Activation of cell stress response pathways by Shiga toxins

Introduction: Shiga toxins

Shiga toxins (Stxs) are genetically and structurally related cytotoxins expressed by the enteric pathogens Shigella dysenteriae serotype 1 and an expanding number of Shiga toxin-producing Escherichia coli (STEC) serotypes (Gyles, 2007). Ingestion of small numbers of Stx-producing bacteria in contaminated food or water may lead to bloody diarrhoea (bacillary dysentery or haemorrhagic colitis). Unfortunately, these patients are at risk for developing life-threatening extra-intestinal complications including acute renal failure and neurological abnormalities such as seizures and paralysis (Tarr et al., 2005; Proulx and Tesh, 2007; Obata, 2010). Shiga toxin, the prototypical member of the toxin family, is expressed by S. dysenteriae serotype 1. Stxs expressed by STEC may be categorized as Shiga toxin type 1 (Stx1), which is essentially identical to Shiga toxin, and Shiga toxin type 2 (Stx2), which is 56% homologous to Shiga toxin/Stx1 at the deduced amino acid sequence level (Jackson et al., 1987). Stx1 and Stx2 genetic variants have been described. Stxs are encoded by late genes of lambdoid bacteriophage (reviewed in Allison, 2007). The Stx2 operon is under control of the P_R and P_R′ promoters, and toxin production is optimal under conditions that induce the phage lytic cycle. An additional iron-regulated promoter adjacent to the Stx1 operon appears sufficient to induce Stx1 transcription, although Stx1 translocates to the bacterial periplasmic space rather than being released into the environment (Wagner et al., 2002).

Stxs are AB₅ holotoxins, consisting of an enzymatic A subunit (∼32 kDa) in non-covalent association with five B subunits, each B subunit protein being ∼7.7 kDa. B subunits pentamerize to form a ring, and the C-terminus of the A subunit inserts into the central pore (Fraser et al., 2004). B subunits of Stxs that cause disease in humans bind to the neutral globo-series glycolipid globotriaosylceramide (Gb₃). Following toxin binding to Gb₃-expressing cells, the holotoxin is internalized via a mechanism that may initially involve B subunit-induced negative curvature of the host cell membrane leading to the formation of toxin-containing membrane invaginations (Römer et al., 2007). The toxin undergoes a process termed retrograde transport whichinvolves routing via early/recycling endosomes to the trans-Golgi network, through the Golgi apparatus, to reach the ER lumen. During transport, the A subunit is cleaved by furin or a furin-like protease to form a 27 kDa A₁ fragment and a 4 kDa A₂ fragment which remain associated via a disulfide bond (Sandvig et al., 1992; Garred et al., 1995). The ER is the site of disulfide bond reduction, unfolding of the A₁ fragment, and retrotranslocation of the A₁ fragment possibly via the Sec61 translocon into the cytoplasm (Lord et al., 2005; Tam and Lingwood, 2007). The A₁ fragment possesses N-glycosidase activity and cleaves a single adenine residue (at position 4324 in the rat) from the 28S rRNA component of eukaryotic ribosomes (Endo et al., 1988). The target adenine residue is unpaired in a region of non-Watson-Crick base pairing called a GAGA tetraloop. This region of the 28S rRNA is also called the α-sarcin/ricin loop since the enzymatic action of these two ribosome-inactivating proteins (RIPs) is also directed to this loop. The depurination reaction blocks association with elongation factors, resulting in protein synthesis inhibition. Because Vero cells (African green monkey renal epithelial cells) are highly sensitive to protein synthesis inhibition and cell death by Stxs, the toxins are also referred to as verotoxins or verocytotoxins. Readers are directed to several reviews for more detailed information on toxin structure, receptor binding, internalization, retrograde transport and retrotranslocation (Johannes and Römer, 2010; Lingwood et al., 2010; Sandvig et al., 2010).

