Active relocation of chromatin and endoplasmic reticulum into blebs in late apoptotic cells

Apoptosis is a well-characterised form of programmed cell death. Its desired end-point is phagocytosis of the dying cell by either immune or non-immune cells without inducing inflammatory or immune responses (Parnaik et al., 2000; Savill et al., 2002; Savill and Fadok, 2000; White and Rosen, 2003). To date, the changes on the apoptotic cell surface that promote binding and uptake are poorly understood, although exposure of plasma membrane phosphatidyl serine (PS) has been shown to play a vital role in the phagocytosis of apoptotic cells by PS receptor-expressing phagocytes (Fadok et al., 2000; Hoffmann et al., 2001). In addition, a battery of phagocyte receptors, including members of the integrin family, scavenger receptors and lectins, have been implicated in the uptake of apoptotic cells in other contexts (Savill and Fadok, 2000).

Apoptosis is accompanied by a series of dramatic changes in cellular morphology that ultimately cause cells to fragment into `apoptotic bodies' (defined here as plasma membrane-enclosed remnants that are completely released from the body of the cell). These are an important intermediate in the clearance of the apoptotic cell. Consistent with this, accumulating evidence suggests that apoptotic bodies and surface protrusions of apoptotic cells are sites for the concentration of specific phagocytic markers (Casciola-Rosen et al., 1996; Korb and Ahearn, 1997; Navratil et al., 2001; Ogden et al., 2001). Apoptotic bodies and surface protrusions are also important subcellular domains that can influence the immune system: they are known to contain condensed chromatin (e.g. Leist and Jaattela, 2001) and several chromatin- and ribonucleoprotein derived auto-antigens (Casciola-Rosen et al., 1994; Rosen and Casciola-Rosen, 1999). Thus, cellular components that might present a particular danger to the immune system are concentrated within peripheral structures that are enriched with phagocytic markers, and hence are more likely to be efficiently engulfed - particularly where there is limited availability of professional phagocytes (de Cathelineau and Henson, 2003; Savill et al., 2002). Therefore, understanding how cellular components and phagocytosis markers are concentrated within these specialised domains is of considerable importance.

Apoptotic body formation requires the actions of caspases - cysteinyl aspartate-specific apoptotic proteases - that cleave selected proteins within the cell (Earnshaw et al., 1999; Fischer et al., 2003). Early in apoptosis, adherent cells lose contact with neighbours and/or substratum, probably via caspase cleavage of cell-cell contact factors, such as β-catenin, (Brancolini et al., 1997) and focal adhesion proteins, such as tensin (Kook et al., 2003), FAK (Levkau et al., 1998) and p130^cas (Kook et al., 2000). Release of the apoptotic cell is closely followed by surface blebbing (zeiosis), where stress-activated protein kinases (e.g. p38/SAPK2) signal the rearrangement of the actin cytoskeleton, prompting myosin II-dependent contractile forces to drive the formation of dynamic plasma membrane swellings that appear suddenly and are rapidly retracted (Huot et al., 1998; McCarthy et al., 1997; Mills et al., 1998a; Mills et al., 1998b). Blebbing depends upon activation of myosin light chain (Mills et al., 1998b) via phosphorylation indirectly or directly, by Rho-activated kinase (ROCK-I), which in turn is cleaved into a constitutively active form by caspases (Coleman et al., 2001; Sebbagh et al., 2001). Microtubules, conversely, are thought to disassemble early in apoptosis (Mills et al., 1998a; Mills et al., 1999), the minus- end microtubule motor dynein is cleaved by caspases (Lane et al., 2001), and intermediate filaments fragment and aggregate at the cell centre (Byun et al., 2001; Caulin et al., 1997). While these cytoskeletal changes are taking place, the nucleus partitions into distinct pyknotic domains of condensed chromatin (karyorhexis) (Kerr et al., 1972), caused by caspase 3-dependent activation of exonucleases (Fraser et al., 1996; Meng et al., 2000) and caspase 6-dependent cleavage of lamins (Ruchaud et al., 2002). The Golgi apparatus fragments (Lane et al., 2002; Sesso et al., 1999) as a result of caspase-dependent cleavage of several proteins, including GRASP65 (Lane et al., 2002), Golgin-160 (Mancini et al., 2000), p115 (Chiu et al., 2002), giantin and syntaxin 5 (Lowe et al., 2004), while the endoplasmic reticulum (ER) becomes distended (Sesso et al., 1999) and possibly vesiculated (Casciola-Rosen et al., 1994). Finally, `condensation' or shrinkage of the apoptotic cell, which occurs some time after surface blebbing stops (Mills et al., 1999), is followed by the generation of apoptotic bodies. There is evidence that fragmentation is dependent on an intact actin cytoskeleton (Cotter et al., 1992).

