Calmodulin regulates dimerization, motility, and lipid binding of Leishmania myosin XXI

Significance

Myosin XXI is the only myosin isoform expressed in the Leishmania parasite. The myosin-XXI homozygous knockout is lethal, and a reduction in expression levels leads to loss of endocytosis and affects other intracellular trafficking processes. In this paper we show that myosin XXI can adopt a monomeric or dimeric state. The states are determined by calmodulin binding to an IQ motif that, when bound, prevents dimerization of a coiled-coil motif. In the monomeric state the motor binds phospholipids and is motile whereas the dimeric state is unable to bind lipids or to generate motility, but can cross-link actin filaments. Regulation of dimerization, motility, and lipid binding by calmodulin is a mechanism for the myosin family of motor proteins.

Abstract

Myosin XXI is the only myosin expressed in Leishmania parasites. Although it is assumed that it performs a variety of motile functions, the motor’s oligomerization states, cargo-binding, and motility are unknown. Here we show that binding of a single calmodulin causes the motor to adopt a monomeric state and to move actin filaments. In the absence of calmodulin, nonmotile dimers that cross-linked actin filaments were formed. Unexpectedly, structural analysis revealed that the dimerization domains include the calmodulin-binding neck region, essential for the generation of force and movement in myosins. Furthermore, monomeric myosin XXI bound to mixed liposomes, whereas the dimers did not. Lipid-binding sections overlapped with the dimerization domains, but also included a phox-homology domain in the converter region. We propose a mechanism of myosin regulation where dimerization, motility, and lipid binding are regulated by calmodulin. Although myosin-XXI dimers might act as nonmotile actin cross-linkers, the calmodulin-binding monomers might transport lipid cargo in the parasite.

Over 12 million people worldwide are affected by leishmaniasis, which is caused by the flagellated protozoan parasite Leishmania. The disease manifests itself in a cutaneous form, which leaves disfiguring scars, or as visceral leishmaniasis, which is potentially fatal (1). The parasite has a life cycle in two stages. The elongated, flagellated, and motile promastigotes are found in the gut of the sand fly host. The egg-shaped, nonmotile amastigotes possess only a rudimentary flagellum, are found in mammalian macrophages, and are responsible for the human disease pathology (2, 3). In contrast to other eukaryotic cells that express at least 11 different isoforms (4), Leishmania donovani seems to express only a single isoform, myosin XXI (5). A later classification assigned myosin XXI to class XIII, a kinetoplastide-specific class of myosins (6). Previous attempts to identify and localize myosin Ib in L. donovani parasites using anti-Leishmania myosin Ib antibodies were unsuccessful (3). Intriguingly, only myosin XXI was shown to be expressed in both the motile promastigote and the nonmotile amastigote stages of the parasite’s life cycle (3). In the promastigote stage, this motor preferentially localized to the proximal part of the flagellum, although it was also found along the entire length of the flagellum and in other cell-body compartments (7). The myosin-XXI homozygous knockout is lethal. The heterozygous cells were unable to form the paraflagellar rod, a structure of unknown function that runs along the length of the flagellum and contains a variety of proteins including actin and myosin XXI (8). It has also been reported that reduced expression levels cause the loss of endocytosis within the flagellar pocket and affect other intracellular trafficking processes (7). This makes myosin XXI an intriguing candidate to study modes of structural adaptation of a single myosin isoform for a variety of cellular acto-myosin–based motile functions. Two distinct cellular fractions of myosin XXI have already been identified: a membrane-bound fraction that localized to the base of the flagellum and a cytosolic fraction possibly involved in transporting proteins within the flagellum (3). However, it is unknown in which way the motor is targeted to different cargo, which oligomerization states it can adopt, and how transitions between different functional states are regulated.

The design of myosin XXI follows the general structure of myosin motors, which comprises a conserved N-terminal motor domain, followed by a neck region including IQ motifs for the binding of light chains of the calmodulin family, and finally a C-terminal cargo-binding tail domain (9) (Fig. 1A). Although the motor domain is responsible for the binding to actin and hydrolysis of ATP, it is usually the tail domain that determines its function within the cell by controlling dimerization or oligomerization, motor anchoring to membrane compartments, and selection and transport of specific cargo. The neck region, mechanically stabilized by binding calmodulins, is thought to serve as a lever arm to transduce force and movement to the cargo. In a previous study we identified six potential calmodulin-binding IQ motifs outside the motor domain of myosin XXI. The data suggested that only the motif closest to the motor domain bound Xenopus/human calcium-calmodulin (10). Intriguingly, sequence analysis indicated a coiled-coil domain (11, 12) with a strong propensity to dimerize in between the end of the motor domain and what we initially thought to be the first calmodulin-binding IQ motif. Closer inspection of this predicted coiled-coil domain, however, suggested a further potential calmodulin-binding IQ motif within the predicted dimerization site and close to the converter (yellow, Fig. 1A). This suggested that this myosin might be able to dimerize, but in doing so would lose its mechanically essential lever arm structure to the formation of a coiled coil and, as a consequence, make a transition from a motile monomeric form to a nonmotile dimer. Consistent with this, we found in a previous study that calcium-calmodulin binding was not required for the ATPase activity of the motor, but for myosin XXI in vitro motility (10).

Fig. 1.

Calmodulin binding prevents the dimerization of myosin XXI. (A) Domain structure of full-length myosin XXI. By comparing the sequence with well-studied myosins, we identified the highly conserved regions of the motor domain (gray). In addition to the calmodulin-binding IQ motifs identified in a previous study (10), we found another IQ motif (yellow) immediately following the converter domain. Sequence analysis revealed four potential coiled-coil regions (in blue). The predicted propensity to oligomerize was scored between 0 and 10 using Scorer 2.0 (11); positive numbers predict dimer and negative numbers predict trimer formation. (B) Size exclusion chromatography using a Superose-6 column and silver staining of SDS/PAGE gels (C) of the elution fractions indicate that the motor is monomeric when expressed in high levels of calmodulin and dimeric or in a higher oligomeric state when expressed at low levels of calmodulin. Aggregated myosin XXI would appear in the void volume. (D) The Western blot does not resolve endogenous calmodulin in the purified myosin-XXI preparations that were expressed in the absence of added calmodulin virus (i.e., at low levels of calmodulin). By varying the amount of calmodulin virus added during expression, the amount of calmodulin bound to the expressed myosin XXI could be controlled. The apparent difference in molecular weight of calmodulin in C and D is due to the differences in buffers.

