Myotubularin-related Proteins 3 and 4 Interact with Polo-like Kinase 1 and Centrosomal Protein of 55 kDa to Ensure Proper Abscission*

The myotubularins are a family of phosphatases that dephosphorylate the phosphatidylinositols phosphatidylinositol-3-phosphate and phosphatidylinositol-3,5-phosphate. Several family members are mutated in disease, yet the biological functions of the majority of myotubularins remain unknown. To gain insight into the roles of the individual enzymes, we have used affinity purification coupled to mass spectrometry to identify protein–protein interactions for the myotubularins. The myotubularin interactome comprises 66 high confidence (false discovery rate ≤1%) interactions, including 18 pairwise interactions between individual myotubularins. The results reveal a number of potential signaling contexts for this family of enzymes, including an intriguing, novel role for myotubularin-related protein 3 and myotubularin-related protein 4 in the regulation of abscission, the final step of mitosis in which the membrane bridge remaining between two daughter cells is cleaved. Both depletion and overexpression of either myotubularin-related protein 3 or myotubularin-related protein 4 result in abnormal midbody morphology and cytokinesis failure. Interestingly, myotubularin-related protein 3 and myotubularin-related protein 4 do not exert their effects through lipid regulation at the midbody, but regulate abscission during early mitosis, by interacting with the mitotic kinase polo-like kinase 1, and with centrosomal protein of 55 kDa (CEP55), an important regulator of abscission. Structure-function analysis reveals that, consistent with known intramyotubularin interactions, myotubularin-related protein 3 and myotubularin-related protein 4 interact through their respective coiled coil domains. The interaction between myotubularin-related protein 3 and polo-like kinase 1 relies on the divergent, nonlipid binding Fab1, YOTB, Vac1, and EEA1 domain of myotubularin-related protein 3, and myotubularin-related protein 4 interacts with CEP55 through a short GPPXXXY motif, analogous to endosomal sorting complex required for transport-I components. Disruption of any of these interactions results in abscission failure, by disrupting the proper recruitment of CEP55, and subsequently, of endosomal sorting complex required for transport-I, to the midbody. Our data suggest that myotubularin-related protein 3 and myotubularin-related protein 4 may act as a bridge between CEP55 and polo-like kinase 1, ensuring proper CEP55 phosphorylation and regulating CEP55 recruitment to the midbody. This work provides a novel role for myotubularin-related protein 3/4 heterodimers, and highlights the temporal and spatial complexity of the regulation of cytokinesis.

To gain insight on the biological functions of myotubularin family phosphatases, we have used affinity purification coupled to mass spectrometry (AP-MS) to identify protein-protein interactions for each myotubularin. The results expand upon the known repertoire of intra-myotubularin interactions, and, critically, identify specific novel interactions for individual myotubularins, providing valuable clues toward their respective functions. Further investigation revealed an unexpected role for MTMR3 and MTMR4 in abscission (21), the fission event at the end of cytokinesis that severs the final membrane link between divided daughter cells. Future studies of additional identified protein-protein interactions will undoubtedly illuminate the cellular roles of myotubularin family phosphatases.
Cell Culture, Cell Cycle Treatments, RNAi, and Transfection-HEK293T, HeLa, and U-2 OS cells were maintained in standard media and conditions. Stable, tetracycline-inducible HEK293 T-REx and U-2 OS T-Rex cell lines were generated as previously described (22). For biochemical assays and MTMR3/4 overexpression analysis, protein expression was induced 24 h before harvest by addition of 1 g/ml tetracycline (Tet). Transient transfection of plasmid DNA was performed using JetPrime (Polyplus-Transfection, New York, NY), and cells were harvested 24 h post-transfection. To arrest cells in S phase, 1.6 g/ml aphidicolin (Sigma-Aldrich, Oakville, ON) was added to cells overnight. To arrest cells in prometaphase, aphidicolin was removed from cells with three PBS washes and cells were incubated overnight in the presence of 20 ng/ml nocodazole (Sigma-Aldrich). For certain experiments, nocodazole-arrested cells were incubated for 30 min with either 1 M BI2536 (Selleck Chemicals, Houston, TX) or 10 M RO-3306 (Sigma-Aldrich). To obtain cells later in mitosis, nocodazole-arrested cells were harvested by mitotic shake off, washed three times in PBS, and incubated in fresh media for the indicated times. For RNAi, cells were transfected by reverse transfection with siRNA (Dharmacon ON-TAR-GETplus, Lafayette, CO, supplemental Table S1) using Lipofectamine RNAiMax (Life Technologies). Cells were fed with fresh media 24 h after transfection, and harvested or fixed 72 h after RNA transfection. For rescue experiments, cells were induced 24 h after RNA transfection with 0.01 g/ml Tet, and incubated an additional 48 h before fixation.
