TMEM106B is a major risk factor for frontotemporal lobar degeneration with TDP‐43 pathology. TMEM106B localizes to lysosomes, but its function remains unclear. We show that TMEM106B knockdown in primary neurons affects lysosomal trafficking and blunts dendritic arborization. We identify microtubule‐associated protein 6 (MAP6) as novel interacting protein for TMEM106B. MAP6 over‐expression inhibits dendritic branching similar to TMEM106B knockdown. MAP6 knockdown fully rescues the dendritic phenotype of TMEM106B knockdown, supporting a functional interaction between TMEM106B and MAP6. Live imaging reveals that TMEM106B knockdown and MAP6 overexpression strongly increase retrograde transport of lysosomes in dendrites. Downregulation of MAP6 in TMEM106B knockdown neurons restores the balance of anterograde and retrograde lysosomal transport and thereby prevents loss of dendrites. To strengthen the link, we enhanced anterograde lysosomal transport by expressing dominant‐negative Rab7‐interacting lysosomal protein (RILP), which also rescues the dendrite loss in TMEM106B knockdown neurons. Thus, TMEM106B/MAP6 interaction is crucial for controlling dendritic trafficking of lysosomes, presumably by acting as a molecular brake for retrograde transport. Lysosomal misrouting may promote neurodegeneration in patients with TMEM106B risk variants.
GWAS studies identified TMEM106B a major risk factor for frontotemporal lobar degeneration (FTLD) with TDP‐43 pathology. This study combines loss‐off‐function experiments, live‐imaging and proteomics to elucidate the function of the lysosomal protein TMEM106B in neurons.
Co‐immunoprecipitation experiments and mass‐spectrometry show that TMEM106B interacts with microtubule‐associated protein 6 (MAP6).
TMEM106B knockdown promotes retrograde transport of dendritic lysosomes and dendrite loss, which is phenocopied by MAP6 overexpression.
MAP6 knockdown rebalances dendritic trafficking of lysosomes and fully rescues the dendritic phenotype of TMEM106B knockdown.
Promoting anterograde lysosomal transport using dominant negative Rab7‐interacting lysosomal protein (RILP) rescues dendrite loss in TMEM106B knockdown neurons.
TMEM106B/MAP6 interaction controls dendritic branching, presumably by providing a molecular brake for retrograde dendritic trafficking of lysosomes.
Frontotemporal lobar degeneration (FTLD) is the third most common neurodegenerative disease after Alzheimer's disease and Parkinson's disease (reviewed in Rademakers et al, 2012). The clinical presentation is diverse and the symptoms include dementia, behavioral changes, as well as speech and language impairment. Additional symptoms of upper or lower motoneuron disease are common and indicate a partial overlap with amyotrophic lateral sclerosis (ALS). Moreover, FTLD is heterogeneous in terms of pathology and genetics. The majority of cases show neuronal cytoplasmic aggregates of the nuclear DNA/RNA‐binding protein TDP‐43 (Neumann et al, 2006). Pathogenic mutations in TARDBP, the gene coding for TDP‐43 are rare and predominantly cause ALS (Sreedharan et al, 2008). Familial forms of FTLD with TDP‐43 pathology are mainly caused by hexanucleotide repeat expansion in C9ORF72 (DeJesus‐Hernandez et al, 2011; Renton et al, 2011) and dominant loss‐of‐function mutations in the growth factor progranulin (GRN) (Cruts et al, 2006).
A multicentric genome‐wide association study identified TMEM106B as a risk factor for FTLD with TDP‐43 pathology (Van Deerlin et al, 2010). Three SNPs in linkage disequilibrium (LD) (rs1020004, rs6966915, and rs1990622) have been identified in coding and non‐coding regions of TMEM106B with genome‐wide significance. These SNPs conferred the strongest risk in patients also carrying a GRN mutation suggesting a functional interaction between TMEM106B and GRN. The genetic association of TMEM106B variants with FTLD‐TDP was replicated with high confidence (Cruchaga et al, 2011; Finch et al, 2011; Van der Zee et al, 2011). Homozygocity of the protective minor allele of the non‐coding SNP rs1990622 (7 kb downstream of the gene) is observed only in 2.6% of patients and 19.1% of controls. This protective allele is associated with higher GRN levels in plasma (Finch et al, 2011). The TMEM106B coding variant T185S, although not significantly associated with the risk of FTLD‐TDP, but in perfect LD with rs1990622, seems to slightly increase TMEM106B protein levels (Nicholson et al, 2013). Although TMEM106B variants are not associated with ALS per se, the risk allele is linked to cognitive impairment in these patients (Vass et al, 2011). The minor allele of rs1990622 protects against hippocampal sclerosis and TDP‐43 pathology in Alzheimer patients suggesting a more general role in neurodegeneration independent of GRN (Rutherford et al, 2012).
TMEM106B is a type 2 transmembrane protein and primarily localizes to late endosomes and lysosomes (Brady et al, 2012; Chen‐Plotkin et al, 2012; Lang et al, 2012). TMEM106B overexpression leads to aberrant vacuole formation and enlarged lysosomes that are less acidic, which may be accompanied by slightly enhanced GRN levels and thus argues against a gain‐of‐function role of TMEM106B risk variants in FTLD (Brady et al, 2012; Chen‐Plotkin et al, 2012). Lysosomal and autophagosomal inhibitors increase TMEM106B expression (Brady et al, 2012; Lang et al, 2012). Thus, the increased TMEM106B expression observed in GRN mutation carriers may be due to impaired lysosomal/autophagosomal function, which has been described in GRN knockout mice (Ahmed et al, 2010; Chen‐Plotkin et al, 2012; Wils et al, 2012). Interestingly, homozygous GRN mutations cause neuronal ceroid lipofuscinosis, a lysosomal storage disorder (Smith et al, 2012). Pathogenic mutations in CHMP2B and VCP further support a strong endo‐lysosomal dysfunction component in FTLD, as they affect membrane fusion and vesicle sorting within the endo‐lysosomal and autophagosomal system (Filimonenko et al, 2007; Ju et al, 2009; Urwin et al, 2010; Ritz et al, 2011).
