Rab35-regulated lipid turnover by myotubularins represses mTORC1 activity and controls myelin growth

Inherited peripheral neuropathies (IPNs) represent a broad group of disorders including Charcot-Marie-Tooth (CMT) neuropathies characterized by defects primarily arising in myelin, axons, or both. The molecular mechanisms by which mutations in nearly 100 identified IPN/CMT genes lead to neuropathies are poorly understood. Here we show that the Ras-related GTPase Rab35 controls myelin growth via complex formation with the myotubularin-related phosphatidylinositol (PI) 3-phosphatases MTMR13 and MTMR2, encoded by genes responsible for CMT-types 4B2 and B1 in humans, and found that it downregulates lipid-mediated mTORC1 activation, a pathway known to crucially regulate myelin biogenesis. Targeted disruption of Rab35 leads to hyperactivation of mTORC1 signaling caused by elevated levels of PI 3-phosphates and to focal hypermyelination in vivo. Pharmacological inhibition of phosphatidylinositol 3,5-bisphosphate synthesis or mTORC1 signaling ameliorates this phenotype. These findings reveal a crucial role for Rab35-regulated lipid turnover by myotubularins to repress mTORC1 activity and to control myelin growth. Charcot-Marie-Tooth (CMT) is an inherited peripheral neuropathy. Here, the authors show that Rab35 forms a complex with genes implicated in CMT, MTMR13 and MTMR2, which regulates myelin growth by controlling mTORC1 signaling through lipid turnover.

I nherited peripheral neuropathies (IPNs) represent a broad group of genetically and clinically heterogeneous disorders, including Charcot-Marie-Tooth (CMT) neuropathies, one of the most common inherited neurological disorders 1 . Within the past decades almost 100 IPN/CMT disease genes have been identified, yet, the molecular mechanisms underlying most of the IPN/CMT neuropathies are incompletely understood. Among CMTs, CMT4B is characterized pathologically by the presence of myelin outfoldings, a form of focal hypermyelination with redundant loops of myelin 2 . CMT4B includes three distinct subtypes. CMT4B1 and CMT4B2 are associated with mutations in the MTMR2 and MTMR13 (myotubularin-related protein 2 and 13, the latter also named SET binding factor 2, SBF2) genes, respectively, and are characterized by a demyelinating neuropathy with early onset 3,4 . CMT4B3 has been more recently associated with mutations in the MTMR5/SBF1 gene but is characterized by different phenotypes with either a pure demyelinating neuropathy or an axonal polyneuropathy complicated by central nervous system involvement 2 . The tissue specificity of CMT4B disease phenotypes suggests that MTMR2, MTMR5, and MTMR13 have cell-type specific functions. MTMR2 is a ubiquitously expressed phosphatidylinositol 3-phosphatase of the myotubularin-related protein family that dephosphorylates both phosphatidylinositol 3-phosphate [PI(3)P] and phosphatidylinositol 3,5-bisphosphate [PI(3,5)P 2 ] phospholipids, which are mainly enriched in the endolysosomal system 5,6 . Consistently, we found that PI(3,5)P 2 levels are increased in primary cells from Mtmr2 KO mutant mice, which recapitulate CMT4B1 in humans, suggesting that this lipid is an important substrate of MTMR2 in Schwann cells in vivo 7 . On the contrary, MTMR5 and MTMR13 are catalytically inactive proteins and associate with MTMR2 to potentiate phosphatase activity and to regulate its subcellular localization 8,9 . The localization of these MTMRs, however, remains to be clearly defined. How elevated levels of phosphatidylinositol (PI) 3phosphates under conditions of loss-of-function of MTMR2 and/ or MTMR5/MTMR13 may perturb myelination in the peripheral nervous system is largely unknown.
Recent data from non-myelin forming cell types suggest that PI (3)P and PI(3,5)P 2 locally facilitate nutrient signaling by mTORC1 at late endosomes and lysosomes [10][11][12][13] . Elevated signaling via the AKT-mTORC1 axis, e.g. upon constitutive AKT1 activation or conditional genetic disruption of PTEN in Schwann cells causes focal hypermyelination consisting of redundant loops of myelin and tomacula 14,15 , while hyperactive mTORC1 during early stages of development delays the onset of myelination 16 . Loss of mTORC1 activity has been shown to hamper myelination 17,18 . These data suggest that mTORC1 signaling plays a dual role in controlling myelination in the peripheral nervous system 19 that may conceivably be modulated by PI 3phosphates that serve as substrates for MTMRs.
The small GTPase Rab35, a central regulator of endosomal function 20,21 has been implicated in a variety of cell physiological pathways that range from the regulation of endosomal trafficking [20][21][22] including secretion of exosomes 23 , actin dynamics 21 and apico-basal polarity 24 to cytokinesis 25,26 and the modulation of cell signaling 27 , and migration 24,28,29 . These various roles have been linked to the ability of Rab35 to bind and recruit effector proteins such as the PI 5-phosphatase OCRL 30,31 , the Arf6 GTPase activating protein ACAP2 32,33 , the oxidoreductase MICAL1 34 and the endosomal protein MICAL-L1 35 . Given the multitude of effector proteins for other endosomal Rabs such as Rab5 it is likely that additional Rab35 effector proteins exist. Rab35 activation is triggered by GEFs including endocytic or endosomal DENN domain-containing proteins 20,30,36 and, possibly, the late endosomal/lysosomal mTORC1 regulator folliculin, which contains a DENN-like module 37,38 .
Here we show that Rab35 controls myelin growth via complex formation with myotubularin-related phosphatidylinositol (PI) 3phosphatases including MTMR13 and MTMR2 implicated in CMT 4B1 and B2, respectively, to downregulate lipid-mediated mTORC1 activation. Our findings reveal a crucial role for Rab35regulated lipid turnover by myotubularins in the control of mTORC1 activity and myelin growth suggesting possible avenues for the treatment of CMT 4B-type neuropathies in humans.