Affinity Purification and Mass Spectrometry-For each myotubularin and FLAG-GFP control, cell pellets from two 150-mm plates were lysed in 50 mM HEPES-KOH (pH 8.0), 100 mM KCl, 2 mM EDTA, 0.1% Nonidet P-40, and 10% glycerol and affinity-purified with M2-FLAG magnetic beads (Sigma-Aldrich), followed by on-bead trypsin digest as described (22). A spray tip was formed on fused silica capillary column (0.75 m ID, 350 m OD) using a laser puller (program ϭ 4; heat ϭ 280, FIL ϭ 0, VEL ϭ 18, DEL ϭ 200). Ten centimeters (Ϯ 1 cm) of C18 reversed-phase material (Reprosil-Pur 120 C18-AQ, 3 m) was packed in the column by pressure bomb (in MeOH). The column was then pre-equilibrated in buffer A (6 l) before being connected in-line to an Eksigent NanoLC-Ultra 2D plus HPLC system (AB-SCIEX, Concord, ON) coupled to a Thermo Electron LTQ-Orbitrap Velos (Thermo Scientific, Mississauga, ON) equipped with a Proxeon Biosystems nanoelectrospray ion source (Thermo Scientific). The LTQ-Orbitrap Velos instrument under Xcalibur 2.0 was operated in the data dependent mode to automatically switch between MS and up to 10 subsequent MS/MS acquisitions. Buffer A was 100% H 2 O, 0.1% formic acid; buffer B was 100 ACN, 0.1% formic acid. The HPLC gradient program delivered an acetonitrile gradient over 125 min. For the first 20 minutes, the flow rate was 400 l/min at 2% B. The flow rate was then reduced to 200 l/min and the fraction of solvent B increased in a linear fashion to 35% until 95.5 min. Solvent B was then increased to 80% over 5 min and maintained at that level until 107 min. The mobile phase was then reduced to 2% B until the end of the run (125 min). The parameters for data depend-ent acquisition on the mass spectrometer were: one centroid MS (mass range 400 -2000) followed by MS/MS on the 10 most abundant ions. General parameters were: activation type ϭ CID, isolation width ϭ 1 m/z, normalized collision energy ϭ 35, activation Q ϭ 0.25, activation time ϭ 10 msec. For data dependent acquisition, the minimum threshold was 500, the repeat count ϭ 1, repeat duration ϭ 30 s, exclusion size list ϭ 500, exclusion duration ϭ 30 s, exclusion mass width (by mass) ϭ low 0.03, high 0.03.
Mass Spectrometry Data Extraction-RAW mass spectrometry files were converted to mzXML using ProteoWizard (3.0.4468; (23)) and analyzed using the iProphet pipeline (24) implemented within ProHits (25) as follows. The database consisted of the human and adenovirus complements of the RefSeq protein database (version 57) supplemented with "common contaminants" from the Max Planck Institute (http://maxquant.org/downloads.htm) and the Global Proteome Machine (GPM; http://www.thegpm.org/crap/index.html). The search database consisted of forward and reversed sequences (labeled "DECOY"); in total 72,226 entries were searched. The search engines used were Mascot (2.3.02; Matrix Science, Boston, MA) and Comet (2012.01 rev.3; (26), with trypsin specificity (two missed cleavages were allowed) and deamidation (NQ) and oxidation (M) as variable modifications. Charges ϩ2, ϩ3, and ϩ4 were allowed, and the parent mass tolerance was set at 12 ppm, while the fragment bin tolerance was set at 0.6 amu. The resulting Comet and Mascot search results were individually processed by PeptideProphet (27), and peptides were assembled into proteins using parsimony rules first described in ProteinProphet (28) into a final iProphet protein output using the Trans-Proteomic Pipeline (TPP; Linux version, v0.0 Development trunk rev 0, Build 201303061711). TPP options were as follows: general options are -p0.05 -x20 -PPM -d"DECOY," iProphet options are -ipPRIME and PeptideProphet options are -OpdP. All proteins with a minimal iProphet protein probability of 0.05 were parsed to the relational module of ProHits. Note that for analysis with SAINT, only proteins with iProphet protein probability Ն0.95 are considered. This corresponds to an estimated protein-level FDR of ϳ0.5%. Furthermore, we required that each protein be detected by at least two unique peptides.
Interaction Scoring for FLAG AP-MS-For each bait, a cell line was prepared for mass spectrometric analysis as described above. Biological duplicates were grown/treated/processed at different times to maximize the variability and increase the robustness in the detection of true interactors. Negative control purifications (consisting of cells expressing GFP fused to the FLAG tag) were processed in parallel periodically throughout the data acquisition phase of the project in order to mitigate "batch effect" artifacts that we and others have reported (29). The quality of each sample was assessed manually by aligning the runs for the biological replicates in ProHits: samples of low quality were discarded, and other biological replicates were acquired. The main data set acquired in HEK293 Flp-In T-REx cells, and the smaller HeLa Flp-In T-REx data sets were analyzed separately with SAINTexpress (30), a computationally efficient reimplementation of the Significance Analysis of INTeractome method described previously (31). SAINT probabilities computed independently for each bait replicate are averaged, and the average probability (AvgP) is reported as the final SAINT score. To model the interactions robustly, the 35 HEK293-specific controls were first compressed to 25 "virtual controls" in which the 25 highest counts for each prey across all controls were used (see (32) for a description of this conservative scoring approach). Preys with AvgP Ն0.9 were considered "true" interactors (estimated FDR of 1%; see (33) for the description of the FDR estimation in SAINTexpress). For the HeLa cells, the six controls were compressed to four "virtual controls," and preys with AvgP Ն0.92 were considered "true" interactors at 1% FDR. Only trypsin and keratins were manually removed from this dataset after running SAINTexpress. The complete list of the interactions with scores is available in a searchable format on our website: http://prohitsweb.lunenfeld.ca. Downloadable files and all raw mass spectrometry files are deposited in the MassIVE repository housed at the Center for Computational Mass Spectrometry at UCSD (http://proteomics. ucsd.edu/ProteoSAFe/datasets.jsp). The data set has been assigned the MassIVE ID MSV000078915 and is available for FTP download at: ftp://MSV000078915@massive.ucsd.edu. The data set was assigned the ProteomeXchange Consortium (http://proteomecentral. proteomexchange.org) identifier PXD001448. Visualization was performed using Cytoscape (34), or a custom dot plot mapping coded in R (35).
