Phosphoinositide Regulation of Integrin Trafficking Required for Muscle Attachment and Maintenance

Muscles must maintain cell compartmentalization when remodeled during development and use. How spatially restricted adhesions are regulated with muscle remodeling is largely unexplored. We show that the myotubularin (mtm) phosphoinositide phosphatase is required for integrin-mediated myofiber attachments in Drosophila melanogaster, and that mtm-depleted myofibers exhibit hallmarks of human XLMTM myopathy. Depletion of mtm leads to increased integrin turnover at the sarcolemma and an accumulation of integrin with PI(3)P on endosomal-related membrane inclusions, indicating a role for Mtm phosphatase activity in endocytic trafficking. The depletion of Class II, but not Class III, PI3-kinase rescued mtm-dependent defects, identifying an important pathway that regulates integrin recycling. Importantly, similar integrin localization defects found in human XLMTM myofibers signify conserved MTM1 function in muscle membrane trafficking. Our results indicate that regulation of distinct phosphoinositide pools plays a central role in maintaining cell compartmentalization and attachments during muscle remodeling, and they suggest involvement of Class II PI3-kinase in MTM-related disease.


Introduction
Myofibers are large, highly differentiated contractile cells that rely on strong extracellular attachments to preserve their integrity during force-generating muscle contractions. Myofiber attachments are mediated by integrin adhesion complexes (IACs) composed of aand btransmembrane heterodimers that associate with cytoskeletal bridging factors, similar to those found in nonmuscle cells [1]. IACs are crucial at myotendinous junctions (MTJs), attaching the ends of myofibers to tendons. In addition, IACs concentrated at costameres associated with repeating sarcomeric Z-lines attach peripheral myofibrils to the extracellular matrix. IACs are known to be essential for invertebrate and vertebrate muscle cell attachments and organization [2,3], but it is unclear how the critical pattern of spatially restricted adhesions is continuously maintained.
In non-muscle cells, integrin turnover through endocytic recycling has clear roles in localization of dynamic adhesion complexes that mediate cell migration and membrane remodeling in cytokinesis. Trafficking pathways that engage specific endocytic adaptors, protein kinases and Rab GTPases for internalization and recycling of specific integrins are emerging, as primarily understood in isolated cells [4]. In contrast, it is not clear how important regulated integrin turnover is in differentiated muscle, or how this turnover is regulated. In isolated myofibers, uptake of markers for endocytic recycling occurred in the vicinity of adhesion sites and trafficked to perinuclear compartments, distinct from a degradative pathway [5], suggesting common trafficking themes shared with non-muscle cells. Experiments using fluorescence recovery after photobleaching (FRAP) in intact flies recently provided the first observation of endocytosis-dependent, growth-regulated mobility of IAC proteins at MTJs [6], underscoring the significance of regulated endosomal integrin trafficking in muscles, as well.
Dynamic membrane compartment identity and functions are in part conveyed through phosphoinositides. Phosphoinositides exist as seven phosphorylated phosphatidylinositol forms interconverted by dedicated lipid kinases and phosphatases [7]. Different phosphoinositide forms can recruit specific binding proteins to distinct membranes in order to elicit spatiotemporal responses that include localized signaling, cytoskeletal reorganization, membrane deformation and trafficking. However, the complex cellular relationships in vivo are less defined.
Here we show that mtm (GenBank NM_078765), encoding the sole D. melanogaster homolog of human MTM1/MTMR2, acts with Class II Pi3K68D (GenBank NM_079304) to maintain attachments upon myofiber remodeling. We found that mtm controls bintegrin turnover and trafficking from perinuclear compartments to maintain spatially restricted adhesions at MTJs and costameres, reflecting a broad mtm requirement for integrin-mediated adhesion also needed in the wing. The defects discovered in flies were substantiated by observing similar integrin mislocalization in human XLMTM myopathy, suggesting a conserved MTM1 function in membrane trafficking and roles for integrin adhesions in maintenance of myofiber organization. Altogether, our results identify specific phosphoinositide regulation important for endocytic recycling and dynamic control of cell compartmentalization.

