Infectious microbes face an unwelcoming environment in their mammalian hosts, which have evolved elaborate multicelluar systems for recognition and elimination of invading pathogens. A common strategy used by pathogenic bacteria to establish infection is to secrete protein factors that block intracellular signalling pathways essential for host defence. Some of these proteins also act as toxins, directly causing pathology associated with disease. Bacillus anthracis, the bacterium that causes anthrax, secretes two plasmid-encoded enzymes, LF (lethal factor) and EF (oedema factor), that are delivered into host cells by a third bacterial protein, PA (protective antigen). The two toxins act on a variety of cell types, disabling the immune system and inevitably killing the host. LF is an extraordinarily selective metalloproteinase that site-specifically cleaves MKKs (mitogen-activated protein kinase kinases). Cleavage of MKKs by LF prevents them from activating their downstream MAPK (mitogen-activated protein kinase) substrates by disrupting a critical docking interaction. Blockade of MAPK signalling functionally impairs cells of both the innate and adaptive immune systems and induces cell death in macrophages. EF is an adenylate cyclase that is activated by calmodulin through a non-canonical mechanism. EF causes sustained and potent activation of host cAMP-dependent signalling pathways, which disables phagocytes. Here I review recent progress in elucidating the mechanisms by which LF and EF influence host signalling and thereby contribute to disease.
Anthrax refers generally to diseases caused by infection with Bacillus anthracis, a spore-forming encapsulated Gram-positive bacterium [1–3]. The disease takes one of several forms, cutaneous, gastrointestinal or inhalational, depending on the route of entry of infectious spores. Cutaneous anthrax, the most common form of the disease in humans, is associated with characteristic skin lesions and is manageable with antibiotics. Gastrointestinal anthrax causes a much more serious systemic disease but primarily affects livestock that ingest spores. Inhalational anthrax is extremely rare in humans, but follows a rapid course and is highly fatal. Lung macrophages and possibly dendritic cells engulf and kill most inhaled spores [4–7]. However, a fraction of spores survive and germinate within the macrophages, which carry them to nearby lymph nodes. Vegetative bacteria then escape into the lymphatic and circulatory systems, dividing extracellularly. The disease is asymptomatic for several weeks following infection and then gives rise to mild non-specific symptoms, including fever, aches and a cough. Within days, bacteria reach high levels in the circulation, causing fulminant disease characterized by respiratory distress, shock and widespread haemorrhage, often with seizures. Death generally occurs within 24 h of reaching the fulminant stage, and antibiotics are without therapeutic benefit from this point onwards .
B. anthracis produces a number of virulence factors that are critical for establishment of infection and pathogenesis [3,9]. Some of the genes particularly important for promoting disease reside on two plasmids, pOX1 and pOX2. pOX2 encodes proteins involved in biosynthesis of the polyglutamic acid capsule, which helps dividing bacteria evade engulfment and killing by phagocytes . pOX1 carries the genes encoding PA (protective antigen), LF (lethal factor) and EF (oedema factor). Binary combinations of these secreted proteins comprise the two anthrax toxins: PA combined with LF is called LeTx (lethal toxin), and PA combined with EF is known as EdTx (oedema toxin). LF and EF are enzymes that act on intracellular substrates: LF is a metalloproteinase and EF is an AC (adenylate cyclase). Both are translocated into the cytosol of target cells by way of endosomes through a pore formed from a heptamer of PA molecules. EdTx and LeTx interfere with important cellular responses to bacterial infection, disabling host immunity and promoting bacterial dissemination. As the disease progresses, the toxins accumulate to high levels in the circulation and are believed to cause pathology directly.
Historically, the general view regarding toxin-induced pathology has been that LeTx is primarily responsible for fatality associated with systemic anthrax, while EdTx causes localized oedema found with cutaneous anthrax [3,9]. Intravenous delivery of LeTx kills rodents and other mammals, albeit with significant differences in potency among different strains and species. Older studies with crude toxin preparations suggest, in contrast, that EdTx is non-lethal, but does produce oedema upon subcutaneous administration. Non-encapsulated strains of B. anthracis engineered to lack LF or PA are highly attenuated in causing mortality in mice infected subcutaneously, whereas EF deletion has a significant but much smaller effect. However, it is not clear to what extent this model recapitulates features of inhalational anthrax caused by more virulent encapsulated strains of B. anthracis. The availability of highly purified recombinant toxin preparations has allowed some of the earlier work to be revisited in a more systematic manner [11–13]. Death from LeTx treatment in mice is associated with shock, vascular collapse and generalized hypoxia, recapitulating some of the symptoms and pathology seen in human inhalational anthrax patients [8,14,15]. Interestingly, intravenous administration of EdTx to mice causes death at lower doses and with a more rapid onset of symptoms than for LeTx. EdTx causes widespread tissue damage and multi-organ failure accompanied by haemorrhage and hypotension, features that are also observed in patients. These observations support the idea that the two toxins co-operate in causing disease-associated fatality. This notion has prompted research into ‘antitoxin’ drugs to be administered as part of combination therapy with antibiotics, which alone have no efficacy against late-stage inhalational anthrax. Antitoxins targeting PA have been developed that block toxin uptake, and specific inhibitors of LF and EF have been identified . Many of these compounds decrease mortality in toxin-treated animals, but few have been tested in models of infection. Among those that have, a small-molecule LF inhibitor and a humanized monoclonal antibody against PA have both shown efficacy in a rabbit inhalational anthrax model [17,18]. These observations both substantiate a direct role for toxins in anthrax-associated fatality and validate LF and PA as therapeutic targets. Further in vivo studies with inhibitors of both LF and EF will be valuable in dissecting the roles of the two factors at varying stages of disease.
