unlikely to be relevant for the role of MTM-6 in Wnt secretion.
The major defect we observe in mtm-6 mutants is a reduction in MIG-14/Wls protein levels, a phenotype reminiscent of the effect of loss of retromer and thus the absence of MIG-14/Wls recycling. How could an excess of PI3P lead to a decrease in levels of MIG-14? We show that SNX-3, a member of the sorting nexin family of PI3P-binding proteins, accumulates on vesicles in mtm-6 mutants. SNX-3 interacts with the retromer complex to mediate endosome-to-Golgi transport of MIG-14/Wls (M Harterink et al, in preparation). Although the exact mechanism remains to be determined, we speculate that a tight balance of SNX-3 recruitment and release is necessary for the efficient retromer-dependent recycling of MIG-14/Wls (Supplementary Figure 5). In the absence of MTM-6, SNX-3 accumulates on endosomes and the efficiency of the retromer-dependent retrieval of MIG-14/Wls is reduced, which ultimately leads to its degradation. This model is consistent with our observation that Wnt signalling in mtm-6 mutants can be restored by MIG-14/Wls overexpression.
Although mutation of mig-14/Wls affects all Wnt-dependent processes in C. elegans (Thorpe et al, 1997; Pan et al, 2008; Yang et al, 2008), loss of mtm-6 results in a more restricted spectrum of Wnt phenotypes. This is similar to the phenotype observed in mutants of the cargo-selective retromer subunits, which affect some, but not all Wnt-dependent processes (Coudreuse et al, 2006; Prasad and Clark, 2006). We have previously shown that blocking MIG-14/Wls recycling results in a reduction, but not a full block of Wnt secretion (Yang et al, 2008). Interfering with MIG-14/Wls recycling will therefore mainly affect Wnt-dependent processes that depend on a high level of Wnt secretion. Consistently, there is a large degree of overlap in the phenotype of mtm-6 and mutants of the cargo-selective retromer subunits (Coudreuse et al, 2006; Prasad and Clark, 2006). The Wnt-signalling phenotype of mtm-6 mutants is, however, weaker than observed in retromer mutants, which is consistent with the less severe reduction in steady state MIG-14/Wls protein levels. A likely explanation of this difference is that loss of mtm-6 only partially interferes with the retromer-dependent recycling of MIG-14/Wls; a conclusion that is supported by our observation that the Wnt-signalling defect of mtm-6 can be strongly enhanced by partial knockdown of the cargo-selective retromer subunits (data not shown).
We provide evidence that the role of mtm-6 in MIG-14/Wls recycling is evolutionarily conserved. Experiments in Drosophila confirm that Dmtm6 is required to maintain Wls levels in the wing disc. Interestingly, unlike in Drosophila retromer mutants, knockdown of Dmtm6 leads to a drop in Wls levels in the whole disc, not only in the stripe of Wnt-producing cells (Belenkaya et al, 2008; Franch-Marro et al, 2008; Port et al, 2008). So far we do not have an explanation for this observation. A recent study has found that in the Drosophila neuromuscular junction, Wls has an unanticipated role in Wnt-receiving cells (Korkut et al, 2009). A possibility is therefore that the trafficking and function of Wls in the Wnt-receiving cells is different from that in Wnt-producing cells, and whereas retromer is required only in the producing cells, Dmtm6 might be required in both.
Our results suggest that the MTM-6/9 myotubularin complex is particularly important for MIG-14/Wls recycling and Wnt secretion. Apart from the coelomocyte uptake defect, mtm-6 mutants do not show obvious defects in other signalling pathways. Furthermore, we found that the secretion of Hh is not affected by Dmtm6 knockdown in Drosophila. As loss of retromer function also mainly affects Wnt signalling (Port et al, 2008; Yang et al, 2008), these results further support our hypothesis that MTM-6 and the retromer complex function together in the MIG-14/Wls recycling pathway. The specificity of the mtm-6 and retromer mutant phenotype for Wnt signalling suggests that other developmentally important signalling pathways are less dependent on retromer-dependent endosome-to-Golgi transport than the Wnt secretion pathway.
