Which begs the question: what does autophagy have to do with patients suffering from pulmonary disease? Interestingly, the answer may be “quite a lot.” The purpose of this review is twofold: to provide a basic introduction to the rapidly expanding field of autophagy research with an emphasis on the most studied autophagic pathway (macroautophagy), and also to review recent studies that link autophagic regulation to pulmonary diseases.

The final pathway, macroautophagy (21), receives the most attention in the literature and its contribution to human disease is the best explored of the three pathways. In fact, macroautophagy is so much better studied that it is often referred to in many papers as simply “autophagy,” even though it is only one of the pathways involved in lysosome-dependent degradation. This is in large part because macroautophagy can be visualized at both the light microscopic level (using fluorescent fusion proteins), and at the EM level making it relatively easy to detect (22). In macroautophagy, a vesicular membrane is constructed around a volume of cytoplasm that is intended for degradation (Figure 1C). This novel structure, called an autophagosome, is distinct from other vesicles on electron micrographs because it contains a double-unit limiting membrane (21). Autophagosomes then deliver their cargo for degradation by fusing with late endosomes and lysosomes, gradually losing their distinctive membrane structure (23). The products of digestion, such as free amino acids, are recycled back to the cytoplasm via lysosomal permeases (8). Of the three known autophagic pathways, macroautophagy is notable for its ability to process large intracellular structures such as organelles (5, 6), invasive bacteria in the cytoplasm (24, 25), and large protein aggregates (10). Macroautophagy is sensitive to nutrient availability (19, 26), and is altered in a variety of nonpulmonary diseases, such as neurodegeneration (27), myopathy (28, 29), and cancer (30). All of the current literature about the role of autophagy in pulmonary disease focuses on macroautophagy, and therefore we focus on this mechanism for the remainder of this review.

The last decade has seen a dramatic increase in our understanding of the molecular mechanisms underlying macroautophagy (Figure 2). The current paradigm is that macroautophagy is mediated by a highly intricate mechanism that requires the contribution of at least 20–30 core proteins that are physically grouped into several distinct multiprotein complexes, each with a distinct functional specialization (31). The most upstream complex described so far is composed of the proteins ATG1 (autophagy-related-1), FIP200 (FAK [focal adhesion kinase] family interacting protein of 200 kD), ATG101, and ATG13 and possesses serine/threonine kinase activity (32, 33). This complex is regulated by mTOR (mammalian target of rapamycin) and AMPK (AMP-activated protein kinase), which sense nutrient and intracellular ATP availability and, through phosphorylation of ATG1, adjust the rate of macroautophagy to the metabolic needs of the cell (34). The primary function of the ATG1-containing complex is to control the activity and localization of another structure called the vps34 multiprotein complex, although the mechanism by which this is accomplished is unknown. The vps34 complex has phosphatidylinositol-3-kinase (PI3K) activity and the production of PI3 moieties is required for autophagosome formation (35, 36). One important constituent of the vps34 complex is the protein ATG6/Beclin-1, which is an adaptor protein that is crucial for the participation of the vps34 complex in autophagosome formation (37, 38). In yeast, the vps34 complex localizes to a distinct perinuclear structure called the phagophore assembly site (39), whereas in mammalian cells this complex localizes to many sites distributed throughout the cell (40). The precise location of these sites is controversial but there are microscopy data suggesting the vps34 complex can nucleate autophagosomes on the surface of both the endoplasmic reticulum (ER) (40) and mitochondria (41), depending on the cell type and experimental conditions. The PI3-phosphate moieties laid down by vps34 recruit binding proteins that are critical for autophagosome maturation and may also intrinsically produce local changes in the membrane physical properties that make it easier to mold (42–44). The result is a cuplike structure that has been termed the “omegasome” (40, 45, 46). The final group of proteins is an ubiquitin-like ligase system composed of the proteins ATG3, ATG7, and ATG10, and a trimeric complex composed of ATG5, ATG12, and ATG16L that localizes to the omegasome (47, 48). The function of this system is to conjugate a protein called ATG8/LC3b to the lipid phosphatidylethanolamine (PE), which is abundant in autophagosome membranes (49). The PE-conjugated form of LC3b (called LC3b-II) is inserted into both sides of the autophagosome membrane (50), and is required for autophagosome membrane elongation (51, 52). LC3b is actually just one of at least five homologous proteins in mammalian cells (53, 54), all of which are targeted to the autophagosome by PE conjugation and cooperate in autophagosome maturation and sealing (52). Luminal-facing LC3b-II is degraded by lysosomal hydrolases (55), while protein on the cytosolic-facing side of the autophagosome is recovered via delipidation (Figure 3) (56). The cytoplasmic, nonlipidated form of LC3b, called LC3b-I, can then be recruited for the construction of new autophagosomes (50, 56). LC3b also plays an important diagnostic role in macroautophagy research: because it is targeted uniquely to autophagosome membranes it serves as a convenient marker for this structure (50).

