Molecular Characterization and Subcellular Localization of Arabidopsis Class VIII Myosin, ATM1*

Background: Molecular properties of class VIII myosin are not characterized. Results: Arabidopsis class VIII myosin, ATM1, has low enzymatic activity and high affinity for actin and is primarily localized at the cell cortex. Conclusion: Our data suggest that ATM1 functions as a tension sensor/generator. Significance: This is the first report of enzymatic and motile properties of class VIII myosin. Land plants possess myosin classes VIII and XI. Although some information is available on the molecular properties of class XI myosins, class VIII myosins are not characterized. Here, we report the first analysis of the enzymatic properties of class VIII myosin. The motor domain of Arabidopsis class VIII myosin, ATM1 (ATM1-MD), and the motor domain plus one IQ motif (ATM1-1IQ) were expressed in a baculovirus system and characterized. ATM1-MD and ATM1-1IQ had low actin-activated Mg2+-ATPase activity (Vmax = 4 s−1), although their affinities for actin were high (Kactin = 4 μm). The actin-sliding velocities of ATM1-MD and ATM1-1IQ were 0.02 and 0.089 μm/s, respectively, from which the value for full-length ATM1 is calculated to be ∼0.2 μm/s. The results of actin co-sedimentation assay showed that the duty ratio of ATM1 was ∼90%. ADP dissociation from the actin·ATM1 complex (acto-ATM1) was extremely slow, which accounts for the low actin-sliding velocity, low actin-activated ATPase activity, and high duty ratio. The rate of ADP dissociation from acto-ATM1 was markedly biphasic with fast and slow phase rates (5.1 and 0.41 s−1, respectively). Physiological concentrations of free Mg2+ modulated actin-sliding velocity and actin-activated ATPase activity by changing the rate of ADP dissociation from acto-ATM1. GFP-fused full-length ATM1 expressed in Arabidopsis was localized to plasmodesmata, plastids, newly formed cell walls, and actin filaments at the cell cortex. Our results suggest that ATM1 functions as a tension sensor/generator at the cell cortex and other structures in Arabidopsis.

Myosin is a motor protein that converts the chemical energy liberated by ATP hydrolysis into a directed movement on actin filaments. Phylogenetic analyses of myosin sequences reveal that there are at least 35 myosin classes (1), and their motor functions such as motility and ATP hydrolysis vary significantly. Only class VIII and XI myosins exist in higher plants. Arabidopsis thaliana possesses 4 and 13 genes encoding class VIII and XI myosins, respectively (2).
Studies on the intracellular functions of class XI myosins using immunolocalization, transfer DNA mutants, RNA interference, overexpression of dominant-negative myosins, and velocity-modified chimeric myosin XI indicated that they are responsible for organelle transport, organization of actin cable, and cell and plant growth (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13). Molecular properties of several class XI myosins have also been characterized. The actinsliding velocities and the actin-activated ATPase activities of class XI myosins are higher than those of other myosin classes (14 -19). The dissociation rates of ADP from acto-class XI myosins are extremely fast and account for their high actin-sliding velocities (15,19,20).
In contrast to class XI myosins, considerably less is known about molecular properties and intracellular functions of class VIII myosins, although class VIII myosin, ATM1, was the first plant myosin to be identified and sequenced (21). The Arabidopsis genome encodes the following four class VIII myosins: ATM1, ATM2, VIIIA, and VIIIB (2). Immunolocalization experiments showed that ATM1 is localized to the plasmodesmata and new cell plates in the root of Arabidopsis (22). Analysis of the expression of GFP-fused tail domain of ATM1 suggests that endogenous ATM1 is localized to the plasmodesmata, endoplasmic reticulum, and plasma membranes (23). Although there is some evidence indicating the subcellular localization of class VIII myosins, definitive information regarding tissue-specific expression and subcellular localization using full-length myosin VIII expressed under the control of its native promoter have not been reported. In addition, molecular properties such as ATPase activity and actin-sliding velocity of class VIII myosin have not been characterized.
In this study, we expressed Arabidopsis class VIII myosin, ATM1, in a baculovirus system and uncovered its molecular properties. Furthermore, its subcellular localization was determined by expressing GFP-fused full-length ATM1 in Arabidopsis under the control of its native promoter.
