The bacterial toxin toolkit

Key Points

• Bacterial protein toxins are becoming powerful tools, which can be exploited by cell biologists to study crucial and partially characterized physiological processes and possibly discover new ones.

• Toxins acting on the plasma membrane are now used to gain access to the intracellular milieu (pore forming toxins) and to label or target plasma membrane components, such as gangliosides, sphingomyelin and GPI-anchored proteins.

• The actin cytoskeleton and its regulatory proteins are targets of a large group of bacterial toxins, which can either act as negative modulators, therefore inducing actin filament disassembly, or trigger actin polymerization.

• Vesicular transport routes and mechanisms can be explored by using toxins with cytoplasmic targets such as cholera toxin and the plant protein ricin, which were the first identified cargos of the retrograde route to the endoplasmic reticulum.

• Vesicular fusion and exocytosis can be specifically blocked by members of the clostridial neurotoxins family by cleaving SNARE proteins that are responsible for membrane targeting fidelity and fusion of lipid bilayers.

• Signal-transduction cascades involving G-protein-coupled receptors can be modulated by cholera and pertussis toxins, whereas lethal factor from Bacillus anthracis acts on mitogen-activated protein kinase kinase (MAPKK)-activated pathways.

• New virulence factors can be now used to trigger cell-cycle arrest and apoptosis, and provide exciting means to tackle the immune system and promote immune response.

Abstract

Pathogenic bacteria and higher eukaryotes have spent a long time together, leading to a precise understanding of one another's way of functioning. Through rapid evolution, bacteria have engineered increasingly sophisticated weapons to hit exactly where it hurts, interfering with fundamental host functions. However, toxins are not only useful to the bacteria — they have also become an essential asset for life scientists, who can now use them as toolkits to explore cellular processes.

Main

One outcome of the evolutionary race between host and microbial pathogens is the development of sophisticated and specific virulence factors, which give a selective advantage to the organism that produces them¹. Over the past decade, our understanding of the mechanism of action of bacterial toxins has increased enormously². These studies have contributed to the molecular definition — and, sometimes, to the discovery — of important pathways in cell biology. A striking example is the finding that tetanus toxin is an endopeptidase, specific for a membrane protein that is localized on synaptic vesicles and secretory granules. This finding was crucial in the definition of the molecular mechanisms that underlie regulated secretion³. Similarly, the importance of guanine-nucleotide-binding (G) proteins in signal transduction was uncovered by studies into the action of the cholera and pertussis toxins, which modify different classes of G proteins⁴.

Now that the mode of action of many toxins has been unravelled, they can be used as highly specific and efficient tools in cell biology. Here, we have made a subjective selection of what might be beneficial to an experimental cell biologist, dividing bacterial toxins into areas of interest (Fig. 1) and highlighting, where appropriate, the possibility of using them in new, exciting applications.

Figure 1: Cellular processes targeted by bacterial toxins.

These have been subdivided into four classes, each represented by examples in Figs 2 to 5.

Plasma membrane permeabilization

A recurrent theme in cell biology is the need to design systems that make it possible to gain access to the cytoplasm while keeping the cell alive during the course of the experiment. Such a semi-intact cell system can be obtained by treating cells with pore-forming toxins⁵ (Table 1). These toxins are secreted as water-soluble proteins and, when added to target cells, they bind to cell-surface components acting as receptors. Many pore-forming toxins must then oligomerize into an AMPHIPATHIC, circular ring-like structure that inserts into biological membranes to form a pore⁶. These structures can be formed from just a few monomers (up to seven), generating small pores, or they may be composed of up to 50 units, creating pores large enough to permit the passage of proteins. The membrane-spanning region of these oligomeric complexes is thought to be formed by β-BARRELS, as shown by the crystal structure of the Staphylococcus aureus α-toxin⁷.

