Myotubularins constitute a ubiquitous family of phosphatidylinositol (PI) 3-phosphatases implicated in several neuromuscular disorders. Myotubularin [myotubular myopathy 1 (MTM1)] PI 3-phosphatase is shown associated with early and late endosomes. Loss of endosomal phosphatidylinositol 3-phosphate [PI(3)P] upon overexpression of wild-type MTM1, but not a phosphatase-dead MTM1C375S mutant, resulted in altered early and late endosomal PI(3)P levels and rapid depletion of early endosome antigen-1. Membrane-bound MTM1 was directly complexed to the hVPS15/hVPS34 [vacuolar protein sorting (VPS)] PI 3-kinase complex with binding mediated by the WD40 domain of the hVPS15 (p150) adapter protein and independent of a GRAM-domain point mutation that blocks PI(3,5)P2 binding. The WD40 domain of hVPS15 also constitutes the binding site for Rab7 and, as shown previously, contributes to Rab5 binding. In vivo, the hVPS15/hVPS34 PI 3-kinase complex forms mutually exclusive complexes with the Rab GTPases (Rab5 or Rab7) or with MTM1, suggesting a competitive binding mechanism. Thus, the Rab GTPases together with MTM1 likely serve as molecular switches for controlling the sequential synthesis and degradation of endosomal PI(3)P. Normal levels of endosomal PI(3)P and PI(3,5)P2 are crucial for both endosomal morphology and function, suggesting that disruption of endosomal sorting and trafficking in skeletal muscle when MTM1 is mutated may be a key factor in precipitating X-linked MTM.
Myotubularins define a diverse family of lipid phosphatases that when mutated result in severe human diseases affecting different tissues such as muscle in X-linked myotubular myopathy (MTM1)(1,2) and the peripheral nervous system in Charcot-Marie-Tooth diseases type 4B1(3) and type 4B2 (2,4). The MTM family consists of 14 highly conserved dual-specificity protein tyrosine phosphatase (PTP)-like enzymes: eight are catalytically active and six contain mutations in their HCX5R PTP catalytic motif, rendering them phosphatase dead (5,6). Although homologues do not compensate for individual family members, cooperation between catalytically active and inactive myotubularins has been suggested to occur (2,7–9). In addition to a PTP domain, each family member has a Suvar 3-9, Enhancer of zeste, Trithorax-interacting domain (10). Most members also have a coiled-coil domain that mediates homodimerization and/or heterodimerization (11,12), and one or more phosphoinositide-binding domains that facilitate myotubularin membrane localization (11,13). The human myotubularin family appears to be expressed in all tissues with the exception of hMTMR7, which is only expressed in brain (14). Thus, differential expression levels, intracellular localization or activation of distinct myotubularin enzymes may account for different disease states observed when closely related myotubularin family members are mutated. Understanding how mutant myotubularins cause disease necessitates a more detailed understanding of their cellular localizations, functions and interacting partners.
Although first described as dual-specificity PTP enzymes, the primary substrates for myotubularins in yeast and mammalian cells are the phosphoinositides PI(3)P (15,16) and PI(3,5)P2(17–19). These phosphoinositide species are localized to early and late endosomes and play important roles in membrane trafficking (20,21). In particular, PI(3)P is required for the recruitment of factors containing Fab1, YOTB/ZK632.12, Vac1 and early endosome antigen-1 (EEA1) FYVE or phox (PX) domains (22–25). These factors include tethering factors (EEA1), phosphoinositide kinases (PIKfyve/Fab1p) and intralumenal vesicle forming machinery (hepatocyte growth factor-regulated tyrosine kinase substrate, Hrs/Vps27p) recruited to early and late endosomes (26,27). The binding of FYVE/PX domain proteins to PI(3)P-containing endosomes provides a platform for the recruitment and assembly of multiprotein complexes that promote endosomal tethering, docking, fusion and membrane remodeling, thereby facilitating endocytic transport and receptor sorting. This has been demonstrated most clearly by numerous studies on the sorting and degradation of epidermal growth factor receptor (EGFR), which is highly dependent on PI(3)P and PI(3,5)P2(13,28,29). Increased degradation of PI(3)P and PI(3,5)P2 by overexpression of MTM1 has been shown to cause aberrant endosomal sorting and result in a profound inhibition of epidermal growth factor (EGF)-stimulated receptor degradation (13). Thus, myotubularins play an important role in integrating cellular phosphoinositide regulation, membrane trafficking and growth control.
In mammalian cells, early and late endosomal PI(3)P are generated by the PI 3-kinase hVPS34 [vacuolar protein sorting (VPS)] (30–33), and the PI(3)P 5-kinase PIKfyve generates PI(3,5)P2 from PI(3)P on late endosomes (34). These phosphoinositide species are thought to be short lived because of rapid modification by kinases, dephosphorylation by phosphatases such as the myotubularins or degradation by lipases. In yeast, genetic studies demonstrate that PI(3)P is transported to vacuoles (the mammalian equivalent of lysosomes), where it is either converted into PI(3,5)P2 by Fab1p (35) or degraded by vacuolar hydrolases (36). In mammalian cells, conversion of PI(3)P to PI(3,5)P2 by PIKfyve on late endosomes is thought to trigger the inward invagination of intraluminal vesicles into multivesicular bodies (MVB) and to facilitate sequestration of cargo such as EGFR into MVB (13,37). Hence, balanced synthesis and consumption of PI(3)P and PI(3,5)P2 species on endosomal compartments ensures proper and efficient endocytic transport.
