Apoptotic Cleavage of Cytoplasmic Dynein Intermediate Chain and P150 ^Glued Stops Dynein-Dependent Membrane Motility

Apoptosis, or programmed cell death, is accompanied by the activation of a family of apoptosis-specific cysteine proteases called caspases (Cohen 1997). Both the signaling pathways that mediate caspase activation, and the substrates for caspases, are highly conserved. The so-called execution phase of apoptosis involves a characteristic series of morphological changes that in many cases are directly coupled to caspase-dependent cleavage of key substrates. These events ultimately lead to the recognition and engulfment of dying cells by their neighbors. The nature of the phagocytic signal generated by the apoptotic cell is not known, but is likely to be complex. This is perhaps best illustrated by the finding that “professional” and “nonprofessional” phagocytic cells clear apoptotic cells at different rates, with the latter recognizing apoptotic cells only after the execution phase has run its full course (Parnaik et al. 2000).

The exposure of phosphatidyl serine on the surface of apoptotic cells is thought to be a key phagocytic signal (Martin et al. 1995; Fadok et al. 2000; Hamon et al. 2000), but other markers are clearly important (Savill et al. 1990; Ren et al. 1995; Platt et al. 1996; Devitt et al. 1998). Indeed, since such widely recognized phagocytic signals must be generic, it is probable that general alterations in the protein composition of the plasma membrane, or in the glycosylation of proteins or lipids, are also involved. Such changes could be brought about by altering the specificity or efficiency of membrane traffic pathways. In support of this hypothesis, there is evidence that both the secretory and endocytic pathways are affected in apoptotic cells. There are profound changes in membrane structure (Wyllie et al. 1980), including fragmentation of the Golgi complex (Mancini et al. 2000) and inhibition of the secretory pathway (Whyte et al. 1993). In addition, homotypic endosome fusion is inhibited in apoptotic cell extracts via the caspase 3–dependent cleavage of Rabaptin 5 (Cosulich et al. 1997; Swanton et al. 1999a).

The secretory and endocytic pathways rely in numerous ways on the function of cytoplasmic dynein and its regulator dynactin. They are required for ER-to-Golgi region traffic (Presley et al. 1997), for maintaining Golgi apparatus organization and position (Burkhardt et al. 1997; Harada et al. 1998; Quintyne et al. 1999; Roghi and Allan 1999), and for moving endocytic organelles and facilitating traffic between them (Blocker et al. 1997; Harada et al. 1998; Valetti et al. 1999). The ER itself is a cargo for cytoplasmic dynein in extracts of Xenopus eggs and early embryos (Allan 1995; Lane and Allan 1999). In addition to these roles in trafficking, cytoplasmic dynein, and dynactin are also needed for maintaining microtubule organization in interphase, as well as for correct spindle assembly, positioning, and chromosome attachment during cell division (for review see Karki and Holzbaur 1999).

The central importance of cytoplasmic dynein in maintaining cellular architecture makes it an attractive target for inactivation during apoptosis. Cytoplasmic dynein is a large molecule (∼1.2 MDa) which consists of two heavy chains (CD-HC), two or three intermediate chains (CD-IC), four light intermediate chains (CD-LIC), and a variety of light chains (for review see Susalka et al. 2000). More than one gene exists for the heavy, intermediate, and light intermediate chains, and in addition, the CD-ICs and the CD-LICs undergo tissue-specific alternate splicing (Susalka et al. 2000). Whether different subsets of these chains associate to give cytoplasmic dynein molecules with distinct function is not clear, but seems likely (Susalka et al. 2000).

Cytoplasmic dynein function generally requires dynactin, first identified as an activator of cytoplasmic dynein–driven vesicle movement (Gill et al. 1991). Dynactin also consists of multiple subunits, including two p150^Glued chains which extend out from a short filament of actin-related protein 1 which associates with a variety of other subunits, including several dynamitin molecules (Schafer et al. 1994; Quintyne et al. 1999). Genetic and biochemical studies have confirmed that the two complexes must interact for virtually all cytoplasmic dynein functions (for review see Allan 1996; Schroer 1996; Karki and Holzbaur 1999). Dynactin is thought to link cytoplasmic dynein to its cargoes (for review see Allan 1996; Schroer 1996; Karki and Holzbaur 1999) and to enhance cytoplasmic dynein's processivity (King and Schroer 1999). How dynactin attaches to cargoes is not clear, but it might bind directly to membrane lipids, or interact with proteins such as beta spectrin on the Golgi apparatus or ZW10 on the kinetochore (Karki and Holzbaur 1999; Muresan et al. 2001). Cytoplasmic dynein then binds to dynactin via an interaction between p150^Glued and the NH₂-terminal domain of CD-IC (Karki and Holzbaur 1995; Vaughan and Vallee 1995).

