Autophagy in the liver


  • Potential conflict of interest: Nothing to report.


A great part of our current understanding of mammalian macroautophagy is derived from studies of the liver. The term “autophagy” was introduced by Christian de Duve in part based on ultrastructural changes in rat liver following glucagon injection. Subsequent morphological, biochemical, and kinetics studies of autophagy in the liver defined the basic process of autophagosome formation, maturation, and degradation and the regulation of autophagy by hormones, phosphoinositide 3-kinases, and mammalian target of rapamycin. It is now clear that macroautophagy in the liver is important for the balance of energy and nutrients for basic cell functions, the removal of misfolded proteins resulting from genetic mutations or pathophysiological stimulations, and the turnover of major subcellular organelles such as mitochondria, endoplasmic reticulum, and peroxisomes under both normal and pathophysiological conditions. Disturbance of autophagy function in the liver could thus have a major impact on liver physiology and liver disease. (HEPATOLOGY 2008.)

The Basics of Macroautophagy

Two major degradation systems exist in eukaryotic cells: the proteasome and the lysosome. They differ in their functional significance and the type of substrates they take in for degradation. In the lysosome system, degradation of extracellular materials is mediated by endocytosis (heterophagy), whereas degradation of intracellular components is mediated by 3 types of autophagy: macroautophagy, microautophagy, and chaperone-mediated autophagy,1–6 which differ in how the cytoplasmic materials are delivered to the lysosome (Fig. 1). In macroautophagy, content is sequestered in a double-membrane structure called an autophagosome (Figs. 1 and 2), which subsequently fuses with the lysosome.

Figure 1.

Three types of autophagy. (A) Macroautophagy begins with a phagophore or isolation membrane, the origin of which is unclear. The phagophore undergoes 3-dimensional elongation driven by Atg genes while engulfing cytosolic components. The membranes are joined to form autophagosomes, which are delimited by a double- or multilayered membrane. Eventually, the autophagosome fuses with the lysosome to form autophagolysosomes, which are single-membrane structures, and the contents are degraded by lysosomal hydrolases. Small molecules such as amino acids and glucose could be transported out of the lysosome through the membrane permease for cellular reuse. (B) In microautophagy the lysosome can directly take in cytosolic components including cytosol, organelles, and even a piece of nuclear material through membrane invagination. This leads to the formation of single-membrane intralumenal vesicles that contain the engulfed contents. The contents are then degraded by lysosomal hydrolases. (C) In chaperone-mediated autophagy, the substrates to be degraded contain a KFERQ motif. They are associated with Hsc70 and its cochaperones including Hip, Hop, Bag-1, Hsp40, and Hsp90. The complex binds to the multiunit LAMP-2a on the lysosomal membrane. The substrate protein, but not the chaperones, is transported into the lysosome through the putative translocon formed by the LAMP-2a transmembrane domains. An intralysosomal hsc70 (lyso-hsc70) is also required for substrate translocation. The substrate protein is then degraded by lysosomal proteases. Chaperone-mediated autophagy differs from the other 2 types of autophagy in the lack of vesicle formation, the type of substrates, and the absence of organelle degradation.

Figure 2.

Morphological and molecular features of autophagosomes. (A-B) Primary mouse hepatocytes were subjected to starvation in Earle's Balanced Salt Solution (EBSS) for 12 hours. Cells were then fixed and examined by electron microscopy. A low-magnification image shows the presence of multiple autophagic vacuoles at different maturation stages in the cytosol (arrows, A). The double membrane structure and the engulfed mitochondrion and other cellular contents of one autophagosome (arrow) are shown in an enlarged image (arrow, B); m, mitochondria; N, nucleus. (C-D) Mouse hepatocytes were infected with an adenovirus expressing GFP-LC3, which demonstrated a uniform cytoplasmic distribution in the presence of Williams' Medium E (C). Following culture in EBSS for 12 hours, GFP-LC3 translocated to autophagosomal structures, exhibiting a punctated pattern (D). GFP-LC3 translocation can be quantified and is a useful marker for the turnover of macroautophagy.

The term “autophagy” comes from Greek, meaning self-eating. Autophagy as a biological phenomenon was first systemically characterized by de Duve and Wattiaux 40 years ago,1 although it seems that as a process of bulk segregation of cellular constituents, the phenomenon was first reported in 1957 in mammalian cells.7 Much of the pioneering work in the autophagy field is conducted in the liver or using isolated hepatocytes, such as determination of the morphology and membrane compositions of autophagosomes and characterization of the kinetics of protein degradation, regulation by hormones and amino acids, and involvement of the mammalian target of rapamycin (mTOR) pathway in induction.2, 8–12

It is now clear that macroautophagy is important for many physiological and pathological processes.4, 6, 13 It is required for normal development and participates in the clearance of apoptotic cells during embryogenesis. In adults, macroautophagy seems to be involved in extending the life span and in protecting cells from the stress response, such as starvation. Autophagic degradation of cellular constituents can efficiently recycle essential nutrients to sustain basic biological processes. Macroautophagy is also used as a defense mechanism to clear intracellular microbes, misfolded proteins, and damaged organelles.

