Introduction
A large number of unconventional myosins appeared early in eukaryotic evolution and these have vital roles in diverse cellular processes including intracellular transport, organization of F‐actin, mitotic spindle regulation and gene transcription (
Berg et al, 2001;
Richards and Cavalier‐Smith, 2005;
Foth et al, 2006;
Woolner and Bement, 2009). Myosins consist of three distinct regions, a head, neck and tail. The heads contain actin‐based motor domains that display homology among different myosins and these structures have been extensively studied (
O'Connell et al, 2007). In sharp contrast to the motor domains, the tails display diversity by combination of a variety of functional domains that mediate cargo recognition, which determine the individual cellular functions of myosins. However, little is known about the tail domain structures and their specific cargo recognition.
Myosin‐X (Myo‐X, myosin‐10, Myo10) is an unconventional myosin implicated in elongation of filopodia, which function as tentacles that explore and interact with cell surroundings to determine the direction of cell movement and to establish cell adhesion such as in the case of synapses (
Sousa and Cheney, 2005). Myosin‐X is localized at filopodia tips containing a ‘filopodial‐tip complex’, which act as fingertips and sensors in processes such as signal perception, cell signalling, actin polymerization inside of filopodia and adhesion (
Berg et al, 2000;
Berg and Cheney, 2002). Elucidating the molecular nature of the filopodial‐tip complex is key for understanding filopodia functions, but has remained a mystery for over 30 years. The manner by which myosin‐X discriminates between cargos for transportation to the tip and how these cargos contribute to filopodial processes at the tip remains unknown. Myosin‐X contains myosin tail homology 4 (MyTH4) and 4.1 and ezrin/radixin/moesin (FERM) domains for cargo recognition (
Figure 1A). The tandem comprising the MyTH4 and FERM domains frequently appears in related myosins (myosin‐IV, VII, X, XII, XIV and XV), and is referred to as the MyTH4–FERM cassette. Interestingly, this cassette is found in one of three ancestral myosins in the earliest eukaryotes, and myosin‐X is conserved in vertebrates (
Richards and Cavalier‐Smith, 2005).
One of the most exciting processes involving myosin‐X relates to the axon pathfinding of neurons, which is essential for proper wiring in the brain. During neural development, axons are navigated by extracellular guidance cues such as those provided by netrins. Deleted in colorectal cancer (DCC) and neogenin are membrane proteins that function as netrin receptors (
Chan et al, 1996;
Keino‐Masu et al, 1996;
Kolodziej et al, 1996). Myosin‐X recognizes these receptors as cargos and redistributes to the cell periphery or to the tips of neurites, where growth cones dynamically develop filopodia (
Zhu et al, 2007). Moreover, of particular interest is that myosin‐X interacts with integrin through its FERM domain and mediates relocalization of integrin to filopodial tips and thereby promotes filopodial extension by serving to form adhesive structures (
Zhang et al, 2004). A very recent report has shown that DCC is also important as a cargo adaptor that mediates local translation events in neurons by anchoring components of the translational machinery such as ribosome subunits at the plasma membrane of growth cones and dendrites (
Tcherkezian et al, 2010).
In addition to mediating the biological function of selective cargo transportation on actin cables, myosin‐X directly interacts with microtubules and has a key role in spindle assembly during meiosis to ensure faithful delivery of replicated chromosomes to daughter cells following cell division (
Weber et al, 2004;
Woolner et al, 2008). This surprising myosin‐X function is mediated by a direct interaction between microtubules and the MyTH4–FERM cassette. However, the manner by which myosin‐X recognizes microtubules has remained unclear. Interestingly, myosin‐X has a role in integrin‐dependent spindle orientation (
Toyoshima and Nishida, 2007).
Here, we report on a series of structural and biochemical/biophysical studies concerning DCC recognition by the myosin‐X MyTH4–FERM cassette. We reveal the presence of a VHS‐like fold within the MyTH4 domain. Our 1.9 Å resolution structure clarifies details of an unexpected binding mode of DCC to the myosin‐X FERM domain, which is distinct from those found in the FERM domain of radixin that links membrane protein/plasma membrane and actin cytoskeletons. We also show that the cassette binds the C‐terminal acidic tails of tubulins and that this binding is obstructed by DCC binding. Moreover, we show that the cassette binds the cytoplasmic tail of the integrin β5‐subunit and that this binding is obstructed by DCC binding. Like DCC, integrin β5 binding also interferes with microtubule binding. Our results reveal the structural mechanism that underlies cargo recognition by the cassette and provide the molecular basis for further structural and functional investigations of biologically and medically important myosin‐X, as well as of the related unconventional myosins containing MyTH4–FERM cassette.