Why do apoptotic cells bleb? An attractive theory is that membrane blebbing ultimately leads to cellular fragmentation, and that surface blebs are therefore the progenitors of apoptotic bodies. In support of this, cellular fragmentation does not occur when blebbing is prevented by addition of the ROCK-I inhibitor Y27632 (Coleman et al., 2001). However, fragmentation occurs some time after the cessation of plasma membrane blebbing in most models of apoptosis (Huot et al., 1998; Mills et al., 1998b; Sebbagh et al., 2001). In fact, surface blebbing can be dramatically extended (up to several days) if caspases are inhibited (Mills et al., 1998b), cells eventually dying by a caspase-independent, non-apoptotic form of cell death (McCarthy et al., 1997). Moreover, viable cells also display myosin II-dependent plasma membrane blebbing when spreading, migrating, dividing, or when under conditions of stress (Bereiter-Hahn et al., 1990; Cunningham, 1995; Fishkind et al., 1991; Huot et al., 1998; Straight et al., 2003). Taken together, these observations raise the possibility that apoptotic blebbing is simply a by-product of cellular reorganisation, perhaps linked to the partial loss of cellular contact with the substratum, and does not play a direct role in apoptotic progression.

To gain a better understanding of the pathway of apoptotic body formation, we have used both fixed and live-cell imaging to follow cells during apoptosis. We have identified two distinct phases of plasma membrane blebbing. The first is restricted to adherent cells, beginning concomitant with release of cells from their substratum and lasting for around 30-40 minutes. It produces small, highly dynamic surface swellings that arise suddenly and are rapidly withdrawn, and represents the early, caspase-independent blebbing phase (zeiosis) identified in previous studies (McCarthy et al., 1997; Mills et al., 1998b). A second, late phase of blebbing begins ∼20 minutes later, and is also observed in non-adherent cells. It results in large, long-lived membrane swellings that contain a distinct layer of cortical endoplasmic reticulum (ER), which often envelops pyknotic chromatin. Formation of late blebs requires active caspase-6, and surprisingly, involves both actomyosin and microtubules. Together, these findings demonstrate that coordinated, cytoskeleton-driven processes continue late into the apoptotic program.

Unless otherwise stated, all reagents were obtained from Sigma (Poole, UK). Stock solutions of anisomycin (5 mg ml^-1 in DMSO), nocodazole (5 mg ml^-1 in H₂O), zVEID.FMK (Calbiochem, Nottingham, UK: 12.5 mM in DMSO), Y27632 (Calbiochem: 100 mM in DMSO), Blebbistatin (Calbiochem: 100 mM in DMSO), ML7 (Alexis, Nottingham, UK: 30 mM in DMSO) and Latrunculin A (Molecular Probes, Oregon, USA: 10 mM in DMSO) were stored at -20°C. Alexa⁵⁹⁴ annexin V and Phalloidin were obtained from Molecular Probes.

The following antibodies were used: monoclonal anti-KDEL (clone 1D3: from David Vaux, University of Oxford, UK); monoclonal and polyclonal anti-calnexin C-terminal and N-terminal (Stressgen, San Diego, CA); monclonal anti-tubulin (B5-1-2, Sigma); monoclonal anti-GM130 (Transduction Labs, Nottingham, UK); monoclonal anti-HSP60 (clone LK1: Sigma); polyclonal anti-LAMP-2 (Affinity Bioreagents, CO, USA); anti-non-muscle myosin IIA (Sigma); anti-phospho myosin light chain II (Thr18, Ser19: Cell Signalling, MA, USA).

HeLa, A431 and SW13.Cl-2 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% serum. Jurkat T-cells and THP-1 cells were maintained in RPMI-1640 with 10% serum. Cells were grown in 5% CO₂ at 37°C.

Human high mobility group protein 1 (HMGB1) cDNA (accession number BQ894442) was obtained from the human EST I.M.A.G.E. clone database (http://www.hgmp.mrc.ac.uk/). HMGB1 was amplified from the EST clone using the following primers: 5′ TACG GAA TTC ATG GGC AAA GGA GAT CCT AG 3′ and 5′ TACG CCG CGG GAA CGA AAA AAG TCG GAA CTG 3′, which introduce EcoRI and SacII sites, respectively. The resultant PCR product (which encodes a protein lacking the terminal 33 amino acids from the highly acidic C-terminus) was fused to YFP at its C-terminus by subcloning into pEYFP-N1 (Clontech). HMGB1-YFP was transiently transfected into HeLa cells using Fugene 6 (Roche, Lewes, UK) according to the manufacturer's instructions.