Whereas autophagy has been mainly studied in mammalian cells, significant breakthroughs in the understanding of its molecular machinery come from studying it in the yeast system. Currently, 31 autophagy-related genes (ATGs) have been identified since the first gene, Atg1, was discovered from a genetic screening in yeast.14 A large portion of the molecular machinery of autophagy is conserved in yeast, Caenorhabditis elegans, Drosophila, and mammals.15

The core machinery seems to be built around 2 ubiquitin-like conjugation systems (Fig. 3).3 In one system, the ubiquitin-like protein Atg12 was first activated by Atg7, an ubiquitin-activating enzyme (E1)–like protein, and then transferred by Atg10, an ubiquitin carrier protein (E2)–like protein, to Atg5 through a covalent bond. The Atg5-Atg12 complex interacts with Atg16 to form a multimer complex, which is localized to membranes of early autophagosomes. It seems that the assembly of this system is independent of autophagy activation. Thus, the complex appears to provide the necessary platform for autophagy activation.

Figure 3.

Basic molecular machinery and signal transduction pathway of mammalian macroautophagy. Macroautophagy in mammalian cells can be activated by various stimuli, some of which are physiological and some pathological. The major known signaling pathway relates to the PI-3 kinase and mTOR, but many stimuli are not known for how they activate autophagy. At the molecular level, the signals may activate the Beclin 1/VPS34 complex, which in turn regulates the targeting of Atg5-Atg12 and Atg8/LC3 to the phagophore. Atg5-Atg12 targets the autophagosomes first but then recycles to the cytosol once LC3/Atg8-PE is translocated to the membranes and the isolation membranes are closed. Matured autophagosomes are eventually fused with the lysosomes to form autophagolysosomes. LC3 remains on the membranes throughout the process and thus can be used as a reliable marker for autophagosomes. Chemicals known to be able to modulate autophagy are highlighted as well. For more details, see the text.

In another system, a ubiquitin-like protein, Atg8, or one of its mammalian orthologs, microtubule-associated protein 1 light chain 3 (LC3), is first cleaved by Atg4 to expose the conserved Gly120 at its C-terminus. Atg8/LC3 is then conjugated to phosphatidylethanolamine (PE), also via Atg7, and Atg3, another ubiquitin carrier protein (E2)–like protein.3, 16 The unconjugated form of Atg8/LC3 (called LC3-I) is in the cytosol, whereas the conjugated form (called LC3-II) targets to the autophagosomal membrane16 following the Atg5-Atg12-Atg16 complex (Fig. 2). This association of Atg8/LC3-PE to autophagosomes is considered important for the membrane extension of the autophagosome and the eventual closure of the membrane to form vesicles. The Atg12-Atg5 complex detaches from mature autophagosomes, whereas Atg8/LC3 stays on the membrane until it is degraded by the lysosome. Atg8/LC3 is thus widely used as a marker for monitoring the autophagy process.

Several other key autophagy genes are important to the initiation and regulation of autophagy (Fig. 3). Atg6 and its mammalian homologue, Beclin 1, are particularly important. Atg6/Beclin 1 forms a complex with VPS34, VPS15, and Atg14. VPS34 is a class III phosphoinositide-3 kinase (PI-3 kinase) that is required for autophagy, and its activity is regulated by the complex. Although so far Atg14 has been identified only in yeast, Bcl-217 and Ambra-118 are only found in mammalian cells that can also interact with Beclin 1. However, Bcl-2 suppresses but Ambra-1 promotes Beclin 1/VPS34 activity and therefore autophagy. Although the exact mechanism is not clear, the recent finding that Beclin 1 contains a BH3-only domain19, 20 suggests that Bcl-2 may sequester Beclin 1 from its interaction with VPS34. Finally, UVRAG stimulates autophagy by interaction with Beclin 1.21 Another recently found molecule, Bif-1, promotes autophagy by interacting with UVRAG.22 Thus, there are extensive molecular interactions in the early stage of the formation of autophagosomes.


AT, α1-antitrypsin; ATZ, α1-antitrypsin z mutant; ER, endoplasmic reticulum; mTOR, mammalian target of rapamycin.

Role of Autophagy in Nutrient and Energy Metabolism in Hepatocytes

Induction of Autophagy by Starvation and Regulation by Amino Acids and Hormones.

