ABSTRACT
One major mechanism of cell-mediated cytolysis is the polarized secretion of lytic granules, a process which is highly dependent on microtubules. We isolated lytic granules from murine cytotoxic T cells and tested their ability to bind to and move along microtubules in vitro. In the presence of a motor-containing supernatant, the granules bound to the microtubules and moved along them at an average maximal rate of 1 m/second. Virtually every granule could bind to microtubules, and about half translocated within a few seconds of binding. Motility required exogenous cytosolic motors, hydrolyzable nucleotides, and an intact granule membrane. Although the motor preparation used to support granule movement contains both plus- and minus-enddirected motor proteins, granule movement was strongly biased toward microtubule plus-ends. Inactivation of cytoplasmic dynein had little effect on granule binding and movement, but immuno-depletion of kinesin from the motor preparation inhibited granule binding by 50%. These results indicate that most granule movement in this assay is mediated by kinesin. The speed and direction of granule movement in vitro are sufficient to account for the release of lytic granules in the intact T cell. This model system should be valuable for studying the interactions of secretory granules with microtubules, and for identifying the regulatory factors involved.
INTRODUCTION
A major mechanism by which natural killer (NK) cells and cytotoxic T lymphocytes (CTL) kill their targets is by regulated exocytosis of specialized granules, termed lytic granules (reviewed by Henkart, 1985). When a cytolytic lymphocyte recognizes a virus-infected cell or a tumor cell as a suitable target, cross-linking of surface receptors triggers a rapid reorganization of the killer cell cytoskeleton and secretory apparatus. Within five minutes, the microtubule organizing center (MTOC), the Golgi complex, and the lytic granules face the bound target (Carpen et al., 1982; Geiger et al., 1982; Kupfer et al., 1985; Yannelli et al., 1986). In this way the lymphocyte assumes polarity with respect to its target. As soon as this polarity has been established, some of the granules move away from the MTOC and fuse with the plasma membrane, releasing their contents in the space between the killer cell and the target (Frey et al., 1982; Zagury, 1982). The proteins released from the granules then induce perforations in the target cell membrane and degrade the target cell’s DNA (Millard et al., 1984; Podack and Konigsberg, 1984; Hayes et al., 1989), thereby killing the target. Agents that depolymerize microtubules, including colchicine, nocodazole, vinblastine and vincristine, inhibit cell-mediated cytotoxicity (Katz et al., 1982; Kupfer et al., 1983). If the microtubules are allowed to repolymerize, cytolytic activity returns (Kupfer et al., 1983). Thus, the exocytosis of lytic granules by CTL represents one of the clearest cases of microtubule involvement in regulated secretion.
In lymphocytes, as in most cell types, microtubules are arrayed with their fast-growing (plus) ends at the periphery and their slow-growing (minus) ends at the cell center (Bergen et al., 1980). This provides a radially polarized cytoskeletal scaffold along which membranous organelles can bind and move. Several proteins which can mediate organelle-microtubule interactions have been identified, primarily as a result of studies that have reconstituted these interactions in vitro. These proteins include several organelle-microtubule binding proteins (Mithieux and Rousset, 1989; Kreis, 1990) and two major motor proteins, kinesin and cytoplasmic dynein (Brady, 1985; Vale et al., 1985a; Lye et al., 1987; Paschal and Vallee, 1987). Kinesin has been shown to support the motility of organelles toward the microtubule plus-end (Vale et al., 1985b; Porter et al., 1987), while cytoplasmic dynein supports organelle motility in the opposite direction (Paschal and Vallee, 1987; Schroer et al., 1989).
Two basic roles have been attributed to the microtubule cytoskeleton in membrane traffic: positioning stationary organelles in the cytosol and mediating the directional movement of motile organelles (reviewed by Vale, 1987; Kelly, 1990). In many cell types the Golgi complex, lysosomes and late endosomes cluster near the MTOC, and depolymerization of microtubules disrupts their distribution (Matteoni and Kreis, 1987; Swanson et al., 1987; Ho et al., 1989). It seems likely that minus-end-directed motors are responsible for establishing the clustering of organelles at the MTOC (Matteoni and Kreis, 1987; Ho et al., 1989; Bomsel et al., 1990). Plus-end-directed motors are likely to play a role in organelle positioning as well, since antibodies to kinesin can disrupt the distribution of tubular lysosomes along microtubules in the cell periphery (Hollenbeck and Swanson, 1990). Microtubules also direct the outward movement of organelles. The best studied example of this is fast axonal transport, where secretory vesicles and mitochondria are transported along microtubules to the nerve terminal (Sheetz et al., 1989). In non-neuronal cells, microtubule-based movement is not required for secretion in general, but it is critical for directed transport of secretory vesicles to specific membrane domains (Rindler et al., 1987; Achler et al., 1989; Eilers et al., 1989; Kreis et al., 1989).