Despite its lysosomal localization, none of the previous publications reported altered lysosomal function or morphology or any effect on GRN expression upon TMEM106B knockdown (Brady et al, 2012; Chen‐Plotkin et al, 2012; Lang et al, 2012). The physiological role of TMEM106B is therefore unknown. Here, we combine loss‐of‐function experiments and proteomics to identify the microtubule‐associated protein 6 (MAP6) as novel interacting protein for TMEM106B. We show that interaction of TMEM106B and MAP6 regulates dendritic trafficking of lysosomes and affects dendrite morphology.
TMEM106B knockdown causes lysosomal clustering in HeLa cells
TMEM106B shows a predominantly lysosomal localization, but its function in lysosomes is unknown. Transfection of HeLa cells with siRNAs against TMEM106B strongly reduced TMEM106B protein expression without affecting the pH‐dependent proteolytic maturation of the lysosomal protease Cathepsin B (Fig 1A). However, immunostaining of lysosomes against the marker protein LAMP2 revealed a striking change in lysosomal localization upon TMEM106B knockdown (Fig 1B). Lysosomes were distributed throughout the cytoplasm in control‐transfected cells, but appeared tightly clustered near the nucleus in cells devoid of TMEM106B, although the cell size was unaltered. Automated quantitative image analysis measuring the average distance of lysosomes to the nucleus and the lysosomal distribution using the Clark aggregation index (Clark & Evans, 1954) confirmed the shift of lysosomes towards the nucleus and the more compact subcellular distribution of lysosomes upon TMEM106B knockdown (Fig 1C). Overexpressing siRNA‐resistant human TMEM106B (HA‐hT106b*) in cells transfected with TMEM106B siRNA rescued normal lysosomal localization (compare HA‐T106b*‐expressing cell to the neighboring cells in supplementary Fig S1) thus confirming siRNA specificity. These data suggest that TMEM106B controls lysosomal localization and trafficking, which may be particularly important in highly polarized cells such as neurons.
TMEM106B localizes to late‐endosomes and lysosomes throughout the somatodendritic compartment in primary neurons
Since accumulation of lysosomes in the soma and proximal dendrites is a common finding in brain tissue of FTLD patients particularly with GRN mutations (Chen‐Plotkin et al, 2012; Busch et al, 2013), we analyzed the localization of TMEM106B in rat primary hippocampal neurons. Consistent with previous reports (Chen‐Plotkin et al, 2012; Lang et al, 2012), endogenous TMEM106B colocalized with LAMP1‐positive late‐endosomal/lysosomal vesicles in the cell body and in dendrites (Fig 2A). Quantitative correlation analysis confirms good colocalization of TMEM106B with LAMP1 (Pearson's coefficient 0.66 ± 0.02), but not synaptic vesicles (labeled with SV2) and early or recycling endosomes (labeled with transferrin receptor TfR) (supplementary Fig S2A and B).
To investigate the role of TMEM106B in neurons, we took advantage of lentiviral knockdown using two specific shRNAs targeting TMEM106B (shT106b#1 and #2). Cortical neurons were transduced at day 7 in vitro (DIV7) with individual shRNAs for 5 days (DIV7+5). Both shRNAs strongly reduced TMEM106B expression compared to the control shRNA (shCtrl), without affecting the expression of β‐actin, βIII‐tubulin or the FTLD‐associated proteins GRN, TDP‐43, FUS and Tau (Fig 2B). We detected no overt toxicity upon TMEM106B knockdown under these conditions using an XTT‐based viability assay (Fig 2C) indicating that loss of TMEM106B alone does not cause neurodegeneration.
To analyze the morphology of individual cells by immunofluorescence, we transfected hippocampal neurons with shRNA constructs targeting TMEM106B. The punctate somatodendritic TMEM106B staining disappeared almost completely in neurons transfected with both TMEM106B‐specific short hairpin constructs (supplementary Fig S3A). While the distribution of lysosomes showed only a subtle trend towards clustering in the soma of TMEM106B shRNA‐transfected neurons (supplementary Fig S3B), the dendritic arborization appeared less complex suggesting TMEM106B may affect the function or distribution of lysosomes particularly in dendrites. Together, these data indicate that in neurons TMEM106B is localized to lysosomes throughout the somatodendritic compartment without grossly affecting their somatic distribution and neuronal viability.
TMEM106B is essential for dendrite branching and maintenance
We observed a blunted dendritic arborization in TMEM106B shRNA‐transfected neurons compared to shCtrl‐transfected cells (supplementary Fig S3A). To quantitatively assess these changes, we cotransfected hippocampal neurons at DIV7 or 14 with the shRNA constructs together with a GFP‐expressing plasmid to outline cell morphology. Five days after transfection the most noticeable effect was significantly reduced complexity of the dendritic arbor (Fig 3A). We quantified this phenotype by Sholl analysis, which measures the number of dendrites crossing concentric circles around the cell body (Sholl, 1953). Despite the nearly normal number of primary dendrites in TMEM106B knockdown neurons, we observed a striking reduction in dendritic branching compared to control transfected neurons. This effect was similar for neurons transfected at DIV7 and 14 (Fig 3B), indicating that TMEM106B is also required for maintenance of already established dendritic arborization in mature neurons. While overall dendritic branching was less complex in TMEM10B knockdown neurons, we found that the remaining principal dendrite was longer in the shRNA‐treated cells (supplementary Fig S4A). To exclude off‐target effects, we used lentiviral co‐expression of shRNA‐resistant rat TMEM106B (T106b*) in knockdown neurons to restore basal TMEM106B levels (Fig 3C), while avoiding TMEM106B aggregation seen at higher expression levels (Brady et al, 2012; Chen‐Plotkin et al, 2012). Expression of shRNA‐resistant T106b* in TMEM106B shRNA‐transfected hippocampal neurons restored dendritic arborization (Fig 3D and E) almost to control levels, while the expression of T106b* together with the control shRNA had no effect on dendritic arborization.