Table 1 The toxin toolkit

The S. aureus α-toxin and Aeromonas sp. aerolysin^5,8 (Fig. 2) are the best-characterized members of this family. These two toxins are suitable for creating hydrophilic pores of fixed size (approximately 1.5 nm in diameter) that allow exchange of small ions as well as nucleotides. The α-toxin has the advantage of being widely used; the drawback, however, is that high doses in the 100-nM range are usually necessary. Doses strongly depend on the cell type, presumably reflecting the distribution of an unknown α-toxin receptor. By contrast, aerolysin has not yet been widely used as a tool in cell-permeabilization experiments, but it has the advantage of being effective on any mammalian cell at picomolar concentrations. This is because aerolysin binds to glycosylphosphatidylinositol (GPI)-anchored proteins, a ubiquitous class of membrane-associated proteins, and binding occurs through the GLYCAN core⁹ that is present in all the lipid anchors of this protein family.

Figure 2: Targeting the plasma membrane.

Cells can be permeabilized with pore-forming toxins such as streptolysin O (SLO) or Staphylococcus aureus α-toxin. GPI-anchored proteins (GPI-APs) can either be tagged using aerolysin or removed using phosphatidylinositol-specific phospholipase C (PI-PLC). Sphingomyelin can be tagged using lysenin or cleaved using sphingomyelinases. Ganglioside GM1 can be detected using cholera toxin (CT) or its pentameric binding subunit (CT-B).

Large pores, which allow the passage of proteins, can be formed by members of the cholesterol-dependent toxin family, which includes streptolysin O (SLO)^5,10 (Fig. 2). Members of this large toxin family require cholesterol for pore formation, and they generate channels of variable size, up to 35 nm in diameter, corresponding to about 50 subunits. During polymerization, each subunit contributes 4 β-strands to the pore-forming β-barrel, generating a transmembrane structure composed of ∼200 β-strands¹¹. Walev et al.¹⁰ recently reported that pore formation can be reversible under appropriate experimental conditions. Indeed, short exposure to SLO, followed by incubation of the cells in a toxin-free buffer in the presence of 1–2-mM calcium, resulted in repair of the membrane lesion.

We must, however, keep in mind that pore-forming toxins induce a wide spectrum of cellular events as a result of the permeabilization process. These include activation of G proteins¹², production of cytokines¹³ or vacuolation of the endoplasmic reticulum (ER)¹⁴. Some of these effects are probably triggered by ion fluxes, calcium entry or depletion of cytoplasmic components, whereas others are still unexplained. These events must be taken into account when interpreting the data from experiments with pore-forming toxins involving plasma-membrane permeabilization of living cells.

Tagging specific molecules

Toxins are generally active even at very low concentrations. This high efficiency is due to the fact that most toxins act catalytically, resulting in an amplification of cell responses, and also because essential surface molecules on target cells are opportunistically used by the toxins as receptors. This high affinity and specificity for their surface receptors make toxins excellent probes to follow the molecules with which they interact. Two examples will be highlighted here: cholera toxin and lysenin, which bind to the ganglioside GM1 (Ref. 15) and sphingomyelin¹⁶, respectively.

The receptor-binding unit of cholera toxin (CT-B) is a homopentamer, in which each subunit can interact with ganglioside GM1, a ubiquitous monosialylated glycosphingolipid found in eukaryotic cells (Fig. 2). This pentavalent binding unit has a higher affinity for clustered GM1 (Ref. 15), which can be found in cholesterol–glycosphingolipid-rich microdomains of the plasma membrane called LIPID RAFTS¹⁷. Lipid rafts are the subject of intensive research because of their role in various fundamental cellular processes such as membrane sorting, signalling and cholesterol homeostasis, and markers for this membrane compartment are scarce. Therefore, CT-B (labelled with peroxidase, colloidal gold or fluorophores) has been an extremely useful tool as it allows the detection of GM1 not only on dot blots, but also on cells by either electron or fluorescence microscopy¹⁸.