The recruitment and activity of the hVPS34 PI 3-kinase on early and late endosomes depends on the upstream activation of the small GTPases Rab5 and Rab7, respectively (32,38). Through their organelle-specific localization and tightly regulated nucleotide binding and hydrolysis cycle, Rab GTPases are ideally poised as pivotal regulators of vesicular trafficking (reviewed in 64). Thus, a delicate balance between Rab GTPases, lipid kinases and lipid phosphatases must exist to regulate endosomal transport. Indirect evidence exists for the co-ordinate regulation of PI(3)P levels in Saccharomyces cerevisiae and S. pombe through the actions of hVPS34 and myotubularin phosphatases (15,39). In mammalian cells, Rab5 has been reported to regulate PI(3)P levels at the plasma membrane and possibly continuing en route to early endosomes through interactions with two PI 3-kinases, as well as PI 4- and 5-phosphatases (40). Our own work showed the direct binding of Rab7 to the hVPS15/hVPS34 PI 3-kinase complex and demonstrated the importance of PI(3)P in late endocytic events (32,41). This prompted us to characterize the interfaces of the hVPS15/hVPS34 PI 3-kinase with MTM1 PI 3-phosphatase on early and late endosomes.
Here, we clearly establish that MTM1 localizes to both early and late endosomes and associates with the hVPS15/hVPS34 complex. In addition, we demonstrate that both Rab7 and MTM1 bind to hVPS15 (formerly called p150), the hVPS34 adapter molecule (42) that regulates hVPS34 activity through its interactions with Rab5 (38) and Rab7 (32,41). The WD40 domain of hVPS15 mediates binding to Rab5, Rab7 and MTM1. Consequently, binding of the hVPS15/hVPS34 complex to the Rab GTPases and MTM1 is mutually exclusive. The data suggest that tight regulation and rapid turnover of the phosphoinositide-mediated signals may be accomplished through the coupling of proteins involved in the synthesis and turnover of PI(3)P on endosomes.
MTM1 is recruited to Rab5- and Rab7-positive early and late endosomes
Overexpressed MTM1 has been observed at the plasma membrane and diffusely dispersed in the cytosol (15,16) and has been shown to diminish early endosomal PI(3)P (43). Recently, however, EGF-dependent recruitment of overexpressed MTM1 to kinetically defined late endosomes has been reported (13). As a first step toward assessing the endosomal localization and recruitment of MTM1 more precisely, endogenous and overexpressed MTM1 and a phosphatase-dead mutant of MTM1 (MTM1C375S) were localized relative to specific early and late endosomal markers.
The endosomal distribution of endogenous MTM1 was evaluated relative to markers of early and late endosomes (Figure 1A). To reveal the membrane-bound pool of MTM1, cells were saponin extracted, a method well known to preserve membrane ultrastructure while extracting cytosolic content (44,45). Endogenous MTM1 was consistently observed outlining early endosomal membranes that were positive for Rab5. The Pearson's correlation coefficient (range −1 to +1) for MTM1 and Rab5 was 0.8 ± 0.03, indicating a high degree of colocalization. The incomplete overlap between EEA1 and MTM1 may reflect localization of EEA1 to discrete early endosomal membrane domains and/or competition between the two proteins for binding to the same PI(3)P-enriched regions. Endogenous MTM1 was also observed on a subset of Rab7-positive late endosomes with a Pearson's correlation coefficient of 0.4 ± 0.1. The MTM1 staining often gave the impression of vesicle chains associated with the cytoskeleton, consistent with the known association between the endosomes and the microtubule network. Tyramide amplification was necessary to definitively identify the endosomal localization of the endogenous, membrane-bound MTM1. The peroxidase-mediated deposition of tyramide during the amplification process resulted in a strong nuclear staining that was not observed in the absence of amplification (data not shown) or when overexpressed MTM1 was immunolocalized (Figures 1B,S1). Therefore, the nuclear staining was not pursued further.
Similar to the endogenous MTM1, overexpressed wild-type (wt) MTM1 was found membrane associated and partially colocalized with early and late endosomes (Figure 1B). Because PI(3)P and PI(3,5)P2 facilitate membrane binding of MTM1 and also serve as substrates, an MTM1-phosphatase-dead mutant (MTM1C375S) was used to further evaluate localization in the absence of catalytic function (Figure 1B). The MTM1C375S was consistently and extensively colocalized with both Rab5 (Rab5wt) and Rab7 (Rab7wt). The combined data demonstrate that MTM1 is present on early and late endosomes. Furthermore, the ectopically expressed MTM1 is shown to reflect the distribution of the native protein, making it a useful tool to further dissect the function and regulation of the phosphatase.