To understand how cells regulate the timing and directionality of particular organelle movements, we set out to reconstitute the binding and movement of a purified population of organelles on microtubules. The lytic granules of CTL are ideally suited to this purpose, for two reasons. First, the granules have a distinct size and structure, which enable their purification with relative ease (Millard et al., 1984; Podack and Konigsberg, 1984). Second, CTL are likely to have a well developed mechanism for regulating microtubule-based granule movements, since the positioning and secretion of the granules is vital for CTL function. Therefore, we isolated lytic granules from CTL and examined in vitro their interaction with microtubules. We show that in the presence of exogenous motors, lytic granules bind to microtubules and translocate along them in vitro. Consistent with their secretory destination, the movement of lytic granules in vitro is toward the plus-end of the microtubule, and is mediated preferentially by kinesin.
MATERIALS AND METHODS
Cell culture
To generate quantities of granular CTL suitable for fractionation, primary cultures of splenic T cells were polyclonally stimulated with lectin and lymphokines by an adaptation of the procedure of Hardt et al. (1985). C57Bl/6 mice were obtained from the National Cancer Institute Animal Program. Spleens were removed aseptically, and splenocytes teased into RPMI-1640 (Sigma) containing 10% fetal calf serum, non-essential amino acids, penicillin, streptomycin (all from JRH Biosciences) and β-mercaptoethanol (Kodak). Precursors of cytotoxic cells were positively selected by panning for CD8 + cells as described by Sprent and Shafer (1985) using monoclonal anti-CD8 antibody 31M (gift from Dr. R. Kurlander, Duke Univ.). Typically, 5-8% of the starting splenocytes were recovered after panning. Cells were seeded for culture at 1 × 106/ml in medium supplemented as described above. Concanavalin A (1 μg/ml; Con A, Aldrich) was added as a polyclonal mitogen. To further stimulate growth and differentiation of cytotoxic T cells, medium was also supplemented with 20 units/ml recombinant human interleukin 2 (Dupont, gift from Dr. A. Hollingsworth, U. Nebraska) and 10% Con A supernatant, produced by culturing rat splenocytes for 2 days at 2 × 106/ml in the presence of 5 μg/ml Con A (gift from Dr. D. Howell, Duke University). These culture conditions have been shown to support the preferential differentiation of CTL from a mixed population of cells (Erard et al., 1985; Hardt et al., 1985). After 2 days in culture, cells were adjusted daily to 3 × 105/ml with medium and growth factors, without additional lectin stimulation. Cells were cultured for a total of 5-6 days, during which time their numbers increased by more than 25-fold and the cells developed granules detectable by fluorescence and electron microscopy. FACS analysis revealed that the expanded cultures contained greater than 95% CD8 positive cells. In parallel with the development of granules, the CTL developed cytolytic activity. Cytolysis was completely dependent on calcium, indicating that granule exocytosis was the main lytic mechanism used (Henkart, 1985). Though these cultures survived for 10 days, cells after 5-6 days were judged to be in optimal condition, and were therefore used for granule preparations.
Primary cultures of chicken embryo fibroblasts (CEF) were prepared from 11-day embryos as described by Kelley and Schlesinger (1978) and maintained in Earle’s minimal essential medium, supplemented with 5% fetal calf serum, penicillin and streptomycin (all from Gibco). After 2 days, primary cultures were frozen in 30% FCS, 10% DMSO, 60% Iscove’s modified Dulbecco’s modified medium (Gibco) and maintained under liquid nitrogen. Secondary cultures were prepared from these frozen cells, and used within 2-3 days of plating.
Preparation of lytic granules
Lytic granules were prepared by a modification of the procedure of Millard et al. (1984). A total 2 × 108 to 1 × 109 CTL were washed with balanced salts solution, adjusted to 2 × 108/ml in disruption buffer (0.25 M sucrose, 4 mM EGTA, 10 mM Hepes, pH 7.4) containing protease inhibitors, and disrupted by 10 passes through a ball-bearing homogenizer at 4°C. Nuclei and unbroken cells were removed by centrifugation at 1000 g in a Sorvall RT6000B centrifuge and the resulting pellet was washed with disruption buffer and centrifuged again. Percoll (Pharmacia) was made isotonic by the addition of 1/10 vol. of 2.5.M sucrose and 1/100 vol. of 1 M Hepes, pH 7.4, and 20 ml gradients of 48% (v/v) isotonic Percoll in disruption buffer were prepared. Each gradient was overlaid with lysate from 1 × 108 to 3 × 108 cells, and gradients were centrifuged at 60,000 g for 30 minutes at 4°C using a 70Ti rotor (Beckman L3-50 ultracentrifuge). Fractions (0.8 ml) were collected from the bottom of the tubes, and analyzed as described below. For dense granule preparations, fractions 3-7 of the gradient (δ = 1.12 to 1.09 g/ml) were pooled, diluted with one volume of disruption buffer, and concentrated by centrifugation for 45 minutes at 85,000 g at 4°C using a Beckman Ti70.1 rotor. The visible granule band was collected, and granules were concentrated again by centrifugation for 30 minutes at 100,000 g at 4°C using a TLA-45 rotor (Beckman TL-100 ultracentrifuge).