A convenient system for studying cytoplasmic dynein function during apoptosis is provided by Xenopus egg extracts, which support active cytoplasmic dynein–driven ER movement (Allan 1995; Niclas et al. 1996; Lane and Allan 1999) and can readily be made apoptotic (Kluck et al. 1997). Here, we show that CD-IC and p150^Glued are cleaved by caspases both in apoptotic egg extracts and during apoptosis in vivo. The implications of these cleavage events for cytoplasmic dynein–dynactin function are described.

Unless otherwise stated, chemicals were obtained from Sigma-Aldrich. Stock solutions were stored at −20°C: anisomycin (5 mg/ml in DMSO), etoposide (50 mM in DMSO), and caspase inhibitors (Ac-DEVD.CHO at 100 μM and zVAD.FMK at 50 μM, both in DMSO; Calbiochem-Novabiochem). Protease inhibitors (leupeptin, chymostatin, pepstatin, and aprotinin) were used at 10 μg/ml final concentration.

For immunoblotting, we used the following monoclonal antibodies: anti–CD-IC (70.1, Sigma-Aldrich; 1618, Chemicon International), anti-p150^Glued (Transduction Labs), antifodrin (ICN Biomedicals), antitubulin (B-5-1-2), antiribophorin (CEL5C, from Birgit Lane, University of Dundee, Dundee, UK), antikinesin II (K 2.4, from John Scholey, University of California at Davis, Davis, CA), and anti-p50 dynamitin (from Richard Vallee, University of Massachusetts, Worcester, MA); and the following polyclonal antibodies: anti-rat CD-IC (from Richard Vallee; Vaughan and Vallee 1995), anti-Xenopus CD-LIC (Addinall et al. 2001), anti–CD-IC NH₂ terminus (Xenopus-specific; Lane and Allan 1999), anti-Xenopus CD-IC peptide antibody (raised and affinity purified against the sequence NRSNKRTPVQRTPLS; underlined in Fig. 3), anti–poly-ADP-ribose polymerase (PARP; Calbiochem), antikinesin heavy chain (KHC; from Ron Vale, University of California at San Francisco, San Francisco, CA), anti–β-COP (D1; from Thomas Kreis, University of Geneva, Geneva, Switzerland), and anti–Arp-1 (centractin; from David Meyer, University of California at Los Angeles, Los Angeles, CA). Secondary antibodies (conjugated to alkaline phosphatase or horse radish peroxidase) were from Jackson ImmunoResearch Laboratories.

HL60 cells were grown in RPMI 1640 plus 10% foetal bovine serum and penicillin/streptomycin. To analyze apoptosis in vivo, cells were treated with 5 μg/ml anisomycin or 50 μM etoposide, with or without a 30 min preincubation with 50 μM zVAD-FMK. Cells were washed three times in PBS, then broken by passing 20 times through a cell cracker (Balch et al. 1984) in acetate/sucrose buffer (100 mM K-acetate, 3 mM Mg-acetate, 5 mM EGTA, 10 mM Hepes, pH 7.4, 150 mM sucrose) containing protease inhibitors. Postnuclear supernatants were prepared and stored at −80°C.

Interphase Xenopus egg extracts were prepared as described (Murray 1991). Apoptotic or control extracts were generated by the addition of cytochrome c (to 10 μM) ± the caspase inhibitor, Ac-DEVD-CHO (2 μM), followed by incubation for 8 h at 25°C. Cytosol and membranes were prepared from control and apoptotic extracts, and membranes were further purified by flotation (Lane and Allan 1999). For biochemical analysis of rebinding to membrane, isolated membranes (10 μl 5 mg/ml protein) were incubated in 90 μl control or apoptotic cytosol at 25°C for 60 min. The mixture was then centrifuged in a rotor (TLA100; Beckman Coulter) for 15 min at 40,000 g_av at 4°C through a 150-μl cushion of 0.75 M sucrose in acetate buffer. The pelleted membranes were analyzed by immunoblotting.

Cytoplasmic dynein and dynactin were isolated from 400 μl control or apoptotic cytosols by microtubule affinity and sucrose density gradient centrifugation as described (Niclas et al. 1996). 50-μl fractions were collected and run on 4–15% gradient acrylamide gels then analyzed by silver staining and immunoblotting. Immunoprecipitation experiments were performed as described (Roghi and Allan 1999). In brief, 20 μl of control or apoptotic egg cytosol was diluted to 500 μl in immunoprecipitation buffer then incubated with protein G Sepharose beads (Zymed Laboratories) prebound with anti–CD-IC (1618) or anti–CD-LIC antibodies.