Discussion
We have provided the first structures of the MyTH4–FERM cassette in the free and cargo‐bound forms. These structures together with the biochemical analyses allow us to delineate the mechanisms that account for the two modes of cargo and microtubule recognition. Of the 37 different types of myosins identified displaying different domain combinations, MyTH4–FERM‐containing myosins appear in several species (
Richards and Cavalier‐Smith, 2005). Vertebrates possess three conserved MyTH4–FERM‐containing myosins, VII, X and XV. Although
Drosophila is thought to lack myosin‐X (
Berg et al, 2001;
Foth et al, 2006), the DCC P3 helix is well conserved in frazzled, which encodes a
Drosophila member of the DCC immunoglobulin subfamily and is required for CNS and motor axon guidance (
Kolodziej et al, 1996) (
Figure 3B). The myosin assigned as myosin‐XV may function as myosin‐X.
Recently, DCC has been suggested to mediate local translation in neurons by anchoring components of the translational machinery such as ribosome subunits at growth cones and dendrites (
Tcherkezian et al, 2010). The machinery‐binding site of DCC is located at the juxtamembrane P1 region, which is separated from the C‐terminal P3 helix by more than 250 residues thought to form a random coil. In fact, the P3 region also exists as a random coil in the free state (
Supplementary Figure S5). The DCC cytoplasmic long tail may function as a molecular glue for secondary cargo binding.
In our structure, the MyTH4 domain has no direct contacts with the DCC peptide, in spite of the yeast two‐hybrid assay which demonstrated full DCC‐binding affinity to the complete MyTH4–FERM cassette but weaker binding affinity to the isolated FERM domain (
Zhu et al, 2007). We speculated that the MyTH4 domain is necessary for folding and/or maintaining the stable structure of the FERM domain. In fact, we experienced difficulties in expressing isolated MyTH4 or FERM domains in a stable form in bacteria as previously mentioned (
Weber et al, 2004). This seems to be in line with our observation of rigidity in the overall structure of the MyTH4–FERM cassette, which exhibits no significant conformational changes in the free and DCC‐bound states or in different crystal packing environments.
An unexpected finding related to the DCC αP3 helix that was docked into the groove between the β5C strand and the α1C helix of subdomain C of the myosin‐X FERM domain (
Figure 3A). In sharp contrast to this α‐helix groove docking (
Figure 6A), binding partners of ERM proteins have been shown to bind this groove of FERM domains by forming an additional anti‐parallel β–β association with the β5C strand (
Hamada et al, 2003;
Takai et al, 2007,
2008;
Terawaki et al, 2007;
Mori et al, 2008) (
Figure 6B). In these structures, binding partners such as adhesion molecule ICAM‐2, CD43 and PSGL‐1 form a short β strand for association with the β5C strand and a 1‐turn 3
10 helix for docking into the groove, and preserve the sequence motif Motif‐1α (RxxTYxVxxA, where x stands for any amino‐acid residue) (
Hamada et al, 2003;
Takai et al, 2007,
2008) (
Figure 6B and E). Another adhesion molecule CD44 and a membrane‐associated protease NEP preserve the Motif‐1β conserved sequence Ile‐Asn (IN), and form a short β strand followed by a reverse turn (
Terawaki et al, 2007;
Mori et al, 2008). These binding motifs differ from the DCC/neogenin motif that forms the P3 α‐helix. Compared with radixin subdomain C, myosin‐X possesses a shorter β5C strand (4 residues versus 6 residues of radixin) and shorter α1C helix (5 turns versus 6 turns of radixin), and these are loosely associated to form a larger groove for α‐helix accommodation, given the absence of bulky nonpolar residues at the groove surface (
Figure 6F).
