To induce cytoskeletal changes, pathogenic microbes must ensure delivery of effector molecules onto or into host cells. Effectors are usually proteins that interface with and influence host-cell pathways, and can facilitate disease. As viruses have no metabolic activity outside their hosts, effectors must already be packaged in the virion or expressed from the viral genome from within host cells. Bacteria use several methods to deliver effector proteins to the host cell. Some effectors, such as toxins, are secreted by bacteria in the vicinity of the host cell, where they bind specific receptors and are taken up by endocytosis1. Other effector proteins can facilitate their own uptake with pore-forming subunits or autotransporter domains. Some Gram-negative pathogenic bacteria have acquired sophisticated 'molecular syringes', such as type III or type IV secretion systems, which are multisubunit molecular machines that span the bacterial and host membranes and translocate effectors directly into host cells. Protozoan parasites use secretory organelles, such as the micronemes, rhoptries and dense granules of Plasmodium and Toxoplasma, to deliver microbial products to the host–pathogen interface.

Forced entry

Viruses and some bacteria and protozoan pathogens are obligate intracellular parasites that can only replicate inside their host cells. Other pathogens can replicate extracellularly, but choose an intracellular lifestyle to obtain a favourable niche within the host. To gain access into non-phagocytic cells, or to enter into a protected niche within phagocytic cells, microbes have developed dedicated strategies that mediate pathogen invasion. Because the cytoskeleton controls surface remodelling events such as phagocytosis2 and macropinocytosis3, it is an obvious target for such invasion mechanisms. Many Gram-negative bacteria utilize a type III secretion system (TTSS) and associated effectors to mediate invasion into non-phagocytic cells4. Salmonella and Shigella species direct their own uptake into host cells using a multifaceted approach, coordinating several signalling pathways that converge to induce transient, actin-rich membrane ruffles that engulf the infecting bacteria (Fig. 1a, b).

Figure 1: Host-cell invasion.
figure 1

a, Salmonella invasion. Entry into host cells is mediated by the Salmonella pathogenicity island-1 (SPI-1) type III secretion system (TTSS) and its effectors. Membrane attachment to the cortical actin cytoskeleton is loosened by SigD/SopB, an inositol phosphatase that acts on PtdIns(4,5)P2. SopE and SopE2 enhance Cdc42 and Rac1 activity directly by acting as guanine-nucleotide-exchange factors. SipA and SipC alter cytoskeletal structure, SipC by nucleating actin and initiating polymerization and SipA by binding actin and modulating actin bundling. These cytoskeletal rearrangements are downregulated by the GAP (GTPase-activating protein) activity of SptP, which inactivates Cdc42 and Rac. SigD also is involved in sealing invaginating regions of the plasma membrane to form intracellular vacuoles. b, Shigella invasion. IpgD loosens membrane/cytoskeletal attachments in a mechanism similar to Salmonella SigD. VirA binds tubulin and promotes microtubule destabilization. This stimulates increased microtubule growth and causes Rac1 activation. Rho and Cdc42 are also activated during Shigella entry, through mechanisms that are still unclear. Activation of the small GTPases triggers filopodia and lamellipodia formation in the vicinity of the bacteria. IpaA binds vinculin and is involved in the transformation of filopodial extensions into membrane leaflets, and IpgD is probably involved in vesicle sealing. c, Trypanosoma cruzi invasion. Binding of an unidentified parasite product to host cells causes an increase in intracellular calcium. This calcium flux destabilizes the cortical actin cytoskeleton and induces the microtubule-mediated recruitment and fusion of lysosomes to the plasma membrane. Trypanosoma cruzi uses this membrane to form a vacuole.

Salmonella directly activate Rho GTPases using secreted effectors and a TTSS encoded within the Salmonella pathogenicity island 1 (SPI-1) locus. The SPI-1-secreted effectors SopE and SopE2 act as guanine-nucleotide-exchange factors (GEFs) for the small GTPases Cdc42 and Rac5. Structural analysis reveals that, despite a lack of sequence and architectural similarity, SopE and eukaryotic GEFs induce virtually identical conformational changes in their target Rho proteins, providing an example of bacterial mimicry of a normal cellular process through convergent evolution6. Additional SPI-1-translocated effectors of Salmonella affect actin dynamics during the invasion process. SipA binds to and stabilizes actin, and SipC, which forms part of the TTSS delivery pore, nucleates and bundles actin while anchored in the host cell membrane5.