Cells were induced to undergo apoptosis by treatment with 5 μg ml^-1 anisomycin or by UV irradiation (100 Jm^-2) (Lane et al., 2002). For analysis of apoptotic phenotype in the presence of inhibitors, HeLa cells were treated with anisomycin and incubated for 5 hours in the absence or presence of latrunculin A (1.0 μM), Y27632 (100 μM), nocodazole (5 μg ml^-1) or zVEID.FMK (12.5 μM).

Conventional epifluorescence images were obtained using an Olympus IX-71 inverted microscope (100× objective, 1.3 N.A.) fitted with a CoolSNAP HQ CCD camera (Photometrics, Tucson, AZ) driven by MetaMorph software (Universal Imaging Corporation, Downington, PA). Confocal images were obtained using a Leica AOBS SP2 microscope (63× objective, 1.32 N.A.) at 0.2 μm z-steps. Images were processed using Adobe Photoshop 7.0 for MacIntosh. For immunofluorescence staining, cells were either fixed for 15 minutes in 2% paraformaldehyde (methanol-free, EM grade; TAAB, Aldermaston, UK) followed by permeabilisation with 0.1% Triton X-100, or by -20°C methanol (5 minutes). For quantitation of late apoptotic blebs, cells were fixed in 2% paraformaldehyde, permeabilised with 0.1% Triton X-100, then labelled with anti-KDEL antibody (1D3) or anti-calnexin (polyclonal) followed by alexa⁵⁹⁴-conjugated anti-mouse and DAPI. Apoptotic cells were identified by chromatin morphology, and cells with chromatin containing late blebs were scored. These cells were then analysed for the presence of ER within surface blebs. For phalloidin staining, cells were fixed in 2% paraformaldehyde, permeabilised then incubated for 20 minutes with 2.5% phalloidin.

Live-cell imaging was carried out using either an Olympus IX-70 inverted microscope equipped with a Pentamax intensified cooled frame transfer camera (Roper Scientific, Trenton, NJ), driven by MetaMorph software (Universal Imaging Co.), using halogen lamp illumination for both transmitted and incidental light (Lane et al., 2002), or using a Leica DMIRBE microscope (63× objective, 1.32 N.A.) fitted with a Hamamatsu CCD camera (Welwyn Garden City, UK), driven by OpenLab 3 software (Improvision, Coventry, UK). For live-cell imaging on the Olympus system, cells were grown in CO₂-independent DMEM (Invitrogen, Paisley, UK), on 3 cm plastic dishes containing cover-glass inserts (MatTek Co., Ashland, MA). For measuring the timing of PS exposure, cells were incubated in DMEM with 1% alexa⁵⁹⁴-conjugated annexin V under 5% CO₂, and were imaged using the Leica DMIRBE/Hamamatsu system under constant camera exposure settings (1 image every 120 seconds at 10 ms exposure for phase contrast; 1 image every 600 seconds at 10 millisecond exposure for fluorescence; 125 digital gain; 2× binning).

Anisomycin-treated HeLa cells were fixed in 2% glutaraldehyde (TAAB) in 0.1 M cacodylate buffer for 30 minutes, washed in 0.1 M cacodylate buffer before post-fixing in 1% OsO₄ and 1.5% K₄(FeCN)₆. Cells were washed, scraped then suspended in a plug of 1% low melting point agarose. Cell pellets were dehydrated then embedded in Epon resin (TAAB). Ultrathin sections were stained with uranyl acetate and lead citrate, and were viewed using an FEI (Phillips) CM100 electron microscope.

Plasma membrane blebbing (zeiosis) is one of the defining characteristics of apoptosis, but its significance is not understood. In an attempt to understand better the kinetics and consequences of apoptotic blebbing, we monitored apoptosis in a selection of adherent cell-lines by phase contrast, time-lapse microscopy. We confirmed that zeiosis arises at the onset of the apoptotic release phase - the point at which adherent cells begin to retract away from their neighbours and partially detach from the substratum (see supplementary material Movie 1) (Mills et al., 1999). This release phase marks the start of apoptotic execution, beginning shortly after caspase-3 activation (Lane et al., 2002), which in HeLa cells can be as little as 20 minutes after cytochrome c is released from the mitochondria (Goldstein et al., 2000). As reported by others (McCarthy et al., 1997; Mills et al., 1998b), zeiosis lasted for only around 30-40 minutes, at which point the cells became relatively static (see supplementary material Movie 2) (Fig. 1A). Interestingly, when we continued to examine adherent cells beyond this point, we found that they subsequently commenced a second phase of surface blebbing (supplementary material Movie 2) (Fig. 1A). The blebs produced during this late phase were morphologically distinct and fewer were seen per cell than during zeiosis. Once formed, they rapidly enlarged and were seldom retracted (supplementary material Movie 2) (Fig. 1A). In HeLa and SW13 cells, these late phase blebs only rarely detached (data not shown), possibly because they often adhered to the cover-glass surface. As a result of late phase blebbing, cells had an asymmetric profile, possessing typically only 2-5 large blebs at any time (supplementary material Movie 2) (Fig. 1A).