Results
Mtm is required to maintain myofibers during remodeling and for adult muscle function Given the role for myotubularins in human myopathy, and our discovery of an mtm requirement in muscle essential for fly viability [18], we investigated the contribution of mtm-dependent phosphoinositide regulation to muscle cell function and compartmentalization.
Loss of mtm function using either null alleles or muscle-directed RNAi had no visible effects on muscle in larvae, which remained mobile and exhibited normal body wall muscle formation, attachments and growth ( Figure 1A and 1B-1B9, Figure S1A-S1C9). However, targeted RNAi depletion revealed muscle requirements for mtm at later developmental stages. Muscle-specific mtm knockdown, as indicated by protein depletion, showed either animal lethality (24B-GAL4) or developmental delay (DMef2-GAL4) around the stage of adult eclosion that was rescued by co-expression of either wildtype mtm or human MTMR2 ( Figure 1A and 1C, Figure S1B).
Metamorphosis occurs inside a rigid pupal case that adult flies escape at eclosion with the help of muscle contractions, including supporting contractions from a subset of abdominal persistent larval muscles (PLMs) [24,25]. The PLMs called dorsal temporary internal oblique muscles (IOMs) are large, individual, multinuceated myofibers that span abdominal segments ( Figure 1D; Figure S1D). Consistent with defects in eclosion, there was a decrease in the number of both dorsal IOMs and ventral PLMs in mtm-depleted abdomens (Figure 1D9-1E; Figure S1D9-S1E). The remaining mtm-depleted myofibers were frequently detached and seen as rounded-up balls or as elongated fibers with one completely detached end ( Figure 1D9, 1F; Figure S1D9, S1F), never observed in controls.
To explore the developmental requirement for an mtm muscle function, we first characterized myofibers using timelapse microscopy in intact animals. With mtm depletion, GFP-labeled IOMs were properly maintained during early pupal stages, when other larval muscles undergo developmentally regulated cell death ( Figure 1A,1G-1G9, 2 days after pupal formation, APF). The mtmdepleted IOMs subsequently underwent normal myofiber thinning (3d APF) and rethickening (4d APF), indicative of developmental turnover and rebuilding of the contractile myofibrils [25]. While no detachment was observed in control animals (n = 19), IOM detachment occurred during remodeling in late pupal stages with mtm knockdown (n = 15) ( Figure 1G9, 4d APF). Thus, mtm is not essential for IOM formation or survival, but is important for muscle attachments and maintenance upon remodeling.
To address whether mtm plays a role in other muscles, we examined different developmental stages and myofiber types. Although adult somatic muscles appeared to form normally with muscle-specific mtm knockdown ( Figure 1D9 and not shown), 100.060.0% of the viable adult flies were flightless (versus 22.265.4% control; n = 10, $124 flies). Visceral muscles that normally migrate to ensheath the testis [26] were present but also disrupted with mtm muscle-depletion (Figure S1G-S1G9, 24B-GAL4). Taken together, mtm function appears dispensable for myogenesis, but is broadly required in both somatic and visceral muscles for myofiber remodeling, maintenance and function.

Mtm disruption models centronuclear myopathy, including T-tubule disorganization
Pathological hallmarks of XLMTM are small, rounded myofibers with nuclei displacement and disorganization of the perinuclear compartment [8]. In wildtype IOMs, myofibrils are normally tightly packed around centrally aligned nuclei following myofiber remodeling [25] ( Figure 1H). In contrast, in mtmdepleted IOMs, central myofibrils were misaligned or absent around a normal number of centrally-displaced nuclei (2.1-fold increased nuclei distance from midline; Figure 1H9-1J). The nuclei were otherwise normal in size and morphology ( Figure S2A-S2A9) and pharate adult IOMs were impermeable to propidium iodide staining ( Figure S2B-S2C9), while ultrastructural analysis confirmed normal mitochondrial integrity ( Figure S2D-S2D9), all indicating viability of mtm-depleted IOM cells. The peripheral