This review will focus on the mechanisms by which LF and EF interfere with host signalling pathways that are critical for proper cellular regulation and thereby promote disease. Interested readers are referred to recent reviews covering mechanisms of toxin uptake [19,20], roles of toxins in pathogenesis [21–23], regulation of toxin genes  and antitoxin development [16,25].
LF is a 90 kDa protein comprising an N-terminal PA-binding domain, a large central domain and a C-terminal catalytic domain (Figure 1) . The catalytic domain has an active site HEXXH sequence that is common to metalloproteinases and constitutes part of the zinc-binding and catalytic machinery. Mutation of these residues eliminates the toxicity of LF in vitro and in vivo, indicating that its metalloproteinase activity is required for function. The overall fold of the catalytic domain is unique, and homologues from other organisms have not been identified. LF is also atypical in that it lacks a propeptide and is not activated by cleavage of a precursor. The N-terminal PA-binding domain is necessary and sufficient for cellular uptake of LF [27,28]. Although non-catalytic and unrelated in primary sequence, it shares a common fold with the metalloproteinase domain. The large central domain has structural homology with bacterial ADP-ribosyltransferases but also lacks a conserved catalytic residue. This domain is likely to be involved in substrate recruitment (discussed below) . A helix-rich insertion within the central domain covers the metalloproteinase active site and may serve to compartmentalize its activity.
LF cleavage and inactivation of MKKs (mitogen-activated protein kinase kinases)
LF is a highly specific protease that exclusively cleaves MKKs [30–34]. MKKs lie in the middle of three-component phosphorylation cascades activated by a wide variety of cellular stimuli, including growth factors, cytokines and stress [35,36]. These signals lead to activation of the upstream MKKKs (mitogen-activated protein kinase kinase kinases), which phosphorylate and activate MKKs, which in turn phosphorylate and activate MAPKs (mitogen-activated protein kinases). The MAPKs themselves phosphorylate a large number of cellular targets on serine and threonine residues and play an important role in mediating responses to extracellular cues, including changes in the transcriptional profile. In mammals, MAPKs activated by canonical cascades belong to four families: ERK1/2 (extracellular-signal-regulated kinase 1/2), p38, JNK (c-Jun N-terminal kinase) and BMK (big MAPK)/ERK5 (in this review, I will use the term ERK to refer exclusively to the ERK1/ERK2 family). The seven MKKs integrate signals from at least 20 MKKKs and are dedicated to particular MAPK families (summarized in Figure 2). LF cleaves all mammalian MKKs, except for MKK5, and thereby shuts down the ERK, JNK and p38 pathways. Interestingly, LF can also cleave MKKs from Drosophila melanogaster, which could provide a genetic model system for studying LF action .
LF cleaves MKKs at specific sites outside of their catalytic domains [30,31] (Figure 3). Cleavage disrupts or removes a D-site (also called a DEJL motif or D-domain), a MAPK-docking motif also found in MAPK substrates and phosphatases [38,39]. MKKs lacking the D-site have reduced binding affinity for their cognate MAPKs, and thus LF cleavage inhibits MAPK phosphorylation by the MKK [40–43]. However, some MKKs (in particular MKK3) largely retain the ability to bind MAPKs subsequent to LF cleavage. In addition, LF cleavage of MKK1 decreases the rate of phosphorylation of saturating concentrations of ERK2 , suggesting that MKK D-sites make additional contributions to MAPK activation beyond simply effecting recruitment. One potential mechanism has been suggested by recent structural studies of MAPK–D-site interactions [44–46]. MKKs phosphorylate two conserved residues, a tyrosine and a threonine, within the MAPK activation loop, a segment that rearranges upon activation to form part of the substrate-binding site (Figure 4). The crystal structure of inactive ERK2 reveals an ordered activation loop that packs against the catalytic domain, burying the side chains of these tyrosine and threonine residues . D-site-derived peptides bind MAPKs within a region that is on the opposite face of the catalytic domain from the activation loop [44–46,48,49]. In most crystal structures of MAPKs in complex with D-site peptides, the activation loop is disordered, suggesting long-range communication between the two sites. Hydrogen-exchange MS experiments also suggest increased flexibility in the activation loops of p38 and ERK2 upon D-site peptide binding . In the case of ERK2, the activation loop can also adopt a conformation similar to the active enzyme, with the tyrosine and threonine residues exposed . These observations suggest that D-site binding increases MAPK phosphorylation to some extent by making the activation loop more accessible to the MKK. LF cleavage may thus inactivate the MKK in part by rendering it incapable of inducing this ‘activatable’ conformation in the MAPK.