Overexpression of MTM1 wild type (MTM1wt) or MTM1C375S was observed to cause changes in early and late endosome morphology. In particular, myotubularin wt and mutant were often colocalized with Rab5 and Rab7 on enlarged endosomes (Figure 1, inset panels). Coexpression of MTM1C375S with the activated Rab5Q79L caused extensive dilation of early endosomes that was more marked than when Rab5Q79L was expressed alone and enabled the detection of Rab5 and MTM1 in discrete puncta while on the same dilated early endosome (Figure S1). Our observations suggest that the Rab proteins and the MTM1 phosphatase are likely present in distinct domains and MTM1 impacts endosome form and function by affecting endosomal PI(3)P levels. To further address these issues, we examined the impact of MTM1 expression on PI(3)P pools and FYVE-domain protein binding. We also undertook to dissect the protein–protein interactions involved in the membrane association of MTM1.
Overexpression of wtMTM1 results in alteration of endosomal PI(3)P pools and loss of FYVE-domain protein binding
Activation of Rab5 is known to cause activation of the hVPS15/hVPS34 PI 3-kinase complex. The resulting enrichment in PI(3)P results in the recruitment of cytosolic proteins with PI(3)P-binding motifs such as EEA1 (46). Therefore, EEA1 membrane association is a useful measure of early endosomal PI(3)P levels. Overexpression of wtMTM1 in BHK cells resulted in the obvious and apparently complete depletion of EEA1 staining from early endosomes relative to adjacent untransfected cells in the same field (Figure 2A). Cells expressing the inactive MTM1C375S had no effect on EEA1 levels compared with adjacent untransfected cells in the same field (Figure 2A). The results are in agreement with those reported using COS7 cells infected with adenoviruses encoding mutant MTM1 (43). Quantification of the decreased endosomal EEA1 staining suggests a two-fold decrease in early endosomal PI(3)P pools (Figure 2C). This is a minimum estimate because the remaining diffuse EEA1 staining most likely represents a cytosolic EEA1 pool that is incompletely extracted by saponin. The MTM1 was also transfected with 2xFYVEHrs-myc as a second reporter for endosomal PI(3)P, and samples were stained for endogenous Rab7. The 2xFYVEHrs-myc staining (red) was detected on Rab7-positive late endosomes (blue) in the absence of MTM1 overexpression or in samples transfected with MTMC375S (seen as magenta in the merged image), but decreased 10-fold when wtMTM1 was overexpressed (Figure 2B, absence of any magenta and Figure 2D, Slidebook quantification of overlap). The combined data demonstrate that wtMTM1 not only localizes to endosomes but also functions to decrease early and late endosomal PI(3)P levels.
MTM1 colocalizes with and binds the hVPS34/hVPS15-kinase complex in vivo
MTM1 displays enzymatic activity for PI(3)P and PI(3,5)P2, phosphoinositide species that are localized to early and late endosomal compartments, and is specifically recruited to membranes in response to the presence of these phosphoinositides (13). In previous work, we demonstrated that the synthesis of late endosomal PI(3)P is governed by the hVPS15/hVPS34 PI 3-kinase complex bound to Rab7 and that hVPS34-kinase activity was dependent on active Rab7 (32). The hVPS15/hVPS34 PI 3-kinase complex is also present on early endosomes where it is activated by Rab5 (31,38). Because specific membrane recruitment of lipid-binding proteins is often dependent on dual lipid–protein and protein–protein interactions, it was of interest to determine if any direct association between the MTM1 and the proteins responsible for coordinating PI(3)P synthesis on late endosomes exists.
Immunofluorescence analysis of wtMTM1 or the phosphatase-dead MTM1C375S mutant with hVPS34 revealed their colocalization on endosomes (Figure 3A). To test if MTM1 was part of the hVPS15/hVPS34 complex, coimmunoprecipitation experiments were performed. The BHK cells overexpressing FLAG-epitope (DYKDDDDK octapeptide from gene-10 product of bacteriophage T7) tagged wtMTM1 or MTM1C375S with hVPS34 were lysed and MTM1 proteins were immunoprecipitated. Coimmunoprecipitated hVPS34 was detected by immunoblotting (Figure 3B). The hVPS34-kinase was bound to wtMTM1 as well as MTM1CS, but the association was more pronounced with wtMTM1. Individually transfected control samples demonstrate the specificity of the coimmunoprecipitation and whole cell lysate controls confirmed uniform protein expression across samples. Analogous results were obtained in reciprocal immunoprecipitations using hVPS34 or MTM1 antibodies and radiolabeled lysates (data not shown). Furthermore, the interaction was also demonstrated between endogenous MTM1 and hVPS34 (Figure 3C). The data indicate that MTM1 associates with the hVPS34 PI 3-kinase on endosomes.
The association of hVPS34 with Rab5 and Rab7 is mediated by the hVPS15 adapter protein (31,32,38,42). Therefore, cells expressing MTM1 and hVPS34 with or without hVPS15 were used for coimmunoprecipitation experiments. Lysates were prepared from radiolabeled samples and the relative amounts of hVPS34 coimmunoprecipitated with MTM1 were determined. As shown in Figure 4A and 4B, endogenous hVPS15 is sufficient for the coimmunoprecipitation of hVPS34 with MTM1, however, the ratio of hVPS34 coimmunoprecipitated with MTM1 increased upon coexpression of exogenous hVPS15. This result suggests the importance of stoichiometry in the interaction between the hVPS15/hVPS34-kinase complex and the MTM1 and hinted that the hVPS15 adapter protein might facilitate MTM1 binding to the hVPS15/hVPS34 complex.