The best-known autophagy inducer is food restriction or starvation. Under this condition, autophagy is activated to provide cells with the necessary nutrients through degradation of intracellular materials. In animals, starvation induces the largest protein loss in the liver. In the first 48 hours, mice and rats can lose 25%-40% of their liver protein.23 The loss seems to be mainly in the cytosol with most subcellular constituents except nuclei and DNA, resulting in a 25% reduction in cell volume in the first 24 hours of starvation.12 The involvement of macroautophagy and the lysosome was realized when protein loss was found to be associated with stimuli that cause autophagy, such as glucagon administration or deprivation of amino acids, insulin, or serum in the perfused rat liver and heart and in cultured cells.12 Amino acid deprivation has since been employed as a standard simulation for autophagic degradation of proteins in mammalian cells. In yeast, nitrogen deprivation is an equivalent stimulus for autophagy.24

Mortimore and Poso were largely responsible for the careful determination of the kinetics of protein degradation in the liver and the concept that long-term protein degradation is mediated by autophagy.12 For example, they found that the process of autophagy stimulated by amino acid deprivation in the liver was rapid, with the autophagosome having a half-life of only 6-9 minutes.12 The turnover rate of the autophagosome is well correlated with the proteolysis rate at different amino acid levels. Many factors could suppress autophagosome formation, including amino acids, insulin, cycloheximide, and 3-methyladenine (3-MA). Together these data have established that macroautophagy can account for the accelerated proteolysis in the liver on amino acid deprivation.

About 8 so-called regulatory amino acids (Leu, Tyr, Phe, Gln, Pro, Met, Trp, and His) could suppress macroautophagy in perfused rat liver at 0.5× and 4× the normal plasma concentration, among which leucine is by far the most effective.12 Leucine is also the only amino acid that can inhibit protein degradation in myocytes and adipocytes. Alanine has a coregulatory effect, in that it could work with regulatory amino acids at 1× the normal plasma concentration to achieve the maximal inhibitory effect but has no impact on proteolysis by itself. The coregulatory effect of alanine could also be mimicked by insulin.

Glucagon was perhaps the first hormone known to simulate autophagy. Ashford and Porter reported “autophagy-like” structures in glucagon-perfused rat liver in 1962.8 Soon it was confirmed that glucagon can stimulate autophagy in the liver.9, 25 The glucagon receptor is G-protein-coupled. The effector of the coupled Gαs is adenylyl cyclase. Thus, cAMP and the downstream protein kinase A mediate the effects of glucagon. Consistently, β-agonists such as epinephrine, which also activates adenylyl cyclase, and cAMP itself have effects similar to glucagon on autophagy induction. Interestingly, their autophagic effect is liver specific. In the muscle they instead inhibit autophagy and protein degradation.12

Glucagon-induced autophagic proteolysis in the liver does not seem to be related to glycogenolysis, which seems to be different from its role in glycogen autophagy (see below). Consistently, neither glucose nor fatty acids have a regulatory role in autophagic protein degradation in the liver.12 It is not clear how glucagon stimulates autophagy in the liver, but it could be related to the regulatory pathway of amino acids. Thus, it could reverse the inhibitory effect of regulatory amino acids at 0.5× the normal plasma concentration to induce autophagy, and its ability to induce autophagy reaches the maximal level at the normal amino acid concentration. However, its autophagic effect could be blocked by amino acids at twice the normal plasma level.26

Very early on, insulin was found to have an inhibitory effect on autophagy in the liver.27, 28 Testing the effect of insulin on autophagy was initially based on the desire to search for antagonists of glucagon and on the observation that insulin could correct the abnormally elevated autophagy level in livers of severely diabetic rats. Interestingly, the inhibitory activity of insulin is universal and can be also observed in cardiac and skeletal muscles, kidney, adipose tissue, and cultured cells despite that the maximal effect varies and the degradation kinetics differ in these tissues.12 The inhibitory effect is likely related to the property of insulin as a growth factor and is mediated by the class I PI-3 kinase and PKB/Akt, which results in activation of mTOR (Fig. 3).29 Suppression of autophagy by class I PI-3 kinases had been demonstrated earlier in a colon cancer cell line.30, 31 In isolated rat hepatocytes, insulin can stimulate mTOP-dependent p70 S6 kinase, leading to increased S6 phosphorylation.29 Rapamycin can block insulin-induced S6 phosphorylation and reverse insulin-induced autophagy suppression, suggesting that insulin's activity can be mediated by mTOR.29 The importance of TOR in suppressing autophagy has been shown in yeast, in which suppression of TOR with rapamycin can induce autophagy even in nutrient-replete conditions.32

The mechanisms by which the regulatory amino acids suppress autophagy seem to be more complicated and have not been worked out. In isolated rat hepatocytes, amino acids are potent activators of p70 S6 kinase via mTOR.29, 33 However, unlike insulin, rapamycin was shown to have no29 or only a partial33 effect on the antiautophagy activity of amino acids despite being able to block amino acid–induced S6 phosphorylation. It thus seems that amino acids may have distinct functions in meeting nutrient demands, which are mediated by mTOR to promote protein synthesis, and in suppressing the catabolic autophagy pathway, which is likely to be independent of mTOR, at least in hepatocytes.