LF cleavage of MKKs
LF cleavage sites in MKKs have conserved features that are also generally found in D-sites, including a cluster of basic residues located three to five residues upstream of two or more aliphatic residues (Figure 3). Peptide library screens and analysis of MKK cleavage site mutants indicate that these features are important for efficient proteolysis by LF [30,51,52]. Although many other proteins, such as MAPK substrates, possess D-sites, none that have been tested appear to be LF substrates . In addition, synthetic peptides spanning MKK cleavages sites are much less efficient LF substrates than full-length MKKs . The unique capacity of MKKs to be cleaved by LF appears to be due to a non-active-site-mediated docking interaction with a region of the MKK distal from the site of proteolysis. This docking interaction was initially hypothesized based on the observation that a truncation mutant of MKK2 lacking the cleavage site interacts with LF in a yeast two-hybrid assay . More recently, the LFIR (LF-interacting region) of MKK1 has been mapped to a site in the large C-terminal lobe of the catalytic domain . The LFIR lies immediately downstream of an insertion unique to MKK1 and MKK2 that mediates interaction with Raf kinases and other regulators. Mutation of residues within the LFIR that are conserved among all MKKs abolishes LF cleavage. These residues are not surface-exposed, but rather map to the hydrophobic interior of MKK1 , suggesting that MKKs may be conformationally flexible in this region. Alternatively, these residues may not contact LF directly, but may be required to maintain the overall LFIR structure. A site within the central domain of LF has been identified that is required for efficient cleavage of MKKs in cultured cells and appears to mediate a direct physical interaction with MKKs . However, mutation of residues within this site do not block MKK cleavage in vitro. This suggests that additional factors may influence MKK cleavage in cells, and that there may be additional points of interaction with the LFIR of MKKs that remain to be identified. Solving the structure of an LF–MKK complex will be an important step in understanding why LF has such exquisite specificity for MKKs as substrates.
Consequences of MAPK down-regulation by LF
In common with many other pathogenic bacteria, B. anthracis uses its toxins to impair host defence and immunity as a means to establish infection, and virtually all cells of the immune system require MAPK pathways for proper function. Phagocytic cells of the innate immune system (macrophages and neutrophils) constitute an important initial line of defence against invading microbes before onset of the adaptive immune response. Neutrophils are highly chemotactic phagocytes that engulf and kill bacteria and are thus a key component of early responses to infection. LeTx blocks both neutrophil chemotaxis and superoxide production in response to bacterial components, thus crippling a key component of innate immunity [54–56]. Inhibition of chemotaxis by LeTx is due to a reduction in filamentous actin assembly and establishment of cell polarity. Neutrophil chemotaxis is also blocked both in vitro and in vivo by p38 inhibitors (some studies also implicate the ERK pathway), suggesting a likely mechanism of action [57–60]. As with chemical inhibitors of the ERK and p38 pathways, LeTx only partially inhibits activation of the NADPH oxidase and subsequent superoxide anion production in response to formyl peptide (N-formylmethionyl-leucylphenylalanine) in neutrophils [58,61,62]. LeTx has a dramatic effect, however, on the amplification of this response by pre-treatment with agents such as TNFα (tumour necrosis factor α), a process termed priming. Neutrophil priming appears to involve distinct pathways depending on the stimulus, but a common mechanism is direct phosphorylation of p47phox, a regulatory subunit of the NADPH oxidase, by either p38 or ERK [63,64]. Induction of chemokines in neutrophils is also blocked by MAPK pathway inhibitors, and is probably impaired by LeTx as well .
Macrophages are important mediators of the inflammatory response, and produce pro-inflammatory cytokines such as TNFα, IL (interleukin)-1β and IL-6 in response to infection. Macrophages are activated directly by conserved molecules found in bacteria and other microbes generally termed PAMPs (pathogen-associated molecular patterns), which signal through cell-surface TLRs (Toll-like receptors) . PAMPs that activate macrophages include LPS (lipopolysaccharide) from Gram-negative bacteria as well as lipoteichoic acid and peptidoglycan from the cell wall of Gram-positive bacteria such as B. anthracis. PAMPs found in the cytoplasm, arising, for example, from intracellular pathogens, signal through an alternative family of cytosolic molecules known as NLR proteins, which are discussed in more detail in a later section. TLR agonists potently activate all MAPK pathways, which are important for inflammatory cytokine production at transcriptional and post-transcriptional levels . For example, ERK directly phosphorylates transcription factors such as Elk-1 that are important for TNFα gene induction, and positively regulates nucleocytoplasmic transport of TNFα mRNA [67–70]. JNK and p38 control TNFα production at the translational level in a manner dependent on the ARE (AU-rich element) found in its 3′-UTRs (untranslated regions). p38 enhances TNFα translation by phosphorylation of MAPKAPK2 (MAPK-activated protein kinase 2), which in turn phosphorylates an ARE-binding protein called hnRNP (heterogeneous nuclear ribonucleoprotein) A0 [71–73]. This induces association of hnRNP A0 with the TNFα mRNA, promoting its translation. Consistent with the requirement for MAPKs at multiple levels in its biosynthesis, LeTx inhibits secretion of TNFα, as well as the cytokines IL-1β, IL-6 and IL-10, and no doubt others [32,74,75]. A recent DNA microarray analysis of LeTx-treated macrophages revealed significant down-regulation of immediate early genes encoding transcription factors including c-Jun, c-Fos and Egr-1, an ERK-dependent transcriptional target of Elk-1 that directly activates TNFα transcription [70,76,77]. LeTx also blocks the p38-dependent activation of IRF3 (interferon regulatory factor 3), a transcription factor that is important for induction of multiple TLR target genes, including the chemokine RANTES (regulated upon activation, normal T-cell expressed and secreted) . As discussed in more detail below, under certain circumstances, LeTx induces cell death in macrophages, which is likely to contribute to impairment of the immune system. Interference with the production of soluble inflammatory mediators by LeTx independently of cell death may provide a selective advantage early on during the course of infection, when intact macrophages are needed for germination of spores.
