Human Rabaptin-5 Is Selectively Cleaved by Caspase-3 during Apoptosis*

We have previously shown that Xenopusrabaptin-5 is cleaved in apoptotic extracts, with a concomitant reduction in the ability of these extracts to support endosomal membrane fusion (Cosulich, S. C., Horiuchi, H., Zerial, M., Clarke, P. R., and Woodman, P. G. (1997) EMBO J.16, 6182–6191). In this report we demonstrate that caspase-dependent cleavage is a conserved feature of rabaptin-5. Human rabaptin-5 is cleaved at two sites (HSLD379 and DESD438) in apoptotic HeLa extracts. Cleavage is effected by caspase-3, since it is prevented when caspase-3 activity is either inhibited by Ac-DEVD-CHO or removed by immunodepletion. Moreover, an identical pattern of cleavage is observed using recombinant caspase-3. The action of caspase-3 is highly selective; neither caspase-2 nor caspase-7 are able to cleave recombinant or cytosolic rabaptin-5. Caspase-dependent cleavage of rabaptin-5 generates two physically separated coiled coil-forming domains, the C-terminal of which retains the ability to bind the Rab5 exchange factor rabex-5.

Programmed cell death (apoptosis) plays a fundamental role in the development and homeostasis of multicellular organisms (1,2). The primary feature of apoptosis is rapid engulfment and degradation of dying cells by their neighbors, so that an inflammatory response can be avoided. Since in many cases the engulfing cells are not specialized for phagocytic uptake (3), signals that expedite engulfment and degradation are likely to arise from the apoptotic cell. A critical event during apoptosis is therefore the expression of surface receptors that permit the specific recognition of a dying cell. One such receptor is probably phosphatidylserine, which is translocated from the inner leaflet to the outer leaflet of the plasma membrane during apoptosis (4). There is considerable evidence, however, that other surface moieties, including carbohydrate, form part of the recognition signal (5).
In addition to changes at the surface, the changes in cellular function that occur in an apoptotic cell are characterized by a variety of striking morphological and biochemical alterations. These include fragmentation of the nucleus and activation of endonuclease(s) (6,7), cell shrinkage and fragmentation, and plasma membrane blebbing (8). A further distinguishing feature of apoptotic cells is a loss of organized endomembrane structure; the nuclear envelope is frequently lost, and other recognizable membrane structures such as the Golgi complex are replaced by a disorganized array of vacuoles and vesicles (9,10). The so-called execution phase of apoptosis is evolutionarily conserved (11), underlining its importance.
It is now widely believed that many (although not all) apoptotic changes are linked to activation of a number of conserved cysteine proteases (caspases), which cleave specific substrates involved in key cellular processes (for review see Ref. 12). Caspases themselves are present as proenzymes that are readily cleaved, either autocatalytically (13,14) or by upstream "activator" caspases (15,16), during apoptosis. This provides the cell with a means to rapidly amplify its apoptotic response. Caspases can be divided into groups based on their sequence-selective protease activity toward peptide substrates. Thus, caspases-2, -3, and -7 (group II caspases) all cleave preferentially after the sequence DEXD, whereas caspases-6, -8, and -9 (group III) prefer the sequence (V/L)E(H/T)D (17). A major question is whether such overlapping substrate specificity within each group indicates that these enzymes represent tissue-specific isoforms or redundant isoforms within the same cell, or whether caspases exhibit far greater specificity toward polypeptide substrates in vivo.
Given the profound morphological changes occurring to membranes within apoptotic cells, and evidence for alterations in the expression of surface receptors, we anticipated that apoptosis would be associated with changes in the dynamics of the endocytic/recycling pathways. On this basis, we examined whether endosomal membrane fusion, an event that is essential for endosomal organization and for transport of receptors through the endocytic recycling pathway, is affected in apoptotic extracts. Endosomal fusion was indeed substantially reduced during apoptosis in Xenopus extracts, and this reduction was associated with specific cleavage of the Rab5 effector rabaptin-5 (18). Cleavage of rabaptin-5 was also apparent in cellular models of apoptosis, and was accompanied by reduced endocytic capacity. Hence, rabaptin-5 cleavage appears to be an important determinant in the abrogation of normal cellular function during apoptosis.