Deletion constructs of hVPS15 were previously used to map the binding site for Rab5 to the hVPS15-WD40 domain (38). Rab7 was also shown to bind to wthVPS15 both in vivo and in vitro(32,41). Coimmunoprecipitation experiments performed using radiolabeled cell lysates overexpressing individual hVPS15 deletion mutants and wtMTM1 revealed that the WD40 domain also plays an important role in MTM1 binding to hVPS15 (Figure 4C,D). The WD40 hVPS15 deletion mutant was consistently coprecipitated with MTM1 at 2.5- to 3-fold lower ratios than wthVPS15 or other deletion mutants (Figure 4D) despite similar levels of expression of all hVPS15 constructs (Figure 4C, lower panel). As a control for proper folding of the deletion constructs, various hVPS15 deletion mutants were tested for retention of their ability to bind to hVPS34 relative to the wthVPS15 (Figure 4E,F). The hVPS34-kinase was immunoprecipitated with a polyclonal antibody directed against hVPS34 and the coprecipitation of the hVPS15 adapter was scored by immunoblotting with an antibody directed against the V5-epitope tag. The wthVPS15 and hVPS15ΔWD40 mutant were indistinguishable in their binding to hVPS34 (0.85 and 0.90, respectively) and exhibited nearly one to one stoichiometry, confirming that the lack of binding of hVPS15ΔWD40 was not because of misfolding of the mutant protein. There were also no statistically significant differences between wthVPS15 and either of the two other deletion mutants (hVPS15ΔPKD, 0.67 and hVPS15ΔHEAT, 0.80). The combined data suggest that hVPS15 is pivotal in binding MTM1 and likely important in MTM1 recruitment to sites of hVPS34 localization. Furthermore, specifically the WD40 domain appeared important for MTM1 binding to the hVPS15. The findings raised the question as to whether or not MTM1 may compete with Rab5 and Rab7 for binding to the hVPS15/hVPS34-kinase complex.
Rab7 and MTM1 bind to the hVPS15-WD40 domain in vitro
An in vitro binding assay was used to confirm the interaction between hVPS15 and MTM1 that was observed in vivo(41). Wild-type hVPS15, hVPS15 deletion mutants and the hVPS15-WD40 domain were transcribed and translated in vitro. Binding of the 35S-labeled hVPS15 proteins to equimolar amounts of purified glutathione S-transferase (GST), GST–Rab7 or GST–MTM1 was analyzed as detailed in Materials and Methods (Figure 5A–D). Data presented are expressed as the fold increase in hVPS15 binding to specific GST fusion proteins relative to binding to GST alone (a value of one indicates no binding above that detected for GST alone). Deletion of the WD40 domain of hVPS15 reduced binding to both Rab7 (Figure 5A) and MTM1 (Figure 5B) to background levels. Deletion of either the HEAT [Huntington, Elongation Factor 3, PR65/A, TOR (HEAT)] or the PKD (protein kinase domain) domain of hVPS15 actually increased binding of mutant hVPS15 proteins above the levels observed for wthVPS15 to both Rab7 and MTM1. Both Rab7 and MTM1 bound to the WD40 domain expressed alone. The suboptimal binding of the WD40 domain to Rab7 may suggest that additional interactions between hVPS15 and Rab7 contribute to wthVPS15 binding. Alternatively, the isolated WD40 domain may adopt an altered conformation that is less effective in binding to Rab7. The results of the in vitro binding studies confirm the binding of hVPS15 to MTM1 and Rab7 observed in vivo and suggest that Rab7, MTM1 and Rab5 (38) share the hVPS15-WD40 domain as a common binding site.
Binding of Rab GTPases or MTM1 to the hVPS15/hVPS34 complex is mutually exclusive
The established interaction of Rab5 with the hVPS15/hVPS34 PI 3-kinase complex, as well as with PI 5- and PI 4-phosphatases demonstrates that phosphoinositide generation can be directly linked to phosphoinositide turnover (40). Similarly, the results presented here suggest that hVPS15/hVPS34 PI 3-kinase activity may be directly coupled to PI(3)P degradation. To test if a three-way complex between the Rab GTPases, the PI 3-kinase complex and MTM1 might exist in vivo (as opposed to mutually exclusive MTM1/hVPS15/hVPS34 and Rab/hVPS15/hVPS34 complexes), coimmunoprecipitation experiments were performed with cells simultaneously expressing Rab5 or Rab7, together with MTM1 and hVPS15. Immunoprecipitation using anti-Rab5 or anti-Rab7 antibodies demonstrated that the Rab GTPases and hVPS15 are in a complex that does not include MTM1 (Figure 6). Conversely, immunoprecipitation of MTM1 with an anti-FLAG antibody demonstrated that MTM1 bound to hVPS15 but not to Rab5 or Rab7. Because hVPS15 is virtually completely bound to hVPS34 in vivo (see Figure 4E–F), the results suggest that the PI 3-kinase complex binds to the Rab GTPases and MTM1, but not to both simultaneously.