Motility assays were performed using video-enhanced differential interference contrast (VE-DIC) microscopy using a light microscope (BX60; Olympus) (Allan 1998). To compensate for differences in microtubule assembly between control and apoptotic cytosols, a uniform field of stable microtubules was assembled. Flow cells were perfused with diluted control egg cytosol (1:20 in acetate/sucrose) for 5 min. After two washes with 10 μl acetate/sucrose, taxol-stabilized microtubule seeds (Allan and Vale 1991) were added. After 10 min, soluble tubulin (5 mg/ml) plus 1 mM GTP was introduced, and microtubules were allowed to grow for 15 min. Microtubules were stabilized by flowing through 20 μM taxol in acetate/sucrose, washed once with acetate/sucrose, and then a mix of cytosol and membranes was added. Over time, ER networks assemble on the coverslip surface in a cytoplasmic dynein–dependent fashion. Network formation was quantitated after 45 min by counting the number of three way junctions in 20 random fields, which correlate directly with the quantity of cytoplasmic dynein–based movement (Allan 1995). Motility experiments were carried out in the presence of 2 μM Ac-DEVD-CHO to prevent further caspase action.

The first 288 bases of Xenopus CD-IC were obtained from a frog oocyte cDNA library (Nicolás et al. 1997) by degenerate PCR using the following primers: forward, GGGGATCCATGTC(I)GA(CT)AA(AG)AG-(CT)GA(I)(CT)T(I)AA; reverse, GGAATTC(AG)TC(CT)TG(AG)C-T(I)CC(I)GC(AG)CT(I)GG. The PCR product was cloned into pBluescript (pBSIIsk+; Stratagene), and was used to prepare a probe to screen the oocyte cDNA library to obtain a full-length Xenopus CD-IC clone. A single clone was isolated, the sequence of which was most similar to CD-IC2B (these sequence data are available from EMBL/GenBank/DDBJ under accession nos. AF319780 and AF319781). The clone lacked the first 90 bases from the 5′ end, so to obtain a full-length construct it was joined to the original PCR product using a conserved NsiI site within the overlapping region. Site-directed mutagenesis was performed using the QuikChange kit (Stratagene). All mutations were confirmed by sequencing.

The full-length Xenopus CD-IC construct was subcloned into the expression vector, pCDNA3.1 (Invitrogen). [³⁵S]methionine-labeled protein was produced using a coupled transcription/translation kit (Promega). Translations were analyzed by autoradiography. Translated Xenopus CD-IC appeared as several 60–85-kD bands (data not shown), which were probably derived from random initiation events, so CD-IC was routinely immunoprecipitated using an NH₂-terminal anti–CD-IC antibody (1618). The immunecomplex was then used in caspase cleavage assays. Recombinant PARP and caspase 2 were prepared as described (Swanton et al. 1999b).

Assays using recombinant caspases 2, 3, and 7 (Alexis Biochemicals; and from Don Nicholson and Sophie Roy [Merk Frosst Centre for Therapeutic Research, Quebec, Canada]) were performed in caspase buffer (for caspases 3 and 7; 50 mM Hepes-KOH, pH 7.0, 2 mM EGTA, 0.1% Chaps, 10% sucrose, 5 mM DTT; for caspase 2, 50 mM Na-citrate, pH 5.5 replaced the Hepes-KOH). Fractions enriched in native Xenopus or pig brain cytoplasmic dynein–dynactin were incubated with caspases for 3 h at 37°C and analyzed by silver staining and immunoblotting.

Cleavage of in vitro–translated, immunoprecipitated, [³⁵S]methionine-labeled Xenopus CD-IC was carried out either by incubating beads in control or apoptotic egg cytosol for 6 h at 25°C, or by incubating beads in the presence of recombinant caspases for 2 h at 30°C. Radiolabeled products were detected by autoradiography. For some experiments the caspase 3 and 7 inhibitor, Casputin™ (Biomol Research Laboratories, Inc.; 80 μg/ml final) was included.