Salmonella also alters the actin cytoskeleton through manipulation of phosphoinositides. The plasma membrane is intimately associated with the actin cytoskeleton, and this interaction depends on phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P2)7. SigD/SopB is an SPI-1-translocated inositol phosphatase that induces the rapid disappearance of PtdIns(4,5)P2 from invaginating regions of the membrane during Salmonella invasion. This increases elasticity to facilitate the remodelling of the plasma membrane associated with Salmonella entry8. PtdIns(4,5)P2 has also been implicated in vesicle fission during the creation of phagosomes and clathrin-coated vesicles, and accordingly, SigD also is involved in sealing plasma membrane invaginations to form bona fide vacuoles8. After invasion, an additional SPI-1 effector, SptP, acts as a GTPase-activating protein (GAP) for Cdc42 and Rac1, thereby inactivating these G proteins and returning cell morphology to a relatively normal state9.

SptP is a bifunctional protein, with its GAP domain at the amino terminus, and a protein tyrosine phosphatase domain at the carboxy terminus5. A potential target for the tyrosine phosphatase activity of SptP is the intermediate filament protein vimentin, which is recruited to the membrane ruffles stimulated by Salmonella10. Other studies have also identified another intermediate filament protein involved in Salmonella entry: SipC binds cytokeratins and expression of dominant-negative cytokeratin-18 inhibits Salmonella entry into HEp2 cells11. As virtually nothing is known about the disruption of intermediate filaments by pathogens, this emerging area of investigation shows promise for future advances.

In a mechanism very similar to Salmonella, Shigella uses a TTSS to deliver effectors that activate Cdc42 and Rac12 and deplete PtdIns(4,5)P2 (ref. 13) to mediate entry into non-phagocytic cells (Fig. 1b). However, compared to Salmonella, Shigella utilizes an additional means to affect the actin cytoskeleton. The Shigella effector VirA binds tubulin and promotes microtubule destabilization14. This stimulates increased microtubule growth, which has recently been shown to activate Rac1 in other systems15. Indeed, injection of VirA directly into cells induces membrane ruffling, which can be inhibited by dominant-negative Rac1 (ref. 14). The molecular mechanisms linking microtubule growth to Rac activation are still being elucidated and VirA may be a valuable probe for such studies.

Viruses enter cells through fusion of their membrane with the host-cell plasma membrane or via endocytosis16. But this is not a passive process, as binding of viral proteins to cellular receptors can initiate specific signalling cascades17. Several viruses bind to and initiate signalling through integrins to induce cellular uptake17, and although the role of actin in endocytosis is still controversial18, actin rearrangements are necessary for the entry of several viruses, including adenovirus and certain forms of vaccinia16. In addition, emerging evidence shows that many pathogens enter host cells through caveolae or lipid rafts, and viruses are no exception19. Studies of SV40 entry into cells have recently revealed a new microtubule-dependent transport pathway from plasma-membrane caveolae to the endoplasmic reticulum, the details of which are still under investigation20. As such, the study of viral entry into host cells represents a fertile area for future discoveries.

The significantly larger protozoan parasites such as Plasmodium falciparum and Toxoplasma gondii do not rely on host-cell actin-dependent internalization machinery for host-cell invasion. Instead, they utilize an actinomyosin motor present in their own cytoskeleton to generate the motile force necessary to actively propel themselves inside cells21,22. Host-cell actin polymerization is necessary, however, for the invasion process of some protozoan parasites such as Cryptosporidium parvum23, but nothing is known about the mechanisms used by this parasite to initiate its cellular uptake. Trypanosoma cruzi invasion is actually enhanced by inhibitors of actin polymerization. This parasite uses a novel calcium-regulated microtubule-mediated pathway that directs recruitment and fusion of lysosomes to the plasma membrane of host cells. T. cruzi then co-opts this lysosomal-derived membrane to form a parasitophorous vacuole inside the host cell24. Binding of T. cruzi to cells causes a transient increase in intracellular calcium that induces actin disruption and the mobilization of lysosomes towards the site of parasite attachment, and this is mediated by microtubules and associated kinesin motors (Fig. 1c). Studies in uninfected host cells have subsequently revealed that the calcium-induced fusion of lysosomes to the plasma membrane comprises a previously undiscovered pathway to recruit new membrane to the plasmalemma during wound repair24.

