The myotubularins (MTMs) constitute a large family of phosphoinositide lipid 3-phosphatases with specificity for PtdIns3P and PtdIns (3,5)P2. Mutations in MTM proteins are associated with inherited conditions such as myotubular myopathy and Charcot–Marie–Tooth syndrome. The substrate lipids are known to be regulators of the endosomal pathway through recruitment of specific effector proteins. Hydrolysis of PtdIns (3,5)P2 provides a biosynthetic pathway to the production of PtdIns5P, which itself can allosterically activate MTMs. We review the properties of this intriguing family of proteins and discuss potential physiological functions that include regulation of the endocytic pathway.
Myotubularins are a Large Family of Phosphatases Mutated in Hereditary Conditions
The myotubularin (MTM) family consists of 14 members containing a conserved protein tyrosine phosphatase (PTP) domain (Figure 1). In eight of the family members, the active site motif CSDGWDR is absolutely conserved. The remaining six contain mutations within this sequence, which render them catalytically inactive ‘pseudophosphatases’. Collectively they constitute one of the largest PTP families in the genome.
The prototypical family member, MTM1, was identified as a gene on the X-chromosome, which when mutated leads to a condition known as myotubular myopathy (1). The defining feature of this condition is the preponderance of skeletal muscle myotubes with centrally located nuclei, which resemble foetal myotubes. Other MTMs are also associated with chronic conditions; MTMR2 and MTMR13/Sbf2 mutations lead to clinically indistinguishable forms of a neuropathy called Charcot–Marie–Tooth syndrome (types 4B1 and 4B2), which is caused by defective myelination and myelin outfoldings (2–4). MTMR2 and MTMR5/Sbf1 mutations lead to impaired spermatogenesis and azoospermia (5–7). MTM1 and MTMR2 knock-out mice have recapitulated these conditions (6,8) and have shown that the MTM1 defect corresponds to a failure to maintain the mature myotube phenotype rather than myogenesis per se.
Myotubularins are Phosphoinositide 3-Phosphatases
Following earlier attempts to identify phospho-protein substrates for MTMs (9), two papers made the key observation that the phosphoinositide lipid PtdIns3P provided a good substrate for MTM1 and MTMR3 (10,11). Shortly thereafter, Walker et al. (12) showed that PtdIns (3,5)P2, but none of the other five remaining phospho-phosphoinositide lipids, provided an equally good substrate for MTMR3, in addition to PtdIns3P. It is now widely accepted that specificity for these two lipid substrates is conserved throughout the MTM family of enzymes, based on further characterization of MTMR2 (13–15), MTM1 and MTMR6 (14). As most MTMs are ubiquitously expressed, the conservation of substrate raises the question as to whether MTMs may exhibit functional redundancy, and what may be the origins of the different clinical conditions associated with different MTMs.
Hydrolysis of PtdIns (3,5)P2 provides a potential biosynthetic pathway for the production of PtdIns5P, an enigmatic phosphoinositide with respect to its cellular functions. This was first demonstrated by ectopic expression of MTMR3 and subsequently MTM1 in Saccharomyces cerevisiae (12,14). Such yeast do not normally produce measurable amounts of PtdIns5P, but following expression of MTMR3, a clear peak corresponding to PtdIns5P is obtained through HPLC separation of the lipid headgroups. This peak is increased when the MTM-expressing yeast have been subjected to hyper-osmotic shock, because this treatment leads to elevation of the substrate lipid PtdIns (3,5)P2 (12,16). Corresponding rises in PtdIns5P levels have since been obtained following over-expression of MTM1 in mammalian cells (17). To date, this represents the only established pathway for PtdIns5P generation as no convincing demonstration of a PtdIns 5-kinase pathway has been provided. Over-expression of the inositol 5-kinase PIKfyve has been linked with increased PtdIns5P synthesis (18), but this effect could be indirect, resulting from phosphorylation of its preferred substrate PtdIns3P to generate PtdIns (3,5)P2, which can then serve as a substrate for MTMs.
MTMs and the Endocytic Pathway
PtdIns3P and PtdIns (3,5)P2 are both key regulators of the endocytic pathway. PtdIns3P is concentrated on sorting endosomes as evidenced by labelling with a 2xFYVE domain probe specific for this lipid (19). It is constitutively produced by a class III PtdIns 3-kinase enzyme (Vps34) (20). PtdIns3P levels govern the endosomal association of several key membrane-trafficking regulators (e.g. EEA1 and Hrs), containing specific lipid-binding domains such as the FYVE or PX domains [reviewed in (21)]. The PtdIns3P 5-kinase, Fab1 (PIKFYVE in mammals), which generates PtdIns (3,5)P2, has been implicated in the multivesicular body-sorting pathway in yeast (22), and several endosome-localized effector proteins of this lipid have now been identified (23–26). It might be expected then, that MTMs influence the endocytic pathway. One immediate problem is that MTMs, which have been characterized so far, do not co-localize with endosomal markers. With the caveat that these studies have exclusively relied upon over-expressed epitope-tagged markers, it is clear that the MTM proteins are often cytosolic, but may also associate with plasma membrane ruffles and with ill-defined subcellular structures that frequently present a reticular aspect (27). For MTMR3, this reticulum was shown to overlap with ER markers (12). For reasons unknown, expression of a catalytically inactive form of MTMR3 (C413S) leads to the formation of structures resembling autophagosomes and an accompanying loss of the Golgi complex (12,28).