Results
Rab35•GTP recruits MTMR13-based lipid phosphatase complexes. While Rab35 has been implicated in a multitude of cell physiological functions 20,21 , we know comparably little about the precise molecular mechanisms and protein effectors, e.g. proteins associated with active Rab35-GTP, that underly these roles. To fill this gap, we conducted a non-biased proteomic screen for Rab35 interacting proteins based on BioID 39 , a technique that harnesses the ability of a promiscuous biotin ligase to biotinylate proteins in its close proximity. We expressed a chimeric protein comprised of Rab35 fused to a mutant version of the bacterial BirA* biotin ligase in biotin-fed HEK293T cells (Fig. S1a) and analyzed affinity-purified biotinylated proteins co-enriched with Rab35-BirA* over BirA* by quantitative mass spectrometry (Supplementary Table 1 and Supplementary Data 1). This analysis revealed a striking association of Rab35 with MTMR13 (Supplementary Data 1) and MTMR5 (Supplementary Table 1), two myotubularin-related catalytically inactive PI 3-phosphatases implicated in myelin biogenesis and CMT in humans (Fig. 1a). Affinity chromatography experiments using GST-Rab35•GTP (active) or -Rab35•GDP (inactive) as baits confirmed specific binding of MTMR13 and, less efficiently, MTMR5 to Rab35•GTP in vitro (Fig. 1b, c). A similar preference of MTMR13 for active Rab35•GTP was observed in co-immunoprecipitation experiments employing GDP-vs. GTP-locked variants of GFP-Rab35 and FLAG-MTMR13 co-expressed in HEK293T cells (Fig. 1d). Complex formation between Rab35 and MTMR13 or MTMR5 (consistent with data from Drosophila 40 ) was specific as other myotubularin family members such as MTM1, MTMR1, or MTMR2 when expressed on their own failed to bind to GST-Rab35•GTP (Fig. 1b, c). As Rab35•GTP associated with MTMR13 with preference over MTMR5 we focussed on the Rab35/ MTMR13 interaction. First, we observed that MTMR13 associates with Rab35 but not with GST-fused Rab1A, Rab5, Rab7, or Rab11 irrespective of their nucleotide loading status ( Supplementary  Fig. 1b). Second, we made use of HeLa knockin (KI) cells expressing GFP-Rab35 from its endogenous locus 30 .We found that GFP-Rab35 endo captured on GFP-nanotrap beads co-purified with mCherry-MTMR13 expressed in these cells (Fig. 1f). Further biochemical mapping experiments revealed that MTMR13 complex formation with Rab35 was mediated, in part, by its catalytically inactive PTP phosphatase domain (Fig. 1f), although other domains may also contribute to binding. Consistently, a mutant of MTMR13 lacking its DENN domain (and, hence, devoid of its putative GEF activity), associated with constitutively active GFP-Rab35•GTP but not with dominant-negative GFP-Rab35•GDP in transfected HEK293cells (Fig. 1e). These data suggest that Rab35•GTP associates with myotubularin family proteins including MTMR13 possibly to aid its recruitment to intracellular membranes. As no antibodies suitable for detection of endogenous MTMR13 are available, we probed this possibility by analyzing the subcellular distribution of overexpressed MTMR13 in transfected HeLa cells. When expressed on its own MTMR13 exhibited a largely cytoplasmic distribution ( Supplementary  Fig. 2c). Endogenous GFP-Rab35 endo in addition to its known localization to peripheral endosomes and the plasma membrane partially colocalized with the late endosomal/lysosomal marker LAMP-1 (Fig. 2a), a localization pattern that was even more pronouncedly observed for GTP-locked Rab35 (Rab35 CA) (Fig. 2b, Supplementary Fig. 2a). Interestingly, expression of constitutively active mutant Rab35 sufficed to induce the recruitment of co-expressed MTMR13 to the cell surface (Fig. 2c,  Supplementary Fig. 2b) and to LAMP1-positive late endosomes/ lysosomes, while inactive GDP-locked Rab35 was less potent (Fig. 2d, e; Supplementary Fig. 2c). Active Rab35-mediated recruitment to LAMP1-positive late endosomes/lysosomes was also seen for the PTP domain of MTMR13 (Fig. 2f,g; Supplementary Fig. 2d), further confirming our biochemical mapping results (Fig. 1f). Rab35•GTP, thus, can recruit MTMR13 to intracellular membranes including late endosomes or lysosomes.
Although MTMR13 is catalytically inactive, it is known to form tetrameric assemblies with MTMR2 (another CMT gene) and,   Collectively, these findings uncover Rab35 as an interactor of MTMR13-based lipid phosphatase complexes.

Rab35 loss in Schwann cells causes focal hypermyelination.
Given the prominent genetic and functional association of MTMR13 and its catalytic subunit MTMR2 with CMT and demyelinating neuropathies in humans and in defective myelin biogenesis in mice 2,4,41-43 , we hypothesized that Rab35 may regulate myelin biogenesis in vivo. Consistent with this hypothesis, we found Rab35 to be highly expressed in sciatic nerve lysates from newborn mice at P2, i.e. around the onset of myelination, whereas its expression levels declined during postnatal PNS development (Fig. 3a). To explore the possible function of Rab35 in myelin biogenesis in the PNS, we targeted the Rab35 locus for genetic manipulation by introducing loxP recombination sites that allow for the conditional disruption of Rab35 expression ( Supplementary Fig. 3). Constitutive disruption of the Rab35 locus in β-actin Cre expressing strains resulted in embryonic lethality. To bypass this, we crossed Rab35 flox/flox mice with a transgenic mouse line expressing Cre recombinase under the control of the myelin protein zero (P0) promoter resulting in the Schwann cell (SC)-specific downregulation of Rab35 expression as early as E13.5 44 . Rab35 flox/flox P0-Cre mice (hereafter termed cKO SC ) were born according to a Mendelian ratio and did not display any evident clinical phenotype. Sciatic nerve lysates from Rab35 cKO SC showed a strong reduction of Rab35 protein ( Fig. 3b; the remaining Rab35 signal likely originates from axons and/or fibroblasts present in the sciatic nerve). Ultrastructural or semithin section analyses from their sciatic nerves at postnatal day 20 (P20) (Fig. 3c, d), P30 and P90 (Fig. 3c, e) revealed a normal number of myelinated axons and unaltered myelin thickness. Only at P90 large fibers (>5 µm in diameter) displayed reduced myelin thickness (Fig. 3e). Interestingly, myelinated fibers from Rab35 cKO SC sciatic nerves showed abundant signs of focal hypermyelination including tomacula and myelin outfoldings, in addition to myelin degeneration (Fig. 3f, g). These myelin abnormalities, which were never observed in control nerves (Fig. 3c, g) progressively worsened with the age of the animals (Fig. 3c, f). These findings indicate that loss of Rab35 expression in Schwann cells recapitulates morphological defects in myelin architecture observed in other IPN/CMT models including loss of MTMR2 and MTMR13 in mice 42,45 . Our data reveal a prominent role for Rab35 in the control of myelin growth that is consistent with a physical and functional interaction with the CMT-associated lipid phosphatases MTMR13 and MTMR2.