To investigate whether the inactivation of microtuble motors might contribute to the changes in cell morphology associated with apoptosis, we assessed the stability of a variety of molecules after activation of caspases in intact cells. We used a number of cell lines (including human HL60, jurkat and A431, and rat NRK) treated with a selection of drugs (anisomycin, etoposide, staurosporine, or camptothecin). Most molecules tested (namely KHC, kinesin II, p50 dynamitin, β-COP, and tubulin [Fig. 1 A]) were not altered during apoptosis. However, loss of immunoreactivity of both CD-IC and the p150^Glued component of the dynactin complex was seen in apoptotic cells (Fig. 1A and Fig. B). In HL60 cells undergoing apoptosis in response to 5 μg/ml anisomycin or 50 μM etoposide (Fig. 1 B), CD-IC rapidly became undetectable at about the same rate that the characterized caspase substrate PARP was cleaved. This suggests that CD-IC may be an early target for caspases. The reduction in p150^Glued signal was more gradual, mirroring the loss of full-length fodrin immunoreactivity (Fig. 1 B). In each case, loss of immunoreactivity was greatly reduced by the cell permeable caspase inhibitor zVAD-FMK.

To study how caspase-dependent cleavage of CD-IC and p150^Glued affect the function of the cytoplasmic dynein–dynactin complex, we examined the cleavage of both proteins in vitro using Xenopus egg extracts, since they are well-suited for studying motor proteins (Allan and Vale 1991) and apoptosis (Newmeyer et al. 1994). Apoptosis can be triggered in cell-free extracts by the addition of cytochrome c (Liu et al. 1996; Kluck et al. 1997), whereupon effector caspases are activated (e.g., Swanton et al. 1999b). A clear reduction in recognition of CD-IC by NH₂-terminal monoclonal antibodies was seen upon addition of cytochrome c to Xenopus egg extracts, but no cleavage products were detected (Fig. 1 C). This loss of signal occurred over a similar time scale to the cleavage of other caspase substrates in Xenopus extracts (Newmeyer et al. 1994; Cosulich et al. 1996) and was prevented by the caspase inhibitor Ac-DEVD-CHO (Fig. 1 C). In contrast, the monoclonal antibody to p150^Glued detected a cleavage product in apoptotic egg extracts, ∼10–20 kD smaller than the native molecule (Fig. 1 C); no such product was seen in HL60 extracts, although the p150^Glued antibody reactivity was lost with time (Fig. 1A and Fig. B). We do not yet know why the pattern of p150^Glued cleavage differs in these two systems.

To drive the movement of a variety of cargoes towards the minus ends of microtubules, cytoplasmic dynein interacts with dynactin through binding of the NH₂ terminus of CD-IC to p150^Glued (Fig. 2A and Fig. B; Karki and Holzbaur 1995; Vaughan and Vallee 1995). This region of CD-IC is recognized by both monoclonal antibodies 70.1 and 1618 (data not shown). Since neither antibody recognized cleavage products in apoptotic cells and extracts (Fig. 1), this suggested that caspase-dependent cleavage occurs within this region. To test this hypothesis, we used a polyclonal antiserum raised against full-length rat CD-IC (Vaughan and Vallee 1995), and a peptide antibody raised against a sequence towards the COOH-terminus of CD-IC (see Materials and Methods; Fig. 3). In immunoblots of apoptotic Xenopus egg extracts, these antibodies revealed a cleavage product ∼12 kD smaller than the molecular mass of the full-length molecule (Fig. 2 C). These data suggest that CD-IC is cleaved by caspases within its NH₂-terminal p150^Glued binding domain.

Comparison of multiple CD-IC sequences revealed several conserved aspartic acids towards the NH₂ terminus of the molecule (Fig. 3). To identify the caspase cleavage site, we used site-directed mutagenesis to alter these aspartic acids to alanine, and examined the cleavage of in vitro–translated, [³⁵S]methionine-labeled protein when added to apoptotic egg cytosol. Two cleaved products were generated from in vitro–translated wild-type Xenopus CD-IC (Fig. 4 A): one at ∼74 kD (C₁) and one at ∼66 kD (C₂). C₁ matches the size of the product seen in immunoblots of apoptotic egg extracts using the polyclonal anti–CD-IC (Fig. 2 C), whereas we have never observed a band of the size of C₂, perhaps because the C₂ cleavage site is inaccessible when CD-IC is in the native cytoplasmic dynein complex. Mutating the putative site DSGD⁹⁹G (D→A) resulted in the loss of the upper cleavage product (C₁) (Fig. 4 A). None of the other mutations altered the pattern of cleavage of Xenopus CD-IC (data not shown).

