Apoptosis is an evolutionarily conserved process in multicellular organisms that leads to the death and subsequent removal of redundant or excess cells. In the dying cell, a family of cysteine-proteases called caspases (reviewed in reference1) are responsible for the ‘execution’ phase which is characterised by morphological changes, including cell contraction and dynamic membrane blebbing (Figure 1), one of the earliest described and most obvious aspects of apoptotic cell death.2 Contractile force generated by actin-myosin II cytoskeletal structures has been implicated as the driving power behind cell contraction and the formation of membrane blebs and apoptotic bodies (reviewed in reference3). Ultimately, the dead cell is packaged into membrane-clad apoptotic bodies that facilitate uptake by neighbouring cells or by specialised phagocytic cells. Dynamic re-arrangements of the actin cytoskeleton in the phagocyte are required for the internalisation of apoptotic cell fragments. Recent research has revealed the importance of signal transduction pathways controlled by Rho family GTPases in regulating the marked changes in cell morphology observed in the processes of apoptosis and phagocytosis.

Figure 1
figure 1

Changes to cell morphology and actin cytoskeletal structures associated with apoptosis. 1 Normal; 2 Early apoptosis; 3 Blebbing; 4 Late apoptosis; 5 Engulfment of apoptotic bodies by phagocyte

The Rho GTPases are a family of proteins (RhoA, RhoB, RhoC, RhoD, RhoE/Rnd3, RhoG, RhoH/TTF, Rnd1, Rnd2, Rac1, Rac2, Rac3, Cdc42/G25K, Wrch-1, TC10, TCL, Chp, Rif) that act as molecular switches in intracellular signal transduction pathways (reviewed in reference4). Rho activation results from a combination of reduced association with GDP-Dissociation Inhibitors (GDIs) and enhanced exchange of GDP for GTP promoted by Guanine nucleotide Exchange Factors (GEFs). Activated GTP-bound Rho proteins then transduce signals to downstream effector proteins and finally, through association with GTPase Accelerating Proteins (GAPs), return to the inactive GDP-bound form by hydrolysis of the bound GTP (Figure 2). One of the key functions of Rho proteins is to regulate the architecture of the actin cytoskeleton (reviewed in reference5). The best characterised proteins of this family are RhoA, which leads to the formation of actin stress fibres and actin-myosin II contractile force generation,6,7 Rac1, which promotes the formation of lamellipodia and membrane ruffles8 and Cdc42/G25K, which drives the formation of actin-rich filopodia.9,10

Figure 2
figure 2

Regulation of Rho GTPase cycle. Cycling of Rho GTPase between inactive GDP-bound and active-GTP bound states. GDI, GDP-Dissociation Inhibitor; GEF, Guanine nucleotide Exchange Factor; GAP, GTPase Accelerating Protein

One theory that attempted to explain the cell contraction, membrane blebbing and apoptotic body formation observed in apoptotic cells was that caspase-mediated proteolysis of structural and cell adhesion proteins (Table 1) leads to a release from points of cell attachment followed by collapse of the cell. However, numerous lines of evidence have demonstrated that the first phase of cell contraction and membrane blebbing (Figure 1) is a dynamic process associated with, and dependent upon, the presence of filamentous actin,11,12,13,14,15,16,17,18,19,20,21,22,23 increased myosin light chain (MLC) phosphorylation15,16,19,22,24 and myosin ATPase activity.15,16 After the initial phase of contraction and blebbing, a second phase of actin filament disassembly occurs via depolymerisation18,20 and possibly through caspase-mediated cleavage of actin monomers.25,26,27

Table 1 Cytoskeletal, structural and regulatory proteins cleaved by caspases during apoptosis

First phase: contraction and blebbing

Given the well-characterised effects of RhoA on promoting actin filament bundling and actin-myosin II contractile force generation, it was proposed that the activation of RhoA is responsible for the contraction observed in apoptotic cells.3 Consistent with this proposal, introduction of a constitutively-active form of RhoA was shown to be sufficient for cell contraction and membrane blebbing.28,29