To further characterize whether impairment of lysosomal function inhibits dendritic branching we transfected hippocampal neurons (DIV7+5) with wild‐type Rab7a, constitutively active Rab7a (Q67L) or dominant‐negative Rab7a (T22N), which blocks cargo transport from early to late endosomes and lysosomes and inhibits lysosomal biogenesis (Mukhopadhyay et al, 1997; Press et al, 1998; Bucci et al, 2000). Neurons transfected with wild‐type or constitutively active Rab7a appeared normal. In contrast, dendrites in neurons expressing the dominant‐negative Rab7a showed a less complex branching pattern reminiscent of TMEM106B knockdown suggesting that proper lysosomal function is required for dendrite development (supplementary Fig S4B and C, compare to Fig 3A and B).
Moreover, compared to the abundant mushroom‐shaped spines in control neurons, the dendritic protrusions in TMEM106B knockdown cells appeared less dense and thinner (supplementary Fig S4D). Reduction of the pre‐ and post‐synaptic marker proteins (synaptophysin and PSD‐95) in cortical neurons transduced with TMEM106B knockdown virus corroborate synapse loss at a biochemical level (supplementary Fig S4E and F).
We detected some endogenous TMEM106B also in the axonal compartment in developing neurons at DIV4 (supplementary Fig S5A), which allows better morphological analysis of axons than in mature neurons. Interestingly, developmental TMEM106B knockdown (DIV0+4) increased axon length by 40% (supplementary Fig S5B and C), confirming that TMEM106B knockdown has no general toxic effect, but differentially affects axons, dendrites and synapses. Due to its predominant somatodendritic localization, we focused our further analysis on the dendritic phenotype.
Taken together, the late endosomal/lysosomal protein TMEM106B is required for proper maintenance of all except the primary dendrites possibly through a specific effect on lysosomes in the dendritic compartment.
TMEM106B interacts with the microtubule‐binding protein MAP6 in the brain
In order to tie TMEM106B function to known cellular pathways, we performed immunoprecipitation experiments from P15 rat brain to identify interacting proteins using a proteomics approach. In three independent experiments with a total of five replicates the only common protein specifically identified in TMEM106B immunoprecipitates under these conditions was the microtubule‐associated protein 6 (MAP6), also known as stable tubule‐only polypeptide (STOP) (Bosc et al, 1996). Apart from microtubules, MAP6 is known to interact with the actin cytoskeleton and the Golgi apparatus (Baratier et al, 2006; Gory‐Faure et al, 2006). LC‐MS/MS analysis identified a total of 11 peptides unique for MAP6 (23% sequence coverage) in TMEM106B immunoprecipitates, but none in controls (supplementary Fig S6A). To confirm the proteomics data, we repeated the experiment and analyzed TMEM106B and MAP6 immunoprecipitates by immunoblotting. Coimmunoprecipitation in both directions corroborates the specific interaction of TMEM106B and MAP6 (Fig 4A and B).
To determine the interacting domains we transfected variants of MAP6 and TMEM106B in HEK293 cells. We expressed only the cytoplasmic N‐terminus of TMEM106B tagged with GFP to avoid aggregation of the full‐length construct. We could robustly detect GFP‐N‐TMEM106B in immunoprecipitates of rat MAP6 and the long neuron‐enriched isoform 1 of human MAP6, but not in the C‐terminally truncated isoform 2, indicating that the C‐terminal domain of MAP6 interacts with the TMEM106B N‐terminus (Fig 4C).
Consistent with previous observations (Baratier et al, 2006; Gory‐Faure et al, 2006), double immunofluorescence shows predominant microtubule‐like staining of MAP6 in dendrites and axons and vesicular staining in the soma with some overlap with TMEM106B‐positive vesicles (supplementary Fig S6B).
To further support the biological relevance of the interaction between TMEM106B and MAP6, we performed subcellular fractionation of rat brain on a discontinuous iodixanol density gradient. While ßIII‐tubulin was broadly distributed over the whole gradient, TMEM106B and MAP6 levels peaked in endosomal/lysosomal fractions (marker protein Rab7) supporting close contact of the two proteins in the same cellular compartment (Fig 4D).
Together, our data establish the physical interaction of TMEM106B and MAP6 and imply a role of TMEM106B in microtubule‐dependent processes, which are central to neurite development and maintenance (Hoogenraad & Bradke, 2009).
MAP6 knockdown rescues TMEM106B knockdown
Since TMEM106B knockdown strongly reduced dendritic branching, we asked whether MAP6 and the interaction between TMEM106B and MAP6 in particular affects dendrite morphology. Interestingly, MAP6 overexpression in hippocampal neurons (DIV7+5) strongly reduced dendrite branching and thus phenocopies the effect of TMEM106B knockdown (Fig 5A and B). We generated a shRNA construct against MAP6 to perform loss‐of‐function studies (supplementary Fig S6C). MAP6 knockdown enhanced dendritic complexity and particularly the distal branching (supplementary Fig S7A and B). This effect was fully rescued by overexpressing shRNA‐resistant human MAP6 indicating a specific effect of the MAP6 shRNA.
We therefore speculated that MAP6 knockdown might alleviate the impact of TMEM106B knockdown. MAP6 knockdown alone enhanced branching preferentially in the distal part of the dendrite (Fig 5 C and D, but see also supplemental Fig S7A and B). Strikingly, combined knockdown of MAP6 and TMEM106B fully restored also proximal dendritic branching compared to TMEM106B knockdown alone suggesting the two interacting proteins act in a common pathway (Fig 5C and D).