Since DSGD⁹⁹G fits with consensus sites for caspase 3 (DXXD motif), we tested the effects of the selective caspase 3 and 7 inhibitor, Casputin™. This inhibitor prevented cleavage of in vitro–translated wild-type CD-IC to C₁ (Fig. 4 A), suggesting that this product is generated by caspase 3 or caspase 7 in apoptotic egg extracts. We then tested the ability of recombinant caspases 2, 3, and 7 to cleave in vitro–translated, immunoprecipitated CD-IC (Fig. 4, B–D). These caspases were shown to be active by incubating them with known substrates. All three caspases cleaved CD-IC to generate C₁, to varying degrees, with caspase 3 being most effective. From these data, caspase 3 is the best candidate for cleaving CD-IC at DSGD⁹⁹G.

To confirm that CD-IC within native cytoplasmic dynein could be cleaved by caspase 3, we carried out in vitro cleavage assays using recombinant caspases 2 and 3, and cytoplasmic dynein–dynactin fractions prepared from Xenopus eggs (Fig. 4 E). Using immunoblotting, we demonstrated that endogenous Xenopus CD-IC was degraded in vitro by caspase 3, but much less so by caspase 2 (Fig. 4 E). Caspase 3, but not 2, also cleaved some of the p150^Glued present within the same fraction (Fig. 4 E). The inclusion of Ac-DEVD-CHO in these assays prevented caspase-mediated CD-IC and p150^Glued cleavage. Identical results were obtained using pig brain cytoplasmic dynein–dynactin (data not shown).

Cytoplasmic dynein and dynactin sediment as separate complexes of ∼20S. To test if caspase cleavage of CD-IC and p150^Glued affected the stability of the complexes, we compared their sedimentation properties when isolated from control and apoptotic egg cytosol. Motors bind statically to microtubules in the absence of ATP, so cytoplasmic dynein can be enriched from cytosol relatively simply. Dynactin copurifies with cytoplasmic dynein, since the p150^Glued subunit also possesses a microtubule binding domain (between amino acids 39–150; Waterman-Storer et al. 1995). Silver staining and immunoblotting of sucrose gradients of cytoplasmic dynein–dynactin-enriched fractions from control egg cytosol revealed that CD-HC, CD-IC, and CD-LIC cosedimented at 20S (in fractions 5 and 6; Fig. 5 A). In fractions enriched from apoptotic cytosol, CD-HC and CD-LIC were still found at 20S, even though full-length CD-IC could not be detected using 70.1 (Fig. 5 A). Hence, cleavage of CD-IC does not appear to destabilize CD-HC and CD-LIC interactions.

We next asked whether the large COOH-terminal CD-IC fragment also remained in the complex using an immunoprecipitation approach (Fig. 5B and Fig. C). In many apoptotic egg extracts, cleavage of CD-IC is only ∼60–70% efficient (Fig. 5 B, lanes 1 and 2). When we immunoprecipitated using anti–CD-LIC antibodies, we found that the CD-IC cleavage product, intact CD-IC (Fig. 5 B, lane 4), and CD-HC (not shown) coprecipitated efficiently. We wanted to know whether dynein complexes existed which only contained the cleaved CD-IC, and to address this we first depleted dynein containing intact CD-IC by three rounds of immunoprecipitation with an NH₂-terminal monoclonal antibody (1618), which does not recognize cleaved CD-IC (Fig. 5 C). Interestingly, cleaved CD-IC did not coprecipitate with intact CD-IC (Fig. 5 C, lanes 3–6), but remained in the unbound fraction (Fig. 5 C, lanes 7 and 8). Moreover, when we used anti–CD-LIC beads to immunoprecipitate the remaining dynein from full-length CD-IC–depleted apoptotic cytosol, the immune complexes contained the cleaved form of CD-IC along with CD-HC (Fig. 5 C, lane 10). These experiments suggest that, in Xenopus egg cytosol, cleavage of CD-IC chains does not cause the dynein components to disassociate. Strikingly, the dynein complexes contain either all intact or all cleaved CD-ICs.

Sucrose density gradient analysis also provided information on the dynactin complex. Using the monoclonal anti-p150^Glued antibody, we found that in control cytosol p150^Glued sedimented at 20S as expected (Fig. 5 A). In apoptotic cytosol, we detected some full-length p150^Glued at 20S (Fig. 5 A), since cleavage of p150^Glued is incomplete (Fig. 1 C). Although some cleaved p150^Glued also migrated to this position (Fig. 5 A), the majority was detected at the top of the gradient, suggesting that it had dissociated from the dynactin complex (Fig. 5 A). The method for isolating cleaved (and intact) p150^Glued relies on microtubule affinity, suggesting that the truncated product contains a functional microtubule binding domain (although some may be complexed with the small amount of intact, free p150^Glued). Moreover, the monoclonal p150^Glued antibody used was raised against the NH₂-terminal microtubule binding domain of p150^Glued, so it seems likely that the caspase cleavage site(s) within Xenopus p150^Glued is located towards its COOH terminus.

