Shiga toxin-producing E. coli O157:H7 (STEC) is a widespread pathogen presenting severe risk to human health, causing haemorrhagic colitis and haemolytic uremic syndrome (Riley et al., 1983; Ko et al., 2016). Annually, the prevalence of acute STEC infection is thought to be ∼2.8 million cases worldwide, with infection progressing to HUS in 3,890 cases and resulting in death in 230 cases (Majowicz et al., 2014).

Virulent strains of STEC have been shown to target the ISR to induce host cell death mediated via a secreted virulence factor termed subtilase cytotoxin (SubAB; Tsutsuki et al., 2016; Figure 2A). SubAB is a secreted toxin consisting of two subunits; the B subunit binds the host extracellular toxin receptor and facilitates toxin internalisation, whereas the A subunit is a serine protease, which, in conjunction with the B subunit, facilitates the intracellular virulent effects of the pathogen (Morinaga et al., 2007). The main target of SubAB is the cleavage of the PERK chaperone binding immunoglobulin protein (BiP; Figure 2B), resulting in the dimerisation and activation of PERK, inducing eIF2α phosphorylation (Tsutsuki et al., 2016; Figure 2C). ISR activation triggered in this manner causes stress granule formation, which is dependent upon death-associated protein 1 activation (Tsutsuki et al., 2016). Inhibition of protein kinase C δ (PKCδ) and phosphoinositide-dependent kinase 1 (PDK1) are implicated in the formation of these stress granules, as chemical inhibition of both also heightens stress granule formation in response to SubAB (Tsutsuki et al., 2016). Furthermore, death-associated protein 1 knockdown increased basal levels of phospho-PDK1(S196), thereby inhibiting stress granule formation, further implicating PKD1 inhibition in formation of stress granules in response to SubAB (Tsutsuki et al., 2016; Figure 2D). Interestingly, in rat intestinal epithelioid cells, PDK1 has been shown to inhibit cell death in response to H₂O₂ (Song et al., 2009). Therefore, the inhibition of PDK1 coupled with prolonged PERK activation (Lin et al., 2009), which is known to promote apoptosis, may push the host cell towards death (Figure 2E; Lin et al., 2009; Tsutsuki et al., 2016). In contrast, in human lung cancer cells, PKCδ activation induces cell death via the CHOP-ATF3 arm of the ISR (Xu et al., 2012). Hence, the exact role PKCδ inhibition by SubAB in cell death requires further attention. Although driving the host cell towards death may seem counterproductive for bacterial survival, it is thought that Shiga-toxic E. coli displays altruism (Loś et al., 2012); in this context, ISR-mediated destruction of host cell and internalised bacteria could provide nutrients to the wider STEC community.

Shigella is a genus of gram-negative, facultative anaerobic bacteria that primarily infect the gastrointestinal tract, causing acute shigellosis (Fernandez and Sansonetti, 2003). Whilst closely related to E. coli, Shigella possesses unique methods of pathogenicity (Ud-Din and Wahid, 2014). Diarrhoea is an early symptom of infection as the bacteria moves through the small intestine, but the primary target of Shigella is the invasion of colonic epithelial cells from the basolateral surface (Phalipon and Sansonetti, 2007). Once internalised, the bacteria replicate and spread from cell to cell. The infection also causes inflammatory colitis via secreted toxins (Eashida et al., 2015). The mechanism of Shigella invasion has been reviewed elsewhere (Carayol and Van Nhieu, 2013; Liu et al., 2019).

Infection with Group B serogroup S. flexneri has been shown to robustly induce the ISR, resulting in the activation of two eIF2α kinases, GCN2, and HRI (Tattoli et al., 2012; Abdel-Nour et al., 2019; Figure 3). During the initial stage of infection, S. flexneri induces AA starvation through membrane damage, which results in the activation of GCN2 (Tattoli et al., 2012; Figure 3A). In its active form, GCN2 inhibits mTORC1 (Figure 3F), as demonstrated via its dispersal from LAMP2, and increases the transcription of the AA stress-related gene asparagine synthetase, a response that increases for up to 4 h post-infection (Tattoli et al., 2012; Abdel-Nour et al., 2019). However, S. flexneri is able to activate mTORC1 via direct delivery of its OspB effector into host’s cellular cytoplasm using the S. flexneri Type III secretion system (T3SS), which interacts with the IQ motif of the GTPase-activating protein 1 (Lu et al., 2015), an upstream regulator of mTORC1 (Tekletsadik et al., 2012; Figure 3I), ultimately resulting in increased host cell proliferation around the infection foci during the later stages of early infection. This increased cellular proliferation reduces S. flexneri spread but is thought to provide a preferential intracellular niche acting as a protective measure (Lu et al., 2015).

