Real-time Visualization of Processive Myosin 5a-mediated Vesicle Movement in Living Astrocytes* 210

Recycling endosomes in astrocytes show hormone-regulated, actin fiber-dependent delivery to the endosomal sorting pool. Recycling vesicle trafficking was followed in real time using a fusion protein composed of green florescent protein coupled to the 29-kDa subunit of the short-lived, membrane-bound enzyme type 2 deiodinase. Primary endosomes budded from the plasma membrane and oscillated near the cell periphery for 1–4 min. The addition of thyroid hormone triggered the processive, centripetal movement of the recycling vesicle in linear bursts at velocities of up to 200 nm/s. Vesicle migration was hormone-specific and blocked by inhibitors of actin polymerization and myosin ATPase. Domain mapping confirmed that the hormone-dependent vesicle-binding domain was located at the C terminus of the motor. In addition, the interruption of normal dimerization of native myosin 5a monomers inactivated vesicle transport, indicating that single-headed myosin 5a motors do not transport cargo in situ. This is the first demonstration of processive hormone-dependent myosin 5a movement in living cells.

The trafficking of intracellular organelles utilizes both microtubules composed of polymerized tubulin and microfilaments composed of polymerized actin (1). Dyneins and kinesins provide bidirectional movement of organelles along individual microtubules, and members of the myosin superfamily use microfilaments for this function. Recent work done in melanocytes revealed that individual organelles can possess both microtubule-and microfilament-based motors and that the two cargo carrier systems cooperate in organelle trafficking (2,3). Microtubule-based vesicle carriers are responsible for long range vesicle movement in these cells, whereas myosin 5amediated vesicle movement is restricted to short-range shuttling along the cortical actin cytoskeleton in the dendritic processes (4).
Four of the 18 identified classes of myosin motor proteins appear to transport vesicles along microfilaments in vitro (5), and loss of the two-headed myosin 5a motor disrupts vesicle trafficking in rodents and yeast (1, 2, 6 -13). In melanocytes from the myosin 5a null mouse, dilute, pigment granules remain centrally distributed and fail to accumulate in dendritic arbors (4,14). The short distances traveled by cargo-laden myosin 5a in melanocytes limit the direct analysis of myosin 5a-mediated movement in living cells. In vitro, myosin 5a is a processive cargo-carrying motor and undergoes multiple catalytic cycles that allow the motor to move long distances along actin fiber tracks by taking large steps of ϳ36 nm (15)(16)(17). These estimates of the stepping size of myosin 5a closely agree with the distance between the two motor heads (18). The globular tail of the motor appears to bind tightly to the surface of melanosomes (4), synaptic vesicles (19,20), and recycling vesicles in astrocytes (21). The tail of myosin 5a also shows hormone-regulated vesicle binding (21) and Ca 2ϩ -modulated interactions with two synaptic vesicle proteins, synaptophysin and synaptobrevin II, in isolated synaptosomes (22). However, the short distances traveled and the presence of cooperating microtubule-based cargo carriers have confounded the direct analysis of the processive movement and vesicle-docking reactions in living cells.
Astrocytes provide a unique model for the study of myosin 5a-mediated vesicle trafficking, because the dynamic hormonedependent regulation of the short-lived membrane-bound enzyme type II deiodinase (D2) 1 is an actin-based endocytotic event (21,(23)(24)(25) that results in the delivery of recycling vesicles from the cell periphery to the endosomal storage pool (26). In vitro, the C-terminal 22 residues of myosin 5a serve as the hormone-dependent vesicle-binding domain (21). This long range actin-based trafficking of recycling vesicles can be exploited to define the properties of myosin 5a-mediated vesicle movement in living cells without the participation of the microtubule cargo carriers.