To investigate the mechanism of dimerization and the mechanical properties of myosin XXI, we expressed L. donovani full-length myosin XXI, a truncated minimal motor domain, and a series of tail constructs. We also expressed L. donovani-specific calmodulin-like (CamL) proteins to characterize calcium-dependent binding to target peptides on the myosin-XXI heavy chain. Using size exclusion chromatography (SEC), analytical ultracentrifugation, Förster resonance energy transfer (FRET), motility assays, and electron microscopy (EM), we studied the effect of calmodulin binding on motor dimerization and motility. We used liposome pull-down experiments and phospholipid blot assays combined with sequence analysis to identify specific phospholipid-binding motifs in the head, neck, and tail domain of the motor. The data suggest a form of myosin regulation where dimerization, motility, and phospholipid binding of the motor are determined by binding of calmodulin.

Results

Calmodulin Binding Prevents Dimerization of Myosin XXI.

By comparing the sequence of the myosin-XXI motor domain with myosins for which crystal structures are available (e.g., refs. 13–15), we localized the highly conserved elements involved in chemo-mechanical energy transduction, including the ATP-binding pocket, switches I and II, and the relay loop (Fig. 1A). The converter domain (645–747) is thought to transform small conformational changes in the nucleotide-binding pocket to changes in the orientation of the calmodulin-binding neck domain that serves as a mechanical lever arm. Intriguingly, the sequence suggests a coiled-coil domain (11, 12) with a strong propensity to dimerize between the end of the motor domain and what we previously thought to be the first calmodulin-binding IQ motif (809–823) (10). This predicted coiled-coil region (730–810) revealed a further potential calmodulin-binding site (754–769, yellow in Fig. 1A) close to the converter. Downstream of that coiled-coil region we identified three more sections with potential to form a dimerizing coiled-coil or even to trimerize (blue sections, negative score for predicted trimers in Fig. 1A). To investigate whether calmodulin binding had an effect on dimerization, we coexpressed full-length myosin XXI (FL-XXI) with human calmodulin using a baculovirus/SF21-based expression system and analyzed the oligomerization state by SEC as described previously (Materials and Methods) (10) (Fig. 1 B and C). The SEC study showed that FL-XXI heavy chain, coexpressed with high levels of calmodulin virus, was monomeric. Using the approach of Coluccio (16), we had found in a previous study that the stoichiometry between myosin-XXI heavy chain and human calmodulin was 1:1 for these conditions of expression (10). In contrast, in the absence of added calmodulin virus, most of the purified myosin-XXI sample eluted earlier, indicating the formation of dimers or higher oligomers (Fig. 1B). The eluted fractions were analyzed by SDS/PAGE, and silver staining confirmed that calmodulin only eluted together with the myosin-XXI heavy chain in the monomeric fraction at intermediate or high levels of calmodulin expression (fractions 12–15, Fig. 1C, Lower). In the absence of added calmodulin virus, the levels of endogenous calmodulin in the Sf21 insect cells were too low to be detected by a Western blot (Fig. 1D, low levels of calmodulin expression). This suggests that the expression level of endogenous calmodulin in Sf21 was too low for a 1:1 stoichiometry of calmodulin to myosin-XXI heavy chain. For these conditions, SEC analysis showed a large peak for the dimeric or oligomeric fraction of myosin XXI and a very small peak for the monomeric fraction, presumably with endogenous insect calmodulin bound (Fig. 1B). This small peak was the only peak observed previously with the Superdex 200 column (10). The major peak eluting here at 10 mL was in the void volume on the Superdex 200 column in the previous study. The expression levels of calmodulin could be varied by varying the amount of added calmodulin virus and are reflected in a qualitative fashion in the signal strength of the Western blot in Fig. 1D (high and intermediate levels of calmodulin expression). The experiments showed that a change in the level of coexpressed calmodulin in the Sf21 cells resulted in a different oligomerization state of myosin XXI.

Two Different Leishmania CamL Proteins Bind to the Same Target Sequence on Myosin XXI.

To investigate in more detail how calmodulin binding affects dimerization of myosin XXI, we studied the binding of human calmodulin and of Leishmania-specific calmodulin-like proteins to predicted target sequences on the myosin-XXI heavy chain. Sequence analysis showed that there are only eight CamL proteins in the L. donovani genome. Calmodulin is highly conserved between species, and Xenopus or human calmodulin (CamH, NCBI Gene ID 4502549) is 90% identical to L. donovani CamL1 (NCBI Gene ID 13392113). Using change in tryptophan fluorescence of the target peptide as described previously (17), we investigated CamH and CamL1 binding to the target sequence 754–769 that is close to the myosin-XXI converter domain and within the strongest predicted dimerization domain of myosin XXI (Fig. 2A). Titrations of increasing concentrations of CamH or CamL1 against the target peptide saturated at a stoichiometry of 1:1, consistent with the target peptide binding a single CamL1 or CamH protein. Nonlinear least-squares fitting of the fluorescence signal at increasing CamH or CamL1 concentrations (18) yielded similar dissociation constants with K_d values of 21 ± 5 nM for CamH and 18 ± 3 nM for CamL1. Binding to the target peptide was dependent upon calcium and fully reversible (Fig. 2B). In contrast, another calmodulin-like protein from the L. donovani genome, which shares only 30% sequence identity with human calmodulin CamL2 (Gene ID 13386921), also bound to the target peptide in the presence of calcium, but, once bound, remained bound in a calcium-independent fashion (Fig. 2C). This showed that at least two different Leishmania-specific CamL proteins can bind to the same target peptide in the strongest predicted coiled-coil domain in myosin XXI and could affect myosin-XXI dimerization. To characterize oligomerization of the myosin-XXI tail in the absence of calmodulin in more detail, we expressed a number of myosin-XXI tail constructs. The oligomerization state of the constructs was analyzed by a combination of SEC, analytical ultracentrifugation, and FRET studies (Fig. 3 and Table 1).

Fig. 2.

Analysis of calmodulin binding to predicted calmodulin-binding motifs. Tryptophan fluorescence (excitation at 290 nm, emission at 323 nm) was used to study calmodulin binding to target peptide sequences. (A) The synthetic peptide of the first predicted calmodulin-binding motif C-terminal of the converter (aa 754–769) bound to human calmodulin (CamH) and to Leishmania CamL1 with a 1:1 stoichiometry and a K_d of 20 nM in the presence of calcium (pCa 4). (B) For this target peptide, calmodulin binding was reversible and calcium-dependent. (C) Leishmania CamL2 also bound to the peptide in the presence of calcium, but remained bound when free calcium was subsequently lowered to less than 100 nM.

