ESCRT-III-associated proteins and spastin inhibit protrudin-dependent polarised membrane traffic

Antibodies

Rabbit polyclonal anti-spastin (86-340) was produced as previously described [23]. Mouse monoclonal anti-spastin (6C6) and mouse monoclonal anti-Acetylated tubulin (6-11B-1) were obtained from Sigma-Aldrich. Mouse monoclonal anti-myc (4A6) was obtained from Millipore. Rabbit polyclonal anti-myc tag (ab9106), rabbit anti-GFP (ab6556) and rat anti-alpha-tubulin-tyrosinated clone YL12/2 (ab6160) were obtained from Abcam. Rat anti-RFP (5F8) was obtained from Chromotek. Goat anti-BMPR2 (EB08820) and goat anti-Kinesin-1/KIF5 (EB05492) were obtained from Everest Biotech. Rabbit polyclonal anti-ZFYVE27 (12680-1AP) and anti-IST1 (51002-1-AP) were obtained from Proteintech. Mouse monoclonal anti-TfnR (13-6800) was obtained from Invitrogen. Rabbit polyclonal anti-GAPDH (2118) was obtained from Cell Signaling Technology. Rabbit polyclonal anti-CHMP1B was produced as previously described [15]. Peroxidase conjugated secondary antibodies for western blotting were obtained from Sigma-Aldrich. Alexafluor-labelled secondary antibodies for immunofluorescence were obtained from Molecular Probes.

Constructs and generation of stable cell lines

pLXIN-myc-Spastin M1 wild-type (wt) and M87 wt, K388R and F124D constructs for generation of stable cell lines were made as previously described [15]. Briefly M87 or M87A-M1 Spastin was cloned into pIRESneo2 (Clontech) followed by insertion of a myc tag (mutagenesis of the M87 residue ensures that no M87 transcript is produced from the M1 construct). Myc-spastin was further cloned into SalI-BamHI sites in pLXIN. pLXIN-myc-M87-Spastin-K388R and F124D mutants were then generated and all constructs were made resistant to spastin siRNA 1 and 3 by introducing two mutations into each of the relevant sequences by site-directed mutagenesis. Myc-BMRP2 stable cell lines were generated by PCR of the myc-BMRP2 construct described in [22] and cloning the myc-BMPR2 fragment into pLXIN, followed by retroviral transduction of HeLa cells.

Cell culture, transfection and siRNAs

HelaM and MRC5 cells were maintained as previously described [23]. HeLaM cells stably expressing spastin constructs were additionally cultured in the presence of 500 μg/ml Geneticin (Invitrogen). Mouse embryonic fibroblasts (MEFs) were obtained from E14 mouse embryos and cultured in DMEM containing 10% foetal calf serum, 1% penicillin–streptomycin (Invitrogen), 1% l-glutamine (Invitrogen), 1% MEM non-essential amino acids (Invitrogen) and 0.1% 2-mercaptoethanol (Invitrogen). For transient transfections of plasmid DNA, HeLaM cells were transfected using Effectine (Qiagen) following the manufacturer’s protocol, and incubated for 24 h before analysis. For siRNA transfection, cells were transfected with the relevant siRNAs, using Oligofectamine transfection reagent (Invitrogen), according to a 5-day protocol modified from [24]. Briefly, HeLaM or MRC5 cells were plated into a well of a 6-well plate and transfected with siRNA after 24 h. Cells were harvested 96 h later. The efficiency of siRNA knock-down was verified by immunoblotting cell lysates and/or by immunofluorescence microscopy of fixed cells, with an antibody against the relevant protein, or in the case of some protrudin depletion experiments (performed when an endogenous protrudin antibody was not available), by carrying out parallel knock-down experiments in cells transiently expressing GFP-protrudin and validating protein knock-down using an anti-GFP antibody for immunoblotting. The following siRNAs were obtained from Dharmacon and used at the indicated concentrations:

• Spastin (5 nM), siRNA1: 5′-GAACUUCAACCUUCUAUAA, siRNA3: 5′-UAUAAGUGCUGCAAGUUUA.