Rab35 and its associated MTMRs repress mTORC1 activity. Focal hypermyelination including tomacula and myelin outfoldings have been found to be associated with hyperactive mTORC1 signaling, for example as a consequence of genetic inactivation of the mTORC1 repressor TSC1 or the lipid phosphatase PTEN, a major negative regulator of the PI 3-kinase-AKT-mTORC1 pathway 14,46,47 . Based on these results and the fact that the MTMR substrates PI(3)P and PI(3,5)P 2 can activate mTORC1 signaling 10-12 , we hypothesized that Rab35 via complex formation with MTMR-based lipid phosphatase complexes at late endosomes/lysosomes may locally repress mTORC1 activity. We therefore analyzed mTORC1 signaling in HEK293T cells depleted of endogenous Rab35 and/or MTMR2. Depletion of either Rab35 or MTMR2 caused a significant elevation in the steady-state levels of phospho-mTOR/total mTOR, or phospho-S6 kinase 1 (p-S6K)/total S6K levels, indicative of mTORC1 hyperactivity (Fig. 4a, c, d). A similar increase in mTORC1 activity was observed upon knockdown of either MTMR5 or MTMR13 (Fig. 4e, f; Supplementary Fig. 4a, b). Codepletion of Rab35 and MTMR2 led to an even more robust increase in mTORC1 signaling (Fig. 4a, c, d), in line with the fact that Rab35 can associate with MTMR13-MTMR2 and other lipid phosphatase complexes including MTMR5 and MTMR1 (compare Fig. 1, Supplementary Fig. 1). In contrast, loss of Rab35 and/ or MTMR2 in HEK293 cells did not result in significant alterations in AKT signaling (Fig. 4b) consistent with previous data on MTMR2 in Schwann cells 48 . These data suggest that Rab35 and its associated MTMRs may regulate mTORC1 activity independent of AKT.
To dissect whether genetic loss of Rab35 affects mTORC1 signaling in cells of the nervous system we crossed our Rab35 flox/flox line with CAG-Cre ER mice allowing for the conditional KO (cKO) of Rab35 in primary astrocytes in culture ( Supplementary Fig. 5a) following addition of tamoxifen ( Supplementary Fig. 5b). Rab35 cKO astrocytes displayed significantly elevated levels of p-S6/total S6 ( Supplementary  Fig. 5c, d) and p-S6K/S6K and these partially persisted following serum starvation, i.e. conditions under which p-AKT is inactive   Fig. 5a, b). Increased mTORC1 activity was not a consequence of overactive p-AKT signaling (Fig. 5c, d), which, instead, was reduced in KO astrocytes (similar to recent data from mouse embryonic fibroblasts 29 and different from our results in HEK293 cells discussed above). Moreover, Rab35 cKO astrocytes were significantly larger than their control (Rab35 flox/flox without Cre) counterparts (Fig. 5e, Supplementary Fig. 5g), consistent with the established function of mTORC1 in regulating cell size 14,19 . Surprisingly, cKO of Rab35 did not cause any overt defects in the endolysosomal system as evidenced by unaltered levels and appearance of EEA1-containing early endosomes, Rab7-positive late endosomes, or lysosomes probed by LAMP-1 and LAMP-2 ( Supplementary Fig. 5h-o). We did, however, notice a mild reduction in the expression of the autophagy marker LC3, possibly as a consequence of reduced TFEB-mediated expression of autophagy genes under conditions of elevated mTORC1 activity ( Supplementary Fig. 5e, f). mTORC1 hyperactivity in Rab35 cKO astrocytes was further aggravated by lentiviral codepletion of MTMR2 ( Fig. 5f; Supplementary Fig. 5p, r), akin to results from HEK293T cells, whereas the observed reduction of p-AKT signaling in Rab35 cKO astrocytes was unaffected by MTMR2 loss (Supplementary Fig. 5q). Similar changes in    (Fig. 6a, b) and in sciatic nerve lysates from Schwann cell-specific cKO mice (Rab35 cKO SC ) that display focal hypermyelination (Fig. 6c, d). In contrast, the steady-state levels of p-ERK or p-AKT were unchanged in nerve lysates from cKO mice (Fig. 6e). These collective data suggest that Rab35 via complex formation with MTMR-based lipid phosphatase complexes represses mTORC1 signaling. This model is further supported by the observations that (i) MTMR2 levels are significantly reduced in Rab35 cKO astrocytes (  Fig. 4d). If mTORC1 hyperactivity in the absence of Rab35 causally contributes to focal hypermyelination found in Schwann cellspecific Rab35 cKO SC mice this phenotype might be ameliorated by pharmacological modulation of mTORC1 signaling. To test this, we analyzed the effects of mTORC1 suppression induced by application of the specific mTORC1 inhibitor Rapamycin in myelin-forming Schwann cell/DRG neuron co-culture explants established from "WT" control (Rab35 flox/flox genotype) and    Rab35 cKO SC mouse embryos (Fig. 7a). mTORC1 signaling plays a developmental timing-dependent dual role in peripheral nerve myelination: hyperactive mTORC1 during early nerve development delays the onset of myelination by suppressing differentiation 16 , while mTORC1 hyperactivity in differentiated myelinating Schwann cells causes focal hypermyelination 14,46 . Consistent with this, we observed a significantly reduced number of myelin segments marked by myelin basic protein (Mbp) in co-cultures from Rab35 cKO SC embryos compared to control (Fig. 7b), likely as a consequence of overactivated mTORC1 signaling in these explants particularly at the onset of myelination between two and seven days of ascorbic acid treatment in differentiating conditions (Fig. 7c). Those fibers that had undergone myelination displayed aberrant myelin (similar to our observations in vivo, compare Fig. 3) with a complex shape and, in part, resembling myelin outfoldings of Mtmr2 KO cultures (Fig. 7e). Importantly, treatment of co-culture explants from Rab35 cKO SC embryos with Rapamycin rescued impaired myelin segment formation and ameliorated aberrant myelin (Fig. 7d-f) To corroborate these findings in vivo, we treated Rab35 cKO SC mutants and littermate controls (Rab35 flox/flox ) with Rapamycin. The drug was given from P12, when mice can still tolerate longterm Rapamycin treatment, to age P70, when the clinical phenotype was pronounced. Immunoblotting of sciatic nerve lysates of Rapamycin-treated control or Rab35 cKO SC animals displayed strongly diminished p-S6 levels compared to placebotreated controls, demonstrating efficient blockage of the mTORC1 pathway (Fig. 8a). Rapamycin treatment significantly reduced aberrant myelin, particularly fibers carrying tomacula (Fig. 8d, e) and, to a lesser extent, myelin degeneration (Fig. 8b), while no significant effect on outfoldings was observed (Fig. 8c), possibly as a result of the very early onset of the phenotype ( Supplementary Fig. 6c). Of note, at either P3 or P5, prior to Rapamycin treatment, we observed a consistent number of altered myelinated fibers carrying myelin degeneration or focal hypermyelination ( Supplementary Fig. 6b, e) with a similar number of myelinated fibers (Supplementary Fig. 6a, d) and of myelin thickness between control and Rab35 cKO SC sciatic nerves (gratio as a function of axonal diameter in sciatic nerves at P3: Rab35 cKO SC , 0.8245 ± 0.0059, 192 fibers; controls (Rab35 flox/+ ), 0.834 ± , 0.011, 194 fibers, n = 3 animals per genotype, p = 0.7000. P5: Rab35 cKO SC , 0.7542 ± 0.0052, 443 fibers; controls (Rab35 flox/+ ), 0.749 ± , 0.0085, 527 fibers, n = 5 animals per genotype, p = 0.6905). These findings suggest that elevation of mTORC1 signaling does not impair the onset of myelination and is consistent with the observed extent of mTORC1 overactivity at P30 (compare Fig. 6c, d). Collectively, our data are consistent with the hypothesis that elevated mTORC1 activity in the absence of Rab35 contributes to focal hypermyelination in Rab35 cKO SC mice, a phenotype that is ameliorated by pharmacological inhibition of mTORC1 signaling by Rapamycin.