Lysenin, a newly characterized toxin, could be similarly useful to study lipid rafts, as it binds specifically to the raft component sphingomyelin¹⁶. Little is known about the intracellular distribution of sphingomyelin — not only under normal conditions, but also under pathological conditions, such as, for example, lipid-storage diseases¹⁹. With appropriate labelling, lysenin should allow detection of sphingomyelin on blots as well as on cells (Fig. 2) and permit the study of the distribution and transport of this lipid. So far, lysenin has been successfully used to select mutant cells that are defective in sphingolipid biosynthesis²⁰.

Removing lipids or lipid-anchored molecules

Several bacterial species produce enzymes that affect lipids¹, and are therefore extremely useful in cell biology. These include soluble phosphatidylinositol-specific phospholipase Cs (PI-PLCs), the prototype of which is the enzyme purified from Bacillus cereus. PI-PLCs recognize and cleave both phosphatidylinositol and its glucosyl derivatives, and are therefore exploited to release GPI-anchored proteins from the cell surface (Fig. 2). This treatment has been the method of choice to determine whether or not a protein is GPI-anchored²¹.

A second class of enzymes of great interest are sphingomyelinases¹ (Fig. 2), which cleave sphingomyelin to generate ceramide, a second messenger involved in cell growth, differentiation, apoptosis and membrane transport^22,23. The coupling of ceramide signalling to specific signal-transduction cascades is specific to both the stimulus and the cell type, and seems to be determined by the subcellular topology of its production²³. Interestingly, transfection with Bacillus cereus sphingomyelinase expressed under the control of an inducible promoter led to the identification of a signalling pool of sphingomyelin that is distinct from the pool accessible to exogenous sphingomyelinase²⁴.

Probing intracellular routes

With the notable exception of pore-forming toxins mentioned above, most bacterial toxins have an enzymatic activity towards specific cytoplasmic targets. This implies that the toxin must cross the lipid bilayer and penetrate the cell. But this event rarely occurs at the plasma membrane. Toxins generally undergo endocytosis followed by transport to specific intracellular organelles before they translocate into the cytosol.

Distinct portions or subunits are involved in the different steps at which these toxins act. The B (binding) subunit is involved in receptor binding and translocation into the cytoplasm, whereas the A (active) subunit bears the enzymatic activity. On this functional basis, toxins with intracellular targets have been defined as A–B toxins. Studies of the routes used by these molecules turned out to be extremely interesting to cell biologists as they allowed the characterization of certain transport steps, and sometimes even revealed new intracellular pathways²⁵.

Toxins have hijacked all known entry points into a cell (Fig. 3). Some enter through CLATHRIN-COATED VESICLES (diphtheria toxin²⁶, Pseudomonas exotoxin A); others through the putative CAVEOLAR pathway (cholera toxin¹⁵, heat-labile enterotoxin from Escherichia coli). Interestingly, the plant toxin ricin is internalized by all pathways including a putative clathrin and caveolae-independent route²⁷, probably as a consequence of its binding to several receptors.

Figure 3: Studying intracellular membrane transport.

Various toxins are internalized and transported to intracellular organelles from where they are translocated into the cytoplasm. Different endocytic pathways can be studied using, for example, cholera toxin (CT) and heat-labile enterotoxin (Etx) from Escherichia coli (caveolar pathway), Shiga toxin (clathrin-dependent and clathrin-independent pathways), and ricin (several pathways including clathrin- and caveolae-independent). Clathrin coating is shown by a grey dashed line. Molecular mechanisms of vesicle fusion can be addressed using tetanus neurotoxin (TeNT) and botulinum neurotoxins (BoNTs), which cleave specific SNARE proteins. Also shown is Helicobacter pylori vacuolating toxin (VacA), which triggers vacuolation of late endocytic compartments and has recently been suggested to translocate to mitochondria (Box 1). ER, endoplasmic reticulum; MVB, multivesicular body.

