Filopodia formation and endosome clustering induced by mutant plus-end–directed myosin VI

To investigate the requirement of minus-end–directed movement of MYO6 for cellular function, we reengineered MYO6 into a plus-end–directed motor by replacing the unique insert and neck regions with the myosin V lever arm (Fig. 1A). Similar replacements in porcine MYO6 have been previously shown to lead to plus-end–directed movement in vitro (10). When this construct was transfected into HeLa cells, we observed a dramatic reorganization of the actin cortex with formation of filopodia-like (11) protrusions of actin, where MYO6+ accumulated at the tips (Fig. 1B). This observation is reminiscent of the previously reported accumulation of myosin X (12) and dimeric myosin VIIA (13) at tips of filopodia, indicating that MYO6+ was acting as a processive plus-end–directed motor in cells. A shorter replacement of the unique MYO6 neck region with only two isoleucine–glutamine calmodulin binding motifs (IQ) motifs (retaining the lever arm extension) led to a similar formation of filopodia (Fig. S1 A and B). To confirm whether MYO6+ was indeed accumulating at actin filament plus ends, we cotransfected GFP-MYO6+ with myosin X (MYO10) labeled with-mCherry (Fig. 1C). MYO6+ was observed to colocalize with MYO10, a protein known to induce filopodia and accumulate at filopodia tips (12). MYO6+ had a similar propensity to induce filopodia as MYO10 in HeLa cells (Fig. 1D), indicating that MYO6+ was functioning in a similar capacity. The actin-rich protrusions induced by MYO6+ were positive for fascin along the shaft (Fig. 1E), indicating they are indeed filopodia. We extracted the lengths of filopodia from total internal reflection fluorescence (TIRF) images and found they adopted a log-normal distribution (Fig. 1F), as previously reported for both filopodia (14) and neuronal spines (15). The mode length of this distribution is 5.66 ± 0.14 µm compared with 3.22 ± 0.04 µm in cells transfected with GFP-MYO6. The average length of filopodia in cells expressing GFP-MYO6+ is 6.3 ± 3.7 µm (SD, n = 233) compared with 2.95 ± 1.1 µm (SD, n = 149) in cells expressing wild-type GFP-MYO6 (P < 0.0001) and 5.3 ± 2.9 µm (SD, n = 180) in cells expressing GFP-MYO10 (in agreement with a previous report, ref. 12), indicating the activity of MYO6+ is capable of significantly increasing the length of filopodia. To assess the requirement for endogenous MYO6 in this process, we expressed MYO6+ in HeLa cells rendered null for MYO6 by modification with CRISPR/Cas9 (16). We found no difference in the propensity of GFP-MYO6+ to induce filopodia in these cells compared with wild-type HeLa cells, indicating that indeed expression of MYO6+ alone leads to the formation and elongation of filopodia with MYO6+ localized to the tip (Fig. S1C). Expression of GFP-MYO6+ did not affect the level of endogenous MYO6 (Fig. S1D).

Given that MYO6+ dramatically reorganized cortical actin structures and displayed a very different localization compared with MYO6, we examined the distribution of well-characterized MYO6 binding partners. The endosomal protein GAIP-interacting protein, C terminus (GIPC) (17), which typically colocalizes with MYO6 on adaptor protein, phosphotyrosine interacting with PH domain and leucine zipper 1 (APPL1)-positive signaling endosomes at the cell periphery, accumulated with MYO6+ at the tips of filopodia (Fig. 2A). In contrast, the endosomes themselves, as detected with an antibody against APPL1, clustered at the base of the filopodia (Fig. 2B). This finding indicates that APPL1 endosomes are guided to the base of the filopodia by MYO6+/GIPC complexes, but are unable to move into the filopodial shaft, as it is tightly packed with protein and surrounded by membrane. MYO6+/GIPC complexes then detach from these endosomes and move toward the filopodia tips. Examining these structures by structured illumination microscopy (SIM), we observed many of these tips were apparently extended and curved (Fig. S2A). Furthermore, APPL1 endosomes clustered in an actin-rich region from which filopodia emanate (Fig. S2B).