In this study, we have examined in detail the activity that cleaves rabaptin-5. We have also mapped the site of rabaptin-5 cleavage, in order to understand how it might interfere with rabaptin-5 function and thus contribute to impairment of normal endocytic transport. Our previous data suggested that rabaptin-5 is cleaved by a caspase-related activity (18). We now show that caspase-3 cleaves human rabaptin-5 at two closely positioned and conserved sites to generate physically separated N-and C-terminal domains. The activity of caspase-3 toward rabaptin-5 is surprisingly selective, since neither caspase-2 nor caspase-7 effect cleavage.

EXPERIMENTAL PROCEDURES
Reagents-Cytochrome c was obtained from Roche Molecular Biochemicals Ltd., Lewes, Sussex, United Kingdom (UK). Ac-DEVD-CHO 1 * This work was supported by Medical Research Council Grants G117/153, G9630910, and G9533795MA and by Biotechnology and Biological Sciences Research Council Grant C05969. 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.
‡ To whom correspondence should be addressed. Generation of Extracts-Suspension HeLa cells were grown to a density of 10 6 cells/ml in minimal essential medium modified for suspension cultures (Life Technologies, Inc., Paisley, Scotland) supplemented with 5% fetal calf serum. Cells were harvested from 2 liters of culture, washed twice in KEHM buffer (50 mM KCl, 10 mM EGTA, 50 mM Hepes, pH 7.4, 2 mM MgOAc), and resuspended in 2 volumes of the same buffer. After addition of dithiothreitol (1 mM) and protease inhibitors (1 g/ml apopain, 1 g/ml pepstatin A, 5 g/ml E64, 1 g/ml chymostatin, 40 g/ml phenylmethylsulfonyl fluoride from a 1000x stock in Me 2 SO), cells were homogenized by passing through a 8.02-mm bore in a stainless steel block containing a 8.004-mm diameter ball (19). A cytosol fraction was produced by centrifuging the homogenate at 300,000 ϫ g av for 45 min. The cytosol was snap-frozen and stored at Ϫ80°C. To generate apoptotic extracts, cytosol (20 mg of protein/ml) was incubated for the indicated times at 30°C with an ATP regenerating mixture (1 mM ATP, 5 mM creatine phosphate, 10 g/ml creatine kinase final) and 10 M cytochrome c.
HL-60 cells (grown in RPMI 1640 medium with 5% fetal calf serum) were induced to undergo apoptosis by treatment with 50 M etoposide (20) or with 1 g/ml anisomycin (21). To generate apoptotic HL60 extracts, 1-2-liter cultures of cells were treated with etoposide, then harvested by centrifugation, washed in KEHM buffer, and homogenized as for HeLa cells.
For immunodepletion of caspase-3, cytosol (4 mg of protein) was incubated overnight with gentle rotation in a tube containing 40 l of protein A-Sepharose beads to which had been pre-bound 8 g of anticaspase-3 antibody. As a control, cytosol was incubated with beads pre-treated with a non-relevant antibody.
Xenopus egg extracts were prepared exactly as described (18). Caspase Activity Assays-Recombinant active caspases (17) were diluted into caspase cleavage buffer (50 mM Hepes-KOH, pH 7.4, 2 mM EDTA, 0.1% (w/v) CHAPS, 10% (w/v) sucrose, 5 mM dithiothreitol; with the exception of caspase-2, which was diluted in the same buffer except with 50 mM NaOAc, pH 5.5, to provide a pH optimum), and duplicate samples (5 l) were incubated with 0 -100 M Ac-DEVD-AMC or Ac-LDESD-AMC for 10 min at 30°C. Samples were diluted to 2.5 ml in water and the generation of fluorescent product determined (22). Cleavage of polypeptide substrates by recombinant caspases was carried out in the appropriate buffer for 2 h at 30°C, in a final volume of 10 l.
Recombinant Proteins-Bacterially expressed His 6 -rabaptin-5 was purified as described by Stenmark et al. (23). For some experiments, His 6 -rabaptin-5 containing a C-terminal protein C peptide tag was expressed and purified as above. This reagent was generated by inserting the protein C coding sequence (GAA GAT CAG GTA GAT CCA CGG TTA ATC GAT GGT AAG TAA) immediately downstream from the rabaptin-5 coding sequence. The protein C coding sequence was followed by an in-frame stop codon (underlined) and was inserted using polymerase chain reaction-based site-directed mutagenesis (ExSite; Stratagene) according to the manufacturer's instructions.