Since MAP6 knockdown is predicted to enhance microtubule dynamics (Bosc et al, 1996), we also tested if pharmacological destabilization of microtubules associated with enhanced dynamics could rescue the TMEM106B knockdown phenotype. Prolonged treatment with low concentrations of nocodazole (10 nM, added fresh every 36 h) for 5 days after transfection partially rescued the blunted dendritic morphology in TMEM106B shRNA transfected neurons (Fig 5E and F). The slight difference in the rescue capacity of MAP6 knockdown and nocodazole treatment might be due to additional effects apart from changing lysosomal transport and microtubule dynamics.
We conclude that the dendrite loss by TMEM106B knockdown can be rescued by additionally reducing MAP6 expression or by increasing microtubule dynamics pharmacologically, which strongly suggests that the functional interaction of TMEM106B and MAP6 is microtubule‐dependent.
TMEM106B and MAP6 regulate retrograde lysosomal trafficking in dendrites
Vesicular trafficking in dendrites and axons strongly relies on microtubules and microtubule‐associated proteins (Dehmelt & Halpain, 2005; Kapitein & Hoogenraad, 2011). To test whether the interaction between TMEM106B and MAP6 affects lysosomal trafficking in dendrites we performed live imaging of late endosomes and lysosomes labeled with Rab7a‐GFP. We analyzed TMEM106B knockdown neurons at a time point before dendritic loss becomes apparent to exclude secondary effects (DIV6+3). Accelerated movies of lysosomal trafficking revealed a population of highly mobile lysosomes in the dendrites of TMEM106B knockdown neurons that was not present in controls (supplementary Movies S1 and S2). Quantitative analysis of kymographs visualizing the trafficking of dendritic lysosomes showed a more than two‐fold increase in the number of mobile vesicles in the dendrites of TMEM106B knockdown neurons (Fig 6A and B). This effect was predominantly due to enhanced retrograde motility while the number of anterogradely transported vesicles and vesicles without net‐movement (during the 5 min recording time) remained unaffected (Fig 6B). This imbalance towards retrograde transport implies a progressive net loss of lipid membranes from distal dendrites which may be the cause of progressive dendrite loss. Reintroduction of TMEM106B by viral overexpression of shRNA‐resistant T106b* fully restored the balance of anterograde and retrograde movement and thereby excludes off‐target effects of the TMEM10B shRNA (supplementary Fig S7C and D). Both the total distance a mobile vesicle traveled and the average velocity in the 5 min interval were increased upon TMEM106B knockdown (supplementary Fig S8A and B). This effect was primarily due to increased speed and distance of retrogradely transported lysosomes. The overall density of dendritic Rab7a‐positive vesicles was not affected by TMEM106B knockdown (supplementary Fig S8C).
To exclude unspecific effects on dendritic trafficking in general, we analyzed the transport of mitochondria using dsRed fused to a mitochondrial targeting sequence. Importantly, TMEM106B knockdown had no effect on mitochondrial density and motility in dendrites supporting a specific effect of TMEM106B on lysosomal trafficking (supplementary Fig S8D and E).
As a further control for the specificity of TMEM106B knockdown on dendritic trafficking of lysosomes, we analyzed the movement of Rab7a‐GFP labeled vesicles in axons. Hippocampal neurons were nucleofected directly before plating and analyzed by live imaging at DIV4. We found no apparent difference in the number or the direction of moving lysosomes (supplementary Fig S5D and E).
Since overexpression of MAP6 reduced dendritic arborization similar to knockdown of TMEM106B, we asked whether lysosomal trafficking is similarly affected under both conditions. Indeed, overexpression of MAP6 accelerated retrograde transport of Rab7a vesicles comparable to TMEM106B knockdown without affecting anterograde transport (Fig 6C and D) and overall vesicle density (supplementary Fig S8F). Interestingly, live imaging of co‐expressed MAP6‐GFP and LAMP1‐RFP revealed that some moving lysosomes are labeled with MAP6‐GFP, indicating that excess MAP6‐GFP that can no longer bind to microtubules may still bind to TMEM106B on lysosomes (supplementary Fig S9). This implies a dominant‐negative effect of MAP6 overexpression. Thus, TMEM106B and MAP6 both regulate retrograde transport of lysosomes along dendritic microtubules.
MAP6 knockdown and nocodazole treatment rescue TMEM106B knockdown phenotype by rebalancing lysosomal transport in dendrites
MAP6 knockdown or treatment with low doses of nocodazole rescued the branching deficit in TMEM106B knockdown neurons (see Fig 5). To test whether MAP6 knockdown could also correct the enhanced retrograde transport of lysosomes, we cotransfected TMEM106B shRNA #2 together with MAP6 shRNA and performed live cell imaging of Rab7a‐GFP labeled vesicles. Surprisingly, MAP6 knockdown alone as well as double knockdown of TMEM106B and MAP6 (Fig 7A and C) enhanced lysosome motility. However, retrograde and anterograde transport were both increased to an equal level, suggesting balanced lysosomal transport is important for dendrite development and maintenance.
Similar results were obtained for the rescue with low‐dose nocodazole treatment. While untreated TMEM106B knockdown neurons showed an increased number of retrogradely moving vesicles, neurons which were additionally treated with 10 nM nocodazole for 5 days showed enhanced movement in both directions (Fig 7B and D). Thus, genetic or pharmacologic re‐balancing of dendritic transport of lysosomes rescues dendrite loss in TMEM106B knockdown neurons.
