Intracellular Inclusions Containing Mutant α1-Antitrypsin Z Are Propagated in the Absence of Autophagic Activity*

Mutant α1-antitrypsin Z (α1-ATZ) protein, which has a tendency to form aggregated polymers as it accumulates within the endoplasmic reticulum of the liver cells, is associated with the development of chronic liver injury and hepatocellular carcinoma in hereditary α1-antitrypsin (α1-AT) deficiency. Previous studies have suggested that efficient intracellular degradation of α1-ATZ is correlated with protection from liver disease in α1-AT deficiency and that the ubiquitin-proteasome system accounts for a major route, but not the sole route, of α1-ATZ disposal. Yet another intracellular degradation system, autophagy, has also been implicated in the pathophysiology of α1-AT deficiency. To provide genetic evidence for autophagy-mediated disposal of α1-ATZ, here we used cell lines deleted for the Atg5 gene that is necessary for initiation of autophagy. In the absence of autophagy, the degradation of α1-ATZ was retarded, and the characteristic cellular inclusions of α1-ATZ accumulated. In wild-type cells, colocalization of the autophagosomal membrane marker GFP-LC3 and α1-ATZ was observed, and this colocalization was enhanced when clearance of autophagosomes was prevented by inhibiting fusion between autophagosome and lysosome. By using a transgenic mouse with liver-specific inducible expression of α1-ATZ mated to the GFP-LC3 mouse, we also found that expression of α1-ATZ in the liver in vivo is sufficient to induce autophagy. These data provide definitive evidence that autophagy can participate in the quality control/degradative pathway for α1-ATZ and suggest that autophagic degradation plays a fundamental role in preventing toxic accumulation of α1-ATZ.

prototypic member of serine protease inhibitor (serpin) superfamily proteins and the most abundant of the circulating serpins. The principal role of ␣ 1 -AT in serum is to protect lung tissues from destructive proteases (elastase, cathepsin G, and proteinase 3) released by neutrophils during inflammation. Some genetic alterations in ␣ 1 -AT are responsible for defective secretion and thus cause serum ␣ 1 -AT deficiency (1)(2)(3). The most common causal mutation found in Caucasian populations is the replacement of Glu-342 by Lys that characterizes the Z mutant of ␣ 1 -AT (␣ 1 -ATZ). This substitution is sufficient to cause an abnormality in folding early in the secretory pathway with retention of the mutant ␣ 1 -ATZ molecule in the ER of liver cells. Homozygotes for the ␣ 1 -ATZ mutation (PIZZ) are characterized by serum levels of ␣ 1 -AT that are ϳ10 -15% of those in the general population and are susceptible to two major target organ injuries. Destructive lung disease/emphysema in adults is due to a loss-of-function mechanism. Chronic liver disease often first discovered in childhood, but also affecting adults, is due to a gain-of-toxic-function mechanism in which liver cell injury results from the hepatotoxic effects of retained ␣ 1 -ATZ. However, only 8 -10% of homozygotes develop clinically significant liver disease. This observation has led to the concept that mechanisms by which cells respond to the ER retention of mutant ␣ 1 -ATZ play a role in determining which of these homozygotes develop liver disease and which are protected from it. Because previous studies have shown that a reduction in ␣ 1 -ATZ disposal activity correlates with the presence of liver disease among deficient individuals (4), the mechanisms by which ␣ 1 -ATZ is degraded are thought to be particularly important in determining the liver disease phenotype of patients with ␣ 1 -AT deficiency.
A number of studies have addressed the determinants of the cellular fate of ␣ 1 -ATZ, including retention in the ER and disposal by the quality control/degradative pathways of the ER. Seminal works by Lomas and co-workers (2,5) have shown how the Z mutation confers an unstable polymerogenic intermediate conformation on ␣ 1 -AT so that polymerization is promoted by a reactive loop:␤-sheet A linkage reminiscent of the inhibitory interaction between serpins and cognate proteases. Recent studies of the ER-associated degradation system revealed that immature or misfolded glycoproteins are captured by ER chaperone proteins calnexin/calreticulin via the terminal glucose residue on asparagine-linked oligosaccharide side chains, which is reciprocally added or trimmed by UDP-glucose:glycoprotein glucosyltransferase or glucosidase II, respectively, according to the folding status of the glycoprotein (6,7). Indeed, a stoichiometric interaction was observed between ␣ 1 -ATZ and calnexin (4,8). Terminally misfolded proteins are translocated from ER lumen to the cytoplasm and consequently degraded by the ubiquitin-proteasome system putatively deployed at the cytoplasmic face of ER.