In this report we use rapid acquisition time lapse microscopy to follow the processive, centripetal movement of individual recycling vesicles in living cells and show that a specific hormone(s) triggers the docking of primary endosomes to the C terminus of the actin-bound motor. Once tethered, the cargoladen myosin 5a shows processive, centripetal movement from the cell periphery to the endosomal sorting pool with a velocity of ϳ100 nm/s. Actin depolymerization, myosin ATPase inhibitors, and the expression of motorless, dominant negative myosin 5a truncation mutants arrested the processive hormone-dependent movement of recycling vesicles. These findings provide the first evidence of the processive long range movement of cargo-laden myosin 5a along actin fibers in living cells and identify a specific thyroid hormone-dependent domain responsible for reversible vesicle docking. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. hormones, Triton X-100, ATP, dibutyryl cAMP,  hydrocortisone, colchicine, bovine serum albumin, 2,3-butanedione  monoxime (BDM), and rabbit anti-actin IgG were obtained from Sigma.  Dulbecco's modified Eagle's medium, antibiotics, Hanks' solution, and  trypsin were purchased from Life Technologies, Inc. Restriction endonucleases and DNA-modifying enzymes were purchased from New England Biolabs (Beverly, MA). Anti-myosin 5a was generated as detailed previously (21). The anti-sera used were: anti-SV2 antisera (Dr. Kathleen Buckley, Harvard); anti-Synaptotagmin I and anti-Rab 3 A, B, and C antisera (Dr. Reinhard Jahn, Yale); and anti-GFP IgG (CLONTECH). Anti-D2p29 (27) and anti-p55 (44) (the subunit of protein disulfide isomerase) antibodies were raised in-house.

Materials-Thyroid
Culture Conditions-Astrocytes were prepared from 1-day-old neonatal rats as described previously (45) and grown in growth medium composed of Dulbecco's modified Eagle's medium supplemented with 10% bovine calf serum, 50 units/ml penicillin, and 90 units/ml streptomycin. The cells were grown to confluence in 75-cm 2 culture flasks in a humidified atmosphere of 5% CO 2 and 95% air at 37°C and used at passages 1-3.
Dominant-negative Myosin 5a Mutants-The myosin 5a dimerization mutant, ⌬myo5a coiled-coil encoding the last IQ domain and the coiled-coil region (residues 892-1040, myosin 5a) was generated by reverse transcription-polymerase chain reaction using rat brain mRNA and sitespecific 20-mer oligonucleotides (upstream 5Ј-CAGTGCTGCTTCCG-GCGGAT-3Ј; downstream, 5Ј-GTTGAGGGTCTCCTTTTCCTA-3Ј). The ϳ500-base pair fragment was cloned into the EcoRV site of the prokaryotic expression vector, pThioHis B (Invitrogen, San Diego, CA) and the ⌬myo5a coiled-coil cDNA was then subcloned into the NotI site of the multiple cloning site of the Sinbis viral expression vector, pSinHis A (Invitrogen). The plasmid, containing a promoter for in vitro transcription as well as the ⌬myo5a coiled-coil cDNA, was linearized with PvuI, the mRNA was transcribed using InvitroScript CAP SP6 in vitro transcription kit (Invitrogen), and pseudovirions were produced according to manufacturer instructions in baby hamster kidney cells. Spent medium containing the pseudovirion packaged His 6 -tagged ⌬myo5a coiled-coil mRNA was collected 2-3 days post infection.
The C terminus of myosin 5a (nucleotides 2911-7087, a gift from Dr. Nancy Jenkins) was cloned into the BamHI-NotI site of the multicloning site of AdpRec (27). The ⌬myo5a 1830 (nucleotides 5534 -5660) containing the C-terminal-most 22 amino acids was cloned into the EcoRV site of the multicloning site of AdpRec. For both truncation constructs, the shuttle vector was linearized by EcoRI and cotransfected with ClaI-XbaI linearized Ad-5-␤-gal replication-deficient adenovirus DNA into HEK293 cells. The virus was propagated in HEK293 cells, and cell lysates containing ϳ10 8 plaque-forming units/l replication-deficient Ad5-⌬myo5a tail or replication-deficient Ad5-⌬myo5a 1830 were then used as a reagent for in vivo analysis of D2p29 vesicle trafficking.