Interestingly, the MYO6 endosomal binding partner TOM1 (18) can only accumulate at the base of filopodia but not at the tips (Fig. S3A), indicating it does not detach from signaling endosomes. MYO6 and its binding partners often undergo transient interactions, which are difficult to observe in cells. Intriguingly, MYO6+ is able to sequester a selection of binding partners at the tips of filopodia, spatially segregating these complexes at the plasma membrane. In particular, MYO6+ is able to recruit Tax1 binding protein 1 (TAX1BP1) (19) and dedicator of cytokinesis protein 7 (DOCK7) (20) into newly formed filopodia (Fig. S3B), whereas optineurin (21) and nuclear dot 52-kDa protein (NDP52) (19) were not relocalized (Fig. S3C). This finding was quantified by comparing Pearson’s correlation coefficients between cells expressing either GFP-MYO6 or GFP-MYO6+ (Fig. S3D). Thus, MYO6+ may prove to be a useful tool to interrogate the properties of various MYO6 binding partner complexes in intact cells via fluorescence imaging.

Given that MYO6+ was able to relocalize its binding partners, we examined whether sites in the MYO6 tail known to mediate interactions with adaptor proteins (4) were important for filopodia formation and the translocation of MYO6+ to filopodia tips. Mutation of either the RRL motif (R1117A, known to coordinate interactions with GIPC, TAX1BP1, NDP52, and optineurin) or the PI(4,5)P₂ interacting motif in the tail of MYO6+ led to a dramatic reduction in the accumulation of MYO6+ at the tips of filopodia (Fig. 3 A and B). In contrast, mutation of the tryptophan–tryptophan–tyrosine (WWY) site [W1202L, which binds to disabled homolog 2 (DAB2), target of Myb1 (TOM1), and lemur tyrosine kinase 2, Fig. 3 A and B] or the ubiquitin binding site (A1013G) had no effect on this accumulation. Thus, the RRL site appears to be the key site for formation of the processive (multimeric) form of MYO6 in cells. To test whether processive movement on actin is required, we made mutations in the motor domain of MYO6+. The rigor mutant K157R, which is unable to bind ATP and so remains tightly bound to actin, also prevented accumulation in filopodia (Fig. 3 A and B). A mutation (D179Y), which has been shown to accelerate phosphate release from MYO6 and thereby prevents processive runs on actin (the mutation increases the rate of phosphate release before actin binding) (22), also leads to a complete loss of MYO6+-induced filopodia (Fig. 3 A and B). Thus, processive movement on actin is required for accumulation of MYO6+ at the tips of filopodia. In contrast, mutation of the putative phosphorylation site in the motor domain Thr405 (23), to either alanine or glutamic acid, had no effect on filopodia formation.

As only a specific subset of tail motifs was required for filopodia formation and only certain binding partners were reorganized by MYO6+, we next examined the role of these partners in the capacity of MYO6+ to induce filopodia. Knockdown of the endosomal protein GIPC (Fig. 4 A–C) led to a dramatic reduction in the number of cells with filopodia induced by MYO6+. Likewise, ablation of the characteristic marker of signaling endosomes, the GIPC-binding partner APPL1, displaced GIPC from endosomes and prevented accumulation of MYO6+ at tips of filopodia (Fig. 4 A–C). In contrast, depletion of either NDP52 or DOCK7 had no effect on filopodia formation and movement of MYO6+ to tips of filopodia (Fig. 4B). When APPL1 was depleted, GIPC became cytosolic but was still stable and not degraded (Fig. 4D). However, MYO6+ no longer induces formation of filopodia, indicating that the MYO6 tail can only interact with GIPC on endosomes. This finding is consistent with the requirement for an intact PI(4,5)P₂ binding site in the MYO6 tail to ensure membrane binding.

To further probe the requirement for processive movement in formation of filopodia, we transfected cells with a cargo binding domain (CBD)-tailless version of MYO6+, comprising the head and lever arm of MYO6+ and a leucine zipper at the C terminus to enforce dimerizations (Fig. 4E). Expression of a leucine zippered MYO6+ lacking the CBD tail still led to the formation of filopodia, but without recruitment of GIPC into the tips and without clustering of endosomes at the base of filopodia (Fig. 4F and Fig. S4). Further deletion of the leucine zipper led to a cytoplasmic construct (Fig. S4). This finding indicates that key to the formation of filopodia is the multimerization of MYO6+, by assembly of at least a dimer. Furthermore, this experiment indicates that assembly of processive complexes of MYO6 requires targeting to endosomes.