A baculovirus transfer vector for Arabidopsis CAM3 (pFastBac Arabidopsis calmodulin 3) was generated as follows. An Arabidopsis cDNA encoding CAM3 was mutated to create an XbaI site at the 5Ј end and XhoI site at the 3Ј end (after the translation termination codon) using site-directed mutagenesis PCR. PCR products were digested with XbaI and XhoI and ligated to the XbaI and XhoI fragment of pFastBac 1 (Invitrogen). This construct (pFastBac Arabidopsis calmodulin 3) encodes full-length Arabidopsis CAM3. This was expressed in insect cells and purified as described previously (19).
For the promoter-GUS (␤-glucuronidase) assay, 3 kb of 5Ј-flanking sequences containing the first exon of ATM1 was PCR-amplified and subcloned into pENTR-D-TOPO (Invitrogen) and subsequently exchanged into the binary vector pGWB533 (40) using LR Clonase (Invitrogen) according to the manufacturer's instructions.
For transient expression in protoplasts, an ATM1 cDNA (AT3G19960) was amplified from total RNA purified from Arabidopsis seedlings (7-day-old) and subcloned into pENTR/D-TOPO (Invitrogen). The ATM1 sequence in the pENTR-D-TOPO cloning vector was exchanged into the binary vector pUGW0-sGFP using the LR reaction. For expression in Arabidopsis, full-length genomic ATM1 DNA, including the cDNA encoding sGFP upstream of the translational start codon, was generated by fluorescent tagging (41). The ATM1 fragment contains 3 kb of 5Ј-and 1 kb of 3Ј-flanking sequences. Amplified chimeric fragments were subcloned into the binary vector pGWB501 (40).
ATPase Activities and in Vitro Actin Gliding Assays-Steadystate ATPase activities were measured using a modified malachite green method (42). To avoid the possible inhibitory effect of ADP on binding of ATP to the active site, ATPase activities were measured when the molar ratio of ADP/ATP was less than 0.01.
Actin-sliding velocities were measured using an anti-Myc antibody-based version of the in vitro actin-gliding assay as described (19). To avoid the possible inhibitory effect of ADP on actin-sliding velocities, an ATP regeneration system (0.4 mM phosphocreatine and 25 units/ml creatine phosphokinase) was included. To compare with other myosins (17,19,28,29,33,34,36,(43)(44)(45)(46)(47)(48), the reactions of steady-state ATPase activities, in vitro actin-gliding assay, and kinetic measurements were conducted in the commonly used ionic strength (ϳ50 mM) and temperature (25°C) unless stated otherwise.  (49). ATPase activities and in vitro actin-gliding assays were done at 25°C. For ATPase activities and in vitro actin-gliding assays of ATM1-1IQ, 1 M Arabidopsis calmodulin was added in the assay buffer.
Co-sedimentation Assays-Co-sedimentation assays were performed as described (29), except that the solution was 170 mM in ionic strength (90 mM KCl, 4 mM MgCl 2 , 25 mM Hepes-KOH (pH 7.4), 2 mM ATP, 1 mM EGTA, 5 mM DTT, 10 mM phosphocreatine, and 50 units/ml creatine phosphokinase). In brief, 1 M ATM1-MD was mixed with 0 -96 M actin and centrifuged at 280,000 ϫ g for 20 min at 25°C in the presence of ATP and the ATP generation system. Less than 12 M actin, phalloidin-actin was used. The supernatant of each actin concentration was subjected to SDS-PAGE, and the fraction of ATM1-MD bound to actin was determined by quantifying the amount of ATM1-MD in the supernatant using ImageJ software (National Institutes of Health). About 10% of ATM1 was precipitated in the absence of actin (0 M actin). Thus, it is necessary to remove this artificial effect presumably caused by the aggregations of ATM1 to calculate the actual amount of acto-ATM1 precipitations. So, the amount of the pellet and the supernatant at 0 M actin was set as zero and total amount of ATM1-MD, respectively. Therefore, the fraction unbound at each concentration of actin was calculated as follows: 100% ϫ (the amount of supernatant at each concentration of actin)/(the amount of supernatant at 0 M actin). The fraction bound was plotted as a function of actin concentration to determine the affinity of ATM1-MD for actin in the presence of ATP.
Transient Kinetic Measurements-All kinetic experiments were performed using an Applied Photophysics SX18MV stoppedflow spectrophotometer (dead time, 1.15 ms) as described (19). The solution used for the transient kinetic experiments was the same as that used for the ATPase activities and the in vitro actingliding assays unless stated otherwise.