Redirecting traffic

Once inside a host cell, an intracellular pathogen must use a strategy to avoid or withstand the maturation of its vacuole into a phagolysosome. Some pathogens have adapted to resist and thrive in the harsh phagolysosomal environment, others lyse their vacuole and escape to the cytoplasm, but many actively modify the vacuole to suit their needs25. Vacuole remodelling by pathogens is an active process that involves both blocking of fusion with certain compartments, and promotion of fusion with others. Because the cytoskeleton is involved in membrane traffic events including phagosomal maturation, it follows that manipulation of the cytoskeleton is a potential means for pathogens to influence these processes. Indeed, the well-studied example of Salmonella, where some of the bacterial effectors and host factors involved have now been identified, reveals that both microtubules and actin are used by this pathogen to alter its vacuolar fate (Fig. 2a).

Figure 2: Vacuolar remodelling.
figure 2

a, Salmonella. Intracellular Salmonella reside and replicate within a vacuole known as the SCV (Salmonella-containing vacuole). Maintenance of this vacuole as a permissive environment for the bacteria involves altering normal vacuolar transport pathways. SifA is involved in the formation of tubular extensions to the SCV (Sifs) which, are involved in maintaining vacuolar integrity. Sifs form along microtubules and it is likely that Salmonella uses microtubules and their associated motor proteins to deliver membrane vesicles destined for fusion to the SCV, or to send out membranous SCV tentacles that fuse with vesicular compartments. An unknown SPI-2 effector mediates recruitment of actin to the SCV and is also involved in maintaining vacuolar integrity. This may involve recruitment and fusion of actin-containing or actin-propelled vesicles to the SCV. An actin coat may also protect the SCV from fusion with unfavourable compartments. SpvB modifies actin by ADP-ribosylation, and may be involved in a subsequent disassembly of actin at the SCV and other cellular sites, such as stress fibres. SseJ may be involved in membrane removal from the SCV, through budding or scission. b, Mycobacteria. Coronin is recruited to mycobacterial vacuoles during their formation. Retention of coronin on the vacuolar membrane may serve to protect the mycobacterial vacuole from undergoing unfavourable fusion events. Mycobacteria also disrupt the actin cytoskeleton, through an unknown mechanism. This disruption causes the vacuole to arrest during phagosomal maturation at a stage where it can fuse with early endosomes but not lysosomes. The mechanisms involved in this maturation arrest have not been described, but may involve downregulation of actin-mediated vesicle rocketing.

The Salmonella-containing vacuole (SCV) of epithelial cells and macrophages is a dynamic structure that undergoes a maturation process involving fusion with certain endosomal compartments, while avoiding fusion with others. Salmonella vacuole remodelling requires the concerted effort of several effector proteins of a second TTSS, encoded within SPI-2. Salmonella induce the formation of tubular membranous structures adjoining the SCV that are known as Salmonella-induced filaments or Sifs26. Sif formation requires SifA27, a bacterial effector protein translocated into host cells by the SPI-2 TTSS, and transfection of SifA into uninfected cells induces the formation of Sif-like structures28. sifA mutants lose their vacuolar membrane and are released into the cytosol, leading to attenuation of sifA mutants in macrophages and mouse models29,30. Sifs form along microtubules and the disruption of microtubules with nocodazole blocks Sif formation and inhibits Salmonella replication26,31. Together, these results indicate that SifA is actively involved in the recruitment of membrane to the SCV and that this acquisition is microtubule dependent. It is likely that Salmonella uses microtubules and their associated motor proteins to deliver membrane vesicles destined for fusion to the SCV, or to send out membranous SCV tentacles that fuse with vesicular compartments.