The contribution of toxins to the field of membrane transport has, however, gone far beyond the initial internalization step. Studies of Shiga toxin were the first to show that a molecule can be transported from the cell surface to the ER^25,27. More specifically, Shiga toxin binds to the cell surface by interacting with globotriaosylceramide (Gb3) and is then endocytosed through clathrin-dependent and clathrin-independent mechanisms and routed to the ER through the Golgi apparatus. Whether Shiga toxin remains bound to Gb3 is still not known.

The exact mechanisms that operate in this retrograde route seem to be new, and have not yet been fully characterized²⁸. More than one retrograde transport pathway could exist, however, between the Golgi and the ER, as Pseudomonas exotoxin A and cholera toxin are also transported to the ER through the Golgi, but using a distinct mechanism²⁹. Therefore, these bacterial toxins will continue to be useful tools to dissect retrograde transport pathways. In addition, they have been (and will be) helpful in studies on the still poorly understood communication routes between the endocytic and the biosynthetic pathways such as early endosomes to Golgi²⁹ or plasma membrane to Golgi³⁰.

Whereas cholera toxin can be used on almost any cell type owing to the ubiquitous expression of its receptor, some toxins have a remarkable cell type specificity. This is the case with Shiga toxin, which will affect only Gb3-positive cells such as those of the immune system. Members of the clostridial neurotoxin family, which is composed of tetanus and botulinum neurotoxins, bind at the neuromuscular junction and undergo differential transport in mammalian motor neurons³¹. Botulinum neurotoxins remain at the synapse, allowing the study of site-specific endocytosis, whereas tetanus toxin is retrogradely transported to the cell body and then transcytosed to inhibitory interneurons, making it a marker of choice to study retrograde transport in neurons (G. Lalli and G.S., unpublished data). In addition, botulinum neurotoxins, which are synthesized as large, multisubunit protein complexes, can be used to follow transcytosis in binding-competent epithelial cells such as colon cells, making use of the fact that these toxins can gain entry into an organism through the intestinal tract³².

Delivery of proteins into cells

As mentioned above, many toxins can translocate their catalytic moiety across a membrane into the target cell cytoplasm. The binding and/or translocation domains of toxins can therefore be used as vehicles for intracellular delivery of peptides and proteins. A molecule of choice for such an application is the adenylyl cyclase toxin from Bordetella pertussis³³. The mode of entry of this toxin is unique because the catalytic domain, harbouring the adenylyl cyclase activity, is directly translocated across the plasma membrane of the target cell, independently of any endocytic step.

When making fusion proteins between the selected peptide or protein and an inactive variant of the adenylyl cyclase toxin, efficient translocation of the cargo into the cytoplasm can be monitored with a concomitant, sought biological response. This approach has been successfully used to present peptides to the cytosolic pathway for major histocompatibility (MHC) class-I-restricted antigen presentation³⁴. In this pathway, peptides that reach the cytoplasm are processed by the proteasome, translocated into the ER and loaded onto newly synthesized MHC class I complexes, which are then transported to the cell surface. Similarly, fusion proteins harbouring the binding subunits of Shiga toxin allowed the MHC-class-I-restricted presentation of tumour peptides in dendritic cells³⁵.

The adenylyl cyclase and Shiga toxins are not only useful for studying antigen presentation but they are also attractive vectors for vaccine development as they elicit a response from cytotoxic T lymphocytes (CTLs) in vivo. CTLs are thought to have a key role in the control of virus-infected cells and tumour growth, but immunization with peptides generally fails to trigger this response.