In addition to interfering with innate immunity, LeTx also affects cells involved in the adaptive immune response, which rely on MAPK pathways for proper function as well. LeTx severely impairs dendritic cells, which play an important role in adaptive immunity by acting as antigen-presenting cells for lymphocytes [75,79,80]. LPS-induced up-regulation of co-stimulatory molecules, production of inflammatory cytokines and priming of T-cells in vitro and in vivo are all blocked by LeTx. As with macrophages, combined pharmacological inhibition of the ERK and p38 pathways attenuates cytokine production. LeTx has direct effects on T- and B-cells as well, decreasing production of growth-promoting cytokines and blocking proliferation [81–83].
LeTx affects gene transcription mediated by nuclear hormone receptors, which may have additional immunomodulatory effects in vivo [84–86]. GR (glucocorticoid receptor)-dependent transcription is reduced by LeTx in cultured cells and in living mice in a ligand-independent manner. The effect of LeTx on GR activity in cell culture could be recapitulated by p38 MAPK inhibitors. LeTx does not affect DNA binding to a simple GR target sequence, but does attenuate recruitment to complex promoters. This suggests that p38 may phosphorylate additional transcription factors that bind co-operatively to DNA with the GR. LeTx also diminished the transcriptional activity of other nuclear hormone receptors, including the progesterone receptor and oestrogen receptor-α. Removal of the adrenal gland from mice, which generally increases sensitivity to inflammatory challenge, also increases their susceptibility to killing by injected LeTx. This suggests that, despite attenuation of GR activity, adrenocortical steroids play a protective role in vivo, either through residual GR activity or that of other hormone receptors.
Although curtailing the immune response is an important function for LeTx in promoting infection, it is likely that non-immune cell types contribute to fatality caused directly by circulating LeTx. Cultured human endothelial cells for example undergo apoptosis when exposed to LeTx or MKK1/2 inhibitors [12,87]. LeTx has also been reported to decrease transendothelial electrical resistance in cell monolayers . Enhanced permeability is accompanied by adoption of an elongated morphology with abundant actin stress fibres and a reduction in cell surface VE-cadherin (vascular/endothelial-cadherin). These changes preceded induction of significant cell death, and were not prevented by blocking apoptosis. MAPK pathway inhibitors had opposing effects on the permeability of endothelial cell monolayers, and combined suppression of the ERK, p38 and JNK pathways had no net effect. The role of MKK cleavage in mediating these effects of LF is therefore not clear. The effects of LeTx on endothelial cells may be responsible for increased vascular permeability seen in LeTx-treated mice, which could contribute to toxin-induced mortality.
Mechanisms of macrophage killing by LeTx
Among studies of its cellular activities, the effect of LeTx on macrophages has attracted far and away the most attention and controversy. When treated with LeTx, macrophages and macrophage cell lines derived from some strains of mice undergo rapid cell death accompanied by swelling and lysis, characteristic morphological features of necrosis [3,9]. In contrast, macrophages from other strains of mice are considered to be resistant to LeTx because they do not die when exposed to the toxin alone, despite efficient LF uptake and MKK cleavage. Recent work indicates that activation of even ‘resistant’ mouse macrophages with TLR agonists or inflammatory cytokines renders them sensitive to LeTx-induced cell death [51,89]. Activation-induced macrophage cell death, however, occurs by distinct mechanisms that under some conditions are apoptotic rather than necrotic. LeTx can also cause apoptosis of human monocytes and myelomonocytic cell lines [74,90]. These strain and species differences in macrophage sensitivity also appear to apply to dendritic cells, which are killed by LeTx as well . Although macrophage and dendritic cell death by either apoptosis or necrosis would constitute a general means for disabling host immunity, the specific mechanism by which cell death occurs seems likely to influence the course of disease among various species and strains of infected animals. Although all macrophages are sensitive to killing by LeTx, in the discussion below, I will by convention use the terms ‘susceptible’ and ‘resistant’ to refer to the propensity of macrophages to undergo necrosis when exposed to LeTx alone.
Inflammatory pathways leading to macrophage necrosis by LeTx
LeTx-induced macrophage necrosis is characterized initially by univalent cation permeability, followed by intracellular ATP depletion, loss of mitochondrial function, shutdown of protein synthesis, loss of plasma membrane integrity and eventual cell lysis [92,93]. The process is sensitive to antioxidants and to inhibitors of the proteasome, protein synthesis, calmodulin and calcium channels [94–97]. To date there is no direct evidence relating cleavage of MKKs or MAPK pathways to induction of necrosis; MAPK pathway inhibitors neither cause cytolysis directly, nor do they appear to change the sensitivity of susceptible cells to LeTx. Several MS-based proteomic studies have been conducted recently to globally analyse changes in protein levels induced by LeTx treatment of sensitive macrophage cell lines [98–100]. The three groups used different analytical platforms and observed largely distinct sets of proteins. Conflicting results were obtained for several of the proteins that were common between the studies, possibly due to differences in timing and doses of LeTx used. However, all of the groups observed changes in the levels of proteins involved in energy production and the cellular stress response. Whether the observed changes contribute to or are a consequence of the cell death process is not clear and will require further study.
The pathways involved in triggering macrophage necrosis have been obscure, but recent genetic studies have provided new insight into this phenomenon. The major locus responsible for differences between mouse strains in macrophage sensitivity to LeTx has been mapped to a single gene on chromosome 11 called Nalp1b [101,102]. (Earlier work implicating a nearby gene, encoding the kinesin KIF1C, in controlling susceptibility to LeTx appears to have been an artefact). Nalp1b is one of three genes (along with Nalp1a and Nalp1c) arranged in tandem that are the closest mouse homologues of the single human Nalp1 gene. The Nalp1b gene is highly polymorphic, particularly among the several resistant alleles, one of which is predicted to encode a truncated protein. As macrophage susceptibility is genetically dominant to resistance, it was suggested that susceptibility is caused by the presence of a functional allele. Supporting this idea, a resistant mouse strain carrying a susceptible Nalp1b transgene produces LeTx-sensitive macrophages . In addition, antisense morpholino oligonucleotide-mediated Nalp1b knockdown confers resistance to sensitive macrophages in culture.