To investigate the processes underlying formation of filopodia by expression of MYO6+, we used live-cell TIRF microscopy to observe growing filopodia. By coexpressing GFP-MYO6+ and Ruby-Lifeact (to detect F-actin), growing filopodia could be identified (Fig. 5 A and B and Movie S1). The average rate of growth of a filopodia tip was 0.63 ± 0.02 µm⋅min⁻¹ (SEM, n = 228), similar to the previously reported rate of actin polymerization in filopodia (0.5–1 µm⋅min⁻¹) (24) and about 5–10 times slower than the rate that MYO6 can slide actin filaments in vitro (1). In addition to growth, we also frequently observed movement of GFP-MYO6+ patches away from the filopodia tips toward the base (Fig. 5 C and D and Movie S2) at an average speed of 0.99 ± 0.09 µm⋅min⁻¹ (SEM, n = 62), again similar to the rate of actin polymerization in cells (Fig. 5E). Although most of the MYO6+ accumulates at the tips of filopodia, it forms long lasting complexes with actin and thus can be transported through actin filament treadmilling toward the base of filopodia (opposite to their direction of motion).

In addition to simple growth and retrograde movement, the filopodia were capable of more complex dynamics, including wave propagation (Fig. 5F). These waves have a wavelength of 2.27 µm ± 0.18 µm (the distribution is shown in Fig. 5G, SEM, n = 20) and propagate at a speed of 1.18 ± 0.09 µm⋅min⁻¹ (Fig. 5H, SEM, n = 39), again very similar to the rate of actin treadmilling. On some occasions, complete twisting of the wave into a coil was observed (Fig. S5 A and B). Filopodia were also capable of splitting at the head (Fig. S5 C and D).

We next analyzed whether MYO6+ induced the formation of filopodia by recruiting or activating proteins regulating actin organization (Fig. S6A). We did not observe a significant effect of depleting actin-related protein 3 (ARP3), diaphanous-related formin-3 (DIAPH3), or neural Wiskott–Aldrich syndrome protein (N-WASP) (shown in Fig. S6E) on the percentage of cells with MYO6+-induced filopodia (Fig. S6 A and B). Depletion of ARP3 in cells transfected with GFP-MYO6+, and thus reduced actin branching, led to the formation of slightly more transfected cells forming filopodia (Fig. S6B) and longer individual filopodia (of mean length 13.3 ± 6.5 µm SD, n = 215 from three experiments, vs. 6.6 ± 3.8 µm SD, n = 174 from three experiments, P < 0.0001, Fig. S6C). However, ARP3 depletion significantly reduced the propensity of filopodia induced by MYO6+ to themselves form clusters (Fig. S6D that are not observed for MYO10, which does not require endosomes to promote filopodia formation). This observation indicates that ARP2/3-induced network branching is important for the reorganization of the actin cortex and the clustering of peripheral endosomes. Depletion of the formin DIAPH3 had no effect on filopodia induced by either MYO6+ or MYO10. Whereas depletion of N-WASP had a limited effect on MYO6+-induced filopodia, depletion of N-WASP led to a marked reduction in filopodia formation by MYO10, indicating that differences exist in the mechanism of filopodia formation between the two motors. In summary, our data indicate that formation of filopodia by MYO6+ is a robust phenotype that depends on motor activity and cargo-mediated multimerization, but does not involve the established actin polymerization machinery.

Increasing evidence indicates that myosin motors not only use actin as a track, but can also organize actin filaments. In vitro, MYO6 in a zippered-dimer format can bundle arrays of filaments on a timescale of minutes to hours (26). In vivo, actin organization has been shown to require MYO6 in Drosophila spermatid individualization (27), at cell–cell junctions (28), and on melanosomes (29). GIPC has been identified as having a role in maintaining actin structures in hair cells (30), indicating that the actin organizing function of MYO6 operates through GIPC on endosomal membranes. In the present study, nascent actin filaments formed by MYO6+ activity were apparently stabilized by the accumulation of fascin (Fig. 1E). Thus, the action of actin-associated motor proteins can also affect the recruitment and activity of actin regulators. The underlying mechanism may operate in concert with selective recruitment of tropomyosin isoforms, which can serve to specify actin filament identity (31).