Analysis of membrane fractions
Density of gradient fractions was calculated on the basis of refractive index. The protein profile of the gradients was determined as TCA-precipitable counts from cells metabolically labelled with [35S]methionine. CTL (5 × 106) were labelled for 10 hours (greater than one generation time) in medium supplemented as usual, except that it contained 5% dialyzed newborn calf serum, 1/10 the normal amount of unlabelled methionine, and 75 μCi/ml [35S]methionine (Translabel, ICN). Radiolabelled cells were mixed with an excess of unlabelled cells and fractionated as usual. Samples were adjusted to 0.5% NP-40, 0.1 M Tris, pH 8.0, precipitated with 10% TCA, and relative protein levels were determined by scintillation counting. The granule marker enzymes BLT esterase and perforin/cytolysin, and the mitochondrial marker succinate dehydrogenase were assayed using the microassays described by Young et al. (1987). Plasma membrane was detected by fractionating cells, which had first been surface biotinylated (Lisanti et al., 1989), and blotting with 125I-streptavidin. Endoplasmic reticulum was monitored by blotting with rabbit anti-ribophorin I (gift from Dr. D. Meyer, UCLA).
Electron microscopy
Granule fractions prepared as described above were mixed with an equal volume of ice-cold fixative (2% glutaraldehyde, 2% osmium tetroxide, 0.25 M sodium cacodylate, pH 7.4), and fixed in suspension for 30 minutes at 4°C, as described by Millard et al. (1984). Membranes were then pelleted in an Eppendorf microcentrifuge, washed with cacodylate buffer, and embedded in agar. Membranes were stained en bloc with uranyl acetate, dehydrated, and embedded in EMBED-812. Silver sections were obtained using a Reichert Ultracut E ultramicrotome and observed with a Philips EM -300 electron microscope operating at 80 kV.
Preparation of cytosol from chicken embyro fibroblasts
The cytosolic motor fraction was prepared as previously described (Schroer et al., 1989). Briefly, approximately 4 × 108 secondary passage CEF were collected by trypsinization, and washed with 35 mM PIPES, pH 7.4, 5 mM MgSO4, 5 mM EGTA, 0.5 mM EDTA, 1 mM DTT, and protease inhibitors (PMEE). The pelleted cells were resuspended in an equal volume of ice-cold PMEE, and passed 6 times through a ball-bearing homogenizer. Nuclei and unbroken debris were pelleted at 1000 g, and the supernatant (S1), was centrifuged for 30 minutes at 100,000 g using a TL55 rotor in a Beckman TL100 ultracentrifuge (all at 4°C). 1 mM GTP (Sigma) and 20 μM taxol (gift from Dr. Nancita Lomax, National Cancer Institute) were added to the resulting supernatant (S2), and the mixture was incubated for 15 minutes at 37°C to polymerize endogenous microtubules. Assembled microtubules were removed by centrifugation at 100,000 g for 5 minutes in a Beckman airfuge. The resulting supernatant (S3) was maintained at 4°C and used as the source of cytosolic motors.
Organelle motility assays
Bovine brain tubulin was purified by phosphocellulose chromatography (Williams and Lee, 1982), and stored at −70°C until use. Microtubules were polymerized by incubating tubulin at 2-4 mg/ml in PMEE containing 20 μM taxol and 1 mM GTP at 37°C for 15 minutes. Assays were prepared by spotting 1 μl of diluted microtubules, 3 μl of cytosolic motors, 1.2 μl 10 mM ATP, and 1 μl of organelles directly onto a glass coverslip. Movement was visualized using video-enhanced DIC microscopy (Allen et al., 1981; Kuo et al., 1991). Since the activity of different cytosolic motor preparations was somewhat variable, all comparisons were made from samples analysed in parallel. When indicated, substitutions were made for the cytosolic motor preparation. These included purified kinesin isolated from either chick brain (Schroer et al., 1988) or squid optic lobe (Vale et al., 1985a), cytoplasmic dynein isolated from chick brain (Schroer et al., 1989), and 0.5 mg/ml casein (Sigma). Motility rates were determined by identifying short time intervals during which organelles travelled in a straight line, and dividing the distance travelled by the time interval.
For some experiments, organelles were trapped and presented to microtubules using the laser trap described by Kuo et al. (1991). Laser power was adjusted to minimal levels, such that tightly bound organelles could not be removed from microtubules, and motile organelles pulled themselves out of the trap. The behavior of granules within 30 seconds of binding was observed and recorded.
To determine the orientation of granule movement, granules were mixed with kinesin-coated beads and the direction of their movement on the same microtubule was compared. Kinesincoated beads were prepared by mixing carboxylated beads (Polysciences, 0.143 μm diameter) with purified squid kinesin in the presence of NaPMEE (PMEE containing 80 mM NaCl) and incubating 5 minutes at 25°C. The binding was blocked by adding FCS to 10% and incubating an additional 5 minutes. Unbound protein was removed by pelleting beads through a cushion of 5% sucrose in NaPMEE. Coated beads were resuspended in PMEE containing 150 μg/ml casein as a stabilizing agent. The usefulness of these beads as standards for directionality relies on knowing that their movements are due to kinesin, even in the presence of the cytosolic motor preparation. We therefore prepared control beads, coated in parallel with heat-inactivated kinesin. These control beads exhibited very low binding and movement along microtubules, even in the presence of the mixed motor extract.
As an alternative assay for directed granule movement, sea urchin axonemes were prepared according to Bell et al. (1982), and used as seeds for microtubule polymerization. A mixture of NEM-conjugated tubulin and unconjugated tubulin was used to favor growth from the plus-end of the axoneme (Vale and Toyoshima, 1988). After the addition of taxol to stabilize microtubules, these axonemes were used in the standard motility assay in place of randomly oriented microtubules.