Nalp1b encodes a member of the NLR family (also referred to as the NOD-LRR, NACHT-LRR or CATERPILLER family) of intracellular inflammatory mediators [103,104]. This family shares a common central ATPase domain (the NACHT or NOD domain) followed by a variable number of LRRs (leucine-rich repeats) (Figure 5). Most members of this family belong to the NOD and NALP subfamilies that generally have a CARD (caspase activation and recruitment domain) or PYRIN domain respectively at their N-termini. NALP1 proteins have an additional CARD at the C-terminus. NLR proteins act as intracellular sensors for PAMPs and more general ‘danger’ signals that accompany infection by activating pro-inflammatory signalling pathways. NALP3, for example, is activated by intracellular bacterial RNA , and NOD1/2 respond to peptidoglycan. NLR proteins are thus regarded as intracellular analogues of cell-surface TLRs, whose ectodomains that recognize extracellular PAMPs consist largely of LRRs. Some NLR proteins, including human NALP1, are central components of so-called inflammasomes, large multiprotein complexes that activate inflammatory caspases (e.g. caspase 1) that are important for processing cytokines such as IL-1β and IL-18 [105–109]. This complex functions similarly to the apoptosome, which serves as an oligomerization platform to induce processing and activation of apoptotic caspases. NALPs are believed to play a role analogous to that of the apoptosome protein Apaf-1 (apoptotic protease-activating factor 1), which possesses a NACHT-related ATPase domain, and an N-terminal CARD domain (Figure 5).
Representative members of the NLR family
The pro-inflammatory activity of NALP1B appears to be important for LeTx-induced macrophage necrosis. LeTx treatment induces caspase 1 activation in susceptible, but not resistant, macrophages [101,110]. Furthermore, genetic deletion of caspase 1 in mice bearing sensitive alleles of Nalp1b renders their macrophages LeTx-resistant. The mechanism by which NALP1B activates caspase 1 is not yet clear. In humans, an obligate adaptor protein called ASC (apoptosis-associated speck-like protein containing a CARD) interacts with NALP1 through its PYRIN domain and recruits caspase 1 to the inflammasome complex . Mouse NALP1B, however, lacks a PYRIN domain (Figure 5), and it is not yet known whether ASC is required for LeTx-induced caspase 1 activation in mice. Another outstanding question is how LeTx triggers activation of NALP1B. As other NLRs are rendered constitutively active by deletion of the LLRs , direct cleavage of NALP1B by LF is an intriguing possibility that remains to be tested. Interestingly, lowering the intracellular potassium concentration by treatment with extracellular ATP or pore-forming toxins activates caspase 1 by means of the NALP3 inflammasome . As plasma membrane permeability to univalent cations occurs early in LeTx-treated sensitive macrophages , it is possible that NALP1B activation is secondary to potassium stress. Although no direct involvement of MKK cleavage in macrophage necrosis has been described, it is possible that MAPKs hold NALP1B in check under basal conditions, either through direct phosphorylation or by an indirect mechanism. Alternatively, it may be that cleavage of unidentified novel LF substrates leads to NALP1B activation. Lastly, it remains formally possible that LeTx cannot activate NALP1B alone, but must co-operate with inflammatory PAMPs that could contaminate recombinant toxin preparations produced in bacteria.
The dependence on an inflammatory pathway for LeTx-induced necrosis provides a likely explanation for the restriction of this phenomenon to macrophages and dendritic cells. Inflammasome and caspase 1 activation are in fact common hallmarks of non-apoptotic macrophage cell death caused by bacterial toxins. Infection of cultured macrophages with Shigella flexneri and Salmonella species causes rapid caspase-1-dependent necrosis under some conditions [112–114]. Caspase 1 activation and cell death are dependent on protein virulence factors (Salmonella SipB or Shigella IpaB) introduced into target cells by means of the bacterial TTSS (type III secretion system). These factors activate caspase 1 through an inflammasome comprising the NLR protein Ipaf (IL-1β-converting enzyme protease-activating factor) and the ASC adaptor . The mechanism of inflammasome activation by these agents is not clear, although an interaction between Salmonella SipB and caspase 1 has been described.
Macrophage susceptibility to anthrax toxin and virulence factors from other pathogenic bacteria is likely to have an impact on disease pathology. Somewhat counterintuitively, it appears that this phenomenon may in fact be a host response to curtail infection, rather than a mechanism used by the pathogen to promote disease. Mouse strains that harbour sensitive macrophages generally die more slowly when infected with virulent B. anthracis, and are much less susceptible to attenuated strains . Shigella flexneri causes a more lethal disease in mice lacking caspase 1, while the impact of caspase 1 activation on Salmonella Typhimurium virulence is variable and may be mouse strain-dependent [117–119]. Consequences of caspase 1 deficiency, however, may be more related to a generally blunted inflammatory response than to macrophage necrosis itself. The fact that all bacteria known to cause caspase-1-dependent macrophage necrosis are intestinal pathogens raises the question of whether this phenomenon may be most relevant to the pathology of gastrointestinal anthrax as opposed to other forms of the disease. Macrophages from non-human primates are LeTx-resistant, suggesting that the necrotic response is not relevant to human anthrax in general. However, as with mice, it is possible that genetic differences among humans influence macrophage susceptibility and disease pathology.