Author Summary
Muscles require strong extracellular attachments to preserve cellular integrity during force-generating contractions. Integrin transmembrane receptors mediate muscle attachments at highly localized sites, but how this pattern of attachments is continuously maintained with muscle use is not understood. Human X-linked myotubular myopathy (XLMTM), a frequently fatal muscle disease, is associated with mutations in the MTM1 lipid regulator. Myotubularin (MTM) lipid phosphatases are implicated in endocytosis, a process of cellular uptake that can traffic transmembrane receptors for redelivery to the plasma membrane or to protein destruction. Here, we address MTM roles in muscle, using the genetically tractable fruit fly for detailed investigation of muscle cellular organization and functions. We show that fly muscle cells depleted for mtm function exhibit hallmarks of human XLMTM. We found that mtm regulates integrin localization through endocytosis and, in this role, is needed to maintain muscle attachments. Co-depletion of Class II PI3-kinase with mtm restores normal integrin localization at muscle attachment sites and fly survival, identifying a potential therapy target in MTM-related disease. Importantly, we show that integrin localization is also disrupted in human XLMTM. Our work shows conservation of MTM function in integrin trafficking and reveals insights into regulation of muscle cell maintenance and human disease.  Figure S2E-S2H), suggesting that mtm is unlikely to function directly in sarcomere assembly.
We also found that transverse (T)-tubules were disrupted in mtmdepleted myofibers, consistent with defects recently described in vertebrate XLMTM [14,27]. T-tubules are an extensive membrane network, continuous with the sarcolemma, which mediates excitation-contraction coupling throughout the myofiber interior. Although critical for force-generating contractions, there is little understanding of T-tubule biogenesis and structural regulation. We found that both the Amphiphysin (Amph) BAR-domain protein and Dlg1 membrane-associated guanylate kinase scaffold protein localize to T-tubules in wildtype abdominal myofibers ( Figure S3A, S3B, S3C), as in flight muscles [28]. In mtm-depleted IOMs, although longitudinal elements of T-tubule membranes were present, lack of Amph and Dlg indicated that transversal membranes were specifically disorganized or absent (Figure S3A9, S3B9; 9.5% control versus 96.3% mtm RNAi with transversal membranes in ,half of IOM; n$21). These conclusions were confirmed by transmission electron microscopy ( Figure S3D-S3D9). Altogether, the conserved mutant phenotypes and timing of onset suggests that mtm-depleted muscles in flies model hallmarks of XLMTM.

