Oxidative stress caused by exposure to cigarette smoke can trigger the ISR. In vitro studies have shown that cigarette smoke extract induces apoptosis in a CHOP-dependent manner that can be antagonized by antioxidants (7, 8). ER stress appears to be a major mediator of the response to cigarette smoke extract, because inhibition of the kinase, PERK, blocks induction of CHOP (7). Consistent with this, heightened levels of ER stress have been observed within smokers’ lungs (9) and patients with chronic obstructive pulmonary disease (COPD) (10). Even a single cigarette can impair oxidative protein folding in murine lungs through impaired function of the ER enzyme, protein disulphide isomerase (PDI) (11).

Recently, an airway gene expression signature associated with COPD was described, and included 98 genes, many of which are targets of ATF4 (12). When ATF4 is overexpressed in epithelial cells in vitro, this COPD gene signature is recapitulated. The observation that many of the COPD-associated changes in gene expression in bronchial biopsies can be reversed by inhaled corticosteroid therapy suggests that treatments might be targeted to this mechanism in future (12).

When mice are exposed to inhaled particulate pollution or cultured human cells are treated with diesel exhaust particles, ER stress triggers the ISR (13, 14). Because the effects of particulate matter can be antagonized by antioxidants, this also likely reflects oxidative stress (13). In fact, the toxicity of particulates may be mediated by the ISR, because CHOP-deficient cells are protected from particulate-induced apoptosis.

It would be wrong to suggest that the ISR is purely toxic during pulmonary oxidative stress or that PERK is the sole kinase involved. For example, when mice are exposed to hyperoxia (95% O₂), phosphorylation of eIF2α does not involve ER stress (5). Instead, PKR is activated by unknown mechanisms to induce ATF4 and CHOP. In this instance, CHOP is protective, because CHOP-deficient mice reared in 95% O₂ develop more severe lung injury, and even modest hyperoxia (80% O₂) causes higher mortality in CHOP-deficient animals (5). The mechanism for this protection is unclear, but CHOP appears to prevent increased epithelial permeability. However, the complex role of CHOP in hyperoxic lung injury requires more study, because, in newborn mice with hyperoxia-induced bronchopulmonary dysplasia, CHOP appears to mediate pathological apoptosis (15).

Point mutations of secreted proteins can cause ER stress and activate PERK. Surfactant protein C (SFTPC) secreted by type II pneumocytes is mutated in some cases of familial interstitial lung disease (16). A splice-site mutation (c.460+1A > G) that causes deletion of SFTPC exon4 results in protein misfolding and familial interstitial pneumonia. The Sftpc-null mouse has a nonlethal phenotype, suggesting that the disease-associated mutations in man may be caused by toxic gain of function. Accordingly, when the Δexon4 mutant is expressed in cells, it causes ER stress and phosphorylation of eIF2α (17). It remains to be tested, however, whether CHOP or GADD34 genotype can alter the toxicity of the SFTPC mutants. Although currently only a correlation, it is intriguing that elevated levels of ATF4 and CHOP protein have been detected in lung homogenates and in type II cells from patients with idiopathic pulmonary fibrosis (18). ISR activation might, therefore, represent a common pathway in interstitial lung disease.

Mutations of the SERPINA1 gene encoding α₁-antitrypsin can cause its accumulation within the ER; however, surprisingly, most of these mutants fail to cause ER stress, but instead enhance a cell’s susceptibility to further insults (19). Most commonly, α₁-antitrypsin deficiency results from homozygosity for the Z allele (Glu342Lys) that destabilizes the protein such that most is degraded, but some polymerizes within the ER, resulting in serum deficiency and early-onset emphysema. Approximately 10% of homozygotes also develop clinically significant liver disease. The increased sensitivity of hepatocytes to ER stress may be involved, as mice made to express the Z variant are far more susceptible to developing hepatic fibrosis after bile duct ligation, which causes ER stress, than are animals expressing wild-type α₁-antitrypsin (20). In this model, Z-expressing mice show an exaggerated ISR with enhanced induction of CHOP (20), and it is known that CHOP expression is essential for cholestasis-induced hepatic fibrosis (21). These effects are dependent on the level of Z α₁-antitrypsin expressed by a cell, because primary bronchial epithelial cells, which secrete low levels of α₁-antitrypsin, fail to achieve concentrations of Z α₁-antitrypsin within their ER to allow polymerization (22). As a result, in contrast to hepatocytes, airway epithelial cells show no augmentation of CHOP or GADD34 expression upon a second hit of mild ER-stress (22). Each tissue must, therefore, be considered individually, because monocytes, another α₁-antitrypsin–expressing cell type, have been reported to display enhanced expression of ATF4 when purified from Z homozygous individuals (23).

The lung is exposed daily to many inhaled infectious micro-organisms that can trigger the ISR. Viral dsRNA can trigger PKR to inhibit protein translation and impedes viral replication (24). Because of the need to recover translation to synthesize inflammatory cytokines, GADD34 is required for an efficient host response, as demonstrated by the increased susceptibility of GADD34-deficient cells and mice to viral infection (25). However, there is redundancy within the system—for example, the coronavirus, infectious bronchitis virus, can activate both PKR and PERK, leading to expression of ATF4 and CHOP (26). Surprisingly, both kinases appear to mediate the cytotoxicity of this infection via CHOP-dependent apoptosis.