MTM1 GRAM domain-mutant lacking PI(3,5)P2 binding is membrane localized and retains hVPS15-binding activity
The GRAM domain of MTM1 has been shown to facilitate PI(3)P- and PI(3,5)P2-dependent membrane association in liposome-binding assays and a single GRAM-domain point mutant, V49F, blocked PI(3,5)P2 binding (13). GRAM-domain-mediated binding to PI(5)P and PI(3,4,5)P3 have also been reported (11,47). The importance of GRAM-domain-mediated phosphoinositide binding in endosomal membrane localization and hVPS15 binding was therefore tested. Coimmunoprecipitation assays show hVPS15 binding of the V49F mutant was identical to that of the wthVPS15 (Figure 7A). Early and late endosomal localization of the mutant V49F protein was also unperturbed as evidenced by the depletion of EEA1 membrane localization in cells expressing MTM1V49F and the colocalization with Rab7 (Figure 7B).
Endosomal PI(3)P is a critical signaling molecule required for effective cargo sorting and transport on the endocytic pathway. Thus, PI(3)P levels must be tightly regulated. Although the mechanistic details are still incomplete, the demonstration that Rab5 interacts with PI 4- and 5-phosphatases at the cell cortex suggests coordinated control of phosphoinositide synthesis and degradation (40). In the present study, we provide new evidence that endosomal PI(3)P levels are regulated through coordinated synthesis and degradation brought about by the formation of a specific complex between the myotublarin MTM1 lipid phosphatase and the hVPS34/hVPS15 lipid kinase. Biosynthesis of endosomal PI(3)P requires activation of the Rab5 and Rab7 GTPases on early and late endosomes and their direct association with the hVPS15/hVPS34 PI 3-kinase complex (38,41). Interaction of both Rab GTPases with the lipid kinase is mediated principally through the WD40 domain of the hVPS15 adapter protein [present study and (38)]. Although Rab5 also interacts with the hVPS15 HEAT domain (38), this was not observed for Rab7. As shown here, the same WD40 domain of hVPS15 is also responsible for binding to the myotubularin (MTM1) 3-phosphatase. Consequently, the hVPS15/hVPS34 complex associates with the Rab GTPases, Rab5 and Rab7, or the MTM1 phosphatase in a mutually exclusive manner. Thus, on early and late endosomes, PI(3)P synthesis and degradation are most likely coordinately and possibly sequentially regulated. We envision that the activated Rab5 and Rab7 GTPases bind to and activate the hVPS15/hVPS34 complex leading to the production of PI(3)P. The presence of active hVPS15/hVPS34 and PI(3)P in turn serves to recruit MTM1 to turn off the PI(3)P signal and downstream effector actions. Degradation of the PI(3)P signal may also contribute to the release of the MTM1 from membranes because of decreased GRAM-dependent phosphoinositide binding.
The present data support the idea that the MTM1 phosphatase is localized to both early and late endosomes through dual binding specificities for the hVPS15/hVPS34 protein complex and the phosphoinositide lipids, PI(3)P and PI(3,5)P2. Specific lipid binding was previously shown to be mediated by the GRAM domain of MTM1 (11,13,47). Deletion of the GRAM domain or substitution of valine at position 49 for phenylalanine (V49F) resulted in a failure of stimulated membrane recruitment in response to activated EGFR (13). In the present work, we reveal for the first time that both endogenous and overexpressed MTM1 are endosome associated even under basal, unstimulated conditions. The importance of phosphoinositide binding for stable MTM1 membrane recruitment is underscored by our findings that active MTM1 depleted endosomal PI(3)P pools and exhibited reduced membrane binding relative to the inactive MTMC375S mutant. Under conditions where endosomal PI(3)P synthesis was simultaneously enhanced through coexpression of hVPS34 or activated Rab7 (not shown), membrane retention of active MTM1 was enhanced. In contrast to what was observed in EGF-stimulated cells, the V49F mutant was indistinguishable from wtMTM1 in its ability to associate with endosomes under unstimulated conditions. Most likely, this is because as shown in the present work, the V49F mutant retains its hVPS15-binding activity and therefore can still be recruited to endosomal membranes. The membrane-associated pool in unstimulated cells was not detected by Tsujita and coworkers because of the large cytosolic pool of MTM1 present in the absence of saponin pre-extraction. Yet, upon ligand stimulation, the greatly expanded PI(3,5)P2 levels appear to contribute significantly to the enhanced membrane binding of MTM1 reported by Tsujita et al. (13). Taken together, the data suggest that both phosphoinositide and protein–protein interactions are important in the endosomal membrane association of the myotubularin MTM1 phosphatase.
Accumulating evidence indicates that endosomal sorting and cargo selection entails segregation into discrete membrane domains that are marked by different Rab GTPases. For example, Rab5, Rab4 and Rab11 cooperatively mediate early endosomal sorting of recycling cargo (48). Recent real-time studies along with our previous work suggest that Rab7 is recruited to Rab5-positive endosomes and participates in the selection of cargo destined for late endosomes and lysosomes (40,49–51). When present on Rab5-positive endosomes, Rab7 is always observed in discrete patches that do not contain Rab5 (51). Because Rab7 association with early endosomes appears to be extremely transient, detection is enhanced upon disruption of microtubules or following the expression of mutant Rab5Q79L (51). Expression of Rab5Q79L also has been shown to cause the accumulation of hVPS15/hVPS34 on early endosomes, usually on budding structures where lower levels of Rab5 are observed [(38) and our own unpublished observations]. Therefore, it is notable that the association of wtMTM1 with early endosomes appeared to be transient and was enhanced by expression of the phosphatase-dead MTM1C375S mutant or Rab5Q79L. Rab7 itself is shown to be recruited to early endosomes and to bud and move away from Rab5-positive structures (52). Based on the cumulative data, Rab7 may facilitate the segregation of the hVPS15/hVPS34-kinase complex and consequently the MTM1, into discrete early endosome domains destined for transport to late endosomes. Such a mechanism may ensure that these components are present at the observed high levels on late endosomes where continued tight regulation of PI(3)P synthesis and degradation is essential. Because Rab5 interacts directly with PI 4- and 5-phosphatases (40), the data also imply that there may be a hierarchy of coordination of phosphoinositide kinases and phosphatases.