Transient Expression in Protoplasts of Arabidopsis-The transient expression of GFP-ATM1 in protoplasts prepared from suspension cultures of Arabidopsis cells was conducted as follows: 2 g of cultured cells was incubated in 25 ml of enzyme solution (0.4 M mannitol, 5 mM EGTA, 1% cellulase R-10, and 0.05% pectolyase Y-23) for 1-2 h at 30°C and filtered through a nylon mesh Plant Transformation-Plasmids were electroporated into Agrobacterium tumefaciens strain GV3101::pMP90 using a Gene Pulser (Bio-Rad). They were introduced into an Arabidopsis atm-1 knock-out line (SAIL_405_B08; AT3G19960) using the floral dipping method. T1 plants with resistance to hygromycin were selected. T2 generation plants were used for the promoter-GUS assay (50), and homozygous T3 plants were used for imaging full-length ATM1.
Confocal Laser Scanning Microscopy-GFP fluorescence in cells was detected using a spinning disk confocal laser scanning microscope (CSU10, Yokogawa, Kanazawa, Japan) equipped with a high resolution CCD camera (ORCA-AG, Hamamatsu Photonics, Hamamatsu, Japan). The images were processed using iVision Macintosh software (BioVision Technologies, Exton, PA).

RESULTS
Constructs-A schematic diagram of native ATM1 deduced from its amino acid sequence is shown in Fig. 1A, Native ATM1. Native ATM1 includes motor and neck domains with four IQ motifs and coiled-coil and globular tail domains. It is likely that calmodulin or calmodulin-like light chains bind to each IQ motif and that native ATM1 forms a dimer through the interaction of its coiled-coil domain. We generated two recombinant ATM1 constructs called ATM1-MD and ATM1-1IQ (Fig.  1A, Recombinant ATM1). ATM1-MD contains a single motor domain, and ATM1-1IQ includes a motor domain and the first IQ motif. The expected lever arm lengths of ATM-1MD, ATM1-1IQ, and native ATM1 are 3, 7, and 19 nm, respectively. ATM1-MD and ATM1-1IQ were expressed using a baculovirus system. ATM1-1IQ was expressed along with Arabidopsis calmodulin. ATM-1MD and ATM1-1IQ were purified by coprecipitation with actin and nickel-affinity resin. For ATM1-1IQ purification, 1 M Arabidopsis calmodulin was added to the purification buffer throughout the purification. Purified ATM1-1IQ contained calmodulin with stoichiometry 1:1 (Fig.  1B, with CaM). When external calmodulin was deleted before washing the Ni-NTA column, the stoichiometry decreased below 1:1 (Fig. 1B, wo CaM). These results suggest that calmodulin was weakly bound to the first IQ motif as a light chain, which was in equilibrium with the external calmodulin. It is also possible that the IQ motif might bind other light chains, not just calmodulin.
Actin-sliding Velocity-The actin-sliding velocities of ATM1-MD and ATM1-1IQ measured using an in vitro actin-gliding assay were 0.02 Ϯ 0.003 and 0.089 Ϯ 0.013 m/s, respectively, at 25°C (Fig. 3, 1.8 mM free Mg 2ϩ ). The difference may be explained by the differences in their lever arm length because the velocities of myosins are proportional to their lever arm length, if their motor domains are the same (18,19,44). Based on these results and lever arm length of native ATM1, we estimated that the velocity of native ATM1 is ϳ0.2 m/s at 25°C. This value is only 1% of those of plant-specific class XI myosins (15,17,18) and is similar to those of unconventional animal slow myosins such as classes V(30), VI (51), and X (52). These results suggest that ATM1 functions as a slow transporter and/or a tension generator.
The relationship between velocities and free [Mg 2ϩ ] was the same for ATM1-MD and ATM1-IQ. When free [Mg 2ϩ ] increased, actin-sliding velocity decreased. Actin-sliding velocities in the presence of 0.46 and 0.17 mM free [Mg 2ϩ ] were higher by factors of 2 and 3, respectively, than in the presence of 1.8 mM (Fig. 3). Actin-activated ATPase activities also decreased when free [Mg 2ϩ ] increased. V max of actin-activated ATPase activities of ATM1-MD at 0.46 and 0.17 mM [Mg 2ϩ ] were higher by factors of 1.7 and 2.1, respectively, than those in the presence of 1.8 mM [Mg 2ϩ ] (Fig. 2).