The appearance of both types of surface blebs was timed in relation to phosphatidyl serine (PS) exposure - an important and almost universal feature of apoptotic cell death (Martin et al., 1996) that is readily measured by binding of fluorescent annexin V. The length of time between the onset of zeiosis and PS exposure varied between individual cells (typically 75-125 minutes: see Fig. 1A), consistent with previous studies (Lane et al., 2002). However, PS exposure always occurred after the onset of late blebbing (typically 30-70 minutes later) (Fig. 1A) (see supplementary material Fig. S1). Hence, the order of occurrence of these apoptotic features was always consistent in adherent cells: release → zeiosis → pause → late blebbing → PS exposure (Fig. 1A).

If zeiosis is associated with the release of cells from their neighbours and substratum, then non-adherent cells might omit this phase of blebbing. To test this theory we observed non-adherent cells from the point of apoptotic induction through to PS exposure (Fig. 1B) (see supplementary material Fig. S1). As predicted, non-adherent Jurkat and THP-1 cells displayed only one phase of surface blebbing (Fig. 1B) (see supplementary material Movie 3). Non-adherent cell blebbing was morphologically indistinguishable from the late phase of blebbing observed in adherent cells (compare Fig. 1A,B) (see supplementary material Movies 2, 3), and occurred with similar kinetics relative to PS exposure (compare Fig. 1A,B). In non-adherent cells, PS exposure was first detected 40-60 minutes after the initiation of blebbing (Fig. 1B) (see supplementary material Fig. S1), similar to the delay between the onset of late phase blebbing and PS exposure in adherent cells (Fig. 1A) (see supplementary material Fig. S1). Hence, only the second, late phase of surface blebbing is universally associated with apoptosis, and its timing is fixed relative to other conserved apoptotic events.

If late blebs are an intrinsic feature of apoptosis, then an obvious question is what is their function? One important consequence of apoptosis is the redistribution of condensed chromatin and associated caspase-cleaved autoantigens into apoptotic bodies (e.g. Casciola-Rosen et al., 1994; Leist and Jaattela, 2001). It has been noted previously that zeiosis terminates too soon to feature in the generation of apoptotic bodies (Mills et al., 1999). It seemed possible, by contrast, that the late blebs represent an intermediate in apoptotic body formation, and so might accumulate chromatin. To test this, we imaged apoptotic chromatin dynamics and cellular morphology simultaneously by fluorescence/phase contrast time-lapse microscopy. HeLa cells were transiently transfected with human high mobility group 1 protein fused to YFP (HMBG1-YFP), and were induced into apoptosis by treatment with anisomycin. HMGB1 binds to chromatin, but dissociates during necrosis (due to increased chromatin acetylation) (Scaffidi et al., 2002), providing a useful marker to distinguish between these types of cell death. Throughout zeiosis, chromatin underwent condensation while remaining static at the centre of the cell (Fig. 2) (see supplementary material Movie 4). Importantly, redistribution of chromatin towards the cell periphery was restricted to the late phase of surface blebbing (Fig. 2) (see supplementary material Movie 4). Often, the large, late blebs appeared to form as a direct consequence of chromatin `squeezing' through the cell cortex (Fig. 2) (see supplementary material Movie 4).

Next, we tested whether chromatin entered blebs alone or in combination with other cellular components. Confocal fluorescence microscopy of apoptotic HeLa cells containing clearly identifiable late blebs demonstrated that every one of these structures contained a cortical membrane layer that stained for KDEL receptor, a marker for the ER (Fig. 3A). Conversely, mitochondria, lysosomes and apoptotic fragments of the Golgi apparatus (Lane et al., 2002) were excluded from late blebs in the vast majority of cases (Fig. 3B). Hence, even at such late stages of apoptosis, movement of ER and chromatin into surface blebs is selective and therefore likely to be regulated.

The selective redistribution of ER into apoptotic blebs implies that it is actively remodelled from the reticular network of interphase cells into sheets that underlie the plasma membrane. Further evidence for active remodelling, rather than for the passive movement of ER to fill the volume of apoptotic blebs, was provided by examining cells that were apparently at an intermediate stage of apoptosis (Fig. 4). In these cells the ER could often be seen in extended patches underlying areas of plasma membrane that had not yet formed blebs (Fig. 4, arrowheads). These areas were often devoid of chromatin, suggesting that remodelling of the ER may occur independently of nuclear fragmentation. In some instances, sheets of ER ended abruptly, consistent with them being formed by the active extension or accumulation of the ER along the underside of the plasma membrane (Fig. 4, arrows).