In addition to its role in invasion, actin is also used in SCV remodelling by Salmonella. Approximately four hours after bacterial uptake, Salmonella induces the formation of an actin meshwork around the SCV32. This event requires the SPI-2 TTSS, although the translocated effectors involved have not yet been identified. Treatment of infected cells with actin-depolymerizing agents inhibits Salmonella replication in macrophages and results in loss of the SCV membrane and the release of bacteria into the cytoplasm32. SPI-2 directs actin assembly and Sif formation at the SCV, but only bacteria with an intact SPI-2 TTSS require actin and SifA to maintain vacuolar integrity29,32. This suggests that a complex, sequential series of events is involved in SCV remodelling. At later times, Salmonella can mediate disruption of actin around the SCV and at other host-cell sites31, a phenotype that is attributed to spvB, which ADP-ribosylates actin33. Another effector protein of the SPI-2 TTSS, SseJ, is proposed to be involved in budding or scission of the membrane from the SCV30. Therefore, it appears that remodelling and maintenance of the SCV membrane by Salmonella requires coordinated regulation of membrane acquisition and removal, and involves both microtubules and actin.

The role of actin in maintaining the SCV membrane remains to be elucidated, but recent papers have described a role for actin polymerization in movement and/or fusion of several endomembrane vesicle systems34. This suggests that actin may be used for the recruitment and/or fusion of membranous compartments to the SCV. As Salmonella replicate, the amount of SCV membrane needs to increase to accommodate a growing population of bacteria, and this is probably accomplished by fusion of vesicles to the SCV. The nature of the compartments fusing with the SCV at later time points is a matter of controversy and remains to be determined. However, the recent observations that the SCV recruits actin32 and accumulates large amounts of cholesterol36 at late time points may provide important clues to the mechanisms of SCV remodelling. Actin-mediated motility of endocytic and Golgi-derived vesicles has been described in fibroblast cells and is preferentially induced on membranes enriched in cholesterol/sphingolipid microdomains35. Together, these observations suggest that actin-mediated vesicle rocketing may be involved in the recruitment and fusion of cholesterol-rich vesicles to the SCV. In contrast to a role in promoting vesicle fusion, in certain cases actin filaments may also be important in blocking fusion between particular compartments37, raising the possibility that an actin meshwork may be involved in the SCV's avoidance of fusion with NADPH-oxidase-containing vesicles38 or other compartments.

Vacuolar remodelling has been most extensively described for Salmonella, but as we learn more about the mechanisms used by other pathogens, it is likely that hijacking of the cytoskeleton for the manipulation of membrane transport will emerge as a common theme. Studies of Mycobacteria have also linked a pathogen-directed disruption of vacuolar transport with the actin cytoskeleton (Fig. 2b). Vacuoles containing virulent Mycobacteria do not fully mature into phagolysosomes, and appear arrested during phagosomal maturation at a stage where they can fuse with early endosomes but not lysosomes. This maturation arrest occurs simultaneously with a bacterially induced disruption of the host-cell actin cytoskeleton, suggesting that these two events could be related39. Indeed, disruption of actin filaments with inhibitors can induce a similar maturation arrest in latex-bead phagosomes40. A second link between mycobacterial vacuole maturation and the actin cytoskeleton has also been described: the actin-binding protein coronin is recruited to and retained on mycobacterial vacuoles and has been implicated in their inability to fuse with other compartments41, although this is controversial42. Further characterization of these events should provide some insights into the relationship between the cytoskeleton and membrane transport.

Hitching a ride

As opposed to intracellular pathogens that live in membrane-bound vacuoles, several pathogens survive and replicate in the cell cytosol. These include pathogens that actively lyse their vacuole and those that enter the cytosol directly or through extrusion from an endosome. Once inside the cytosol, many of these pathogens harness and utilize the host cell's cytoskeletal machinery to move around.

Listeria, Shigella and vaccinia virus have provided cell biologists with valuable tools to study actin-based transport processes43,44. All use actin-based motility to move within cells and/or spread between cells (Fig. 3). Enteropathogenic Escherichia coli (EPEC) represents a variation on this theme. It directs actin polymerization from an extracellular position to form an actin pedestal underneath extracellular adherent bacteria. Studies using cell extracts, reconstituted purified components and, more recently, cells from knockout mice genetically null for various cytoskeletal or signalling proteins, have facilitated detailed dissection of the roles of many components involved in actin polymerization. What is striking about the motility of these pathogens is that they have independently evolved mechanisms to harness the activity of the actin cytoskeleton at different points, yet their strategies all converge on the Arp2/3 complex.