Detailed studies of proteasome-mediated ␣ 1 -ATZ disposal, including cell-free assays using ER-derived microsomes, have suggested the existence of one disposal pathway in which the ␣ 1 -ATZ-calnexin complex is ubiquitinated on calnexin and subsequently degraded (8 -10). It is conceivable that the asparagine-linked Glc 1 Man 8 GlcNAc 2 on ␣ 1 -ATZ determines the proteasomal degradation pathway via its physical interaction with calnexin and that diversion to other degradation pathways is brought about by further mannose trimming of the oligosaccharide chain during ER retention (11). An unknown protease activity sensitive to tyrosine phosphatase inhibitors has also been reported as a potential mechanism for nonproteasomal intramicrosomal degradation of ␣ 1 -ATZ (11).
Autophagy (synonymously used here as macroautophagy) is a major intracellular degradation pathway mediated by proteins of the evolutionarily conserved Atg family unique to this function (12)(13)(14)(15)(16)(17)(18). Autophagy is characterized by bulk sequestration of cytoplasmic constituents within a double-membrane-bound vesicle, called an autophagosome, and their subsequent degradation upon fusion of this vesicle with lysosome. This process accounts for a major portion of the cellular turnover of long lived proteins and organelles such as ER and mitochondria. In addition to constitutive bulk turnover at steady state, autophagic sequestration is induced by specific physiological perturbations, such as nutrient deprivation. Several lines of evidence have implicated autophagy as a physiological response to cope with accumulation of ␣ 1 -ATZ in the ER in vivo (19 -21). The liver lesion in PIZZ patients as well as in the PiZ mouse model of ␣ 1 -AT deficiency is accompanied by a marked autophagic response as determined by ultrastructural studies. When cells engineered to express ␣ 1 -ATZ were treated with 3-methyladenine, an inhibitor of autophagy, the degradation of ␣ 1 -ATZ was attenuated (19).
Traditional methods for monitoring autophagy by means of ultrastructural criteria or the use of chemical inhibitors can be criticized because it is sometimes difficult to discriminate autophagic vacuoles from other organelles especially in degenerating cells, and because 3-methyladenine, a relatively nonspecific inhibitor of autophagy, could potentially suppress degradation pathways other than autophagy (14).
Here we tried to provide molecular evidence for autophagy-mediated disposal of ␣ 1 -ATZ by two approaches. First, we used cell lines deleted for the Atg5 gene, a target molecule for the ubiquitin-like Atg12 conjugation that is necessary for the initial steps of autophagic sequestration (12). Second, we used GFP-LC3, a defined marker for autophagosome membrane (13,22), for colocalization studies in cell lines and in the liver of novel mouse models of ␣ 1 -AT deficiency.
Pulse-Chase Experiments-Cell lines engineered for expression of ␣ 1 -ATZ either by transient or stable transfection were subjected to pulse-chase studies. The transiently transfected cell lines were studied 24 h after transfection. Separate monolayers were incubated in serumand methionine/cysteine-free medium for 1 h at 37°C followed by pulse labeling with 150 Ci/ml of 35 S-labeled EasyTag Express protein labeling mix (NEG-772; PerkinElmer Life Sciences) for 2 h at 37°C. Cells were then rinsed with chase medium and chased for several different time periods. Cells were lysed in lysis buffer containing 50 mM Tris-HCl (pH 7.5), 1% Triton X-100, 3 mg/ml BSA, 1 mM PMSF, and protease inhibitor mixture (Roche Applied Science). Cell lysates were subjected to immunoprecipitation using anti-␣ 1 -AT polyclonal antibody and protein G-Sepharose 4FF (Amersham Biosciences), followed by analysis with SDS-PAGE (10% gel) and autoradiography using LAS-3000 bioimage analyzer (Fuji Film).