Resent research, however, has shown that activation of RhoA is unlikely to be a general mechanism responsible for apoptotic contraction and blebbing. RhoA was not activated in response to a pro-apoptotic stimulus in NIH 3T3 or Swiss 3T3 cells (EA Sahai and MF Olson, unpublished observation and reference23). In addition, inhibition of Rho with C. botulinum C3 toxin did not inhibit membrane blebbing23,24 or MLC phosphorylation.24 Signalling downstream of Rho, however, was essential, as pharmacological inhibition of the Rho effector kinase ROCK prevented membrane blebbing in a range of cell types,23,24 which was accompanied by diminished MLC phosphorylation.24 Active ROCK I was shown to be sufficient for cell contraction and membrane blebbing in NIH 3T3 mouse fibroblasts, NIE-115 human neuroblastoma and Jurkat human T cells.23,24,30 Taken together, these data are consistent with increased actin-myosin II contractile force being driven by Rho-independent ROCK activity that results in membrane blebbing during apoptosis.

In collaboration with other Rho effector proteins, ROCK contributes to agonist-induced changes to the actin cytoskeleton without necessarily producing dramatic contraction and blebbing.31,32 The ROCK I and ROCK II isoforms bind to Rho–GTP, which activates the ROCK catalytic domain by displacing the carboxy-terminal autoinhibitory domain (Figure 3).33,34,35 Deletion of the inhibitory region increases kinase activity both in vitro and in vivo.34,36,37,38 During apoptosis, ROCK I, but not ROCK II, is cleaved by caspase-3 at a conserved sequence that removes the autoinhibitory domain.23,24 The truncated kinase has an eightfold higher specific activity in vitro relative to full-length protein in the absence of Rho.23 The enhanced kinase activity is sufficient to drive caspase-independent cell contraction and membrane blebbing (Figure 3),23,24 consistent with a direct effect of ROCK on the development of the apoptotic morphology.

Figure 3
figure 3

ROCK activation and phosphorylation of downstream targets. Loss of autoinhibition by Rho binding, or caspase cleavage results in kinase activation. ROCK phosphorylates a number of substrates that regulate actin-myosin-II contractile force generation. MLC, Myosin Light Chain

ROCK activation positively contributes to actin-myosin force generation through the phosphorylation of a number of downstream target proteins (Figure 3). ROCK-dependent activation of LIM kinases-1 and -239,40,41,42 results in phosphorylation and inactivation of the actin-severing protein cofilin, thereby stabilising filamentous actin. ROCK also directly phosphorylates MLC,43,44,45 and phosphorylates and consequently inhibits MLC phosphatase.46,47 ROCK phosphorylation of calponin48 and CPI-1749 relieves their inhibition of myosin ATPase activity thereby promoting increased actin-myosin contractile force. Taken together, ROCK activation leads to a concerted series of events that promotes actin filament stabilisation, increased interaction with myosin and contractile force generation, which together drive cell contraction.

Membrane blebbing occurs where the strength of interactions that tether the plasma membrane to cytoskeletal structures is exceeded by the hydrodynamic force within the cell.50,51 Application of negative pressure to the exterior of cells allows bleb protrusion powered by the positive interior pressure,52 indicating that a pressure gradient between inside and outside of a cell is sufficient for blebbing. Alternatively, a reduction in the strength of interaction between the plasma membrane and the cytoskeleton may promote membrane blebbing, for example melanoma cells deficient in filamin, which links actin filaments to plasma membrane-associated proteins, bleb continuously.53 During apoptosis, caspase-mediated proteolysis of structural and regulatory proteins (Table 1) diminishes the interactions that tethers the plasma membrane to the cytoskeleton and allows the increased hydrodynamic forces generated by ROCK-induced cell contraction to drive bleb protrusion at points of weakness (Figure 1). Although required for apoptotic body formation,11,23 blebbing can be sustained for prolonged periods of time and will not directly lead to a breakdown of the cell.15,23,24,27,30 Therefore, additional events are required for the blebbing cell to be broken down into apoptotic bodies.