Figure 3: Intracellular transport.
figure 3

Vaccinia virus reaches the plasma membrane by transport on microtubules. At the plasma membrane, the viral protein A36R is tyrosine-phosphorylated, and this forms a binding site for the SH2 domain of the adaptor protein Nck. Nck recruitment to vaccinia leads to subsequent recruitment and activation of neuronal Wiskott–Aldrich syndrome protein (N-WASP), which brings together Arp2/3 and an actin monomer to initiate actin polymerization, directing the virus away from the cell. Enteropathogenic Escherichia coli (EPEC) inserts Tir into the plasma membrane, where it is tyrosine-phosphorylated and directs a recruitment cascade remarkably similar to vaccinia. This leads to the formation of an actin pedestal underneath adherent EPEC, the purpose of which is not known. Shigella IcsA binds and activates N-WASP, which facilitates binding of Arp2/3, while Listeria ActA mimics N-WASP, binding Arp2/3, actin monomers and vasodilator-simulated phosphoprotein (VASP) directly. Shigella and Listeria use actin-based motility for transport through the cytoplasm and from cell to cell.

The Arp2/3 complex is a seven-protein complex that, when activated, nucleates de novo actin polymerization45. Arp2/3 is activated by the Wiskott–Aldrich syndrome protein (WASP) family, which consists of haematopoietic cell-specific WASP, ubiquitous neuronal WASP (N-WASP) and three Scar/WAVE family proteins. WASP proteins serve a scaffolding function to bring together actin monomers and Arp2/3 to form a nucleation core, the rate-limiting step in actin polymerization. N-WASP exists in an autoinhibited conformation, but is activated by the binding of a variety of proteins and/or lipids that relieve the autoinhibition or serve an accessory function to increase the rate of actin polymerization by Arp2/3 (ref. 45). The Listeria protein ActA binds the Arp2/3 complex directly, and has additional domains that bind actin monomers and vasodilator- simulated phosphoprotein (VASP), thus mimicking the scaffolding function of N-WASP to activate Arp2/3 (refs 43,44). In contrast, IcsA/VirG of Shigella binds and activates N-WASP, which in turn recruits and activates the Arp2/3 complex43,44. EPEC Tir, a TTSS-translocated protein of EPEC that is inserted into the host-cell plasma membrane, and the vaccinia protein A36R, are tyrosine phosphorylated in host cells and then bind the adaptor protein Nck, which recruits and activates N-WASP, which then recruits Arp2/346,47.

Thus, microbes have intersected the pathway from tyrosine phosphorylation to actin polymerization at virtually each step, providing a 'nested set' of reagents to study actin polymerization. Two other pathogens that show tantalizing promise for additional discovery are Rickettsia, which has actin-based motility that seems to be independent of Arp2/3 (ref. 48), and enterohaemorrhagic E. coli, which forms an actin pedestal without tyrosine phosphorylation or Nck43,44. Future work on these pathogens should reveal more about alternate pathways that can induce actin polymerization.

While it was previously thought that vaccinia use actin-based motility to propel themselves through the cytosol of infected cells, recent, more detailed observations have revealed that vaccinia actually utilize microtubules and associated kinesins for transport to the cell periphery and then switch to actin-based motility at the plasma membrane, where Src family kinases phosphorylate A36R to create a Nck-binding site49. It is not entirely unexpected that vaccinia uses microtubules and motor proteins for directed transport to the cell periphery, as several other viruses also use microtubules for transport16. Herpesviruses, adenoviruses, and HIV all use minus-end-directed transport along microtubules to get from their point of entry at the cell periphery to their site of replication at the nucleus, and vaccinia most likely uses minus-end-directed transport to reach its perinuclear replication centre as well. After replication, plus-end-directed transport is used by several viruses for directed movement to the cell periphery for release. Some viruses seem to engage both plus- and minus-end-directed motors simultaneously, regulating their activity to favour motion in the desired direction. Future work aimed at understanding the mechanisms viruses use to engage and regulate molecular motors should shed light on motor function.