Mtm is required for bPS-integrin trafficking for adhesions in muscle and wing
Given the muscle detachment and myofibril misalignment observed in mtm mutant myofibers, we considered a possible defect in IACs at MTJs and costameres (Figure 2A-2C). We found that bPS-integrin, the single D. melanogaster b-integrin subunit encoded by mys (GenBank NM_080054), was dramatically mislocalized in mtm-depleted muscles ( Figure 2B9). In contrast to wildtype muscle, bPS-integrin was absent at the ends of detached myofibers ( Figure 2B0) and from costameres ( Figure 2C9), consistent with detachment due to disruption of integrin adhesions. Although an intracellular pool of bPS-integrin protein was detected as small punctae within wildtype myofibers ( Figure 2D), upon mtm knockdown, bPS-integrin became enriched along abnormal vacuolar inclusions within the myofiber center ( Figure 2D9). Ultrastructural analyses revealed large, lucent membrane-bound compartments within the central regions of the mtm-depleted but not control myofibers ( Figure 2E-2E0). Other proteins of the integrin adhesion complex, aPS2-integrin and Talin, were both detected at MTJs but not at the inclusions, suggesting bPS-integrin as a primary target of mtm function ( Figure S4A-S4B9).
To address the possible relationship between the appearance of membrane inclusions and muscle detachment, we examined bPSintegrin localization at earlier developmental stages. In mtm null or RNAi depleted larval muscles, normal integrin localization was detected at myofiber attachments, without any bPS-integrincontaining central inclusions ( Figure S4C, S4C9). This indicates that appearance of inclusions coincides with detachment, and that mtm function is not needed for initial IAC formation. Although detected at the larval myofiber surface, bPS-integrin was not organized into uniform striations in either wildtype or mutant myofibers, as compared to the costameres observed in wildtype pharate adult IOMs. To address whether mtm function affects integrin trafficking prior to myofiber remodeling and detachment, we performed FRAP analysis of bPS-integrin:YFP along MTJs in intact larvae [6]. The mobile fraction of bPS-integrin:YFP was significantly increased in mtm mutant larval muscles ( Figure 2F), indicating that mtm is required to stabilize sarcolemmal bPSintegrin localization, preceding myofiber remodeling.
To explore a basis for the sensitivity to mtm loss of function in pupal stages, we investigated integrin localization during IOM remodeling in metamorphosis. In wildtype myofibers at 2-3 days after pupal formation (APF), we discovered that there was a normal loss of integrin from the cell surface, along with detectable presence of integrin-marked inclusions ( Figure 2G-2H). By 4 days APF, integrin was again predominantly absent in the myofiber center, with reappearance at costameres. This result reveals a normal redistribution of integrin that occurs with IOM remodeling, and suggests a distinct requirement for myofiber integrin regulation in pupal stages. To test a temporal requirement for mtm function specifically in pupal stages, we performed temperature shift experiments to induce conditional mtm knockdown. Due to the temperature sensitivity of the GAL4 transcription factor, flies with muscle-targeted mtm hairpin expression maintained normal myofiber attachments when raised continuously at 18uC with low GAL4 activity ( Figure S4D). However, when flies were shifted during metamorphosis to 29uC for 1-2 days with increased GAL4 activity, the pharate adults then exhibited myofiber detachment ( Figure S4D9-S4E, 0% versus 71% mtm pharates, respectively). Similarly, flies also carrying the temperature sensitive GAL80 ts , an inhibitor of GAL4, raised continuously at 18uC did not exhibit integrin-containing inclusions ( Figure S4F, 0%). In contrast, flies shifted to 29uC for 3 days (with shorter metamorphosis at higher temperatures), exhibited myofibers with integrin-containing inclusions ( Figure S4F9-S4G, 58%). These results indicate a requirement for mtm function in pupal stages that is important for integrin localization at the cell surface following myofiber remodeling, and further supports a primary role for mtm in integrin trafficking.
The requirement for mtm function in myofiber remodeling during development raised the question whether there is a similar mtm requirement during cellular remodeling that may occur with ongoing adult muscle use, repair or ageing. We investigated integrin localization in adult abdominal myofibers, which are derived from a different developmental program from the persistent larval muscles. The long, thin adult ventral abdominal muscles, called lateral transversal muscles, normally exhibit a striated pattern of intense integrin localization at repeating costameres ( Figure 2I). In contrast, integrin deviated from this pattern with a diffuse distribution in portions of mtm-depleted myofibers in both six and ten day old adult flies ( Figure 2I9). This result points to an important role for mtm in the maintenance of integrin adhesions with ongoing muscle use in adult flies.
To test whether mtm function has a specific role for integrin localization in broader developmental contexts, we assayed function of integrin-mediated adhesions in the epithelial bilayer of the developing fly wing [1]. The low frequency of wing blisters  resulting from a hypomorphic allele of aPS2-integrin, if 3 [29], was dominantly enhanced when in combination with heterozygous mtm null alleles with reduced function ( Figure 2J), similar to interactions seen with components known to be required for integrin adhesions [30]. The interaction in two different tissues suggests a specific and fundamental role for mtm in maintenance of integrin-mediated attachments.

Integrin adhesions and T-tubules have independent requirements for Mtm function
The disruption in mtm mutants of both IACs and T-tubules, and their normal proximity along the sarcolemma, raised the question whether a structural or functional relationship between the two compartments normally exists or is relevant to the abnormal membrane inclusions ( Figure 3A). Moreover, in mtm-depleted myofibers, we noted that Dlg, similar to bPS-integrin, appeared along abnormal central inclusions ( Figure 3B, 3B9). To characterize the membrane identity of the inclusions, we first tested whether bPS-integrin and Dlg or Amph co-localized, either in normal or mtm mutant muscles. In wildtype myofibers, Dlg and Amph were not detected at IACs. However, internal bPS-integrin frequently co-localized with Dlg ( Figure 3C) and occasionally with Amph ( Figure S5A, S5B, S5C) along apparent longitudinal elements of T-tubules. Upon mtm depletion, bPS-integrin extensively colocalized with Dlg and Amph on longitudinal T-tubules ( Figure  S5A, S5B9) and along the central inclusions ( Figure 3C9, Figure  S5B0), suggesting accumulation of a possible common precursor membrane or trafficking compartment in mtm-depleted muscles. We next considered whether there is functional co-dependence between integrin adhesions and T-tubules. In amph 26 null mutants that lack T-tubules in IOMs ( Figure S5D), as in adult flight muscles [28], we observed normal muscle attachments, normal bPS-integrin localization to costameres, and no bPS-integrinor Dlg-inclusions ( Figure 3D-3D9). This suggests that T-tubules are not required for bPS-integrin trafficking in the formation or maintenance of IACs, and that the inclusion defects in mtm mutants do not reflect a general consequence of failed T-tubule formation. Furthermore, we found bPS-integrin mislocalized to internal membrane inclusions upon mtm depletion in amph 26 null mutants ( Figure 3E), signifying that the abnormal inclusions are independent of transverse tubule membrane and possible misregulation of amph function. Conversely, transverse tubules were present normally in abdominal muscles with hypomorphic mys conditions that were pharate lethal ( Figure 3F), suggesting that T-tubule organization does not require normal levels of bPS-integrin protein or IACs. The dramatic defects in both IACs and T-tubule organization upon mtm depletion therefore appear to reflect a requirement for two independent mtm functions.