Second phase: breakdown of actin structures and apoptotic body formation

After the initial phase of cell contraction and membrane blebbing driven by ROCK-mediated actin-myosin contractile force generation, there is a second phase in which actin filaments are depolymerised.18,20 Signalling downstream of Rho GTPases may contribute to both phases of actin reorganisation. During apoptosis, caspase-mediated cleavage of the Rho effector kinases PRK1 (also known as PKN)54 and PRK255 as well as the Rac/Cdc42 effector kinase PAK2 (also known as γ-PAK)56,57 releases constitutively active kinase fragments. Catalytically active PRK1/PKN has been reported to promote actin stress fibre disassembly,58 possibly resulting from the phosphorylation of the actin binding protein α-actinin as well as the phosphorylation of monomeric actin.59 A C-terminal fragment of PRK2 released by caspase-mediated proteolysis may feed forward to promote apoptosis by inhibiting the catalytic activity of the anti-apoptotic protein kinase Akt/PKB and reduce its inhibitory phosphorylation of the pro-apoptotic Bcl-2 family member Bad.60 PAK isoforms also have been shown to lead to actin stress fibre disassembly,23,61,62,63 possibly through the phosphorylation of the myosin-II heavy chain,64 which is thought to promote actin-myosin destabilisation,65 and through the phosphorylation and consequent inhibition of the MLC kinase.66,67 Overexpression of a dominant-negative form of PAK2 blocked Fas-induced apoptotic body formation in Jurkat T cells suggesting that PAK-mediated effects on actin filament depolymerisation and dismantling of cytoskeleton structures may be required for the final breakdown of the apoptotic cell.56 Interestingly, PAK1 and PAK2 isoforms have been shown to protect cells from apoptosis through the phosphorylation of the pro-apoptotic Bad protein.68,69,70 Thus, in some situations the activation of PAK2 by caspase-cleavage may antagonise both the morphological and biochemical events in a cell not fully committed to apoptosis.

Caspase-cleaved Rho GTPase signalling proteins

In addition to the proteins listed above, four additional proteins involved in Rho GTPase signalling have been identified as being caspase-cleaved in apoptotic cells. Given the importance of the Rho GTPase family in the regulation of the actin cytoskeleton, any significant changes to their signalling activities would likely influence the apoptotic morphology.


The Rho family GTPases Cdc42 and Rac1 were shown to be cleaved by caspases-3 and -7 in a variety of cell lines during Fas-induced apoptosis.71 The cleavage occurs at a position that separates the N-terminal portion of the protein containing the principal effector-interaction domain from the C-terminal portion responsible for the essential localisation of the protein to membranous structures. Therefore, the most likely outcome of Cdc42 and Rac1 cleavage is the termination of downstream signalling. It has been suggested previously that Cdc42 and Rac1 signalling promote cell survival through the PAK-mediated phosphorylation of Bad68,69,70 and through the activation of NF-κB.72 Therefore, the cleavage of Cdc42 and Rac1 may further the apoptotic process by eliminating their normal function as pro-survival signalling proteins. In addition, by eliminating the actin filament destabilising actions of Cdc42 and Rac161,63,64 the effects of ROCK I on promoting actin-myosin II cell contractility would be accentuated.


In haematopoietic cells, Rho GTPases may be activated during apoptosis when they are released from the inhibitory actions of the D4–GDI protein, a GDI for Rho family proteins, which is cleaved by caspase-3 (Figure 2 and Table 2).73,74,75 Since RhoA, Rac1 and Cdc42 all bind to D4–GDI in vitro, it is not clear what the effect of D4–GDI cleavage and potential simultaneous activation of these multiple pathways during apoptosis would be.

Table 2 Rho GTPase signalling proteins cleaved by caspases during apoptosis


The haematopoietic-specific RhoGEF Vav1, which is essential for normal T and B cell function,78,79 was shown to be cleaved by caspase-3 at a site that is also conserved in the related Vav2 and Vav3 proteins.80 The consequences of caspase-mediated cleavage of Vav1 are not clear, however, ectopic expression of N-terminally deleted versions of Vav1 was sufficient to induce Rho-dependent actin stress fibres (MF Olson, unpublished observations and references81,82). Thus, it is possible that cleavage of Vav1 serves to disrupt signalling from the T and B cell receptors and to promote rearrangements of the actin cytoskeleton that lead to the morphological changes in apoptotic immune cells.