Whether macrophage necrosis is directly responsible for fatality associated with circulating LeTx, as was once generally believed, now seems questionable. Strains of mice that are killed most potently by LeTx have sensitive macrophages, supporting a contributing role, but there is no general correlation between macrophage susceptibility and overall mortality [120,121]. In addition, there is substantial variability among mouse strains with resistant macrophages in their vulnerability to LeTx, and LeTx-induced death is associated with similar pathology, regardless of macrophage sensitivity . LeTx-induced fatality may be most directly attributable to its impact on other cell types, such as endothelial cells.
Other pathways impacted by LeTx in susceptible macrophages
Transcriptional profiling of LeTx-treated susceptible macrophages unexpectedly revealed modulation of genes involved in the Wnt signalling pathway and up-regulation of Wnt target genes . This led to the finding that LeTx treatment of sensitive, but not resistant, macrophages dramatically decreases levels of GSK-3β (glycogen synthase kinase 3β), a serine/threonine kinase inactivated by the Wnt pathway . GSK-3β inhibition sensitized both susceptible and resistant macrophages to LeTx, suggesting that inactivation of GSK-3β contributes to LeTx-mediated necrosis. GSK-3β is not a direct substrate of LF, and it is unclear why it is degraded in response to LeTx treatment. GSK-3β may simply be a short-lived protein that disappears as a consequence of the LeTx-induced block in protein synthesis. How loss of GSK-3β promotes macrophage necrosis is not clear. GSK-3β inhibition by Akt/PKB (protein kinase B)-mediated phosphorylation is generally correlated with cell survival rather than death, although GSK-3β deficiency sensitizes multiple cell types to apoptosis through down-regulation of NF-κB (nuclear factor κB) . Macrophage activation also leads to Akt/PKB-dependent GSK-3β inhibition that could contribute to inflammasome activation . The involvement of GSK-3β is interesting in the light of the recent discovery of the Wnt co-receptor LRP6 (low-density lipoprotein receptor-related protein 6) as a critical factor for cellular anthrax toxin uptake . The indirect interaction between LRP6 and PA could account for modulation of Wnt target genes in LeTx-treated cells, and may indicate an additional role for LRP6 subsequent to toxin uptake.
LeTx blockade of survival pathways in activated macrophages
In the context of infection, host cells do not encounter anthrax toxins in isolation, but rather alongside inflammatory molecules shed by dividing and dying bacteria. Macrophages are activated by crude B. anthracis cell wall preparations, which are rich in typical Gram-positive bacterial PAMPs . In addition, anthrolysin O, a protein secreted by B. anthracis, is a potent agonist for the LPS receptor TLR4 in macrophages . Two groups have reported that, regardless of genetic background, LeTx is cytotoxic to macrophages when activated by inflammatory TLR agonists [51,89]. There is disagreement over the mechanism by which death occurs under these conditions. One group reported that LPS combined with LeTx triggers non-apoptotic cell death that is mediated by autocrine TNFα production , while the other reported that death occurred by TNFα-independent apoptosis . Other groups have reported apoptosis of susceptible macrophage cell lines at sub-necrotic doses of LeTx either alone  or more robustly in the presence of the phosphatase inhibitor calyculin A . Since high LeTx concentrations completely abolish TNFα secretion , it is possible that activation-induced cell death depends on autocrine TNFα signalling at intermediate amounts of LeTx, but is TNFα-independent with higher amounts. If this is the case, both mechanisms could operate in vivo at distinct stages of disease.
TNFα can substitute for direct signalling through TLRs in causing macrophage cell death when combined with LeTx . Pharmacological inhibition of MAPK pathways either alone or in combination could not substitute for LeTx, although blocking the ERK or JNK pathway potentiated death from TNFα in the presence of LeTx. Interestingly, cell death caused by TNFα plus LeTx could be partially blocked by rapamycin treatment, suggesting a role for the mTOR (mammalian target of rapamycin) kinase in this pathway. It is tempting to speculate that in this context, TNFα provides a signal for inflammasome activation, promoting caspase 1 processing that leads to necrotic death of resistant macrophages. However, it is not yet known whether TNFα-induced cell death is similar phenotypically to that caused by LeTx alone in susceptible macrophages, or if it occurs by a distinct non-apoptotic pathway.
Inflammatory stimuli trigger a number of pro-apoptotic pathways, and cell survival is attributed to concomitant activation of survival pathways, most prominently those impinging on the transcription factor NF-κB (Figure 6). LeTx does not affect NF-κB activity directly, but a number of NF-κB target genes also require p38 activity for induction [51,130]. Accordingly, inactivation of the p38-α MAPK through MKK6 cleavage appears to be the principal mechanism by which LeTx causes TLR4-dependent apoptosis . One of the downstream targets for p38 that co-operates with NF-κB is CREB (cAMP-response element-binding protein), a bZIP (basic leucine zipper) transcription factor. CREB is activated by phosphorylation at Ser-133 within its transcriptional activation domain, promoting recruitment of transcriptional co-activators . Although classically mediated by PKA (cAMP-dependent protein kinase), a large number of cellular stimuli induce CREB Ser-133 phosphorylation, acting through several distinct protein kinases. p38 phosphorylates and activates MSK (mitogen- and stress-activated kinase) 1 and 2, members of the RSK (p90 ribosomal S6 kinase)/MAPKAPK family, which are responsible for CREB phosphorylation in response to growth factors and UV irradiation . These kinases also appear to phosphorylate CREB in LPS-stimulated macrophages, and thus represent critical targets of the p38 pathway in LeTx-treated cells .