PI(3)P role in Mtm-dependent muscle compartmentalization
The disrupted bPS-integrin localization together with the enlarged membrane inclusions suggested defective membrane trafficking in mtm mutant myofibers. Characterization of the central inclusions could point to a specific compartment or trafficking step that normally requires Mtm phosphatase activity in muscle remodeling. The inclusions did not noticeably contain markers of endoplasmic reticulum, the trans-Golgi network or autophagosomes ( Figure S5E-S5E9 KDEL; Figure S5F-S5F9 PH-FAPP1; Figure S5G-S5G9 Atg8). In contrast, the majority of inclusions were decorated by the endosome-lysosomal marker, GFP:LAMP ( Figure 4A-4A0). The inclusions were frequently colocalized with an indicator of early  Figure S5I-S5I9). Together, these results suggest a relationship between the inclusions and early endocytic traffic, and that mtm depletion disrupts endocytic traffic. PI(3)P is normally enriched at endosomal membranes. We have recently shown that the normal Mtm PI(3)P phosphatase activity promotes membrane efflux, effecting both endosomal homeostasis and cortical remodeling in macrophages [18]. We therefore explored PI(3)P distribution in muscle with respect to integrin adhesion and T-tubule compartments. In wildtype animals, muscle expression of the PI(3)P biosensor, GFP:2xFYVE, was detected along the sarcolemma and localized to punctae distributed throughout abdominal myofibers, with the greatest concentration in the perinuclear area without obvious overlap with bPS-integrin ( Figure 4C) or Dlg ( Figure 4D). Upon mtm-depletion, enlarged and more erratically positioned PI(3)P-containing compartments were detected ( Figure 4C9, 4D9). In addition, GFP:2xFYVE co-localized with bPS-integrin ( Figure 4C0) and with Dlg ( Figure 4D0) along the abnormal inclusions in mtmdepleted myofibers, suggesting a possible role for Mtm phosphatase activity in PI(3)P turnover involved in integrin trafficking.