In humans, members of the myotubularin family have been implicated in several disorders, such as myotubular myopathy (Laporte et al, 1996), peripheral neuropathies (Bolino et al, 2000; Azzedine et al, 2003; Senderek et al, 2003) and azoospermia (Firestein et al, 2002). A common mechanism underlying these pathologies is defective membrane transport and homeostasis (for review, see Nicot and Laporte, 2008). However, the exact mechanism through which these myotubularin family members cause the different pathologies is not understood. We provide a new link between myotubularin phosphatases and the Wnt-signalling pathway, raising the interesting possibility that some of the pathologies associated with myotubularin mutations may be caused by defects in this essential signalling pathway.
Molecular biology, germline transformation and RNA interference
C. elegans cell-specific expression constructs and transcriptional reporters were generated using standard molecular biology techniques and vectors provided by Dr A Fire. cDNA of the mtm-6 A isoform was amplified by PCR from Bristol N2 mixed stage cDNA. To construct Pegl-20∷mtm-6∷gfp, 4.4 kb of egl-20 promoter and the mtm-6 cDNA were cloned in frame with GFP in the pPD95.81 vector. To construct Pcdh-3∷mtm-6∷gfp, the mtm-6 cDNA was inserted behind a 6-kb fragment of the cdh-3 promoter in pJP38 (Pettitt et al, 1996). To construct Pmyo-2∷mtm-6∷gfp, the myo-2 promoter was cut from pPD118.33 and was inserted together with the mtm-6 cDNA into pPD95.81. To construct Pmtm-9∷gfp, 4-kb of mtm-9 promoter sequence was PCR-amplified and inserted into pPD95.81. To construct Pegl-20∷snx-3∷gfp, 4.4 kb of egl-20 promoter sequence was inserted together with the snx-3-coding sequence into pPD95.75. Pegl-20∷snx-3∷mcherry was derived from Pegl-20∷snx-3∷gfp by replacing the gfp sequence with the mcherry sequence. To construct Pegl-20∷gfp∷2xFYVE, the GFP∷2 × FYVE fragment was released from the 2 × T10G3.5(FYVE) construct described previously (Roggo et al, 2002; a kind gift of Dr Fritz Müller, University of Fribourg, Switzerland) and cloned behind 4.4 kb of egl-20 promoter sequence in vector pPD49.26. To generate Pegl-20∷mtm-6(C335S)∷gfp and Pegl-20∷snx-3(R70A)∷gfp, point mutations were introduced into the corresponding constructs using the QuikChange Site-Directed Mutagenesis kit (Stratagene). To generate stable extrachromosomal arrays, constructs were injected into the germline of young adults at a concentration of 1–10 ng/μl (Mello and Fire, 1995). Pmyo-2∷tdTomato was used at a concentration of 10–15 ng/μl as a co-injection marker. RNAi was performed by feeding using clones from the Ahringer or Vidal libraries (Kamath et al, 2003; Rual et al, 2004). L4 larvae were placed on RNAi plates and phenotypes were scored in their progeny after 6–10 days incubation at 15 or 20°C.
Experiments with Phsp16-2∷mig-14 were performed on synchronized L1 larvae as described (Yang et al, 2008). In short, L1 larvae were collected that had hatched within a 1-h interval. Each batch was divided into two groups, one was exposed to a 10-min heat shock at 33°C, the other was left as a control. Animals were left to develop at 20°C and the position of the QL.d was scored at the L4 or early adult stage.
Detection of mab-5∷lacZ expression was performed essentially as described previously (Fire et al, 1990). Briefly, synchronized L1 larvae were collected 4–5 h after hatching in M9 buffer, washed with distiled water and dried. Larvae were fixed with cold acetone and incubated in β-galactosidase staining solution (Fire et al, 1990) at room temperature. Larvae were counterstained with DAPI for better cell recognition.
To obtain synchronized populations of L1 larvae, populations of gravid hermaphrodites were subjected to hypochlorite treatment, embryos were allowed to hatch overnight in M9 buffer and the synchronized larvae were allowed to feed for 5–6 h. Larvae were then collected in TX-114 buffer (25 mM Tris–HCl (pH 7.5), 150 mM NaCl, 0.5 mM CaCl2, 1% Triton X-114), frozen in liquid nitrogen and subsequently ground in liquid nitrogen to break open the cuticle. Lysates were centrifuged at 20 800 g at 4°C for 30 min, mixed with Laemmli sample buffer, separated on 8% SDS–PAGE and transferred onto PVDF membrane (Bio-Rad). GFP was detected with monoclonal anti-GFP antibody (clone JL-8, Clontech) and equal loading was assessed by staining with monoclonal anti-α-tubulin antibody (clone DM1A, Sigma). Densitometric analysis was performed on scanned images using the Gels plug-in of ImageJ (NIH).