ERM proteins bind the Na
+/H
+ exchanger regulatory factor‐1, 2 (NHERF‐1 and ‐2), which are key cytoplasmic proteins involved in the anchoring of ion channels and receptors to the actin cytoskeleton through binding to ERM proteins. These adaptor proteins preserve FERM‐binding Motif‐2 (MDWxxxxx(L/I)Fxx(L/F)) and form an α‐helix that docks into the groove formed by the β‐sandwich loops of subdomain C (
Terawaki et al, 2006), which displays no similarity to the DCC/neogenin P3 motif (
Figure 6C). The myosin‐X FERM domain lacks the PtdIns(4,5)P
2‐binding site found in the radixin FERM domain as previously described (
Figure 6D).
It was one of the most exciting findings in cell biology of cytoskeletons that myosin‐X has the ability to function as a motorized link between actin filaments and microtubules in spindle assembly during meiosis (
Weber et al, 2004;
Toyoshima and Nishida, 2007;
Woolner et al, 2008). However, to date, only a single report of
Xenopus myosin‐X has shown the direct binding of myosin‐X to microtubules (
Weber et al, 2004). Our present studies confirmed that result in a mammalian myosin‐X but also revealed that at least the major part of the novel myosin‐X–microtubule interaction is mediated by the MyTH4 domain and the C‐terminal acidic tails of tubulins. Interestingly, these acidic tails of αβ‐tubulins are also collectively referred to as E‐hooks, which interact with the kinesin K‐loop for processive motor movement (
Okada and Hirokawa, 2000;
Hirokawa et al, 2009). It is well known that α‐tubulin acidic tails possess a terminal tyrosine residue at the C‐terminus, which is recognized by the CAP‐Gly domains of microtubule‐binding proteins such as CLIP‐170 (
Honnappa et al, 2006;
Mishima et al, 2007). The MyTH4–FERM cassette has no homologous region to the CAP‐Gly domain and binds both α‐ and β‐tubulin acidic tails.
We showed that the interactions of the MyTH4–FERM domain with microtubules and DCC are mutually exclusive, in spite of the fact that the binding sites are not overlapping. This is particularly important as myosin‐X is now reported to have several different binding partners, implying that under different cellular conditions, a different subset of proteins may be associated with myosin‐X. It is noteworthy that integrin binding also interferes microtubule binding. We speculate that the primary motor function carrying cargos and the linker function linking an actin filament and a microtubule are two alternative functions of myosin‐X.
At present, it is unclear how the interference between tubulin tail and DCC P3 binding occurs, although the C‐terminal flexible region of the DCC αP3 helix may interfere with tubulin tail binding to the MyTH4 domain (
Supplementary Figure S6). Our present data suggest no significant changes in the overall structure of the MyTH4–FERM cassette on DCC binding. However, it is possible that the acidic tail or microtubule bindings may cause some induced‐fit changes in the overall structure. We also speculated that negatively charged residues of the N‐terminal flanking region of the DCC P3 peptide may cause electrostatic repulsion with the tubulin acidic tails and the negatively charged surfaces of microtubules.
Recently, the crystal structure of a fusion protein of the myosin‐X MyTH4–FERM cassette (human residues 1503–2047) and the DCC P3 peptide (rat 1409–1445) at 2.5 Å resolution has appeared (
Wei et al, 2011). The fusion protein contains an artificial covalent link between the C‐terminus of the subdomain C α1C helix and the N‐terminus of the DCC peptide. Compared with our structure, the MyTH4–FERM cassette fused to the DCC P3 peptide displays a relatively large r.m.s. deviation (1.9 Å) of the overall structures (
Supplementary Figure S7A). This unexpectedly large deviation is due to a change in orientation of subdomain C relative to subdomains A and B with an ∼5° rotation. Contrary to the overall structures, each FERM subdomain and the MyTH4 domain exhibit relatively small r.m.s. deviations (∼1 Å). Like our structure, the fused DCC peptide binds subdomain C with the DCC αP3 helix docked into the subdomain C groove. Side‐chain packing at the groove involving conserved nonpolar residues is similar in the two complexes. Nevertheless, superimposition of subdomain C reveals notable local conformational deviations in this subdomain and the bound DCC peptide (
Supplementary Figure S7B). Compared with our DCC peptide, the αP3′ helix of the fused DCC peptide is moved by 4.5 Å away from the α1C helix, while the C‐terminal region of the α1C helix is shifted away from the DCC αP3′ helix. These movements seem to be induced by the artificial covalent link, which exhibits high mobility with high‐temperature factors and seems to push these helices away from each other. These movements result in looser helix packing between the N‐terminal region of the α1C helix and the C‐terminal region of the αP3 helix of the fused protein.