Nevertheless, over-expression of MTM1 but not MTMR2 has been shown to deplete endosomal pools of PtdIns3P as evidenced by the elimination of labelling with the 2xFYVE probe for PtdIns3P (29). Intriguingly, Chaussade et al. (30) found that despite this depletion of endosomal PtdIns3P pools by over-expression of MTM1 in muscle cell lines, the overall cellular mass of this lipid is little altered. This suggests that whilst PtdIns3P may be concentrated on the endosomal pathway, generating affinity for the 2xFYVE probe, the bulk of the lipid lies elsewhere in the cell, presumably at lower concentrations and largely untouched by MTM1. In our hands, over-expression of several MTMs in HeLa cells has not lead to dissociation of the endogenous FYVE domain containing protein EEA1 from endosomes (OL and MC, unpublished results), which is diagnostic for depletion of PtdIns3P levels through the application of PtdIns 3-kinase inhibitors.
An expected consequence of MTM activity at endosomes would be the negative regulation of receptor downregulation, which requires both substrates. Tsujita et al. have reported that MTMR2 translocates to EGF receptor containing endosomes some time after EGF stimulation leading to a block in receptor degradation at late endosomes (31). Translocation requires an intact N-terminal PH-GRAM domain, which they interpret to reflect binding to PtdIns (3,5)P2 based on its proclivities in an in vitro binding assay and the kinetic profile of lipid accumulation following EGF stimulation.
In S. cerevisiae, the sole myotubularin, Ymr1p, plays an overlapping role in PtdIns3P regulation with members of the synaptojanin-like family of proteins, Sjl2p and Sjl3p. Only in the context of a double mutant is a phenotype apparent. Thus, a ymr1▵Sjl3▵ strain accumulates PtdIns3P on the vacuolar membrane and suffers from vacuolar sorting defects, whilst a triple deletion is lethal (32).
The Caenorhabditis elegans genome contains six MTM genes, three of which code for active enzymes. Xue et al. have shown that siRNA knockdown of ceMTM-6 can suppress the lethality of let-512 mutants, lacking a class III PtdIns 3-kinase enzyme (Vps34 homologue) that generates the MTM substrate PtdIns3P. Similarly, ceMTM-1 knockdown could correct an endocytic defect in let-512 worms (33). The authors propose that alternative PtdIns 3-kinase enzymes in worms can generate sufficient compensatory PtdIns3P in the absence of a class III enzyme, provided the pathway specific lipid phosphatase activity is knocked down. So far, the role of MTMs is consistent with negative regulation. However, a screen for mutants in C. elegans with defective fluid-phase endocytosis identified positive roles for three MTM family members on the endocytic pathway. Mutations or knockdown of MTM-3, MTM-6 and MTM-9 (an inactive enzyme) lead to reduced uptake of a secreted form of EGFP into coelomocytes (33,34). This is a particularly intriguing result, which hints at functions beyond the negative regulation of PtdIns3P levels, or alternatively, a negative influence of this lipid at some point on the endocytic pathway.
Structure of MTMs
Domain profiling of the MTM sequences revealed several other conserved features in addition to the phosphatase domain. Most notably, each MTM has a proposed PH-GRAM (Pleckstrin Homology-Glucosyl transferases, Rab-like GTPase activators and myotubularins) domain and a coiled-coil region, respectively, N- and C-terminal to the phosphatase domain (Figure 1). The GRAM domain motif, which overlaps with a PH domain, is approximately 70 amino acids and has been predicted to function in protein–protein or protein–lipid interactions (35). Most MTMs have additional domains such as PH and FYVE with the potential to bind phosphoinositide lipids, or PDZ domains, which may mediate protein–protein interactions.
A crystal structure of MTMR2 (amino acids 74–586) has provided valuable new insight (15). This clearly revealed the PH domain-like structure at the N-terminus, which had previously been proposed to be present in MTMR3 (12). The 376 residue C-terminal phosphatase domain has a similar core structure to other PTPs but is somewhat larger than expected (approximately 250 residues). The PTP domain contains a SID (SET-interacting domain) sequence that is conserved in myotubularins, and which was originally proposed to mediate protein–protein interactions (9,36). However, this domain is in fact largely buried within the PTP domain structure and may rather be required for its structural integrity. The PH-G domain is predicted to fold autonomously of the PTP domain. They are separated by a 20-residue linker sequence and form an extensive interface (1350 Å2 per domain).