PARP, caspase-3, and caspase-2 cDNAs were present in pcDNA3 vectors (24). [ 35 S]Methionine-labeled proteins were generated by incubating 1 g of cDNA with a combined transcription/translation kit (Promega) supplemented with T7 RNA polymerase and 60 Ci of [ 35 S]methionine. Rabaptin-5 cleavage site mutants were generated using the QuikChange site-directed mutagenesis kit (Stratagene). To examine cleavage of in vitro translated His 6 -rabaptin-5 in extracts, translation product was diluted 10-fold into Xenopus or HeLa extracts and incubated at 25°C or 30°C respectively. Samples were denatured by boiling for 5 min in 1% SDS, then diluted to 200 l with immunoprecipitation buffer (10 mM Tris-HCl, pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100) and precipitated with 0.5 l of appropriate antibody with protein G-or protein A-agarose.
Other Methods-Gel filtration chromatography to separate rabaptin-5 from procaspases 2 and 3 was performed using a 24-ml Sepha-rose-6 column (Amersham Pharmacia Biotech) on a Beckman BioSys 510 HPLC system. Analytical fractionation of apoptotic cytosol was achieved using a Superose 6 column on a Smart System (Amersham Pharmacia Biotech). For immunoprecipitation experiments, polyclonal anti-rabaptin-5 serum (50 l) or monoclonal anti-rabaptin-5 (50 l of ascites) were preincubated with 100 l of protein A-or protein Gagarose, respectively, and the antibody was covalently attached to the beads using dimethylpimelimidate according to published methods (25). These beads were then used to immunoprecipitate rabaptin-5 or its fragments from 2 mg of cytosol. For determination of N-terminal sequences, protein C-tagged rabaptin-5 was incubated with 10 nM caspase-3 for 4 h at 37°C, then precipitated in the presence of 1 mM CaCl 2 onto protein G-agarose beads to which anti-protein C antibody (HPC4; Roche Molecular Biochemicals, Lewes, UK) had been covalently attached. Residual full-length rabaptin-5 and C-terminal fragments were eluted with 5 mM EDTA. The products were separated by SDS-PAGE, transferred to polyvinylidene difluoride membrane, and subjected to N-terminal sequence analysis.
Endosome fusion was assayed as described previously (18), using for each sample 5 l of donor membranes, 7 l of acceptor membranes, 10 l of cytosol in a total volume of 40 l.

RESULTS
Human Rabaptin-5 Is Cleaved during Apoptosis-Our previous work had demonstrated that Xenopus rabaptin-5 is cleaved in egg extracts to yield a C-terminal fragment of 45-50 kDa (18). Similar extracts have been shown to undergo several apoptotic events, including activation of caspases and endonucleases (26,27). To establish whether cleavage by apoptotic proteases is a conserved feature of rabaptin-5, human His 6rabaptin-5 was translated in vitro and then combined with a Xenopus egg extract. After appropriate incubation, rabaptin-5 cleavage products of approximately 62 (Fig. 1A) and 47 kDa (Fig. 1B) were produced, which could be immunoprecipitated by anti-His antibody (Fig. 1A) or an antibody recognizing the C-terminal portion of rabaptin-5 ( Fig. 1B), respectively. The time course of cleavage was similar to that previously reported for cleavage of endogenous Xenopus rabaptin-5 (18) (data not shown), and was prevented by inclusion of the specific caspase inhibitor Ac-DEVD-CHO ( Fig. 1). Hence, human His 6 -rabaptin-5 behaves in apoptotic Xenopus extracts in the same way as Xenopus rabaptin-5, confirming that cleavage is a conserved function.