While the enzymatic action of Stxs is well characterized, the precise relationship between rRNA depurination/protein synthesis inhibition and cell death remains unclear. It has become evident that Stxs induce apoptosis or programmed cell death in many cell types in vitro and in vivo (reviewed in Tesh, 2010). Thus, recent studies have focused on the exploration of cell death signalling mechanisms activated by the toxins. Stxs are effective signalling molecules activating multiple stress responses in eukaryotic cells. While protein synthesis inhibition may contribute to cell death, Stx-induced protein synthesis inhibition may be dissociated from cell death signalling in some cell types. This Microreview examines cell stress responses activated by Stxs following the depurination reaction (ribotoxic stress response) or by the presence of unfolded proteins within the ER (unfolded protein response). Signalling through these pathways may be involved in the induction of cytokine/chemokine expression and programmed cell death, processes which contribute to the pathogenesis of disease caused by Stxs.

Shiga toxins activate the ribotoxic stress response

The term ribotoxic stress response was introduced by Iordanov et al. (1997) who showed that site-specific modifications to the ribosomal peptidyl transferase reaction centre activated signalling through the c-Jun N-terminal (JNK) mitogen-activated protein kinase (MAPK) pathway. Rat-1 fibroblasts were treated with protein synthesis inhibitors acting at different ribosomal sites to disrupt translation. Anisomycin, which binds to domain V of 28S rRNA, rapidly activated the JNK1 isoform with a greater than 15-fold activation within 15 min of treatment. The response was sensitive to anisomycin with half maximal activation occurring at doses mediating < 10% reduction in [³H]-leucine incorporation into polypeptides. Other RIPs such as pactamycin and emetine effectively inhibited protein synthesis (> 95%), but failed to activate JNK signalling. Thus, signal transduction and protein synthesis inhibitory activities could be dissociated. Anisomycin-specific signalling required alteration of functional ribosomes rather than directly activating the JNK pathway since pre-treatment of Rat-1 cells with pactamycin or emetine blocked the capacity of anisomycin, but not IL-1α, to activate JNK1. The analysis was extended to RIPs that act on domain VI (α-sarcin/ricin loop) of 28S rRNA including the fungal ribonuclease α-sarcin and the plant toxin ricin A-chain. While α-sarcin cleaved the phosphodiester bond adjacent to A4324, and ricin A-chain depurinated 28S rRNA at position A4324, both RIPs activated JNK1 and the upstream MAPK kinase (MAP2K) SEK1/MKK4. In contrast, toxins which ADP-ribosylated EF-2 but failed to mediate rRNA damage in the α-sarcin/ricin loop, failed to activate JNK1.

One might predict that Stxs, like ricin, activate the ribotoxic stress response since these toxins share identical enzymatic activities. Treatment of human macrophage-like (differentiated) THP-1 cells with Stx1, ricin or anisomycin activated JNK and p38 MAPK signalling cascades (Foster and Tesh, 2002; Cherla et al., 2006). Activation was rapid, with JNK1, JNK2 and p38 MAPK phosphorylation peaking 3–6 h after toxin exposure. Within this short time frame, exposure of THP-1 cells to Stx1 transiently increased total protein synthesis and the cells were relatively resistant to the rapid onset of cytotoxicity characterized using other cell types such as Vero cells (Foster et al., 2000; Harrison et al., 2005). Thus, Stx1 induction of the ribotoxic stress response in macrophage-like cells did not appear to require rapid protein synthesis inhibition or cell death. In contrast to stress-activated protein kinases, JNK and p38, Stx1 induced modest and transient activation of extracellular signal-regulated kinases (ERK). Patients infected with STEC may have elevated serum titres of anti-STEC lipopolysaccharide (LPS) antibodies (Karmali, 1998) and LPS bound to blood cells (Ståhl et al., 2009), suggesting that intestinal damage may be sufficient to allow LPS to enter the circulation. LPS are effective activators of innate immunity, and treatment of macrophage-like cells with Stx1 and LPS significantly increased activation of all three MAPK cascades (Cherla et al., 2006), suggesting that the presence of both Stxs and LPS in the circulation may be an important determinant in the progression of disease caused by Stx-producing bacteria.