Microscopy was the first technique used to study macroautophagy and the discovery of autophagosomes on EM images is what gave rise to the field in the first place (58). At the ultrastructural level, early-stage autophagosomes (AV_is) can be distinguished by their double-unit membrane structure and by the consistency of its luminal material, which resembles that of the adjacent cytoplasm (59). However, EM image analysis is less sensitive for picking up late-stage autophagosomes (AV_ds), as they tend to have lost their distinguishing physical characteristics. Fluorescence-based methods get around this problem by tagging specific proteins that are found on both early- and late-stage autophagosomes. The most commonly used reporter is green fluorescent protein–tagged microtubule-associated protein-1 light chain-3 (GFP–LC3) (22). On induction of macroautophagy by nutrient starvation, GFP–LC3 changes its cellular localization from a diffuse cytosolic pattern to a punctate pattern as it is recruited to newly formed autophagosomes (22). This phenomenon can be observed both in mammalian tissue culture cells and in vivo, using GFP–LC3–overexpressing transgenic mice (60). Fluorescence-based reporter systems can be used in fixed cells or observed dynamically by video microscopy (61).

The most widely used assay to examine macroautophagy is Western blot analysis of LC3b protein (62). During the autophagosome maturation process LC3b is conjugated to the lipid phosphatidylethanolamine (PE) and then inserted into the inner and outer leaflets of the autophagosome membrane. Despite its higher molecular weight the PE-conjugated form of LC3b (called LC3b-II) migrates more rapidly in sodium dodecyl sulfate–polyacrylamide gels than the unconjugated version (LC3b-I), and can be easily distinguished on Western blot (Figure 3). Levels of LC3b-II on Western blot were shown to correlate with autophagosome abundance and so this method provides a convenient alternative to microscopy and can also be applied to tissue biopsies (50, 63). Other macroautophagy markers that are amenable to Western blot analysis include p62 (64), which is an adaptor protein that targets ubiquitinated proteins to the autophagosome by interacting with LC3b-II (65). Under starvation conditions, the steady state level of p62 tends to fall as a result of increased consumption of this protein within autophagosomes (64), and so levels of p62 are generally assumed to inversely correlate with macroautophagic activity (or flux).

A common feature of the assays mentioned so far is that they are primarily static estimates of the levels of autophagosomes or macroautophagy proteins at a given point in time. However, some authors have begun to question the assumption that such static measurements reliably equate with increased protein turnover (66, 67), citing examples where increased amounts of autophagosomes and LC3b-II proved instead to represent a decrease in the fusion of autophagosomes with lysosomes and hence a lower rate of flux (68, 69). As such, there has been an increased emphasis on dynamic “activity assays” that directly measure macroautophagic flux.

At present the most popular method for measuring macroautophagic flux is the “LC3b turnover assay,” in which cells are cultured in the presence or absence of an inhibitor of lysosome proteases for a fixed length of time (67). Because LC3b-II is inserted on the luminal surface of autophagosomes as well as the exterior, a portion of LC3b-II is degraded when autophagosomes fuse with lysosomes (Figure 3). By comparing the difference in the amount of LC3b-II on Western blot in the presence versus the absence of lysosome inhibitors, macroautophagic flux can be measured as the amount of LC3b-II degraded per unit time (67). Other groups have produced turnover assays that are based on specific substrates of macroautophagy besides LC3b-II (64, 70–72).

The first indication that macroautophagy was regulated in patients with COPD came from Western blot analysis of lung biopsy specimens, which showed elevated levels of LC3b-II protein in patients with COPD compared with non-COPD control patients (78). The amount of LC3b-II induction correlated positively with clinical severity as measured by GOLD (Global Initiative for Chronic Obstructive Pulmonary Disease) score (78). To further investigate this, our laboratory turned to smoke exposure models in tissue culture cells and in mice, given that cigarette smoke represents the major risk factor for COPD development. Our group as well as others found that exposure of lung epithelial cell lines and fibroblasts to cigarette smoke extract (CSE) induced the accumulation of autophagosomes on electron micrographs and enhanced levels of LC3b-II protein, similar to what was seen in lung biopsies from patients with COPD (79, 80). The appearance of autophagosomes occurred quickly, on the order of a few hours, and preceded cell death, which is seen with prolonged exposure to CSE (79). Importantly, genetic depletion of two macroautophagy pathway members, Beclin-1 and LC3b, reduced the rate of cell death in CSE-exposed cells (78, 79), and mice deficient in LC3b were resistant to emphysematous changes caused by chronic cigarette smoke exposure (81). The overall interpretation was that the macroautophagy pathway facilitated lung epithelial and fibroblast cell death in response to cigarette smoke, which then contributed to the development of emphysema. Given that cigarette smoke exposure stimulated the formation of the death-inducing signaling complex, a complex containing the protein Fas receptor, we concluded that macroautophagy must mediate emphysematous change through programmed cell death (79).