Western Blotting-Cells were collected, rinsed with PBS, and lysed in PBS containing 1% Triton X-100, 1 mM PMSF, and protease inhibitor mixture (Roche Applied Science) on ice for 30 min. Triton X-100-soluble and -insoluble fractions were obtained by centrifuging cell lysates at 15,000 rpm for 10 min at 4°C. Alternatively, cells were rinsed with PBS and directly lysed in SDS sample buffer. Samples were resolved by SDS-PAGE and transferred to polyvinylidene difluoride membrane. The membranes were blocked with 5% skim milk in 0.1% Tween 20/Tris-buffered saline and then incubated with primary antibodies. Immunoreactive bands were detected using horseradish peroxidaseconjugated secondary antibodies (The Jackson Laboratories) and luminol solution (1.25 mM luminol, 65 mM Tris-HCl (pH 8.0), 0.2 mM coumaric acid, 0.01% H 2 O 2 ).
Immunofluorescence Microscopy and Flow Cytometry-For immunofluorescence microscopy, cells cultured on coverslips were fixed with 4% paraformaldehyde/PBS, permeabilized with 0.1% Triton X-100/PBS, and blocked with 3% BSA/PBS. Primary antibodies were diluted 1:100, and secondary antibodies were diluted 1:200 in 1% BSA, 0.1% Triton X-100/PBS. Coverslips were successively incubated with primary antibodies, Alexa-conjugated secondary antibodies (Invitrogen), and 1 g/ml Hoechst 33342/PBS (Sigma) with intervening washes with PBS. Samples were examined by using Olympus FV1000 confocal microscopy. For flow cytometry, cells were collected, rinsed with PBS, and fixed with 4% paraformaldehyde/PBS. Cells in suspension were stained using anti-␣ 1 -AT polyclonal antibody and Alexa 488-conjugated secondary antibody as in immunofluorescence microscopy samples and analyzed by BD FACScan.
Mice-GFP-LC3 transgenic mice were described previously (22). Z mice, in which expression of human ␣ 1 -ATZ is induced only in liver parenchymal cells upon removal of doxycycline, were produced by using the Tet-Off gene expression system and TALap 2 mice (26). These mice were crossed to produce Z ϫ GFP-LC3 mice. PiZ mice with constitutive expression of ␣ 1 -ATZ (27) were also crossed to produce PiZ ϫ GFP-LC3 mice. Liver sections were viewed by confocal microscopy. Quantitative morphometry was carried out by counting green vacuoles in 10 cells with many red globules and 10 cells with few or no red globules in three random areas of the liver each from two different liver sections. The results were analyzed using the MetaMorph software program.

RESULTS
The Degradation Rate of ␣ 1 -ATZ Is Attenuated in Atg5 Ϫ/Ϫ Cells-A previous study demonstrated that disruption of the Atg5 gene in an ES cell line resulted in complete abrogation of autophagy, as confirmed by both morphological and biochemical analyses (14). To investigate whether ␣ 1 -ATZ is degraded via the autophagic pathway, we first carried out pulse-chase analysis in wild-type and Atg5 Ϫ/Ϫ ES cells transiently transfected with the ␣ 1 -ATZ expression plasmid. Twenty four hours after transfection, cells were metabolically radiolabeled for 2 h with [ 35 S]methionine/cysteine and chased for 0, 4, 6, and 8 h. Cell lysates were subjected to immunoprecipitation using anti-␣ 1 -AT antibody, and immunoprecipitated samples were resolved by SDS-PAGE followed by autoradiography. A semi-logarithmic plot of ␣ 1 -ATZ-specific signals against time indicated that there was more than a 2-fold decrease in the degradation rate of ␣ 1 -ATZ in Atg5 Ϫ/Ϫ cells in comparison with that in wild-type cells (Fig. 1A). Cell culture fluid from