Rapid Acquisition Digital Imaging Microscopy-Rat astrocytes were grown on poly-D-lysine-coated (10 g/ml) coverslips and infected with Ad5-D2p29 GFP (multiplicity of infection ϭ 5-6), and expression of the GFP fusion protein was visually confirmed 24 h later. The cells were then grown in serum-free medium for 24 h, and D2p29 GFP translocation to the plasma membrane was done by an overnight incubation with 100 nM hydrocortisone and 1 mM dibutyryl cAMP (45). Twenty minutes prior to study the microtubules were disassembled with 10 M colchicine. Coverslips were then placed on a heated stage (37°C), the individual cells were isolated, and the treatment medium containing 10 nM thyroid hormone (T 4 , rT 3 , or T 3 ) or 0.1% bovine serum albumin vehicle control was added. The movement of the D2p29 GFP molecule was followed by collecting image data sets composed of 40 images taken at 250-nm focal increments. Each data set was collected in 200 msec and repeated every 15 s for a total of 10 min. Computer reconstructions (46) created three-dimensional images that were assembled into time-lapse movies. Representative QuickTime movies may be viewed in the on-line version (http://www.jbc.org).
Statistics-All experiments were done a minimum of three times, and where appropriate, statistical analysis was performed using Student's t test.

RESULTS
Real-time Analysis of Hormone-dependent Vesicle Trafficking-Real-time rapid-acquisition digital microscopy was used to visualize the hormone-dependent centripetal movement of the recycling vesicles in astrocytes expressing a catalytically active green fluorescent fusion protein of the 29-kDa subunit of D2 (D2p29 GFP ) (27). Prior end-point analysis revealed that the principal circulating hormone T 4 , but not the transcriptionally active metabolite T 3 , initiated the delivery of D2 vesicles from the cell periphery to the perinuclear space over 20 min (23)(24)(25). D2p29 GFP was delivered to the cell periphery prior to the vesicle trafficking study by cAMP stimulation, and the microtubules, which have no effect on recycling vesicle internalization, were disassembled with colchicine (23). Recycling vesicle movement was initiated by hormone(s), and the centripetal movement of the GFP reporter was monitored by collecting stacked image data sets at 250-nm focus increments through the cell every 15 s.
Three-dimensional image reconstructions were assembled into time-lapse movies, and the photomicrographs in Fig. 1 show representative "snapshots" of D2p29 GFP -expressing cells at 0, 4, and 8 min after adding T 4 , T 3 , or no hormone. A two-dimensional projection of the three-dimensional data set (th1.qt) may be viewed in the on-line version (http://www.jbc. org). At the beginning of the experiment discrete D2p29 GFP signals were distributed around the cell periphery in all treatment groups. In hormone-free cells, the fluorescent reporter wobbled in the same position on the plasma membrane throughout the experimental period. In T 4 -treated cells, the fusion protein remained at the periphery of the cell for up to 4 min and then showed episodic centripetal movement toward the cell nucleus. Individual vesicles in transit showed frequent perpendicular shifts preceded by short pauses, and by 8 min Ͼ80% of the recycling vesicles were deposited in the perinuclear space. In contrast, T 3 -treated cells showed no directed movement, and the fluorescent reporter remained at the cell periphery as observed in hormone-free cells. Vector analysis was then used to determine the path, distance traveled, and the velocity of 30 -45 individual recycling vesicles. Fig. 2 shows the vector profiles for D2p29 GFP vesicles in hormone-free and thyroid hormone-treated cells. In hor-mone-free cells or cells treated with T 3 , the D2p29 GFP vesicles oscillated within 500 nm of the plasma membrane with average velocities (Ϯ S.E.) of 46 Ϯ 8 nm/sec (n ϭ 36, no hormone) and 40 Ϯ 5 nm/sec (n ϭ 30, T 3 -treated), respectively. In contrast, both T 4 -and rT 3 -treated cells showed centripetal but interrupted vesicle movement from the cell periphery to the perinuclear space. In both T 4 and rT 3 cells, individual vesicles showed 3-8 episodes of linear centripetal movement over the 10-min evaluation period, and the average of peak velocities (mean Ϯ S.E.) during linear vesicle movement was 88 Ϯ 16 nm/sec (n ϭ 30) in the T 4 -treated cells and 105 Ϯ 26 nm/sec (n ϭ 32) in the rT 3 -treated cells.
T 4 -initiated Centripetal Vesicle Movement Uses Actin Cables-We next examined the effects of inhibitors of microfilament depolymerization and the myosin ATPase inhibitor, BDM, on recycling vesicle trafficking. Cyclic AMP-induced D2p29 GFP -expressing astrocytes were pretreated for 20 min with 10 M colchicine alone or in combination with either 10 M dihydrocytochalasin B or 10 mM BDM, and vesicle movement was initiated by T 4 . As shown in Fig. 3, the colchicine-treated control cells showed that T 4 initiated vesicle trafficking from the cell periphery to the perinuclear region of the astrocyte like that observed above (see Fig. 2). Consistent with earlier biochemical studies (21,23), both 10 M dihydrocytochalasin B and 10 mM BDM completely arrested T 4 -initiated vesicle movement (Fig. 3), and the vesicles remained within 500 nm of the cell periphery.