• Protrudin (20 nm), siRNA3: 5′-GUAACCAGACCUUGAGCAA; siRNA4: 5′-ACAAGAGGCUGGAGAUCUA.

• KIF5B (10 nm) siRNA1: 5′-GCAGUCAGGUCAAAGAAUA; siRNA custom2: 5′-UGAAUUGCUUAGUGAUGAAUUUU.

• IST1 (25 nM), siRNA1: 5′-CCAAGUAUAGCAAGGAAUA-3′ (D-020977-01); siRNA3, 5′-GCAAAUACGCCUUUCUCAU-3′ (D-020977-03).

• CHMP1B (10 nM), siRNA: 5′-GAAGAUUUCUGCUUUGAUGUU-3′ (D-004698-01); siRNA6: 5′-GUCGAUGGCUGGUGUGGUU-3′ (J-004698-06); siRNA7: 5′-CCUUCGGGAUCAAGUGUGA-3′ (J-004698-07).

Immunofluorescence microscopy of HeLa cells and MEFs

For immunofluorescence co-localisation cells were either fixed at room temperature in 3.7% (v/v) formaldehyde in phosphate-buffered saline (PBS) and permeabilised in PBS containing 0.1% (v/v) Triton-X100 (Sigma) or fixed and permeabilised for 5 min with ice-cold Methanol (Fisher Scientific). Where pre-fixation extraction of cytosolic protein was required, this was carried out using a saponin-based extraction buffer as previously described [25]. Fixed cells on coverslips were blocked in 5% foetal calf serum in PBS for 1 h and incubated with primary antibodies and Alexa Fluor^® conjugated secondary antibodies in blocking buffer and mounted onto glass slides using Prolong^® Gold antifade reagent with DAPI (Molecular Probes). Cells were visualised using a Zeiss LSM880 confocal microscope using a 63 × oil immersion objective and images captured using Zen software. To quantify the degree of co-localisation between individual pixels in different fluorescent channels, a Pearson’s correlation coefficient with Costes automatic thresholding was calculated using Volocity Software (Quorum Technologies) for individual cells in each image. Images of SNX1 immunofluorescence (i.e., for endosomal tubule quantification) were obtained with a Zeiss AxioImager Z2 Motorized Upright Microscope (63 × NA 1.40 oil immersion objective, room temperature, Zeiss Axiocam 506) and deconvolved with Huygens Professional version 18.04 (Scientific Volume Imaging, The Netherlands, http://svi.nl). Images were subsequently processed using Adobe Photoshop and presented using Adobe Illustrator.

Protrusion quantification in HeLa Cells and MEFS

HelaM cells were plated at low density (1 × 10⁴) onto poly-d-lysine (Sigma-Aldrich)-coated coverslips in six-well plates. 24 h later, cells were transfected with spastin siRNA with or without rescue siRNA oligonucleotides and incubated for a further 96 h. Cells were then fixed and permeabilised in ice-cold methanol and labelled with anti-acetylated tubulin primary antibodies and Alexa Fluor^® conjugated secondary antibodies. Coverslips were mounted onto glass slides using Prolong^® Gold antifade reagent with DAPI (Molecular Probes) and cells were visualised using a Zeiss Axioimager Z2 Wide-field upright microscope with a motorised stage and 40 × oil immersion objective. Some images were deconvolved as above. Stitched images of 5 × 5 individual tiles were automatically captured using Zen Pro analysis software and cells in 25 automatically captured fields were analysed per coverslip. Protrusions were defined as cell extensions which were acetylated tubulin positive with a length greater than 2 × the widest diameter of the nucleus, a stringent criterion compared to previous publications that have used a process length greater than 1 × the nuclear diameter [26]. In cells with more than one protrusion, the longest protrusion only was counted. The length of each protrusion was manually measured using the spline curve tool in Zen from the edge of the nucleus to the longest point away from the nucleus identified by acetylated tubulin immunofluorescence. The total number of cells in 25 fields per coverslip was then automatically counted using a nucleus detection protocol or using cellular detection with a whole cell stain (Thermo Scientific) in Zen and the percentage of total cells with protrusions was calculated. For each independent experiment, between 300 and 1000 cells were analysed per siRNA and at least three independent experiments were carried out. For quantification of protrusions in MEFs, the same protocol and measurement parameters were used and in addition protrusions were only counted if the diameter of the half way point along the length of each protrusion was equal to or less than 10 µm.