these cells was also subjected to immunoprecipitation and autoradiography at the same time, but secreted ␣ 1 -ATZ was barely detectable in either wild-type or Atg5 Ϫ/Ϫ cells (data not shown), indicating that secretion is not the cause for more rapid disappearance of ␣ 1 -ATZ in wild-type cells. The intracellular half-life of ␣ 1 -ATZ was calculated as 122 min (r 2 ϭ 0.99) and 274 min (r 2 ϭ 0.97) in wild-type and Atg5 Ϫ/Ϫ cells, respectively. The value in wild-type ES cells is well within the range of values that have been described previously in transfected fibroblasts and hepatoma cell lines (8,10). Next, we examined the fate of ␣ 1 -ATZ by pulse-chase experiments in wild-type and Atg5 Ϫ/Ϫ MEF cells engineered for stable expression of ␣ 1 -ATZ (Fig. 1B). The results show that there is a delay in the disappearance of ␣ 1 -ATZ in the Atg5 Ϫ/Ϫ cells compared with the wild-type MEFs. The difference in degradation of 52-kDa precursor ␣ 1 -ATZ in the presence and absence of autophagic activity in stable transfected MEFs was almost identical to that in transiently transfected ES cells. A trace amount of mature 55-kDa ␣ 1 -ATZ was secreted into extracellular fluid. Most interestingly, the abrogation of autophagy was associated with a slight increase in the secretion of ␣ 1 -ATZ in these cell lines, the significance of which is not yet known.
To compare the relative contribution of autophagy and the proteasomal pathway in ␣ 1 -ATZ degradation in these cells, we performed pulsechase analysis in the presence of various proteasome inhibitors. Twenty four hours after transient transfection with the ␣ 1 -ATZ expression plasmid, cells were preincubated with Met/Cys-free medium for 1 h, metabolically labeled for 2 h, and chased for 4 h in the presence of proteasome inhibitors MG115 (20 M), epoxomicin (10 M), lactacystin (10 or 30 M), or vehicle control. The results show that degradation of ␣ 1 -ATZ was inhibited by all of the proteasome inhibitors in the wild-type cells but not in the Atg5 Ϫ/Ϫ cells. The ␣ 1 -ATZ degradation rate in wild-type cells treated with proteasome inhibitor was almost same as that in Atg5 Ϫ/Ϫ cells treated with vehicle control (Fig. 1C), indicating that the inhibitory effects of autophagy and the proteasomal pathway were nearly equivalent in these experiments. To exclude the possibility that delayed ␣ 1 -ATZ degradation in Atg5 Ϫ/Ϫ cells is secondary to proteasome inhibition in these cells, we analyzed proteasome activity by using Ub-G76V-EGFP, a model substrate for the ubiquitin-fusion degradation pathway (23). Wild-type and Atg5 Ϫ/Ϫ MEF cell lines were engineered for stable expression of Ub-G76V-EGFP and then the cell lines were subjected to flow cytometric analysis for the fluorescent signal. The results showed that there were no differences between the two cell lines in the absence or presence of MG132. Furthermore, transient transfection of the ␣ 1 -ATZ expression plasmid had no effect on the levels of the proteasomal substrate (Fig. 1D). There was also no difference in the level of polyubiquitinated proteins in the two cell lines as determined by Western blot analysis for ubiquitin (data not shown). These data indicate that expression of ␣ 1 -ATZ does not inhibit proteasomal activity and furthermore that inhibition of proteasomal activity cannot be an explanation for the delayed degradation in the Atg5-deficient background. Thus, the results of Fig. 1 provide definitive evidence that autophagy can contribute to degradation of ␣ 1 -ATZ.