The path of least resistance

Many pathogens, including bacteria, viruses and protozoan parasites disrupt tight junctions during infection50. Tight junctions seal the space between adjacent cells, limiting diffusion of solutes through the intercellular space and creating a boundary between the apical and basolateral sides of cellular barriers such as epithelia51. Tight junctions consist of integral membrane proteins (such as occludin, claudins and junctional adhesion molecule) and cytoplasmic PDZ-domain-containing proteins (zonula occludens (ZO)-1, -2, -3, MAGUK family and PAR family), the latter of which act as adaptors at the cytoplasmic surface of tight junctions, binding each other, as well as occludin, claudins, actin and several cytoplasmic proteins. ZO-1 and ZO-2 link tight junctions to the actin cytoskeleton by binding the tight junction transmembrane proteins and actin with their N- and C termini, respectively. This complex links tight junctions to a perijunctional actomyosin ring, which supports and regulates tight junction permeability. The strategies used by pathogens for altering tight junction permeability are as numerous as the pathogens themselves50. Whereas some pathogens bind and modify junction components directly, others exert their effects through the actin cytoskeleton, which controls the integrity of tight junctions.

Rho family proteins are implicated in the assembly and maintenance of tight junctions, and many bacteria produce toxins that modify these small GTP-binding proteins, leading to disruption of tight junctions52. One example is the diarrhoeagenic pathogen Clostridium difficile, which produces two toxins (A and B) that inactivate RhoA by glucosylation. This causes F-actin restructuring, dissociation of actin from ZO-1, dissociation of occludin, ZO-1 and ZO-2 from the tight junction, and a decrease in transepithelial resistance of epithelial monolayers50.

Enteropathogenic E. coli also targets tight junctions through manipulation of the actin cytoskeleton. Although much work has been done to identify the host and bacterial players involved in this phenomenon, the exact sequence of events and their relation to each other is not yet fully understood. Decreases in transepithelial resistance during EPEC infection can be correlated with myosin light chain (MLC) phosphorylation53, ezrin phosphorylation54, occludin dephosphorylation55 and occludin55 and ZO-1 (ref. 56) dissociation from tight junctions (Fig. 4). MLC phosphorylation causes contraction of the perijunctional actomyosin ring, which opens tight junctions, presumably by increasing the tension on the tight junction, resulting in a decrease of the transepithelial resistance. Inhibitors of MLC kinase partially prevent the decrease in transepithelial resistance seen during EPEC infection53.

Figure 4: Disruption of tight junctions.
figure 4

a, Tight junctions consist of integral membrane proteins and cytoplasmic PDZ-domain-containing proteins that act as adaptors at the cytoplasmic surface, binding each other, as well as occludin, claudins, actin and several cytoplasmic proteins. Extracellular domains of tight junction membrane proteins from one cell bind those on adjacent cells, forming the basis of the intercellular seal. Zonula occludens (ZO)-1 and -2 link tight junctions to the actin cytoskeleton by binding the tight junction transmembrane proteins and actin. Ezrin may be important directly or indirectly in tight junction integrity (see text). b, Enteropathogenic Escherichia coli (EPEC)-induced disruption of tight junctions requires the type III secretion system (TTSS)-translocated effector EspF. The mechanism of action of EspF remains to be determined, but disruption of tight junctions by EPEC is associated with a variety of effects: myosin light chain (MLC) and ezrin phosphorylation, occludin dephosphorylation, and dissociation of occludin and ZO-1 from tight junctions. MLC is phosphorylated by myosin light chain kinase (MLCK), which causes contraction of the perijunctional actomyosin ring. This opens tight junctions, presumably by increasing the tension on the tight junction, resulting in a decrease of the transepithelial resistance. Occludin is dephosphorylated and occludin and ZO-1 dissociate from tight junctions. EPEC's effects on tight junctions may be mediated by ezrin, either by sequestration of ezrin away from tight junctions by its recruitment to the pedestal, or ezrin effects on signalling through Rho-family proteins (see text).