Pi3KC2 suppresses Mtm-related defects in bPS-integrin localization and muscle maintenance
To test if PI(3)P regulation is involved in mtm muscle functions, we investigated the contribution of Class II and III Pi3-kinases (Pi3K68D and Vps34, respectively), known to synthesize PI(3)P, to abdominal muscle maintenance. Muscle-targeted knockdown of Pi3K68D or expression of dominant negative kinase-dead Vps34-KD did not individually disrupt eclosion or animal viability. However, Pi3K68D depletion in combination with mtm RNAi was able to rescue the lethality and delayed development; in contrast, Vps34-KD expression enhanced lethality in combination with mtm depletion (Figure S6A, S6B). Neither Pi3K68D nor Vps34 knockdown rescued the loss of T-tubules with mtm-depletion ( Figure S6C), and accordingly adult flies remained flightless ( Figure S6D). These results indicate separable Pi3K68D-independent and dependent mtm muscle functions required for normal Ttubules and viability, respectively.
A similar functional relationship was seen between Pi3K68D and mtm for roles related to integrin adhesions, as with viability. Importantly, Pi3K68D, but not Vps34, depletion rescued muscle detachment ( Figure 5A, 5B) and loss of bPS-integrin localization at costameres ( Figure 5C, 5C9, 5D) that occurs with loss of mtm function. Consistent with rescue of the IACs, co-depletion of mtm and Pi3K68D, and not Vps34, also eliminated the bPS-integrinand Dlg-containing membrane inclusions ( Figure 5E, 5E9, 5F, Figure S6E), indicating a functional relationship between the abnormal central inclusions and IACs at the sarcolemma. The testis visceral muscle function was also restored to normal with Pi3K68D and mtm co-depletion, implicating turnover of integrinmediated adhesions in the gonadal muscle. Altogether, these results signify that Pi3K68D function mediates mtm RNAi mutant defects in maintenance of IACs, and suggest that Pi3K68D may synthesize a PI(3)P subpool co-regulated by Mtm important for integrin trafficking and localization.
Interestingly, muscle-targeted disruption of Vps34 shared with mtm depletion a similar staged semi-lethality ( Figure S6A), IOM detachment ( Figure 5A9, 5B) and loss of bPS-integrin at costameres ( Figure 5C0, 5D), however, without inducing abnormal inclusions ( Figure 5E0, 5F, Figure S6E). This points to possible shared or sequential roles for Vps34 and Mtm in phosphoinositide-mediated steps in IAC maintenance, distinct from trafficking points that involve antagonistic Pi3K68D and Mtm co-regulation. Vps34 is broadly attributed with roles in PI(3)P-regulated endocytosis and autophagy. We found that inhibition of autophagy upon depletion of the central regulator, Atg1, phenocopied the Vps34 integrin defects ( Figure S6F, S6G, S6H), indicating an important role for autophagy in IOM remodeling. We tested whether the phenotypes associated with mtm phosphatase depletion are a result of increased PI(3)P-mediated autophagy. The myofiber detachment and integrin localization defects, including integrin-containing inclusions, persisted with co-depletion of Atg1 and mtm ( Figure S6F9, S6G9, S6H9), indicating that autophagy is not responsible for the integrin-related defects upon mtm depletion.

Integrin is mislocalized in human myofibers with XLMTM myopathy
Given shared defects observed in mtmand MTM1-disrupted myofibers in flies and human XLMTM, respectively ( Figure 1H9, Figure S3A9-S3D9), we asked whether an mtm function required for integrin adhesions is also shared with MTM1 in human muscle. b1-integrin, the major b-integrin isoform found in vertebrate muscle, was detected along the myofiber sarcolemma in crosssections of skeletal muscle from control subjects, as expected ( Figure 6A). In contrast, b1-integrin localized throughout the perinuclear compartment of centronucleated myofibers in muscle from neonates with XLMTM ( Figure 6B). The Dystroglycan adhesion complex (DAC) is a second complex localized to MTJs and costameres with key roles in muscle attachments, and mutations of DAC components are frequently associated with muscular dystrophy. Unlike integrin, the dystroglycan transmembrane protein exhibited only the expected peripheral staining along the sarcolemma in both control and XLMTM myofibers, without any abnormal centronuclear localization or inclusion ( Figure 6C-6D). These results show that MTM1 is specifically required for normal b1-integrin localization in human myofibers, and suggests that disruption of integrin trafficking and adhesion complex function is important in XLMTM.