␣ 1 -ATZ Accumulation in Atg5 Ϫ/Ϫ Cells Is Augmented Over Time-Next, we examined steady-state levels of ␣ 1 -ATZ in wild-type and Atg5 Ϫ/Ϫ cells ( Fig. 2A). The results show a significant increase in levels of ␣ 1 -ATZ in the Atg5 Ϫ/Ϫ cells. ␣ 1 -ATZ degradation in Atg5 Ϫ/Ϫ cells was restored by cotransfection of wild-type Atg5 but not mutant Atg5 K130R (14), indicating that ␣ 1 -ATZ disposal requires functional Atg5 that is covalently modified by Atg12 to form autophagosomes and to promote the conversion of LC3-I to faster migrating LC3-II (13, 16) ( Fig. 2A). To examine whether the degradation of any exogenously expressed protein would be inhibited in the autophagy-deficient background, we employed EYFP as a control cytosolic protein. When ␣ 1 -ATZ and EYFP were cotransfected in wild-type and Atg5 Ϫ/Ϫ cells, ␣ 1 -ATZ accumulated in Atg5 Ϫ/Ϫ cells, but there was no difference in the amounts of EYFP in wild-type as compared with Atg5 Ϫ/Ϫ cells (Fig.  2B). At least a portion of the increased ␣ 1 -ATZ levels was found in Triton X-100-insoluble fractions, recapitulating the previous observations that polymerogenic ␣ 1 -ATZ forms detergent-insoluble aggregates (10,25,28) (Fig. 2B).
To examine the accumulation profile of ␣ 1 -ATZ over time, wild-type and Atg5 Ϫ/Ϫ MEF cells were transiently transfected with the ␣ 1 -ATZ expression plasmid and then examined 24, 48, and 72 h after transfection. Cells were directly lysed into SDS sample buffer, and whole cell lysates were analyzed for the amount of ␣ 1 -ATZ by Western blotting and densitometry (Fig. 2C). Atg5 Ϫ/Ϫ cells contained higher amounts of ␣ 1 -ATZ than wild-type cells at each time point. The total level of ␣ 1 -ATZ peaked around 24 h and then gradually decreased in both wildtype and Atg5 Ϫ/Ϫ cells. The difference in accumulation of ␣ 1 -ATZ in Atg5 Ϫ/Ϫ cells compared with that in wild-type cells was progressively greater at later times, possibly reflecting the differential degradation rate. A previous study demonstrated that the conversion of cytosolic LC3-I to membrane-bound LC3-II represents a good marker for cellular autophagic activity (13,16). Although in this experiment LC3-II levels were not significantly changed between mock-and ␣ 1 -ATZ-transfected wild-type MEF cells (Fig. 2C), in vivo studies revealed autophagic induction in the liver following ␣ 1 -ATZ expression (as demonstrated below).
In both wild-type and Atg5 Ϫ/Ϫ cells, ␣ 1 -ATZ migrated as a single 52-kDa band during chase periods (Fig. 1) and longer time intervals of ␣ 1 -ATZ Inclusion in the Absence of Autophagy FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7 expression (Fig. 2C), suggesting that ␣ 1 -ATZ accumulates as an ERglycosylated form in either the presence or the absence of autophagic activity. This was further confirmed by the endoglycosidase-H sensitivity assay. ␣ 1 -ATZ expressed in Atg5 Ϫ/Ϫ cells was immunoprecipitated and digested with endoglycosidase-H, followed by immunoblot analysis using anti-␣ 1 -AT. The 52-kDa polypeptide was cleaved to 46 kDa, indicating that it represents the ER-glycosylated form of ␣ 1 -ATZ (Fig. 2D) (29). These data indicate that glycosylation and the degradative intermediate profile of ␣ 1 -ATZ are not altered in the absence of autophagy.
To exclude the possibility that the reduction in the level of ␣ 1 -ATZ over time is due to a reduction in the percentage of cells expressing ␣ 1 -ATZ, flow cytometric analysis was performed on MEF cells 24, 48, and 72 h after transfection with the ␣ 1 -ATZ expression plasmid (Fig.   2E). The results show that there is a decrease in ␣ 1 -ATZ-positive cells in both cases over time with greater ␣ 1 -AT fluorescence in the Atg5 Ϫ/Ϫ than in the wild-type background at each time point. The results were almost identical to the steady-state levels of ␣ 1 -ATZ determined by densitometric analysis as shown above. These data also indicated that initial transfection efficiencies achieved similarly high levels of expression of ␣ 1 -ATZ in wild-type and Atg5 Ϫ/Ϫ cells. Because transfected wild-type and Atg5 Ϫ/Ϫ cells divided at a nearly equivalent rate during the 72-h time course (data not shown), the more rapid disappearance of ␣ 1 -ATZ in wild-type cells could not be attributed to the propagation of untransfected cells or the dilution of ␣ 1 -ATZ because of more rapid cell division. The data were consistent with the idea that ␣ 1 -ATZ is synthesized at a roughly equivalent rate but is degraded at a slower rate in Atg5 Ϫ/Ϫ cells compared with parental wild-type cells, confirming the pivotal role of autophagic pathway in ␣ 1 -ATZ disposal.