How could amino acids fulfill these 2 functions at the same time? One clue may come from the finding that activation of mTOR by amino acids is not through the traditional class I PI-3 kinase pathway and is thus independent of PKB/Atk, TSC1/TSC2, and RheB (Fig. 3).34 Rather, it is mediated by the class III PI-3 kinase, VPS34.34 But VPS34, together with its regulatory component, VPS15, is also known to be involved in autophagy induction in both yeast and mammalian cells. The substance 3-methyladenine was the first known suppressor of autophagic proteolysis as demonstrated in isolated hepatocytes35 and was subsequently shown to be a PI-3 kinase inhibitor.36 The importance of the class III PI-3 kinase (VPS34), but not the class I PI-3 kinase, which actually has opposite effects, in promoting autophagy was then demonstrated in a colon cancer cell line30 and in yeast.37 In yeast, VPS34 forms 2 different complexes, with complex I (containing Atg6/Beclin 1 and Atg14) functioning in the autophagy pathway. Thus, it is possible that amino acids may direct VPS34 from the catabolic autophagy pathway to the synthetic mTOR pathway, resulting in suppression of the former and activation of the latter. Neither the mechanism by which amino acids activate VPS34 nor how VPS34 subsequently turns on mTOR is known at the present time. But revealing these mechanisms would likely lead to the resolution of how VPS34 may possess opposite functions.

Amino acids could employ other mechanisms to suppress autophagy. Recently it has been found that both insulin and amino acids, such as phenylalanine, could also work through a pathway related to Gαi3 because deletion of Gαi3 eliminates their protective effects against protein degradation in perfused liver.38 Gαi3 is a Gα subunit of heterotrimeric G proteins, and its positive role (in guanosine diphosphate [GDP]-binding status) in autophagy induction was shown previously in a colon cancer cell line.39 Consistently, an early study implicated the involvement of guanosine triphosphatase (GTPase) activity in autophagy induction in permeabilized hepatocytes.40 How insulin and amino acids affect Gαi3 and how Gαi3 in turn affects autophagy are not known. In other types of cells, it has been found that amino acids may modulate activation of Raf-1 kinase and therefore reduce extracellular signal-regulated kinase 1/2 (ERK1/2) activation, which in turn reduces the phosphorylation and activity of GAIP.41, 42 GAIP is a GTPase-activating protein (GAP) for Gαi3 and elevates its GTPase activity. As such, GAIP promotes the GDP status of Gαi3 and is proautophagic. There is also some evidence suggesting that the Gαi3-mediated mechanism may interface with the VPS34-mediated mechanism, because activation of the latter could rescue autophagy caused by the defect in the former.30 It will be interesting to determine how these events are intertwined.

Effect of Energy Level and AMPK in Autophagy Induction in Hepatocytes.

Early studies have shown that the autophagy process is ATP dependent during both the sequestration and postsequestration steps.5 Thus, depletion of ATP will hamper autophagy. However, direct suppression of energy supply may have complicated outcomes. An obvious target of such manipulations is adenosine monophosphate-activated protein kinase (AMPK), which is activated by a high AMP/ATP ratio and in turn switches off ATP-dependent processes. An initial study found that an analog of AMP, AICAR, suppressed autophagic sequestration of lactose dehydrogenase in isolated hepatocytes.43 Although this suppressive effect of AMPK indicated by the use of AICAR as well as metformin, another indirect AMPK activator, could be confirmed in both hepatocytes and nonhepatocytes, complementary approaches employing an AMPK inhibitor, compound C, or a dominant-negative AMPK mutant instead indicate that AMPK is required for autophagy induced by amino acid deprivation.44 This latter finding that AMPK is a positive regulator of autophagy is supported by other studies in yeast45 and in mammalian cells46, 47 under various conditions, including ischemia. In general, these latest observations are consistent with the dogma that AMPK can suppress TOR and TOR suppresses autophagy (Fig. 3).

Thus, the suppressive effects of AICAR and metformin on nutrient deprivation-induced autophagy in hepatocytes may not be related to their effects on AMPK.44 Indeed, the AMPK-independent effects of AICAR have been observed in other cases.48, 49 Furthermore, this suppressive effect may be cell-type specific because in both murine embryonic fibroblasts and HCT116 colon cancer cells, AICAR and metformin are able to induce autophagy in a p53-dependent manner.50

Glycogen Autophagy and Neonatal Survival.

Autophagy activity is essential for the survival of newborn animals. Within 1 day of birth, genetic deletion of Atg5 or Atg7 causes the death of neonatal mice, which can be rescued by forced milk feeding.51, 52 In the mouse, based on changes in the subcellular localization of Atg8/LC3, it could be determined that autophagy induction increases in the first half hour after birth, peaks at 3-6 hours, and returns to baseline level by 2.5 days after birth in the liver, heart, diaphragm, gastrocnemius muscle, lung, and pancreas.51 Autophagy seems to be required for providing the necessary nutrients from the stored cellular supply during this critical period as the newborn adapts to the nutrients from the milk.