In addition, like STEC infection, Shigella infections also result in the aggregation of stalled messenger ribonucleoproteins (mRNPs) into stress granules (Tattoli et al., 2012; Vonaesch et al., 2016; Abdel-Nour et al., 2019; Figure 3E). The activation of the ISR leads to the upregulation of ATF3, ATF4, and GADD34 (Abdel-Nour et al., 2019) and consequently the robust upregulation of the transcription and expression of ISR and inflammatory-related genes (Tattoli et al., 2012; Abdel-Nour et al., 2019). Interestingly, in the presence of ISR-inducing exogenous stresses such as mitochondrial, oxidative and heat shock stress, an increase in frequency and a decrease in an area of stress granules formed are observed in S. flexneri-infected cells at 2–3.5 h post-infection (Vonaesch et al., 2016). The composition of the stress granules is also altered with the selective exclusion of eIF3B, eIF4G, and eIF4B via a mechanism downstream of eIF2α (Vonaesch et al., 2016). Since the phenotype observed is similar to the hindered assembly of stress granules seen following chemical disruption of the tubulin network with nocodazole (Fujimura et al., 2009; Kolobova et al., 2009; Vonaesch et al., 2016) and the movement of eIF3B and eIF4B is controlled by microtubule assembly (Shanina et al., 2001; Harris et al., 2006; Figure 3E), this differential stress granule composition may be dependent upon microtubule dysregulation. Interestingly, stresses such as selenite and hydrogen peroxide, which bypass eIF2α phosphorylation instead inhibiting mTORC1 function, also result in the formation of atypical stress granules lacking components such as eIF3 (Emara et al., 2012; Fujimura et al., 2012), thereby also implicating the S. flexneri infection-induced inhibition of mTORC1 in the differential stress granule formation (Tattoli et al., 2012; Vonaesch et al., 2016). As the formation of stress granules during S. flexneri infection has only been investigated up to 5 h (Abdel-Nour et al., 2019), whether the formation of stress granules is altered similarly during later-stage infection when mTORC1 is reactivated remains unknown. If this does not occur later when mTORC1 is reactivated, this would aid the hypothesis that modulation is at least partially dependent on mTORC1 inhibition. Furthermore, whether this modulation to stress granule formation during infection may provide any evolutionary benefit to S. flexneri or is simply a downstream effect of ISR modulation remains unknown.

In addition to ISR activation via membrane damage, intracellular sensing of bacterial peptidoglycan by the host PRR, nucleotide-binding oligomerisation domain-containing protein 1 (NOD1), also induces ISR activation and expression of the pro-inflammatory cytokine nuclear factor kappa-light chain-enhancer of activated B cells (NF-κB) in an HRI-dependent manner (Abdel-Nour et al., 2019; Figure 3J). NOD1 activation results in dissociation of the HRI chaperone, heat shock protein beta-r 8 (HSPB8), from HRI, consequently activating HRI (Abdel-Nour et al., 2019; Figure 3K). Activated HRI induces robust eIF2α phosphorylation (Figure 3C), resulting in heightened HSPB8 transcription in a manner dependent on ATF4 and ATF3 signalling (Abdel-Nour et al., 2019; Figure 3L). This nascent HSPB8 can interact with the previously dissociated HSPB8 and NOD1 to form a signalosome (Figure 3M) and during S. flexneri infection causes the upregulation of host immune inflammatory responses and macrophage activation through the NF-κB pathway (Abdel-Nour et al., 2019; Figure 3N). As this pathway is also triggered by misfolded proteins within the cytosol and is comparable with the UPR in the ER, it was coined the cytosolic UPR (cUPR; Abdel-Nour et al., 2019).