␣ 1 -ATZ Inclusion Formation Is Accelerated in Atg5 Ϫ/Ϫ Cells-The accumulation profile of ␣ 1 -ATZ in transiently transfected cell lines over time revealed that the net amount of ␣ 1 -ATZ progressively decreased after 24 h and that ␣ 1 -ATZ disposal was significantly delayed in Atg5 Ϫ/Ϫ cells. Next, we compared ␣ 1 -ATZ localization in wild-type to Atg5 Ϫ/Ϫ MEF cells 24 and 72 h after transfection. We used immunostaining with anti-KDEL antibody as a marker of the ER, because it recognizes a series of ER-soluble proteins that have the C-terminal KDEL tetrapeptide ER retrieval signal, and immunostaining for cotransfected GFP-LC3 as a marker for autophagosomes. The results showed that 24 h after transfection in both wild-type and Atg5 Ϫ/Ϫ cells and 72 h after transfection in wild-type cells, ␣ 1 -ATZ primarily localized to the ER as evidenced by its colocalization with anti-KDEL staining (Fig. 3A, left panel). This ER localization of ␣ 1 -ATZ is consistent with that described in previous studies. Seventy two hours after transfection of Atg5 Ϫ/Ϫ cells, ␣ 1 -ATZ was also prominently localized to cytoplasmic inclusion body-like structures distinct from the original ER pattern. The size and morphology of these inclusions ranged from small dots to large aggregates. Although these structures did not colocalize with anti-KDEL staining (Fig. 3A, middle and right panels), they did colocalize with an ER membrane chaperone protein calnexin (Fig. 3B), which directly interacts with ␣ 1 -ATZ (4,8). Similar cytoplasmic inclusions were also found in wildtype cells but at lower frequency. GFP-LC3, which labeled autophagosomes in wild-type cells and diffusely stained the cytoplasm in Atg5 Ϫ/Ϫ cells, did not significantly overlap with ␣ 1 -ATZ inclusions in either wild-type or Atg5 Ϫ/Ϫ cells (Fig. 3C). These inclusion body-like structures and the ER pattern often coexisted in individual cells, suggesting that ␣ 1 -ATZ inclusions gradually emerge in the shadow of the ER pat-

-ATZ accumulation in Atg5 ؊/؊ cells relative to that in wild-type (WT) cells is augmented over time.
A, wild-type and Atg5 Ϫ/Ϫ ES cells were lysed for Western blot analysis 24 h after transfection with ␣ 1 -ATZ expression plasmid together with mock vector (ϩVec), expression plasmid for wild-type Atg5 (ϩAtg5), or Atg5 K130R (ϩKR). B, wild-type and Atg5 Ϫ/Ϫ ES cells were lysed 24 h after transfection with expression plasmids for ␣ 1 -ATZ and EYFP. Cells were lysed in 1% Triton X-100-containing lysis buffer, and insoluble materials were lysed into SDS sample buffer. C, wild-type and Atg5 Ϫ/Ϫ MEF cells were directly lysed in SDS sample buffer for Western blot analysis 24, 48, and 72 h after transient transfection with the ␣ 1 -ATZ expression plasmid. Equal sample loading was confirmed by Coomassie staining of the blot (not shown). The bar graph in the lower panel shows the relative density units of tern and persist after the disappearance of ␣ 1 -ATZ from the ER localization.