Cell Lysis, Co-Immunoprecipitation, SDS-PAGE, Western Blotting-Cells were washed with PBS and then lysed (50 mM HEPES-NaOH, pH 8.0, 100 mM KCl, 2 mM EDTA, 0.1% Nonidet P-40, 10% glycerol, 1 mM PMSF, 1 mM DTT, Sigma protease inhibitor mixture, P8340, 1:500, 40 mM ␤-glycerophosphate, 10 mM NaF, 0.3 mM sodium vanadate, and 100 nM okadiac acid) for 15 min on ice. Lysates were cleared by centrifugation for 20 min at 14,000 rpm at 4°C. Lysates treated with phosphatase (New England Biolabs, Whitby, ON) were lysed without phosphatase inhibitors, and incubated with 400 U phosphatase in 1ϫ Protein MetalloPhosphatase buffer (New England Biolabs) and 1 mM manganese chloride for 30 min at 30°C. Cell lysates were boiled in Laemmli buffer for analysis by SDS-PAGE. After lysis of cells expressing FLAG-tagged proteins as above, cleared lysates were incubated with 5 l of FLAG M2 magnetic beads (Sigma-Aldrich) with rotation for 1.5 h at 4°C. Beads were washed twice in lysis buffer, and boiled in Laemmli sample buffer. Proteins were separated by electrophoresis on Criterion TGX 4 -15% gradient gels (BioRad, Mississauga, ON), and then transferred to Protran nitrocellulose (GE Healthcare, Baie d'Urle, QC). Membranes were blocked in ddH 2 O with 5% skim milk powder and incubated with primary and secondary antibodies (supplemental Table S1) in 50:50 PBS:LiCor Blocking Buffer (LiCor Biosciences, Lincoln, NE). Blots were washed in TBST, with a final wash in TBS. Antibody signals were visualized on a LiCor imaging system (LiCor Biosciences, Lincoln NE), with the exception of co-immunoprecipitated endogenous CEP55, which was incubated with a Donkey anti-mouse TrueBlot secondary conjugated to HRP (Rockland Immunochemicals, Limerick, PA; used to block the IgG heavy chain signal), and visualized using Lumiglo ECL (Cell Signaling Technologies, Danvers, MA).
Immunofluorescence, Fixed Imaging, Quantitative Imaging-Cells were seeded onto coverslips (Electron Microscopy Services, Hatfield, PA) or custom-made teflon printed coverslips (Scientific Devices Laboratory, Des Plaines, IL), and fixed with ice-cold methanol for 20 min at Ϫ20°C. Samples were blocked in PBS with 2.5% BSA, and all antibodies (supplemental Table S1) were diluted in the same blocking buffer. Blocking, primary antibody, and secondary antibody incubations were performed for one hour at 37°C, and coverslips were washed with PBS between incubations. DNA was counterstained with DAPI (Sigma), and the coverslips were mounted on glass slides by inverting them into mounting solution (ProLong Gold Antifade, Life Technologies). The samples were allowed to cure for 24 -48 h. 3D deconvolution images were acquired on a DeltaVision Core imaging system (Applied Precision, Mississauga, ON) equipped with an IX71 microscope (Olympus, Richmond Hill, ON), CCD camera (CoolSNAP HQ2 1024 ϫ 1024, Roper Scientific, Martinsried, Germany) and ϫ60/ 1.42 NA plan-Apochromat oil-immersion and 20ϫ air objectives (Olympus). Z stacks (0.2 m apart) were collected, deconvolved using softWoRx (v5.0, Applied Precision) and shown as maximum intensity projections (pixel size 0.1075 m). Images were cropped in ImageJ (National Institutes of Health, Bethesda, MD). For all quantitatively compared images, identical imaging conditions (including exposure times) were used, and maximum intensity projections of Z-stacks were analyzed. Image quantitation was performed on unaltered projected Z-stacks using custom written MATLAB (MathWorks Inc., Natick, MA) scripts to mask the midbody regions and calculate integrated intensity for different markers.
Live Cell Imaging-HeLa cells with stable expression of either GFP-Tubulin and RFP-H2B, or GFP-CEP55 were transfected with control, MTMR3, MTMR4, or CEP55 RNA and plated in LabTek II imaging chambers (Thermo Scientific). Cells were fed with fresh media 24 h later, and arrested in G2 with 10 M RO-3066 (Sigma) overnight. GFP-CEP55 cells were also transfected with mCherry-Tubulin (a kind gift from Geoffrey Hesketh) at the time of G2 arrest. Before imaging, the cells were released into mitosis by washing three times with PBS, and fed with phenol red-free DMEM supplemented with 10% FBS, 1 mM sodium pyruvate, and 1ϫ Glutamax (all Life Technologies). Cells were imaged at 37°C in 5% CO 2 at 60ϫ magnification, with 2ϫ2 binning. Image stacks of 24 m were taken at 2 m intervals, and 40 stage positions were imaged every 5 mins for 18 h. Exposure times were 100 ms each, at 10 and 32% transmission for GFP-Tubulin/GFP-CEP55 and H2B-RFP, respectively. Images were deconvolved and processed as above.
Statistical Methods-p values were obtained from two-tailed unpaired student t-tests. In the figures, * denotes p Ͻ 0.05, ** p Ͻ 0.01, and *** p Ͻ 0.001. In all box-whisker plots in Fig. 6, the red horizontal line denotes the mean, the pink box represents the 95% confidence interval, and the grayed region represents one standard deviation of the data.