Interestingly, we also observed rotational waves and coiling within MYO6+-induced filopodia (Fig. 5). Traveling waves along filopodia have been recently described by a helical buckling model (32). This phenomenon was hypothesized to require a rotational force at one end and a fixed point at the other end of the filopodia. We propose that a rotation could also be generated by the coupling of MYO6+ to both the filopodial membrane through binding to PI(4,5)P₂ and actin through the motor domain. Consistent with this notion, helical movement of MYO6 around actin filaments has been reported in vitro (33). Furthermore, the hook-like shape of the tips of MYO6+ filopodia, as observed in our SIM images (Fig. S2), may indicate that the accumulation of a high density of MYO6+ motors at the tip may exert signification bending or torsional forces on the actin core. This process may in turn be coupled to the initiation of waves (Fig. 5). Retrograde movement of MYO6+ was seen to occur in large patches (significantly brighter than a single molecule), as previously observed for MYO6 (34) and myosin XV(MYO15) (35), indicating that formation of large teams of motors is an important and conserved mechanism of motor regulation.

Why do filopodia form at the plasma membrane? Similar observations of motor translocation to the tips of filopodia have been made when expressing other plus-ended myosins such as MYO10 (12), MYO7A (13), MYO3 (36), and MYO15 (37). MYO19 is even capable of translating mitochondria to the tips of filopodia (38). Taken together, our data and these reports indicate that formation of filopodia, or at least condensation of actin filaments, may be a general property of multimeric plus-ended myosins that must be regulated by cells. We speculate that one factor allowing growth of filopodia in the MYO6+ system is the protection of actin filaments from severing factors (e.g., gelsolin, villin, severin) and depolymerizing factors (ADF/cofilin), caused by sequestration of F-actin away from the cytosol and into filopodia. It is interesting to note that despite much investigation of the mechanisms underlying filopodia formation by MYO10, a similar phenomenon can be generated by a motor that is usually not associated with this process and is thus unlikely to contain any specific structural adaptations. Filopodia formation thus appears to be a universal property for multimeric plus-end–directed myosins.

In conclusion, we have shown that reversing the direction of MYO6 leads to the reorganization of peripheral actin and endosomes, reinforcing the link between MYO6 function and actin organization. This reversal adds to the list of situations in which MYO6 has been shown to affect actin organization. Many questions remain about the interaction of MYO6 with actin in its numerous functions, which will be probed in concert with the development of novel tools and methods.

HeLa cells were grown in RPMI-1640 (Sigma) supplemented with 10% (vol/vol) FBS and penicillin/streptomycin. Twenty-four hours before transfection, cells were plated in a 35-mm dish at a density of 100,000 cells per well. Cells were transfected with Fugene 6 according to manufacturer’s instructions. Knockdown of specific proteins with siRNA was performed twice with Oligofectamine reagent and the respective Smartpool siRNA (Dharmacon) before transfection, in accordance with manufacturer’s instructions.

Primary antibodies used in this study are as follows: mouse anti-GFP (Abcam ab1218) 1:400; rabbit anti-GFP (Invitrogen A-11122) 1:400; mouse anti-fascin (Millipore MAB3582) 1:100; rabbit anti-GIPC (Proteus Bioscience 25–6792) 1:100; rabbit anti-APPL1, (Santa Cruz sc-67402) 1:100; rabbit anti-DAB2 (Santa Cruz, sc-13982) 1:100; rabbit anti-EF-2 (Santa Cruz, sc-166415); rabbit anti-TOM1L2 (Abcam ab96320) 1:100; rabbit-anti TAX1BP1 (19) 1:100; rabbit anti-optineurin (21) 1:100; and rabbit anti-NDP52 (19) 1:100. MYO10-mCherry was constructed by cloning mCherry into EGFPC1-hMyoX, a gift from Emanuel Strehler, Mayo Clinic, Rochester, MN (Addgene plasmid 47608 (39). A rabbit polyclonal antibody was raised against the C-terminal region of DOCK7.