Our present studies and those of others highlight the importance of MTM1 in endosome function through its impact on PI(3)P levels. The EEA1 is recruited to membranes by active Rab5 and is sensitive to endosomal PI(3)P levels because of the specific recognition of PI(3)P by its FYVE domain. Expression of wtMTM1 was previously shown to result in dissociation of EEA1 from early endosomes and PI(3)P consumption (43,53). We have repeated this result and show that MTM1 also depletes late endosomal PI(3)P levels. Changes in endosomal PI(3)P levels most likely affect membrane recruitment not only of EEA1 but also of Hrs and other factors required for early endosome fusion (24). The resulting changes in the trafficking and fusion machinery in turn precipitate alterations in endosome morphology and function. Coexpression of MTM1C375S and Rab5Q79L resulted in massively dilated early endosomes most likely because of the enhanced fusion precipitated by the increased synthesis and failure to degrade PI(3)P precipitated under these conditions (data not shown). Tsujita et al. showed that overexpression of wtMTM1 caused the dilation of an unidentified endosomal compartment and impaired EGFR degradation (13). The dilated endosome morphology upon overexpression of MTM1 was attributed to depletion of the substrate or product of the PIKfyve PI 5-kinase by MTM, leading to impaired endosome invagination (13). Late endosomal PI(3)P is known to facilitate the recruitment of the PIKfyve PI 5-kinase and is a requisite substrate in the synthesis of PI(3,5)P2. In turn, PI(3,5)P2 is necessary for late endosomal intraluminal vesicle formation and growth factor receptor sorting (13,34). The recruitment of MTM1 provides a negative feedback mechanism at sites where PIKfyve is active, reducing the concentration of PI(3,5)P2 and generating the MTM1 allosteric activator PI(5)P on endosomal membranes (47). Thus, endosome form and function are highly dependent on PI(3)P and PI(3,5)P2 levels both of which are substrates of MTM1.
In conclusion, the present studies demonstrate for the first time an interaction between a lipid kinase complex and a lipid phosphatase, mediated by the kinase adapter hVPS15. Furthermore, it appears that a dynamic equilibrium must exist between PI 3-kinase activation, mediated by the endosomal Rab5 and Rab7 GTPases, and PI 3-kinase inactivation, mediated by association with the MTM1 lipid phosphatase, which in turn promotes endosomal phosphoinositide homeostasis.
Materials and Methods
Cells and reagents
Cell lines were purchased from the American Tissue Culture Collection (ATCC, Rockville, MD, USA). The BHK cell line BHK21 was cultured in Glasgow's minimal essential medium containing 10% FBS, L-glutamine, penicillin, streptomycin and tryptose-phosphate broth. The A431 cells were cultured in DMEM supplemented with 10% FBS, 2 mm L-glutamine, penicillin/streptomycin. For vaccinia virus infection and transfection assays, cells were plated in six-well plates or 100-mm dishes at 80–90% confluence 1 day prior to use. For transfection alone, cells were plated at 60% confluence 1 day prior to transfection. All reagents were from Sigma unless otherwise noted.
Vectors containing FLAG-tagged wtMTM1 and FLAG-tagged MTM1C375S were as described (54). The FLAG-tagged constructs were subcloned into pcDNA3.1 for use in infection/transfection experiments. N-terminally tagged MTM1wt was prepared as described (16). The pEGFP [enhanced green fluorescent protein (eGFP)] human MTM1wt and V49F mutant have been described elsewhere (13,43). The FLAG-tagged MTM1 and the eGFP-tagged MTM1 proteins localized identically in all experiments performed. The pGEM-Rab5 constructs were the generous gift of Dr Marino Zerial (Max Planck Institute for Molecular Cell Biology and Genetics, Dresden, Germany). The pGEM-2xFYVEHrs-myc vector was the generous gift of Dr Harald Stenmark (Norwegian Radium Hospital, Oslo, Norway). The hVPS15 constructs in pcDNA3.1-V5-His (Invitrogen), (hVPS15wt, hVPS15ΔHEAT, hVPS15ΔWD40, hVPS15ΔPKD and specific domain constructs) were constructed as described (38). The Rab7 and hVPS34 mammalian expression vectors were constructed as described (32,55). The N-terminal GST–Rab7 (canine) fusion construct was prepared in pGEX-5X-2 (Amersham Biosciences, Piscataway, NJ, USA) as described (41). Human MTM1 was amplified by polymerase chain reaction using 5′TCCCCCGGGACAATGGCTTCTGCATCAACTTC and 5′ATAAGAATGCGGCCGCTCAGAAGTGAGTT-TGCACATG as primers and inserted into the pGEX5x.2 vector with SmaI and NotI such that GST was fused to the N-terminus of MTM1.