The C Terminus of Myosin 5a Tethers the Recycling Vesicle to the Actin-based Motor-Because our biochemical studies (21) showed that the C terminus of myosin 5a tethered the D2p29 vesicle to actin fibers in vitro, we examined the myosin 5a domain(s) responsible for the hormone-dependent docking of recycling vesicles in living cells. To deliver the recombinant myosin 5a tail mutants to astrocytes, we used replicationdeficient adenoviral constructs containing the complete myosin 5a tail domain (⌬myo5a tail ), a fragment encoding for the last 22 residues of the myosin 5a (⌬myo5a 1830 ), and a Sinbis pseudoviron containing the mRNA encoding the coiled-coil domain of myosin 5a (⌬myo5a coiled-coil ). Control cells were infected with Ad5-␤-gal to determine the influence, if any, of the adenoviral delivery system on astrocyte vesicle trafficking. Shown in Fig.  4 are representative vector tracings of individual D2p29 GFP vesicles in T 4 -treated astrocytes expressing the individual myosin 5a truncation mutants and velocity diagrams for T 4 -initiated movement of individual vesicles. Representative video files for cells expressing the ⌬myo5a tail (T4_myo5atail.qt) and ⌬myo5a coiled-coil (T4_myo5acc.qt) may be viewed in the on-line version (http://www.jbc.org). All the Ad5-⌬myo5a mutant and Simbis-infected cells remained viable for up to 5 days after infection (the longest time tested) and showed cAMP-induced increases in D2 activity identical to that observed in uninfected cells. As expected, control cells expressing Ad5-␤-gal showed T 4 -initiated centripetal vesicle trafficking identical to uninfected astrocytes, indicating that adenoviral infection did not alter hormone-dependent vesicle trafficking (Fig. 4). Overexpression of the ⌬myo5a tail completely blocked the T 4 -initiated vesicle transport and decreased the maximal velocity of D2p29 GFP -containing vesicles to 38 Ϯ 8 nm/sec (n ϭ 30), a value equal to that in the hormone-free cell (see Fig. 2). Similarly, cells expressing the last 22 residues of myosin 5a (⌬myo5a 1830 ) showed no hormone-dependent directed vesicle movement, and the D2p29 GFP vesicles oscillated within 500 nm of the plasma membrane with average velocities (Ϯ S.E.) of 41 Ϯ 5 nm/sec (n ϭ 30). These data confirm that the last 22 residues of myosin 5a are essential for vesicle trafficking in situ and show that this domain serves as the hormone-dependent To determine whether myosin 5a-mediated vesicle trafficking in living cells required two functional motor domains, myosin 5a dimerization was disrupted with a ⌬myo5a coiled-coil mutant (residues 892-1040), and the consequences on T 4 -initiated vesicle trafficking were examined. In cells expressing the ⌬myo5a coiled-coil , D2p29 vesicle trafficking was divided into two roughly equal populations: one showed no directed movement (immobile), and the other showed directed long range centripetal movement identical to control T 4 -treated cells (mobile) (Fig. 4). Unlike our in vitro findings in which this ϳ20-kDa ⌬myo5a mutant had no effect on vesicle binding to the actinbound myosin 5a motor (21), in living cells it decreased the number of migrating vesicles by ϳ50%. Mobile vesicles had peak velocities (Ϯ S.E.) during linear movement of 90 Ϯ 10 nm/sec (n ϭ 16), which is equal to that of cargo-laden native myosin 5a dimers (see Fig. 2).