Protrusion rescue experiments in stable cell lines

For protrusion rescue experiments in myc-spastin stable cell lines, Zen Pro analysis software was first used to set a fluorescence intensity threshold in the fluorescent channel of the myc tag to identify the total number of stably transfected cells in 25 fields. Protrusions were then counted in all cells identified as stably transfected and the number of stable expressing cells with protrusions was expressed as a percentage of the total.

Western blotting

Cells were washed twice with PBS and lysed in ice-cold lysis buffer (25 mM HEPES pH 7.4, 1 mM EDTA, 150 mM NaCl, 0.5% TX100, protease inhibitors). Lysate was centrifuged at 13,000 rpm for 15 min and protein concentration in the supernatant was quantified using BCA protein assay (Pierce). 30 µg of protein was resolved using SDS-PAGE and protein gels were blotted onto Nitrocellulose membrane and blocked in 5% non-fat dried milk in TBS-Tween. Membrane was incubated with primary antibodies and HRP-conjugated secondary antibodies in blocking buffer and HRP-signal detected with SuperSignal ^® West Pico Chemiluminescent Substrate using X-Ray film (Fuji Medical).

Statistical analysis

Statistical analysis was performed using unpaired two-tailed Student’s t tests or one-way ANOVA with Dunnett’s or Tukey’s post hoc tests for multiple comparisons, using GraphPad Prism 5.01 for Windows (GraphPad Software, San Diego) statistical software. Bars in all histograms represent mean ± standard error of the mean; bars in dot plot represent mean ± standard deviation.

Animals

Mice were maintained in accordance with UK and European Union regulations. Animal work for this study was approved by the University of Cambridge Ethical Review Committee and was performed under project licenses (80/2304 and 70/7888) granted by the UK Home Office under the Animals (Scientific Procedures) Act 1986. The University of Cambridge is a designated establishment for breeding and scientific procedures under the Animals (Scientific Procedures) Act 1986. Spastin knock-in mice were housed, maintained on a C57BL/6J background and bred to generate embryos by heterozygous breeding as previously described [27].

Embryonic fibroblasts from a spastin ATPase-defective knock-in mouse model develop protrusions

We began by examining the morphology of murine embryonic fibroblasts (MEFs) derived from a spastin mouse model (spastin^N386K) knocked in for an HSP-associated mutation that renders the protein ATPase deficient and unable to sever microtubules. These mice express mutant spastin at endogenous levels and so results are physiologically relevant. In addition, homozygous mutant mice develop a gait phenotype consistent with human HSP and cultured primary neurons from these mice show pathological axonal swellings; thus, these mice appear to be an excellent HSP disease model [17, 27].

We found a significant increase in cellular protrusions in MEFs derived from spastin^N384K/N384K mice versus controls (Fig. 1a, b). The protrusions were filled with linear arrays of microtubules and were defined as cell extensions which were tubulin positive and with a length greater than 2 × the diameter of the nucleus of the cell from which it arose.