Endogenous or overexpressed Rab7 was detected using a custom chicken anti-Rab7 antibody generated against a C-terminal Rab7 peptide and prepared by Aves Laboratories (Tigard, OR, USA) or a polyclonal rabbit antibody (50). Rab5 was detected using a polyclonal rabbit antibody that was the kind gift of Dr Marino Zerial (Max Planck Institute for Molecular Cell Biology and Genetics, Dresden, Germany) or a monoclonal antibody (mAb) 4F11 prepared by our laboratory (56,57). A custom polyclonal rabbit anti-hVPS34 antibody was generated against recombinant protein purified from Escherichia coli [all animal work performed by HRP, Inc. (Denver, PA, USA)] and used to monitor the endogenous and overexpressed protein as described previously (32,41). Human autoimmune serum against EEA1 was from Dr Ban-Hok Toh (Monash Medical School, Prahran, Victoria, Australia) (58). Mouse mAb 9E10 to the myc-epitope tag was purchased from Santa Cruz Biotechnologies Inc. (Santa Cruz, CA, USA). A mouse mAb directed against GST was from Santa Cruz Biotechnologies Inc. A rabbit anti-GFP antibody was from Molecular Probes Inc. (Eugene, OR, USA). Monoclonal (IG6) and polyclonal (R929) antibodies to detect human MTM1 were as described (54). Immunofluorescence and immunoprecipitation of MTM1-FLAG constructs were accomplished using the M2 clone of the monoclonal anti-FLAG antibody or a polyclonal rabbit anti-FLAG antibody purchased from Sigma (St Louis, MO, USA). A mouse mAb to the V5-epitope tag was purchased from Invitrogen Life Technologies (Carlsbad, CA, USA). Primary antibodies were detected using fluorescein isothiocyanate (FITC)-, Rhodamine- or Cy-5 (bis-OSU, N,N’-biscarboxypentyl-5,5’-disulfonatoindodicarbocyanine)-conjugated donkey secondary antibodies purchased from Jackson Immunoresearch Laboratories (West Grove, PA, USA).
Vaccinia virus infection and transient transfection of BHK cells
BHK21 cells were infected with the vaccinia virus (vTF7.3) and then transfected with plasmids containing complementary DNA under the control of the T7 promoter as previously described (55). After 5–6 h of transfection, cells were washed with PBS and collected for immunofluorescence or biochemical analysis. Cells were incubated in DMEM medium without cysteine or methionine for 30 minutes, then radiolabeled in the same medium containing 100 μCi/mL Trans 35S-label (ICN Biomedicals Inc., Costa Mesa, CA, USA) for 30 minutes and analyzed.
As observed in previous reports (15,16), visualization of membrane-bound MTM1 in cells overexpressing MTM1 that were fixed prior to permeabilization was not possible because of a significant cytosolic distribution (data not shown). Therefore, we used permeabilization with low levels of saponin prior to fixation of cells, which is known to preserve ultrstructural morphology and allow functional studies yet sufficiently extracts cytosolic proteins to allow for optimal visualization of membrane-bound pools of Rab GTPases, lipid kinases and phosphatases, among others (32,59–62). For immunofluorescence microscopy, cells were permeabilized for 5 minutes in 0.05% saponin in 80 mm PIPES buffer then fixed in 3% paraformaldehyde. Care must be taken with saponin extraction because the membrane pool of MTM1 may also be extracted by increased time or concentration of saponin treatment. Fixed samples were quenched in 50 mm NH4Cl in PBS for 10 minutes, then treated with 0.1% Triton-X-100 in PBS containing 0.2% gelatin, 0.9 mm CaCl2 and 1 mm MgCl2 for 5 minutes. Cells were incubated with primary antibodies diluted in PBS containing 0.2% gelatin, 0.9 mm CaCl2 and 1 mm MgCl2 for 30 minutes and primary antibodies were detected with fluorophore-conjugated secondary antibodies for 30 minutes. Endogenous MTM1 was detected using the Tyramide signal amplification fluorescence system according to the manufacturer's protocols (PerkinElmer, Boston, MA, USA). Coverslips were mounted with Mowiol (41), and samples were viewed on a Zeiss LSM 510 confocal microscope unless otherwise noted. All images were exported as tiff files and compiled in Adobe Photoshop. Quantification of immunofluorescence images was performed using Slidebook software (Intelligent Imaging Innovations Inc., Denver, CO, USA). Colocalization was determined by measuring the Pearson's correlation coefficient for two markers in 10 cells and determining the average and standard deviation.