Shigella flexneri infection results in the induction of the ISR, which can be viewed as a protective response activating pro-inflammatory responses via the cUPR (Abdel-Nour et al., 2019) and autophagy via mTORC1 inhibition (Tattoli et al., 2012), during early infection. However, Abdel-Nour et al. (2019) found that the eIF2α S51A mutant, which cannot be phosphorylated, results in a significantly increased frequency of intracellular S. flexneri compared with cells with phosphorylated eIF2α (Abdel-Nour et al., 2019). Taken together, these data indicate that intracellular S. flexneri replication is heightened during ISR activation, and it is plausible that the phosphorylation state of eIF2α may at least partially control bacterial spread and viability.

However, during later-stage infection, S. flexneri-mediated reactivation of mTORC1 not only increases host cell viability but also decreases bacterial spread around the infection foci (Lu et al., 2015). Whilst the impact of mTORC1 reactivation on bacterial infection has not been investigated, focussing on the ISR may provide further insights into this mechanism. Furthermore, infection results in GADD34 expression; however, whether this occurs during later-stage infection, when mTORC1 is reactivated, is as of yet unknown (Abdel-Nour et al., 2019). As GADD34 induces the dephosphorylation of eIF2α via the activation of PP1 (Connor et al., 2001; Novoa et al., 2001), the potential of sustained GADD34 expression during later stage of S. flexneri infection (Abdel-Nour et al., 2019) may lead to termination of the ISR, potentially aiding the cell in returning to homeostatic conditions.

There is also evidence that inhibition of mTORC1 leads to eIF2α phosphorylation in cancer cell lines (Harvey et al., 2019). If this also occurs during S. flexneri infection, the reactivation of mTORC1 may further push the cellular equilibrium of eIF2α towards the non-phosphorylated form. Therefore, if dephosphorylation of mTORC1 was coupled with the GADD34 expression, this could act as a two-part shift to favour a state with minimal eIF2α phosphorylation. Though counter-intuitive, whilst defective eIF2α signalling favours S. flexneri invasion (Abdel-Nour et al., 2019), it may lead to increased host viability, which has been suggested to benefit S. flexneri, the latter remaining in infected cells for much of its life cycle (Killackey et al., 2016). Thus, whether mTORC1 reactivation or persistent GADD34 expression leads to eIF2α dephosphorylation during infection, and the consequential effects upon host cell viability and S. flexneri persistence and replication, is an area requiring further attention. Given that increased host cellular replication during OspB-mediated mTORC1 reactivation is thought to provide a preferential niche for S. flexneri survival (Lu et al., 2015), it is entirely plausible that this potential ISR termination may feed into this, helping to create an even further immune-privileged environment for S. flexneri. Furthermore, the effect of OspB only occurs later during the infection (Lu et al., 2015), whereas the initial phenotype of GCN2 activation and mTOR inhibition require internalisation and the resulting membrane damage (Tattoli et al., 2012; Abdel-Nour et al., 2019), which is intriguing as internalisation of OspB only requires extracellular contact between the bacteria and host cell (Hueck, 1998). Therefore, elucidation of the interaction between S. flexneri and these pathways may provide valuable insights into the pathogenic mechanisms of S. flexneri in chronic infections.

Salmonella enterica serovar Typhimurium is an enteric pathogen primarily associated with food-borne gastrointestinal disease (Fabrega and Vila, 2013) and is thought to affect 1.3 billion people annually, leading to approximately 3 million deaths globally (Pui et al., 2011). During infection, Salmonella typhimurium adheres to the host’s intestinal epithelium, resulting in extensive cytoskeletal rearrangements (Finlay et al., 1991). These modifications cause membrane ruffles, which eventually engulf the bacteria in large vesicles known as Salmonella-containing vesicles, creating an intracellular compartment in which Salmonella can survive and replicate (Steele-Mortimer, 2008). Salmonella infection is detected by host TLRs and NOD proteins, which initiates the NF-κB signalling cascade and results in cytokine and chemokine production, leading to an inflammatory state (Souvannavong et al., 2007; Spiller et al., 2008; Winter et al., 2009; Keestra et al., 2011).

As with S. flexneri, membrane damage induced by Salmonella invasion causes intracellular AA starvation with the activation of ISR through the eIF2α kinases GCN2 (Tattoli et al., 2012; Figures 4A,B). This AA starvation-induced activation of GCN2 also initially inhibits the activity of mTORC1, via the dispersal from LAMP2 and results in the activation of autophagy (Tattoli et al., 2012; Figures 4C,D). Interestingly, within 4 h of infection, the raptor/rag/regulator pathways can reactivate mTORC1, effectively saving Salmonella from autophagy; however, the mechanism by which this occurs has yet to be fully determined (Tattoli et al., 2012; Figure 4E).