While our group focused on the role of macroautophagy in lung epithelial cells during COPD, others have focused on alveolar macrophages. Macrophages derived from patients with COPD often have impaired antimicrobial activities (82, 83), but at the same time they are prone to hypersecrete proinflammatory cytokines when presented with a stimulant such as bacterial endotoxin (84). The combination of these two functional alterations in macrophages may help to make patients with COPD vulnerable both to airway bacterial colonization and chronic bronchitis (85). Monick and colleagues (86) examined the role of macroautophagy in alveolar macrophages from actively smoking patients (with a greater than 10 pack-year smoking history), or macrophages of nonsmokers who were exposed to CSE. Importantly, their study employed direct measurements of macroautophagic flux and so could directly examine the effect of smoking on protein turnover via this pathway. They found that macroautophagic flux was strongly inhibited in the alveolar macrophages of patients with COPD, and this effect could be reproduced in the macrophages of nonsmokers by exposing them to CSE (86). The level of inhibition was at a downstream step in the pathway: autophagosomes could form but then were not degraded as fast as normal, leading to an overall accumulation of autophagosomes. The inhibition of macroautophagy by smoke exposure coincided with alterations in the energy metabolism of alveolar macrophages: including increased numbers of depolarized, nonfunctional mitochondria, decreased ATP levels, and a greater dependence on anaerobic respiration for ATP production (86). An identical metabolic phenotype can be produced, in the absence of smoke exposure, by inhibiting macroautophagy through genetic or chemical means (87–89). Taken together, the results of Monick and colleagues (86) suggest that a loss of macroautophagic flux is at least partly responsible for the metabolic disturbances seen in the alveolar macrophages of smokers and patients with COPD. Moreover, because mitochondrial damage has been implicated in immune dysfunction in septic patients (90), their results may help to explain the increased susceptibility of patients with COPD to infection.

Macroautophagy inhibition also may help to explain why macrophages from patients with COPD secrete higher amounts of IL-1β compared with healthy individuals. Multiple groups have found that inhibiting macroautophagy leads to excessive secretion of multiple cytokines when cells are challenged with microbe-derived compounds such as endotoxin (91–93). We and others investigated the mechanism by which macroautophagy inhibition leads to increased cytokine production by focusing on caspase-1 activation, which is the rate-limiting step in the secretion of IL-1β and IL-18 in macrophages (88, 89). We found that, in macroautophagy-deficient macrophages, the mitochondrial population was dysfunctional because of a failure in organelle quality control. Mitochondria from macroautophagy-deficient cells produce excessive amounts of reactive oxygen species and are more prone to undergo membrane permeabilization when stressed by endotoxin exposure (89). As a result internal components of mitochondria, such as mitochondrial DNA, leak into the cytoplasm and lead to caspase-1 activation, followed by excessive IL-1β and IL-18 secretion (89). Our overall conclusion was that macroautophagy exerts a moderating effect on at least some aspects of the innate inflammatory response by maintaining mitochondrial health. Given that smoking inhibits macroautophagy and impairs mitochondrial homeostasis, it is tempting to speculate that the inhibition of macroautophagy contributes to the excessive secretion of proinflammatory cytokines observed in some patients with COPD.

The basis of these articles derives from research showing that inactivating cftr mutations lead to exaggerated proinflammatory cytokine secretion in airway epithelial cells when challenged with Pseudomonas aeruginosa (95). To investigate why cystic fibrosis transmembrane conductance regulator (CFTR)–deficient cells are intrinsically prone to airway inflammation, Maiuri and colleagues (96) focused on the most common cftr mutant, Δ508, which folds improperly and is therefore retained in the ER. Expression of Δ508 mutant protein led to ER stress and elevated reactive oxygen species production followed by activation of the enzyme tissue transglutaminase (TG2), an enzyme that can act as a “natural fixative” by forming covalent bridges between lysine and glutamine residues (96). TG2 activation in this context led to the production of protein aggregates that sequestered the antiinflammatory regulatory proteins peroxisome proliferator-activated receptor (PPAR)-γ and IκBα (96, 97). It was proposed that the loss of these proteins explained the proinflammatory phenotype seen in epithelial cells expressing mutant CFTR.