To confirm these light microscopy observations, we examined glutaraldehyde-fixed apoptotic HeLa cells by transmission electron microscopy (TEM). In these cells, we could often identify double membranes enclosing condensed chromatin within large cell surface blebs (Fig. 5A). Their morphology suggested that these membranes were derived from the ER, consistent with the intense and selective staining for ER markers within late blebs observed by fluorescence microscopy, but also contained elements derived from the nuclear envelope: ribosomes could sometimes be seen associated with the outer (but never the inner) membrane leaflet (Fig. 5Aii), and small gaps reminiscent of nuclear pores were sometimes apparent (Fig. 5Aiii). In almost all cases, the double membrane surrounding chromatin in surface blebs formed a continuous layer running just beneath the plasma membrane (Fig. 5A). Interestingly, the outer membrane leaflet was often a remarkably uniform distance from the plasma membrane, such that it occasionally followed the profile of the plasma membrane into short finger-like extensions (Fig. 5Ai). In some blebs, sheets of membrane extended in parallel to the underside of the plasma membrane before terminating (Fig. 5B), while intact loops of membrane could sometimes be found in blebs that apparently did not contain chromatin (data not shown). Although imaging fixed cells by confocal fluorescence microscopy and TEM can provide only a snap-shot of the dynamic events occurring within apoptotic cells, we have been able to discern the following: (1) all late blebs contain membrane derived from the ER and most likely the nuclear envelope, usually in a cortical layer close to the plasma membrane; (2) many but not all late blebs contain condensed chromatin; (3) when present within late blebs, condensed chromatin is always surrounded by cortical ER/nuclear envelope.

Previous studies have implicated actin and myosin in zeiosis (Coleman et al., 2001; Mills et al., 1998b; Sebbagh et al., 2001), condensation (Mills et al., 1999) and fragmentation (Cotter et al., 1992) of apoptotic cells. To investigate the contribution of the cytoskeleton to the movement of ER into late apoptotic blebs, we assessed the effects of cytoskeleton inhibitors on the process. In cells treated for 5 hours with anisomycin, 68.0% of apoptotic cells (identified by chromatin morphology) displayed late apoptotic blebs, with the majority of these (present in 69.2% of apoptotic cells) also staining strongly for cortical ER (Fig. 6A). When Latrunculin A was included at a concentration (1.0 μM) that disrupted actin organisation without causing cells to detach from the cover-glass (supplementary material Fig. S2), very few apoptotic cells displayed late blebs (1.0%) or cortical ER (0.7%) (Fig. 6A). This suggests that actin is required for bleb formation, and not simply for movement of the ER into the blebs.

Given the effect of Latrunculin A on blebbing we predicting that blebbing would be myosin-dependent. Indeed, Blebbistatin, a selective inhibitor of myosin II (Straight et al., 2003) reduced the formation of late blebs in fixed cell assays by greater than 90% (Fig. 6B), as well as preventing zeiosis (data not shown). However, in contrast to its effect on zeiosis, the ROCK I inhibitor Y27632 only partially inhibited the formation of late blebs and cortical ER, which were reduced to 29.1% and 20.4% of control levels, respectively (Fig. 6A). Consistent with this, Y27632 was unable to completely prevent the formation of late blebs when assayed using video microscopy (see supplementary material Movie 5). One possible reason for the partial effect of Y27632 was that myosin light chain is still subject to phosphorylation by Ca²⁺-dependent myosin light chain kinase under these conditions. We therefore tried to examine whether the specific myosin light chain kinase inhibitors ML7 and ML9 blocked blebbing either alone or in combination with Y27632. However, both inhibitors caused healthy cells to round very rapidly unless used at concentrations well below their K_i values (data not shown) and so could not be used reliably to assess apoptosis-specific blebs.

Since movement of ER in interphase cells is critically dependent on microtubule motors, we examined whether inclusion of ER in late blebs was sensitive to the microtubule destabilising agent nocodazole. Surprisingly, inclusion of nocodazole markedly reduced the proportion of apoptotic cells with late blebs (40.2%; Fig. 6A), suggesting that bleb formation itself is microtubule- as well as actin-dependent. Of those cells with blebs, most (35.6%) contained cortical ER. Treatment of cells with both Y27632 and nocodazole inhibited late blebbing and ER redistribution to levels observed for Y27632 alone (27.0% and 18.7%, respectively; Fig. 6A). These data suggest that while actin and myosin II perform a dominant role in late blebbing, the microtubule cytoskeleton also contributes to this process.