Protease, detergent and antibody treatment of granule surfaces
For protease treatment, purified granules were incubated for 1 hour at 4°C with 20 μg/ml trypsin (type XIII, Sigma). Prior to assaying motility, trypsin was inactivated for 10 minutes with an excess (300 μg/ml) Trasylol (FBA Pharmaceuticals). This quantity of Trasylol was sufficient to inhibit 20 μg/ml trypsin completely in a colorimetric enzyme assay (Young et al., 1987). For control digests, trypsin and Trasylol were mixed for 10 minutes, and the mixture was then incubated with granules for 1 h at 4°C. Detergent treatment was performed by mixing granules with 0.02% (v/v) Triton X-100, and incubating on ice for 10 minutes prior to assaying motility. For antibody blocking studies, purified granules were incubated for 1 hour at 4°C with a polyspecific rabbit antigranule antiserum (Reynolds et al., 1987; gift from Dr. P. Henkart, National Institutes of Health, Bethesda, MD) diluted either 1:100 or 1:30 in PMEE. Goat anti-rabbit Ig antiserum was used in parallel as a negative control.
UV photocleavage
Samples of organelles and S3 cytosol were incubated for 15 minutes in the presence of 2 mM ATP and 100 μM sodium vanadate, and exposed to long wave ultraviolet light for 50 minutes as described by Schroer et al. (1989).
Immunodepletion of kinesin
SUK4 (anti-kinesin) Sepharose (Ingold et al., 1988) or Protein A/Sepharose (Sigma) was pre-equilibrated with PMEE, and 300 μl of S3 was added to 50 μl of each packed resin. Slurries were incubated for 1 hour at 4°C, and the supernatant was removed and subjected to a second round of depletion with fresh affinity resin. Further depletion removed no more kinesin. Resin was washed three times with PMEE containing 0.5 mg/ml casein and bound material was eluted with 0.2 M glycine-HCl, pH 2.5.
Gel electrophoresis and Western blotting
7.5% SDS-polyacrylamide gels were run using the Biorad minigel apparatus and the buffers of Laemmli (1970). Western blotting was performed as described previously (Dabora and Sheetz, 1988), using alkaline phosphatase-conjugated secondary reagents.
RESULTS
Purification of lytic granules
As a source of lytic granules, we used murine CTL grown in large-scale cultures. CTL homogenates were fractionated by Percoll density centrifugation, as described by Millard et al. (1984). A representative fractionation is shown in Fig. 1. The gradient fractions were monitored for BLT esterase, which is the enzymatic activity of the major lytic granule protease, granzyme A (Pasternack and Eisen, 1985). As described previously (Millard et al., 1984), this procedure yielded two peaks of lytic granules, a dense peak centered around δ = 1.105 ± 0.02 g/ml, and a light peak around δ = 1.075 ± 0.02 g/ ml (Fig. 1A). In addition to granzyme A, the dense peak was enriched for perforin/cytolysin, as measured by hemolytic activity. A pool of fractions 3-7 of this peak, containing 5 × 106 cell equivalents, was sufficient to completely lyse 1 × 107 red blood cells in 1 hour at 37°C. The light peak of BLT-esterase activity comigrated with a major peak of total proteins (Fig. 1B). In addition to granule proteins, this peak contained the mitochondrial protein succinate dehydrogenase (Fig. 1A), as well as ER and plasma membrane markers (detected immunologically, data not shown). None of these markers was detectable in the dense granule fractions.
The purity of granules in the dense peak is illustrated in the electron micrograph shown in Fig. 2. The pooled material contained almost exclusively granules with electrondense cores. The size of the granules was relatively uniform (mean diameter: 0.45 ± 0.1 μm). Occasional granules exhibited multivesicular regions surrounding the dense core (Fig. 2, asterisk), although these are difficult to distinguish with the fixation procedure used. The granules were surrounded by a membrane bilayer, or by several membrane lamellae (Fig. 2, arrow), which in most granules appeared to be intact. The size and morphology of the dense core granules were very similar to the type I granules as observed in CTL and NK cells (e.g. see Fig. 1 of Burkhardt et al., 1990). No mitochondria, Golgi elements, endoplasmic reticulum or other contaminating organelles were observed in these fractions. Thus, the dense core granules isolated by this procedure are highly purified and morphologically intact. On the basis of the specific activity of BLT esterase, we estimate that the dense granules in fractions 3-7 were enriched by more than 200-fold. These pooled fractions were therefore used as the source of lytic granules for the following in vitro motility studies.