Rab35 and MTMRs regulate mTORC1 by PI 3-phosphate hydrolysis. mTORC1 activation at late endosomes or lysosomes is potently stimulated by PI 3-phosphates such as the MTMR substrates PI(3)P and PI(3,5)P 2 . Hence, the observed elevation in mTORC1 signaling in the absence of Rab35 might conceivably be a consequence of increased PI 3-phosphate levels due to impaired hydrolysis mediated by MTMR2/MTMR13 and related MTMR complexes. While no reliable probes exist to quantitatively monitor the levels and distribution of PI(3,5)P 2 in eukaryotic cells, PI(3)P can be semi-quantitatively measured by specific PI(3) P-binding probes such as recombinant eGFP2xFYVE. We used eGFP2xFYVE as a surrogate measure for the possible accumulation of MTMR substrates in Rab35 depleted cells. Loss of Rab35 in cKO astrocytes caused a significant, more than two-fold elevation in the levels of PI(3)P (Fig. 9a, b). Vps34 lipid kinase produces the bulk of PI(3)P in mammalian cells, which then serves as a precursor for PI(3,5)P 2 synthesis by the PIKFYVE complex (comprised of the active PI 5-kinase PIKFYVE, the lipid phosphatase FIG4, and Vac14 protein). Consistently, we found that PI(3)P was strongly depleted following application of the selective Vps34-class III PI 3-kinase inhibitor VPS34-IN1 (Supplementary Fig. 7b). Similar observations were made in HeLa cells depleted of either Rab35, MTMR2, or of both proteins in combination, where PI(3)P was seen to accumulate on CD63-positive late endosomes or lysosomes (Fig. 9c, Supplementary Fig. 7), e.g. compartments where active mTORC1 is located. Next, we probed whether elevated PI 3-phosphate levels were causal for mTORC1 hyperactivity in Rab35 cKO cells. Treatment of Rab35 cKO astrocytes with either VPS34-IN1 49 , SAR405 50 , or Compound 19, selective inhibitors of Vps34-mediated PI(3)P synthesis, potently suppressed elevated mTORC1 signaling (Fig. 9d, e). Pharmacological inhibition of Vps34 activity also suppressed mTORC1 hyperactivation in Rab35 cKO astrocytes co-depleted of MTMR2 (Fig. 9f, g). We conclude that loss of Rab35 and/or its downstream effector MTMR2 causes the accumulation of PI 3phosphates including PI(3)P at late endosomes/lysosomes, resulting in local mTORC1 hyperactivation.
These combined data open the possibility that uncontrolled myelin growth due to lack of Rab35 might be rescued by pharmacological manipulation of PI 3-phosphate levels in Schwann to WT controls and these were robustly suppressed by the mTORC1 inhibitor Rapamycin (Fig. 10a, b, Supplementary  Fig. 8a). Increased mTORC1 activity in cKO Schwann cells was paralleled by increased levels of PI(3)P (Fig. 10c, d, Supplementary Fig. 8b).
If the elevated mTORC1 activity in Rab35 cKO Schwann cells was a consequence of elevated PI(3)P levels, it should be rescued by pharmacological inhibition of Vps34-mediated synthesis of PI (3)P but not by selective inhibition of PI(3,5)P 2 synthesis in the presence of the specific PIKFYVE inhibitor Apilimod. Rescue of elevated mTORC1 signaling in Rab35 cKO Schwann cells by either Vps34 inhibitors or Apilimod, however, would indicate a major role for PI(3,5)P 2 in mTORC1 hyperactivation in these cells. To test these hypotheses cultured differentiating Schwann cells were chronically treated with either DMSO (as a solvent control), low-doses of Rapamycin, the specific Vps34 inhibitors VPS34-IN1, or compound 19, or the selective PIKFYVE inhibitor Apilimod 51 . Elevated mTORC1 signaling in Rab35 cKO Schwann cells was potently rescued by pharmacological inhibition of Vps34-mediated synthesis of PI(3)P or inhibition of the PI(3,5)P 2 synthesizing enzyme PIKFYVE (Fig. 10a, b; Supplementary  Fig. 8a). Semi-quantitative determination of PI(3)P levels in Schwann cell cultures confirmed the near complete elimination of PI(3)P production by Vps34 inhibition, while PI(3)P levels were elevated under conditions of impaired synthesis of PI(3,5)P 2 from PI(3)P in the presence of Apilimod (Fig. 10c, d; Supplementary  Fig. 8b). The levels of PI(4)P analyzed as a control remained unchanged ( Supplementary Fig. 8c, d). Elevated mTORC1 activity was also observed in WT and Rab35 cKO Schwann cells lentivirally depleted of MTMR2 ( Supplementary Fig. 8e-h). These findings suggest that elevated mTORC1 activity in the absence of Rab35 is largely caused by the accumulation of PI(3,5)P 2 in Schwann cells, possibly as a consequence of perturbed PI(3,5)P 2 hydrolysis by MTMRs including MTMR2/MTMR13 consistent with previous data 41 , although more indirect mechanisms (e.g. via recycling) cannot be ruled out. Finally, we assessed the effects of pharmacological manipulation of PI 3-phosphate synthesis on myelin protein production by Rab35 flox/flox WT control or cKO Schwann cells in culture. Acute tamoxifen-induced cKO of Rab35 in cultured differentiated Schwann cells from Rab35 flox/flox mice resulted in increased production of myelin protein zero (P0) compared to WT controls and this defect was rescued by inhibition of mTORC1 signaling in the presence of Rapamycin. Strikingly, we found that increased myelin protein P0 synthesis was equally well suppressed by sustained pharmacological inhibition of either PI(3)P or PI(3,5)P 2 synthesis by Apilimod (Fig. 10e, f; Supplementary Fig. 8i).
These collective data indicate that uncontrolled myelin growth in Rab35-depleted Schwann cells largely results from mTORC1 hyperactivation caused by the accumulation of the MTMR substrate PI(3,5)P 2 , and possibly also PI(3)P, at late endosomes or lysosomes. They further suggest that pharmacological manipulation of mTORC1 signaling and/or of PI 3-phosphate metabolism may represent a potential avenue for the treatment of CMT4B1/ 2-type hereditary neuropathies caused by loss of myotubularin family phosphatases.