Fig. 3.

Analysis of myosin-XXI tail constructs to investigate domains involved in dimerization. (A) SEC data of myosin-XXI tail constructs. The UV signals (in arbitrary units) were normalized to the highest peak for each experiment. Each construct was loaded at ∼30 μM. (B) SEC showed that, in the presence of the N-terminal RFP tag, both monomer and dimer conformations were observed for the tail constructs. The ratio of monomer to dimer was dependent on protein concentration. The calculated molecular weight for RFP and CFP is 27 kDa. (C) For the RFP-830–930 construct, the presence of the N-terminal RFP tag was insufficient to induce dimerization. (D) FRET study of a control CFP-RFP fusion protein containing a cleavable TEV site. CFP-donor excitation was performed at 445 nm. The emission spectrum was recorded from 450 to 650 nm (in gray for the uncleaved fusion protein and in black for the fusion protein following enzymatic cleavage). Protein concentration before cleavage was 7 μM; measurements were done at 22 °C. (E) CFP-donor emission at 475 nm (blue) and RFP-acceptor emission at 605 nm (red) for a mixture of 2-μM CFP-730 tails and 7-μM RFP-730 tail constructs in the presence of 100 μM calmodulin and 1 mM CaCl₂. (F) In the absence of calcium calmodulin, the mixture of 2-μM CFP-730-tails and 7-μM RFP-730 tail constructs resulted in a FRET signal due to the formation of heterodimers. (G) Summary of the FRET experiments with 1:1 mixtures of RFP-730 and CFP-730 tails (2 μM each) in the presence and absence of 100 μM calmodulin and of the control CFP-RFP fusion protein (2 μM) before and after enzymatic cleavage. The FRET efficiency was calculated as described in Materials and Methods. Each experiment was performed at least three times with three different preparations.

Table 1.

Summary of hydrodynamic studies on myosin-XXI constructs to determine their oligomerization state

Tail construct	Stokes radius from SEC/AUC* (nm)	Estimated Stokes radius from molecular weight (nm)	Molecular weight from amino acid sequence and MS† (kDa)	Calculated molecular weight (kDa)	Monomer/dimer
730-tail	3.77	3.12	41	80	D
800-tail	2.81*	2.9	32.1†	28	M
830-tail	2.78*	2.81	28.6†	25	M
930-tail	2.73*	2.43	18†	19.5	M
RFP-730 tail
Peak 1	4.18			117	D
Peak 2	3.42	3.61	68	56	M
RFP-830 tail
Peak 1	4.38			141	D
Peak 2	3.48	3.42	56	59	M
RFP-930 tail
Peak 1	4.06			106	D
Peak 2	3.33	3.23	46	50	M
RFP-730–830
Peak 1	4.11			110	D
Peak 2	3.2	3.19	44	44	M
RFP-830–930	3.17	3.17	43	51	M
FL-XXI
Low CamH	10.43	4.65	147		D
High CamH	5.29	4.8	164		M
Trunc XXI (amino acids 1–800)	4.17	4.20	120		M

The hydrodynamic properties of full-length and truncated myosin XXI and myosin-XXI tail constructs were determined using SEC and AUC. For the full-length and truncated myosin-XXI constructs, the native molecular weight had been determined previously (10) from their Stokes radius measured by SEC and their sedimentation coefficient, which was determined by sucrose gradient centrifugation as described by Post et al. (19). For the tail constructs, a near-globular shape was assumed. A native molecular weight was estimated by comparing the Stokes radius determined from the SEC elution volume, AUC, and mass spectrometry (MS) to the calculated Stokes radius determined for the theoretical molecular weight of the constructs. For RFP-730, a smaller Stokes radius was determined compared with RFP-830, which has a lower molecular weight. This suggests that the RFP-730 construct can adopt a more compact conformation. D, dimer; M, monomer.

Dimerization Involves Neck and Tail Domains of Myosin XXI.

To simplify, we will call all constructs “tails” that begin downstream of the converter region (645–747) of myosin XXI. The tail constructs in Fig. 3A were designed to study the dimerization propensity of the four predicted coiled-coil segments. The predicted propensity was scored between 0 and 10 using Scorer 2.0 (11) (Fig. 1A; positive numbers predict dimer and negative numbers predict trimer formation). The SEC experiment in Fig. 3A was carried out in the absence of calcium-calmodulin and shows that all four constructs eluted as a single peak, consistent with the samples adopting predominantly a single oligomeric state. As expected for proteins of near-globular shape, the 800-, 830-, and 930-tail constructs ran through the SEC columns approximately to their calculated molecular weight, following the calibration curves for standard proteins of approximately globular shape, as described previously (10) (Materials and Methods). The Stokes radii of these constructs were determined by a combination of SEC and analytical ultracentrifugation (Materials and Methods and Table 1). The results from both methods were consistent and suggested that the distal 800-, 830-, and 930-tail constructs were monomeric. Following this argument we also inferred the oligomeric states of the largest tail construct directly from the Stokes radius, which was calculated from the elution volume (19) (Table 1 and Materials and Methods). This 730-tail construct eluted too early to be monomeric and, when sized, strongly suggested a dimeric state. The 730-tail constructs composed the strongest predicted coiled-coil segment in full. To study the oligomeric state of the different sections of the myosin-XXI tail at lower protein concentrations, we used FRET. For these studies, we generated tail constructs with an N-terminal red fluorescent protein (RFP) or cyan fluorescent protein (CFP) tag. SEC of the RFP-tagged constructs showed that the RFP-730, RFP-830, and RFP-930 tails eluted with two peaks, consistent with a mixed monomeric and dimeric population (Fig. 3 B and C and Table 1). In the presence of the N-terminal RFP tag, all tail constructs in fact adopted mixed monomeric and dimeric populations (Table 1). The only exception was RFP-830–930, which remained purely monomeric. The coiled-coil segment of this construct seemed too weak to dimerize, even in the presence of the RFP tag. This showed that the RFP tag on its own was not able to induce dimerization. The RFP-tagged sequences needed a propensity to dimerize with a score >0.6 to induce the formation of dimers.