Lytic granules move on microtubules in vitro
To determine whether lytic granules can bind to and move along microtubules in vitro, granules were added to a motility assay containing microtubules polymerized from bovine brain tubulin, ATP, and a cytosolic extract from CEF cells. This extract contains both kinesin and cytoplasmic dynein, and has been shown to support the motility of mixed membranous organelles (Dabora and Sheetz, 1988; Schroer et al., 1989). In the presence of the cytosolic extract, the dense core granules bound to the microtubules and translocated along them. Fig. 3 shows a series of video images from one such assay. During this sequence, one granule (arrow) moves several micrometers along a microtubule. The second granule in the field has encountered an area where several microtubules intersect; it begins to move down a microtubule between Fig. 3E and F. Granules frequently switched from one microtubule to another, but they did not reverse direction. From sequences like the one shown in Fig. 3, the average rate of granule translocation was calculated to be 0.7 ± 0.3 μm/second, with a range of 0.3 to 1.2 μm/second (at 25°C). Much of the variation in this average rate is due to retardation of granule movement by obstacles in the web of microtubules. Because of this, a more useful parameter is the maximal rate of granule movement, which was 1.0 ± 0.2 μm/second. Using either value, the rate of lytic granule translocation is in good agreement with rates reported for kinesin- and cytoplasmic dynein-driven motility of organelles and motor-coated beads (Vale et al., 1985a; Schroer et al., 1989).
Motility requires cytosol, hypotonic buffer and hydrolyzable nucleotides
The binding of granules to microtubules was absolutely dependent on the addition of exogenous cytosolic extract; if BSA was substituted for cytosol, the granules remained free or became stuck to the glass coverslip. Substitution of casein for cytosol prevented the granules from binding non-specifically to the coverslip, as previously reported for squid axoplasmic organelles (Schnapp et al., 1991). However, binding of granules to microtubules was not observed in the presence of casein unless cytosol was also added (data not shown).
As in other organelle motility systems (Vale et al., 1985a; Dabora and Sheetz, 1988), granule motility was favored by hypotonic buffer conditions, and was inhibited by the addition of KCl. A concentration of 5-10 mM KCl in the assay did not inhibit granule binding, but inhibited granule movement by over 70%. Addition of 50 mM KCl to the assay diminished binding to 20% of control levels, and completely abolished movement. Since the motor-dependent binding of microtubules to the coverslip was also inhibited by KCl, it is likely that KCl acts directly on the motor proteins, as opposed to the granule membranes.
ATP was not required for granule binding, but was required for motility. This effect was more pronounced if a non-hydrolyzable ATP analog, AMP-PNP, was substituted for ATP in the assay (Fig. 4). In the presence of AMP-PNP, granule binding was permitted or even enhanced, but motility was abolished. Thus, ATP hydrolysis, and not simply ATP binding, is necessary for granule movement. The effects of AMP-PNP were reversible; perfusion of ATP into an AMP-PNP-arrested assay restored motility after a lag of several seconds. This result is consistent with motor-mediated movement in other systems (Lasek and Brady, 1985; Vale et al., 1985c). GTP could substitute for ATP in supporting granule motility. The numbers of granules that bound to microtubules and moved in the presence of GTP were very similar to those with ATP (Fig. 4), but the rate of granule movement was consistently slower with GTP (data not shown). This finding is consistent with kinesinmediated granule motility, since the ability to utilize GTP for force production is a characteristic of kinesin (Warner and McIntosh, 1989). The slowed rate of granule movement is expected, since kinesin hydrolyses GTP more slowly that ATP (Kuznetsov and Gelfand, 1986).
Virtually every granule can interact with microtubules
Our standard motility assay necessarily samples only a portion of the lytic granules, i.e. that population near enough to the coverslip to bind to the attached microtubules. Since at any given time the majority of granules in the chamber are not bound to microtubules, we wished to determine what proportion of granules was capable of interacting with microtubules in vitro. Using a sample prepared as for the standard assay, unbound granules diffusing above the plane of the coverslip were optically trapped with a single-beam gradient laser trap (Ashkin, 1992), and brought into apposition with microtubules. As shown in Table 1, 90% of the randomly sampled granules bound to microtubules when given an opportunity. Most of the organelles bound tightly, as determined by our inability to pull them away from the microtubule using the laser trap. Of those granules that bound, 40% translocated along the microtubule within a few seconds. This level of motility is very comparable to that observed in our standard assay (e.g. see Fig. 4, ATP). These results demonstrate conclusively that the vast majority of lytic granules isolated from CTL can interact directly with microtubules.
Integrity of the granule membrane is required for microtubule binding
To determine whether an intact granule membrane is important for motility, granules were incubated with Triton X-100 for 10 minutes at 4°C, and their movement on microtubules was assessed using the laser trap assay. Incubation of the granules with as little as 0.02% Triton X-100 was sufficient to completely abolish granule motility (data not shown). Limited proteolysis of the granule preparation was used to test whether the interaction with microtubules requires the participation of granule membrane proteins. The granules were incubated with trypsin as described in Materials and methods, and the enzyme was inactivated prior to setting up the assay. Trypsin treatment greatly diminished granule binding, and the few granules that did bind, did not move (Table 1). This inhibition was due to proteolysis of the granule membranes, since control granules that were incubated with previously inactivated trypsin bound and moved at near-normal frequencies. Similar results were obtained if granules were treated with chymotrypsin or proteinase K (data not shown).
As an independent method of disrupting interactions of granule membrane proteins with microtubule motors, granules were preincubated with a polyspecific rabbit anti-granule antiserum (Reynolds et al., 1987). At a dilution of 1:100, this antiserum reduced granule binding in the laser trap assay to 25% of control levels. Preincubation of granules with a 1:30 dilution of the antiserum completely abolished granule binding. In contrast, a control antiserum at the same dilution showed minimal inhibition. Taken together, these results show that the interaction of granules with microtubules and microtubule motors requires an intact and accessible granule membrane.