Discussion
Previous work had revealed critical roles for the myotubularinfamily PI 3-phosphatases MTMR2 and MTMR13, and for the AKT-mTORC1 signaling pathway in the control of myelin growth and architecture. In our study we identify the small Rasfamily GTPase Rab35 as a regulator of myelination that by   and, likely, PI(3,5)P 2 (although current methodology does not allow us to directly assess this in Schwann cells), which facilitate mTORC1 activation (consistent with [10][11][12] ). Such a model agrees well with recent findings that suggest defective PI 3-phosphate metabolism or recognition, e.g. due to mutations in PTEN, MTMR2, MTMR13, or in the FYVE domain (i.e. a module that binds PI 3-phosphates)-containing protein Frabin/FGD4, as a common feature of neuropathies with myelin overgrowth 3,4,14,43 .
Our findings are further consistent with a critical dual function for mTORC1 signaling -a process crucially regulated by PI 3phosphate metabolizing enzymes such as PTEN -in controling the differentiation of myelin producing cells and myelin growth 14,16,46 . How exactly the different forms of overgrown myelin such as myelin outfoldings and tomacula arise, remains poorly understood. Akin to results for PTEN mutants 14 we found that mTORC1 inhibition ameliorated tomacula but much less well myelin outfoldings in Rab35 cKO SC mice, suggesting that mTORC1 activity primarily controls bulk membrane addition. It is therefore conceivable that other pathways beyond mTORC1 signaling contribute to focal hypermyelination in Rab35 cKO SC mice. These may include dysregulation of exosome secretion, a process known to be important for myelin membrane growth that is regulated by the Rab35 GTPase activating proteins TBC1D10A-C 23 , phosphoinositide regulation of actin dynamics by Rab35 21,52 and/or MTMRs, or the endocytic recycling of signaling receptors, for example as a consequence of alterations in the activity of the small GTPase Arf6 21,53 via the Rab35 effector ACAP2 20,21,28,54 . The latter possibility is consistent with the finding that depletion of Rab35 or ACAP2 promotes myelin protein synthesis in oligodendrocytes in culture 32 , while Arf6 itself has been implicated in Schwann cell differentiation 55 . Unlike its interaction partners MTMR13 or MTMR2 (bound to Rab35 via MTMR13 or MTMR5) Rab35 does not appear to be an inherited neuropathy disease gene. This is likely owed to the fact that Rab35 function is essential for survival, at least in mice, and that mutations within Rab35 are unlikely to affect a single effector protein and, hence, would exhibit pleiotropic effects in many different cell types and organs. Given the phenotypic similarity of Rab35 cKO SC in Schwann cells to focal hypermyelination observed in MTMR2 and MTMR13 mouse models 42,45,56 and the physical and functional interaction of Rab35•GTP with these proteins, we propose that Rab35 is a major upstream regulator of MTMR function in Schwann cells and other cell types. The fact that Rab35•GTP can bind to both MTMR13 and MTMR5 and either of these catalytically inactive proteins is capable of associating with several active MTMRs 57 including MTMR2 and MTMR1(compare Fig. 1, Supplementary Fig. 1) suggests that Rab35 acts as a master regulator for MTMR-mediated hydrolysis of PI 3-phosphates at late endosomes or lysosomes. Consistent with this hypothesis we observe that focal hypermyelination in peripheral nerves from Rab35 cKO SC mice is even more severe than in Mtmr2 or Mtmr13 KO models 42,45,56,58 . Rab35-bound MTMRs (e.g. including MTMR13, MTMR5 and their associated active MTMs such as MTMR2, MTMR1, and possibly others) in this model would counteract the role of the PI(3,5)P 2 -metabolizing PIKFYVE-Vac14-FIG4 complex, a genetic interactor of Mtmr2 7 . Indeed, PIKFYVE-Vac14-FIG4 are required for oligodendrocyte differentiation and myelination in the central nervous system 7,59 .
In conclusion, our findings suggest an intimate relationship between the synthesis and turnover of PI 3-phosphates mediated by Rab35-associated MTMRs and the activity of late endosomal/ lysosomal mTORC1 that crucially controls myelin growth, thereby unraveling a common effector pathway for myelination in the peripheral nervous system. Based on this model it appears that interference with the synthesis of PI 3-phosphates, in particular of PI(3,5)P 2 and/or pharmacological perturbation of mTORC1 signaling may represent viable options for the treatment of CMT patients suffering from inherited neuropathies with abnormal myelin growth, e.g. by local administration of mTORC1 or PIKFYVE inhibitors.

Methods
Antibodies. Primary and secondary antibodies and probes used for immunoblotting and immunocytochemistry are specified in Supplementary Data 2.
Primers. Primers used for analytical (genotyping) and preparative (cloning) PCRamplification are listed in Supplementary Data 2. shRNAs. pLKO.1 encoded shRNAs, non-targeting scrambled with parallel GFPexpression (RHS6848), mouse Mtmr2 targeting shRNA (TRCN0000030098), and mouse targeting Rab35 (TRCN0000100532) were obtained from Dharmacon. Inhibitors. All inhibitors were dissolved in DMSO to stock solutions as indicated in Supplementary Table 2. Working concentrations are indicated in the method section for each experiment.
Animals. Rab35 flox/flox were generated by homologous recombination using mouse genomic DNA harboring the Rab35 locus isolated from the BAC clone RP23-192F6 (http://bacpac.chori.org/). LoxP recombination sites and a neomycin selection cassette were introduced flanking the second and third coding exons of the mouse Rab35 genomic locus. The resulting targeting vector was linearized by NotI digestion, and ES cells (F1 from 129Sv/C57BL6J) were electroporated. After positive and negative selection with Geneticin and ganciclovir, respectively, genomic DNA of surviving ES cell colonies were screened by nested long-range PCR using primers outside the 5′ (Scr-F1, Scr-F2, PL452-LoxP-Sc1R and PL452_Sc5R2) and 3′ (PL452_3′-Sc1F, PL451_3′-Sc2F, Scr-R1 and Scr-R2) arms. Targeted ES clones were expanded and confirmed by long-range PCR. Correctly targeted ES cell clones were used to generate chimeric mice at the NCI gene targeting facility. Chimeric mice were bred with β-actin-flippase (flp) transgenic mice (The Jackson Laboratory) to remove the PGKneo cassette, resulting in heterozygously floxed Rab35 flox/+ mice that were confirmed by PCR products using LoxP5F, PL452-LoxP-Sc1R, and Frt-3R primer and the lack of a product from NeoF and NeoR primers which recognize the PGKneo cassette. Litters were intercrossed to obtain homozygotes (Rab35 flox/flox ). The mice were housed in a specific pathogen-free facility and cared for in accordance with National Institutes of Health guidelines, and all protocols were approved by the National Cancer Institute Animal Care and Use Committee.