Although degradation of cellular proteins by autophagy may be required to maintain plasma amino acids level in the newborn, which is lower in Atg5- or Atg7-deficient mice,51, 52 amino acids may not be the major nutrient required for survival. What is required most immediately after birth is probably glucose for such energy-demanding organs as the heart and diaphragm. Indeed, Atg5-deficient mice have increased AMPK activation in the heart, suggesting a shortage in ATP production.51 However, gluconeogenic mechanisms are not fully established in the newborn,53 and they may need a direct supply of glucose. Degradation of glycogen could meet this demand.

Glycogen can be found in 2 spatially separated cellular locations in hepatocytes, the cytosol (hyaloplasmic pool), and autophagosomes. Autophagy actively participates in the overall breakdown of cellular glycogen and can selectively degrade polysaccharides. Autophagic degradation of glycogen in the newborn liver has long been recognized as an important survival mechanism for the newborn to adapt to the postnatal environment.53 Glycogen autophagy is likely stimulated by glucagon, which is secreted in response to hypoglycemia normally occurring in the newborn. In newborn hepatocytes, most of the autophagosomes are spatially and functionally related to glycogen, appearing at the borders of glycogen stores and usually containing various amounts of glycogen.54 Acid glucosidase in the lysosome is a major glycogen-hydrolyzing enzyme, and its activity increases rapidly after birth in both the liver and the heart, peaking at 6 hours in the livers of newborn rats, which is coincident with the abundance of autophagosomes around the glycogen stores. These early kinetic observations seem to be remarkably consistent with the new findings using GFP-LC3.51 This increased glycogen-hydrolyzing activity seems to be relatively specific, as there are no changes in other lysosomal enzyme activities in the liver or heart at this time of life. On the other hand, glycogen autophagy is also seen in the heart and muscles of the newborn.

Glycogen autophagy in the liver seems to be regulated by the same mechanisms already explained for protein degradation. It can thus be stimulated by glucagon, cAMP, β-adrenergic agonists, or rapamycin but suppressed by β-antagonists or insulin.53 Uniquely, glycogen autophagy could be suppressed by parenteral glucose. It is not clear how the relative selectivity of glycogen degradation versus protein degradation under the same hormonal regulation is achieved. In addition, because glucagon can promote breakdown of hyaloplasmic glycogen via the action of glycogen phosphorylase, how autophagic degradation of glycogen is coordinated with nonautophagic degradation is not clear. Perhaps the unique neonatal cellular milieu provides conditions favoring glycogen autophagy, which may be a more efficient or a more rapid way to produce nonphosphorylated glucose to the hungry newborn.

Role of Autophagy in Clearing Misfolded Proteins in Liver Diseases

Mice deficient in Atg7 in the liver accumulate a significant amount of polyubiquitinated proteins in hepatocytes and present notable hepatomegaly.52 Thus, autophagy is required for protein homeostasis in the normal liver. The identity of proteins to be degraded by autophagy in the liver under physiological condition and why this degradation is important are not clear. However, some clues might be derived from the understanding of autophagic degradation of misfolded proteins under pathological conditions.55

Alpha-1-Antitrypsin Deficiency and Hypofibrinogenemia.

Alpha-1-antitrypsin (AT), the archetype of the Serpin supergene family, is the principal blood-borne inhibitor of destructive neutrophil proteases. The normal AT protein is secreted from hepatocytes into the bloodstream, where it inhibits neutrophil proteases. However, mutations in the AT gene can result in misfolding of the molecule, which leads to its degradation and AT deficiency. The classical form of AT deficiency is caused by the z mutation (α1-antitrypsin z mutant, ATZ). Homozygous ATZ mutation causes the early onset of pulmonary emphysema, chronic liver inflammation, and hepatocellular carcinoma. In fact, AT deficiency is the most common genetic cause of liver disease in children.

The liver manifestation of the disease is caused by aggregation of the mutant ATZ in hepatocytes. As a secretory protein, AT is matured in the endoplasmic reticulum (ER). Although other forms of AT mutants could be degraded by the proteasome via the ER-associated degradation pathway (ERAD), ATZ is prone to aggregate formation, leading to its retention in the ER and subsequent ER stress and cell death.56 A prompt removal of misfolded ATZ is thus important for reduction of the toxicity. Recent studies indicate that autophagy may play a significant role in this process.56 In livers of ATZ patients as well as of ATZ transgenic mice, an increasing number of autophagosomes can be observed.57 In addition, an important adaptor protein for autophagy, p62/SQSTM1, can be detected in aggregates in the ATZ liver (see below).58 In cells overexpressing ATZ, autophagosomes seem to be closely juxtaposed with the ATZ aggregates, and suppressing autophagy retards the removal of ATZ.59 Furthermore, in the yeast model of ATZ degradation, it has also been found that aggregated ATZ is targeted to the autophagy pathway for degradation.60 It is thus conceivable that autophagy can play an important role in the liver pathogenesis of AT deficiency.