ATP-induced Dissociation of Acto-ATM1-To unravel the molecular mechanism underlying slow actin-sliding velocity and the dependence of actin-activated ATPase activity and actin-sliding velocity on free [Mg 2ϩ ], we investigated the kinetic properties of ATM1. We used ATM1-MD for kinetic analyses because its steady-state ATPase activity and modulation of actin-sliding velocity by free [Mg 2ϩ ] were the same as those of ATM1-1IQ (Table 1 and Fig. 3), although it is possible that there could still be differences between ATM1-MD and ATM1-1IQ. Kinetic modeling and simulations of actomyosin were performed according to the simplified reaction mechanism as shown in Scheme 1, which has been used in kinetic studies of many myosins (28,29,34,35,56,(73)(74)(75).
We first investigated the molecular basis underlying slow actin-sliding velocity. In general, the actin-sliding velocities of myosins depend on the time when they are strongly bound to actin (t s ). t s is primarily determined by the ADP dissociation rate from acto-myosin and the acto-myosin dissociation rate upon ATP binding at saturating ATP concentrations (76,77). Therefore, we determined these rates.   The basal ATPase rate (V 0 ) was measured in the absence of actin. The actin-activated ATPase results were fit to the Michaelis-Menten equation to determine the maximum ATPase rate (V max ) and actin concentration at which the ATPase rate was half-maximal (K actin ).  The acto-myosin dissociation reaction upon ATP binding can be described as a two-step process (Scheme 2), and the rate constant is expressed as The first step in Scheme 2 (where A is actin, and M is myosin) is the formation of a collision complex between acto-myosin and ATP, which are in rapid equilibrium. The second step is the dissociation of myosin-ATP from actin following the isomerization of myosin. To determine KЈ 1 and kЈ ϩ2 , a pyrene⅐actin⅐ATM1-MD complex was dissociated by mixing with 10 -3000 M ATP using a stopped-flow apparatus, and the acto-ATM1-MD dissociation was monitored by the increase in pyrene fluorescence (Fig. 4) (78). The values of KЈ 1 and kЈ ϩ2 were obtained by fitting the data to . The values of 1/KЈ 1 and kЈ ϩ2 were 450 Ϯ 70 M and 620 Ϯ 40 s Ϫ1 , respectively ( Fig. 4 and Table 4). The acto-ATM1-MD dissociation rate in the presence of a physiological concentration of ATP (2 mM) was ϳ500 s Ϫ1 , which is extremely fast to explain the slow actinsliding velocity of ATM1 (see "Discussion").
ADP Dissociation from Acto-ATM1-The rate of ADP dissociation from acto-ATM1-MD was determined by measuring the decrease in the fluorescence intensity of mant 3 -ADP (28,32,33,75). Time course of fluorescence change followed a dou-  Tables 2 and 4). These results show that there are two acto-myosin-ADP states in ATM1. In animal, slow myosins such as myosin III, V, and VII, similar two actomyosin-ADP states were reported (29,35,54,55,79). In these studies, it is proposed that two states in the active site, namely open and closed pockets, are in equilibrium. According to these studies, we interpret our results as follows: kЈ ϩ5, fast and kЈ ϩ5, slow exhibit ADP dissociation rates from the open and closed pockets, respectively. kЈ ϩ5, slow was much slower than the actin-activated V max value (4.1 s Ϫ1 , Table 1). Because no transition on the main pathway can be slower than the transition that limits the overall ATPase cycle, it is more likely that the slow transition is indicative of a branched or a parallel pathway (Scheme 3). The kЈ ϩ5, fast value is similar to kЈ ϩ5 for animal slow myosins (28,33,34) and most probably explains the slow actin-sliding velocity of ATM1 (see under "Discussion"). The value of kЈ ϩ5, fast is nearly equal to the V max value of the actin-activated ATPase (4.1 s Ϫ1 ; Fig. 2 ADP Binding to Acto-ATM1-The rate of mant-ADP binding to acto-ATM1-MD was biphasic with observed fast and slow rate constants (Fig. 5B), consistent with the two acto-myosin⅐ADP states described above. The observed fast rate constants depended linearly on mant-ADP concentrations. The association rate constant of ADP with acto-ATM1-MD of the fast phase (kЈ Ϫ5, fast ) was 1.6 Ϯ 0.11 M Ϫ1 s Ϫ1 (Fig. 5C and Table 4). The observed slow rate constants were almost independent of mant-ADP concentrations in the concentration range measured (5-25 M) and were 3.0 Ϯ 0.5 s Ϫ1 (Fig. 5C, kЈ Ϫ5, slow ). The relative amplitudes of each fast and slow phase were similar at every mant-ADP concentration, and the average values were 49 Ϯ 5 and 51 Ϯ 5% for fast and slow phases, respectively.