For immunofluorescence, cells were plated on 24-mm no.1 coverslips 24 h before transfection (overnight). Cells were washed once in Dulbecco’s PBS and fixed in 4% (vol/vol) formaldehyde for 15 min, followed by permeabilization with 0.2% Triton for 5 min and blocking with 1% BSA in PBS for 1 h. Coverslips were incubated with primary antibodies at 1:100 dilution in BSA on for 1 h, washed three times in PBS, incubated with Alexa-dye–conjugated secondary antibodies (and phalloidin at a concentration of 13 nM to label F-actin if desired) at 1:500 dilution in BSA for 1 h, washed three times in PBS, rinsed with deionized water, and mounted on slides with Prolong Gold. Widefield images were acquired on a Zeiss Axioimager Z2 upright microscope. Confocal microscopy was performed on a Zeiss LSM880. Fields of view were selected by randomly scanning the stage.

Structured illumination microscopy (SIM) was performed on a Zeiss Elyra PS1 microscope. Samples were prepared by growing cells on acid-washed 20-mm square high-performance coverslips (no. 1.5, 170 ± 5 µm; Schott). Cells were fixed in formaldehyde, permeabilized with 0.2% Triton in PBS, and washed with 0.02 M glycine (two times for 5 min each) to quench background fluorescence. Proteins of interest were antibody labeled as described above; however, F-actin was stained with phalloidin at a concentration of 1.06 µM (eight times the concentration used for standard widefield or confocal microscopy). Following secondary antibody labeling, coverslips were washed extensively with water, mounted with ProLong Gold (Life Technologies) on unfrosted glass microscope slides (Thermo Scientific), and cured for 3 d in the dark at room temperature. Image stacks were acquired for five grating rotations, with reconstruction and alignment executed using ZEN Black ELYRA edition software (Carl Zeiss Microscopy). The presented data are a single slice, out of a z-stack, containing the largest number of filopodia and endosomes.

Imaging was performed on a Zeiss SD-TIRF system. Cells were grown in 35-mm glass-bottom dishes (ibidi). Before mounting on the microscope, media was exchanged to live cell imaging solution (Invitrogen, essentially Ringer’s solution buffered with Hepes) containing 10% (vol/vol) FBS.

Cell lysates were generated from HeLa cell cultures seeded at 150,000 cells per well in 35-mm dishes (six-well plate, Greiner) 2 d before lysis in 200 μL Nonidet P-40 buffer (150 mM NaCl, 50 mM Tris, 1 mM EDTA, 1% Nonidet P-40, Roche protease inhibitor mixture) on ice. Cells were scraped and whole cell lysates were centrifuged at 16,000 × g for 15 min. Supernatant was removed and boiled in sample loading buffer containing β-mercaptoethanol. Samples were loaded on 8% (vol/vol) polyacrylamide gels and run using the MiniProtean system (Bio-Rad). Proteins were transferred to PVDF membrane (Millipore) using wet transfer on ice (Bio-Rad). Membranes were blocked in 5% (wt/vol) milk, incubated with primary antibodies in milk at 4 °C overnight and HRP-linked secondary antibodies at room temperature for 1 h. Membranes were incubated with ECL detection reagent (GE Healthcare) for 1 min, and developed using X-ray film (Kodak). Films were digitalized using a scanner (Epson). Elongation factor 2 (EF-2), molecular weight 93 kDa, was used as a loading control in most experiments, detected by rabbit polyclonal C-9 (Santa Cruz, sc-166415).

All image analysis was performed using Fiji (ImageJ). Filopodia lengths were measured manually using the line function. Colocalization was assessed by calculating Pearson’s correlation coefficient using the ImageJ plugin coloc2 (for which cells were manually segmented and background was subtracted using inbuilt ImageJ functions). Filopodia formation induced by MYO6+ was assessed by counting the proportion of cells with MYO6+ accumulated at the tips of filopodia in widefield images with a field of view 211 µm × 211 µm (which typically contained 10–30 cells each). Ten fields of view were counted per experiment. For analyzing waves, wavelength was calculated from the peak-to-peak distance obtained from a line scan of TIRF images; velocity was calculated for each peak from the distance traveled down the filopodia divided by the imaging time. As a consequence, we made slightly more observations of wave velocity than wavelength.