Activation of the ribotoxic stress response by Stx1 played a critical role in the induction of cytokine/chemokine expression in macrophage-like cells. Inhibitors of JNK, p38 and ERK signalling partially blocked, and simultaneous use of all three inhibitors extensively blocked, toxin induced expression of soluble IL-1β and IL-8 (Cherla et al., 2006). An inhibitor of MAPK-interacting kinase 1 (Mnk1), a downstream substrate of p38 MAPK, also inhibited cytokine expression. Primary human peripheral blood monocytes treated with Stx1 or Stx2 showed a slightly different pattern of MAPK activation, with the JNK pathway being transiently activated and the ERK and p38 MAPK pathways more strongly activated (Cameron et al., 2003). Inhibitors of p38 MAPK signalling were effective in blocking TNF-α and GM-CSF production in response to the toxins. Gray et al. (2008) showed that Stx1 treatment of the human monocytic cell line U937 increased IL-8 production, which was reduced ∼80% by pre-treatment of cells with PKR inhibitors. A similar phenomenon was noted using ribotoxic stress inducers ricin and deoxynivalenol (a trichothecene mycotoxin). When U937 cells stably transfected with a non-functional PKR mutant were used, elevated IL-8 levels were not detected following treatment with Stx1, ricin or deoxynivalenol. Optimal IL-8 expression induced by deoxynivalenol required a second kinase, haematopoietic cell kinase (Hck) which associates with the 40S ribosomal subunit and triggers activation of ASK1, MKK3/6 and p38 MAPK (Bae et al., 2010). Additional studies will be required to determine whether multiple kinases must interact with ribosomes for Stx activation of the ribotoxic stress response.

While the capacity of Stxs to activate MAPKs is well defined, and some of the upstream MAP2Ks and MAP3Ks involved in the ribotoxic stress response have been identified, proximal signalling events linking the depurination reaction with the initiation of signalling cascades remain to be fully characterized. Recently, Stx A₁ fragments and the ricin A-chain have been shown to associate with acidic ribosomal phosphoproteins that comprise the ribosomal stalk, a protuberance of the large ribosomal subunit involved in the recruitment and binding of initiation and elongation factors. The interaction of Stxs and ricin with the ribosomal stalk appears to be required for the depurination reaction within the α-sarcin/ricin loop (Chiou et al., 2008; McCluskey et al., 2008). Gray et al. (2008) hypothesized that the interaction of Stx A₁ fragments with ribosomes may alter ribosomal tertiary structure and/or toxin-mediated 28S rRNA damage may alter rRNA secondary structure. PKR is a serine/threonine kinase which binds to, and is activated by, damaged ribosomes via interaction with two dsRNA-binding domains (Nallagatla et al., 2011). Activated PKR phosphorylates eIF-2α at Ser⁵¹, leading to inhibition of overall translation, although the expression of genes involved in the host cell response to stress is maintained or upregulated. Furthermore, there is evidence that the association of PKR with damaged ribosomes creates an activation scaffold for the direct binding and activation of JNK and p38 MAPKs (Iordanov et al., 2000; Alisi et al., 2008; Zang et al., 2009). Thus, the observation that PKR is activated in Stx1-treated U937 cells implicates this kinase as an immediate sensor of ribosomal damage and suggests that Stxs may inhibit protein synthesis through multiple mechanisms.

Once activated, the ribotoxic stress response regulates cytokine/chemokine expression at the transcriptional level through the activation of transcription factors such as NF-κB, AP-1 and Egr-1 (Sakiri et al., 1998; Zoja et al., 2002; Leyva-Illades et al., 2010), and through post-transcriptional mechanisms involving mRNA stabilization, and increased ribogenesis and translation initiation (Thorpe et al., 2001; Harrison et al., 2004; Cherla et al., 2006). Stxs also induce expression of dual specificity phosphatases (Kojima et al., 2000; Leyva-Illades et al., 2010). Thus, the ribotoxic stress response induced by Stxs may include the activation of mechanisms that ultimately downregulate the response.