ADP Dissociation from ATM1-Kinetic parameters in the absence of actin were also measured. The rate of ADP dissociation from ATM1-MD was determined by measuring the decrease in fluorescence intensity of mant-ADP and followed by a double exponential with fast (kЈ ϩ5, fast , 2.4 Ϯ 0.1 s Ϫ1 ) and slow (kЈ ϩ5, slow , 0.16 Ϯ 0.03 s Ϫ1 ) rate constants in the presence of 1.8 mM free [Mg 2ϩ ] (Fig. 6A and Tables 3 and  4). The relative amplitudes of fast and slow phases were 65 Ϯ 4 and 36 Ϯ 4%, respectively. Similar to the ADP dissociation rate from acto-ATM1-MD, the dissociation rate of ADP from ATM1-MD was also dependent on free [Mg 2ϩ ] concentrations (Table 3).
ADP Binding to ATM1-The rate of ADP binding to ATM1-MD was monitored using mant-ADP. The time courses of reactions were biphasic with observed fast and slow rate constants (Fig. 6B). The observed fast rate constants linearly depended on mant-ADP concentrations. The association rate constant of ADP with ATM1-MD of the fast phase (kЈ Ϫ5, fast ) was 2.3 Ϯ 0.11 M Ϫ1 s Ϫ1 (Fig. 6C and Table 4). The observed slow rate constants (kЈ Ϫ5, slow ) were almost independent of mant-ADP concentrations in the concentration range measured (10 -50 M) and were 5.3 Ϯ 1.4 s Ϫ1 (Fig. 6C and Table 4). The relative amplitudes of each fast and slow phase were almost SCHEME 3 Mean Ϯ standard deviation of 3-4 independent measurements from two independent protein preparations. a Relative amplitude of k ϩ5´fast and k ϩ5´slow , respectively. b Average of values from mant-ADP and pyrene-actin.
the same at every mant-ADP concentration, and the average values were 59 Ϯ 5 and 41 Ϯ 5% for fast and slow phases, respectively.
The equilibrium dissociation constant for ADP binding to ATM1 of the slow phase (K 5 slow ) is calculated as k ϩ5 slow / k Ϫ5 slow and is 0.03 (0.16/5.3 s Ϫ1 ). The equilibrium dissociation constant for ADP binding to ATM1 of the fast phase (K 5 fast ) is calculated as k ϩ5 fast /k Ϫ5 fast and is 1.0 M (2.4/2.3 M Ϫ1 s Ϫ1 ). The overall equilibrium dissociation constant for ADP binding to ATM1 (K 5 overall ) is calculated as K 5 fast (K 5 slow /(1 ϩ K 5 slow )) (35,55) and is 0.029 M (Table 4).
Actin-ATM1 Interaction in the Presence of ATP-To determine the affinity of ATM1 for actin in the presence of ATP, actin co-sedimentation assays (29) were performed under physiological ionic conditions. Fig. 7 shows the fraction of ATM1-MD bound to actin as a function of actin concentration. The data were fit to a hyperbola to determine the actin affinity in the presence of ATP. The equilibrium dissociation constant for actin binding to ATM1 in the presence of ATP (K 8 ) was 4.5 Ϯ 0.7 M (Table 4). This value is similar to that of animal myosin III (29), which is known to have high affinity for actin. The maximum percentage bound to actin was 90 Ϯ 4.5%, showing that ATM1 is a high duty ratio (ϳ90%) myosin.