Rab35 flox/flox animals were either crossed with a tamoxifen inducible Cre ER mouse line (B6.Cg-Tg (CAG-Cre/Esr1*)5Amc/J; The Jackson Laboratory) for in vitro recombination or a P0-Cre line (ref. 44 ) for in vivo recombination. The presence of Rab35 wild-type allele in the heterozygotes of the F1 generation was detected with 'Lox5F' as forward primer and 'Lox5R' as reverse primer resulting in an amplification of 573 bp from the intron 1 region. Replacing the reverse primer by 'PL452-Loxp-Sc1R' allowed for detection of the inserted loxP site (647 bp) in the mutant floxed allele. CAG-Cre ER transgene was detected by PCR-amplification of 445 bp using 'TM63_EllaCre_fw' and 'TM64_EllaCre_rev' primer. For P0-Cre detection the following primers were used: P0_fw and P0_rev. For PCR, we isolated DNA from tail biopsies using DirectPCR lysis reagent (Viagen Biotech), following manufacturer's directions. Mtmr2 −/− mice have been already reported and characterized 45,56 .
All animal experiments were conducted according to the guidelines of the "Landesamt für Gesundheit und Soziales" (LAGeSo) and with their permission under the license S0313/17 and T0243/08, the French national regulations according to guidelines 2010/63/UE, and the Italian national regulations and covered by experimental protocols reviewed by local Institutional Animal Care and Use Committees (IACUC #894).
In vivo analyses. Animals were randomly included into experimental groups according to genotyping, age and sex. Rab35 cKO SC are mice with conditional ablation of Rab35 in Schwann cells (Rab35 flox/flox P0-Cre). These mutants were compared with controls, which correspond to Rab35 flox/flox or Rab35 flox/+ mice which are phenotypically normal. No animals had to be excluded due to illness in all the experiment performed. Investigators performing animal handling, sampling, euthanasia, and raw data analysis were not blinded.
Semithin section and ultrastructural analyses of sciatic nerves were performed as follows: 45,56 Tissues were removed and fixed with 2% glutaraldehyde (vol/vol) in 0.12 M phosphate buffer, postfixed with 1% osmium tetroxide (vol/vol), and embedded in Epon (Fluka). Semi-thin sections (0.5-1 μm thick) were stained with toluidine blue and examined by light microscopy. Ultrathin sections (70-90 nm thick) were stained with uranile acetate and lead citrate and examined by electron microscopy.To perform morphometric analysis, digitalized images cross sections were obtained from corresponding levels of sciatic nerves with a ×100 objective and Leica DFC300F digital camera (Milan, Italy). At least five images per animal were analyzed using the Leica QWin software (Leica Microsystem) to calculate the gratio that is the ratio between the mean diameter of an axon (without myelin) and the mean diameter of the same axon including the myelin sheath. To estimate the percentage of myelinated fibers carrying alterations the entire nerve section was reconstructed and the total number of myelinated fibers was assessed. For morphometric analysis on ultrastructural sections, 20-40 images per animal were taken using TALOS L120C transmission electron microscope (Thermo Fisher Scientific,Waltham, MA, USA) at 120 kV using 3400×(for P20) and the g-ratio values were determined by measuring axon and fiber diameters.
Sciatic nerve lysates from Rab35 cKO SC and controls for western blot analysis were prepared using a lysis buffer containing 2% SDS, 50 mM Tris buffer pH 8.0, 150 mM NaCl, 10 mM NaF, 1 mM NaVO 3 , complete protease and phosphatase inhibitors (Roche). Protein quantification was performed using BCA assay (Pierce, Thermo Fisher Scientific). Rapamycin (LC Laboratories) was dissolved in Ethanol and administered QD by i.p. injection 5 days a week at a final concentration of 10 µg/g in vehicle solution containing 5% Polyethylen glycol 400 (PEG 400), 5% TWEEN 80. and NaCl 0.9%, as already reported 14 .
Cell lines. GFP-Rab35 endo HeLa cells were derived from a TALEN-edited HeLa knock in (KI) cell line, which expresses endogeneous Rab35 tagged with GFP 30 . HEK293T and HeLa cells were obtained from ATCC. Cells were cultured in DMEM with 4.5 g/L glucose (Thermo Fisher) containing 10% heat-inactivated FBS (Gibco) and 100 U/mL penicillin and 100 µg/mL streptomycin (Gibco) during experimental procedures and were routinely tested for mycoplasma contamination.
Preparation of primary astrocytic cultures. Primary astrocytic cultures were prepared by isolating cortices from tamoxifen-inducible cKO Rab35 mice (Rab35 flox/flox CAG-Cre ER ) for knockout cultures and from littermates without Creexpression for WT cultures at postnatal days P2 to P5 as follows. Cerebral cortices were isolated and meninges were removed in ice-cold HBSS (Thermo Fisher). The tissue was digested for 15 min at 37°C in TrypLE (Thermo Fisher) supplemented with 150 U/mL DNase I (Sigma). Cells were dissociated by trituration in DMEM with 4.5 g/l glucose (Thermo Fisher) supplemented with 10% (v/v) fetal bovine serum (FBS) and 100 U/mL penicillin and 100 µg/mL streptomycin ('culturing medium') with 375 U/mL DNase I. Cells were pelleted by centrifugation, filtered through a 70-µM Nylon cell strainer (Sigma) and plated in culturing medium in 10 cm dishes. At day in vitro (DIV) 1, cells were washed with D-PBS (Thermo Fisher) and WT and cKO cultures were supplied with fresh culturing medium supplemented with 0.4 µM tamoxifen ((Z)-4-hydroxytamoxifen, Sigma) for knockout induction. Culturing medium supplemented with tamoxifen was exchanged every 2-3 days. Cells were passaged at DIV8 to well plates or PLLcoated coverslips at low densities to reach confluency at the day of experiment, performed at DIV20 to DIV22. For protein quantification under serum-deprived conditions, the day before lysis cells were rinsed once with D-PBS and supplied with culturing medium without FBS ('-FBS') over night. For protein quantification upon acute inhibitor treatment, astrocytic cultures were supplied with DMSO, 5 or 10 µM VPS34-IN1, 30 µM SAR405 or 30 µM Cmp-19 1 h before cell lysis.
Preparation of primary Schwann cell cultures. Schwann cells were purified from sciatic nerve and Brachius plexus, isolated from P4 to P6 mice after decapitation. Nerves were collected in Leibovitz's L-15 medium (Thermo Fisher) and remaining connective tissue and endoneurium were removed prior to digestion in L-15 supplemented with 0.1 % (w/v) Collagenase A (Sigma) and 0.25 % (w/v) Trypsin (Thermo Fisher) for 1 hr at 37°C and 5% CO 2 . After dissociation by trituration, nerve tissue was pelleted in DMEM with 4.5 g/L glucose, GlutaMAX and pyruvate (Thermo Fisher) supplemented with 5% (v/v) heat-inactivated horse serum (Thermo Fisher) and 100 U/mL penicillin and 100 µg/mL streptomycin (Gibco) ('plating medium'). One 6-well for each four dissociated nerves was coated for 3 h at room temperature (RT) with 20 µg/mL PLL (Sigma) in water and subsequently for 1 h at 37°C with 10 µg/mL Laminin (Sigma) in DMEM, before Schwann cells were added in plating medium supplemented with 5 µM cytosine β-darabinofuranoside (AraC, Sigma).