Consistent with these functional studies, we observed that both the actin and microtubule cytoskeletons persisted in late stage apoptotic cells. Alexa⁴⁸⁸ phalloidin staining revealed a concentration of f-actin around the periphery of apoptotic HeLa cells, in addition to a mass of f-actin at the cell centre and in retraction cables at the base of the cell (Fig. 7A,B). In some examples, f-actin could be found transecting the base of chromatin-containing late blebs (see arrowhead in Fig. 7A), suggesting that closure of an actin ring structure might be responsible for the eventual scission of a bleb to release an apoptotic body. Myosin II was also found within late blebs (see Fig. 7B), and labelling with an antibody against phospho-myosin regulatory light chain II (phospho Th18/Ser19) suggested that active myosin II accumulates within chromatin-containing late blebs (see Fig. 7C). Furthermore, high-resolution confocal microscopy of apoptotic HeLa cells labelled with anti-tubulin antibodies clearly demonstrated that microtubules persist late into the execution phase of apoptosis, up to and beyond the late blebbing stage (Fig. 7D). This observation was confirmed using a number of cell-lines (including human A431 and SW13-Cl.2, and rat F111 cells), following treatment with various reagents including anisomycin and UV (data not shown).

Recent evidence obtained using isolated nuclei incubated in apoptotic cytosol suggested that caspase-6, which is required for lamin A cleavage, is essential for late stages in chromatin condensation and nuclear fragmentation (Ruchaud et al., 2002). We reasoned that lamin A cleavage and consequent nuclear lamina disruption might, therefore, be necessary for dispersal of chromatin into late apoptotic blebs in intact cells. To test this, we induced HeLa cells to undergo apoptosis in the presence of the cell-permeable caspase-6 inhibitor zVEID.fmk, and analysed apoptotic morphology by fixed and live-cell microscopy. zVEID-treated apoptotic HeLa cells underwent zeiosis in a similar manner to control apoptotic cells, and processed caspase-3 and PARP as normal (data not shown). Surprisingly, however, formation of late blebs was inhibited dramatically by the absence of caspase-6 activity: the proportion of apoptotic HeLa cells with late blebs was reduced from 68.0% (n=434) to 7.8% (n=229) in the presence of zVEID. Confocal imaging of fixed HeLa cells labelled for an ER marker and chromatin suggested that chromatin condensation still occurred, but that both condensed chromatin and the majority of ER remained in the cell body (Fig. 8A). Ultrastructural analysis revealed that condensed chromatin remained confined within an intact nuclear envelope (Fig. 8B,C), and confirmed that large chromatin-containing late blebs were entirely absent from zVEID-treated cells (Fig. 8B). The organisation of ER within zVEID-treated apoptotic HeLa cells was clearly distinct from that in either control interphase or in apoptotic cells. ER tubules were present throughout the cytoplasm, often as whorls, and sometimes in the form of convoluted, interconnected networks (Fig. 8D). Although the ER was often located towards the cell periphery, it was never found in close proximity to chromatin. Together, these data indicate that caspase-6 is required not only for dispersal of chromatin, but also for relocalisation of ER and remodelling of the plasma membrane of the apoptotic cell.

In summary, our findings suggest that formation of late phase apoptotic blebs involves the redistribution of ER and chromatin in an active process that requires caspase-6. Generation of these blebs also requires an intact actin cytoskeleton, active myosin II, and the maintenance of an array of microtubules.

We have examined many of the morphological changes that occur during the execution phase of apoptosis, leading to the inclusion of condensed chromatin and ER within large plasma membrane blebs. We believe that these blebs are the progenitors of apoptotic bodies, based on the following pieces of evidence: (1) the majority of these blebs contain aggregates of condensed chromatin and stain positively with annexin V (i.e. contain PS in the external PM leaflet), which are defining properties of apoptotic bodies; (2) they arise late during the execution phase of apoptosis, and considerably later than cell rounding; (3) they require the presence of an intact actin cytoskeleton and myosin II activity, in common with apoptotic body formation (Cotter et al., 1992; Croft et al., 1995); (4) their formation is sensitive to (though not completely abolished by) the ROCK-I inhibitor, Y27632, as is apoptotic body formation (Coleman et al., 2001); (5) occasionally, these late blebs are released from apoptotic cells, though most remain attached to the cover glass (data not shown); and (6) they are present in all types of apoptotic cells that we have examined following stimulation by a variety of apoptosis inducers [including the cell surface death ligand, TNFα (data not shown)], suggesting that they are a universally conserved feature of apoptosis. Although the most likely functional consequence of forming these blebs is assisting in cellular fragmentation, it is also possible that blebbing is coupled to the presentation of phagocytic signals on the surface of the dying cell. However, our preliminary investigations suggest that phosphatidylserine exposure, one of the best characterised of these phagocytic signals, still occurs when bleb formation is prevented by inhibiting caspase-6 activity (J.D.L., unpublished).