Ezrin is a member of the closely related ezrin–radixin–moesin (ERM) family of proteins that mediate membrane–cytoskeletal linkages. Phosphorylation and increased cytoskeletal association of ezrin are seen during EPEC infection, and transfection of dominant-negative ezrin partially blocks the effects of EPEC on tight junctions54. The relationship between ezrin and tight junctions in normal or EPEC-infected cells is currently not well defined. Ezrin is one of the main components of EPEC actin pedestals57, which form at the apical surface of infected cells underneath adherent EPEC. An attractive hypothesis is that some proportion of ezrin localizes to tight junctions, and is sequestered away from this site by its recruitment to the EPEC pedestal, although additional factors are required (see below). Alternatively, EPEC's use of ezrin to disrupt tight junctions may involve signalling through Rho. Ezrin has been implicated in signalling both upstream and downstream of Rho, and recent studies in Drosophila demonstrate that ERM proteins control epithelial polarity and integrity through antagonistic effects on Rho signalling58. However, the mechanisms of cell–cell adhesion in Drosophila are distinct from those in mammals, as Drosophila do not have tight junctions. Clarification of ezrin's role in the dynamics of mammalian tight junctions awaits further experimentation, and EPEC may be a key tool in unravelling this relationship.

The bacterial protein implicated in the effects of EPEC on tight junctions is a TTSS effector protein called EspF59. EspF is required in a dose-dependent manner for the decrease in transepithelial resistance and occludin redistribution seen during EPEC infection59, and ezrin activation is attenuated in an EPEC strain lacking EspF54. Little is known about the function of EspF, but it contains proline-rich domains that resemble Src-homology 3 (SH3)-binding or tryptophan–tryptophan (WW)-binding domains, suggesting that it may interact with one or more host-cell proteins via these domains. Several components of tight junctions have SH3 domains, but no binding partners have yet been described for EspF. Identification of EspF binding partners and precise subcellular localization of EspF inside host cells should clarify the manner in which this bacterial effector affects tight junctions.

Polymorphonuclear disarmament

In contrast to activation of cytoskeletal remodelling for entry into host cells, some pathogens paralyse the cytoskeleton to avoid uptake by phagocytes. Phagocytic cells such as macrophages and polymorphonuclear cells (neutrophils) possess several pathways of phagocytosis2. Opsonin-dependent pathways are mediated by FcγR and complement receptors, and opsonin-independent uptake can be mediated by a variety of receptors, including scavenger receptors, mannose receptors and integrins. Regardless of the specific pathways or particles involved, all phagocytic processes are driven by complex, controlled rearrangements of the actin cytoskeleton2, and disruption of the cytoskeleton is an effective means by which pathogens can block their own uptake.

Yersinia can trigger several key phagocytic pathways: in the absence of opsonization, Yersinia adherence is mediated by the adhesins YadA and invasin, which trigger internalization via β1-integrins in phagocytic and non-phagocytic cells60. Opsonization of Yersinia facilitates cell binding via FcγR and complement receptors in professional phagocytes61. Regardless of the adherence mechanism, Yersinia counteracts its own receptor binding and activation by injecting effector proteins that disrupt the actin cytoskeleton and defuse the triggered phagocytic pathways (Fig. 5). Remarkably, of the six known effectors of the Yersinia TTSS, four target the actin cytoskeleton and contribute to blocking phagocytosis. Yersinia outer proteins (Yops) E, H, T and O all have so-called 'anti-phagocytic' effects61.

Figure 5: Yersinia avoidance of phagocytosis.
figure 5

Binding of Yersinia to host-cell receptors triggers phagocytic pathways that would normally result in bacterial uptake. The rapid translocation of several effectors by Yersinia disarms these pathways, facilitating bacterial avoidance of phagocytosis. Yersinia outer protein H (YopH) dephosphorylates a number of tyrosine-phosphorylated signalling proteins including Fyb, SKAP-HOM and p130cas, thereby disrupting their abilities to mediate further downstream signalling events in the cytoskeletal pathway. YopE disrupts actin filaments by acting as a GTPase-activating protein for the GTPases Rac1, Rho and Cdc42, while YopT proteolytically cleaves this family of proteins, resulting in their release from the membrane. YopO blocks the activation of Rho through a mechanism that is not fully understood.