For the FRET studies, we exploited the fact that mixed conformations were obtained with the RFP-tagged tail constructs. This enabled us to study the formation of heterodimers by mixing CFP- and RFP-tagged tails in the presence and absence of calmodulin. The formation of heterodimers was detected by FRET between the N-terminal CFP and RFP tags. As a control experiment, we generated a CFP-RFP fusion protein with a cleavable TEV site connecting the two fluorophores (20). Fluorescence excitation of the CFP donor was induced at 445 nm, and the emission spectrum was recorded between 450 and 650 nm (Fig. 3D). Following cleavage at the TEV site, CFP-donor emission at 475 nm was increased and RFP-acceptor emission at 605 nm was reduced. The increased RFP-acceptor emission at 605 nm for the uncleaved protein indicated FRET between the linked CFP-RFP fluorophores. The energy transfer efficiency E was determined as described (21) from the quenching of fluorescence of the donor molecule by the relation E = [1 − (F_da/F_dn)] × 100, with F_da and F_dn the donor fluorescence intensities at 475 nm in the presence and absence of the energy acceptor, respectively (Fig. 3G). When CFP- and RFP-730 tails were mixed in the presence of ∼10-fold molar excess of CamH, a FRET efficiency of only about 1% was obtained, similar to the FRET signal for the fully cleaved CFP-RFP fusion protein (0.4%, Fig. 3 E and G). This suggested that the binding of CamH to the monomeric CFP- and RFP-730 tails prevented the formation of heterodimers. When the tails were mixed in the absence of CamH, we observed a fast increase in FRET efficiency, followed by a slower rate of signal change over about 1 h until a steady state was reached (Fig. 3F). The fast signal change was probably due to free monomers forming heterodimers, whereas the slow rate was likely due to an exchange between homo- and heterodimers. As illustrated in the cartoon in Fig. 3F, this result further supports the hypothesis that calmodulin binding regulates myosin-XXI dimerization. The next question was whether the monomeric and/or dimeric conformations of the motor would bind to lipids and therefore could account for the membrane-bound fraction of myosin XXI observed in the parasite.

Lipid-Binding Sections of Myosin XXI Include the Converter, Neck, and Tail Domains.

To test whether myosin XXI would bind to lipids incorporated into a bilayer maintaining geometry and curvature of a physiological membrane, we performed pull-down experiments with liposomes produced from bovine brain extract [Folch fraction I (22); Sigma] following standard procedures (23, 24). Several freeze–thaw cycles and extrusions through 100-nm pore filters were performed to obtain unilamellar vesicles with a size distribution of about 150 ± 30 nm. The lipid-binding studies with FL-XXI were carried out in the presence and those with the tail constructs in the absence of calcium-CamH. Remarkably, the purely monomeric constructs, such as the 830-tail and the truncated motor domain (Trunc XXI) (10), were pulled down completely with the liposomes (Fig. 4A). In contrast, the mixtures of monomers and dimers, such as FL-XXI expressed at intermediate levels of CamH (Fig. 1D), and the RFP-tagged tails, were partially pulled down with the liposomes whereas a fraction remained in the supernatant. This suggested that dimerization and lipid binding were mutually exclusive for myosin XXI. The same result was obtained with a protein lipid overlay (PLO) assay where lipids were blotted onto a nitrocellulose membrane and protein binding was detected using antibodies against the His tag of the recombinant proteins (Fig. 4B). The dimeric 730-tail construct did not bind to the Folch lipid, whereas the monomer/dimer mixtures of the RFP-tagged constructs bound in a concentration-dependent fashion. These assays in addition confirmed that the tails bound only to the mixed lipids of the Folch preparation, but not to the pure, main Folch constituents, namely phosphatidylcholine (PC) or phosphoethanolamine (PE). Furthermore, we found that the proximal (730–830), the medial (830–930), and also the distal (930-tail) myosin-XXI tail constructs bound to mixed lipids. To check which specific lipids the tails would bind to, we used commercial lipid strips (Fig. 4C). These experiments showed phospholipid binding along the neck and tail of the motor (Fig. 4D). Although the binding of phosphatidyl inositol monophosphates was found along the entire molecule outside the motor domain, binding of di- and triphosphates seemed to be more localized at the proximal part of the tail (Trunc XXI and 730–930 construct, Fig. 4C). Comparing phospholipid binding of monomeric (fraction 7) and dimeric (fraction 2) RFP-730–830 constructs further confirmed that only the monomeric fraction of myosin XXI bound to the lipids.

Fig. 4.

Lipid binding of myosin XXI. (A) Pull-down experiments of Folch mixed liposomes and myosin-XXI constructs showed that monomeric constructs bound to liposomes. (B) PLO experiments indicated lipid binding along the entire tail of monomeric myosin XXI, whereas dimers (730-Tail) did not bind to lipids. The PLO data also showed that myosin XXI binds to Folch mixed lipid preparations but not to pure PC or PE lipids. (C) Binding to lipid membrane strips showed that all constructs bound to PIP monophosphates. Binding to di- and triphosphates was found N-terminally of amino acid 830. Binding studies of RFP-730–830 confirmed that the monomeric form (fraction 7) bound to phospholipids, whereas the dimeric form (fraction 2) did not. The fractions collected from a SEC experiment on the RFP-730–830 construct were analyzed using silver-stained SDS/PAGE (Inset) and confirm that both peaks originate from a single protein. (D) The cartoon indicates capacity for phospholipid binding along the entire myosin-XXI tail, including the converter region (645–747).

Phospholipid-Binding Motifs in the Neck and Tail of Myosin XXI Overlap with the Dimerization Domains.

To identify probable phospholipid-binding motifs in the neck and tail of myosin XXI, we compared the sequence to known motifs that have been described for other myosins. We found six nonspecific phospho-lipid-binding domains (LBDs) (Fig. 5A) with basic residues flanking a central hydrophobic patch (Fig. 5C), as described for class I myosins (25, 26). Two of those domains overlapped with potentially unstructured lipid-binding domains (ULBDs in Fig. 5A) rich in basic and hydrophobic residues, which were identified by calculating a basic-hydrophobic score over a running window of 20-amino-acid residues along the sequence, as described by Brzeska et al. (27) (Fig. 5D). Consistent with the experimental results, the predicted lipid-binding sites were found along the entire sequence C-terminal of the myosin-XXI motor domain. Furthermore, they nearly fully overlapped with the dimerization domains (predicted coiled-coil, Fig. 5B).

Fig. 5.

Sequence analysis of myosin XXI to localize lipid binding and dimerization. (A) We identified a number of potential lipid-binding sites, including a PX domain that overlapped with the converter region of the motor. (B) The potential lipid-binding sites (green) overlapped with the potential dimerization domains (blue). (C) Six lipid-binding domains (LBD1–6) with basic residues flanking a hydrophobic patch of amino acids (25, 26) were found. (D) Using the basic-hydrophic scale (25), we identified a potential lipid-binding site following the PX domain and confirmed LBD5 as a site with high potential for phospholipid binding.