Activation of Toll-like receptors (TLRs), specifically TLR3 by dsRNA and TLR4 by LPS, can also trigger phosphorylation of eIF2α (24). The mechanism for this is not entirely clear, but may plausibly reflect increased ER activity, due to the need to secrete cytokines and antimicrobial peptides upon TLR activation. Interestingly, the ISR shows specific modulation by TLR signaling that goes beyond simple activation. In macrophages, when phosphorylation of eIF2α is induced by other stimuli, for example ER-stress, activation of TLR4 by LPS blocks induction of ATF4, CHOP, and GADD34 (27). It is now becoming clear how this can be explained. Normally, when eIF2α is phosphorylated, it binds to and inhibits eIF2B, an enzyme responsible for maintaining eIF2α in its active GTP-bound state. Indeed, binding of phospo-eIF2α to eIF2B is responsible both for the inhibition of translation seen during the ISR and for the up-regulation of ATF4 and CHOP. However, activated TLR4 stimulates eIF2B, thus overcoming the effects of eIF2α phosphorylation (28). It has been proposed that this mechanism may enable cells to selectively inhibit the ISR during chronic ER stress caused by infection, and thus avoid the toxic effects of inducing CHOP and its target, GADD34.

However, the phosphorylation of eIF2α and subsequent activation of the ISR can have negative consequences during chronic infection, and so may represent a therapeutic target. In cystic fibrosis (CF), most causative mutations of CF transmembrane conductance regulator do not cause robust ER stress directly, because these mutations lie within its cytosolic portion (29, 30). However, loss of CF transmembrane conductance regulator function leads to the generation of thickened airway mucus with reduced clearance, which is prone to chronic colonization by bacteria, such as Pseudomonas aeruginosa. This induces local ER stress, and can be recapitulated in normal airway epithelia by application of CF mucus directly (30), and explains why primary CF epithelia recover from ER stress if allowed to grow in vitro in the absence of colonized mucus (29). It has been noted, however, that the ER stress seen in models of CF in response to P. aeruginosa is deficient in PERK–eIF2α signaling (31). This may contribute to the chronic inflammation seen in CF, as pharmacological induction of eIF2α phosphorylation with the drug, salubrinal, was found to lessen the inflammatory response of CF cells. The mechanism by which ER stress fails to activate the ISR in this context has not been studied in detail, but may involve the novel TLR–eIF2B pathway introduced previously here.

Where on-going translation is essential for efficient immune responses—for example, in PKR-activated dendritic cells—a modified ISR is activated in which robust induction of GADD34 prevents significant translational attenuation (32). This and other observations, including those of the TLR–eIF2B pathway, have given rise to the concept of a specific “microbial stress response,” which shares many features of the canonical ISR, but with modifications, such as less robust induction of CHOP (reviewed in Ref. 24). However, the induction of components of the ISR, including CHOP, can differ substantially between models, perhaps because of cell type or the stimuli used, and thus many “flavors” of the ISR may exist. For example, when dendritic cells are challenged either with ER stress or TLR agonists, efficient induction of CHOP is required for secretion of IL-23 (33), which appears to contrast with the response elicited by PKR activation. It should also be noted that viruses and bacteria have evolved numerous mechanisms to escape the antimicrobial effects of the ISR. For example, respiratory syncytial virus can sequester activated PKR and prevent phosphorylation of eIF2α (34).

Being the oldest in evolutionary terms, GCN2 is ubiquitous in eukaryotes. In terms of pulmonary pathology, it has been studied most in the context of tumor biology, as it functions to match amino acid supplies with demand, which can be a limiting factor in cancer growth. The induction of ATF4 during nutrient stress promotes the induction of many amino acid transporters and synthetic enzymes in tumor cells, such as asparagine synthetase (35). As a result, inhibition of the GCN2–ATF4 pathway can reduce proliferation and cell survival (35). This accounts for the up-regulation of GCN2 (both activated and total) in lung cancer samples compared with healthy controls or the surrounding nontumorous cells (35).

In addition to limiting the levels of nutrients required for macromolecule biosynthesis, starvation profoundly affects the metabolism of a tumor. A consequence of this is a heightened level of ER stress that is seen in many hypoxic tumors (36, 37). It is therefore unsurprising that ER stress and the unfolded protein response have been implicated in the pathophysiology of solid tumors, including human lung cancer (8). Evidence of ER stress is associated with more aggressive tumors and resistance to chemotherapy (36). As the mediator of eIF2α phosphorylation during ER stress, PERK has proven necessary for the growth of larger solid tumors (37), and contributes to resistance to therapy (36). It ensures that protein synthesis remains at a level consistent with the supply of nutrients and energy, and, through ISR-mediated activation of autophagy, it also promotes the recycling of nutrients from old and damaged organelles (38). Under such conditions, the recovery of translation mediated by CHOP and GADD34 would be expected to be toxic. Indeed, loss of GADD34 in tumors may prevent apoptosis and promote cell survival in hypoxic conditions (39). Accordingly, GADD34 expression correlates with the degree of differentiation of malignant mesothelioma, with lower levels of expression being seen in the more aggressive sarcomatoid subtype (40). It is therefore tempting to speculate that GADD34 may be a tumor suppressor in ER-stressed environments, but this has yet to be tested formally.