The active site sequence CSDGWDR forms a loop (the ‘P’ loop) at the base of the substrate-binding pocket that is characteristic of PTPs. This pocket is significantly deeper (13 Å) and wider than other PTPs in order to accommodate the large lipid head-group. Protein tyrosine phosphatases also possess an invariant aspartic acid on a separate loop called the ‘WPD’ loop, which acts as a general acid and then a general base within the catalytic cycle (37). Alignment of structures revealed this to be missing in MTMR2 (and by extension other MTMs), suggesting a unique catalytic mechanism. Begley et al. (15) propose that the second aspartate on the P-loop most likely fulfils this function. This mechanism may also be shared with the Sac1 family of phosphoinositide phosphatases, which contain an equivalent aspartic acid within the CX5R catalytic motif.
Owing to the high degree of sequence homology, disease causing missense mutations of both MTM1 and MTMR2 can be mapped on the MTMR2 structure (15). Most mutations are poorly accessible to solvent and are predicted to disrupt the core protein structure. As might be expected several solvent accessible mutations cluster around the active site of the enzyme. Intriguingly some mutations also occur in the PH domain, for which it is difficult to envision a mechanism by which they might affect basal enzyme activity directly.
PtdIns5P is an Allosteric Regulator of MTMs
In following up deviations of MTM activity from Michaelis–Menten behaviour in an in vitro assay, Schaletzky et al. tested each of the five nonsubstrate phosphoinositides for regulation of MTM activity. Uniquely, PtdIns5P was found to be a potent activator of MTM1 and MTMR3 activity towards both PtdIns3P and PtdIns (3,5)P2 substrates in an in vitro assay, using either short-chain water soluble lipids or lipids embedded in a liposome bilayer matrix (14). As PtdIns5P is the product of PtdIns (3,5)P2 hydrolysis, it has been proposed that it may function as an allosteric activator component of a positive feedback loop (14,38). Where then might the allosteric binding site be located?
Phosphoinositide-Binding Properties of the PH-G Domain
There are several hundred PH domain, containing proteins in the genome, many of which display phosphoinositide-binding activity. A wide spectrum of specificities for different phosphoinositides has been reported by surveys of representative proteins (39). Schaletzky et al. made a mutation in the MTM1 PH-G domain (R69C), corresponding to a missense mutation that results in a moderately severe disease phenotype. The enzyme has a moderately reduced basal activity and a significantly dampened response to PtdIns5P stimulation (14). The authors propose that the PH-G domain may constitute the site of allosteric activation by PtdIns5P and that reduced responsiveness to this regulation may underlie disease-causing mutations within the PH-G domain.
Several groups have now directly tested binding of MTM PH-G domains with phosphoinositides. Conflicting results may reflect differences in the methodologies adopted. In particular, a simple reliance on lipid-blotting technique is known to risk misassignments. Berger et al. report a preference for PtdIns5P and PtdIns (3,5)P2 using this technique. They further argue that the PH-G domain of MTMR2 recognizes PtdIns (3,5)P2 based on their observation that MTMR2 translocates to vacuolar structures when COS cells are subjected to hypo-osmotic stress and that this redistribution of MTMR2 is dependent upon an intact PH-G domain (40). Tsujita et al. (31) have used an enzyme-linked immunosorbent based assay to again suggest MTM1 specificity for PtdIns (3,5)P2. They propose that this interaction leads to translocation of MTM1 to late endosomes following EGF stimulation.
We have recently examined the phosphoinositide-binding properties of the MTMR3 PH-G domain (28). Lipid-blotting assays indicated binding to all three monophosphoinositides as well as PtdIns (3,5)P2. An alternative assay, monitoring the binding of a tandem repeat of the MTMR3-PHG domain to unilamellar liposomes, showed similar promiscuity, when the phosphoinositide lipids were incorporated at levels ≥1 mol%. However, by titrating down this mole fraction, PtdIns5P emerged as the preferred binding partner. At 0.25 mol%, only PtdIns5P-containing liposomes bound to the MTMR3-PHG domain. Further evidence that PtdIns5P can be used to confer a specific subcellular localization on MTMR3 was provided by the observation that MTMR3 translocates to the plasma membrane following artificial generation of PtdIns5P from plasma membrane-localized PtdIns (4,5)P2, by expression of the Shigella PtdIns (4,5)P2 4-phosphatase IpgD (28,41). This translocation is lost upon partial deletion of the PH-G domain.