Actin-ATM1 Interaction in the Absence of ATP-The affinity of ATM1 for actin in the absence of ATP (strong binding state) was measured using pyrene-actin (29,75). The transients were fit to a single exponential function at each actin concentration (Fig. 8A, inset). The association rate constant linearly increased with actin concentration to yield the second-order rate con-    Fig. 4. b Data are from Fig. 5 and Table 2. c Data are from Table 2. d Data are from Fig. 5. e Data are from Fig. 6 and Table 3. f Data are from Fig. 6. g Data are from Fig. 8. h Data are from Fig. 7. (Fig. 8A and Table 4). The intercept of Fig. 8A yields the acto-ATM1-MD dissociation rate constants (k ϩ6 ) but is subject to a large uncertainty when it is low (34,35), so dissociation was measured directly by competition with unlabeled actin filaments (Fig. 8B). The observed rates were fit to a single exponential function, which yielded a rate constant of 0.0057 Ϯ 0.001 s Ϫ1 of k ϩ6 (Fig. 8B and Table 4). The equilibrium dissociation constant for actin binding to ATM1 in the absence of ATP (K 6 ) is calculated as k ϩ6 /k Ϫ6 and is 0.66 nM (0.0057 s Ϫ1 /8.6 M Ϫ1 s Ϫ1 , Table 4). This value is lower by a factor of 50 than that of animal skeletal myosin II (80) and by a factor of 10 than that of animal myosins III and VII that are known to have high affinity for actin (29,35,75,81). This indicates that ATM1 binds actin with very high affinity in the strong binding state. Expression Pattern of ATM1 in Arabidopsis-To determine the tissue-specific expression of ATM1, we created a transgenic Arabidopsis plant that expressed a GUS fusion protein under the transcriptional control of the ATM1 promoter. GUS staining indicated that ATM1 was abundantly expressed in seedlings, apices of shoots, shoots, and flowers (Fig. 9, A-D) but not in pollen (Fig. 9E).

Enzymatic Properties of Class VIII Myosin
Localization of ATM1 in Arabidopsis-In earlier studies using an antibody against a peptide corresponding to a tail domain of ATM1, it was suggested that ATM1 is localized to the plasmodesmata and newly formed cell walls in root cells of maize and Arabidopsis (22). The same antibody was used to show the localization of ATM1 on plastid (82). Moreover, when the GFP-fused tail domain of ATM1 was expressed in Arabidopsis, it was localized to the plasma membrane and plasmodesmata (23,83). To obtain accurate information about the localization of ATM1, we used GFP-fused full-length ATM1 (GFP-ATM1) in this study because tail domain expression sometimes acts as a dominant negative, leading to artifactual localization (84). First, GFP-ATM1 was transiently expressed in protoplasts of cultured Arabidopsis cells and was observed using a high speed confocal laser scanning microscope (Fig. 10, A-C). Fluorescent dots were observed at the cell cortex (Fig. 10,  A and B), similar to the results of previous studies using the GFP-fused tail domain of ATM1 (23). However, in contrast to other studies, punctate structures were present on filamentous structures such as actin bundles (Fig. 10B). To ascertain that this structure was formed by F-actin, Lifeact-TagRFP was expressed together with GFP-ATM1. ATM1 was co-localized with Lifeact-TagRFP at the cell cortex (Fig. 10C), indicating that ATM1 associated with F-actin at the cell cortex. Live imaging showed little movement of the dots on F-actin (supplemental Movie S1). This is consistent with the results of the in vitro actin-gliding assays, which showed that ATM1 is an extremely slow myosin (Fig. 3).
Next, GFP-ATM1 was expressed in Arabidopsis plants (Fig.  10, D-G). To avoid artifacts arising from differences in expression level and pattern, a construct expressing GFP-ATM1 under the control of the native ATM1 promoter was transformed into an Arabidopsis ATM1-knock-out mutant (atm1). Fig. 10D shows the expression of GFP-ATM1 in epidermal cells of root tip meristems. GFP-ATM1 was localized to newly formed cell plates, consistent with another study (22). In a growing elongating zone of epidermal cells, GFP-ATM1 was observed as fluorescent dots on filamentous structures (Fig.  10E) similar to protoplasts (Fig. 10B). To determine whether these filamentous structures were composed of F-actin, we treated cells with cytoskeletal inhibitors. The filamentous structures did not form in the presence of latrunculin B (Fig.  11B) but persisted when the cells were exposed to oryzalin (Fig.  11A). These data confirm that the filamentous structures con- is shown in Fig. 10, F and G. Focusing on the center of the cell, ATM1 was localized to the punctate structures on plasma membranes that border neighboring cells (Fig. 10F, arrowheads), supporting an earlier observation that an anti-ATM1 antibody stains the plasmodesmata (85). ATM1 was also localized to structures presumed to be plastids (Fig. 10F, arrows), consistent with the results of a published study (82). Focusing on the subcortical region of the cell (Fig. 10G), ATM1 was observed at filamentous structures, most likely F-actin.