Myosin-XXI Converter Contains a Phospholipid-Binding Phox-Homology Domain.

We also identified a phox-homology (PX) consensus sequence [R (Y/F) x_23–30 K x_13–23 R] for phosphatidylinositol 3-phosphate [PI(3)P] binding (28) that overlapped with the end of the converter domain of the motor head (Fig. 5A). The amino acid residues of the PX consensus sequence are thought to form an electro-positive basic patch to bind the negatively charged phosphate groups of phosphoinositides (29). To investigate whether the critical residues of the PX consensus sequence were located at the surface of the converter domain and potentially accessible for interaction with lipid membranes, we used the crystal structure of scallop muscle myosin II (30) and replaced the sequence in the converter domain by the sequence of L. donovani (Fig. 6). The four critical residues R (Y/F) x_23–30 K x_13–23 R were in nonconserved regions and were all located at the surface of the converter domain and should therefore be able to form a patch for lipid binding without interfering with the structure and function of the motor. We analyzed 130 sequences of the converter domain of 21 other myosin classes (Dataset S1) but only myosin XXI contained the PX-domain consensus sequence for phospholipid binding in this region (Fig. 6). We expressed a construct comprising the myosin-XXI converter domain (600–758), including the WT PX domain, as well as a PX-mutant construct with the first two amino acids of the consensus sequence, RY, replaced by LS. We found that the WT converter construct bound not only PI(3)P monophosphate, which is specific for PX domains (28), but also the two other monophosphates, PI(4)P and PI(5)P, as well as the diphosphate PI(3,5)P₂. A weak interaction was also found for other anionic lipids, such as phosphatidylserine (PS, Fig. 4C). For other PX domains, this has been related to the presence of adjacent hydrophobic residues that might form an additional positively charged surface patch (28, 29). For the PX-mutant construct, however, the binding of the monophosphates was retained whereas the binding of the diphosphate PI(3,5)P₂ and of PS was completely abolished. This strongly supports the hypothesis that the PX domain in the converter is involved in phospholipid binding.

Fig. 6.

Structural analysis of the PX domain in the myosin-XXI converter. (A) We used the crystal structure of scallop muscle myosin II (30) and replaced the sequence in the converter domain with the sequence of L. donovani. The amino acids specific to the PX domain (in green) form a patch on the surface of the converter (in red), consistent with these residues interacting with membranes without interfering with the general structure and function of the converter of this myosin motor. Comparison of 130 sequences from 21 myosin classes (Dataset S1) suggest that myosin XXI is exceptional within the myosin family in terms of featuring a PX domain in the converter region. (B) PIP strips showing that site-directed mutagenesis of arginine tyrosine (RY) in the expressed PX mutant converter construct completely abolished the characteristic Pi(3,5)P₂ binding of the PX domain.

Mechanical Properties of Monomeric and Dimeric Myosin XXI.

Finally, we studied the mechanical properties of monomeric and dimeric myosin XXI in the presence of actin using in vitro motility assays and electron microscopy. The model in Fig. 7A summarizes our current results. At least two different calmodulin-like proteins (CamL1 and CamL2) in L. donovani can bind to the same calmodulin-binding motif (754–769) following the converter and interfere with dimerization. Their binding differs in K_d and calcium sensitivity. Monomeric, CamH-binding myosin XXI (Fig. 1C, Lower, fraction 14) is motile and binds to phospholipids. Immobilized unspecifically on a nitrocellulose surface (31) in the presence of a fivefold molar excess of calcium-CamH and at 2 mM ATP, monomeric myosin XXI moved rabbit skeletal actin filaments (n = 52) at 18 ± 3 nm⋅s⁻¹ (Fig. 7B and Movie S1). Intriguingly, the velocity was about four times higher when the monomeric myosin XXI was bound to Folch membranes. The concentrations of myosin XXI required to produce smooth motility were reduced 30-fold, suggesting that, when bound to lipids via the tail, myosin XXI is in an improved configuration to support motility (Fig. 7B and Movie S2). In contrast, dimeric myosin-XXI motors (Fig. 1C, Upper, fraction 10) immobilized on nitrocellulose bound actin filaments, but were unable to move them. In the presence of 2 mM ATP, the actin filaments were broken into smaller fragments by the interaction with the dimeric myosin-XXI motors (Movie S3). To obtain structural information on monomeric myosin XXI, we performed negative-stain electron microscopy on monomeric full-length myosin-XXI molecules, adsorbed to carbon-coated EM grids in rigor (ATP < 1 μM). The molecules were classified as described previously (32). The two representative class averages (52 and 56 images, respectively) in Fig. 7C are shown with overlaid crystal structures of a truncated myosin-V motor domain (red) in rigor with a single bound calmodulin (green) [Protein Data Bank (PDB) 1OE9]. Although the class averages of the myosin-XXI motor domain are consistent with different surface projections of the myosin-V crystal structure, the calmodulin-binding region is more difficult to localize. And, although in image (i) in Fig. 7C this domain seems clearly visible, a large part of it seems to be hidden in image (ii). The same holds true for the class average of 400 images of myosin XXI bound to actin in image (iii) (PDB 1OE9 and PDB 4A7F). This suggests that at least under these conditions the calmodulin-binding domain either is found in variable orientations with respect to the motor domain or is folded back onto the motor domain. Negatively stained EM images of dimeric myosin XXI in rigor and in the presence of actin showed that this myosin had a tendency to cross-link actin filaments as shown in Fig. 7D.

Fig. 7.

Model of myosin-XXI regulation of dimerization, lipid binding, and motility. (A) The model summarizes our findings. At least two different calmodulin-like proteins (CamL1 and CamL2) in L. donovani can bind to the same calmodulin-binding motif (754–769) following the converter and prevent dimerization. Their binding differs in K_d and calcium sensitivity. Monomeric, calmodulin-binding myosin XXI is motile and binds to phospholipids, whereas the dimeric motor does not and is nonmotile. (B) The velocity of actin filaments driven by monomeric myosin XXI unspecifically adsorbed to nitrocellulose is four times lower compared with motors bound to Folch lipid bilayers. (C) Negatively stained electron micrographs of single, monomeric myosin XXI in rigor (ATP < 1 μM) adsorbed to carbon-coated EM grid on their own and when bound to actin. Representative class averages are shown. Crystal structures of truncated myosin V in rigor with a single light chain bound (green) have been overlaid. The numbers state the number of images contributing to the class average. (D) Representative negatively stained electron micrograph of two actin filaments cross-linked by dimeric myosin XXI in rigor (ATP < 1 μM).