Enhancing anterograde transport of lysosomes promotes dendritic branching and rescues the TMEM106B phenotype
To further confirm that TMEM106B and MAP6 affect dendritic branching through their effect on lysosomal trafficking we aimed to manipulate lysosomal transport independent of TMEM106B and MAP6. Rab7‐interacting lysosomal protein (RILP) is known to recruit dynein‐dynactin motor complexes to lysosomes and thus promotes trafficking towards the minus end of microtubules (Cantalupo et al, 2001; Jordens et al, 2001). To investigate how RILP affects lysosomes in neurons, we co‐expressed the dominant‐negative C‐terminal fragment of RILP together with Rab7‐GFP and quantified lysosomal trafficking in dendrites. dnRILP expression specifically enhanced anterograde movement, without affecting retrograde motility (Fig 9A and B). Moreover, prolonged expression of dnRILP resulted in enhanced dendritic complexity indicating that anterograde lysosomal transport is required for dendritic branching (Fig 8C and D).
We therefore asked whether promoting anterograde lysosomal transport may rescue the impaired branching in TMEM106B knockdown neurons. Expression of dnRILP restored the balance of anterograde and retrograde lysosomal transport in neurons transfected with TMEM106B shRNA (Fig 8E and F). Importantly, dnRILP expression also enhanced dendritic complexity in TMEM106B knockdown neurons (Fig 8G and H). The functional rescue of TMEM106B knockdown by promoting anterograde lysosomal transport indicates that the misbalanced lysosomal trafficking directly reduces dendrite complexity.
Together, our data implicate TMEM106B in dendrite morphogenesis and maintenance by controlling lysosomal trafficking through its novel interacting partner MAP6. Vesicular trafficking, microtubule dynamics, and neurite development and maintenance are tightly connected and all have been linked to neurodegeneration (Garcia & Cleveland, 2001; Hoogenraad & Bradke, 2009). Thus, TMEM106B risk variants may contribute to FTLD pathogenesis through lysosomal misrouting particularly in GRN mutation carriers.
Neuronal phenotype of TMEM106B knockdown
Our data confirm the predominant localization of TMEM106B to late‐endosomes/lysosomes in primary neurons (Brady et al, 2012; Chen‐Plotkin et al, 2012; Lang et al, 2012). TMEM106B overexpression is reported to inhibit lysosomal function (Brady et al, 2012). We find that TMEM106B knockdown has no effect on the maturation of the lysosomal protease Cathepsin B indicating that lysosomal acidification is normal in the absence of TMEM106B. Thus, impaired lysosomal function upon TMEM106B overexpression may be attributed to the pronounced TMEM106B aggregation seen upon strong overexpression (Brady et al, 2012). To avoid unspecific effects, we carefully titrated the TMEM106B levels back to endogenous levels for rescue experiments in HeLa cells and neurons.
TMEM106B knockdown dramatically alters the neuronal architecture without affecting viability. While axon length in immature neurons is increased by 40%, dendritic arborization is severely blunted in mature neurons. The dominant‐negative Rab7a T22N inhibits lysosomal biogenesis and leads to dendrite loss similar to TMEM106B knockdown showing that functional lysosomes are required for dendritic arborization. In contrast, enhancing anterograde transport of lysosomes in dendrites by expression of dnRILP promotes dendritic growth. Moreover, dendritic spines appear less mature and the levels of synaptic marker proteins are reduced in TMEM106B knockdown neurons, which has also been observed in the brains of FTLD patients (Clare et al, 2010). These changes imply weakened synaptic strength and impaired synaptic plasticity, which are common attributes of neurodegenerative diseases and awaits further study in FTLD/ALS (Tackenberg et al, 2009; Petkau et al, 2012). These findings argue for an important role of TMEM106B in regulating neuronal morphology through subtle alteration in lysosomal function, most likely through altered transport.
We identified MAP6 as TMEM106B interacting partner by mass spectrometry and confirmed the interaction using bidirectional coimmunoprecipitation, immunoblotting and live imaging of lysosomal trafficking. The C‐terminal repeat region of the neuron‐enriched splice variant of MAP6 binds to the cytoplasmic N‐terminus of TMEM106B preferentially. MAP6 is a microtubule‐associated protein that has been implicated in the cold‐stability of microtubules, cellular morphology, cognition and mood (Bosc et al, 1996; Andrieux et al, 2002; Arama et al, 2012; Fournet et al, 2012). Posttranslational modifications, such as phosphorylation and palmitoylation, as well as overall protein levels have been shown to regulate MAP6 localization and target the protein also to cellular compartments other than microtubules i.e. Golgi apparatus, dendritic branch points and spines (Baratier et al, 2006; Gory‐Faure et al, 2006). Moreover, MAP6 coaggregates with neurofilaments in spheroid axonal aggregates in ALS (Letournel et al, 2003). Given the dramatic changes upon TMEM106B knockdown on dendritic arborization, linking TMEM106B to microtubules is intriguing. We show that overexpressing MAP6 phenocopies the effect of TMEM106B knockdown on dendrite morphology and that knockdown of MAP6 fully restores dendritic arborization upon TMEM106B knockdown, suggesting these morphological changes depend on the interaction of TMEM106B with MAP6.
Lysosome trafficking in dendrites
TMEM106B knockdown and MAP6 overexpression increase the number and speed of retrogradely moving lysosomes in dendrites suggesting this may be the common cause for the impaired dendritic branching. In HeLa cells, altered trafficking even leads to dramatic perinuclear clustering of lysosomes at the microtubule‐organizing center, where most of the microtubules nucleate in these cells (Akhmanova & Steinmetz, 2008; Tamura & Draviam, 2012). Restoring the balance of dendritic trafficking of lysosomes through MAP6 knockdown and nocodazole treatment rescue dendritic branching. Interestingly, comparable low concentrations of nocodazole also enhance transport of adenovirus particles (Giannakakou et al, 2002). Moreover, promoting anterograde lysosomal transport in dendrites independent of TMEM106B and MAP6 also rescues dendritic branching in TMEM106B knockdown cells, indicating that imbalanced lysosomal transport causes this phenotype.