According to current models, caspase-mediated cleavage of CD-IC and p150^Glued should uncouple cytoplasmic dynein from its cargo and therefore prevent movement along microtubules. We tested this hypothesis using two approaches. First, we determined whether cytoplasmic dynein and dynactin remained membrane-associated in egg extracts during apoptosis, and then we looked at the consequences of apoptotic CD-IC and p150^Glued cleavage on membrane motility using an in vitro assay. Floated Xenopus egg membranes are rich in both cytoplasmic dynein and dynactin (Fig. 6; Lane and Allan 1999). Although silver staining revealed very little difference in the overall protein pattern in apoptotic versus control cytosol or membranes (Fig. 6 A), the apoptotic membranes had lost most of their associated CD-LIC and CD-IC (Fig. 6 B). Since the cytoplasmic dynein complex remains intact during apoptosis (Fig. 5), it is likely that the entire complex is removed from membranes during this process.

We then looked at the dynactin complex. The amount of full-length and cleaved p150^Glued on apoptotic membranes was greatly reduced, and p50 dynamitin also appeared to be depleted (Fig. 6 B). Membrane-associated Arp-1 was partially degraded to an ∼32-kD product in apoptotic egg extracts, which remained associated with membranes (no degradation was observed in apoptotic cytosol; Fig. 6 B). We do not know if the Arp-1 degradation product is due solely to the action of caspases, because we always observed a similar band in immunoblots of control membranes (Fig. 6 B). Taken together, these results suggest that at least part of the dynactin complex stays on apoptotic membranes. We also compared KHC and kinesin II on control and apoptotic membranes (Fig. 6 B). Interestingly, while the amounts of kinesin II (p85 subunit) did not change, we found that KHC, although uncleaved (Fig. 1 A and 6 B), was lost from apoptotic egg membranes, an observation which merits further investigation.

ER membranes present in crude frog egg extracts move exclusively using cytoplasmic dynein in egg cytosol and are easily identified by VE-DIC microscopy (Allan 1995; Niclas et al. 1996; Lane and Allan 1999). Therefore, we used in vitro motility assays to assess cytoplasmic dynein–based membrane motility in control and apoptotic egg extracts. Extensive ER networks formed from membranes and cytosol derived from control extracts treated with cytochrome c and caspase inhibitors (Fig. 7 A, left). The motility of these networks was identical to that seen in untreated extracts. However, upon combining membranes and cytosol from apoptotic extracts, we saw no network formation (Fig. 7 A, right). Instead, membrane clumps rarely associated with microtubules (Fig. 7 A), and tubule extension events were never observed. We have not assessed whether ER membrane fusion is also inhibited, but this is a possibility given that homotypic endosome fusion is a target for apoptotic regulation (via caspase 3 cleavage of Rabaptin 5; Cosulich et al. 1997; Swanton et al. 1999a).

These data suggest that apoptotic membranes are unable to form networks in apoptotic cytosol because they have lost the capacity to associate with and move along microtubules in a cytoplasmic dynein–dependent fashion. However, when apoptotic membranes were incubated in control cytosol, they regained the ability to move along microtubules and undergo homotypic membrane fusion (Fig. 7 B, left), generating networks which were ∼75% as extensive as seen in nonapoptotic conditions. When we carried out the reciprocal experiment (control membranes in apoptotic cytosol), we found that the extent of network formation was similar (Fig. 7 B, right). All experiments were performed in the presence of caspase inhibitor to prevent further caspase action.

When the membranes from these incubations were reisolated and immunoblotted for cytoplasmic dynein and dynactin, we found that apoptotic membranes effectively recruited full-length CD-IC and CD-LIC, as well as p150^Glued, to ∼50% of control levels in just 1 h (Fig. 8), although under identical conditions KHC was not recruited (data not shown). Control membranes lost a proportion of their associated cytoplasmic dynein and dynactin when incubated in apoptotic cytosol for 1 h (Fig. 8). This probably reflects the dynamic nature of cytoplasmic dynein's affinity for membranes and is not due to caspase activity, because Ac-DEVD-CHO was included in these assays.

We next asked whether membrane movement in apoptotic extracts could be restored by the addition of purified cytoplasmic dynein–dynactin. We incubated apoptotic membranes and cytosol (plus Ac-DEVD-CHO) in the presence of pig brain cytoplasmic dynein–dynactin (Fig. 9). This fraction restored the network-forming ability of apoptotic membranes to ∼25% of control activity (Fig. 9 A). The addition of pig brain cytoplasmic dynein brought the amount of intact CD-IC present in the cytosol/membrane mixture to approximately control levels (Fig. 9 B). These results suggest that cytoplasmic dynein–driven ER motility is halted during apoptosis primarily through caspase-mediated cleavage of cytoplasmic dynein and dynactin components.