Discussion
We found that mtm regulates integrin adhesions in muscle and in the developing wing, and that integrin localization was disrupted in human XLMTM, pointing to a central role for Mtm/MTM1 in a trafficking pathway important for localization of b-integrin at the plasma membrane. It is well-established that integrin turnover contributes to cell motility, whereby targeted integrin recycling and reassembly of localized adhesions mediate polarized matrix attachments and signaling responses [4]. Our results reveal that regulated integrin turnover is also important for integrin adhesions in non-motile myofibers, after the establishment of attachments. Importantly, mtm disruption uncovered a demand for bPS-integrin trafficking in the maintenance of adhesions both at MTJs as well as at costameres, a less-understood adhesion site with putative roles in muscle integrity, mechanotransduction, and myofibril assembly [31,32]. Although integrin was destabilized at larval MTJs in mtm mutants, the most severe consequences occurred later with specific loss of pupal or adult mtm function during developmental myofiber remodeling or adult muscle use, respectively. This is consistent with costamere sensitivity to integrin depletion in adult muscle [33] and the possibility that mtm similarly regulates integrin turnover with myofiber remodeling that occurs both in development and with demands in adult muscle growth, repair and aging.
In fly macrophages, Class II Pi3K68D and mtm co-depletion could revert both an imbalance in PI(3)P and defects in cortical remodeling that impaired macrophage shape and in vivo immune cell distribution [18]. Here, we found Pi3K68D disruption is also a specific and potent suppressor of integrin adhesion defects in mtm-depleted muscle. Despite distinct macrophage and myofiber morphology and function, a shared requirement for a PI3KC2/Mtm pathway highlights common functions during cellular remodeling.  Loss of Mtm phosphatase activity could be considered a gain of function condition, analogous to ectopic kinase activity, leading to inappropriate phosphoinositide accumulation. In line with this, either mtm depletion (this study) or Pi3K68D overexpression [34] disrupted integrin adhesion in the fly wing, presumably through imbalanced responses to an accumulation of the same phosphoinositide pool. PI3KC2 and Mtm family members in vertebrates have been associated with antagonistic functions related to regulation of traffic to the plasma membrane. PI3KC2 isoforms are required to promote while overexpression of MTM1 impairs GLUT4 trafficking [20,35] and integrin-mediated cell motility [21,36]. Together, the observations point to a broad and conserved relationship for PI3KC2/Mtm co-regulation at the plasma membrane.
How might PI3KC2 and Mtm co-regulate integrin trafficking? One possibility is that the cycle of phosphoinositides co-regulated by PI3KC2/Mtm tunes the balance between endocytic-exocytic flux. The strong genetic interaction between mtm and Pi3K68D, in conjunction with Pi3KC2 ability to create PI(3)P in vivo [18,[20][21][22][23], supports the possibility that Pi3K68D could generate a PI(3)P substrate pool acted on by Mtm phosphatase. Alternatively, Pi3K68D could act more distantly on an interrelated phosphoinositide pool. We envision that Pi3K68D mediates early endocytic trafficking, tethering or sorting of integrin-containing vesicles. The integrin detected on large inclusions in mtm-depleted and XLMTM muscles in flies and humans, respectively, and evidence that mtm promotes membrane tubulation from PI(3)P compartments [18,37], point to an Mtm/MTM1 role in membrane efflux for delivery of integrin to the plasma membrane. Mtm phosphatase could act to promote recycling or to negatively regulate retention, for example, through a PI(3)P-mediated fusion of integrin-containing vesicles with endosomes-lysosomes. An accumulation of b1-integrin on enlarged, perinuclear compartments has been observed with certain genetic manipulations in non-muscle cells. These results raise the possibility that normal Mtm phosphatase activity functions antagonistically to Rab21 GTPase or in concert with PKCe kinase, Rab11 and/or Arf6 GTPase, respectively, to control redelivery of b-integrin to the plasma membrane [38][39][40]. We found that class III PI3K, Vps34, also contributes to integrin localization upon myofiber remodeling, but with no effect on integrin-containing inclusions. A requirement for class III Pi3K could be at a shared step with the early endosomal Rab5 GTPase shown to be involved in integrin turnover at larval MTJs [6]. Thus, regulation of distinct PI(3)P pools is important for differential regulation of integrin endosomal trafficking, whereby Pi3KC2 and Mtm are dedicated to specific paired antagonistic functions.
We discovered that mtm is required in muscle for both integrinmediated adhesions and T-tubule organization. The T-tubule requirement for mtm was similar to but not as severe as that for amph, the sole homolog of human AMPH2 that is also associated with centronuclear myopathy [41]. However, unlike mtm, null alleles of amph did not share a defect of myofiber detachment. Despite localization of bPS-integrin at T-tubules, and the dual requirements for mtm, we found that normal integrin adhesions and abnormal bPS-integrin localization on inclusions are independent of T-tubule organization. This suggests that mtm may serve a common function for integrin turnover and T-tubule formation at a shared precursor compartment, for example, at recycling endosomes, or alternatively, act independently at two distinct sites. b-integrin, Dlg and Amph are known to functionally interact at postsynaptic junctions [42][43][44], and MTMR2 has been shown to interact with Dlg1/SAP-97 and Dlg4/PSD-95 to promote postsynaptic function [45,46]. Thus, the shared accumulation of bPS-integrin, Dlg and Amph on central membrane inclusions in mtm-depleted myofibers, and their elimination with Pi3K68D co-depletion, points to a possible role for a PI3KC2/ Mtm pathway in endocytic recycling at neuromuscular junctions, as well as at MTJs.
Many of the defects observed in mtm mutant muscle parallel those associated with the human disease, XLMTM, demonstrating that the fly offers a tractable model for the cellular basis of centronuclear myopathy. Importantly, the discovery that mtm broadly regulates bPS-integrin turnover through endocytic trafficking led us to uncover a previously untested defect in b1integrin localization in human XLMTM myofibers. Normal myofiber organization and function rely on integrin adhesions in vertebrate muscle [2,3,31,47]. Thus, disruption of integrin regulation provides a basis for aspects of the severity of myofiber disorganization and dysfunction observed in XLMTM. The conservation between fly mtm and human MTM1 functions brings further significance to the potent interaction demonstrated between mtm and class II Pi3K68D for integrin regulation in flies. Whereas Class I and III PI3-kinases have been the focus of intense study as potential therapeutic targets of specific inhibitory compounds, the Class II PI3-kinases have received little attention. The knowledge of PI3KC2 contributions to specific MTM pathways is significant towards motivating similar studies for potential strategies addressing MTM-related disease.
Mtm is the single fly homolog related to both human MTM1 and MTMR2, and human MTMR2 expression was able to rescue integrin-related defects in mtm-depleted fly myofibers. An mtm pathway function in endocytic trafficking is therefore relevant to a more general understanding of the cell biological functions employed by MTM subfamily members. Mutations in MTMR2 associated with CMT4B neuropathy affect the morphology and function of myelinating Schwann cells [11], which like myofibers, share features of having an extensive plasma membrane and a reliance on integrin adhesions [48]. The regulation of integrin trafficking under the control of a conserved PI3KC2/Mtm pathway may be an important mechanism for controlling cell compartmentalization more broadly in different contexts, and relevant to different MTM-related human disease.