Most granule movements are plus-end-directed
In order to determine the direction of granule movement, we analyzed the movement of granules under conditions where the polarity of the microtubules was known. Microtubules were grown from axonemal seeds under conditions where growth at the plus-end is favored (Vale and Toyoshima, 1988), and used in an assay containing standard proportions of granules, ATP and cytosolic motors. Since the axonemes were sparse, granules were applied to the microtubules using the laser trap. In each of 4 cases where granules bound to the axonemal microtubules, the granule moved toward the free (plus) end of the microtubule (Fig. 5).
For reasons which we do not yet understand, granule binding to the axonemal microtubules was always poor, making this assay unsuitable for quantitative analysis. We therefore adapted our standard motility assay to allow quantitation of granule direction on randomly oriented microtubules. The standard motility assay was performed, except that carboxylated beads coated with purified kinesin were included as an internal standard for plus-end-directed movement. The polarity of individual microtubules was assigned, on the basis of the movement of the beads, and the direction of granule movements on the microtubules was then scored. To ensure that bead movement was kinesin-mediated, the beads were coated under conditions that minimize further binding of cytosolic motors, as described in Materials and methods. We estimate that no more than 5% of beads moved toward the microtubule minus-ends. Wherever possible, the polarity of a given microtubule was assigned, based upon the movement of multiple beads. Of the 29 granules scored, 25 (86%) moved in the same direction as the kinesin-coated beads (toward the microtubule plus-end). Of the four granules that appeared to move in the opposite direction, two were scored on the basis of a single bead movement. Thus, 86% is a lower estimate for plus-end-directed granule movement in this assay. This result using randomly oriented microtubules agrees well with the data obtained using the axoneme assay. It is important to note that the cytosolic extract used in these assays contains both plus-end- and minus-end-directed motors (see Figs 6A and 7A) and that it supports bidirectional movement of mixed organelles (Schroer et al., 1989, and data not shown). Therefore, the granules must preferentially utilize a plus-end-directed motor selected from the mixture.
Granule motility requires kinesin, but not cytoplasmic dynein
On the basis of the directionality data, it seems likely that cytoplasmic dynein, the major minus-end-directed motor in the CEF cytosol, contributes little to granule motility in vitro. To assess the contribution of cytoplasmic dynein, the cytosolic extract and the granules were incubated with 100 μM sodium vanadate and irradiated with ultraviolet light, a procedure which covalently cleaves dyneins at their nucleotide binding sites (Schroer et al., 1989). To ensure that the UV-vanadate treatment successfully cleaved the cytoplasmic dynein, the cytosol used in these experiments was immunoblotted with an antibody that recognizes dynein heavy chain. As expected, the uncleaved dynein heavy chain migrated as a polypeptide of Mr ≥ 400 × 103 (Fig. 6A, S3). After UV-treatment, it shifted to a mobility of 230 × 103Mr (Fig. 6A, S3 + UV-VO4), the so-called heavy UV-fragment (the antibody employed does not detect the 200 × 103Mr light UV-fragment). The ability of treated granules to bind and move in the presence of treated motors was then assessed using the random microtubule assay. As shown in Fig. 6B, granule binding and motility proceeded normally using the cleaved motor preparation. UV-vanadate treatment of granules also had no effect. Since dynein cleaved in this way is inactive as a motor, we conclude that lytic granule motility in vitro is largely independent of functional cytoplasmic dynein.
The major plus-end-directed motor in the CEF cytosolic extract is kinesin (Dabora and Sheetz, 1988; Schroer et al., 1989). To ask whether granule motility requires kinesin, we used an anti-kinesin antibody coupled to Sepharose to immunodeplete kinesin from the cytosol. As a control for the effects of the immunodepletion procedure itself, Protein A/Sepharose was used in parallel. As shown in Fig. 7A, kinesin was specifically removed by the SUK4 anti-kinesin Sepharose, but not by the Protein A/Sepharose. Fig. 7B shows the effects of kinesin depletion on granule motility. In the presence of the kinesin-depleted motor preparation, granule binding was reduced by 50%, and the number of motile organelles was proportionately decreased. The effect was specific to the anti-kinesin affinity resin, since the cytosol incubated with Protein A/Sepharose supported normal levels of granule binding. A small but significant number of granules continued to bind and move along microtubules in the presence of kinesin-depleted cytosol. These granules could represent a subset of organelles that utilize a second motor protein. More likely, this result is due to small amounts of residual kinesin, either on the granule membranes or in the motor preparation. Traces of kinesin are still detectable in the cytosol (Fig. 7A), even after three rounds of immunodepletion.
To determine whether purified motor proteins are sufficient to support granule movement, kinesin and cytoplasmic dynein isolated from embryonic chick brain and kinesin isolated from squid optic lobe were used in place of cytosol in the standard motility assay. In some cases, casein was added as a stabilizing carrier protein. Although each of the motor preparations was active as judged by its ability to support microtubule gliding, none could support granulemicrotubule interactions (data not shown). This result was expected for cytoplasmic dynein, based on the results of UV-inactivation. In the case of kinesin, this result indicates that other factors are required, in addition to kinesin, to support granule movement. As shown in Fig. 7B, purified chick brain kinesin was also unable to support motility if it was added back to the kinesin-depleted cytosol. This result suggests that the missing co-factor is a kinesin binding protein that is removed from the cytosol along with kinesin.