The formation of large chromatin-containing blebs is preceded in adherent cells by a transient phase of plasma membrane blebbing (zeiosis), closely resembling that described by other groups (Coleman et al., 2001; Mills et al., 1998b; Mills et al., 1999). We believe that these early blebs are unlikely to contribute directly to apoptotic body formation, since they occur significantly earlier during the execution phase of apoptosis and are not accompanied by movement of chromatin to the periphery. Indeed, we have not observed their formation in several non-adherent cell lines undergoing apoptosis, indicating that they may be a general feature of apoptosis only in adherent cells. From their morphology, the dynamic nature of their formation and retraction, and their dependence on filamentous actin and myosin II, these early apoptotic blebs appear to be equivalent to the surface blebs produced by cells that are stressed or are undergoing major structural rearrangements such as cell spreading, movement and division (Bereiter-Hahn et al., 1990; Cunningham, 1995; Fishkind et al., 1991; Huot et al., 1998; Straight et al., 2003). Hence, while they no doubt perform an important function in apoptotic cells, their restriction to adherent cells indicates that their main role might be in mediating and/or maintaining the detachment of apoptotic cells from their substratum or neighbours. An incidental consequence of this phase of blebbing being absent from many non-adherent cells is that the time period between the detection of morphological changes and PS exposure is shorter for non-adherent compared with adherent cells. Therefore, PS exposure might appear at first glance to be a particularly early feature of non-adherent cell apoptosis, but a relatively late one in adherent cells.

Two types of apoptotic blebs that could be distinguished by their relative size and content have previously been identified in studies on apoptotic keratinocytes (Casciola-Rosen et al., 1994), though the kinetics of their formation and their mode of generation were not considered in detail. Small (1.3 μm) blebs contained RNA and the ribonucleoprotein autoantigen Ro, while less abundant, larger (2.7 μm) blebs contained chromatin and nuclear-derived autoantigens (Casciola-Rosen et al., 1994). These are likely to correspond, respectively, to the early (zeiotic) and late blebs that we describe herein. Likewise, Bonanno and colleagues (Bonanno et al., 2000) described the formation of large chromatin-containing blebs in late apoptotic U937 cells, though these lacked the characteristic sheets of ER that we have described.

We have now identified several components in the formation of chromatin-containing late blebs. In particular, we have provided evidence for the central involvement of caspase-6, ER, and both actomyosin and tubulin. We anticipated the involvement of the actomyosin cytoskeleton in late blebbing, given its previously identified roles in other forms of cell reorganisation during apoptosis (Coleman et al., 2001; Cotter et al., 1992; Mills et al., 1998b; Sebbagh et al., 2001). More surprising was the finding that microtubules are also involved, since it has been reported that these disassemble at a relatively early stage during apoptosis (see Mills et al., 1999). However, an extensive array of microtubules persists late into apoptosis in all cell-types we tested. In support of this, the presence of a stable subset of microtubules had previously been described in apoptotic cells of a leukaemic T-cell-line (Pittman et al., 1997; Pittman et al., 1994).

What role might microtubules play in late bleb formation? An obvious possibility is that microtubule motors are needed to transport ER and chromatin along microtubules and into blebs. However, since in the absence of microtubules about 50% of cells still form late blebs that contain chromatin and ER, microtubule motors cannot provide the sole means of transporting these components into blebs. It is possible that retraction (squeezing) forces generated by actin and myosin II can drive this relocation as well, or that myosin V plays a role in moving ER. Alternatively, the lack of microtubules may somehow prevent the initiation of late blebbing in some cells. Clearly, further studies are needed to characterise the dynamics and function of the apoptotic microtubule array in the substantial rearrangement of plasma membrane and organelles that occurs at such a late stage of the apoptotic execution phase, just prior to PS exposure.

During the preparation of this manuscript Croft et al. (Croft et al., 2005) also reported the involvement of actomyosin in the movement of condensed chromatin into blebs generated during TNF-induced apoptosis, but did not observe a role for microtubules in this process. The reason for this discrepancy is not clear, but could be related to our finding that nocodazole prevented bleb formation by only about 50%. Hence, while both actomyosin and microtubule cytoskeletons may contribute to chromatin/ER movement, the relative importance of each may vary according to the cell type and/or the mechanism by which apoptosis is induced. In addition, TNF causes p38-dependent phosphorylation of kinesin light chains and subsequent inactivation of kinesin (De Vos et al., 2000), thereby possibly obscuring any role for microtubule motors in bleb formation in TNF-treated cells.