Recent work from Wang and colleagues (16,28,29) has shown that cytosolic extracts from mammalian cells can be triggered to enter apoptosis by addition of cytochrome c. This process can be sensitized by, although it is not strictly dependent upon, inclusion of dATP (24,28). We examined whether this system would be convenient for studying the apoptotic cleavage of rabaptin-5. Previous work has shown that activation of caspase-3 in HeLa extracts occurs via cytochrome c-dependent activation of caspase-9 (16), and is maximal within 30 min of addition of cytochrome c (24). Substrates for caspase-3 and/or downstream caspases are cleaved more slowly. For example, caspase-2, whose cleavage requires caspase-3 activity (24,30), is cleaved in these extracts between 1 and 3 h after addition of cytochrome c to yield an immunoreactive 12/13-kDa fragment (24). Rabaptin-5 is also cleaved, although somewhat more slowly than caspase-2, to yield a C-terminal fragment similar in size to that generated within apoptotic Xenopus extracts. Using a monoclonal antibody that recognizes an epitope within the N-terminal portion of rabaptin-5, the 62-kDa N-terminal fragment was also identified (Fig. 2B). An additional minor product of approximately 53 kDa (N-53) was also seen with this antibody, indicating the presence of a further cleavage site. This was confirmed by inclusion of in vitro translated rabaptin-5 (see Fig. 3). Prolonged incubation of apoptotic extracts resulted in the disappearance of the N-62 fragment, while the N-53 fragment remained resistant to further proteolysis (Fig.  2C, left panel), consistent with it being the result of a second, slower, cleavage event. Importantly, appearance of a 53-kDa polypeptide reactive against this antibody was observed in apoptotic cells and correlated with the disappearance of fulllength rabaptin-5 (Fig. 2C, right panel). These results suggest that N-53 is a final cleavage product in apoptotic cells. Moreover, they demonstrate that the cleaving activity that is present in apoptotic cells is likely to be similar, if not identical, to that present in cytochrome c-activated extracts.
The amino acid sequence of rabaptin-5 (23) and the size of the C-terminal cleavage product in Xenopus extracts (18) had suggested to us that caspase-dependent cleavage was most likely to occur after aspartate 438 or aspartate 446. Indeed, inclusion of a 22-mer peptide that included both aspartate residues retarded cleavage in Xenopus extracts by 1-2 h (data not shown). To identify the exact cleavage site, rabaptin-5 mutants were prepared with either Asp 438 or Asp 446 replaced by alanine, and were incubated in apoptotic HeLa extracts. Compared with wild-type rabaptin-5 (Fig. 3A), Asp 438 rabaptin-5 did not produce N-62 or C-47 fragments (Fig. 3B) while Asp 446 rabaptin-5 behaved identically to the wild-type (Fig.  3C). Hence, human rabaptin-5 is cleaved at the sequence DESD 438 F. Asp 438 rabaptin-5 still gave rise to the N-53 fragment, confirming that it is the product of a second cleavage event that is not dependent on prior cleavage at Asp 438 .
Rabaptin-5 Is Cleaved Selectively by Caspase-3-Systematic studies of caspase cleavage sites using synthetic modified tetrapeptides have identified the amino acid preferences at the P2-P4 positions for each caspase (17). Based on these studies, caspases have been divided into several groups. The group II "effector" caspases, caspase-3 and caspase-7, both cleave effectively after the sequence DEVD. Although this is similar to the primary cleavage site within rabaptin-5 (DESD), the activity of both caspases toward peptides is reduced when serine is placed at P2. In contrast, DESD is a preferred cleavage site, second only to DEHD, for the group II "activator" caspase, caspase-2 (17). Based on these studies, and our previous observation that concentrations of recombinant caspase-3 just sufficient to accelerate apoptotic changes in Xenopus egg extracts were unable to cleave rabaptin-5 directly (18), it seemed likely that a caspase-2-like activity would be responsible for cleaving rabaptin-5. However, we undertook a detailed examination to establish the true identity of the cleaving activity. First, we examined whether cleavage of cytosolic rabaptin-5 in HeLa extracts is dependent on caspase-3 activity. We have already used this approach to establish that cleavage of caspase-2 occurs via caspase-3 (24). We first examined the sensitivity of cleavage to the specific caspase-3 inhibitor Ac-DEVD-CHO, and found that cleavage of rabaptin-5 was prevented by inclusion of 50 nM Ac-DEVD-CHO (data not shown), similar to those concentrations that prevent cleavage of the caspase-3 substrates PARP and caspase-2 (24). To further establish a dependence on caspase-3 activity, extracts were depleted of caspase-3 prior to addition of cytochrome c. Cytosols pre-treated with an antibody to caspase-3 were depleted of caspase-3 precursor by at least 90% compared with mock-depleted cytosols (data not shown). When incubated with cytochrome c, these extracts were unable to cleave rabaptin-5 (Fig. 4A, left panel). In contrast, rabaptin-5 was cleaved almost to completion within 4 h when mock-depleted extracts were incubated with cytochrome c (Fig. 4A, right panel).