Analysis of the individual vesicle paths in hormone-free cells, T 4 -treated cells, and the T 4 -treated cells expressing D2p29 GFP -expressing rat astrocytes co-expressing were infected with either Ad5 ␤-gal , ⌬myo5a tail , ⌬myo5a 1830 , or ⌬myo5a coiled-coil 2 days prior to an overnight treatment with dibutyryl cAMP and hydrocortisone in hormone-free medium. Vesicle recycling was initiated by treatment with 10 nM T 4 , and images were processed as detailed under "Experimental Procedures." Vesicle paths are shown for representative cells (left). N, nucleus. Velocity diagrams were constructed for three representative vesicles by measuring the distance traveled during each 15-s interval. Velocity diagrams are present for both mobile and immobile vesicle pools in cells expressing ⌬myo5a coiled-coil . Representative two-dimensional video projections of the three-dimensional data for the cells expressing the ⌬myo5a tail (T4_myo5atail.qt) and the ⌬myo5a coiled-coil (T4_myo5acc.qt) may be viewed in the on-line version (http://www.jbc.org). ⌬myo5a coiled-coil is shown in Fig. 5. In hormone-free cells, individual vesicles oscillated within 500 nm of the plasma membrane and did not migrate to the cell interior during the 10-min observation period. In T 4 -treated cells, individual vesicles showed 3-8 episodes of linear centripetal movement of up to 3 m that lasted up to 30 s with velocities ranging from 70 to 200 nm/s. All vesicles examined also showed several abrupt lateral shifts and occasional retrograde movement that bracketed the long range vesicle movements. Closer examination of individual members of the two pools of D2p29 vesicles in ⌬myo5a coiled-coil -expressing cells showed vesicles with no directed movement similar to that in hormone-free cells (immobile vesicles) and vesicles with directed long range episodic linear migration and cable shifts that are identical to those of D2p29 vesicles docked to the native myosin 5a dimers in T 4treated cells (mobile vesicles) (Fig. 5).

Thyroid Hormone Initiates the Formation of a Vesicle-Myosin 5a Complex Without Affecting the Vesicle Budding Reaction-
Both in vitro and in situ, T 4 initiates the formation of a detergent-insoluble complex between vesicles containing D2p29 and actin fibers (25,26,28). To determine whether the dominant negative ⌬myo5a constructs altered the affinity of native myosin 5a for the actin cytoskeleton or competed with native myosin 5a for docking to the D2p29 containing vesicles, we examined the association of both native myosin 5a and the ⌬myo5a constructs with the Triton-insoluble actin cytoskeleton (25,26,28). Because synaptophysin forms a reversible complex with myosin 5a in synaptic vesicles (22), we also examined whether this membrane-bound protein and selected other vesicle proteins were also part of this hormone-initiated actinbound complex. Shown in Fig. 6A are the results of the immunoblot analysis of the T 4 -initiated interactions between selected vesicle proteins, affinity-radiolabeled D2p29 (30), and native myosin 5a with the Triton-insoluble actin cytoskeleton. In cell lysates, T 4 treatment had little if any influence on any of the proteins evaluated, indicating that T 4 did not alter the formation of primary endosomes but promoted the formation of a complex between vesicle-docking protein(s) on the primary endosome and myosin 5a, presumably in preparation for centripetal movement of the vesicle to the cell interior. Rab3, synaptophysin, synaptotagmin, and D2p29 were all present in the Triton-insoluble complex of T 4 -treated cells but lacking in this fraction from control hormone-free cells. In contrast, the full-length ϳ190-kDa myosin 5a was always found in the Triton-insoluble fraction from both hormone-free and T 4 -treated astrocytes.
Direct analysis of the distribution of the native myosin 5a and the truncation mutants of myosin 5a is shown in Fig. 6B. Cells expressing ⌬myo5a tail , ⌬myo5a 1830 , and ⌬myo5a coiled-coil were treated with T 4 for 20 min, and the Triton-soluble and -insoluble fractions were prepared as detailed previously (25,26,28). As expected, the native myosin 5a was localized exclusively to the Triton-insoluble fraction of astrocytes expressing the ⌬myo5a tail , whereas the ⌬myo5a tail lacking the actin binding head was exclusively found in the Triton-soluble fraction. Similarly, Ͼ90% of the native 190-kDa myosin 5a in ⌬myo5a 1830 -expressing cells was found in the Triton-insoluble fraction, suggesting that these two dominant negative myosin 5a constructs derived from the vesicle binding tail of myosin 5a do not alter the binding of native myosin 5a to the actin cytoskeleton.