DISCUSSION
ATM1 Exhibits Low Enzymatic Activity and High Affinity for Actin-In this study, we reported the first analysis of the enzymatic properties of plant-specific class VIII myosin. We showed that actin-sliding velocity (Fig. 3) and actin-activated ATPase activity of Arabidopsis class VIII myosin, ATM1 (Fig.  2), were lower by a factor of 10 -100 compared with those of class plant-specific XI myosins (15,17,18) and were similar to those of animal slow myosins (30,51,52). The affinity of ATM1 for actin was very high both in the presence (Fig. 7) and absence (Fig. 8) of ATP. However, it must be noted that the kinetics might be different due to phosphorylation or other post-translational modifications to the myosin.
Slow ADP Dissociation Accounts for Low Velocity, Low Enzymatic Activity, and High Duty Ratio-The ADP dissociation rate from acto-ATM1 (kЈ ϩ5, fast Fig. 5 and Table 4) was almost the same as the V max of the actin-activated Mg 2ϩ -ATPase activity (4.1 s Ϫ1 ; Fig. 2). This indicates that the acto-myosin ATPase cycle primarily exists in the AM⅐ADP state and that ADP release from acto-ATM1 is the rate-limiting step of actin-activated ATPase. This was also supported by actin co-sedimentation experiments, which showed that duty ratio of ATM1 was ϳ90% (Fig. 7).
The dissociation rates of ADP from myosins are accelerated by actin binding, and the acceleration (kЈ ϩ5 /k ϩ5 ) differs between myosins. The value of the acceleration for fast class XI myosin is ϳ1000-fold (19) and that for slow animal myosins is only 1-10-fold (28,33). The value for ATM1 was 2.3-fold (Table 4) showing that, similar to slow animal myosins, dissociation of ADP was not markedly influenced by actin binding.
ATM1 Has the Kinetic Properties of a Tension Sensor-We have shown in this study that ATM1 has two ADP states, open ADP pocket and closed ADP pocket. These two ADP states have been observed only in slow myosins but not fast myosins (29,35,54,55,79). Nyitrai and Geeves (86) suggested that the closed and open ADP states are in equilibrium in all myosins and that the equilibrium was dependent on the myosin types; almost all the ADP state are in the open state for fast myosins, whereas a considerable portion of ADP states are in the closed state for slow myosins. Thus two ADP states (the closed ADP pocket) can be observed only in slow myosins.
The swinging lever arm model proposes that the large free energy associated with P i release drives the power stroke of the cross-bridge (a swing of the converter and neck of the crossbridge) to generate force. In some slow myosins, this swing is followed by a further swing of the converter and neck of the cross-bridge in the same direction as the power stroke when ADP is released (87)(88)(89). This additional swing concomitant with ADP release has been observed only in slow myosins but not fast myosins. Geeves et al. (26) suggested that this additional swing of the neck must be complete before the ADP binding pocket opens to allow ADP to escape. Such a mechanism provides a simple, elegant way of coupling ADP release to the load on the head because the load inhibits the swing and then ADP release. Thus, slow myosins in which a considerable por- tion of ADP states are in the closed state have tension-bearing and tension-sensing mechanisms (86) Because ATM1 falls into this category, it presumably senses tension (strain) and func-tions in Arabidopsis cells. It is possible that the kinetic rates could be gated by strain in a full molecule but that future experiments are required to address this possibility. It has been reported that ATP binding to acto-myosin follows two phases (86). So we also tried to fit the transients of ATP binding to acto-ATM1-MD with a double exponential curve. Some could be fitted, but most could not be well fitted. The uncertainty may arise from the relatively noisy signals of ATP binding to acto-ATM1-MD Although we could not observe this, it is possible that two phases exist in transients of ATP binding to acto-ATM1.