Cytoplasmic delivery of heterologous proteins and peptides has also been successful with anthrax toxin³⁶, the binding subunits of Clostridium botulinum C2 binary toxin³⁷, the E. coli heat-labile enterotoxin³⁸, and diphtheria toxin³⁹. Most of these systems take advantage of a relatively low selectivity for the target cells — a property that reflects the ubiquitous nature of their respective receptors. These are the ganglioside GD1 (Ref. 40) for the enterotoxin and the heparin-binding epidermal-growth-factor-like growth factor precursor^41,42 for diphtheria toxin, whereas the receptor for anthrax toxin is unknown. Specificity can, however, be generated in some cases by the use of chimaeras. For example, neuronal delivery can be obtained by fusing the protein of interest with the non-toxic C fragment of tetanus toxin, which is involved in binding, as demonstrated by the targeting of CuZn superoxide dismutase to spinal cord motor neurons⁴³.

Interfering with cell signalling

The enzymatic activity provided by A–B toxins can also be very useful for altering specific signal-transduction pathways within the cell. In particular, pertussis toxin was instrumental in the discovery of inhibitory heterotrimeric G proteins, and it is extensively used in the definition of the initial steps in signalling cascades in plant and animal cells. This toxin ADP-ribosylates the α-subunits of several G proteins, including G_i, G_o and G_gust⁴⁴, uncoupling them from their cognate receptors. In the case of G_i, this modification causes the silencing of the inhibitory input of the signalling cascade, which induces the indirect activation of downstream effectors, such as adenylyl cyclase, phosphodiesterase and ion channels⁴ (Fig. 4). In contrast to pertussis toxin, cholera toxin and E. coli enterotoxin ADP-ribosylate activatory G proteins (G_s), including G_t and G_olf⁴⁴, by blocking their GTPase activity. This leads to the constitutive activation of adenylyl cyclase and the rapid elevation of cyclic AMP levels in the cell (Fig. 4).

Figure 4: Interfering with cell signalling.

Cholera toxin (CT) and Escherichia coli enterotoxin (Etx) activate the α-subunit of heterotrimeric G_s proteins downstream to stimulatory G-protein-coupled receptors (GPCR_s), leading to the constitutive activation of adenylyl cyclase and rapid elevation of cellular levels of cyclic AMP. Pertussis toxin (PT) instead inactivates the α-subunits of G proteins coupled to inhibitory GPCR (GPCR_i). This modification causes the silencing of the inhibitory input and induces the indirect activation of downstream effectors. Lethal factor (LF) requires the anthrax protective antigen (PA) to reach the cell cytoplasm, where it cleaves mitogen-activated protein kinase kinases (MAPKKs). PA is also essential for the binding and translocation of the oedema factor (EF), a powerful adenylyl cyclase, which elevates the cytoplasmic concentration of cAMP.

The 'lethal factor' produced by Bacillus anthracis (LF) can also affect signalling, as this zinc-endopeptidase specifically cleaves several members of the mitogen-activated protein kinase kinase (MAPKK) family^45,46 (Fig. 4), with the notable exception⁴⁷ of MAPKK5. Lethal factor elicits the hyper-stimulation of macrophage inflammatory responses and can be used to induce cytokine expression and oxidative burst in these cells⁴⁸. Similarly, the oedema factor produced by the same bacterium transiently increases intracellular cAMP by means of its adenylyl cyclase activity, and modulates the inflammatory immune response⁴⁸ (Fig. 4).

Inhibition of membrane fusion

In addition to their use in studying intracellular transport routes, botulinum and tetanus neurotoxins have been crucial in understanding membrane fusion, in particular during neurotransmitter release. These clostridial neurotoxins are characterized by extensive sequence similarity and are typical examples of A–B toxins. Their catalytic subunits contain a zinc-endopeptidase activity that is specific for synaptic members of the SNARE (soluble NSF attachment protein receptor, where NSF stands for N-ethyl-maleimide-sensitive fusion protein) superfamily: VAMP (vesicle-associated membrane protein), syntaxin and SNAP-25 (Table 1)^31,49. SNAREs are not only involved in the fusion of synaptic vesicles with the presynaptic plasma membrane, but they are also implicated at all stages of vesicular transport, during which they mediate targeting fidelity and the fusion of lipid bilayers⁵⁰. The action of these toxins is restricted to neurons and neuronally differentiated cells, but cell permeabilization by means of pore-forming toxins or transfection with the active domains of tetanus and botulinum neurotoxins allow the inactivation of specific SNAREs within any cell type⁵¹. In addition, targeted expression of the active subunit of clostridial neurotoxins can be used to ablate specific SNARE proteins in selected tissues or even in entire organisms^52,53.