In ⌬myo5a coiled-coil -expressing cells, the native myosin 5a was found in both the detergent-soluble and detergent-insolu-

FIG. 6. Characterization of the T 4 -dependent association of vesicle-associated proteins and myosin 5a to the actin cytoskeleton in control rat astrocytes and rat astrocytes expressing dominant negative myosin 5a mutants.
A, astrocytes were grown in hormone-free medium for 24 h, D2 expression was induced as detailed elsewhere (45), and cells were treated with vehicle or 10 nM T 4 for 20 min. The cell lysates were prepared in Triton X-100, and the Tritoninsoluble actin cytoskeleton was isolated as detailed elsewhere (25) and resolved by SDS-polyacrylamide gel electrophoresis. Individual proteins were identified by immunoblot analysis. B, astrocytes expressing ⌬myo5a(s) were grown for 2 days, and D2 expression was induced as detailed above. Triton X-100 cell lysates (L), Triton-soluble supernatant (S), and Triton-insoluble pellet (P) were isolated and resolved by SDSpolyacrylamide gel electrophoresis, and the immunoreactive myo5a identified with anti-myo5a antibodies directed against the coiled-coil and the C terminus of rat myo5a as detailed under "Experimental Procedures." ble fractions (Fig. 6B), with 25-30% of the full-length myosin 5a present in the Triton-soluble fraction and the remaining 70 -75% associated with the Triton-insoluble actin cytoskeleton. Unlike the native myosin 5a, the ϳ20-kDa ⌬myo5a coiled-coil was found only in the Triton-soluble fraction, suggesting that heterodimers formed between native myosin 5a and ⌬myo5a coiled-coil are not bound tightly to the actin cables in astrocytes.
To determine the effects of T 4 on the binding of native myosin 5a and our cohort of ⌬myo5a mutants to D2p29 vesicles, we affinity-radiolabeled (30) native D2p29 recycling vesicles in situ, and D2p29 vesicle binding myosin 5a was initiated by T 4 . D2p29 vesicles were then immunopurified using anti-D2p29 IgG conjugated to magnetic Dynal ® beads, and selected vesicleassociated proteins were identified by immunoblot analysis. The data in Fig. 7A show that three vesicle proteins frequently found in recycling vesicles, rab3, synaptotagmin, and synaptophysin, were all present in the purified D2p29 vesicles from both hormone-free control and T 4 -treated cells, consistent with the lack of effect of T 4 on the formation of primary endosomes. In contrast, T 4 treatment caused the native myosin 5a to bind to the recycling D2p29 vesicles, similar to our previous in vitro results (21). Little if any endoplasmic reticulum contaminated these immunopurified vesicle preparations as judged by the lack of the endoplasmic reticulum marker protein, protein disulfide isomerase.
Finally, we examined the effects of ⌬myo5a(s) on the T 4initiated interaction(s) between the native motor protein and the D2p29 vesicle. D2p29 GFP -expressing astrocytes co-expressing ⌬myo5a tail , ⌬myo5a 1830 , or ⌬myo5a coiled-coil were treated with 10 nM T 4 for 20 min, and the D2p29 GFP -containing vesicles were immunopurified using anti-GFP IgG conjugated to immobilized protein A beads. Native myosin 5a and ⌬myo5a(s) were then identified by immunoblot analysis. The data summarized in Fig. 7B show that the immunopurified D2p29-containing vesicles from cells expressing either the ⌬myo5a tail or ⌬myo5a 1830 do not contain the native 190-kDa myosin 5a, consistent with the idea that these ⌬myo5a mutants compete directly with the native myosin 5a for the recycling vesicle. As expected, abundant ⌬myo5a tail protein was found associated with the D2p29 vesicles in T 4 -treated D2p29 vesicles from ⌬myo5a tail -expressing cells (Fig. 7B). In contrast, both the fulllength native myosin 5a and the 20-kDa ⌬myo5a coiled-coil proteins were found associated with D2p29 vesicles from cells expressing ⌬myo5a coiled-coil , indicating that ⌬myo5a coiled-coilnative myosin 5a heterodimers can bind vesicles. These data confirm our in vitro findings that the vesicle-binding domain of myosin 5a is localized to the globular tail, specifically the last 21 amino acids of the motor protein in living cells. Further, they show that the influence of the dominant negative ⌬myo5a tail mutants on T 4 -initiated vesicle trafficking depended on competition for vesicle binding sites rather than disruption of the binding of native myosin 5a dimers to the actin cytoskeleton. DISCUSSION The long range (up to 10 m) myosin 5a-mediated vesicle movement in astrocytes differs from the more regionalized movement observed in cultured melanocytes. In Xenopus melanophores the long range transport of pigment vesicles relies on both the microtubule motor protein kinesin as well as myosin 5a (2,31,32). Actin depolymerization in fish melanophores prevents the stimulus-induced dispersal of melanosomes but not the long range microtubule-based movement of pigment granules (33). More recently, myosin 5a was shown to be responsible for the short range shuttling of pigment vesicles along filaments of the cortical actin cytoskeleton (1,2,4,32,34). Because the velocity of microtubule-based vesicle movement in dilute melanocytes is almost twice as fast as that in normal cells, it seems that vesicle-bound myosin 5a retards the kinesin-driven centrifugal progress by dynamic interactions with neighboring actin filaments alongside the microtubules (1,4). Unlike the complex interactions between the microtubuleand microfilament-based motors observed in melanocytes, we show that long range centripetal movement of recycling vesicles in astrocytes is (i) mediated by myosin 5a, (ii) processive, (iii) hormone regulated, and (iv) unaffected by the microtubules or their associated motor proteins.