The most prominent differences are found in polar interactions. In our complex, Lys2034 from the α1C helix has a central role in forming the hydrogen bonding network involving DCC residues (Asn1440, Thr1443 and Ser1445) from the C‐terminal end of the αP3 helix (
Figure 4B). This network is absent in the fused protein and, instead, Lys2031 forms a single hydrogen bond with Ser1445 (
Supplementary Figure S7C). This difference is caused by side‐chains rearrangement, which is induced by reorientation of subdomain C, which shifts away from subdomain A with breaking the inter‐subdomain salt bridge between Lys2034 and Glu1756 and the hydrogen bond between Lys2031 and the β4A strand (
Figure 4B). Lack of these inter‐subdomain interactions destabilizes the hydrogen bonding network, resulting in side‐chain rearrangement with destruction of the network. In addition to these movements, the β5C–β6C loop of subdomain C is also moved away from the fused DCC P3 helix and the hydrogen bond involving DCC Gln1438 is missing or neglected probably because of the marginal value (3.4 Å) of the distance. All these differences imply that the DCC binding found in the fused protein may be in a metastable state of binding.
The MyTH4–FERM cassette structure displays a sharp contrast to the recently reported structure of the myosin‐VIIa MyTH4–FERM‐SH3 cassette bound to cargos the central domain (CEN) of Sans determined at 2.8 Å resolution (
Wu et al, 2011). The myosin‐VIIa MyTH4–FERM cassette is distantly related to the myosin‐X MyTH4–FERM cassette with low (∼14%) sequence identity, which is reflected in relatively large deviations of pairwise superimpositions of each subdomains A–C and the MyTH4 domain (r.m.s. deviations ranging from 1.9 to 2.4 Å). Remarkably, superimposition of the overall structures of MyTH4–FERM cassettes of these myosins exhibits a large r.m.s. deviation (6.2 Å for 432 residues) (
Supplementary Figure S8). This large deviation is due to reorientations of subdomains B and C, in addition to no structural similarity between the MyTH4 extensions of two myosins. These reorientations could be induced by Sans CEN binding to all three inter‐subdomain interfaces formed by subdomains A, B and C. This binding site that is distinct from our DCC‐binding site reflects versatility in protein–protein interactions of the FERM domain. Despite these deviations, the MyTH4 domain and subdomain A display a relatively better overlap (the r.m.s. deviation of 2.8 Å), indicating rigidity of the interface between the MyTH4 domain and subdomain C as we discussed with our myosin‐X MyTH4–FERM structure. Subdomain arrangements in FERM domains may be more flexibly than we expected before. In fact, the recent structure of the talin FERM domain shows a novel extended conformation (
Elliott et al, 2010).
In addition to netrin receptors, integrins and microtubules, several reports have suggested other binding partners of myosin‐X such as bone morphogenetic protein 6 receptor ALK6 (
Pi et al, 2007), adhesion molecules such as VE‐cadherin (
Almagro et al, 2010), regulators of actin polymerization such as Mena/VASP (
Tokuo and Ikebe, 2004) and spindle‐pole assembly factor TPX2 (
Woolner et al, 2008). The fact that Mena/VASP is a cargo of myosin‐X may be related to the reported observation that both the MyTH4 domain and motor function of myosin‐X are crucial for actin organization at the leading edge and the promotion of filopodia formation (
Bohil et al, 2006;
Tokuo et al, 2007). At present, our sequence analysis failed to detect candidates for myosin‐X‐binding regions that resemble our DCC P3 peptide or tubulin acidic tails. Lack of sequence similarity between the integrin β5 cytoplasmic peptide and the DCC P3 peptide implies that the myosin‐X subdomain C is a protein‐binding module of dual or multiple modes. It should be noted that the MyTH4–FERM cassette may possess uncovered binding sites for these binding partners or that these bind other domains of myosin‐X.