To confirm the apparent tendency for Atg5 Ϫ/Ϫ cells to bear ␣ 1 -ATZ inclusions, we used morphometric analysis to quantify the inclusions in wild-type and Atg5 Ϫ/Ϫ MEF cells transfected with the ␣ 1 -ATZ expression plasmid. Seventy two hours after transfection, cells were stained for ␣ 1 -ATZ, and a series of confocal fluorescent images was obtained by using parameters of acquisition that were kept constant (Fig. 3D). ␣ 1 -ATZ-positive cells were counted (Fig. 3E, blue bars) and grouped into one of two categories according to the following criteria. Cells that contain ␣ 1 -ATZ inclusions with visually discernible boundaries with little or no ER staining were counted as inclusions (Fig. 3E, red bars). Other cells in which ␣ 1 -ATZ was primarily localized to the ER were counted under the "ER pattern" category. This quantitative morphometric analysis showed that ␣ 1 -ATZ inclusions were clearly favored in Atg5 Ϫ/Ϫ cells (Fig. 3E), indicating that the formation of cytoplasmic inclusions is increased in association with the absence of autophagic activity.
GFP-LC3-labeled Autophagosomes Colocalize with ␣ 1 -ATZ-By using GFP-LC3 as a marker for autophagosomes (13,16,22), we next examined the possibility that ␣ 1 -ATZ is sequestered within autophagosomes. We already observed in Atg5 Ϫ/Ϫ cells that ␣ 1 -ATZ forms inclusions in the cytoplasm that have very little overlap with staining by anti-KDEL and do not colocalize with GFP-LC3 (Fig. 3C, lower panel). In wild-type cells ␣ 1 -ATZ inclusions were much less frequent, and there was very little colocalization with anti-KDEL staining or staining with GFP-LC3 (Fig. 3C, upper panel). Here we found in wild-type MEF cells a few concentrated areas in the cytoplasm in which staining for ␣ 1 -ATZ colocalized with anti-KDEL and GFP-LC3 staining (Fig. 4A), suggesting the possibility that these are autophagosomes sequestering ␣ 1 -ATZ. To confirm this, we examined whether the inhibition of fusion between autophagosome and lysosome resulted in an accumulation of autophagosomes containing ␣ 1 -ATZ; once the fusion occurs, they disappear by degradation. The small GTPase Rab7 is known to play a role in autophagosome-lysosome fusion as well as in vesicular transport to late endosomes and biogenesis of lysosomes so that the GDP-form mutant Rab7 T22N has a dominant-negative effect on the fusion event (30,31). In the presence of Rab7 T22N, the colocalization between GFP-LC3 and ␣ 1 -ATZ was readily detected in wild-type MEF cells 36 h after transfection (Fig. 4B). Similar results were obtained when cells were treated with vacuolar H ϩ -ATPase inhibitor bafilomycin A 1 , another means to inhibit autophagosome-lysosome fusion (32) (data not shown). Because such colocalization is not found in Atg5 Ϫ/Ϫ cells (Fig. 4C), the data suggest that the colocalizing structures in the wild-type cells represent genuine autophagosomes. Together, these results not only demonstrate the sequestration of ␣ 1 -ATZ into autophagosomes but also further confirm that the autophagic pathway contributes to the efficient disposal of ␣ 1 -ATZ.
Previous studies have shown that there is a statistically significant increase in autophagosomes in liver cells in mouse models of ␣ 1 -AT deficiency as well as in liver cells of biopsy specimens from patients with ␣ 1 -AT deficiency (19). To determine whether the autophagic response is specifically induced by accumulation of ␣ 1 -ATZ in the ER of liver cells in vivo, we generated the Z ϫ GFP-LC3 mouse model by mating the Z mouse with liver-specific inducible expression of ␣ 1 -ATZ to the GFP-LC3 mouse, which renders autophagosomes green under fluorescent microscopy. Separate groups of Z ϫ GFP-LC3 mice that had aged to 3 months of age with doxycycline in their drinking water (␣ 1 -ATZ gene expression suppressed) or without doxycycline (␣ 1 -ATZ gene expression induced) were sacrificed, and their livers were examined under fluorescent microscopy (Fig. 5, C and D, respectively) in comparison to the liver of 3-month-old GFP-LC3 mice that were either fed or starved for 24 h prior to sacrifice (Fig. 5, A and B, respectively). The results show that withdrawal of doxycycline and therein induction of the mutant protein is sufficient to elicit GFP-LC3-labeled structures (Fig. 5, D compared with C). This pattern of labeling was almost identical to that seen in the GFP-LC3 mouse but only after starvation (Fig. 5, B compared with  A), indicating that the labeled structures were indeed autophagosomes. These data provide further confirmation of the induction of the hepatic autophagic response in ␣ 1 -AT deficiency and, moreover, that the expression of the mutant ␣ 1 -ATZ molecule is sufficient to induce hepatic autophagy in vivo.