Myotubularins Are Highly Interconnected, but Have Distinct
Interaction Profiles-To shed light on possible functional roles for myotubularin family phosphatases, we performed AP-MS to identify protein-protein interactions for 15 of the myotubularins (MTMR10 was excluded because of low protein expression). We generated human embryonic kidney (HEK) 293 Flp-In T-REx cells stably expressing 3X-FLAG-tagged versions of each myotubularin, and performed FLAG immunoprecipitation on lysates from asynchronous cells. After onbead trypsin digest, peptides were sequenced on a Velos Orbitrap mass spectrometer. Using the data from two inde-pendent biological replicates for each myotubularin, compared with 35 negative controls (3X-FLAG-GFP), we generated a probability score for each protein interaction using SAINTexpress (Significance Analysis of INTeractome) (30,31). Generally, the myotubularins displayed a high degree of interconnectivity ( Fig. 1, supplemental Tables S2 and S3, http:// prohits-web.lunenfeld.ca). The majority of intra-family interactions in our dataset were between active and inactive members. The enzymes interact in several subclusters. For example, the active myotubularins MTMR1 and MTMR2, and inactive myotubularins SBF1 and SBF2, all interacted with each other, while each formed independent interactions with a variety of other proteins. The inactive enzymes MTMR9, 10, 11, and 12 sit on the periphery of this cluster, with all four forming interactions with the active phosphatase MTMR2. MTMR9 also interacted with three additional active myotubularins, MTMR6, 7, and 8, each of which displayed distinct interaction profiles. The intramyotubularin interactions were quite consistent with those previously determined by yeast two-hybrid (4) but also provided a few novel myotubularinmyotubularin interactions (supplemental Fig. S1B). Despite the interconnectivity of the majority of myotubularins, a few formed interactions outside of the main network, including the active myotubularin MTMR14 and inactive myotubularin FAN1 (MTMR15). Interestingly, MTMR14 and FAN1 are the only myotubularins that do not contain coiled coil domains, which are thought to mediate intramyotubularin interactions (36,37). Dot plot analyses of all high-confidence interactions found for each myotubularin demonstrates that, outside of intra-myotubularin interactions, there was very little overlap in proteinprotein interactions between myotubularins (supplemental Fig. S1C). This suggests that individual myotubularins have specialized functions, and the enzyme family may regulate a variety of cellular processes. MTMR3 and MTMR4 Interact with PLK1 and CEP55-Similar to MTMR14 and FAN1, MTMR3 and MTMR4 formed interactions independently of the main myotubularin network. MTMR3 and MTMR4 formed clearly detectable reciprocal interactions by AP-MS. This is consistent with a previous yeast two-hybrid study, in which MTMR3 and MTMR4 were the only active myotubularins found to directly interact (4). In addition, MTMR3 interacted with the mitotic kinase PLK1, whereas MTMR4 interacted with CEP55, an important regulator of cytokinesis, the final stage of mitosis (38) (Fig. 1). To validate the interactions, we performed reciprocal AP-MS of PLK1 and CEP55. (Fig. 2A, supplemental Tables S2 and S3). PLK1 associated with a variety of proteins, including MTMR3 (for details on additional interacting proteins, see prohitsweb.lunenfeld.ca). CEP55 interacted with several components of the ESCRT-I complex, and also interacted with MTMR4, although because of low spectral counts, this interaction was not significant. No PLK1 peptides were detected in CEP55 AP-MS, and vice versa. In addition, PLK1 peptides were only detected with MTMR3, and CEP55 peptides were only detected with MTMR4. The interactions were also observed after AP-MS of MTMR3, MTMR4, PLK1, and CEP55 in HeLa T-REx cell lines (supplemental Fig. S2A, supplemental Tables S4 and S5). To confirm the interactions between these proteins, we performed co-immunoprecipitation coupled to immunoblotting experiments. As expected, MTMR3 and MTMR4 established both homomeric and heteromeric interactions. However, in contrast to the apparent specificity of the MTMR3-PLK1 and MTMR4-CEP55 interactions in our mass spectrometry experiments, CEP55 and PLK1 recovered both myotubularins by co-immunoprecipitation/immunoblotting (Fig. 2B). The homo and heteromeric interactions and the interaction of both MTMR3 and 4 with PLK1 were also detected in U-2 OS cells (supplemental Fig. S2B; CEP55 was expressed at much lower levels in these cells, precluding conclusions as to its interactions with the myotubularins).
MTMR3 and MTMR4 Are Both Required for Cytokinesis-PLK1 is a master regulator of several stages of mitosis (27), including abscission. Chemical inhibition of PLK1 leads to cytokinesis defects, which include delays in cytokinesis and the formation of bi/multinucleated daughter cells (21). CEP55 is a crucial component of the spindle midbody, the microtubule-based structure that forms at the membrane bridge and coordinates abscission, and both depletion and overexpression of CEP55 results in abnormal midbody morphology and abscission defects, particularly the formation of bi-and multinucleated daughter cells upon mitotic exit (26). We therefore evaluated whether MTMR3 or MTMR4 are required for abscission. In both HeLa and U-2 OS cells, depletion of MTMR3, MTMR4 (supplemental Fig. S3A and S3B), or both myotubularins together led to levels of bi/multinucleated cells comparable to CEP55 depletion (Fig. 3A, supplemental Fig. S3C). Overexpression of MTMR3 or MTMR4 also resulted in significant binucleation (Fig. 3B), indicating that the levels of both myotubularins must be tightly regulated to ensure proper abscission. With the exception of small but significant increases in binucleation upon depletion of MTMR7 and MTMR9 (3.9% (p ϭ 0.0305) and 3.5% (p ϭ 0.0427), respectively), none of the other myotubularins affected binucleation (Fig. 3C). The mild phenotype of MTMR7 and MTMR9 in our assay was not because of inefficient silencing as MTMR7 and MTMR9 were depleted to comparable levels as MTMR3 and MTMR4 (supplemental Fig. S3D), and unlike MTMR3 and MTMR4, overexpression of MTMR7 nor MTMR9 resulted in increased binucleation (supplemental Fig. S3E). These experiments confirm that the observed phenotype is specific to MTMR3 and MTMR4 (with approximately double the number of binucleated cells; both p Ͻ 0.0001), and is not a general consequence of myotubularin loss (for example, through global changes in phospholipid levels). The increased number of binucleated cells upon depletion of MTMR3 or MTMR4 was rescued by expression of respective siRNA-resistant variants, indicating that the defects are not caused by off-target effects ( Fig. 3D; Note that 0.01 g/ml tetracycline was used to induce a lower amount of MTMR3 and MTMR4 in rescue experiments, as induction with 1 g/ml tetracycline resulted in an overexpression phenotype). By contrast, GFP-MTMR3 was not able to rescue the MTMR4 siRNA phenotype, or vice versa. This suggests that MTMR3 and MTMR4 functions are not interchangeable, and are both required to ensure genomic integrity in daughter cells.