Taken together with the directionality data, our results with motor inactivation and depletion indicate that granule motility in vitro depends on kinesin and another, as yet unidentified, activating factor.
DISCUSSION
We have reconstituted in vitro the microtubule-mediated movement of lytic granules purified from CTL. The behavior of the granules in vitro is remarkable homogeneous, and is consistent with their behavior in vivo. In the in vitro assay, nearly every granule could interact with microtubules and virtually all movement was toward the plus-ends of microtubules. This directional preference is in keeping with the secretory nature of lytic granules, since in intact T cells, microtubules are oriented with their minus-ends at the MTOC and their plus-ends at the cell surface (Bergen et al., 1980). Granule movement in vitro required ATP, as does the lytic cycle in intact CTL (Roder et al., 1980). Moreover, the speed of granule movement in vitro (about 1 μm/second) is sufficient to account for the release of granules within a few seconds, well within the observed range for delivery of the “lethal hit” in cytolysis. Thus, the properties of granule movement in vitro agree well with the function of lytic granules in vivo.
Directional specificity by selective motor use
It has long been assumed that organelles of different types move with directional specificity on microtubules; i.e. secretory granules move towards microtubule plus-ends while lysosomes move toward microtubule minus-ends. Directional preference has been demonstrated on a population level in axons, where secretory vesicles and endocytic organelles accumulate on opposite sides of a ligation (Smith, 1980; Tsukita and Ishikawa, 1980). However, to our knowledge this is the first study to show that a homogeneous population of organelles exhibits directional specificity in vitro. We now show that purified lytic granules move preferentially toward the microtubule plus-ends, and that they do so by selectively utilizing kinesin.
Several pieces of evidence indicate that granule motility in this assay is driven almost exclusively by kinesin. First, the vast majority of granule movements are toward the microtubule plus-end, as expected for kinesin-mediated motility (Vale et al., 1985b; Porter et al., 1987). Second, GTP can substitute for ATP in supporting granule motility, and kinesin, unlike the dynein family of motors, can utilize GTP for force production (Warner and McIntosh, 1989). Third, motility is not affected by specific inactivation of dynein-like motors (Gibbons et al., 1987; Lye et al., 1987). Finally, immuno-depletion of kinesin from the cytosolic motor preparation substantially inhibits granule motility.
Urrutia et al. (1991) recently showed that in the presence of purified kinesin, disrupted chromaffin granule membranes can move on microtubules in vitro. Our work extends these findings by showing that lytic granules also utilize kinesin, and that the granules selectively utilize kinesin when presented with a mixture of microtubule motors. The fibroblast cytosolic extract used in the motility assay contains a mixture of two oppositely directed motor proteins, kinesin and cytoplasmic dynein. Both motors are active, as indicated by the ability of the extract to support bidirectional movement of mixed organelle preparations. Therefore, the predominantly plus-enddirected movement we observed cannot be explained by the properties of the motor preparation alone. Indeed, if the extract has a tendency to predispose directional movement, it is toward the minus-end, since Schroer et al. (1989). observed 90% minus-end-directed movement of mixed organelles using the same preparation. Thus, some property of the lytic granules themselves dictates their preferential use of kinesin in this assay. The granule membrane undoubtedly contributes to this specificity, since binding and motility are abolished by treatment of the granules with protease, detergent, or a polyspecific anti-granule antiserum.
Association of kinesin with granule membranes
As motility systems using well-characterized organelles become available, it will be important to understand how motor proteins are partitioned between membrane-bound and soluble pools, and between active and inactive forms. Our findings indicate that the lytic granules prepared by Percoll density centrifugation do not bear sufficient levels of active motor proteins to mediate their motility in the absence of exogenous cytosol. During the isolation procedure, the granules were not exposed to high salt or to other conditions that would release peripherally associated proteins such as kinesin. We therefore think it likely that the granules are not decorated with active motors in the resting T cell. In support of this conclusion, immunofluorescence microscopy of T cells stained with anti-kinesin antibodies does not reveal a granular distribution (J. Burkhardt and S. Hester, unpublished results). As with chromaffin granules (Urrutia et al., 1991) and mixed organelle preparations from CEF cells (Dabora and Sheetz, 1988; Schroer et al., 1989), it appears that the major pool of active motor proteins is soluble, with a relatively minor fraction bound to organelle membranes. Perhaps kinesin becomes bound to the granule membrane only after the cell is stimulated to secrete. This hypothesis can be readily addressed using the CTL system, since these cells can be stimulated to degranulate in a variety of ways.