Preparation of Schwann cell/DRG organotypic explants. Myelin-forming Schwann cell/DRG neuron co-cultures were established from E13.5 mouse embryos generated from Rab35 flox/flox mice crossed with Rab35 flox/flox P0-Cre. To obtain efficient P0-Cre mediated recombination of the Rab35 flox locus ex vivo, organotypic explants were kept in Neurobasal (NB) medium supplemented with 2.5 S NGF (Nerve growth factor) at 50 ng/ml final concentration and B27 (Thermo Fisher) for 8-9 days prior to induce myelination. For myelination, C-media supplemented with ascorbic acid for additional 15 days (50 μg/ml, SIGMA) was used 60 . To perform immunohistochemistry, co-culture explants were fixed for 15 min in 4% paraformaldehyde, permeabilized for 5 min in ice-cold methanol at -20°C, blocked for 20 min with 10% NGS, 1% BSA and then incubated with primary antibodies for 1 hour at room temperature in PBS 1×(anti-Mbp) and overnight in 5%BSA, 1% NGS, 0.6% Triton in PBS1x at 4°C (anti-NF-M). After washing, the coverslips were incubated with the secondary antibody for 30 min, washed and mounted. Transduction of Schwann cell/DRG neuron co-culture explants was performed as already reported 61 . To quantify the amount of myelinated segments, using a fluorescence microscope at least 5-10 fields/coverslip were randomly acquired and Mbp-positive myelinated fibers were counted per field. Means of each coverslip/DRG have been used as different "n" for statistical analysis. To quantify myelinated fibers carrying abnormalities, at least 300 Mbp-positive myelinated fibers were evaluated, from "n" different DRG explants/coverslips using a TCS SP5 laser-scanning confocal microscope (Leica). The percentage of Mbp-positive fibers showing altered myelin among the total number of Mbp-positive fibers was indicated. Two independent experiments were performed.
Cell transfection. HeLa cells were transfected with plasmid-DNA using Jet-PRIME® (Polyplus). For siRNA-mediated knockdown in HEK293T or HeLa cells, reverse transfection using the same reagent was performed and cells were harvested after 96 h. For Large-scale transfection with plasmids for transient overexpression of proteins in HEK293T or eGFP-Rab35 endo KI HeLa cells for biochemical analysis, calcium phosphate transfection was used. For rescue experiments upon siRNA-mediated protein depletion, HEK293T cells were transfected with plasmid-DNA 48 h after reverse transfection with siRNA and harvested after transient overexpression for 48 h.
Preparation of lysates from cell cultures. Mammalian cell cultures were rinsed once with ice-cold PBS before cell lysis buffer (20 mM HEPES pH 7.4, 100 mM KCl, 2 mM MgCl 2 , 1 % (v/v) Triton X-100) freshly supplemented with 1 mM PMSF, 0.3% protease inhibitor cocktail (Sigma) and phosphatase inhibitors (coctails 2 and 3, Sigma) was applied. Cells were collected and incubated for 20 min on ice followed by centrifugation for 5 min at 17,000×g. Lysates used for affinity chromatography or immunoprecipitation were further cleared by ultracentrifugation at 180,000 × g for 15 min at 4°C in a TLA110 rotor (Sorvall).
Protein concentration was determined using Bradford assay. Upon addition of Laemmli-sample buffer (boiled 5 min at 95°C, pelleted 5 min at 17,000 × g), samples containing 10 µg to 40 µg protein amount were resolved by SDS-PAGE and analysed by subsequent immunoblotting, using HRP-coupled or LI-COR infrared-fluorescent secondary antibodies.
Immunocytochemistry of cell cultures. Cell cultures grown on coverslips were rinsed with 1x phosphate buffered saline (PBS) and fixed with 4% (w/v) paraformaldehyde (PFA) with 4 % (w/v) sucrose for 15 min at room temperature (RT). Cells were simultaneously permeabilized and blocked for 1 h in PBS supplemented with 0.3% (v/v) Triton X-100 and 10 % (v/v) normal goat serum (NGS). Primary and secondary antibodies were applied in blocking solution for 1 h at RT with PBS washes in between. Coverslips were mounted using Immu-Mount (Thermo Fisher) supplemented with 1 mg/mL DAPI (Sigma) to label cell nuclei. LAMP-2 was detected using a modified protocol: Triton X-100 was omitted. Permeabilization was performed with 20 µM digitonin in PBS for 5 min after fixation. PI(4)P and PI (3)P immunodetection was done as described 62 , for the latter using a reduced concentration of 0.05 µg/mL recombinantly expressed and purified eGFP-2xFYVE (Hrs) or a GST-tagged phox domain of P40 ('phoxP40'); a kind gift from Dr. I. Ganley, MRC Protein Phosphorylation and Ubiquitylation Unit, College of Life Sciences University of Dundee, Dundee, UK. In brief, coverslips were rinsed with 1x PBS and fixed with 2 % (w/v) paraformaldehyde (PFA) with 2% (w/v) sucrose for 15 min at room temperature (RT). Cells were washed 3x with PBS, permeabilized for 5 min in a PIPES-based buffer (PIB; 20 mM PIPES, 137 mM NaCl, 2.7 mM KCl, pH 6.8) for 5 min at RT with 20 µM Digitonin, followed by three thorough washes with PIB. The cells were simultaneously blocked and labeled for 45 min with eGFP-2xFYVE (Hrs) or phoxP40 in PIB supplemented with 5% (v/v) NGS and 50 mM ammonium chloride. After washing in PIB, primary and secondary antibodies were applied in 5% (v/v) NGS in PIB for 1 h and 45 min, respectively. The coverslips were post-fixed for 5 min with 2% PFA fixative, and washed with PBS, three times with 50 mM ammonium chloride, and once without. Mounting was performed as described above.
Confocal microscopy. Images of immunostained cell cultures were acquired on a Zeiss Laser Scanning confocal microscope (LSM) 710, routinely using a ×63 oil immersion (1.4 NA) DIC objective, and analyzed using ImageJ software, by normalizing sum fluorescence intensities of protein signals over Phalloidin positive cell area. P0-protein immunostaining in Schwann cell mono-cultures was imaged with a 10x dry (0.3 NA) objective. Sum intensities per image (10 images per condition and experiment) were normalized by the number of DAPI-positive cell nuclei per image. Whole cell images of HeLa cells or astrocytes immunolabeled for lipids were acquired with a z-stack series of usually seven slices with 0.5 µm intervals and summed up using the ImageJ 'z-projection' tool. For PI(3)P level analysis, a threshold was set to extract unspecific fluorescent signals.