Time-lapse motion studies enabled us to examine the T 4initiated centripetal movement of recycling vesicles in living cells and to establish that myosin 5a delivered the vesicle cargo to the endosomal sorting pool along actin fibers. Immunoblot analysis of this recycling vesicle pool revealed that T 4 promoted the formation of a complex of vesicle proteins with myosin 5a in situ without altering the formation of primary endosomes. Using a dominant negative experimental paradigm, we also found that blocking myosin 5a dimer formation partially arrested vesicle trafficking, whereas overexpression of the entire myosin 5a tail or the recently identified vesicle-binding domain (21) completely halted centripetal vesicle movement. These data identify a hormone-dependent vesicle-binding domain located within the last 22 residues of the myosin 5a motor protein and show that this region of the molecular motor is responsible for tethering the recycling vesicle in living astrocytes.
The hormone-dependent binding of recycling vesicles to myosin 5a in astrocytes is similar to the reversible interactions between several vesicle-docking proteins (35)(36)(37) and this motor protein (20,22). A growing number of membrane-bound vesicle proteins that facilitate localization, docking, and degradation of the vesicle have been identified (35)(36)(37)(38), and at least three of these vesicle proteins (39) form T 4 -dependent complexes with actin-bound myosin 5a in living astrocytes. Importantly, T 4 does not alter the basic endocytic process, and the vesicle proteins rab3, synaptotagmin, and synaptophysin were all found in the detergent-resistant complex that presumably tethered the vesicle cargo to the actin-based motor.
The vesicle proteins that constitute this complex are fundamental to vesicle sorting in neurons and somatic cells. Rab3 is FIG. 7. Immunoblot analysis of immunopurified D2p29 vesicles. A, T 4 -dependent changes in D2p29 vesicle-associated proteins. Rat astrocytes were grown in serum-free medium for 24 h. D2 expression was induced as described elsewhere (45). The cells were treated in the presence and absence of T 4 , and D2p29 vesicles were immunoprecipitated from cell lysates using anti-D2p29 IgG. B, D2p29 GFP -expressing astrocytes were infected with ⌬myo5a mutants as detailed under "Experimental Procedures," D2 activity was induced as described previously (45), and vesicle recycling was induced with T 4 . D2p29 GFP vesicles were isolated from cell lysates by immunoprecipitation with anti-GFP IgG (10 g/ml) and separated by SDS-polyacrylamide gel electrophoresis. Vesicle-associated myosin 5a was identified by specific anti-myo5a antisera. a member of a group of GTP-binding proteins that target vesicles in transit to different intracellular destinations and along with synaptotagmin is found in recaptured synaptic vesicles in the presynaptic nerve terminal (39). Synaptophysin, the third vesicle protein that appears in the detergent-insoluble T 4 -dependent complex also forms a Ca 2ϩ -modulated complex with myosin 5a in brain synaptosomes (22). Thus, the ability of T 4 to promote interactions between key membrane vesicle-signaling proteins and myosin 5a is a potentially important function of this hormone that has been overlooked. Whether these three vesicle proteins are sufficient to form the docking complex for myosin 5a or other components are required remains to be established; no thyroid hormone binding domain has been reported on any of the members of this complex except for D2p29. However, it is unlikely that D2p29 initiates the formation of the docking complex, because alkylating substrate analogs covalently modify the T 4 binding site without initiating actinbased endocytosis (30). This suggests that an as yet unidentified T 4 -binding protein links the vesicle cargo to this binding domain of myosin 5a.