The importance of autophagy in removing misfolded proteins in the liver is also illustrated in another liver ER storage disease, hypofibrinogenemia. A R375W mutant of fibrinogen γ chain (Aguadilla γD) causes misfolding of the molecule, some of which precipitates in the ER, resulting in a reduced level of functional protein in the blood (hypofibrinogenemia). Although the soluble form of the mutant can be degraded by the proteasome via ERAD, autophagy is required to degrade the excessive amount of soluble aberrant protein and most importantly the insoluble aggregates in the ER.61

How the misfolded proteins are recognized and removed by the autophagosome is not fully understood.55 The unfolded protein response (UPR) pathway can be involved. Using generic ER stress inducers, such as A23187 and thapsigargin, and proteasome inhibitors, which cause the accumulation of misfolded polyubiquitinated proteins and ER stress, we and others found that the IRE1-JNK pathway is important for autophagy induction.62, 63 In addition, overexpression of the polyglutamine expansion mutant of huntingtin elicits autophagy in a PERK-eIF2α-dependent manner.64 However, it is interesting to note that ATZ does not seem to stimulate an obvious UPR despite that it apparently induces ER stress.65 Thus, neither the IRE1 pathway nor the PERK pathway may be involved in autophagy induction in this case. Instead, it was found that a G-protein regulator, RGS16, is highly induced by ATZ.66 RGS16 could participate in autophagy induction through a mechanism that involves Gαi3, as described above. Finally, ER calcium leakage during ER stress could also trigger autophagy via several potential mechanisms (Fig. 3).67

Mallory Bodies and Alcohol-Induced Liver Pathogenesis.

Earlier studies showed that feeding alcohol to rats produced hepatomegaly associated with enlargement of the hepatocytes and protein accumulation.68–70 The mechanisms for the alcohol-induced protein accumulation in hepatocytes are not completely known. It has been suggested that alcohol exposure can alter proteolytic activity of hepatic lysosomes,71 trafficking of lysosomal enzymes,72 and microtubule structures.73

Accumulated proteins may contribute to the formation of inclusion bodies known as Mallory bodies in hepatocytes, which was frequently observed in alcoholic hepatitis and cirrhosis.74 These structures contain cytokeratin 8, cytokeratin 18, and ubiquitin-positive protein aggregates, sharing many characteristics with other inclusion bodies found in neuronal degenerative diseases, such as Lewy bodies in Parkinson's disease and huntingtin aggregates in Huntington's diseases.75 Essentially these inclusion bodies are aggregates of polyubiquitinated misfolded proteins that could not be removed by the proteasome. They can be toxic and may contribute to the pathogenesis.

Recent studies have indicated that autophagy plays important roles in clearing these protein aggregates.55 One particularly interesting common feature of these inclusion bodies is that they are all stained positive for p62/SQSTM1.58, 76 p62/SQSTM1 could serve as an important adaptor molecule, binding both the ubiquitin moiety of the misfolded proteins and LC3/Atg8 on the autophagosome.77, 78 LC3 and p62 have been found to colocalize with mutant huntingtin, which seem to be codegraded in autophagolysosomes.77, 78 Deletion of p62 affects the efficient removal of the misfolded mutant huntingtin by autophagy.77, 78

That p62 could be detected in both Mallory bodies and in aggregates in the ATZ liver58 suggests they could be subjected to autophagic removal. Whereas this has been independently demonstrated for the ATZ protein, any role of autophagy in removing Mallory bodies has yet to be determined. On the other hand, the presence of Mallory bodies may suggest that autophagy is impaired in the liver following alcohol exposure.70

Selective Organelle Degradation by Autophagy in the Liver

Sequestration of subcellular organelles, including mitochondria, ER membranes, ribosomes, Golgi apparatus and peroxisomes has been well documented in hepatocytes under various pathophysiological conditions.27, 28, 79 ER was probably the most common organelle engulfed, followed by mitochondria. By following the removal of these organelle-containing autophagosomes on insulin treatment, it was found that the “decay rate” of these autophagosomes differed from each other and from the autophagosomes containing just cytoplasm. This finding may imply a degree of specificity and selectivity in autophagic removal of various cellular constituents. Autophagy clearly plays an important, although not exclusive, role in organelle turnover in the normal liver. Under pathophysiological conditions, a significant change in autophagic removal of a certain cellular constituent can occur. Conversely, changes in the autophagic breakdown of organelles could have an impact on certain pathological processes. For example, Atg7-deficient mice can accumulate excessive numbers of peroxisomes, deformed mitochondria, and concentric membranous structures, in addition to ubiquitin-positive aggregates, in hepatocytes.52 All these could be anticipated to cause liver dysfunction in the long run.

Autophagic Degradation of Mitochondria—Mitophagy.