Affecting the cytoskeleton

Clostridial species not only produce the above-mentioned SNARE-specific endopeptidases, but also several cytotoxins that affect the cytoskeleton (Fig. 5). A great deal of information on the cytoskeleton and its ability to undergo dynamic changes — the basis of numerous cellular processes, such as cell migration, polarization and cytokinesis — has been gathered by studying the effects of these toxins (Table 1). Their targets include actin itself as well as several regulatory components of the actin cytoskeleton^1,54,55.

Figure 5: Affecting the cytoskeleton or cell polarization.

a | Cell treatment with large clostridial toxins (LCTs) or clostridial binary toxins, such as Clostridium botulinum C2, lead to disassembly of the actin cytoskeleton caused by the modification of actin or the inhibition of small GTPases. Rho inactivation can also be triggered by the smaller exoenzyme C3 and C3-like toxins. By contrast, Escherichia coli cytotoxic necrotic factor (CNF) and Bordetella dermonecrotic toxin (DNT) activate Rho, which can in turn promote the appearance of stress fibres and Rho-dependent macropinocytosis. Bacteroides fragilis enterotoxin cleaves E-cadherin, causing the disassembly of zonula adherens and indirectly the redistribution of actin cytoskeleton. b,c | Cytopathic effects induced by Rho-modifying LCTs in epithelial cells (b) and fibroblasts (c). (Courtesy of E.C. Olarte, E. Freer and M. Thelestam.)

The large family of clostridial binary toxins (CBTs) comprises several bi-chain toxins (C. botulinum C2, iota and iota-like toxins), produced by different species of Clostridium sp., but all belonging to a class of ADP-ribosylating enzymes that act on monomeric G actin (Table 1). ADP-ribosylation of actin inhibits its polymerization and leads to the dissociation of actin microfilaments (Fig. 5). By contrast, most other bacterial toxins that affect the actin cytoskeleton modify small GTP-binding proteins of the Rho subtype, which seem to be the most important regulators of actin microfilament dynamics^54,55. Large clostridial cytotoxins (LCTs) form the biggest family in this toxin subgroup and are characterized by extensive sequence homology, similar structural organization and identical glycosyl-transferase activity. Target specificity varies among members of the LCT family. It can be restricted to Rho, Rac and Cdc42, as in the case of the A and B toxins from Clostridium difficile^54,55. Or it can cover a wider spectrum, including Ras, Rap and Ral, but with the notable exception of Rho, as, for example, the lethal toxin from Clostridium sordellii⁵⁶ (Table 1).

Finally, selective inactivation of Rho GTPases can be achieved by treatment of cells with members of a heterogeneous group of single-chain, low-molecular-weight ADP-ribosylating toxins of common ancestral origin. These comprise C. botulinum C3 toxin, staphylococcal epidermal cell differentiation inhibitor (EDIN), staphylococcal C3stau2 (which also modifies RhoE)⁵⁷ and other, similar, C3-like toxins produced by various bacteria including Clostridium limosum and Bacillus cereus. The general effect of cell intoxication with these negative regulators of small GTP-binding proteins is a depolymerization of actin filaments (F-actin) with a consequent collapse of the actin cytoskeleton⁵⁶.

By contrast, the cytotoxic necrotizing factor (CNF) from E. coli and the dermonecrotic toxin (DNT) from different Bordetella species activate Rho GTPases by specific deamidation^58,59,60 (or transglutamination^61,62) of an essential glutamine residue, therefore increasing the formation of actin filaments. This process leads to the appearance of stress fibres and promotes macropinocytosis^54,55 (Fig. 5).