We have investigated the fate of cytoplasmic dynein during apoptosis, both in vivo and in vitro, as part of an ongoing study into alterations in membrane traffic and cellular organization. We have used Xenopus egg extracts as a well-characterized model system, as they support regulated cytoplasmic dynein–driven ER movement (Allan 1995; Niclas et al. 1996; Lane and Allan 1999) and can be driven into apoptosis by the addition of cytochrome c (Kluck et al. 1997).

We have identified CD-IC as a target for caspase-dependent cleavage both in vivo and in cell-free extracts, and have demonstrated by site-directed mutagenesis and the use of specific antibodies that cleavage results in the loss of the NH₂-terminal p150^Glued-binding domain. In contrast, CD-HC and CD-LIC are unaffected and remain as a complex containing the COOH-terminal cleavage product of CD-IC. Interestingly, we found that dynein complexes either contained only full length or only cleaved CD-IC, suggesting that once a dynein molecule is targeted by the caspase (probably caspase 3), all the CD-IC chains are cleaved.

The p150^Glued component of dynactin is also cleaved, and most of the cleaved form no longer migrates as part of the 20S dynactin complex on sucrose density gradients. Since cleaved p150^Glued seems to maintain its ability to bind to microtubules and is still recognized by an antibody to its NH₂-terminal domain, it is likely that the cleavage occurs in its COOH-terminal domain, through which it interacts with the rest of the dynactin complex (Waterman-Storer et al. 1995). This behavior is remarkably similar to that of the COOH-terminally truncated product of the dominant Glued mutant gene in Drosophila (McGrail et al. 1995). It will clearly be of interest to map the p150^Glued cleavage site in future studies. Since some dynamitin and Arp1 stay bound to the membrane, this may mean that the dynactin complex minus p150^Glued is left behind. However, lack of antibodies recognizing other Xenopus dynactin components has prevented a thorough investigation of this question.

The cleavage of these two key components, CD-IC and p150^Glued, would be predicted to destroy cytoplasmic dynein's ability to bind to and therefore move cargo (Fig. 10), and we have shown this to be the case for cytoplasmic dynein–driven ER motility in apoptotic Xenopus egg extracts. Interestingly, apoptotic membranes recruit intact cytoplasmic dynein and dynactin from control cytoplasm, and regain the ability to form ER networks. Moreover, the addition of a pig brain cytoplasmic dynein fraction (which also contains dynactin) to a mixture of apoptotic cytosol and membranes was sufficient to reconstitute network formation partially. Although the amount of pig brain dynein added matched that present in control cytosol, motility was not fully restored. There are many possible explanations for this: first, we do not know how much of this dynein is recruited to the membrane; second, the purified dynein fraction may be partially inactive, or take time to become activated when added to cytosol; third, the concentration of dynactin may be too low; and fourth, there may be other factors that are absent in apoptotic cytosol that are required to restore full motility. Together, however, our data suggest that cytoplasmic dynein and dynactin are primary targets for the caspase-dependent abrogation of ER tubule movement described here. Any cleavage of integral membrane proteins by caspases cannot be inhibiting cytoplasmic dynein binding in our assay, as such changes would not be reversible.

Our results clearly demonstrate that the interaction of cytoplasmic dynein and dynactin with the membrane is dynamic, and suggests that there is rapid turnover, at least in the presence of apoptotic cytosol. This is in keeping with other studies in Xenopus, which have revealed that both cytoplasmic dynein and dynactin are rapidly released from membranes on entry into metaphase (Niclas et al. 1996; Addinall et al. 2001), or when anti–CD-IC antibodies are added to interphase extracts (Steffen et al. 1997). Similar loss of cytoplasmic dynein from membranes occurred after the addition of anti-p150^Glued antibodies to squid axoplasm (Waterman-Storer et al. 1997) and under certain conditions in cultured mammalian cells (Lin et al. 1994). In addition, overexpression of the dynamitin subunit of dynactin causes loss of cytoplasmic dynein from kinetochores (Echeverri et al. 1996) and the Golgi apparatus (Roghi and Allan 1999), again demonstrating that cytoplasmic dynein–cargo interaction is dynamic.