Evidence for kinesin-regulatory factors
Our results indicate that kinesin is necessary, but not sufficient, for lytic granule motility. When added in place of the cytosolic extract, purified kinesin did not support granule movement. Purified kinesin also failed to reconstitute granule motility when added back to the kinesin-depleted cytosol. Similar results were obtained by Schroer et al. (1988) using a mixed organelle preparation. There are two possible interpretations of these findings. First, granule motility may require an additional activating factor which is removed from the cytosol during immunodepletion with anti-kinesin, and which is not supplied with highly purified kinesin. This would be analogous to the requirement of purified cytoplasmic dynein for the activating factor dynactin (Gill et al., 1991; Schroer and Sheetz, 1991). Indeed, one study indicates that a single activating factor may stimulate both kinesin and cytoplasmic dynein, and that this factor partially co-purifies with kinesin (Schroer and Sheetz, 1991). An alternate possibility is that post-translational modification of kinesin regulates its interaction with organelles, and the purification procedure that we employ does not preserve the active kinesin species. Phosphorylation is a likely candidate, since it has been recently shown that kinesin heavy and light chains are phosphorylated in vivo (Farshori and Goode, 1991; Hollenbeck, 1991). These two possibilities are not mutually exclusive, and studies are underway to test their relative importance for granule-microtubule interactions.
Regulation of granule-microtubule interactions in vivo
In the intact T cell, lytic granules probably interact with microtubules in multiple ways depending on the physiological state of the cell. Prior to 1982; Geiger et al., 1982; Kupfer et al., 1985; McKinnon et al., 1988). Similar clustering of late endocytic organelles has been attributed to the action of microtubule binding proteins (Mithieux and Rousset, 1989; Kreis, 1990) and minus-end-directed microtubule motors (Matteoni and Kreis, 1987; Bomsel et al., 1990). The granules closely resemble these endocytic organelles in function and composition (Burkhardt et al., 1990), and may share common protein machinery for maintaining their resting distribution. Since all these organelles are acidic (Mellman et al., 1986; Burkhardt et al., 1990), one attractive possibility is that interaction with minus-end-directed motors is linked to acidification. Consistent with this idea, treatment of intact CTLs with the protonophore CCCP neutralizes granule pH and disrupts granule reorientation during cytolysis (McKinnon et al., 1988).
When a T cell is stimulated by binding to a target cell, the MTOC and the clustered granules reorient to face the target, and the granules move to the cell surface for secretion (Kupfer et al., 1985; Yannelli et al., 1986; Lye et al., 1987). Presumably, granule-microtubule interactions must change from a state where quiescent granules cluster at the minus-ends of microtubules to one where they rapidly move toward the microtubule plus-ends. This dramatic shift is likely to be regulated by signal transduction events that occur after engagement of the T cell receptor, including an activation of protein kinases and a transient rise in cytosolic free calcium (Weiss et al., 1986; Hsi et al., 1989). As discussed above, phosphorylation of kinesin or its accessory factors may be important for activating movement of granules toward the cell surface. Proteins responsible for clustering of granules at the MTOC may also be targets for regulation, since at least one microtubule binding protein is released from microtubules by phosphorylation (Rickard and Kreis, 1991). The increase in cytosolic calcium levels could also stimulate granule motility. Haverstick et al. (1991) have shown that treatment of CTL with calcium ionophore induces cytoplasmic redistribution of granules, and our own preliminary data indicate that granule motility is stimulated by small increases in calcium concentration. Thus, taken together, the effects of increased kinase activity and elevated free calcium are excellent candidates for regulators of granule motility in vivo.
It is important to point out that the behavior of granules in the presence of CEF cytosol resembles that expected for granules in CTL that have been stimulated to secrete. Since the CTL were not stimulated prior to lysis, we assume that this effect is due to the use of CEF cytosol, which must somehow mimic the cytosol of stimulated CTL. For unknown reasons, CEF cells are an exceptionally rich source of active motors, superior to other cell types, including other fibroblasts and lymphocytes. Nonetheless, we have recently found that cytosol from CTL and T cell hybridomas can also support granule motility. As with the CEF cytosol, granule movement is constitutive, suggesting that regulatory factors that inhibit granule motility (or that direct granules towards the centriolar region) are missing from the reconstituted system.
While the CEF extract supports high levels of motility of all organelle preparations tested, not all organelles behave like the dense granules in this assay. For example, when the fractions of the Percoll gradient which contain light granules and other organelles are assayed in the presence of the CEF cytosol, the organelles move in a bidirectional fashion, and ER-like networks form with time. It therefore seems that the motile behavior of the dense granules is produced by specific interaction of cytosolic factors (determining the secretory state of activity) and granule membrane components (determining the direction of movement).
The regulation of granule-microtubule interactions is likely to be a complex process involving a variety of cellular factors. The relative ease with which lytic granules can be isolated and the wealth of information about signal transduction in lymphocytes makes this system well suited for determining how granule motility is regulated in vivo.
ACKNOWLEDGEMENTS
We thank P. Henkart and D. Meyer for gifts of antibodies, R. Kurlander and D. Howell for help with culturing CTL, and D. McClay for providing sea urchin sperm. Purified kinesin and cytoplasmic dynein were generously provided by S. Kuo, S. HammAlvarez and C. Martenson. S. Hester and R. Phang provided valuable assistance with electron microscopy and video analysis. Finally, we thank D.B. Amos and members of the Sheetz and Argon laboratories for helpful advice and criticism. This work was supported in part by grants from the American Cancer Society to Y.A. and the National Institutes of Health (GM36277) to M.P.S., and J.K.B. is a postdoctoral fellow of the Irvington Institute for Medical Research.