Materials and methods
Protein expression and purification
A DNA fragment encoding the MyTH4–FERM cassette (residues 1486–2058) of human myosin‐X was amplified by the polymerase chain reaction (PCR) and cloned into the pET47b [+] vector (Novagen). To prevent degradation, nucleotides encoding residues 1872–1891 of the extra loop were deleted from the plasmid using inverse PCR. The PCR‐amplified nucleotides encoding human DCC (residues 1390–1447), α1A‐tubulin (399–451), β2B‐tubulin (390–445) and integrin β5 cytoplasmic tail (743–799) were cloned into the pET49b [+] vectors (Novagen). All plasmids were verified by DNA sequencing and transformed into Escherichia coli strain BL21Star (DE3) (Invitrogen) cells for protein expression.
Protein expression was performed at 20°C in Luria‐Bertani medium containing 0.1 mM isopropyl‐β‐d‐thiogalactopyranoside. Cells expressing the myosin‐X MyTH4–FERM cassette were harvested, suspended in 20 mM Tris–HCl buffer (pH 8.5) containing 150 mM NaCl and disrupted by sonication. After ultracentrifugation, the supernatant was applied onto a Ni‐NTA resin (Qiagen) and treated with HRV3C protease to remove the N‐terminal hexahistidine tag. Eluted proteins were further purified by cation exchange (HiTrap SP HP, GE Healthcare) and gel filtration (Superdex 200 pg, GE healthcare) chromatography. For preparation of the DCC P3, tubulin tail and integrin peptides, wet cells expressing proteins were suspended in 20 mM Tris–HCl buffer (pH 8.0) containing 150 mM NaCl and disrupted by sonication. Following ultracentrifugation, the supernatant was purified by glutathione‐Sepharose 4B column (GE Healthcare), anion exchange (HiTrap Q HP, GE Healthcare) and gel filtration (Superdex 75 pg, GE Healthcare) chromatography. For crystallization, the GST‐fused DCC P3 peptide was treated with HRV3C protease to remove the N‐terminal GST tag. Separately purified myosin‐X MyTH4–FERM cassette and DCC P3 peptide were mixed and the 1:1 complex was purified using gel filtration chromatography.
For structure determination, protein expression of the selenomethione (SeMet)‐labelled MyTH4–FERM cassette was performed in M9 medium containing SeMet under conditions inhibiting the methionine biosynthesis pathway (
Doublié, 1997). The expression conditions and purification procedures were the same as those used for the native protein. The purified proteins were verified using matrix‐assisted laser desorption/ionization time‐of‐flight mass spectroscopy (MALDI–TOF‐MS; Bruker Daltonics).
Crystallization and data collection
Initial crystallization conditions were screened using a Hydra II Plus One crystallization robot (Matrix Technology) with commercial crystallization‐solution kits, JCSG Core Suite I–IV and PACT Suite (Qiagen). The best crystals of the complex between myosin‐X MyTH4–FERM cassette and DCC P3 peptide were obtained from solutions containing 5 mg/ml of the complex and reservoir solution containing 100 mM malic acid‐MES‐Tris (MMT) buffer (pH 8.0) and 2% polyethylene glycol (PEG) 1500 at 20°C. Crystals of the free myosin‐X MyTH4–FERM cassette were obtained using reservoir solution containing 10% PEG 8000, 5% MPD and 0.1 M HEPES (pH 7.5) from protein solutions containing the myosin‐X MyTH4–FERM cassette and β‐tubulin (390–445) in an equimolar ratio (112 μM). These crystals appeared to contain no β‐tubulin within the crystals. The crystals obtained were transferred stepwise into a cryoprotective solution containing 30% glycerol (for the MyTH4–FERM/DCC complex) or 25% ethylene glycol (for the MyTH4–FERM cassette) and flash cooled at 100 K. X‐ray diffraction data were collected at 100 K on a BL41XU or BL44XU beamline at SPring‐8. All data were processed and scaled using
HKL‐2000 (
Otwinowski and Minor, 1997). The crystal data are summarized in
Table I.