Although cleavage of cytosolic rabaptin-5 is dependent on caspase-3, it is possible that it is cleaved directly by a downstream effector of caspase-3, such as caspase-2. To address this question, HeLa extracts were preincubated for a period sufficient to activate caspase-3, as well as potential downstream caspases. They were then depleted of activated caspase-3. Western blotting of extracts confirmed that greater than 90% of activated caspase-3 had been removed (data not shown). When excess bacterially expressed rabaptin-5 was added to depleted extract, no cleavage product was detected before 3 h, and significant cleavage did not occur until after 5 h incubation (Fig.  4B, top left). In contrast, cleavage of rabaptin-5 was observed within 1 h of its addition to mock-depleted extract, and the majority of recombinant rabaptin-5 was cleaved after 4 -5 h (Fig. 4B, top right). The rate at which caspase-3-depleted extract could cleave rabaptin-5 was increased significantly by inclusion of recombinant caspase-3 (Fig. 4B, bottom left) or the beads isolated from the immunodepletion step (Fig. 4B, bottom  right), confirming that the major cleaving activity was caspase-3.
To demonstrate that the caspase activated within apoptotic cells most likely to cleave rabaptin-5 is caspase-3, cytosol was prepared from apoptotic HL60 cells. These cytosols had substantial rabaptin-5 cleaving activity, since significant cleavage of recombinant rabaptin-5 was observed within 1-2 h (Fig. 4C,  left panel). Again, prior depletion of active caspase-3 from these extracts significantly reduced the rate at which rabaptin-5 was cleaved (Fig. 4C, right panel).
Caspase-2 has a preference for the cleavage site DESD over DEVD, so it was expected that rabaptin-5 would be a good substrate for caspase-2 (17). Surprisingly, no cleavage of rabaptin-5 was observed at concentrations of recombinant caspase-2 as high as 2 M (Fig. 5E). The activity of the caspase-2 preparation was confirmed by examining its ability to cleave caspase-3, which was observed above 0.5 nM recombinant caspase-2 (Fig. 5F). Importantly, no self-cleavage by in vitro translated caspase-3 was observed. Likewise, the activity of caspase-2 and its preference for the sequence DESD over DEVD (17)  Studies so far have suggested that the substrate specificity of caspase-3 and caspase-7 are very similar (see Ref. 12 for review), tempting speculation that they may be redundant activities. However, we observed no cleavage of rabaptin-5 by recombinant caspase-7, even at concentrations as high as 580 nM (Fig. 5G). Again, the activity of the caspase-7 preparation was confirmed by examining its ability to cleave PARP (Fig. 5H) or Ac-DEVD-AMC (K m ϭ 27 M; V max ϭ 1608 fluorescence units/min/nmol).
The activity of recombinant caspase-3 toward rabaptin-5 was exploited to obtain the N-terminal sequences of the major cleavage fragments. Recombinant rabaptin-5, tagged at the C terminus with a peptide derived from protein C, was incubated with caspase-3 before being immunoprecipitated with antiprotein C beads. When the cleavage products were eluted and sequenced, the C-47 fragment was found to contain the N-terminal sequence FGPLVGADSV. Since this sequence corresponds exactly to that starting at amino acid 439 within full-length rabaptin-5, this confirms that the primary caspase-3 cleavage site is after DESD 438 . A higher molecular mass product of approximately 54 kDa gave the N-terminal sequence AGL(-)(-)PS-GDP (where (-) was not resolved), corresponding to the sequence immediately after HSLD 379 . The slower cleavage observed at this site within extracts is consistent with it being further removed than DESD from the optimal caspase-3 cleavage site.