We validated the relevance of these findings to humans using HeLa cells, a line that has already been shown to generate protrudin-dependent protrusions [19]. Using two different siRNA oligonucleotides, we saw a highly significant increase in the proportion of cells with protrusions in cells lacking spastin (Fig. 1c–e). These protrusions were typically many times longer than the diameter of the cell body from which they originated (> 60% were longer than 80 μm, i.e., representing 4 × the typical nuclear diameter) and were filled with a linear array of microtubules (Fig. 1c). The microtubules were labelled strongly with antibodies for tyrosinated tubulin, an indicator of dynamic microtubules that are prone to growth and shrinkage, and especially for acetylated tubulin, an indicator of long-lived structures that are associated with long-range transport (Fig. 1f) [28]. These tubulin modifications in protrusions were not associated with alterations in total cellular content of acetylated or tyrosinated tubulin, as measured by immunoblotting (Fig. 1g). The protrusions were also clearly visible upon cellular labelling with a whole cell stain or with the actin marker phalloidin (Supplementary Figure 1). Markers of endosomes accumulated at protrusion tips, including RAB11 (which was present in every protrusion where it was analysed) (Fig. 1h) and transferrin receptor, as well as SNX1 (a marker of endosomal tubules) and RAB5 and 7 (early and late endosomal markers, respectively; not shown). We also observed protrusion formation when spastin was depleted from MRC5 cells, a human fibroblast line, confirming the relevance of this phenotype to cells from at least two human germ layers (Supplementary Figure 1).

We concluded that spastin inhibits protrusion formation in mice and humans, and that this requires its ATPase activity, and hence its ability to modify microtubules.

ESCRT interaction as well as ATPase activity is required for spastin to regulate polarised membrane traffic in human cells

To analyse further the functional requirements for spastin to prevent protrusion formation, we carried out siRNA-rescue experiments in HeLa cell lines stably expressing siRNA-resistant forms of human spastin.

Mammalian cell types express at least two spastin isoforms, a less-abundant full-length form (M1-spastin) and a more highly expressed isoform (M87-spastin) that lacks the N-terminal 86 amino acids of the full-length protein (Fig. 2a) [29]. M1-spastin localises to the ER via a predicted hairpin membrane domain that is specific to this isoform [23, 30]. Here, it participates with the endosomal ESCRT-III protein IST1 in ER–endosome contacts that promote endosomal tubule fission [17]. This M1 isoform (but not the M87 isoform) interacts with protrudin and several ER-resident proteins that have been implicated in HSP, very likely via hairpin membrane domain interactions [30–34]. Although M87-spastin is predominantly cytosolic at steady state, it can also be recruited to endosomes (and presumably ER–endosome contact sites) by ESCRT-III interactions [23, 35].

SiRNA-rescue experiments in HeLa cells have shown that either isoform is sufficient to rescue endosomal tubule fission defects in cells lacking endogenous spastin [15, 17]. Consistent with this, the cellular protrusion phenotype was rescued by either M1- or M87-spastin (Fig. 2b–d, g, h). In contrast, and consistent with the MEF data above, a form of M87-spastin containing the ATPase-deficient disease-associated mutant K388R was unable to rescue the protrusion phenotype (Fig. 2a, e, g, h). This mutant is able to bind but not hydrolyse ATP and so is unable to modify microtubules, nor is it able to promote efficient endosomal tubule fission [13, 15]. Spastin binds to ESCRT-III-associated proteins via interactions with its microtubule interacting and trafficking (MIT) domain (Fig. 2a), and introduction of the MIT F124D mutation, which both blocks these interactions and fails to support endosomal tubule fission, similarly caused failure to rescue protrusion formation (Fig. 2f–h) [36]. Finally, a form of M87 spastin lacking a region that is necessary and sufficient for microtubule binding (myc-M87-spastin∆MTB, Fig. 2a) was also unable to rescue protrusion formation (Fig. 2i–m) [37]. We concluded that either spastin isoform is sufficient to inhibit protrusion formation, although neither was specifically necessary, and that this property of spastin requires its ability to bind microtubules, hydrolyse ATP (and so modify microtubules), and to interact with ESCRT-III. Thus, the properties of spastin that are required to prevent protrusions are the same as those previously identified as being required for it to promote endosomal tubule fission [15, 17].

Spastin-binding ESCRT-III-associated proteins inhibit protrusion formation

The spastin MIT domain is able to interact with two ESCRT-III-associated proteins, IST1 and CHMP1B [35, 36, 38]. As spastin required the capacity to interact with ESCRT-III to suppress protrusion formation, we tested the effect of cellular depletion of these proteins and found that lack of either protein caused cellular protrusions to develop (Fig. 3a–c).