Discussion

In summary, our experiments have provided insights into the structural and functional properties of the unusual myosin motor in the protozoan parasite Leishmania, myosin XXI. In less ancient eukaryotic cells, such as mammalian cells, more than 10 different isoforms are expressed and responsible for various forms of motility (33). Loss of any particular myosin may or may not reveal a clear phenotype owing to the functional compensation by other myosins. With only two myosins in the Leishmania genome and with only a single isoform expressed, which, however, is vital for parasite survival (3), myosin XXI provides an intriguing opportunity to study modes of regulation and structural-mechanical adaptation for diverse motile functions of a single myosin. In this study we have focused on myosin-XXI oligomerization and lipid cargo binding.

Myosin-XXI Dimerization.

An unexpected finding was revealed by sequence analysis. The motor has eight potential calmodulin-binding sites, only one of which bound human calmodulin in our previous study (10). We have now identified a binding site close to the converter domain that is located within an extended dimerization site covering the entire C terminus of the molecule from the converter region to the C-terminal end. This site bound the L. donovani-specific calmodulin-like proteins CamL1 and CamL2 in a calcium-dependent fashion. This was unexpected because in other myosins that dimerize or oligomerize, for example, myosin class II, V, or VI (34, 35), the predicted coiled-coil domain starts only downstream of the calmodulin-binding neck region, which serves as a mechanical lever arm to transduce force and movement to the cargo-binding C terminus. The lever arm structure is thus unaffected by motor dimerization. For myosin XXI, however, we found that calcium-calmodulin binding not only prevented dimerization by precluding the formation of a dimerizing coiled-coil but also seemed to enable motility in the first place by generating a functional lever arm. Consistent with this hypothesis, we found in a previous study that, in contrast to monomeric full-length myosin XXI with calmodulin bound, the truncated construct (1–800) was also monomeric, but did not bind calmodulin and was not motile, although it had an actin-activated ATPase activity similar to that of the full-length motor (10). This construct comprised the catalytic head domain including 53 amino acids downstream of the converter. However, it did not bind CamH and therefore probably could not form a functional lever arm. Its inability to bind calmodulin might be related to some sort of backfolding of the C-terminal sequence onto the catalytic domain, such as the EM images suggested for the full-length motor (Fig. 7C). In the absence of calmodulin, the entire myosin-XXI C terminus downstream of the converter region seemed to dimerize. Here, loss of the lever arm due to the formation of a coiled-coil could explain why the dimer was able to bind to and cross-link actin filaments, but unable to generate any movement. Still binding of actin filaments in the presence of ATP caused the actin filaments to break, and the biochemical actin-activated ATPase cycle of the dimer seemed to be at least partly uncoupled from the mechanical output of the motor.

Myosin-XXI Dimerization and Lipid Binding Are Mutually Exclusive.

The second finding was that the predicted coiled-coil sections of the molecule nearly fully overlapped with lipid-binding domains. Furthermore, lipid binding and dimerization were mutually exclusive. This suggested a mechanism of myosin regulation and targeting to cargo: in the absence of calmodulin, the myosin-XXI motor seemed to be dimeric and nonmotile, but able to bind to and cross-link actin filaments. The dimer is expected to be cytosolic because it was unable to bind to lipids. In contrast, calmodulin-binding brought the motor into a monomeric state that can generate force and movement and bind to lipid cargo. Lipid binding at various sites along the neck and tail of the molecule might target the monomer to specific lipid compartments. Interestingly, the lipid-binding studies with the tail constructs suggested that calcium-CamL binding to different target sites along the tail might have a regulatory effect on tail binding to specifc compartments.

The third finding was that membrane binding of myosin XXI was not necessarily mediated in a stereospecific fashion, as described, for example, for myosin-I pleckstrin homology domains that interact specifically with PI(4,5)P₂ (36–38). For other myosin-I isoforms, however, membrane targeting does not seem to require stereospecific phosphoinositide recognition (25, 26, 39). They seem to bind via more unspecific electrostatic interactions with a variety of charged phospholipids. This is consistent with our observation for myosin XXI, where all three monophosphates bound to the converter (645–747) and all of the tail constructs. The diphosphates PI(3,5)P₂ and PI(4,5)P₂ seemed to bind N-terminally of amino acid 830 and the triphosphate PI(3,4,5)P₃ between 758 and 830. Furthermore, our sequence analysis predicted six lipid-binding domains along the myosin-XXI C terminus following the converter, with basic residues flanking a central hydrophobic patch, as described for nonspecific charged interactions with phospholipids for Acanthamoeba, Dictyostelium, and human myosin class I (25–27). Two of those regions scored significantly on the basic-hydrophobic scale, as described by Brzeska et al. (27), to identify putative unstructured lipid-binding sites in myosin tail sequences. The consensus sequence for a phospholipid-binding PX domain that overlaps with the converter region of myosin XXI seemed to be unique among myosins. PX domains in Saccharomyces cerevisiae reportedly have high specificity, and mammalian ones show a preference for PI(3)P monophosphate (40). However, for some PX domains, binding of mono, di-, and triphosphoinositides have been reported (28, 40). This is consistent with our finding for myosin XXI where the respective sequence (600–758 construct) bound to all three monophosphates, PI(3,5)P₂, and also to phosphatidyl serine (PS) whereas for the PX-mutant binding to the PI(3,5)P₂ diphosphate and to PS was completely abolished. For the kinesin family of motors, a PX domain has been reported for KIF16B (41). This PX domain was located at the C terminus of the molecule and targeted the motor to early endosomes via binding to PI(3)P monophosphate. As in our case, however, the PX domain in KIF16B did not show high selectivity for PI(3)P. It also bound other phosphoinositides with nanomolar affinity, in this case PI(3,4)P₂ and PI(3,4,5)P₃, which are typically found at the plasma membrane.

Potential Cellular Functions of Myosin-XXI Monomers and Dimers.