Although the data presented above are consistent with selective cleavage of rabaptin-5 by caspase-3, it remained possible that the same specificity would not be observed for cytosolic rabaptin-5. This point seemed particularly important, given that cytosolic rabaptin-5 is found as part of a large complex, which also includes the Rab5 guanine nucleotide exchange factor rabex-5 (31). Components of this complex might influence caspase specificity by influencing the orientation or exposure of the cleavage sites within rabaptin-5, or by recruiting alternative caspases. To address this issue, cytosolic rabaptin-5 was fractionated away from endogenous caspase-2 and caspase-3 by gel filtration chromatography. Its ability to act as a substrate for caspase-2 or caspase-3 was then compared with that of recombinant rabaptin-5. As shown in Fig. 6 (upper panels), cleavage of both cytosolic and recombinant rabaptin-5 was first observed at the same concentration of caspase-3 (0.5-1 nM). Similarly, neither preparation of rabaptin-5 was cleaved by caspase-2 concentrations as high as 2 M (Fig. 6, lower panels). For these experiments rabaptin-5 was added well in excess, so that product formation would be linear in relation to enzymic activity.
Rabaptin-5 Cleavage Separates N-and C-terminal Fragments-Rabaptin-5 forms part of a high molecular weight complex containing rabex-5 (31), and has been shown to bind to a number of effectors including Rab5 (23) and Rab4 (32). We sought to establish, therefore, whether caspase-dependent cleavage of rabaptin-5 would give rise to physically separate subcomplexes and thus potentially provide a means to uncouple the action of various effectors. This was investigated first by co-immunoprecipitation using antibodies specific for either Nor C-terminal fragments of rabaptin-5. In control extracts, immunodepletion of rabaptin-5 from cytosol using an anti-N-terminal antibody led to complete loss of rabaptin-5, as detected both by anti-N-terminal and anti-C-terminal antibodies, as expected (Fig. 7A). Likewise, immunodepletion with anti-Cterminal antibody led to complete loss of anti-N-and anti-Cterminal reactivity. When apoptotic extracts were depleted with anti-N-terminal antibody, complete loss of anti-N-terminal reactivity from the supernatant was observed. However, significant amounts of the C-47 fragment remained. Conversely, immunoprecipitation with anti-C-terminal antibody led to complete depletion of this fragment, while the N-terminal fragment remained in the supernatant.
The fate of the rabaptin-5 fragments was also followed by gel filtration chromatography (Fig. 7B). When apoptotic cytosol was applied to a Superose 6 column, residual full-length rabaptin-5 migrated in fraction 6, and could be detected with both anti-N-and anti-C-terminal antibodies. Rabaptin-5 from control extracts migrated to a similar position (data not shown). The band in fraction 8 that cross-reacted with the anti-Cterminal antibody was identified as rabaptin-5␤, a rabaptin-5related protein (33), by the use of an antibody specific for rabaptin-5␤ (data not shown). The rabaptin-5 fragments migrated as distinct but overlapping lower molecular weight species. Both N-62 and N-53 fragments migrated to fraction 9, and to some extent to fraction 10 (Fig. 7B, top panel). The C-47 fragment migrated slightly more slowly, with most appearing in fraction 10 and some in fraction 11 (Fig. 7B, middle panel). Thus, caspase cleavage results in physical separation of the two halves of rabaptin-5 into lower molecular weight complexes.
Physical separation of the rabaptin-5 fragments allowed us to determine which portion of the protein binds to rabex-5. When fractions from the Superose 6 column were analyzed with antibodies to rabex-5, two peaks of reactivity were observed (Fig. 7B, bottom panel). The first peak was found in fraction 8, coincident with rabaptin-5␤, which is known to bind directly to rabex-5 (31,33). The second peak was found in fraction 10, with the bulk of the C-47 reactivity. Although this suggested that rabex-5 interacted with the C-terminal portion of rabaptin-5, the data were complicated by the presence of rabaptin-5␤. To further examine the binding of rabex-5, coimmunoprecipitation of rabex-5 with the anti-N-terminal rabaptin-5 antibody was measured. As shown in Fig. 7C, rabex-5 co-immunoprecipitated with anti-rabaptin-5 N-terminal antibody in control cytosol. However, efficient co-immunoprecipitation was not observed after rabaptin-5 was cleaved in apoptotic extracts. This provides evidence that the C-terminal portion of rabaptin-5 is required for efficient rabex-5 binding. The converse experiment, of demonstrating that rabex-5 could still be precipitated from apoptotic cytosol with anti-C-terminal rabaptin-5 antibody, could not be performed since this antibody also recognizes rabaptin-5␤.