PtdIns5P is a minor component of the cellular mass of phosphoinositide lipids, about which little is known (42). It has been proposed to play a role in regulation of Akt signalling (43) and of nuclear responses to DNA damage (44). Activation of MTMs now provides the most clearly resolved effector function of this lipid, which may be coupled to correct microlocalization of the enzyme. Our data support the notion that the PH-G domain provides a plausible allosteric-binding site for the activation of MTMR3 by PtdIns5P. Certainly, this domain is required for enzymatic activity, as its partial deletion renders the enzyme inactive (28).
Oligomerization of MTM Proteins
Several studies have described heteromeric interactions between MTM family members, following identification of co-immunoprecipitating proteins. To date, these have paired an ‘active’ family member with a ‘pseudophosphatase’ MTM. Thus, MTM1 and MTMR2 interact with MTMR12/3-PAP (36), MTMR6 and MTMR7 with MTMR9 (45), whilst MTMR2 also interacts with MTMR5/Sbf1 and MTMR13/Sbf2 (46,47). All of these associations are mediated by interaction between coiled-coil domains. MTMR2 has also been shown to self-associate through its coiled-coil domain. Sedimentation, cross-linking and FRET analysis have revealed that MTMR2 forms parallel dimers (40,46).
Hetero-oligomerization can directly enhance enzymatic activity. Kim et al. (45) used an in vitro assay with purified components to obtain three- to fourfold increases in enzymatic activity of MTMR2 due to MTMR5/Sbf1 (46); similarly, MTMR9 enhances the activity of MTMR7. In each case, further regulation may be derived through re-directing the subcellular localization of the active enzyme. This has been reported for MTMR2/MTMR5 and MTM1/MTMR12 but not seen with the MTMR7/MTMR9 combination. Co-expression of FLAG-MTMR5 redistributes EGFP-MTMR2 towards a perinuclear distribution when compared with a coiled-coil deletion mutant of FLAG-MTMR5 which does not interact (46). The fact that mutations in either MTMR2 or MTMR13/Sbf2 leads to clinically indistinguishable forms of Charcot–Marie–Tooth syndrome suggests that the interaction between these two proteins is of physiological significance (47). Along similar lines, mutations in either of the established binding partners MTMR2 and MTMR5 lead to defective spermatogenesis (5,6). On the other hand, no mutations in the coiled-coil domain of MTM1 have so far been associated with myotubular myopathy.
Schaletzky et al. (14) noticed a highly nonlinear dependence of the initial rate of reaction upon enzyme concentration (over a particular concentration range), when purified His6-MTM1 was incubated with substrate lipid. This indicates a high degree of co-operativity. Furthermore, supplementation with catalytically inactive MTM1 (C375S) could also stimulate the reaction. When MTM1 (C375S) was incubated with substrate (PtdIns3P) and allosteric activator (PtdIns5P), they could show by negative stain electron microscopy that the protein assembles into heptameric ring structures of about 12.5-nm diameter. They have proposed a model, in which PtdIns5P promotes heptameric ring formation of substrate-bound MTM1, necessary for a highly co-operative hydrolysis of substrate, coupled to disassembly of the ring – reminiscent of the hydrolysis of GTP by the GTPase dynamin (14,48). As yet, no information has been procured on the structural requirements for heptameric ring formation.
Alternative Cellular Functions of MTMs
Whilst substrate lipids are concentrated on endosomes, most MTMs are not found on endosomal structures. Accumulation of particular phosphoinositides may confer or maintain the identity of discrete organelles. Hence, one role of myotubularins may be to suppress concentration of substrate lipids at inappropriate compartments, for example, the cell may need to limit PtdIns3P accumulation on the secretory pathway to avoid recruitment of endosomal proteins. We have recently observed that a small fraction of MTMR3 associates with the Golgi complex and that this association can be stabilized through combined mutations in the active site and PH-G domain (28). The subcellular distribution of PtdIns5P is unknown. However, one hint that there may be a Golgi pool comes from the characterization of a Golgi-localized inositide phosphatase, PLIP, which favours PtdIns5P as a substrate in vitro (49).
Recent studies using an siRNA library targeting the entire complement of cellular phosphatases have revealed an unexpected role for several MTMs (MTMR1, MTMR6, MTMR7, MTMR8 and MTMR12) in promoting cell survival (50) that may be consistent with an idea that PtdIns5P levels may regulate a PtdIns (3,4,5)P3 specific 5-phosphatase, which in turn controls PKB/Akt activation (43). Srivastatava et al. have identified a Ca2+-activated K+ channel (KCa3.1) as a binding partner for MTMR6, for which negative regulation of channel activity requires an intact phosphatase domain (51). Loss of MTMR2 may lead to defective adherens junctions between Sertoli cells and developing spermatocytes (6). We and others have also noted the localization of MTMs to sites of membrane ruffling (28,52) and suspect that this may be a component of the complex interplay between phosphoinositide metabolism and cytoskeletal dynamics.
We thank the Wellcome Trust for financial support.