With lipid-binding sections along the converter, neck, and tail domain of the molecule, monomeric myosin XXI seems to be well suited to be targeted to lipid compartments with diverse lipid compositions. This might include the plasma membrane, but also organelles or vesicular cargo involved in intraflagellar transport, where myosin XXI might anchor these structures to and transport along the actin cytoskeleton (7). As shown in the cartoon in Fig. 6A, we speculate that the monomeric, motile, and lipid-binding conformation corresponds to a myosin-XXI fraction that has been localized at the base of the Leishmania flagellum (3, 7). In contrast, the free, cytosolic dimeric fraction of myosin XXI might correspond to the detergent-labile component that has been localized within the flagellum where it could contribute to the structural organization of the actin network (3, 7). Apart from demonstrating the presence of actin filaments in the main cell body and the flagellum using immuno-labeling of fixed Leishmania parasites, to date little is known about the structure and dynamics of the actin cytoskeleton in this system (3). Future studies investigating colocalization and distribution of actin and myosin-XXI monomers and dimers in the parasite will be revealing. As illustrated in the cartoon, the monomeric form might bind to single monophosphates or form ensembles at membrane clusters of PIP₂ with its different lipid-binding domains. Our lipid-binding studies with full-length myosin XXI and tail constructs in the presence or absence of calcium-calmodulin indicated that calmodulin binding was not required for the tail domains to bind to lipids. However, calcium-calmodulin binding might be involved in sorting the motor to different target membranes. Consistent with this, it was recently reported that calmodulin is required for paraflagellar rod assembly and flagellum–cell-body attachment in trypanosomes (42).

In conclusion, we found that Leishmania myosin XXI is a system, where calcium-calmodulin regulates dimerization, motility, and lipid binding of the motor. The following observations suggest that expression levels of different Leishmania-specific CamLs and cytosolic calcium might regulate myosin function in this system: (i) two very different Leishmania CamL proteins bound to a single-peptide sequence in the strongest dimerization domain on myosin XXI, and these proteins differ in K_d and calcium sensitivity; (ii) in vitro (human) calmodulin binding affected myosin-XXI dimerization, lipid binding, and motility; (iii) cytosolic calcium levels in the parasite are reported to be kept very low and tightly regulated (43); (iv) the expression levels of CamLs in the parasite are still largely unknown. However, down-regulation of cytosolic calmodulin affects the parasite’s paraflagellar rod assembly (42), which is where myosin XXI has been localized (3, 7). Intriguingly, we have now identified eight other calmodulin-like proteins in the fully sequenced genome of the Leishmania parasite. They might interact with the seven additional calmodulin-binding sequences outside the converter. Future studies will show which other conformations and oligomerization states the myosin-XXI motor can adopt, what the mechanical properties of these complexes are, and how they might be used for specific cellular functions in the parasite. Apart from the interest in the basic mechanism of energy transduction and regulation of myosin motors, myosin XXI might also be an interesting potential drug target as it has been shown to be vital for Leishmania parasite survival.

Materials and Methods

Plasmids and Generation of Recombinant Baculovirus.

FL-XXI was chemically synthesized, cloned into pUC57 (Genscript USA Inc.), and subcloned into the 6 × His-tagged vector pFastBacHb (Invitrogen) using standard PCR methods as described previously (10). An amino acid 1–800 truncation of myosin XXI (Trunc XXI) was created using PCR, and the N terminus of FL-XXI and Trunc XXI modified by addition of EGFP (for details, see ref. 10). Recombinant baculovirus DNA was generated using the Bac-to-Bac method before transfer into Spodoptera frugiperda (SF21) cells. Human calmodulin (Gene ID 4502549; identical sequence in Xenopus laevis) was amplified from a P4 stock before being transferred into SF21 cells for coexpression. Calmodulin sequences are highly conserved between different species. Compared with Leishmania calmodulin-like protein CamL1 (Gene ID 13392113), the sequence of human calmodulin differs only in 15 amino acids, with eight conservative differences. Therefore, we expected similar binding properties for human calmodulin compared with the endogenous Leishmania CamL1. We also expressed Leishmania CamL2 (Gene ID 13386921) that shares only 30% sequence identity with human calmodulin. However, like CamL1, CamL2 colocalizes with the myosin-XXI heavy chain when coexpressed in HeLa cells. Nonfluorescent myosin-XXI tail fragments were cloned into a pET28a vector (Invitrogen) via HindIII/XhoI restriction sites using standard cloning techniques.

The following oligonucleotides were used for PCR (boldface indicates restriction enzyme sites): XXI-730FP-nonF (GGGAAGCTTTAGGCAAGACGAAGGTGTTCCTCC); XXI-800FP-nonF (GGGAAGCTTTA GACGCCGCCAATGGTGTGTGC); XXI-830FP-nonF (GGGAAGCTTTAGCCGTCGAGGCGG ACACGCGCG); XXI-930FP-nonF (GGGAAGCTTTAGGCACGGACAGCGAATATA TGCC); XXI-830RP (ACTCGAGCTAGCGCGTGTCCGCCTCGACGGC); XXI-930RP (ACTCGAGCT AGGCATATTCGCTGTCCGTGCC); and XXI-ENDRP (ACTCGAGCTAGCTCACCTTGAACAGC). monomeric RFP-tagged myosin-XXI tail constructs were cloned into pET28a vector containing N-terminal monomeric RFP via the following NotI/XhoI restriction sites: XXI-730FP (AGAGGCGGCCGCCGGCAAGACGAG GTGTTCCTCC); XXI-830FP (AGAGGCGGCCGCCGCCGTCGAGGCGGACACGCGCG); and XXI-930FP (AGAGGCGGCCGCCGGCACGCGGACAGCGAATATGCC). The tail fragments were expressed in Escherichia coli and purified as described previously (10).

Protein Expression and Purification.

Protein expression and purification are based on previously published protocols (9).

Liposome Cosedimentation.

Liposomes were prepared from bovine brain extract, type I, Folch fraction I (Sigma) following the protocol by Spudich et al. (2007) (23). Five freeze–thaw cycles were applied to obtain unilamellar vesicles, and the liposomes extruded 11 times using a 100-nm pore filter to obtain a size distribution of 150 ± 30 nm. The liposomes (1 mg⋅ml⁻¹) were mixed with myosin XXI (250 nM) in liposome buffer (in mM: 20 Hepes, pH 7.4, 150 NaCl, 1 DTT) and incubated at room temperature for 10 min before centrifugation at 160,000 × g for 15 min at 4 °C. The pellet (resuspended in liposome buffer) and supernatant were run on SDS/PAGE and stained with Coomassie.

PLO.