Macrophage apoptosis pathways
Two of the genes co-operatively induced by the NF-κB and p38 pathways, Bfl-1/A1 and Pai-2, appear to contribute to the survival of activated macrophages . BFL-1/A1 is an anti-apoptotic member of the Bcl-2 family. Like Bcl-2, it is likely to promote cell survival by antagonizing pro-apoptotic members of the family [such as BAD (Bcl-2/Bcl-XL-antagonist, causing cell death) and BAX (Bcl-2-associated X protein)] that mediate release of cytochrome c and other apoptotic proteins from the mitochondria . PAI-2 (plasminogen-activator inhibitor 2) is a serpin-type serine protease inhibitor that had been implicated previously as a cell survival factor . The anti-apoptotic activity of PAI-2 depends on residues in its reactive loop, suggesting that it functions as a protease inhibitor in this context, but the relevant target is not known.
Pharmacological inhibition of the ERK and JNK pathways alone do not cause activation-induced apoptosis in macrophages, and their contributions to cell survival may be redundant in the context of LeTx activity. The ERK pathway has a well-established anti-apoptotic role mediated by RSK, a direct target of ERK. RSK phosphorylation of the pro-apoptotic Bcl-2 family member BAD triggers sequestration by 14-3-3 proteins, thus promoting cell survival [135,136]. In LPS-stimulated macrophages, ERK phosphorylates and destabilizes another pro-apoptotic Bcl-2 homologue, BIM (Bcl-2-interacting mediator of cell death) . As in other cell types, JNK can be either pro- or anti-apoptotic in macrophages depending on the activating stimulus [138–141]. Direct phosphorylation of BIM (at a site distinct from those phosphorylated by ERK) has been implicated in mediating anti-apoptotic effects of JNK signalling . In addition, phosphorylation by JNK activates transcription factors such as Ets-2, PU.1 and JunD, leading to induction of anti-apoptotic genes [141,143].
What are the pro-apoptotic pathways that contribute to activation-induced macrophage death? Apoptosis induced by LeTx accompanied by LPS is diminished in macrophages lacking PKR (protein kinase R) . PKR is a member of a small family of stress-responsive eIF-2α (eukaryotic initiation factor-2α) kinases that share a homologous catalytic domain yet are activated by distinct inputs . PKR is activated by direct binding of double-stranded RNA and can induce apoptosis as an antiviral strategy. General inflammatory stimuli such as LPS also induce rapid and transient activation of PKR through an uncharacterized mechanism. Ser-51 phosphorylation of eIF-2α by PKR results in an overall decrease in general protein synthesis. However, eIF-2α phosphorylation also induces selective translational up-regulation of specific mRNAs that are normally repressed through non-productive translation of sequences in the 5′-UTR. It is likely that some of these translational targets are pro-apoptotic factors that trigger cell death in macrophages that lack necessary survival signals.
B. anthracis is one of several bacteria known to induce activation-dependent apoptosis in macrophages. Interestingly, MKK inhibition is also one of the mechanisms used by another bacterial virulence factor, YopJ from Yersinia. Yersinia infect macrophages, and introduce YopJ among an array of proteins into the host cytosol by means of a TTSS. YopJ inactivates all MKKs, as well as the IKKs [IκB (inhibitory κB) kinases] that mediate activation of NF-κB, by a novel mechanism that was elucidated only recently . YopJ was found to be an acetyltransferase that modifies its targets on the same activation loop serine and threonine residues phosphorylated by upstream kinases. By blocking phosphorylation, YopJ maintains its targets in an inactive state. Infection of macrophages with Yersinia inhibits inflammatory cytokine production and causes apoptosis in a YopJ-dependent manner . The capacity of YopJ to promote cell death has generally been attributed to its inactivation of NF-κB, but blockade of MKKs also appears to have an important role, and may operate through similar mechanisms as described for LF above . As with B. anthracis, Yersinia requires TLR signalling and PKR to cause apoptosis . That killing of macrophages is a common strategy used by numerous bacteria underscores the critical role that these cells play in defending the host against infection.
Many pathogenic bacteria produce toxins that act by increasing intracellular cAMP in host cells . Classical ADP-ribosyltransferase toxins, such as cholera toxin and pertussis toxin, do so indirectly through modification of heterotrimeric G-protein α subunits, while other bacterial toxins are themselves AC enzymes. EF is an extraordinarily active Ca2+ and CaM (calmodulin)-dependent AC consisting of an N-terminal PA-binding domain, a central catalytic core and a C-terminal helical domain [150,151]. The N-terminus is homologous in sequence and structure to the corresponding domain from LF, and residues conserved between the two are required for PA binding and toxin uptake . The catalytic domain shares homology with other bacterial AC toxins, including Bordetella pertussis CyaA and Pseudomonas aerunginosa ExoY. The unique helical domain of EF packs against the catalytic core in the inactive state. CaM inserts between the helical and catalytic domains in an extended conformation, separating them and inducing a 30° rotation of the two domains relative to one another (Figure 7). The extensive interactions between CaM and EF induce conformational changes in the linker region between the two domains and within two regions of the catalytic domain. This correctly positions many of the residues involved in substrate binding, thus activating the enzyme. Interestingly, Bordetella pertussis CyaA, which lacks the helical domain, interacts with CaM by means of a distinct set of non-conserved residues in the catalytic core, indicating that EF has a unique mode of binding to CaM .