The ability of HeLa cytosol to support endosome fusion was reduced with increasing time in the presence of cytochrome c, and this reduction was accompanied by cleavage of rabaptin-5 (Fig. 8, A and B). Inhibition of fusion was only partial and, in contrast to previous studies using Xenopus egg extracts (18), could be overcome to some extent by increasing the cytosol concentration (data not shown). This was most likely due to the resistance of rabaptin-5␤ to apoptotic cleavage (data not shown). In any case, it did not allow for reconstitution experiments to demonstrate whether cleavage of rabaptin-5, or other effector(s) of fusion, is solely responsible for the reduction in fusion activity, as appears to be the case in Xenopus egg extracts. Cleavage of rabaptin-5 remains the most likely cause of the reduction in fusion activity, since rabex-5 is not cleaved in apoptotic extracts and the Rab5 effector EEA1 (8, 34) concentration is reduced only slightly (data not shown). However, cleavage of other effectors of endosome fusion cannot be ruled out. DISCUSSION In this report we have investigated the caspase-specific cleavage of rabaptin-5 in cytochrome c-activated human cell extracts. We have demonstrated directly and by mutagenesis experiments (for one site only) that rabaptin-5 is cleaved at two sites; cleavage in extracts occurs more rapidly at the sequence DESD 438 , but also occurs at HSLD 379 . Cleavage of rabaptin-5 at this second site is followed to completion in apoptotic cells, most likely by caspase-3, since examination of apoptotic cells revealed a stable N-terminal product co-migrating precisely with rabaptin-5 1-379. Further data indicate that caspase-3 is responsible for rabaptin-5 cleavage in apoptotic cells, since immunodepletion of caspase-3 substantially reduced rabaptin-5 cleaving activity from extracts made directly from apoptotic cells.
Both caspase cleavage sites are conserved between all rabaptin-5 sequences currently on the data base. Therefore, it is probable that the cleavage of Xenopus rabaptin-5 that we previously observed corresponded to that at DESD 438 , underlining the fact that caspase-dependent cleavage is likely to be a conserved feature of this protein. The absence of cleavage at HSLD 379 in Xenopus extracts may reflect somewhat lower caspase activity, or a slightly different substrate specificity of Xenopus versus human caspase-3. Analysis of rabaptin-5 reveals two extensive domains capable of forming coiled coils, which generate a rodlike parallel homodimer (23). Between the N-and C-terminal helical domains is a non-helical linker region (approximately amino acids 350 -530). Thus, cleavage at Asp 379 and Asp 438 would be expected to cause separation of these domains, at least in the purified protein. Despite the fact that it forms part of a high molecular weight complex (23), our data indicate that physical separation of these domains also occurs within cytosolic rabaptin-5.
Rabaptin-5 has previously been identified as a downstream Rab5 effector, which is essential for endosome fusion and which binds Rab5 within its C-terminal domain (23,32). Rab5 is a member of the family of small GTP-binding proteins that participate in intracellular membrane docking/fusion reactions (35). It cycles between a cytosolic GDP-bound pool and a membrane-associated GTP-bound form (36). It is in the latter conformation that Rab5 is active and able to recruit effectors of endosome fusion, including rabaptin-5 and EEA1 (23,37,38). The precise role that Rab effectors such as these play in membrane docking/fusion is not clear, although evidence indicates that they participate in peripheral "tethering" of membranes (39) and may in addition activate the appropriate docking receptors (SNAREs (soluble N-ethylmaleimide factor attachment protein receptor)) within opposing membranes (40,41).