The assay was performed essentially as described by Dowler et al. (44). We prepared 1-mM stocks of the lipids (Avanti) in chloroform. A total of 500, 350, or 200 pmol of the lipids and 0.5 μg of the target protein were blotted onto the nitrocellulose membrane (Hybond-C Exra, Healthcare). The membrane was allowed to dry for 1 h at room temperature before being incubated for 1 h in blocking buffer [in mM: 50 Tris⋅HCl, pH 7.5, 150 NaCl, 0.1% Tween 20 plus 2 mg⋅ml⁻¹ fatty acid-free BSA (Sigma)]. This was followed by an overnight incubation with the target protein (5–10 nM) in blocking buffer plus 2 mg⋅ml⁻¹ BSA at 4 °C. The membrane was then washed 10 times in blocking buffer before incubation for 1 h at room temperature with anti–His-HRP antibody (Abcam ab1187) at a 1:2,000 dilution. The membrane was washed again 12 times with blocking buffer. Antibody binding was detected using an ECL kit (Invitrogen) and imaged using a Biorad Geldoc system. Target protein binding to lipid membrane strips and phosphatidyl-inositol phosphate (PIP) strips (Echelon Inc.) were carried out as described above for reconstituted liposomes. Here the incubation with the target protein was reduced to 1 h at room temperature. For the PIP strips, analysis of the RFP-tail constructs was limited because of unspecific interactions of RFP with the PIP strip. Therefore, we focused the experiments on the untagged constructs.

SEC.

Purified proteins (∼30–50 μM) were loaded onto a Superdex-200 (10/300 GL) analytical column (GE Healthcare) or a Superose-6 column (10/300 GL). The native molecular weight of the proteins was calculated from their Stokes radius measured by SEC, and their sedimentation coefficient was determined by sucrose gradient centrifugation as described previously (10, 19) and by analytical ultracentrifugation (AUC).

AUC.

Sedimentation velocity experiments were performed on an Optima XL-I analytical ultracentrifuge (Beckman Inc.) using an An 60 Ti rotor and double-sector epon centerpieces. The proteins were studied in 50 mM Tris⋅HCl buffer, pH 7.5, with 150 mM NaCl at 0.3 mg/mL. Buffer density and viscosity was measured using a DMA 5000 densitometer and an AMVn viscosimeter, respectively (both from Anton Paar). Protein concentration distribution was monitored at 280 nm at 54,000 × g and 20 °C. The time-derivative analysis was computed using the SEDFIT software package, version 12.1b (45), resulting in a continuous size distribution and an estimate for the molecular weight. This was estimated from the sedimentation coefficient and the diffusion coefficient, as inferred from the broadening of the sedimentation boundary, assuming that all observed species share the same frictional coefficient f/f0) (45).

FRET Studies.

We mixed CFP-730 tails and RFP-730 tails at a 1:1 molar ratio (2 μM each) in the presence or absence of 100 μM calmodulin and measured the FRET signal due to heterodimer formation under equilibrium conditions. To obtain a larger proportion of heterodimers, these experiments were also carried out in excess of RFP-730 tails (i.e., 2 μM CFP-730 tail mixed with 7 μM RFP-730 tail). The proteins were diluted in a buffer containing (in mM) 50 Tris, pH 7.4, 150 NaCl, 1 DTT, and 1 CaCl₂. A fluorescence spectrophotometer (Varian Cary Eclipse) was used with a wavelength of 445 nm for donor excitation and of 475 nm for donor emission and 605 nm for acceptor emission, respectively; for acceptor excitation, 585 nm was used. The excitation and emission slits were 10 and 5 nm, respectively; the detector gain was set to 630 V. Data were collected every 30 s over 60 min at 22 °C using a 2-s integration time. All experiments were carried out in duplicate and repeated using at least four different protein preps. The CFP-TEV-RFP control experiment was performed in TEV-cleavage buffer that contained (in mM) 50 Tris⋅HCl, pH 8.0, 0.5 EDTA, and 1 DTT. Ten units of TEV protease (Invitrogen) were added and spectra were taken every 30 min until the CFP emission peak at 475 nm reached a plateau. The energy transfer efficiency E was determined as described (21) from the quenching of fluorescence of the donor molecule by the relation E = [1 − (F_da/F_dn)] × 100, with F_da and F_dn the donor fluorescence intensities in the presence and absence of the energy acceptor, respectively. This experiment was repeated three times using a 1:1 ratio of RFP:CFP-730 tail.

Tryptophan Fluorescence.

Titrations of target peptide sequences with calmodulin-like proteins were performed at 20 °C in buffer (in mM) of 25 Tris (pH 8), 100 KCl, and 1 DTT with 1 mM CaCl₂ or 0.2 mM EDTA, using a Varian Cary Eclipse fluorescence spectrophotometer; λ_ex = 290 nm and λ_em = 323 nm as described (17, 18). The dissociation constants K_d for the Trp-containing peptides were determined by direct titration, and the data were analyzed as described (17).

Electron Microscopy.

Images were recorded on a JEOL JEM-1011 transmission electron microscope (Joachim Rädler, Ludwig-Maximilians-Universität, Munich). Myosin XXI was diluted to 100 nM in a buffer containing (in mM) 25 KCl, 50 Tris, pH 7.5, 0.1 MgCl2, and 0.1 DTT. The protein was applied to UV-treated, continuous carbon-coated copper grids (Science Services) and negatively stained with 1% uranyl acetate or 1% uranyl formate as described (9).

Motility Assays.

Procedures were adapted from those described by Kron et al. (1991) (46) and Lister et al. (2004) (31). In brief, myosin preps (∼150 μg⋅ml⁻¹) were immobilized on a nitrocellulose-coated surface (0.1% vol/vol nitrocellulose in amylacetate) of the experimental chamber. To prevent unspecific binding of actin filaments, the surface was then blocked with 0.5 mg⋅ml⁻¹ BSA in assay buffer [AB buffer (mM): 25 KCl, 4 MgCl₂, 1 EGTA, 25 imidazole, pH 7.4]. Subsequently, the rhodamine-phalloidin–labeled and stabilized rabbit skeletal actin filaments were introduced into the chamber in AB buffer supplemented with a scavenger system (10 mM DTT, 0.01 mg⋅ml⁻¹ catalase, 0.05 mg⋅ml⁻¹ glucose oxidase, 1.5 mg⋅ml⁻¹ glucose) and 2 mM ATP. For fluorescence imaging, the rhodamine label on the actin filaments was excited with a 532-nm laser (50 mW). Images were recorded every 0.4 s for a total period of 300 s. Only filaments moving continuously for at least 20 frames were included in the data analysis to determine the sliding velocity. For lipid-based motility assays, before adding the motor, Folch vesicles were flowed into the chamber and allowed to bind for 10 min. The velocities were calculated using the imaging motion analysis software GMimPro (Gregory Mashanov; www.nimr.mrc.ac.uk/gmimpro). All assays were carried out at 22 °C.