The first characterized anti-phagocytic effector of Yersinia is YopH, a protein tyrosine phosphatase62. Yersinia interaction with host cells leads to rapid phosphorylation of several host proteins that are then rapidly dephosphorylated by YopH. Several substrates have been identified for YopH: Crk-associated substrate (Cas), focal adhesion kinase (FAK), paxillin, Fyn-T-binding protein/SLP-76-associated protein/ADAP (Fyb/SLAP/ADAP), the scaffolding protein SKAP-HOM63 and Pyk2 (ref. 64). Which of these substrates is involved in blocking phagocytosis has not been analysed systematically, but, conceivably, a role in actin rearrangements and phagocytic uptake could be proposed for each of them. Blocking either Cas or FAK activity decreases uptake of Yersinia pseudotuberculosis in fibroblasts64, and FAK may be involved in FcγR-mediated phagocytosis, although this remains controversial2. Paxillin is a FAK substrate that localizes to FcγR- and complement-mediated phagosomes and is strongly phosphorylated during FcγR-mediated uptake65,66. Fyb/SLAP/ADAP and SKAP-HOM bind each other in a multi-protein complex that is implicated in integrin-mediated adhesion during T-cell signalling67. A similar complex containing Fyb/SLAP/ADAP has recently been described in FcγR-mediated phagocytosis68. The relevance of multiple YopH substrates has not been investigated thoroughly, but the observations that Yersinia uptake can occur via multiple phagocytic pathways, and that YopH can act in trans to inhibit phagocytosis through Fcγ receptors69, raise the possibility that multiple substrates of YopH facilitate the blockage of several pathways of phagocytosis simultaneously.

Not all phagocytic pathways require tyrosine phosphorylation. However, other anti-phagocytic Yops target the cytoskeleton through their effects on Rho family GTPases. Rho GTPases have key roles in phagocytosis, but the requirement for different family members varies with the mode of uptake2. Complement receptor-mediated phagocytosis is Rho-dependent, whereas FcγR-mediated phagocytosis requires both Cdc42 and Rac. Rac1 has been implicated in integrin-mediated uptake of Yersinia into HeLa cells, although other Rho family members may also be involved70. To disarm these various pathways, Yersinia encodes three effectors that inactivate Rho GTPases in distinct ways. YopE is a GAP that has a broad specificity for Rho, Rac and Cdc42 in vitro71, although there is evidence that in vivo it may specifically inactivate either Rac or RhoA in certain contexts71,72. Using a novel strategy for inactivating Rho-family proteins, YopT recognizes isoprenylated Rho, Rac and Cdc42 and cleaves them near the C terminus, resulting in their release from the membrane73. YopO (called YpkA in Yersinia pseudotuberculosis) is a serine/threonine kinase that binds and is activated by actin, and it also interacts with Rho and Rac. YopO blocks the activation of RhoA in Yersinia-infected cells and has sequence similarity with eukaryotic Rho- binding kinases, but the mechanism by which YopO mediates its anti-phagocytic effect is not fully understood60.

The presence of four anti-phagocytic Yops establishes Yersinia as an anti-phagocytic specialist, although several other pathogens also inhibit their own phagocytosis63. At least two other pathogens act through the cytoskeleton to block their own uptake. Pseudomonas aeruginosa resists internalization through manipulation of Rho-family GTPases through mechanisms similar to Yersinia, while EPEC inhibits the phosphatidylinositol-3-OH kinase signalling pathways that are essential for actin polymerization and resultant phagocytic uptake63.

Future directions

The collection of known microbial effectors provides an extensive set of reagents for cell biologists and those wishing to understand the physiological causes of infectious disease. The versatility of the cytoskeleton makes it a particularly attractive target for microbes, which use it as a multipurpose target to achieve many ends. As genetics, genomics and proteomics lead to the discovery of more microbial effectors, the number of reagents will continue to expand. Additionally, as our knowledge of cell biology increases, so will our ability to understand the activities of these effectors inside host cells. It is reasonable to speculate that for every host-cell cytoskeletal pathway, there is a microbial effector protein that exploits it, and the combined investigation of these pathways and their effectors should continue to foster insights that are mutually beneficial to both cell biologists and those studying infectious processes.