Affinity chromatography with GST-coupled Rab proteins. GST-Rab fusion proteins were expressed in a E. coli Rosetta™ (DE3) (Novagen) under optimized expression conditions with 100 µM Isopropyl-β-D-thiogalactopyranosid (IPTG) at 20°C for 20 h. Bacteria pellets were either stored in PBS at −20°C or directly processed for protein purification. Nucleotide-free purification of GST-coupled Rab proteins was performed in PBS supplemented with 2 mM EDTA, 0.02% (v/v) Cyanase, 0.5 mg/mL Lysozyme, 1 mM PMSF, 1 tablet/50 mL of EDTA-free Protease Inhibitor Cocktail (Sigma), and 150 mM NaCl. Cells were disrupted by sonification with subsequent addition of 1% (v/v) Triton X-100, followed by 15 min rotation at 4°C. Bacterial lysates were cleared from cell debris by centrifugation at 35,000 × g for 15 min in a SS-34 rotor (Sorvall) at 4°C, and subjected to affinity purification. GST-bind resin (Novagen) was used according to manufacturer's instructions. In brief, 300 µl beads-slurry were washed with PBS and incubated with cleared bacterial lysates for 2 h at 4°C while rotating. Beads with bound GSTfusion proteins were stored for 20 h at maximum in a phosphate-free buffer (20 mM HEPES pH 7.5, 5 mM EDTA).
10 cm culture dishes of HEK293T cells were lysed 24 h post-transfection and cleared by ultracentrifugation as described above. For each sample, 100 µg of beadcoupled GST-fusion protein were washed three times in ice-cold 500 µl cell lysis buffer (without MgCl 2 ) and supplemented with 5 mM EDTA. GST-Rab proteins were loaded with either 1 mM GTPyS or GDP (Sigma) in 50 µl cell lysis buffer (without MgCl 2 ) at 30°C for 15 min with slight agitation. Control beads with GST alone were supplemented with 5 mM EDTA instead of nucleotides. In all, 10 mM MgCl 2 was added, followed by an incubation of 10 min at 4°C. Lysates were freshly supplemented with PMSF and 1 mM GTPyS, GDP, or EDTA, and applied to the beads. After 2 h rotation at 4°C, bead-coupled proteins were washed three times for 10 min using lysis buffer with reduced concentrations of 50 mM KCl and 0.5% (v/v) Triton X-100 and supplemented with 0.1 mM nucleotides, and a fourth time without detergent. Proteins were eluted in Laemmli-sample buffer at 95°C for 5 min, and analyzed by SDS-PAGE and immunoblotting.
Immunoprecipitation assays. Immunoprecipitation of eGFP-Rab35 from 15 cm culture dishes of eGFP-Rab35 endo KI HeLa cells, or overexpressed eGFP-Rab35CA or -DN from 10 cm culture dishes of HEK293T cells 24 h after transfection was performed using eGFP-Trap ® _MA beads (Chromotek). Cleared lysates were applied to 20 µl lysis buffer-washed eGFP-Trap ® _MA beads and incubated for 90 min rotating at 4°C. Protein-coupled beads were washed three times with lysis buffer, with reduced concentrations of 50 mM KCl and 0.5% (v/v) Triton X-100, and once without detergent. Proteins were eluted in Laemmli-sample buffer for 5 min at 95°C, analyzed by SDS-PAGE and immunoblotting.
Lentiviral production in HEK293T cells. The production of lentiviral particles containing shRNA-encoding plasmids (pLKO.1 backbone, Dharmacon), using a 2nd generation lentiviral system according to the manufacturer's guidelines (Dharmacon), and transduction of mammalian cells was done in accordance with the S2 guidelines (LAGESO). In brief, 10 cm dishes of HEK293T cells were cotransfected over night with 15 µg pLKO.1_GFP_shRNA-scrambled or pLKO.1_shRNA-msMTMR2, 10.5 µg lentiviral envelope VSV-G encoding plasmid pMD2.G and 4.5 µg lentiviral packaging-encoding plasmid psPAX2 using calciumphosphate transfection. Cells were washed with D-PBS (Thermo Fisher) and fresh cell culture medium was applied. Viral particles were harvested by collecting the culturing medium after 24 h and 48 h with subsequent centrifugation at 0.2 × g for 5 min at room temperature. The virus-containing supernatant was filtered using a 0.45-µm falcon filter and viral particles were ×100 concentrated by centrifugation in Amicon Ultra-15 100 kDa filter columns (Merck) at 5000 × g at 4°C, aliquoted and stored at −80°C. The number of transducing particles were calculated by the number of GFP-positive HEK293T cells 72 h post-transduction. Titer concentration was usually between 10 8 and 10 9 Units/mL. Viral transduction of primary cultures. Viral transduction of astrocytic cultures was performed with a multiplicity of infection (MOI) of 20. Cells were transduced in two consecutive rounds, at DIV11-12 and DIV16-17, before harvest at DIV20-22. Primary Schwann cells in monocultures were transduced with a MOI of 50 in two consecutive rounds, at DIV3-4 and DIV7-8, and fixed on DIV11-12.
Mass spectrometry of BioID samples. For liquid chromatography (LC)-mass spectrometry (MS)/MS analysis, Coomassie-stained lanes of label free BioID samples (BirA*Rab35 vs. Rab35) were cut into each 15 gel pieces per sample and digested in-gel by trypsinization. Peptides were analyzed by a reversed-phase capillary liquid chromatography system (Ultimate 3000 nanoLC system, Thermo Scientific) connected to an Orbitrap Elite mass spectrometer (Thermo Scientific). Identification and quantification of proteins were performed using MaxQuant (version 1.5.2.8) software, searched against the HUMANswiss database (April, 2014) and summarized with the software Scaffold (Proteome Software). The mass tolerance of precursor and sequence ions was set to 10 ppm and 0.35 Da, respectively. Methionine oxidation and the acrylamide modification of cysteine were used as variable modifications. Proteins with at least two detected sequenced peptides (razor unique peptides) with a minimum length of 7 amino acids were counted as biotinylated and specifically bound. False discovery rates were estimated to be <1% based on matches to reversed sequences in the concatenated target-decoy database. Label-free quantification (LFQ)-intensity ratio of BirA*-Rab35-over BirA*-biotinylated proteins was used to rank the detected proteins according to their specific Rab35-association (enrichment). Proteins that associated with BirA*-Rab35 only were considered as "selective" (Supplementary Data 1), proteins with a LFQintensity ratio (BirA*-Rab35-over BirA*) >5 as "enriched" (Supplementary Table 1).
Data and reproducibility. All data are represented as mean ± standard deviation (SD) for n = 2 independent experiments or mean ± standard error of the mean (SEM) for n > 2 independent experiments or number of animals, and normalized where indicated. GraphPad Prism 5 and 'R' software was used to perform statistical analysis. If not otherwise indicated, normalized data were analyzed using onesample two-tailed Student's t-tests. For data sets from more than two different genotypes, depicted p-values were corrected for multiple testing using Holm's multiple comparison test 63 . Level of significance are depicted by asterisks in the figures: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. n.s. non-significant. Exact p, t, and df values are provided in the respective figure legends.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support these findings are available from the authors on request. Numerical source data for Figs