Coimmunoprecipitation from radiolabeled lysates
Lysates from cells radiolabeled and overexpressing MTM1wt-FLAG or MTM1C375S-FLAG with hVPS34, and/or hVPS15wt or hVPS15 mutants were permeabilized in 0.05% saponin in 80 mm PIPES buffer for 5 minutes, washed again with PBS and lysed in ice-cold RIPA (radioimmunoprecipitation assay) buffer (150 mm NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 50 mm Tris, pH 8.0) containing protease inhibitors. Lysates were clarified by centrifugation at 20 817 g in an Eppendorf microfuge for 10 minutes at 4°C. Monoclonal anti-FLAG (clone M2) antibody was added to each lysate and samples were incubated rotating for 1.5 h at 4°C. A rabbit anti-mouse immunoglobulin G (IgG) linker antibody was added for 30 minutes prior to the addition of protein A–Sepharose to those lysate samples immunoprecipitated with mAb anti-FLAG. Samples were washed and solubilized in SDS–PAGE sample buffer. Samples were resolved on 7.5, 10, 12.5 and 4–16% gradient gels, dried and bands were visualized using a Molecular Dynamics Storm Phosphorimager (Sunnyvale, CA, USA). Quantification of coimmunoprecipitated proteins was performed using Molecular Dynamics ImageQuant software. All values represent the ratio of coimmunoprecipitated protein as a fraction of the specifically immunoprecipitated protein and average ratios and error bars denoting the standard error of the mean were generated using GraphPad Prism software (San Diego, CA, USA).
Immunoprecipitation and Western blot
Cells were plated in 60- or 100-mm dishes and infected/transfected as described above. After 6 h transfection, cells were washed once with PBS, placed on ice and lysed in ice-cold RIPA buffer containing protease inhibitors. The lysate was clarified by centrifugation for 10 minutes at 20 817 g in an Eppendorf microfuge at 4°C. The mAb directed against the FLAG epitope (Clone M2), a rabbit anti-GFP antibody, a monoclonal anti-Rab5, a polyclonal anti-hVPS34 antibody or a rabbit anti-Rab7 antibody and protein A–Sepharose were individually added to supernatants to immunoprecipitate protein complexes. Immunoprecipitated samples were washed repeatedly with RIPA buffer and solubilized in SDS–PAGE sample buffer for analysis. Proteins were resolved on 10% SDS–PAGE and transferred to nitrocellulose. Filters were blocked with 5% milk in Tris-buffered saline (TBS) prior to incubation with the primary antibody. After extensive washing in TBS containing 1% Tween-20 (TBST), blots were incubated with horseradish peroxidase-conjugated anti-rabbit or anti-mouse antibody in milk-TBST, and bands were visualized using the Pierce SuperSignal chemiluminescent detection kit according to the manufacturer's instructions (Pierce Biotechnology Inc. Rockford, IL, USA).
GST–Rab7 and GST–MTM1 fusion protein expression and purification
Recombinant Rab7 and MTM1 were expressed as a GST fusion proteins in E. coli BL21 as described (41). Briefly, cells were grown to A600 of 0.5, chilled on ice for 20 minutes and then induced with 50 μm IPTG [isopropyl β-d-1 thiogalactopyranoside overnight at room temperature (63). Bacteria were harvested and lysed in GST Lysis Buffer (25 mm Tris pH 7.4, 2 mm ethylenediaminetetraacetic acid, 137 mm NaCl, 2.6 mm KCl, 1 mm dithiothreitol, 1 mm phenylmethylsulfonyl fluoride, 10 mg/mL each chymostatin, leupeptin, antipain, pepstatin) by sonication. Samples were Triton-X-100 solubilized by addition of Triton-X-100 to a final concentration of 1% (v/v) and incubation for 30 minutes at 4°C while rotating. The GST fusion proteins were affinity purified using Glutathione–Sepharose 4B (Amersham Biosciences). Protein concentration was measured using BioRad DC Protein Assay (BioRad Laboratories Inc., Hercules, CA, USA) according to manufacturer's instruction.
Full-length (hVPS15wt) deletion mutants or isolated hVPS15 domains (38) were synthesized in vitro by TNT Quick Coupled Transcription/Translation Systems (Promega, Madison, WI, USA) as previously described (41). In vitro translated hVPS15 samples were added to equimolar GST (as a control), GTP-bound GST–Rab7 or GST–MTM1 immobilized on Glutathione–Sepharose beads and incubated at 4°C for 2 h. Beads were washed with GST Lysis Buffer, boiled in SDS/PAGE sample buffer and proteins were resolved by SDS/PAGE. Bound proteins were quantified using a Molecular Dynamics Phosphorimager.
We thank Dr Rebecca Lee and Ms Genevieve Phillips for expert technical support and equipment maintenance that allowed us to acquire the images presented in this study. We also gratefully acknowledge Melanie Lenhart and Elsa G. Romero for their expert technical assistance and all those who kindly provided plasmids and antibodies for this project. This work was generously supported by the National Science Foundation under grant numbers MCB9982161 and MCB MCB0446179 to A. W. N. This work was also supported by grants from the American Diabetes Association and National Institute of Diabetes and Digestive Diseases (DK070679) to J. M. B. and by INSERM, CNRS, Collège de France and grants from the Association Française contre les Myopathies to J. L. Partial salary support was provided to M. P. S. through a grant from the University of New Mexico (UNM) Cancer Center while a postdoctoral fellow. Images in this article were generated in the UNM Fluorescence Microscopy Facility, which received extramural support from National Center for Research Resources (P20RR11830, S10RR14668 and S10RR016918), NEM-sensitive factor (MCB9982161), National Cancer Institute (R24 CA88339) and intramural funding from the University of New Mexico Health Sciences Center and the University of New Mexico Cancer Center.