We showed previously that IST1 is involved with spastin in regulation of endosomal tubule fission [15, 17]. This caused cells lacking IST1 to develop increased numbers of SNX1-positive endosomal tubules. As IST1 and CHMP1B have been proposed to act together as a tubule fission machinery, somewhat surprisingly in this previous work, we found no significant effect of CHMP1B depletion on endosomal tubule numbers [39]. We revisited this using a more sensitive assay, reporting the percentage of cells with at least one long endosomal tubule, and found that cells lacking CHMP1B also showed an increase in endosomal tubulation (Supplementary Figure 2). Thus, the effects of spastin, IST1 and CHMP1B depletion in inhibiting endosomal tubule fission mirror their effects on inhibiting protrusion formation.

Protrusion formation in cells lacking spastin requires protrudin and KIF5A

We examined whether the protrusions seen in cells lacking spastin were generated by the protrudin-KIF5-dependent mechanism proposed by Raiborg et al. [19]. We began by examining whether the subcellular localization of protrudin and KIF5 was affected by spastin depletion. Protrudin and KIF5 were both present in protrusions that developed in HeLa cells lacking spastin (Fig. 4a; Supplementary Figure 3). KIF5 and protrudin showed minimal co-localisation in wild-type cells, but punctate co-localisation was observed in cells lacking spastin (Fig. 4a, b; Supplementary Figure 3). Co-localised KIF5 and protrudin puncta were often arranged in linear arrays within the protrusions, probably representing alignment along a microtubule, and in some cases co-localised with an endosomal marker (Fig. 4a, c; Supplementary Figure 3). Thus, spastin depletion promotes co-localisation of protrudin and KIF5A in cellular protrusions.

We then analysed whether protrudin or KIF5 are required for the protrusion formation in cells lacking spastin. Using double siRNA knock down experiments in HeLa cells, we found that the spastin protrusion phenotype was efficiently rescued by concurrent knock-down of KIF5B or protrudin, using two independent siRNAs to target each transcript (Fig. 5a–c; Supplementary Figure 4).

Spastin controls the polarised traffic of membrane receptors implicated in HSP

Having shown that cells lacking spastin develop a polarised morphology, we examined whether this affected polarisation of cell surface receptors. Upregulated Bone Morphogenetic Protein (BMP) signalling, caused by altered endosomal or secretory pathway trafficking of BMP receptors (BMPRs), has been proposed as a potential unifying mechanism in several subtypes of HSP [3, 22]. With specific reference to spastin, zebrafish lacking the M1-spastin isoform have motor axonal mistargeting that is caused by upregulated BMP signalling [40], and we have shown that HeLa cells lacking the protein have upregulated BMP signalling, without alteration in the total cellular amount of the BMP receptor BMPRII [22]. We, therefore, examined whether this upregulated BMP signalling might be associated with a more polarised distribution of BMP receptors.

At steady state in wild-type cells, epitope-tagged BMPRII is present at the plasma membrane and in early and recycling endosomes [22]. Consistent with this, in wild-type HeLa cells both endogenous BMPRII and stably expressed myc-tagged BMPRII were present at the plasma membrane in a non-polarised distribution, and also in intracellular vesicles that showed occasional alignment with microtubules (Fig. 6a, b). However, in cells lacking spastin, both endogenous and stably expressed BMPRII became much more polarised, with strong signal being present on the plasma membrane of protrusions, especially at their tip (Fig. 6c, d). BMPRII signal was also present intracellularly within the protrusions, where it aligned with markers of microtubules and KIF5. Although we cannot completely exclude this overlap as being the result of the concentration of the markers within a confined cellular space, it is consistent with the idea that BMPR-containing endosomes are subject to KIF5-dependent microtubule-based transport into the protrusions (Fig. 6e, f). In addition, in cells lacking spastin, there was generally increased colocalisation between endogenous BMPRII and protrudin (Supplementary Figure 5).

We confirmed the physiological relevance of these observations in MEFs from spastin^N384K/N384K mice, where we found that endogenous BMPRII had a highly polarised distribution at the tips of the protrusions that developed in heterozygous, and especially homozygous, knock-in mice (Fig. 7).