Intracellular protein aggregates are typically cleared by macroautophagy, and the failure to do so is critical for the pathogenesis of important CNS diseases such as Huntington's disease (98). Therefore Luciani and colleagues (99) examined why protein aggregates induced by the Δ508CFTR mutant were not disposed of by macroautophagy. They found that that expressing Δ508CFTR in epithelial cells inhibited macroautophagic flux (99). The level of inhibition was at the level of autophagosome formation and was caused by depletion of the protein Beclin-1. Beclin-1 is a multifunctional adaptor protein that is a critical component of the vps34 complex that is responsible for autophagosome membrane initiation (100). Luciani and colleagues found that Beclin-1 was a substrate for TG2 and, in the context of Δ508CFTR expression, was cross-linked by TG2 and sequestered in protein aggregates similar to PPAR-γ (99). This led to the sequestration of the entire vps34 complex and inhibition of macroautophagy (99). The loss of macroautophagy was significant because this pathway appears to be responsible for degrading CFTR protein that becomes aggregated within the ER (101). As such macroautophagy inhibition leads to a vicious cycle wherein Δ508CFTR protein accumulates unchecked and leads to continual TG2 activity, protein aggregation in the cytoplasm, and excessive inflammation. Importantly, restoring macroautophagy by overexpressing Beclin-1 led to protein aggregate clearance in the cytoplasm and marked reduction in airway inflammation in a mouse model of CF (99). Presumably, restoring macroautophagy somehow led to improved folding of the Δ508 mutant in the ER, because Beclin-1 overexpression allowed some of this mutant protein, which is otherwise functional, to reach the cell surface.

The initial stages of MTb infection occur within alveolar macrophages that encounter inhaled droplets containing the organism. These previously unstimulated macrophages succeed in internalizing MTb bacteria by phagocytosis. However, the phagocytic vesicles containing MTb fail to fuse with lysosomes (102), because of MTb virulence factors that alter the phospholipid structure of the phagosome membrane (103–105). As such the vesicles become protected niches that support unchecked MTb replication, which leads to the eventual lysis of the infected macrophages, often within mediastinal lymph nodes. In immunocompetent hosts this ultimately leads to an inflammatory response directed by helper T cell type 1–polarized CD4⁺ T cells, which contains or eliminates the MTb infection (106). Crucial to the success of the cellular immune response to MTb is the elaboration of IFN-γ, which activates macrophages and enables them to kill MTb bacteria or else sequester them in granulomas (107). Macrophages stimulated by IFN-γ acquire several antimicrobial activities not observed in naive cells, but most notably they are able to overcome the block in fusion between MTb-containing phagocytic vesicles and lysosomes (108). As a result IFN-γ–activated macrophages are able to kill the MTb bacteria they ingest, which contributes to clinical resolution.

Interestingly, the ability of IFN-γ to stimulate the clearance of MTb organisms within macrophages requires an intact macroautophagy pathway. IFN-γ was shown to stimulate macroautophagy and inhibition of this process by chemical or genetic means leads to elevated bacterial MTb titers within macrophages (108, 109). Conversely, activation of macroautophagy by means other than IFN-γ exposure (such as nutrient starvation or rapamycin), similarly leads to bacterial clearance (108, 110). Interestingly, the bulk macroautophagy pathway alone is insufficient to promote MTb clearance. Rather, autophagosomes export specific cytosolic proteins to the MTb-containing vesicle, which is mediated by the adaptor protein p62 (109). Among the proteins that are delivered are ubiquitin and a ribosomal protein (rps30) that, when cleaved by lysosomal hydrolases, are degraded to cationic peptides that are potent MTb antibiotics (109, 111). This novel mechanism of reprocessing cytosolic proteins to generate endogenous antimicrobial peptides may apply to other infections besides MTb.

Furthermore there are genetic data to suggest that macroautophagy contributes to MTb immunity in humans. IFN-γ appears to activate macroautophagy by up-regulating a 47-kD GTPase called IRGM-1 (108, 112–114). IRGM-1 is required for macroautophagy stimulated by both IFN-γ and starvation and appears to work by manipulating the life cycle and redox state of mitochondria (114). Interestingly, polymorphisms in the IRGM-1 gene were linked to increased susceptibility to MTb infection, at least in some cohorts (115, 116). Other polymorphisms in the same gene were also linked to Crohn's disease, another disorder in which macroautophagy is thought to influence pathogenesis (117).