Autophagic degradation of mitochondria (mitophagy) in hepatocytes could be observed under both normal and pathological conditions. For example, mitochondria were among the most commonly engulfed organelles, second to ER, in vinblastine-treated liver.79 Isolated hepatocytes cultured in glucagon-supplemented amino-acid-free medium rapidly developed mitophagy.80, 81 Mitochondrial destruction through autophagy was found in liver biopsies of patients with Reye's syndrome82 and was also observed in a murine model of Reye's syndrome using influenza B virus.83

It is not clear how mitochondria are recognized by the autophagosome. The term mitophagy was proposed to imply a selective process.81 Induction of the mitochondria permeability transition (MPT) and mitochondrial depolarization may contribute to the process in some cases. Salicylates implicated in the pathogenesis of Reye's syndrome can induce MPT, which is blocked by cyclosporin A (CsA).84 In photodamage-induced mitophagy, mitochondrial depolarization occurs before engulfment takes place,81 implicating the role of MPT, which is activated on photodamage. On the other hand, in amino acid deprivation–induced mitophagy, mitochondria can retain transmembrane potential until they are fully sequestered by the GFP-LC3-labeled autophagosomes, and this process cannot be blocked by CsA.81 CsA, instead, blocks the depolarization and acidification of the autophagosome after sequestration. These results suggest that in this model of mitophagy, activation of the permeability transition may not constitute an early signal to trigger selective engulfment of mitochondria by the autophagosome but can contribute to the subsequent degradation.85 Thus, additional signals could be required for the initial recognition.

In yeast, mitophagy could potentially be triggered by osmotic swelling of the mitochondria.86 In addition, microautophagy, rather than macroautophagy, may be preferentially utilized to remove mitochondria following nitrogen deprivation. Both selective and nonselective mitophagy can occur, and a mitochondrial outer membrane protein, Uth1, is required for selective degradation.87 These findings may provide some clues to the understanding of mammalian mitophagy.

Mitophagy may be important for normal mitochondrial turnover.81 Failure to promptly remove mitochondria through autophagy may lead to accumulation of aged and dysfunctional mitochondria, with increased generation of reactive oxygen species and DNA mutations, as indicated by a study in autophagy-defective yeast.88 Consequently, mitophagy may play a critical role in inhibiting aging and tumorigenesis (see below).

Autophagic Degradation of ER-Reticulophagy (ER-phagy).

A major function of liver is drug metabolism. There is often a significant induction of genes and molecules involved in the catabolic pathway in response to drug administration. A well-studied example is the induction of the cytochrome P-450 system by phenobarbital (PB). Notably, this is usually accompanied by extensive proliferation of smooth ER, where the catabolic reactions occur. The cessation of PB treatment is followed by reduction of the ER membrane, which in a matter of several days resumes its original size. Interestingly, smooth ER membranes could be selectively sequestrated by autophagic vacuoles.89 This is perhaps the first demonstrated example of selective autophagy. The removal of extra ER membranes through autophagy was later on confirmed by detailed morphological studies90 as well as biochemical studies. In the latter, 2 typical ER membrane proteins, PB-inducible cytochrome P-450 and NADPH-cytochrome P-450 reductase, were found to be selectively degraded by autophagy in rat liver following phenobarbital withdrawal.91 Thus, autophagy is important for the recovery of normal ER structure and function after drug treatment in order to avoid potential cellular dysfunction.

How the expanded ER is selectively removed by autophagy is not known. A similar phenomenon was observed recently in yeast, which was under ER stress.92 It is possible that the unfolded protein response pathway and/or ER calcium leakage may also be involved in the induction of autophagy, as in the case of ER-retained misfolded proteins (see above).

Autophagic Degradation of Peroxisomes—Pexophagy.

Peroxisomes are single-membrane-bound organelles enriched with peroxidase and catalase.93 They are involved in multiple types of metabolism, most notably the β-oxidation of very long-chain fatty acids. Peroxisomes, mitochondria, and microsomes are the organelles for fatty acid oxidation in the liver. Peroxisomes can incur dramatic changes in number, size, and content on stimulation of peroxisome proliferators via transcriptional activation of PPARα-regulated genes. These changes are implicated in the pathogenesis of liver steatosis, steatohepatitis, and hepatocellular carcinoma.94

It seems that autophagy plays an important role in the normal homeostasis of peroxisomes. Selective removal of peroxisomes by autophagy (pexophagy) was first described in yeast, where the excessive peroxisomes required for growth under one type of medium can be rapidly removed on switching to another type of medium.95 Both macroautophagy and microautophagy could be involved in peroxisome degradation in yeast. For example, in P. pastoris, glucose-induced peroxisome degradation occurs mainly through microautophagy, whereas ethanol-induced pexophagy is mediated by macroautophagy.96

The mammalian example of pexophagy was first demonstrated in the liver following the treatment of clofibrate or phthalate esters, both peroxisome proliferators that activate PPARα.97, 98 Abundant amounts of autophagosomes were found in the livers of treated rats. In addition, pexophagy as monitored by the degradation of peroxisomal enzymes could be suppressed by 3-MA. The most definitive evidence that autophagy is required for removing excessive peroxisomes comes from a study using Atg7-deficient mice.99 Selective degradation of peroxisomes following discontinuation of phthalate ester treatment was exclusively impaired in the Atg7-deficient livers. Autophagic sequestration of peroxisomes was only observed in the wild-type, not in the Atg7-deficient, livers. It will be interesting to determine whether the deficient mice might have an altered pathological response to the peroxisomes proliferators in the liver.