Binucleation in daughter cells does not necessarily indicate problems in abscission, as issues with either chromosome segregation in anaphase or cleavage furrow ingression in telophase can also produce this phenotype (39). Therefore, we performed live cell imaging in HeLa cells expressing both GFP-Tubulin and H2B-RFP. In cells depleted of either MTMR3 or MTMR4, chromosome segregation and cleavage furrow ingression occurred normally (supplemental Fig. S3F, supplemental Movies S1-S3), but cells displayed a variety of mitotic phenotypes. In addition to forming binucleated daughter cells, many cells died, during cytokinesis, and, in the case of already-binucleated cells, during subsequent early mitosis (Fig. 3E).
The spindle midbody is a highly organized structure, and depletion of CEP55 results in abscission failure because of the formation of disorganized, nonfunctional midbodies (40). As with CEP55, depletion of MTMR3 or MTMR4 resulted in significantly longer midbody microtubule bridges (13.1 m and 13.6 m, respectively, compared with 7.94 m for control cells; Fig. 3F). Postmitosis, the abnormal membrane fusions remaining between daughter cells after MTMR3/4 depletion contained disorganized midbody components, suggesting that these bridges were the remnants of nonfunctional midbodies (Fig. 3G). The presence of these membrane fusions was rescued upon expression of siRNA-resistant MTMR3 and MTMR4 (Fig. 3H). Taken together, these observations suggest that loss of either MTMR3 or MTMR4 results in abscission defects caused by aberrant midbody formation.
The Roles of MTMR3 and MTMR4 in Abscission are Lipid Independent-Recently, it has been demonstrated that correct PI3P concentration at the midbody is crucial for proper abscission, as depletion of PI3 kinase components VPS34 or BECN1 results in abscission failure (41). However, whether PI3P phosphatases are involved in the regulation of abscission has not been investigated. MTMR3 and MTMR4 are both active enzymes, with demonstrated phosphatase activity against both PI3P and PI3,5P (42,43). As both CEP55 and PLK1 localize to the midbody (38), we postulated that interaction with PLK1 and CEP55 may serve to recruit MTMR3 and/or MTMR4 to the midbody, where they might regulate midbody PI3P concentration. Despite the ability to rescue the loss of MTMR3 or MTMR4 function in abscission (Fig. 3D), we noted that neither GFP-MTMR3 nor GFP-MTMR4 visibly colocalized with CEP55 at the spindle midbody ( Fig. 4A; note that, like others in the field (4), we have not found suitable antibodies for the detection of endogenous MTMR3 and MTMR4 by immunofluorescence). However, it is possible that a small, undetectable fraction of the proteins could localize to the midbody and affect PI3P concentration. Therefore, to further investigate this possibility, we performed rescue experiments with mutants of MTMR3 and MTMR4.
MTMR3 and MTMR4, like other myotubularins, contain both a PH-GRAM phosphatidylinositol-binding domain, and a specialized protein tyrosine phosphatase (PTP) domain that dephosphorylates phosphatidylinositols (Fig. 4B, 4C). In addition (and unique across the myotubularin family), MTMR3 and MTMR4 also contain phosphatidylinositol-binding FYVE domains. The FYVE domain of MTMR4 is responsible for its localization to endosomes in interphase cells (44). The MTMR3 FYVE domain, however, is divergent in sequence from most FYVE domains, does not bind phosphatidylinositols or localize MTMR3 to endosomes, and is therefore considered nonfunctional (45). To assess whether the lipid binding and/or phosphatase activities of MTMR3 or MTMR4 were required for their cytokinetic functions, function-disrupting point mutations were introduced in each of the conserved domains. Phosphatase-dead versions (PTPmut, C413S for MTMR3, C407S for MTMR4 (44,46)) and FYVE domain-inactive mutants (MTMR3-FYVEmut, C1174S (46), MTMR4-FYVEmut, C1169S (4)) have been previously described. Point mutations to disrupt lipid-binding through the PH-GRAM domains (PHmut, L73P for both enzymes) target a conserved residue which, when mutated, abrogates the lipid-binding of the MTM1 PH-GRAM domain (47). GFP-tagged, siRNA-resistant forms of each mutant protein were evaluated for the ability to rescue MTMR3/MTMR4 RNAi-induced binucleation in U-2 OS T-REx stable cell lines (Fig. 4B, 4C). Importantly, expression of each MTMR3/4 mutant was comparable to the wildtype proteins (Fig. 4D), and all mutants were expressed as full-length fusion proteins (supplemental Fig. S4). Phosphatase-dead versions (PTPmut) of both enzymes rescued RNAiinduced binucleation, indicating that lipid phosphatase activity is not required for proper abscission (this also rules out a hypothetical protein target for MTMR3/4 phosphatase activity). In addition, abscission occurred normally upon disruption of the PH-GRAM domain (PHmut) of either phosphatase, or the FYVE domain of MTMR4 (MTMR4-FYVEmut), indicating that lipid binding ability is also not required for abscission. Curiously, disruption of the MTMR3 FYVE domain (MTMR3-FYVEmut), which does not bind lipids (45) failed to rescue MTMR3 RNAi-induced binucleation, suggesting that it may play a (lipid-independent) role in regulating abscission. As none of the conserved domains known to bind phospholipids were required to rescue abscission defects caused by MTMR3/4 depletion, the roles of MTMR3 and MTMR4 in abscission appear to be lipid-independent.