In later stages of infection, e.g., 12–24 h post-infection, Salmonella has also been shown to induce ER stress (Figure 4F), robustly activating the UPR and leading to the activation of X-box binding protein 1 and transcription factor 6 (Antoniou et al., 2018; Figure 4G), which is known to increase lipid biogenesis and increase ER expansion (Sriburi et al., 2004, 2007; Figure 4H). During Salmonella-induced ER stress, human leukocyte antigen (HLA)-B27 becomes misfolded, causing SCV to move away from host golgi apparatus (Antoniou et al., 2018). This, coupled with ER expansion, is thought to allow for increased space for the SCV and is supported by observations that ER stress induction by thapsigargin and misfolded HLA-B27 increase intracellular Salmonella viability and replication (Antoniou et al., 2018; Figure 4I). Thus, Salmonella effectively utilises the ISR and UPR in two opposing ways, firstly by reversing the autophagic responses brought on by the ISR during early-stage infection via mTORC1 reactivation and then inducing ER stress to allow for preferential replication conditions in later stages of infection.

Pseudomonas aeruginosa, a gram-negative, rod-shaped, mono-flagellated bacterium, is one of the most frequent causative agents for hospital-acquired infections resulting in loss of life (Buhl et al., 2015), with immunocompromised patient’s survival rates being disproportionately lowered (Migiyama et al., 2013). Chronic lung infections caused by P. aeruginosa are a common cause of death in patients with cystic fibrosis and chronic obstructive pulmonary disease, with those affected often experiencing recurrent infections (Murphy et al., 2008; Yum et al., 2014).

During infection, P. aeruginosa secretes a wide variety of proteins including the extracellular adhesin CdrA (Borlee et al., 2010), the diffusible quorum-sensing molecule N-(3-oxododecanoyl)-homoserine lactone (HSL; Smith et al., 2002), and virulence factors [e.g., alkaline protease A (ArpA; Vasil and Ochsner, 1999) and HasAp (Létoffé et al., 1996)], all of which are known to induce ER stress (Grabiner et al., 2014; van‘t Wout et al., 2015; Figure 5A). In mouse embryonic fibroblasts (MEFs), HSL, which has a key role in P. aeruginosa cell-to-cell communication within the structurally ordered biofilm (Smith et al., 2002), induces ER stress via the release of Ca²⁺ from ER stores (Figure 5B). This causes an imbalance in ER homeostasis and activates PERK (Figure 5C), which phosphorylates eIF2α and results in a global shutdown of protein synthesis (Grabiner et al., 2014; Figures 5D,E). Interestingly, this translational stalling reduces the expression and secretion of the pro-inflammatory chemokine keratinocyte chemoattractant (KC), the mouse equivalent of interleukin 8 (IL-8; Grabiner et al., 2014; Figure 5F). Thus, HSL may lead to the suppression of KC secretion through eliciting the host ISR, aiding P. aeruginosa to evade host inflammatory and antibacterial responses during the early stages of infection (Grabiner et al., 2014). This observation contrasts with the robust expression of IL-8 seen in S. flexneri infection (Abdel-Nour et al., 2019) and suggests a species-specific response (Abdel-Nour et al., 2019). However, further studies are required to probe this and ascertain whether this response is species specific or cell dependent due to the differential approaches used with one study using mouse-derived MEFs (Grabiner et al., 2014) and the other using the human cervical epithelial cell line HeLa (Abdel-Nour et al., 2019).

Another example of P. aeruginosa-mediated ISR manipulation is through the secretion of ArpA (Figure 5G), a protease involved in, amongst other pathways, hosts siderophore-mediated iron scavenging (Vasil and Ochsner, 1999; Kim et al., 2006; van‘t Wout et al., 2015). Whereas HSL induces ER stress through activation of the p53 MAPK pathway, ArpA specifically activates HRI (Figure 5H), which induces the expression of GADD34 (van‘t Wout et al., 2015; Figure 5I) and is protective against P. aeruginosa cytotoxicity, allowing for prolonged host cell survival (van‘t Wout et al., 2015; Figure 5J). The mechanism by which HRI activation and GADD34 expression increase host viability is currently unknown, but increased GADD34 expression could lead to increased PP1 activity and consequently dephosphorylation of eIF2α (Connor et al., 2001; Novoa et al., 2001). This deactivation of the ISR could prove to be a promising system for P. aeruginosa to push the cell towards survival, thereby increasing viability. However, given the recent findings of Abdel-Nour et al. (2019), it is entirely plausible that activation of HRI by ArpA may activate the cUPR, which has been shown to be protective against S. flexneri infection. Whether this increased viability is due to eIF2α dephosphorylation or cUPR activation, or a combination of both, remains to be elucidated and requires further attention.