In vitro myosin 5a shows the properties of a processive motor protein and undergoes multiple ATP hydrolysis cycles coupled with mechanical displacement before dissociating from actin (16,17). Rapid-acquisition digital image microscopy allowed us to determine the path taken, the distance traveled in a finite period, and the velocity of individual vesicles in living cells. Once captured by myosin 5a, individual vesicles showed episodic movement with a centripetal bias and a velocity of ϳ100 nm/sec, in close agreement with that determined for the short range movement of pigment-laden vesicles in melanocytes (4). Close inspection of the reconstructed time-lapse sequences revealed that vesicles showed frequent abrupt lateral shifts between episodes of linear movement. Such patterns of movement are observed for a single myosin 5a motor moving along actin fibers in vitro (16,17) and are consistent with the ability of a processive motor to move between fiber track(s) in midstep.
As reported previously in vitro (21), expression of recombinant ⌬myo5a 1830 arrested hormone-initiated centripetal vesicle movement in living cells indicating that the recycling vesicles are tethered to myosin 5a through the last 22 residues of the motor protein. The myosin 5a tail domain has been implicated in vesicle transport from yeast to mice (8,40). At least two spontaneously occurring mouse mutants have lost segments of the myosin 5a tail and show altered phenotypes that illustrate the consequences of disrupted myosin 5a vesicle trafficking. Mice homozygous for the dilute-neurological mutation (Myo5a d-n ) lack the last 14 residues of myosin 5a and show postnatal neurological defects that are similar to those of the dilute lethal mouse (8), in which the loss of this motor protein disrupts the synaptic vesicle cycle in maturing cerebellar granule neurons (20). Similarly, mice carrying the Myo5a d-n2J mutation that eliminates the last 92 residues of the myosin V tail also show severe neurological defects during postnatal life. These defects in cerebellar development are reminiscent of the cerebellar defects and developmental delays in brain maturation that are often associated with neonatal hypothyroidism (41,42), and the mapping of the neurological defects in dilute strains to the hormone-dependent vesicle-binding domain of myosin 5a provides an important clue to one of the molecular events that mediates the morphogenic effects of T 4 in brain.
In vitro, myosin 5a monomers transport actin fibers (29,43). Assembly of "crippled" myosin 5a motors composed of a native full-length monomer and a truncated mutant made from the coiled-coil dimerization domain decreased by ϳ50% the number of vesicles showing hormone-dependent centripetal movement, suggesting that the loss of one motor head inactivates vesicle trafficking by myosin 5a. On the other hand, the remaining 50% of the recycling vesicles in cells expressing ⌬myo5a coiled-coil showed centripetal movement with velocities, centripetal bias, and distances traveled that did not differ from control cells. These data suggest that a motor composed of two native myosin 5a monomers carried the mobile vesicles. Because no recycling vesicles were found that showed a slowed rate of travel, our findings in living cells also indicate that myosin 5a molecules with only a single motor domain are poor cargo carriers. This is consistent with the recent model of processive movement by myosin 5a proposed by Walker et al. (18). Alternatively, failure of the recycling vesicle to tether to the C terminus of the crippled myosin 5a would also reduce the number of mobile vesicles. The findings in living cells cannot distinguish between these two possibilities.
In summary we show that T 4 initiates the tethering of recycling vesicles to the C terminus of myosin 5a and that the tethered vesicle moves down actin fiber cables to the endosomal sorting pool. Time-lapse sequences of reconstructed digital images allowed us to determine the velocity and distance traveled for individual recycling vesicles and to identify and characterize the hormone-dependent vesicle-binding domain of myosin 5a in a living cell. The ability to monitor the processive, centripetal movement of individual recycling vesicles in living astrocytes established the mechanism of myosin 5a vesicle trafficking in living cells and will prove invaluable for the characterization of the individual components that participate in the tethering of cargo vesicles to an actin-based motor protein.