Given the large and complex pattern of protein degradation during apoptosis (see Tables 1 and 2), some apparently opposing activities would seem to be simultaneously in effect, for example the increased actin-myosin interactions and contractile force generation driven by cleaved ROCK I versus the actin-myosin disassembly and cell spreading from cleaved PAK2. These seemingly contradictory actions are likely influenced by differences in when and where a given protein is cleaved. In addition, signal intensity would determine which outcome predominates and the ultimate effects would change as relative signal strengths shifted over time. Many protein cleavage events probably do not contribute directly to the morphological changes or to the apoptotic process, rather these proteins are likely ‘innocent bystanders’ caught up in caspase-crossfire.

Rho GTPases in phagocytosis

For normal tissue homeostasis to be maintained in the presence of cells undergoing apoptosis, the associated cell debris must be efficiently removed and destroyed. Therefore, programmed cell death in vivo must occur hand-in-hand with corpse clearance; for this reason phagocytosis is essential as the final step of the apoptotic process. Removal of dying cells prior to their lysis prevents the exposure of surrounding cells and tissue to potentially toxic cell contents, thus protecting from inflammatory injury.83,84 Phagocytosis of cell corpses occurs by both ‘professional’ phagocytes such as macrophages and neutrophils and ‘amateur’ phagocytes that include epithelial cells and fibroblasts.

Actin cytoskeleton

A universal requirement for the process of phagocytosis is the integrity of the actin cytoskeleton. The requirement for polymerised actin is highlighted by the inhibition of phagocytosis by actin disrupting agents such as cytochalasins.85,86,87 Bound particles are surrounded by a phagocytic cup lined by newly polymerised actin microfilaments that provide the driving force for engulfment and subsequent phagosome formation (Figure 1).88,89 Following internalisation of the particle, F-actin is depolymerised and dissociated from the phagosome allowing subsequent endosomal fusion and lysosomal degradation of the ingested particle.90

Labelling cell corpses and recognition by phagocytes

Efficient clearance of apoptotic bodies also depends upon their being labelled with ‘eat me’ signals, which appears to be an integral part of the apoptotic process.91 The best characterised engulfment signal is the appearance of phosphatidylserine on the outer leaflet of the plasma membrane, an event that may be a key determinant in phagocyte recognition through its binding to an evolutionarily-conserved receptor.92 Less well characterised phagocytic markers include sites that bind ‘bridging’ molecules from the extracellular fluid such as β2 glycoprotein 1,93 thrombospondin94 and the complement factors iC3b and C1q.95,96,97

Engulfment signals such as phosphatidylserine and bridging factors mediate phagocytosis through their interaction with specific receptors present on the approaching phagocyte. A variety of engulfment receptors involved in the clearance of dying cells have been identified including phagocyte lectins, integrins, scavenger receptors and macrosialin.89,98 Other factors may also influence the engulfment process, such as local changes in the lipid composition of the phagocyte plasma membrane. These changes in lipid composition are controlled in part by the ABC1 transporter,99 which is required for efficient recognition of apoptotic cells.100

Genetic analysis of phagocytosis in Caenorhabditis elegans

ABC1 is the mammalian homologue of the ced-7 gene from Caenorhabditis elegans, one of six genes identified in genetic screens for mutants defective in the clearance of apoptotic cells.101,102,103 These genes comprise two parallel pathways organised into the epistatic groups ced-2, ced-5, ced-10 and ced-1, ced-6, ced-7 (Table 3).102 Of the other components of the ced-7 pathway, ced-1 encodes a transmembrane receptor with homology to human SREC (Scavenger Receptor from Endothelial Cells)104 and ced-6 encodes a functionally conserved signalling adaptor molecule composed of a phosphotyrosine-binding domain and potential Src-homology domain 3 (SH3) binding sites.105,106 How the ced-1/6/7 pathway conveys signals downstream of activated engulfment receptors to the subsequent internalisation of the apoptotic cell is poorly understood. This is in contrast to our relatively thorough understanding of the ced-2/5/10 pathway, the components of which are not only required for cell corpse engulfment, but also for migration of the distal tip cells of the nematode gonad.107,108 This ‘engulfment cassette’ comprises a conserved signalling pathway previously implicated in regulation of the actin cytoskeleton and cell migration in mammalian cells

Table 3 C. elegans genes involved in phagocytosis of cell corpses and their mammalian homologues