Because myosin VIII has a coiled-coil domain, it may form a dimer like myosin V. However, the number of IQ motifs of ATM1 is four, although that of myosin V and myosin XI, which are known for the processive myosins, is six. The number of IQ motifs of myosin V defines the step size (90). The processivity of myosin V is also dependent on the number of IQ motifs, and six IQ motifs are most effective for the processivity (91). Thus ATM1 may be inadequate for the processive movement. Kinetic and morphological features of ATM1 suggest a tensionsensor role over a transport protein.
Free [Mg 2ϩ ] Regulates Actin Velocity and Enzymatic Activity-Mg 2ϩ is the most abundant divalent cation in animal and plant cells and is an essential cofactor required for the metabolism of ATP and nucleic acids as well as for the catalytic activities of several cellular enzymes. [Mg 2ϩ ] i is in high millimolar range, and cytosolic Mg 2ϩ mainly exists in complexes with ATP and other molecules. Therefore, the cytoplasmic free [Mg 2ϩ ] is only a small percentage of the total (92). Many studies have been conducted on measuring free [Mg 2ϩ ] i in animal cells and have shown as follows: free [Mg 2ϩ ] i in animal cells is cell type-specific and in the range of 0.2-2 mM (57-68); Mg 2ϩ is not uniformly distributed within the cells and is enriched in the vicinity of negatively charged phospholipids residing in cellular and intracellular membranes, and this enriched Mg 2ϩ is related to free Mg 2ϩ (57); and free [Mg 2ϩ ] i changes over time (63,(67)(68)(69)(70). There is growing evidence that free [Mg 2ϩ ] i provides important functions in animal cells. Recently, it has been shown that free [Mg 2ϩ ] i acts as a second messenger in human T-cells (92).
Although studies on the role of free [Mg 2ϩ ] i in plant cells are lacking, it was reported that the range of free [Mg 2ϩ ] i is similar to that in animal cells (0.2-2 mM) (71,72). It is probable that free [Mg 2ϩ ] i play an important function in plant cells as well as animal cells. We demonstrate that free [Mg 2ϩ ] regulates the velocity and ATPase activity of ATM1 within the range of physiological concentrations. Actin velocity and ATPase activity determined using 0.17 mM free [Mg 2ϩ ] were higher by factors of 2 and 3, respectively, compared with those determined using 1.8 mM [Mg 2ϩ ] (Figs. 2 and 3). Furthermore, we demonstrate that these effects were exerted through free [Mg 2ϩ ]-dependent modulation of ADP dissociation from acto-ATM1 (Table 2). It could be that the temporal and spatial changes in free [Mg 2ϩ ] regulate the intracellular functions mediated by ATM1.
Functions of ATM1 in Arabidopsis-Expression of GUS under the control of the ATM1 promoter in Arabidopsis suggests that ATM1 is abundantly expressed in seedlings, apices of shoots, shoots, and flowers but not in pollen (Fig. 9). This result suggests that ATM1 functions in all tissues during all stages of the plant's life cycle. Earlier studies using an antibody raised against a peptide corresponding to a tail domain of ATM1 suggested that ATM1 is localized at plasmodesmata, plasma membrane of newly formed cell walls, and plastid (22,82). This was also supported by live cell imaging using the GFP-fused tail domain of ATM1 (23,83). In this study, we further examined the localization of ATM1 in Arabidopsis using GFP-fused full-length ATM1 (GFP-ATM1). Furthermore, to avoid artifacts arising from differences of expression levels and pattern, GFP-ATM1 was expressed using the native promoter of Arabidopsis ATM1 in an ATM1-knock-out mutant (atm1). Consistent with the studies cited above, GFP-ATM1 was localized to the plasmodesmata (Fig. 10F), plasma membrane of newly formed cell walls (Fig. 10D), and plastids (Fig. 10F). However, in contrast to other studies, we observed that ATM1 was primarily localized to F-actin at the cell cortex in several tissues (Figs. 10, F and G, and  11). This was also confirmed by analysis of protoplasts prepared from suspension cultures of Arabidopsis cells (Fig. 10, A-C). ATM1 showed little movement on F-actin at the cell cortex (supplemental Movies S1 and S2), consistent with the results of the in vitro actin gliding assays.
Taken together, our enzymatic and localization studies suggest that in Arabidopsis class VIII myosin, ATM1 functions at various intracellular structures such as cell cortex, newly formed cell wall, plasmodesmata, and plastids as a tension sensor/generator.