Ample evidence suggests that an imbalance between anterograde and retrograde transport of lysosomes may impair protein and membrane turnover in the distal dendrites and thus impair dendrite and spine maintenance. The secretory pathway in general and endosomal/lysosomal trafficking in particular are important for neurite outgrowth (Horton et al, 2005; Sann et al, 2009; Jan & Jan, 2010). Interestingly, ALS‐causing mutants in the ER‐targeted protein VAPB impair membrane delivery to dendrites through its interacting protein YIF1A and thereby inhibit dendritic branching (Nishimura et al, 2004; Kuijpers et al, 2013). These findings further support overlapping pathomechanisms in the FTLD/ALS disease spectrum. Moreover, fusion of lysosomes with the plasma membrane (and secretion of lysosomal content) supplies membrane components during wound sealing and cell growth (Reddy et al, 2001; Chakrabarti et al, 2003; Huynh et al, 2004). Promoting anterograde lysosomal transport using dnRILP strongly enhances dendritic complexity. Thus, a shift towards retrograde transport of lysosomes may contribute to dendrite withering through loss of lipid membranes in distal dendrites in TMEM106B knockdown neurons.
Additionally, disturbed lysosomal trafficking may impair growth factor signaling through altered transport of receptors, such as the signaling of GRN via sortilin or tumor necrosis factor receptor (Hu et al, 2010; Tang et al, 2011). Intriguingly, MAP6 was recently also reported to interact with Intersectin1 (Morderer et al, 2012), a protein important for endocytosis and signal transduction, suggesting MAP6 may play a broader role in endosomal and lysosomal trafficking. Altered signal transduction may also contribute to increased axonal length upon TMEM106B knockdown, since axons are less dependent on membrane supply through the secretory pathway than dendrites but strongly respond to guidance cues (Ye et al, 2007).
We propose the following model for TMEM106B action centered on its role on lysosomal trafficking (Fig 9). Under normal conditions, binding of lysosomal TMEM106B to microtubule‐attached MAP6 inhibits active retrograde transport of lysosomes along dendrites (Fig 9A). Upon TMEM106B knockdown, retrogradely moving vesicles no longer bind to MAP6 and are transported with increased speed and fewer stops along the dendrites. This imbalance in late‐endosomal/lysosomal trafficking with a shift towards the somatic compartment may impair dendrite and synapse stability through loss of membranes (Fig 9B). Overexpression of MAP6 saturates the binding sites on microtubules and excess MAP6 binds to TMEM106B on lysosomes without anchoring them to microtubules which causes a dominant‐negative effect mimicking the knockdown of TMEM106B and leads to enhanced retrograde transport (Fig 9B). This model is supported by live imaging experiments showing MAP6‐GFP moving together with LAMP1‐RFP labeled lysosomes (Fig S9). Knockdown of MAP6 alone or double knockdown of TMEM106B and MAP6 slightly accelerates lysosome trafficking, but affects retrograde and anterograde trafficking to the same extent, thus preventing membrane loss in the dendrite (Fig 9C). Therefore, additional mechanisms (such as the interaction of Rab7a and RILP) likely control bidirectional transport of lysosomes in dendrites. Enhanced microtubule dynamics upon MAP6 knockdown (Bosc et al, 1996) may add to this effect because increasing microtubule dynamics with nocodazole also enhances overall motility of lysosomes and restores the balance of retrograde and anterograde trafficking in TMEM106B knockdown neurons. Thus, low concentrations of the anti‐cancer drug Vincristine, which also enhances microtubule dynamics, may be beneficial for patients with TMEM106B variants or GRN mutations, where lysosomes accumulate in the soma and proximal dendrites (Chen‐Plotkin et al, 2012).
A similar transport mechanism has already been shown for the docking of mitochondria to the axonal cytoskeleton. Interaction of the mitochondrial protein Syntaphilin with dynein light chain LC8 (independently of its motor function) stalls mitochondrial transport along axonal microtubules (Kang et al, 2008; Chen et al, 2009). Our data suggest that TMEM106B together with its binding partner MAP6 act as a molecular brake for retrogradely moving late endosomes and lysosomes in a similar manner without affecting mitochondrial transport. Since TMEM106B is predominantly localized in the late‐endosomal/lysosomal compartment, the trafficking function of MAP6/TMEM106B is likely specific to this compartment.
We and others find no effect of TMEM106B knockdown on the expression of FTLD‐associated proteins GRN, TDP‐43, FUS and Tau or neuronal viability suggesting that TMEM106B risk variants may sensitize neurons to GRN or TDP‐43‐dependent pathomechanisms, rather than causing neurodegeneration directly (Lang et al, 2012). Emerging pathological and genetic evidence supports lysosomal impairment in FTLD, particularly in patients with GRN mutations in whom TMEM106B variants confer the strongest risk (Van Deerlin et al, 2010). Homozygous loss of GRN in mice and patients leads to lipofuscin accumulation, a common feature of lysosomal dysfunction (Ahmed et al, 2010; Smith et al, 2012). Moreover, CLN3 mutations that cause juvenile onset neuronal ceroid lipofuscinosis induce perinuclear clustering of lysosomes similar to the phenotype of TMEM106B knockdown in HeLa cells (Tuxworth et al, 2009; Uusi‐Rauva et al, 2012). Interestingly, TMEM106B and presumably lysosomes accumulate in the soma and proximal dendrites in FTLD patients, most prominently in those with GRN mutations (Chen‐Plotkin et al, 2012). Thus, lysosomal misrouting may be a common pathomechanism of FTLD that is further boosted by TMEM106B risk variants.
We identified a molecular interaction partner and a very specific function in lysosomal transport for TMEM106B, a protein of previously unknown function, opening new therapeutic strategies for FTLD treatment.