PLK1/MTMR3/MTMR4/CEP55 Interactions Are Required for Proper Cytokinesis-In addition to various lipid-binding domains, myotubularins typically contain a coiled coil domain, which mediates the formation of myotubularin dimers (36,37). We used MultiCoil (48) to identify important hydrophobic residues in the coiled coil regions of MTMR3 and MTMR4, and mutated these to proline to disrupt coiled coil structure (MTMR3-CCmut, V1049/L1052/L1056P, MTMR4-CCmut, V1036/L1039/V1043/L1046P). Expression of MTMR4-CCmut failed to rescue the binucleation observed upon depletion (Fig. 4C), and expression of MTMR3-CCmut partially rescued the phenotype, with a 9% decrease in binucleation (p ϭ 0.0173) (Fig. 4B). Disruption of the coiled coil domain of either myotubularin disrupted the MTMR3-MTMR4 interaction (Fig.  4E). In addition, although disruption of the MTMR3 coiled coil domain did not affect the interaction between MTMR3 and PLK1, it did decrease the interaction between MTMR3 and CEP55, implying that the MTMR3-MTMR4 interaction is required for MTMR3 and CEP55 to associate. Similarly, disruption of the MTMR4 coiled coil domain affected the MTMR4-PLK1 interaction, but not the MTMR4-CEP55 interaction, suggesting that MTMR4-PLK1 interactions are dependent on the presence of MTMR3.
Because MTMR3-FYVEmut could not rescue MTMR3 RNAi-induced binucleation (Fig. 4B), we investigated whether it may also be involved in association with its binding partners. MTMR3-FYVEmut co-immunoprecipitated less PLK1 than wild-type MTMR3, but had no effect on the amount of coimmunoprecipitated CEP55 (Fig. 4F), suggesting that the role of the MTMR3 FYVE domain may be to facilitate interaction with PLK1. The interaction between CEP55 and the ESCRT-I components ALIX and TSG101 is dependent on the presence of a GPPXXXY (where X is any amino acid) motif in the ESCRT-I components (49). Interestingly, such a motif is also located at the extreme N terminus of MTMR4, but is not present in the MTMR3 N terminus (Fig. 4G) or in any of the other myotubularins. Mutating this motif (MTMR4-GPPYmut, Y11A) resulted in failure to rescue MTMR4 RNAi-induced binucleation (Fig. 4C). This mutant also failed to interact with CEP55, confirming that MTMR4 interacts with CEP55 in a manner similar to the ESCRT-I/CEP55 interactions. However, the interaction between MTMR4 and PLK1 was maintained in the presence of this mutation (Fig. 4F). Taken together, these results suggest that the PLK1-MTMR3, MTMR3-MTMR4, and MTMR4-CEP55 interactions occur through different binding interfaces, and, as disruption of any of these binding interfaces failed to rescue RNAi-induced abscission defects, we conclude that interactions between PLK1, MTMR3, MTMR4, and CEP55 are required for proper abscission.
Phosphorylated MTMR3 and MTMR4 Interact with PLK1 and CEP55 During Early Mitosis-Mutation of the coiled coil domains of either MTMR3 or MTMR4 resulted not only in loss of the MTMR3-MTMR4 interaction, but also affected interactions with PLK1 and CEP55 (Fig. 4E), implying that the MTMR3-MTMR4 heterodimer may act as a bridge to facilitate additional interactions. To investigate this possibility, we performed immunoprecipitation experiments after depletion of either MTMR3 or MTMR4, and assayed interactions with PLK1 and CEP55. Upon depletion of MTMR4, FLAG-MTMR3 copurified less CEP55, but not less PLK1, than in the presence of MTMR4 (Fig. 5A). Conversely, upon MTMR3 depletion, FLAG-MTMR4 showed decreased interaction with PLK1, but not CEP55 (Fig. 5B). Taken together with the mutant experiments in Fig. 4E and 4F, we propose associations between PLK1, MTMR3, MTMR4, and CEP55, in which MTMR3 and MTMR4 dimerize via their respective coiled coil domains, MTMR4 interacts with CEP55 through its N-terminal GPPY motif, analogous to ESCRT-I components, and MTMR3 relies on its non-lipid binding FYVE domain to interact with PLK1.
The results above suggest that interactions between MTMR3, MTMR4, PLK1, and CEP55 are required for proper cytokinesis, yet unlike CEP55 and PLK1, MTMR3, and MTMR4 do not localize to the midbody. To determine if these interactions were regulated in a cell cycle-dependent manner, cells expressing FLAG-MTMR4 were synchronized first in S phase with aphidicolin, then released into G2 and arrested in prometaphase with nocodazole. Cells were released from nocodazole arrest and harvested at intervals afterward. Synchronization was monitored by blotting for phospho-histone 3 serine 10 in the lysates; note that, as expected (50, 51), PLK1 total levels were lowest in aphidicolin-arrested cells and highest in nocodazole arrested cells. FLAG-MTMR4 was immunoprecipitated to assay interactions with MTMR3, PLK1, and CEP55. No change was observed in MTMR3-MTMR4 interaction through the cell cycle, suggesting that at least a portion of the two myotubularins may interact constitutively. However, the interactions between MTMR4 and both CEP55 and PLK1 were strongest in prometaphase lysates, and were lost as cells progressed through mitosis (Fig. 5C). Intriguingly, PLK1 has been reported to phosphorylate CEP55 during early mitosis, preventing early recruitment of CEP55 to the developing midbody (21). This suggests that the interactions reported here may play a role in regulating PLK1 phosphorylation of CEP55. We also observed that FLAG-MTMR4 undergoes a mobility shift in mitotic lysates (Fig. 5C). In addition, HA-MTMR3 undergoes a similar shift upon co-expression of PLK1 (see Fig.  2B). This may suggest that MTMR3 and MTMR4 are phosphorylated during mitosis, potentially by PLK1. To confirm that these mobility shifts are indeed the result of phosphorylation, lysates from stable HeLa lines expressing either FLAG-MTMR3 or FLAG-MTMR4 were arrested in either S phase or mitosis as above. As shown in Fig. 5D, both MTMR3 and MTMR4 displayed mobility shifts in cells arrested in prometaphase. When mitotic lysates were treated with phosphatase, the mobility shifts were lost, confirming that MTMR3 and MTMR4 are phosphorylated in early mitosis. To assess the impact of PLK1 on these phosphorylation events, mitotic cells were incubated with BI2536, an inhibitor of PLK1. Interestingly, we observed little to no change in MTMR3 or MTMR4 mobility shifts upon PLK1 inhibition (Fig. 5E). In contrast, inhibition of CDK1 decreased both mobility shifts dramati-cally, suggesting that the mitotic phosphorylation of MTMR3 and MTMR4 depends not on PLK1 activity, but on CDK1 activity.