Smith et al. (2003) linked apoptosis with Stx-induced signalling through the ribotoxic stress response using the human epithelial cell line Hct8. Stx1, but not an enzymatic Stx1 mutant, triggered caspase-3 activation and DNA fragmentation. JNK and p38 MAPKs were activated by Stx1 and pharmacological inhibition of p38 MAPK signalling reduced caspase-3 activation and DNA fragmentation, and partially protected Hct8 cells from apoptosis. Anisomycin and UV-light are effective ribotoxic stress response inducers, and an inhibitor of the MAP3K zipper sterile-α-motif kinase (ZAK), or the knock-down of ZAK expression using small interfering (si)RNAs, protected COS-7 cells from apoptosis, and blocked JNK and p38 MAPK phosphorylation, induced by anisomycin or UV-light treatment. Neither ZAK inhibitor treatment nor transfection with ZAK siRNAs blocked JNK and p38 MAPK activation induced by TNF-α or IL-1β, specifically linking ZAK as an upstream signalling molecule in the ribotoxic stress response leading to apoptosis (Wang et al., 2005). The ZAK inhibitor and ZAK siRNAs blocked Stx2- and ricin-mediated activation of stress-activated protein kinases and partially protected Hct8 and Vero cells from apoptosis (Jandhyala et al., 2008). Inhibition of signalling through ZAK blocked Stx2-induced caspase-3 activation, but DNA fragmentation was not altered, suggesting that additional signalling pathways may be activated by Stxs to trigger DNA scission events. Treatment with the ZAK inhibitor did not alter Stx2-mediated protein synthesis inhibition. In contrast to intestinal epithelial cells, the Ramos Burkitt's lymphoma cell line appears to express basal levels of activated p38 MAPK, and Stx1 treatment did not increase p38 MAPK activation above these levels. Inhibitors of p38 MAPKs actually increased apoptosis induced by Stx1 treatment of Burkitt's lymphoma cells (Garibal et al., 2010). Thus, in lymphoid cells, prolonged signalling through p38 MAPKs may induce survival pathways that protect the cells from toxin-induced apoptosis. In summary, Stx-induced signalling through the ribotoxic stress response may differentially activate MAPK cascades leading to cell death or cell survival signalling in different cell types (Fig. 1).

Shiga toxins activate the unfolded protein response (UPR)

The ER is the intracellular site for the correct folding and post-translational processing of proteins destined to be transported to locations within the cell or secreted. The ER is also involved in Ca²⁺ homeostasis and storage. Recently, an intracellular ‘quality control’ mechanism has been defined which assesses the status of protein folding and Ca²⁺ storage (Bernales et al., 2006; Malhotra and Kaufman, 2007). Three ER-localized transmembrane proteins ‘sense’ levels of unfolded proteins: RNA-dependent protein kinase-like ER kinase (PERK), inositol-requiring ER to nucleus signal kinase-1 (IRE1) and activating transcription factor-6 (ATF6). PERK is a serine/threonine kinase, IRE1 is a multifunctional protein with kinase and endoribonuclease activities, and ATF6 is a transcription factor. The sensor molecules associate with the chaperone binding immunoglobulin protein (BiP, also known as GRP78). When unfolded or improperly processed proteins accumulate in the ER, BiP dissociates from the sensor molecules (Bertolotti et al., 2000). Subsequently, PERK and IRE1 are activated by homo-oligomerization and proximity-dependent autophosphorylation, while ATF6 activation requires translocation from the ER membrane to the Golgi apparatus and proteolytic cleavage by site 1 and site 2 proteases. Activation of the sensor molecules leads to a transient, co-ordinated response involving the attenuation of overall protein translation coupled with the transcriptional activation of a subset of genes encoding chaperones (for correct protein folding) and proteins involved in degradation of unfolded proteins via the ER-associated protein degradation (ERAD) pathway. This co-ordinated response is called the unfolded protein response (UPR). Failure to correct protein folding defects and maintain Ca²⁺ homeostasis may lead to prolonged signalling through the UPR, which in turn, may activate apoptotic signalling events (Tabas and Ron, 2011). A key transcriptional factor in the UPR is C/EBP homologous protein (CHOP, also called Gadd 153). CHOP both positively and negatively regulates the expression of genes involved in apoptosis (McCullogh et al., 2001).