Mammalian homologues of ced engulfment genes

Ced-2 encodes the homologue of the mammalian SH2/SH3-containing adaptor protein CrkII, ced-5 is homologous to the large adaptor protein DOCK180, whilst ced-10 is the nematode orthologue of Rac1 (Table 3).107,108 DOCK180 has been implicated in the activation of Rac since it has been shown to bind Rac–GDP, but not Cdc42 or RhoA, and overexpression of membrane-targeted DOCK180 increases Rac–GTP levels in 293T cells.109,110,111 Membrane targeting of DOCK180 induces spreading of NIH 3T3 cells112 that is dependent on Rac function.109 Although DOCK180 can increase Rac–GTP levels, it contains no discernible Dbl-homology GEF domain and is therefore unlikely to act as a Rac GEF itself.109 Instead, DOCK180 may act as an adaptor that recruits Rac to the plasma membrane where it becomes activated (Figure 4). Although the Rac exchange factor(s) that mediates DOCK180-dependent Rac activation is unknown, it has been shown that DOCK180 can enhance the activation of Rac by the GEF Vav1.109

Figure 4
figure 4

Mammalian homologues of the Ced-2, -5, and -10 engulfment genes form a signalling cassette that drives actin polymerisation at the site of cell corpse ingestion. Left, Cell undergoing apoptosis displays ‘eat me’ signal on outer leaflet of plasma membrane. Minimal actin structures in the resting phagocye due to low levels of active Rac–GTP. Right, upon ligation of the ‘eat me’ signal to an engulfment receptor such as αVβ3 integrin expressed on the approaching phagocyte, a number of events occur that result in actin polymerisation. Integrin ligation results in tyrosine phosphorylation of the adaptor protein p130Cas, which subsequently recruits CrkII, and the Rac-GDP binding protein DOCK180, to the plasma membrane. Now proximal to its GEF, Rac exchanges GDP for GTP resulting in actin polymerisation at the ingestion site. The newly polymerised actin filaments form pseudopods that migrate around the cell corpse and eventually fuse to complete the process of engulfment. GEF, Guanine nucleotide Exchange Factor

Translocation of DOCK180/Rac to the plasma membrane is likely achieved through the interaction of DOCK180 with the amino terminal SH3 domain of the ced-2 homologue, CrkII. CrkII localises to focal adhesions upon integrin stimulation, a process that is dependent on its interaction with a protein called p130Cas.113,114 Integrin stimulation results in tyrosine phosphorylation of p130Cas115,116,117 and subsequent recruitment of the CrkII/DOCK180/Rac complex to focal adhesions109,110,118 via the CrkII SH2 domain (Figure 4).119,120 In addition, coexpression of p130Cas and CrkII enhances the DOCK180-dependent activation of Rac and membrane spreading.109,110 It would also be intriguing to determine whether the MER receptor tyrosine kinase, which has been suggested to be essential for the phagocytosis and clearance of apoptotic cells by macrophages, signals to the CrkII/DOCK180/Rac pathway.121

The mammalian homologues of the ced-2, -5 and -10 genes previously implicated in adhesion-dependent signalling and migration as described above, have recently been confirmed as important players in the clearance of apoptotic mammalian cells.122,123,124 Interestingly, it would appear that integrin-mediated formation of the CrkII/DOCK180 complex and subsequent Rac activation are common features of both cell adhesion and phagocyte signalling. Integrins such as αvβ5 and αvβ3 have previously been implicated as engulfment receptors in professional and amateur phagocytes.125,126,127 For example, phagocytosis of cell corpses by dendritic cells is mediated by the αvβ5 integrin,125 a process that was subsequently shown to depend on recruitment of the p130Cas/CrkII/DOCK180 complex and subsequent Rac activation.122

Other evidence exists that also implicates integrin-mediated Rac activation in the phagocytosis of dying mammalian cells. The engulfment of apoptotic Baf-3 cells by bone marrow-derived macrophages was dependent on the αvβ3 integrin and could be blocked by dominant negative versions of Rac and Cdc42.124 Interestingly, whereas inhibition of Rac or Cdc42 signalling significantly blocked phagocytic uptake, inhibition of Rho actually enhanced the clearance of cell corpses.124