EF activation by CaM
Like LeTx, a major role for EdTx appears to be disabling of host defence by impairing the function of phagocytic cells. Increased intracellular cAMP induced by EF or other means inhibits neutrophil chemotaxis, phagocytosis, superoxide production and microbicidal activity (Figure 8) [56,154–157]. These effects have been generally attributed to activation of PKA, although its downstream targets have not been identified. PKA activity in normal physiology is subject to tight spatiotemporal regulation [158–160]. An asymmetric distribution of active PKA, for example, is required for establishment of polarity and migration of neutrophils . EdTx, in contrast, induces waves of activated PKA emanating from perinuclear foci, probably due to transit of EF through late endosomes . Thus EdTx will effectively inhibit some cAMP-dependent processes by generating high levels of mislocalized cAMP.
Signalling pathways affected by EF in phagocytes
Monocytes and macrophages are also likely targets for EdTx. Cytokine production by activated monocytes is differentially affected by EdTx, with inhibition of TNFα accompanied by increased IL-6 production . Raising intracellular cAMP levels blocks LPS-induced activation of the ERK and JNK MAPK pathways in monocytes through activation of PKA, which could contribute to suppression of cytokine production by EdTx [164–166]. Activation of the other major target for cAMP, the guanine-nucleotide-exchange factor Epac1 (exchange protein directly activated by cAMP 1), inhibits phagocytosis by both alveolar macrophages and monocyte-derived macrophages and probably depends on its small G-protein effector, Rap1 [167,168]. Like LeTx, EdTx also inhibits T-cell proliferation and accompanying cytokine production, which is accompanied by inhibition of ERK and JNK MAPK pathways , indicating that EdTx can interfere with adaptive immunity as well.
The mechanisms by which intravenous delivery of EdTx causes death in mice are not known. Prolonged treatment with EdTx causes apoptosis in macrophage cell lines and 3T3 fibroblasts, but not in human endothelial cells or CHO (Chinese-hamster ovary) cells [87,169]. Direct cytotoxicity of EdTx to other unidentified cell types may therefore mediate some of the pathological features of systemic anthrax infection.
INTERACTIONS BETWEEN EF AND LF
Because their expression is co-ordinately regulated, EdTx and LeTx circulate together during infection. The two toxins target a similar spectrum of cell types yet operate by distinct mechanisms, suggesting that they may synergistically impair cellular function. Although generally studied separately, there is evidence for cooperativity between EF and LF. Superoxide production by neutrophils and cytokine production by dendritic cells, for example, are both inhibited co-operatively by EF and LF [56,80]. Interestingly, EF causes up-regulation of the PA receptors TEM8 (tumour endothelial marker 8) and CMG2 (capillary morphogenesis gene 2) specifically in macrophage cell lines, and increases their sensitivity to LF cytotoxicity [76,170]. Melanoma cell lines produce melanin in response to either LeTx treatment or MKK1 inhibition, and this effect is boosted by EdTx . This phenomenon may be relevant to the black pigmentation characteristic of skin lesions associated with cutaneous anthrax.
EF appears to counteract macrophage apoptosis induced by LF . Ser-133 of the transcription factor CREB is phosphorylated directly by PKA as well as by p38-dependent kinases. EF thereby maintains CREB phosphorylation and transcriptional activity in the absence of p38 activity, allowing for continued expression of anti-apoptotic CREB targets (see Figure 6). Macrophage survival during infection may therefore depend on the relative abundance of EF and LF at a particular stage of disease or site within the organism.
Much insight into the impact of anthrax toxins on cellular signalling pathways has been obtained in recent years, but a number of important questions remain unanswered. For example, what is the mechanism for NALP1B activation by LeTx? It will be important to resolve whether MKK cleavage is involved in this process, and, if not, what the relevant LF substrates are. Other cellular factors involved in caspase 1 activation that act as components of a NALP1B inflammasome also remain to be identified. How caspase 1 causes cell death is not understood, and it will be of interest to know whether MAPK pathways are also important for cell survival in this context. Regarding EF, we have little detailed understanding of the ways in which increased cAMP disrupts transcription-independent processes such as activation of the NADPH oxidase. Does superoxide generation, like cell migration, depend on localized PKA activity? Does PKA phosphorylate components of the oxidase complex directly? One outcome of a detailed understanding of the signalling pathways affected by EF and LF is that it allows the generation of hypothesis regarding the molecular basis of disease that are testable in animal models. For example, does PKR deficiency in mice affect their susceptibility to infection? Does caspase 1 promote mortality associated with LeTx administration in mice? Answering these questions will help in shaping our understanding of the fundamental mechanisms underlying the pathogenesis of anthrax.
I am grateful to Derek Abbott and Lise Thomas for providing helpful comments on the manuscript. This work was supported by U.S. Army Medical Research and Materiel Command Award DAMD17-03-1-0062.
apoptotic protease-activating factor 1
apoptosis-associated speck-like protein containing a caspase activation and recruitment domain
Bcl-2/Bcl-XL-antagonist, causing cell death
Bcl-2-interacting mediator of cell death
caspase activation and recruitment domain
cAMP-response element-binding protein
eukaryotic initiation factor-2α
glycogen synthase kinase
heterogeneous nuclear ribonucleoprotein
c-Jun N-terminal kinase
low-density lipoprotein receptor-related protein 6
mitogen-activated protein kinase
MAPK-activated protein kinase
nuclear factor κB
plasminogen-activator inhibitor 2
pathogen-associated molecular pattern
cAMP-dependent protein kinase
protein kinase B
protein kinase R
p90 ribosomal S6 kinase
tumour necrosis factor α
type III secretion system