Ethics statement
Human samples were obtained and used as per institutional IRB accepted protocol.

Muscle preparations and immunofluorescence
Staged pharate adults were removed from pupal case fastened to double-sided tape and pinned on a sylgard covered petri dish in dissecting buffer (5 mM HEPES, 128 mM NaCl, 2 mM KCl, 4 mM MgCl 2 , 36 mM sucrose, pH 7.2). Abdomens were opened with longitudinal and two lateral incisions, pinned flat, washed and fixed 30 min.

Timelapse microscopy
White pupae were positioned on double-sided tape on a coverslip, placed in a petri dish with water-soaked filter and incubated at 25uC. At each time point, the coverslip was flipped over for imaging on Leica DMI 6000B, as above.
FRAP FRAP was carried out as described [6] in living 3rd instar larvae. Integrin:YFP was heterozygous (Int/+) in all experiments. A total of 21 and 19 individual FRAP experiments from multiple larvae were carried out for int/+ and mtm z2-4747 /mtm D77 ;int/+ respectively.

Viability and flight assays
To assay viability, pre-cleared vials were counted for surviving adults, dead adults post-eclosion on food and mid-eclosion, dead pharates and dead pupae at 13 or 17 days after egg laying. Flies 1-6 days old were tested for flight at least 24 hours after CO 2 anesthesia, by releasing 12 females in a 2 L cylinder (50.8 cm high). Flies that landed below 0.6 L (14.6 cm) were scored flightless.

Statistical analyses
Visual quantification was made for number of IOMs in tergites 3 and 4, number of detached IOMs per abdomen, number of IOMs displaying bPS-integrin costameres or bPS-integrinor Dlgmarked inclusions, and number of nuclei per IOM. ImageJ software was used to draw and measure nuclei distance to IOM midline and sarcomere length. CellProfiler was used to segment and quantify nuclei morphology. Statistical analysis in Prism software used to determine mean, standard error and Student's ttest, where possible.

Genotypes
Full genotypes used are as shown in Figure Legends