MTMR3 and MTMR4 Regulate CEP55 Recruitment to the Midbody-CEP55 is a crucial component of the midbody, and is required for the subsequent sequential recruitment of first ESCRT-I and then ESCRT-III, which is responsible for the severing of the membrane link between the newly-formed daughter cells (52)(53)(54). Because aberrant recruitment of CEP55 to the midbody results in abscission failure (21), we investigated whether depletion of MTMR3 or MTMR4 affected the recruitment of CEP55 to the midbody. Quantitative imaging of CEP55 demonstrates that upon depletion of MTMR3, CEP55 intensity at the midbody increases ϳeightfold (p ϭ 1.6982*10 Ϫ20 ) (Fig. 6A). This increase in CEP55 intensity may correlate with early recruitment of CEP55 to the midbody during telophase in MTMR3 RNAi-treated cells (Fig. 6B), similarly to loss of CEP55 phosphorylation by PLK1 (21). To investigate this further, we performed live cell imaging in HeLa cells expressing GFP-CEP55 and mCherry-Tubulin (supplemental Fig. S5, supplemental Movies S4 -S6). Live cell imaging confirmed that depletion of MTMR3 resulted in increased CEP55 at the midbody, however, we were unable to conclusively show that this was because of early recruitment of CEP55. Intriguingly, MTMR4 depletion had the opposite effect, as CEP55 intensity at the midbody was significantly decreased compared with control cells (37% reduction; p ϭ 9.8*10 Ϫ5 ) (Fig. 6A). Importantly, changes in CEP55 intensity at the midbody were not caused by changes in total CEP55 protein level upon depletion of MTMR3 or MTMR4 (supplemental Fig. S3A).
As CEP55 is required for the recruitment of ESCRT-I components to the midbody (49), we next investigated whether ESCRT-I recruitment to the midbody was similarly affected by MTMR3/4 depletion. As expected, given the decreased amount of CEP55 present at the midbody, depletion of MTMR4 also resulted in a significant decrease (ϳ35%) in ALIX intensity at the midbody (p ϭ 0.0055) (Fig. 6C). However, despite the large increase in CEP55 intensity at the midbody after MTMR3 depletion, ALIX intensity was comparable to control cells, suggesting that loss of MTMR3 is not sufficient to deregulate ALIX levels at the midbody. To ensure that the recruitment defects observed upon MTMR3/4 depletion were caused by misregulation of CEP55, and not general defects in midbody formation, we also quantitated the midbody intensity of MKLP1, a component of the central spindalin complex, which is recruited before CEP55 (40). As expected, depletion of neither MTMR3 nor MTMR4 (nor CEP55) affected the amount of MKLP1 at the midbody (Fig. 6D).
Finally, to confirm that interactions between MTMR3, MTMR4, PLK1, and CEP55 are crucial for proper CEP55 Midbodies were imaged quantitatively, and total CEP55 intensity at the midbody was measured. For all quantitative imaging experiments, results shown depict one of three independent replicates, with at least 60 midbodies measured per condition per replicate. The red line indicates the mean, the pink box indicates the 95% confidence interval, and the gray box indicates one standard deviation. recruitment to the midbody, we examined whether expression of siRNA-resistant MTMR3/4 containing interaction-disrupting point mutants (MTMR3-CCmut, MTMR3-FYVEmut, MTMR4-CCmut, and MTMR4-GPPYmut) could rescue aberrant CEP55 recruitment to the midbody. Consistent with their inability to rescue abscission defects, expression of GFP-MTMR3-CCmut and GFP-MTMR3-FYVEmut (which cannot bind MTMR4 and PLK1, respectively) did not resolve the increased CEP55 intensity observed with MTMR3 depletion (p ϭ 0.014 and 0.0064, respectively, compared with control CEP55 intensity), whereas expression of wild-type GFP-MTMR3 returned CEP55 levels to control levels (Fig. 6E). Expression of wild-type GFP-MTMR4 returned CEP55 intensity to control levels, whereas expression of GFP-MTMR4-GPPYmut (which cannot bind CEP55) resulted in only partial rescue of the decreased CEP55 intensity observed with MTMR4 depletion (ϳ15% lower than control levels, p ϭ 0.06). On the other hand, expression of GFP-MTMR4-CCmut (which cannot bind MTMR3 or PLK1) reversed the phenotype, resulting in increased CEP55 intensity at the midbody (p ϭ 0.0068), as is observed with MTMR3 depletion and PLK1 inhibition. From this, we conclude these interactions may occur to ensure continued phosphorylation of CEP55 by PLK1 during early mitosis, leading to proper temporal recruitment of appropriate amounts of CEP55, and subsequently ESCRT components, to the midbody. DISCUSSION In this study, we have systematically identified proteinprotein interactions established by a relatively understudied family of protein phosphatases, the myotubularins. Consistent with previous studies, the myotubularins displayed extensive intra-family interactions, yet the protein-protein interactions made by each enzyme outside the family were quite divergent, suggesting that each myotubularin may have unique roles in the cell. Consistent with roles in membrane trafficking, a few myotubularins associated with transport proteins, including MTM1 with the sorting nexins SNX2 and SNX17, and MTMR7 with VPS33B, and the VPS33B-interacting protein C14orf133 (also known as VIPAS39). Several myotubularins also interacted with cellular signaling proteins, including several involved in the regulation of growth, DNA repair, and cell death. Even more striking, however, were the many interactions made with transcriptional regulators. This included interactions between the inactive myotubularin SBF1 and both components of the FACT complex (SSRP1 and SUPT16H), which acts in transcriptional elongation and, along with protein kinase CK2, regulates p53 after UV-induced DNA damage (55,56). Several other SBF1 interactors also interact with p53 and/or CK2, including WDR55 (57), TLE1 (58), and ZNHIT1 (which also significantly interacts with MTMR1) (59), suggesting that MTMR1/SBF1-containing complexes may play a role in p53-mediated transcriptional responses to DNA damage. One of the products of myotubularin dephosphorylation, PI5P, was recently demonstrated to affect transcription in Arabidopsis thaliana, by binding to the transcriptional regulator ATX1 (60), but to our knowledge this is the first indication that animal myotubularins may regulate transcription.