Materials and Methods
Antibodies and reagents
Rabbits were immunized with an MBP‐TMEM106B‐NT fusion protein (rat amino acids 1–91) or with MBP‐MAP6‐CT (amino acids 793‐952 according to NP_058900). The serum was purified with corresponding GST fusion proteins cross‐linked to glutathione sepharose with Bis(sulfosuccinimidyl)suberate (BS3; Pierce, Rockford, IL, USA), eluted by pH shift (0.1 M glycine 0.5 M NaCl, pH 2.5) and immediately neutralized (1/10 volume 1 M Tris, pH 9.5). The monoclonal TMEM106B antibody (6F2) was described previously (Lang et al, 2012). Antibodies for LAMP1 (Enzo Life Sciences, Lörrach, Germany), LAMP2 (Developmental Studies Hybridoma Bank), β‐catenin (Sigma‐Aldrich, St. Louis, MO, USA), Cathepsin B (Santa‐Cruz Biotechnology, Dallas, TX, USA), myc (9E10, Santa‐Cruz Biotechnology), HA (3F10; Roche Applied Science, Mannheim, Germany), GFP (Neuromab, Davis, CA, USA), Rab7 (Cell Signaling Technologies, Danver, MA, USA), MAP6 (Cell Signaling Technologies and Abcam), βIII‐tubulin (Sigma‐Aldrich), β‐actin (Sigma‐Aldrich), FUS (Bethyl), TDP‐43 (Cosmo Bio, Tokyo, Japan), total‐Tau (Dako, Hamburg, Germany), Tau‐1 (Millipore, Billerica, MA, USA), Synaptophysin (Millipore), PSD‐95 (Neuromab), Na+/K+‐ATPase (Developmental Studies Hybridoma Bank), Tom‐20 (Santa‐Cruz Biotechnology), calnexin (Enzo Life Sciences), γ‐adaptin (BD Transduction Laboratories, San Jose, CA, USA), TfR (Invitrogen, Grand Island, NY, USA), SV2 (Developmental Studies Hybridoma Bank) are commercially available. Nocodazole was from Sigma‐Aldrich.
DNA constructs and lentivirus production
Rat TMEM106B cDNA was expressed from a lentiviral vector driven by human synapsin promoter. shRNAs were cloned into pSUPER (target sequences: TMEM106B #1 GCAGATTGATTATACGGTA, TMEM106B #2 GTGGAAGGAACACGACTTA, MAP6 GGTGCAGATCAGCGTGACA and luciferase control CGTACGCGGAATACTTCGA). For lentiviral knockdown the H1 promoter shRNA cassette was subcloned in a modified pLentilox 3.7 with a GFP‐cassette driven by human synapsin promoter (Edbauer et al, 2010). The TMEM106B N‐terminal fragment (AA 1‐93) was cloned into the pEGFP‐C1 vector. Rat MAP6 (NP_058900.1), human Map6 (isoform 1: NP_149052.1, isoform 2: NP_997460.1) and dnRILP (rat AA 201‐401) were cloned into GW1‐myc expression vector driven by the CMV promoter (British Biotechnology, Oxford, UK). All constructs were verified by sequencing. siRNA against human TMEM106B was purchased from Thermo Scientific (Waltham, MA, USA) (target sequence GATCAGAGATTAAGGCCAA).
Lentivirus was produced in HEK293FT cells cotransfected with psPAX2, pVSVg and the respective overexpression or knockdown constructs (Orozco et al, 2012). After harvest, the supernatant was concentrated by ultracentrifugation and the virus particles resuspended in Neurobasal medium.
Cell culture and transfection
Hippocampal and cortical neurons were prepared from embryonic day 18 Sprague–Dawley rats and transfected with Lipofectamine 2000 or transduced with lentivirus as described before (Orozco et al, 2012). For the analysis of axonal morphology and lysosomal trafficking in axons, hippocampal neurons were held on astrocyte feeder cultures using N2 medium and transfected before plating with an Amaxa 4D‐Nucleofector (Lonza, Basel, Switzerland) with primary culture kit P3 (Orozco et al, 2012). Immunofluorescence and immunoblotting was performed as described previously (Orozco et al, 2012). Images were taken on Zeiss LSM 510 or 710 laser scanning microscopes (Zeiss, Jena, Germany) using 40× and 63× objectives with 1 Airy unit pinhole. HeLa cells were transfected with X‐tremeGENE 9 DNA Transfection Reagent (Roche Applied Science) for 72 h and either processed for immunoblotting or immunocytochemistry as described before (Lang et al, 2012). For coimmunoprecipitation experiments, HEK293FT cells were transfected with Lipofectamine 2000 according to the manufacturer's instructions.
Cell viability assay
Neurons were plated in 96‐well plates and transduced with the indicated lentiviruses for 5 days. XTT assay (Roche Applied Science) was performed according to the manufacturer's instructions.
Quantitative analysis of colocalization
The amount of colocalization of two fluorescent signals in a confocal image was analyzed using Pearson′s coefficient estimated with the JaCoP plugin (Bolte & Cordelieres, 2006) of ImageJ.
Live cell imaging and quantification of organelle movement
Time lapse images were taken on a Zeiss Cell observer SD spinning‐disc microscope with an air‐cooled Evolve 512 EMCCD camera at 1 Hz for 5 min. Neurons were kept in a climate chamber (37°C, 5% CO2) during image acquisition. Kymographs of vesicular movement from the axon or 3–5 dendrite segments per cell were generated and manually analyzed using ImageJ software (Multiple Kymograph plugin by J. Rietdorf and A. Seitz). For the analysis of average run length and velocity only vesicles with more than 2.5 μm run length were included.