Once again, how peroxisomes can be selectively degraded by autophagy is not known. In yeast (H. polymorpha), pexophagy is affected by peroxin 3 and peroxin 14, which are present on the peroxisome membranes and are required for peroxisome biogenesis.100, 101 They may provide the signals for the selective recognition of peroxisomes by the autophagy machinery. Several Atg genes, such as Atg25, Atg26, Atg28, and Atg30, have also been defined as being specifically required for pexophagy in yeast.101 However, mammalian homologues have not been defined yet. On the other hand, it now seems clear that peroxisomes are derived from the ER,93 and it may thus be useful to examine whether signals involved in ER-phagy and ER-stress-induced autophagy could be also implicated in pexophagy in mammalian cells.

Autophagy and Cell Death, Implications in Tissue Injury, Tumorigenesis, and Cancer Therapy

As an essential mechanism controlling cellular homeostasis, autophagy is intimately associated with the regulation of cell survival and cell death.102 In general, autophagy seems to promote survival by removing unwanted cellular substances and by providing nutrients under starvation conditions. It could, however, also contribute to cell death if the process is deregulated, resulting in excessive catabolism, misrecognition of cargo, and/or hijack of the apoptosis machinery.103–105 For example, overexpressed Atg5 could promote apoptosis via its interaction with FADD104 or Bcl-xL.105 The complicated relationship of autophagy and cell death has not been well understood, particularly in the in vivo setting.102 Often a stress or injury signal could activate both the cell death and autophagy pathways, in which the role of autophagy could vary depending on the context. In particular, the role of autophagy in liver injury and in liver cancer biology/therapy has yet to be explored.

Genetic evidence suggests that autophagy could serve a tumor-suppressor function. Beclin 1 is monoallelically deleted in 40%-75% of sporadic breast and ovarian cancers.106 In addition, loss of heterozygosity in the Beclin 1–interacting protein UVRAG (Fig. 3) is frequently observed in colon cancer, and introducing UVRAG into colon cancer cells suppressed their proliferation.21 Furthermore, Beclin 1/Atg6 heterozygous mice spontaneously develop tumors in multiple organs, including the liver,107 indicating that Beclin 1 is a haplo-insufficient tumor suppressor. Similarly, Atg4C-deficient mice are also prone to chemical carcinogenesis.108 How autophagy may suppress tumorigenesis is not clear, although some evidence indicates that autophagy can help to maintain chromosomal stability,109 perhaps by removing damaged mitochondria and reducing oxidative and metabolic stress.110, 111

Autophagy is likely important for the survival of tumor cells in a less optimal environment. As a general stress response, it can constitute a prominent antideath mechanism against various stress stimulations, including those resulting from cancer therapeutic modalities. It is therefore possible to suppress autophagy in order to enhance the efficacy of these agents. This idea has been tested successfully in tumor cells originated from the liver112 or other organs.62, 113–115 This approach could be particularly useful for cancer cells resistant to standard treatment as a result of a weakened apoptosis response because disabling autophagy could lead to enhanced death stimulation and/or activation of additional death mechanisms, such as necrosis.62, 113, 115 This enhancement could be tumor specific, thus increasing therapeutic efficacy.62


Autophagy research in the liver has been always very active, not only because a great diversity of autophagy processes could be induced in hepatocytes but also because the liver is crucial for the normal metabolic function of the body, to which autophagy is intimately connected. As outlined in this review, autophagy may play significance roles in 3 areas of liver pathophysiology (Fig. 4)—balance of nutrients and energy, removal of misfolded proteins, and mitigation of ER stress—and in maintenance of cellular homeostasis by engaging in organelle turnover. Much has yet to be learned about the causes, signaling events, and impact of autophagy in these areas. In addition, it is likely that additional functional roles of autophagy in liver biology and diseases will be identified. For example, an area that has not been well explored is the relationship of autophagy with cell death in liver injury and in the development and treatment of hepatobiliary neoplasia. It is anticipated that future studies will lead to novel findings in all these areas in hepatology, which will in turn contribute to the general understanding of autophagy in biology and in disease.

Figure 4.

Role of macroautophagy in liver physiology and pathogenesis. Macroautophagy may play important roles in liver physiology and liver diseases in 3 aspects, as denoted by the 3 large spheres: balance of nutrient and energy; removal of misfolded protein and mitigation of ER stress; and maintenance of cellular homeostasis by engaging in organelle turnover. Some of the known examples and/or regulators of each function are indicated by smaller spheres. The main causes, some of which are known and some of which are suspected, are listed in the boxed area above each sphere. The possible signaling events leading to autophagy as well as the potential impact of autophagy on liver disease, which have yet to be fully revealed, are also indicated. See text for more discussion.


The authors are grateful to Dr. Donna Stolz for the assistance in electron microscopy.