Stxs associate with ER-localized chaperone proteins HEDJ/ERdj3 and BiP (Yu and Haslam, 2005; Falguiéres and Johannes, 2006), suggesting that during retrotranslocation through the Sec61 translocon, Stx A₁ fragments probably exist in a transient unfolded state. The induction of apoptosis by Stxs appears to require toxin enzymatic activity in most cell types examined (reviewed in Tesh, 2010). Based on these observations, Lee et al. (2008) reasoned that Stxs may activate the UPR via multiple mechanisms: the transient unfolding of Stx A₁ fragments activates the UPR while the protein synthesis inhibitory activity of the toxins leads to the accumulation of unfolded host proteins within the ER and/or the alteration intracellular Ca²⁺ levels. Stxs may signal apoptosis, therefore, through prolonged UPR signalling. Human monocyte-like (undifferentiated) THP-1 cells are relatively sensitive to killing by Stxs, and Stx1 treatment of the cells activated all UPR sensors within 2 h of intoxication. Stx1 treatment led to the functional activation of the UPR: XBP-1, the mRNA transcript for X-Box Protein-1, was spliced by activated IRE1 to encode the functional transcription factor, eIF-2α was phosphorylated by activated PERK, and ATF6 was cleaved from the inactive 90 kDa form to the active 50 kDa transcription factor. CHOP expression was upregulated within hours of Stx1 treatment of THP-1 cells. CHOP is known to differentially regulate the expression of death receptor 5 (DR5, also known as TRAIL-R2) and the anti-apoptotic protein Bcl-2 (McCullogh et al., 2001; Yamaguchi and Wang, 2004). Stx1 treatment of monocyte-like THP-1 cells upregulated the expression of DR5, and downregulated the expression of Bcl-2. The capacity of Stxs to mediate the release of Ca²⁺ from intracellular stores was associated with the activation of calpains, which in turn, cleaved procaspase-8 and induced apoptosis. Stx1 enzymatic mutant and purified Stx1 B subunits failed to trigger apoptosis or fully activate the UPR so that upregulated expression of CHOP and the differential modulation of DR5 and Bcl-2 expression were not detected. However, there was evidence of IRE1 activation and XBP1 mRNA splicing using these inactive toxin preparations, suggesting that the presence of Stxs within the ER may be sufficient to initiate the UPR, but unfolded Stx A₁ fragments and/or toxin enzymatic activity may be necessary to maintain the UPR leading to apoptosis. A summary of the UPR leading to apoptosis induced by Stx1 treatment of monocytic THP-1 cells is presented in Fig. 2.

The response of macrophage-like (differentiated) THP-1 cells to Stxs is more complex: protein synthesis is transiently upregulated after intoxication, pro-inflammatory cytokine and chemokine expression is induced, and both survival and apoptosis pathways appear to be simultaneously activated (Foster et al., 2000; Harrison et al., 2005; Lee et al., 2007). However, over time macrophage-like cells die in response to intoxication, a phenomenon termed the delayed apoptosis phenotype. Treatment of the cells with Stxs led to the rapid activation of PERK and IRE1, and eIF-2α was phosphorylated and XBP1 mRNA was spliced. However, ATF6 activation was not detected in Stx1 treated macrophage-like THP-1 cells, suggesting that cell maturation correlates with loss of processing and signalling through ATF6. The anti-apoptotic protein Bcl-2 emerged as a critical regulator determining the kinetics of cell death induced by Stxs in THP-1 cells. In contrast to monocyte-like cells, where Bcl-2 expression was downregulated by Stx1, treatment of macrophage-like THP-1 cells with Stx1 increased Bcl-2 expression and mediated an increased translocation of Bcl-2 to mitochondria (Lee et al., 2009). This study also identified ‘cross-talk’ between the ribotoxic stress response and the UPR activated by Stxs. The anti-apoptotic function of Bcl-2 requires JNK-mediated phosphorylation of Bcl-2 at Ser⁷⁰ (Deng et al., 2001). Alternative Bcl-2 phosphorylation reactions, including p38 MAPK-directed phosphorylation of Bcl-2 at amino acid Thr⁵⁶, inhibit Bcl-2 function (DeChiara et al., 2006). Bcl-2 was differentially phosphorylated by Stx1 treatment of monocyte- versus macrophage-like THP-1 cells. Levels of anti-apoptotic Ser⁷⁰-phospho-Bcl-2 molecules were transiently increased in macrophage-like cells, while levels declined in monocyte-like cells. In contrast, levels of Thr⁵⁶-phospho-Bcl-2 declined in toxin-treated monocyte-like cells, and Bcl-2 phosphorylated at position Thr⁵⁶ was not detected in Stx1-treated macrophage-like THP-1 cells (Lee et al., 2009). Thus, the ribotoxic stress response induced by Stxs may regulate the activation of the Bcl-2 family of proteins that, in turn, control apoptosis.