To bypass the various cell-type specificities of these toxins, which are imposed by the presence of selective and largely unidentified cell-surface receptors⁵⁶, the catalytic moiety can be directly delivered into the cytoplasm by permeabilizing cells with SLO¹⁰. Alternatively, the catalytic portion can be fused to the binding and translocation domains of a toxin with ubiquitously distributed receptors. This strategy has been successful for efficient delivery of C3 toxin coupled with the binding fragment of C. botulinum C2 binary toxin³⁷.

By contrast, the enterotoxin of Bacteroides fragilis does not require entry into the cytoplasm to trigger the reorganization of the actin cytoskeleton. This novel toxin is a metalloprotease, which cleaves E-cadherin on the basolateral side of polarized cells⁶³. The specific proteolysis of this essential component of the ZONULA ADHERENS causes the disruption of intercellular junctions, with consequent loss of cell polarization and changes in cell barrier permeability (Fig. 5).

Bacterial two-hybrid screening

The ability of two complementary fragments of the adenylyl cyclase toxin from B. pertussis to reassociate into a fully active enzyme is at the basis of a recently established two-hybrid system for the screening of protein–protein interactions in bacteria^64,65. Re-association of the adenylyl cyclase toxin fragments requires either calmodulin or the genetic fusion of each of the two fragments to two interacting proteins, X and Y. Following the interaction of X and Y, the two toxin fragments are brought into close proximity, resulting in functional complementation. This is followed by production of cAMP, which then triggers transcriptional activation of catabolic operons, yielding a characteristic phenotype used for the selection. The association between the hybrid proteins, which occurs in the cytoplasm, can therefore be spatially separated from the nuclear transcriptional activation readout, offering a more flexible experimental tool to monitor protein–protein interactions occurring in the cytosol⁶⁵.

Concluding remarks

In their never-ending evolutionary search for new ecological niches, bacteria have provided life scientists with invaluable tools to inhibit specifically or modulate key cellular events. This review has presented a subset of bacterial protein toxins that are useful to cell biologists, but several others have been studied (some are mentioned in Box 1 and Table 1). No doubt hitherto-unknown toxins will emerge as many bacterial species remain to be discovered, and others may become pathogenic in the future⁶⁶. Moreover, entire genomes of bacterial pathogens have been or will be sequenced. This should allow the rapid identification of new virulence factors, the mode of action of which will surely be unravelled, thanks to the rising interest in cellular microbiology.

Box 1 | Novel toxins

Cell-cycle arrest. Whereas most bacterial toxins that act inside host cells either destroy or modify host cell proteins, cytolethal distending toxins (CDTs) have a DNase-I-like activity. CDTs cleave DNA during replication, thereby activating host cellular checkpoint mechanisms, which result in cell-cycle arrest^67,68. Although they have not yet been used for such purposes, these toxins may be useful in studies on the mitotic checkpoint and also in membrane transport. For their cytotoxic action, CDTs must indeed undergo endocytosis, be transported through the Golgi and possibly through the endoplasmic reticulum^68,69. However, in contrast to other toxins that undergo retrograde transport and are then translocated into the cytoplasm, CDTs must reach the nucleus. This could occur either through a cytoplasmic intermediate, or directly from the ER through an as-yet-unknown pathway.

Apoptosis. The list of toxins and other bacterial virulence factors that induce apoptosis is increasing exponentially. Recent findings indicate that three bacterial proteins might translocate to host cell mitochondria and cause programmed cell death: Neisseria gonorrhoeae Por B⁷⁰, enteropathogenic Escherichia coli Orf19 (Ref. 71) and Helicobacter pylori vacuolating toxin (VacA)⁷². VacA is also responsible for the well-established alteration of the endocytic pathway resulting in the selective swelling of late endosomes/pre-lysosomal structures⁷³ (Fig. 3).