Structure determination and refinement
Phases of the complex crystal were calculated by a SAD method using data collected at the peak wavelength of selenium. Selenium positions were located using the program BnP (
Weeks et al, 2002) and phase refinement by solvent flattening was performed with RESOLVE (
Terwilliger, 2004). The built model was refined through alternating cycles using the Coot (
Emsley and Cowtan, 2004) and CNS (
Brünger et al, 1998) programs. The model was refined to 1.9 Å resolution. In the Ramachandran plots using MolProbity (
Davis et al, 2007), no outliers were flagged. The free form structure of the MyTH4–FERM cassette was determined by the molecular replacement method using the cassette model of the complex structure and refined at 2.55 Å. The refinement statistics are summarized in
Table I.
Structure and sequence comparison
Multiple sequence alignments of the MyTH4–FERM cassettes and the netrin receptors were performed by CLUSTALW (
Larkin et al, 2007). Pairwise structural comparisons were performed using C
α‐atom positions by the DALI lite server (
Holm and Park, 2000) and structure figures were prepared by PyMOL (
http://www.pymol.org/). In the course of the review process of this article, the crystal structures of the tail domains of myosins‐X and myosin‐VIIa have appeared (
Wei et al, 2011;
Wu et al, 2011). The appearance of structures display a similar fold of the MyTH4–FERM cassette seemed to provide a useful opportunity to review our structures, but significant deviations also appeared among these structures (see text).
Analytical ultracentrifugation
Sedimentation velocity ultracentrifugation experiments were performed at 20°C using a Beckman Coulter Optima XLA analytical ultracentrifuge. Purified samples were dissolved in 20 mM Tris buffer (pH 8.5) containing 150 mM NaCl and 5 mM DTT at a sample concentration of 1 mg/ml (14 μM) and then centrifuged at 25 000 r.p.m. The resultant data were analysed using the programs SEDFIT and SEDNTERP.
Binding studies by isothermal titration calorimetric analysis
ITC analysis was conducted using a calorimeter (MicroCal VP‐ITC, USA) at 20°C. Purified protein samples were dialysed overnight in buffer containing 20 mM Tris–HCl (pH 8.5) and 150 mM NaCl. Data fitting were performed using the ORIGIN™ software program supplied with the instrument.
Pull‐down binding assay
All mutations were produced by site‐directed mutagenesis. For in vitro pull‐down binding assays, the purified myosin‐X MyTH4–FERM domain cassette and GST‐fusion protein were mixed with a slurry of glutathione‐Sepharose 4B and incubated at 4°C. After washing with incubation buffer solution, the eluates were subjected to SDS–PAGE.
Co‐sedimentation assay with microtubules
Forty micromolar tubulin heterodimer in 100 mM PIPES (pH 6.9), 1 mM EGTA and 1 mM MgSO4 (PEM buffer) containing 1 mM GTP and 10 μM taxol was polymerized by incubation at 37°C for 20 min. Protein samples were mixed with the polymerized tubulin at a molar ratio of 1:5 (protein sample to tubulin) and the samples were further incubated at 37°C for 20 min. After centrifugation at 14 000 g at 4°C for 30 min, the resulting pellets and supernatants were subjected to SDS–PAGE. Bands of myosin‐X were analysed by a densitometric analysis for quantitative evaluation of the inhibitory effects by DCC.
Co‐immunoprecipitation
HEK293T cells were transiently transfected with cDNA by the calcium phosphate method. Two days after transfection, cells were washed with PBS and lysed with modified RIPA buffer (50 mM Tris–HCl pH 7.4, 150 mM NaCl, 1% NP‐40, 0.25% sodium‐deoxycholate, 2 mM phenylmethylsulfonyl fluoride and 5 μg/ml leupeptin) and centrifuged at 17 000 g for 10 min at 4°C. The supernatants were incubated with anti‐myc antibody (MBL, M047‐3) for 1 h at 4°C, and immune complexes were then precipitated with protein G‐Sepharose 4B (GE Healthcare) for 1 h at 4°C. After washing, the beads with lysis buffer, immunocomplexes were analysed by immunoblot.
Accession codes
Protein Data Bank: The coordinates and structure factors of the human myosin‐X MyTH4–FERM cassette and its complex with the DCC P3 peptide have been deposited under accession codes 3AU5 and 3AU4, respectively.