Pseudomonas aeruginosa hereby displays a two-part manipulation of the host ISR, both dampening inflammatory responses and increasing host cell viability. These reduced inflammatory and immune responses may act as the critical tipping point, leading to the decreased survival of immunocompromised patients (who already have impaired immune responses), as it could result in unregulated and therefore heightened P. aeruginosa growth.

Porphyromonas gingivalis is a gram-negative, anaerobic bacterium and the “keystone pathogen” of the chronic oral inflammatory gum disease, periodontitis (Socransky et al., 1998). Infection triggers host immune responses resulting in inflammation of the gingival tissues, which in some cases progresses to periodontitis, resulting in alveolar bone resorption and ultimately tooth loss (Pihlstrom et al., 2005). P. gingivalis is known to modulate several host cell responses including the inhibition of antimicrobial responses whilst leaving pro-inflammatory signalling active, thereby providing nutrients from inflammatory spoils (Hajishengallis and Lambris, 2012). To achieve this, P. gingivalis employs a range of virulence factors including lipopolysaccharides (LPSs), fimbriae and lysine- and arginine-specific cysteine proteases, termed gingipains (Jia et al., 2019). Gingipains are cell surface-anchored proteins (Andrian et al., 2006), which can also be excreted in membrane-bound vesicles (Grenier et al., 1989) and therefore can account for up to 85% of proteolytic activity around the P. gingivalis infection site (De Diego et al., 2013).

A recent study using human umbilical vein cells as host cells suggested that the virulence of P. gingivalis (strain 381) may involve the UPR and ISR (Hirasawa and Kurita-Ochiai, 2018; Figure 6). In this study, the authors showed that whilst infection ultimately resulted in apoptosis after 21-h infection, early-stage infection (∼8 h) resulted in ER stress characterised by increased expression of CHOP and BiP at both the transcriptional and translational levels coupled with increased caspase-12 activity (Hirasawa and Kurita-Ochiai, 2018; Figures 6A–C). In addition, enhanced autophagy, characterised by the increased expression of autophagy markers Beclin-1, microtubule-associated protein 1A/1B-light chain 3, and acidic vesicular organelles (Figure 6D), was also observed. This response was inhibited by pretreatment with an ER stress inhibitor salubrinal, an inhibitor of PP1, that results in blockage of eIF2α dephosphorylation (Boyce et al., 2005). Furthermore, siRNA knockdown of LC3 resulted in increased apoptosis, thereby implicating ER stress-induced autophagy as a protective response against P. gingivalis-induced apoptosis (Hirasawa and Kurita-Ochiai, 2018). These results are corroborated by studies in mice where administration of P. gingivalis induced ER stress with increased expression of both CHOP and BiP (Yamada et al., 2015).

Interestingly, the lysine-specific gingipain of P. gingivalis has been shown to degrade mTORC1 and modulate levels of mTORC1-associated proteins in oral epithelial cells after 4 h of infection (Stafford et al., 2013; Figures 6E,F). However, this mTOR degradation requires P. gingivalis internalisation, indicating that these effects are probably not mediated by the secretory fraction of gingipains produced by extracellular P. gingivalis (Stafford et al., 2013). Inactivation of mTOR is known to induce autophagy (Jung et al., 2010), fitting with the early stage autophagy seen by Hirasawa and Kurita-Ochiai (2018; Figure 6G). Furthermore, mTOR inhibition by rapamycin suppresses tunicamycin-induced ER stress, resulting in autophagy (Dong et al., 2015). Therefore, gingipain-mediated degradation of mTOR may help dampen ER stress induced by P. gingivalis infection, aiding host cell survival in the early stages of infection by delaying the onset of apoptosis.