We chose to further investigate some intriguing interactions found between MTMR3, MTMR4, PLK1, and CEP55, and discovered important roles for MTMR3 and MTMR4 in abscission. Taken together, our data support a model in which MTMR3 and MTMR4 form interactions with CEP55 and PLK1 during early mitosis that regulate CEP55 recruitment to the midbody. It will be important to validate these interactions using the endogenous proteins, once suitable antibodies are available. It is tempting to speculate that these interactions participate in the regulation of PLK1-mediated phosphorylation of CEP55. For example, an MTMR3-MTMR4 scaffold could explain how PLK1 is recruited to CEP55. PLK1 typically forms phospho-dependent interactions with substrates via its polo box domain (61). In early mitosis, phospho-binding sites on PLK1 substrates are generally formed through CDK1 phosphorylation. Although there are two CDK1 sites just upstream of the CEP55 PLK1 site, neither site conforms to the consensus requirements for polo box domain binding (62,63). In addition, despite a documented direct kinase-substrate relationship, no study has successfully copurified CEP55 and PLK1 (supplemental Tables S2-S5, (21)). If the MTMR3/4 heterodimer acts as a scaffold, a functional polo box domain binding site in CEP55 would be unnecessary. It will be important to experimentally test this hypothesis.
Along with forming a heterodimeric scaffold with MTMR3, it appears that MTMR4 may play an additional role in CEP55 regulation. With loss of MTMR4, MTMR3 and PLK1 would not be recruited to CEP55, and therefore CEP55 should not be phosphorylated, and should display increased midbody intensity. However, depletion of MTMR4 actually decreased CEP55 recruitment to the midbody (Fig. 5C), suggesting that the effect of MTMR4 depletion is dominant over the phenotypic effects of either MTMR3 depletion or PLK1 inhibition. The exact mechanism behind this effect is unknown, but the CEP55-MTMR4 interaction does appear important for proper CEP55 recruitment. As MTMR4 interacts with CEP55 using a short peptide motif analogous to one in ESCRT-I components ALIX and TSG101 (49), it is possible that MTMR4 competes with ESCRT-I for CEP55 binding. This would explain why loss of MTMR3 results in increased CEP55 at the midbody without a corresponding increase in ESCRT-I: with MTMR3 depletion, MTMR4 would still interact with CEP55, presumably blocking increased CEP55-ESCRT-I interactions.
MTMR3 and MTMR4 may represent two novel examples of interphase membrane trafficking proteins that are co-opted for "mitotic moonlighting" functions (50). During interphase, MTMR4 is primarily localized to early endosomes, where it regulates sorting of endosomal cargo (44), while MTMR3 remains in the cytoplasm (4), and has been implicated in the regulation of autophagy (48) and cell migration (49). These functions-which unlike abscission, all involve the regulation of phosphatidylinositol levels-are presumably not required during mitosis, when endosomal fusion and recycling are reduced (51,52), autophagy is strongly inhibited (53), and the cytoskeleton is drastically reorganized, precluding migration. This raises the possibility that other myotubularins may have mitotic and/or lipid-independent functions, possibly involving a subset of the interactions identified in this study. The switch controlling MTMR3/MTMR4 function may involve phosphorylation. In this study, we have demonstrated that both myotubularins are phosphorylated during mitosis (Fig. 5D), and this phosphorylation is dependent on CDK1 activity (Fig. 5E). Accordingly, both proteins contain potential CDK1 sites (but no clear PLK1 consensus sequences, consistent with the lack of effects observed with PLK1 inhibition). Identification of these phosphorylation sites, as well as examination of their impact on both PLK1/MTMR3/MTMR4/CEP55 interactions and regulation of abscission, will be important in the future to determine if this is indeed the mechanistic switch that diverts MTMR3 and MTMR4 from their interphase roles.
Mitosis is a tightly regulated process with several distinct steps, all of which need to occur properly to ensure genomic integrity in the resultant daughter cells. Defects in abscission, the final step in mitosis, cause aneuploidy that can result in tumorigenesis (64 -66). Both PLK1 and CEP55 have transforming abilities and are overexpressed in a variety of cancers (67)(68)(69)(70), and given the importance of each protein in coordinating abscission, overexpression of either could dramatically accelerate genomic instability. It is possible cells evolved to use MTMR3 and MTMR4 as intermediates linking PLK1 to CEP55 (as opposed to direct binding between CEP55 and the PLK1 polo box domain) as a buffer, ensuring that abscission is not drastically affected by changes in PLK1 activity or CEP55 expression. If so, disruption of these interactions might be a useful therapeutic means of increasing cancer cell death, as the survival of cancer cells depends on maintaining an optimal level of genomic instability (71).