All presented plots were generated using OriginPro 2016 (OriginLab). To obtain P values, two-sample t tests were conducted in the same software.

MYO6+ was constructed by replacing the unique insert (reverse gear), IQ domain and lever arm extension of human myosin VI (gene ID 4646, UniProtKB Q9UM54-5, corresponding to isoform 5 containing no alternative splicing inserts) with the six IQ domain lever arm from mouse myosin V (Myo5, gene ID 17920). MYO6+V2 was constructed by replacing the unique insert (reverse gear), and IQ domain with the first two IQ domains of the myosin V lever arm.

The six IQ domains of Myo5 were cloned into GFP-MYO6-NI (NI indicates the No Insert isoform) (40) by overlap extension PCR. The tail region of MYO6 was amplified from amino acid 913 to the C-terminal end with sense oligo 5′-GCGTGAGCTGAAGAAACTCAAAttacagaaaaaaaaacagcagg-3′ (the uppercase 5′ end corresponds to the 3′ end of the 6IQ region of Myo5 at amino acid 909) and antisense oligo 5′-TTTgcggccgcTTATTTCAACAGGTTCTGCAGC-3′ (containing a NotI site). The Myo5 motor domain and lever arm (amino acids 1–909) was amplified with sense oligo 5′-aaagaattctgatggctgcgtcggagctctacaca-3′ and antisense 5′-cctgctgtttttttttctgtaaTTTGAGTTTCTTCAGCTCACGC-3′ (with the uppercase section corresponding to the reverse complement of the 3′ Myo5 lever arm sequence). These two products were then combined via an overlap extension reaction to create a Myo5(S1)-Myo6(Tail) recombinant fusion construct.

To create MYO6+, the Myo5 motor domain in Myo5(S1)-Myo6(Tail) was replaced by the MYO6 motor domain by a further round of overlap extension PCR. The motor domain of MYO6 (amino acids 1–770) was amplified with sense oligo 5′-caagaattcaaatggaggatggaaagccc-3′ and antisense oligo 5′-GGCAGCCCGAAGTTTGTCAGCcatgatctgatcaaattctgc-3′, where the uppercase region corresponds to the reverse complement of the 5′-end of the Myo5 lever arm. The Myo5 lever arm and MYO6-tail fusion were amplified by sense oligo 5′-gcagaatttgatcagatcatgGCTGACAAACTTCGGGCTGCC-3′ and antisense oligo 5′-TTTgcggccgcTTATTTCAACAGGTTCTGCAGC-3′ (as used in the first round). The two products were then fused by overlap extension to create the final construct: MYO6(1-770):Myo5(763-909):MYO6(913-end), abbreviated as MYO6+. This was further subcloned into the widely used transient expression vector pEGFP-C3 such that a GFP tag was present at the N terminus.

Mutations in MYO6+ were made using site-directed mutagenesis as described previously: R1117A (RRL→RAL) and W1202L (WWY→WLY) (40), ΔPIP2 (WKSKNKKR→WASANNNR) (41), A1013G (18), D179Y (22), K157R (8), and T405A and T405E (42, 43). All amino acid designations are according to the canonical sequence of human MYO6: UniProtKB Q9UM54-3 (isoform 3).

CRISPR/Cas9 MYO6 knockout HeLaM cells were prepared using a published method (16) with some modifications. A 20-bp target in the N terminus of MYO6 (GGAGGATGGAAAGCCCGTTT) was identified using the CRISPR design tool: crispr.mit.edu (Zhang laboratory). Corresponding sgRNA oligos were annealed and ligated to the pSpCas9(BB)-2A-Puro (PX459) vector (48139, Addgene), a gift from Feng Zhang, Massachusetts Institute of Technology, Cambridge, MA (16). HeLaM cells were transfected with the resulting PX459 plasmid encoding Cas9 and sgRNA against MYO6 using FuGENE 6 (Promega). After 24 h of transfection, cells were incubated with 1.5 μg/mL puromycin for an additional 72 h. Single-cell clones were isolated and screened by Western blotting using a rabbit polyclonal MYO6 antibody and Cas9 cleavage of clones lacking MYO6 was assessed for indels by Sanger sequencing.