There is growing evidence to suggest that members of the Rho GTPase family play a universal role in the reorganisation of the actin cytoskeleton during all forms of phagocytosis (reviewed in reference128). For instance, particles opsinised by IgG are recognised by the Fcγ family of receptors for the constant region of immunoglobulin and are subsequently internalised in a Rac and Cdc42-dependent manner. Inhibition of Rac function prevents pseudopod fusion and phagosome closure during FcεRI-mediated phagocytosis, whilst inhibition of Cdc42 interferes with pseudopod extension.129,130,131 In contrast, internalisation of particles opsinised by complement fragments via binding to the αMβ2 integrin requires RhoA function, but not Rac and Cdc42.130 During both Fcγ- and αMβ2-mediated phagocytosis, the actin-nucleating Arp2/3 complex accumulates at the ingestion site to promote actin neo-polymerisation.132

Are Rho GTPases required downstream of the ced-1/6/7 pathway?

Considering the apparent general requirement for Rho GTPase signalling during phagocytosis it seems likely that they also play a role downstream of the ced-1/6/7 pathway. It is unlikely that ced-1 itself signals through ced-2/5 to ced-10 activation and actin-re-organisation however, since mutations to any of the ced-2, -5, or -10 genes significantly co-operate with ced-1 mutations in the inhibition of engulfment.102 However, the possibility that components downstream of ced-1 interact with ced-10 was raised by the observation that overexpression of ced-6 could partially suppress the engulfment defect of ced-10 mutants.106 Since ced-6 has been proposed to act as an adaptor molecule and does not possess an obvious catalytic domain, it is unclear whether it acts upstream or downstream of ced-10.105,106,133,134 It is possible that the ced-1 pathway signals through an alternative Rac-like GTPase even in the presence of a ced-10 mutation since C. elegans express a Rac2 homologue135 and a GTPase termed Mig-2 that is 64% identical to ced-10.136 Interestingly, the scavenger receptor MARCO, which is involved in the macrophage clearance of bacteria, was recently shown to induce morphological changes associated with rearrangement of the actin cytoskeleton when overexpressed.137 These MARCO-induced changes were partially inhibited by dominant negative Rac, but not Cdc42. Perhaps a similar pathway exists downstream of other scavenger receptors or the C. elegans homologue ced-1 to regulate the actin cytoskeleton and phagocytosis of cell corpses via Rho GTPases.

Intriguing recent work suggests that besides their role in phagocytosis, the engulfment genes of C. elegans also may play an active role in the apoptotic process in the dying cell. A screen for genes that could synergise with a partial loss of function of the ced-3 caspase identified all the previously characterised engulfment genes.138,139 A model has been proposed whereby activation of engulfment pathways in the phagocyte promotes a feed forward mechanism that ensures the demise of the associated cell. Since Rho GTPases such as Rac are important effectors in engulfment signalling pathways, this raises the remarkable possibility that Rho proteins may be involved in the homicide of a cell previously thought to have committed suicide.


The actin cytoskeleton rearrangements in both the apoptotic and the phagocytic cell result from activation of signalling pathways associated with Rho GTPases. However, the mode of activation is entirely different in each case. During apoptosis, caspase-mediated cleavage of ROCK I gives rise to a constitutive Rho-independent signal that generates actin-myosin contractile force, membrane blebbing and formation of apoptotic bodies. In marked contrast, phagocytosis requires precise spatio-temporal regulation of Rac and Cdc42 to co-ordinate the dynamic actin remodelling necessary for the engulfment of the apoptotic cell.

It is evident that the actin rearrangements that accompany phagocytosis are critical for the removal of dying cells, thus preventing the exposure of surrounding cells and tissue to potentially toxic cell contents and an inappropriate immune response. However, it is not clear whether the membrane blebbing of an apoptotic cell plays an active part in its subsequent phagocytosis or if, in fact, it has any in vivo physiological purpose at all. Blebbing does appear to be essential for the formation of apoptotic bodies, which may facilitate recognition and clearance by phagocytes. Alternatively, the eventual breakdown of the cell into apoptotic bodies may aid the engulfment process. In addition, the contractile forces generated during the execution phase of apoptosis may be important for pulling adjacent cells with strong cell–cell contacts together, thus maintaining proper tissue organisation and integrity. Having identified the biochemical pathway responsible for the cell contraction and membrane blebbing during apoptosis, it will now be possible to finally determine the physiological function of these processes and answer the question ‘What is blebbing for?’