What is the significance of these cleavage events to the apoptosing cell? As yet, we can only speculate, but it is clear that CD-IC in particular is an early target for cleavage, which proceeds in parallel with PARP cleavage. Given that inhibition of cytoplasmic dynein function disrupts ER-to-Golgi apparatus traffic (Presley et al. 1997) and leads to a scattering of the Golgi apparatus throughout the cell (Burkhardt et al. 1997; Quintyne et al. 1999; Roghi and Allan 1999), it is intriguing to note that Golgi apparatus fragmentation is seen in apoptotic cells (Mancini et al. 2000). It will be of great interest to know if the loss of cytoplasmic dynein from the Golgi apparatus contributes to this phenotype, and whether ER-to-Golgi apparatus traffic is also inhibited due to a lack of motor.

In this study, we have investigated the effects of apoptosis on membrane movement, but since cytoplasmic dynein plays many different roles in the cell, it is plausible that interrupting other cytoplasmic dynein–driven motile events may also be important for apoptosis. Another issue is that the cleavage site that we have identified in Xenopus CD-IC, DSGD⁹⁹G, is conserved in all rat CD-IC2 splice forms but not in rat CD-IC1. The cytoplasmic dynein intermediate chain family is encoded by two genes, of which CD-IC1 is expressed only in brain and testis, while CD-IC2 is found in all tissues (for review see Susalka et al. 2000). Therefore, all cells are likely to contain a cleavable CD-IC, while neurons and testis may also possess a form that lacks this caspase cleavage site. This may provide a means of differentially regulating the function of distinct cytoplasmic dynein complexes during apoptosis.

Cytoplasmic dynein has been proposed to play a direct role in an early, caspase-independent signaling event during apoptosis. The cytoplasmic dynein light chain, LC8, and the proapoptotic protein Bim (O'Connor et al. 1998) are stably associated with each other in healthy cells, and translocate to Bcl-2 shortly after triggering apoptosis (Pakthalakath et al. 1999). Hence, caspase-dependent inactivation of the cytoplasmic dynein–dynactin complex might provide a positive feed-back loop during activation of the apoptotic cascade. However, LC8 is also a component of flagellar dynein, myosin V, IκBα and nitric oxide synthase (for review see King 2000). In addition, the expression of LC8 is increased by cyclooxygenase 2 treatment of PC12 cells, where it is thought to protect from growth factor withdrawal-induced apoptosis by the inhibition of nitric oxide synthesis (Chang et al. 2000). Therefore, because of the apparent promiscuous behavior of LC8, and in the absence of definitive evidence that Bim is part of the native cytoplasmic dynein complex, it remains unclear whether the cleavage of CD-IC that we observe here will have any direct effects on the Bim-LC8 interaction.

Little else is known about the roles that microtubule motors play in apoptosis. An inhibition of the plus end–directed microtubule motor, kinesin, has been implicated in the clustering of mitochondria that occurs on triggering cell death via tumor necrosis factor (De Vos et al. 1998, De Vos et al. 2000), but as a shift in mitochondrial position did not occur after other apoptotic stimuli, it is not clear if this is a general phenomenon. Whether inactivation of kinesin is also responsible for Bax-induced mitochondrial clustering (Desagher and Martinou 2000) remains to be tested. Interestingly, our results show that kinesin, like dynein, was released from apoptotic membranes (Fig. 6), but it was not rerecruited from control cytosol (data not shown). Although KHC is not cleaved in apoptotic egg extracts, we do not yet know if kinesin light chains, which are thought to target kinesin to its cargoes (Hollenbeck 2001), are affected. In addition, since caspase-induced cleavage of kinectin has been seen in apoptotic cultured cells (Machleidt et al. 1998), this raises the possibility that kinectin is also cleaved on apoptotic membranes in our assay system, and that this prevents rerecruitment.

The identification of cytoplasmic dynein and dynactin as targets for degradation adds to the list of key cytoskeletal (Brancolini et al. 1995; Caulín et al. 1997; Mills et al. 1999) and trafficking molecules (Cosulich et al. 1997; Swanton et al. 1999a; Mancini et al. 2000) that are cleaved during the crucial process of apoptosis. A key issue will be to understand how these changes together expedite the recognition and clearance of apoptotic cells.

We thank Birgit Lane, David Meyer, Jon Scholey, Ron Vale, and Richard Vallee for supplying antibodies, and Don Nicholson and Sophie Roy for caspases. We are grateful to Martin Lowe and Lisa Swanton for comments and advice.

This work was supported by the Biotechnology and Biological Sciences Research Council (34/BI11195 to V. Allan and P. Woodman), the Lister Institute for Preventive Medicine (V. Allan) and the Medical Research Council (G117/153 to P. Woodman; and COG grant G9722026).