Recent reports demonstrate that the role of rabaptin-5 during endosome fusion appears more complex than simply that of a downstream effector, however. First, rabaptin-5 itself reduces the GTPase activity of Rab5 and will thus maintain Rab5 in its active form (42). Second, rabaptin-5 forms a cytosolic complex with a Rab5 guanine nucleotide exchange factor, rabex-5 (31). Hence, recruitment of rabaptin-5 by Rab5-GTP may enhance localized exchange activity. Since rabaptin-5 forms a homodimer (32) and could thus recruit two molecules of rabex-5 per Rab5 monomer, this may provide a means to form foci within the membrane where Rab5 activity is retained (31). To date, the site within rabaptin-5 that recruits rabex-5 has not been identified. However, our studies provide strong indications that rabex-5 binds to the C-terminal portion of rabaptin-5, thereby placing Rab5 exchange activity adjacent to Rab5 itself. Our findings therefore indicate that inhibition of endosome fusion in apoptotic extracts is unlikely to be due simply to an inability to maintain Rab5 in its active conformation. It is more likely that cleavage of rabaptin-5 prevents the recruitment of further Rab5 effectors. Indeed, the N-terminal portion of rabaptin-5 behaves as a higher molecular weight species than the C-terminal rabaptin-5/rabex-5 complex, indicating that it is bound to other cytosolic component(s).
Intriguingly, rabaptin-5 will also bind, via its N terminus, Rab4 (32). Rab4, like Rab5, is localized to the early endosome, though the activity of Rab4 seems to oppose that of Rab5. Thus, Rab4 is apparently required for a transport pathway that leads away from the early endosome, most likely the recycling of vesicles to the cell surface (43). The finding that rabaptin-5 will bind Rab proteins involved in both endocytic and exocytic pathways raises the possibility that it acts as a functional linker, co-ordinating the fluxes of both pathways (32). In this way, the ratio of external to internal membrane can be maintained despite the rapid movement of material between these compartments. The notion that rabaptin-5 acts as a linker between endocytic and exocytic pathways is further supported by the finding that rabaptin-5 interacts with the Rab3 effector rabphilin-3 (44). Cleavage of rabaptin-5 by apoptotic proteases will destroy this functional linkage, and may contribute to the changes in membrane dynamics and morphology that are apparent in apoptotic cells. A full explanation of the role of rabaptin-5 cleavage in re-organization of endosomal membranes, and possible compensating effects of other interacting proteins such as rabaptin-5␤, must await examination of these membranes in individual cells undergoing apoptosis.
We have shown that rabaptin-5 is cleaved, both at Asp 379 and Asp 438 , by caspase-3. Cleavage occurs above 0.5 nM caspase-3, somewhat higher than the concentration required to cleave PARP or caspase-2, two well characterized substrates for caspase-3 (45), and is consistent with the 2-3-fold slower rate of cleavage of rabaptin-5 compared with caspase-2 in HeLa extracts. The K cat /K m is however similar to that for other identified caspase-3 substrates, for example focal adhesion kinase (46) and huntingtin (47), as well as being very close to that of caspase-1/ interleukin-1␤-converting enzyme for its cellular substrate, interleukin 1-␤ (47,48). We expected other group II caspases (17) to cleave rabaptin-5, particularly at DESD 438 . However, neither the "effector" caspase, caspase-7, nor the "activator" caspase, caspase-2, cleaved rabaptin-5 when used at concentrations 100fold or greater than caspase-3. These data are fully consistent with our finding that the major cleaving activity in apoptotic cytosols is caspase-3. Crucially, our results rule out the possibility that cytosolic binding partners might influence the specificity of rabaptin-5 cleavage either by changing the conformation of the substrate or by recruiting other caspases. Although it is possible that membrane-bound rabaptin-5 has an altered caspase susceptibility, our preliminary observations indicate that addition of membrane does not influence the pattern or rate of rabaptin-5 cleavage (data not shown).
The inability of caspase-2 or caspase-7 to cleave rabaptin-5 is particularly surprising, given that several polypeptide substrates are cleaved efficiently by all group II caspases (see Ref. 12 for review). In particular, the discrepancy between caspase-2 and -3 is odds with our kinetic data using peptide substrates, where both caspases cleave LDESD effectively. Our finding demonstrates the importance of the tertiary structure of substrates in determining caspase recognition. It further supports the hypothesis that caspases with apparently similar protease activities fulfill selective roles within the apoptotic cell, rather than being merely redundant activities. In the case of caspase-2, such selectivity is consistent with its role as an "activator" caspase, which acts as a functional linker between apoptotic stimuli and downstream "effector" caspases. The inability of caspase-7 to cleave rabaptin-5 is more surprising. Several studies have demonstrated how selectivity of caspase action may be achieved by differential localization of caspases in apoptotic cells (49,50). This is among the first instances where such selectivity appears to be a consequence of biochemical specificity.