Studies to characterize the role of the UPR induced by Stxs in apoptosis in other cell types are limited. Stx1 and Stx2 appear to differentially activate UPR sensors in the human renal tubule epithelial cell line HK-2 so that Stx1 is more effective at activating ATF6, while Stx2 primarily triggered the phosphorylation of PERK and IRE1 (Lentz et al., 2011). Microarray analysis of genes modulated by Stx2 treatment of human brain microvascular endothelial cells revealed that transcripts for genes involved in the UPR were upregulated, including the genes encoding PERK, CHOP and ATF4 (Fujii et al., 2008). Subtilase cytotoxin, a newly described AB₅ holotoxin expressed by STEC, which undergoes retrograde transport to the ER where it selectively cleaves the chaperone BiP, has been shown to activate all UPR stress sensors in Vero cells (Wolfson et al., 2008). In contrast to these studies, Parikh et al. (2008) showed that the transfection of plasmids expressing mature ricin A-chain into yeast expressing a UPR element::lacZ gene reporter construct resulted in decreased β-galactosidase expression, suggesting that the enzymatic activity of ricin may suppress the UPR. Yeast cells lack the ER membrane sensors PERK and ATF-6, so that Ire1 is the sole resident ER membrane protein involved in the UPR (Kohno, 2010). However, Wang et al. (2011) used bovine and human epithelial cell lines to show that ricin A-chain failed to activate Ire1, to mediate XBP-1 mRNA splicing, and to phosphorylate eIF-2α. Furthermore, ricin A-chain inhibited tunicamycin-induced XBP-1 splicing and DTT-induced eIF-2α phosphorylation. The cleavage of procaspase-7 and -3 was dramatically increased in cells treated with ricin A-chain plus the UPR inducers tunicamycin or DTT compared with treatment with ricin A-chain alone, suggesting that ricin-induced inhibition of the UPR sensitized the cells to cytotoxicity. These results highlight the need for caution in formulating generalizations on the role of the UPR in cell death signalling, and justify the need for additional studies to characterize the roles the UPR in the host response to Stxs and other RIPs.

Role of Shiga toxin-activated cell stress responses in pathogenesis

There is limited information available in the literature on the relationship between Stx-induced activation of stress responses characterized in multiple cells types in vitro, and pathogenesis of disease caused by Stxs in humans or in animal models. Furthermore, the cell types which specifically respond to Stxs by inducing the ribotoxic stress response or the UPR have not been characterized in vivo. Psotka et al. (2009) showed that the administration of a caspase inhibitor in mice challenged with Stx2 and LPS reduced the numbers of TUNEL-positive cells detected in renal tissue sections, and reduced indicators of renal failure (BUN and urine osmolality). Stxs are known to activate MAPK cascades and the UPR in epithelial cells in vitro (Smith et al., 2003; Lentz et al., 2011) and it would be interesting to correlate activation of these stress responses with tissue-specific production of cytokines and chemokines and the induction of apoptosis in vivo. Studies employing animal models of ricin intoxication may be informative in the design of these experiments. For example, Korcheva et al. (2005) showed that the intravenous administration of ricin to mice resulted in the activation of all three MAPK cascades. MAPK activation was localized to renal glomerular and peritubular microvascular endothelial cells and nuclei of proximal and distal convoluted tubules, cardiac myocytes, hepatocytes and splenic lymphocytes. Ricin-induced signalling through MAPKs was associated with the significant upregulation of genes encoding cytokines/chemokines and transcriptional regulators, and the development of thrombocytopenia, haemolytic anaemia and renal failure. Reagents necessary to test the role of the ribotoxic stress response and UPR in pathogenesis in vivo are becoming available, and studies to assess the role of cell stress responses in disease are clearly warranted.