We have shown that active caspase-6 plays a crucial role in allowing the condensed chromatin to disperse to the cell periphery. Although we cannot formally exclude other possibilities, the most likely reason for this essential function of caspase-6 is that chromatin dispersal requires cleavage of lamin A, since lamin A is the only known caspase substrate that is absolutely dependent on caspase-6 for its cleavage (Ruchaud et al., 2002). Previous work suggested that protecting lamins from caspase-mediated cleavage delays nuclear envelope disassembly (Rao et al., 1996). These observations are supported by data obtained using nuclei in apoptotic extracts, which suggested that caspase-6 mediated cleavage of lamin A is required for late stages of chromatin condensation, including the formation of spherical chromatin complexes that are released from the lamina (Ruchaud et al., 2002). Our data suggest that this is also the case in intact cells and, importantly, indicate that additional event(s) necessary for chromatin mobilisation to the cell periphery, that are not observed when using isolated nuclei, occur downstream of destabilising the nuclear lamina. Indeed, Croft and co-workers (Croft et al., 2005) have shown independently that reduction of the tensile strength of the nuclear lamina by loss of lamin A allows the actin-dependent dispersal of chromatin to the cell periphery.

Croft and co-workers (Croft et al., 2005) have suggested that destabilisation of the nuclear lamina enables the actomyosin cytoskeleton to tear the apoptotic nucleus apart, and that this process is required to generate apoptotic bodies. Our data are consistent with this, though also suggest that microtubules play a role in generating apoptotic bodies. In addition, our data indicate that selective inclusion of ER-derived membrane, often enveloping chromatin, is an important component in apoptotic body formation. We suggest that destabilisation of the nuclear lamina, involving cleavage of lamin A, might allow the nuclear envelope to mix with the rest of the ER, thereby allowing the condensed chromatin entry into extensive domains of this large organelle. Interestingly, such a scenario is reminiscent of the exchange of membrane between nuclear envelope and ER that occurs upon phosphorylation and subsequent release of lamins during mitosis (Heald and McKeon, 1990; Yang et al., 1997). It could be argued that inclusion of ER in apoptotic blebs is due simply to it being passively carried along with condensed chromatin as a re-sealed `nuclear envelope'. However, we feel that this is unlikely, since sheets of ER can be found within blebs that lack chromatin and, in addition, the structural reorganisation of the ER and its close association with the plasma membrane point towards an active process. Similar reorganisation of ER from a reticular to lamellar structure has been observed in pancreatic acinar cells that have been induced to secrete by exposure to dexamethasone (Rajasekaran et al., 1993), though the underlying mechanism was not investigated. Conversion of ER to lamellae also occurs in nocodazole-treated cells, though this is accompanied by its coalescence towards the cell centre (Terasaki et al., 1986).

One explanation for the remodelling of ER observed in apoptotic cells is that alterations in the balance of membrane fission and fusion lead to a transient vesiculation of the ER reticulum and the subsequent re-building of cortical membrane sheets. Caspase-dependent cleavage of rabaptin-5 (Cosulich et al., 1997), GRASP65 (Lane et al., 2002), golgin 160 (Mancini et al., 2000), giantin and syntaxin 5 (Lowe et al., 2004) contribute to the disruption of vesicle fusion within both the endosomal and secretory pathways during apoptosis. However, the remodelling of ER is more likely to involve reversible, and therefore caspase-independent, changes in fusogenicity. Clearly, further research will be required to understand the mechanism of ER remodelling and its contribution to apoptotic body formation.

In summary, our data support a model in which caspase-6 mediated release of condensed chromatin from the nuclear lamina (Ruchaud et al., 2002) permits the microtubule- and actin-dependent movement of chromatin towards the cell periphery, invariably enclosed within a membrane that is derived from the nuclear envelope/ER. This key event takes place exclusively during late blebbing, leading to the assembly of large surface protrusions that contain membrane-bound condensed chromatin. Interference with any step during this process is likely to affect the ability of the apoptotic cell to relocate chromatin into apoptotic bodies, which is important for the efficient clearance of apoptotic cells and their remnants.

This work was supported by BBSRC project grants 34/C16819 and 34/C11195, MRC COGG grant G9722026, and a Wellcome Trust Career Development Fellowship to J.D.L. (No. 067358). We thank Debbie Carter and Ginnie Tilly for assistance with the EM. We thank the Medical Research Council for providing an Infrastructure Award to establish the School of Medical Sciences Cell Imaging Facility at Bristol University.