C. elegans phenotypes and microscopy
QL.d migration and polarity of the ALM and PLM neurons were assessed in L4 larvae or young adults using the mec-7∷gfp expressing transgenes muIs32 or muIs35 (Ch'ng et al, 2003). All strains were grown at 20°C, except for rme-8(b1023), which was grown at 15°C. For rme-8(b1023), newly hatched L1 larvae were shifted to the restrictive temperature of 25°C and the final positions of the QL.d was determined in L3 larvae. QL.d migration was scored as defective when the PVM daughter of QL was positioned at or anterior to the posterior edge of the vulva. The precise positions of the HSN and Q descendants were scored by DIC microscopy in synchronized L1 larvae as described (Coudreuse et al, 2006). For epi-fluorescence and DIC imaging, animals were mounted on 2% agarose pads containing 10 mM sodium azide. Images were obtained with a Zeiss Axioscop microscope equipped with a Zeiss Axiocam digital camera. For confocal microscopy, animals were immobilized by 10 mM sodium azide or 0.1% tricaine/0.01% tetramisole in M9 and observed with a Leica TCS SPE confocal microscope. For quantification of the SNX-3 vesicular localization, samples were scored blindly under the microscope and animals were assigned into one of three categories: having only cytoplasmic signal with no vesicular structures, having one to five distinguishable vesicles or having more than five clearly visible vesicles.
Drosophila stocks and immunostaining
Flies carrying an inducible Dmtm-6 RNAi transgene were obtained from the Vienna Drosophila RNAi Center (Stock 26216). The hhGal4 and apGal4 driver lines were obtained from the Bloomington Drosophila Stock Center and combined with a UAS-CD8GFP transgene to mark Gal4-expressing cells. DSnx3∷GFP-expressing flies were generated by inserting a ptub-snx3-GFP plasmid into an attB landing site at cytological position 51D using the ϕ31 integrase system. Immunostaining was performed using standard protocols. Briefly, third instar larva were dissected in ice-cold Ringers solution. Discs were fixed and permeabilized in PBS containing 4% paraformaldehyde and 0.05% Triton X-100 for 25 min at room temperature (RT). Discs were washed in PBS containing 0.05% Triton X-100 (PBT) for 1 h at RT and then incubated in primary antibody solution at 4°C overnight. Afterwards, discs were washed in PBT containing 1% goat serum for 1 h at RT and subsequently incubated in secondary antibody solution for 2 h at RT. After a final wash for 1 h in PBT, discs were mounted on cover slips using double-sided tape as a spacer to avoid compression of the discs. Antibodies were diluted in PBT. Snx3∷GFP was visualized in living discs immediately after dissection. Images were collected on a Zeiss LSM710 or Leica SP5 confocal microscope using the sequential scanning mode. Images were analysed using ImageJ (NIH).
We thank Dr Shohei Mitani (National Bioresource Project for the Nematode, Tokyo, Japan) for deletion mutants, Dr Hanna Fares for mtm-6(ar513) and mtm-6(ar515), Dr Michael Hengartner for mtm-1(op309), Dr Andrew Fire for expression vectors, Dr Fritz Müller (University of Fribourg, Switzerland) for the 2 × FYVE construct and the Caenorhabditis Genetic Center (University of Minnesota, Minneapolis) for strains. This study was supported by funds from the Dutch Cancer Society (HUBR 2008-4114), the EU FP6 Program Cells into Organs and NWO VIDI (016.076.317; to HCK), a Boehringer Ingelheim Foundation fellowship (to MH), a grant from the Czech Science foundation (P305/10/0507; to MS) and the Swiss National Science Foundation and the Forschungskredit of the University of Zürich (to FP and KB).
Author's contributions: MS, MH and HCK designed and carried out the C. elegans experiments, FP and KB designed and carried out the Drosophila experiments, MS and HCK wrote the paper.
Conflict of Interest
The authors declare that they have no conflict of interest.