For the automatic analysis of lysosomes, projected images were segmented with Definiens Developer XD (version 2.02; Definiens AG, Munich, Germany). Nuclei regions were identified in the DAPI channel by automatic thresholding. Touching objects were separated by using the shape split function and subsequent erosion and dilation of the nucleus objects. Cell regions were identified based on β‐catenin staining by a watershed algorithm using nucleus objects as seeds. Cell borders were smoothed by subsequent erosion and dilation. Lysosome region was identified in each cell object by automatic thresholding of the LAMP2 channel. The mean distance of lysosomes from nucleus was calculated by averaging the distance of each pixel belonging to the lysosome region to nucleus border. Aggregation of lysosomes was quantified with the Clark aggregation index (CAI). For each object, D(x) denotes the spatial distance to the next neighbor of pixels belonging to the lysosome region, N(x) denotes the number of pixels belonging to the lysosome region and A denotes the number of pixels of the cell object. The CAI is defined as [sum(D(x))/N(x))*sqrt(N(x)/A)].
Dendritic complexity was quantified using manual Sholl analysis using MetaMorph software (Molecular Devices, Sunnyvale, CA, USA). Concentric circles were laid around the cell soma from 12.5 to 112.5 μm (in 12.5 μm intervals) from the soma. The number of dendrites crossing each circle was counted. For morphological analysis of axons, the length of the tau‐1 positive neurite was measured using the AxioVision software. All image acquisition and quantification for morphological analyses were done blind to the experimental conditions.
Immunoprecipitation from rat brain
P15 Sprague‐Dawley‐rat brains were homogenized in 0.32 M sucrose, 4 mM Hepes, 2 mM EDTA (pH 7.4). After centrifugation (75 600 g, 30 min) the pellet was lysed with 1% Triton X‐100 in PBS for 20 min at 4°C. Subsequently, the soluble fraction (100 000 g, 20 min) was subjected to immunoprecipitation with TMEM106B, MAP6 or an unspecific control antibody (anti‐GST) coupled (BS3; Thermo Scientific) to Dynabeads (Life‐Technologies, Carlsbad, CA, USA) for 1.5 h at room temperature. After several washing steps with lysis buffer protein was eluted from the beads with 50 mM glycine, pH 2.8 for Coomassie staining (NOVEX colloidal blue staining kit; Invitrogen) and subsequent LC‐MS/MS analysis, or directly boiled in sample buffer (4% SDS, 20% glycerol, 5% β‐mercaptoethanol, 200 mM sodium phosphate pH 7.4) for immunoblotting. Protease and phosphatase inhibitor were present in all steps of the immunoprecipitation.
Coimmunoprecipitation from HEK293FT cells
HEK293FT (Invitrogen) cells were transfected with the respective constructs for 48 h using Lipofectamine 2000 (Invitrogen). Cells were washed with PBS, lysed with 1% Triton X‐100 in PBS and the lysate centrifuged (40 min, 17 000 g). The supernatant was diluted with PBS to 0.5% Triton X‐100, precleared for 30 min with Protein A sepharose beads and afterwards subjected to immunoprecipitation with 30 μl myc‐agarose beads (Sigma‐Aldrich). Samples were washed four times with 0.5% Triton X‐100 in PBS before loading. In all steps, protease and phosphatase inhibitors were present.
Tryptic in‐gel digestion of the proteins was performed as described (Shevchenko et al, 2006). Peptides were analyzed by an LC‐MS/MS set‐up coupling a Proxeon Easy nLCII (Thermo Fisher Scientific) with in‐house packed 15 cm columns (2.4 μm C18 beads, Dr. Maisch GmbH) to an LTQ Velos Orbitrap mass spectrometer (Thermo Scientific). A bilinear gradient of 60 to 85 min was applied for peptide separation; collision induced dissociation (CID) was applied for fragmentation of TOP14 peptides. Data analysis was performed using the Proteome Discoverer 1.2 (Thermo Scientific) with the embedded SEQUEST algorithm for peptide identification. The International Protein Index database for rat (version 3.87) was used for the database search with carbamidomethylation of cysteine as a static and oxidation of methionine as a dynamic modification. Only full tryptic peptides with a maximum of 2 missed cleavages and an FDR below 5% were allowed.
An adult rat brain was homogenized in 20 ml buffer (10 mM Hepes, 1 mM EDTA, 0.32 M sucrose, pH 7.4). Postnuclear supernatant (1500 g, 10 min) was centrifuged (100 000 g, 30 min) and the membrane fraction was resuspended in homogenization buffer (10 mM Hepes, 1 mM EDTA, 0.25 M sucrose, pH 7.4). 1/10 volume was loaded on a step gradient (2.5, 5, 7.5, 10, 12.5, 15, 17.5, 20 and 30% iodixanol in homogenization buffer). After centrifugation (40,000 rpm, 2.5 h in a TH‐641 rotor) 1 ml fractions were collected by needle puncture at the bottom.
Supplementary Figure 1
Supplementary Figure 2
Supplementary Figure 3
Supplementary Figure 4
Supplementary Figure 5
Supplementary Figure 6
Supplementary Figure 7
Supplementar Figure 8
Supplementary Figure 9
Supplementary Movie 1
Supplementary Movie 2
We thank J. Banzhaf‐Strathmann, Eva Bentmann, S. Hassan, B. Schmid and C. Wahl‐Schott for critical comments. We thank Andrea Wenninger‐Weinzierl, and Nagore Astola for technical assistance. Automated image analysis was performed by C. Moehl at the DZNE light microscopy facility. CML was supported by a fellowship of the Hans and Ilse Breuer Foundation. DE was supported by the Helmholtz Young Investigator program HZ‐NG‐607. We thank the Competence Network for Neurodegenerative Diseases (KNDD) of the Bundesministerium für Bildung und Forschung (BMBF) for support to CH and SFL. The research leading to these results has received funding from the European Research Council under the European Union's Seventh Framework Programme (FP7/2007‐2013) / ERC grant agreement no. 321366‐Amyloid to CH. SFL was supported by the JPND‐RiMOD program.
The authors declare that they have no conflict of interest.
- Received May 31, 2013.
- Revision received October 7, 2013.
- Accepted October 27, 2013.
- © 2013 The Authors