Finally, we examined the localization of autophagosomes in the liver of the Z ϫ GFP-LC3 mouse relative to ␣ 1 -ATZ. It is well known that mutant ␣1-ATZ can be detected by immunostaining in the liver of patients with ␣ 1 -AT deficiency or in mouse models of ␣ 1 -AT deficiency, but wild-type ␣ 1 -AT cannot be detected in the normal human liver (1,33). This has been presumed to be due to the fact that the mutant protein accumulates in liver cells to levels that are well beyond those that occur for the wild-type protein and to levels that reach the threshold for detection by immunostaining. It is also well known that mutant ␣ 1 -ATZ forms large globules in the ER of some, but not all, hepatocytes (27). The globule-devoid hepatocytes are thought to be progenitor, or at least relatively immature, hepatocytes that express ␣ 1 -ATZ but not to the extent that there is enough accumulation to form intracellular globules (27,34). In Fig. 5, E and F, we immunostained liver from the Z ϫ GFP-LC3 mouse shown in Fig. 5D for ␣ 1 -AT by using a red fluorophore. The results show that green autophagosomes are present within hepatocytes that have multiple intracellular globules of ␣ 1 -AT (Fig. 5E) as well as within hepatocytes with lesser numbers of ␣ 1 -AT-containing globules (Fig. 5F). Quantitative morphometric analysis showed that there were no significant differences in the number of globule-containing and globule-devoid hepatocytes that had autophagosomes. It was also notable that in the hepatocytes that had multiple intracellular ␣ 1 -ATZ-containing globules the autophagosomes were often closely adjacent but not colocalized with the globules (Fig. 5E, arrowheads). For reasons not yet determined, the ␣ 1 -AT-containing globules in the liver of the Z mouse model are considerably smaller than the ones that have been detected in the PiZ mouse model, which has constitutive expression of ␣ 1 -ATZ, that is not restricted to hepatocytes (27). In the PiZ ϫ GFP-LC3 mouse, green autophagosomes were also seen in the hepatocytes with these larger globules as well as in globule-devoid hepatocytes in a manner almost identical to the hepatocytes of the Z ϫ GFP-LC3 mouse (data not shown).

DISCUSSION
In the present study we provide the first genetic evidence that autophagy constitutes a major pathway for degradation of ER-retained ␣ 1 -ATZ. This is also the first report to show that autophagosomes can segregate and concentrate a mutant protein ␣ 1 -ATZ as cargo by using GFP-LC3, the most reliable marker for autophagosomes currently available (Figs. 4 and 5). Although it has been widely accepted that autophagic sequestration is a nonspecific bulk process, our data suggest that ␣ 1 -ATZ is delivered to autophagosomes in a more effective manner than a control cytosolic protein EYFP (Fig. 2B). Our recent study also demonstrated that invading pathogenic group A streptococci were spe- . Liver specimens were immunostained with anti-GFP to enhance the fluorescence of GFP-LC3. The sections were then examined under fluorescence microscopy. C and D, Z ϫ GFP-LC3 mice were maintained either in the presence (C) or absence (D) of doxycycline (Dox) from birth until 3 months of age. The ␣ 1 -ATZ gene is only induced in the liver of these mice after withdrawal of doxycycline. These mice were not starved. Sections of liver were immunostained with anti-GFP and then examined by fluorescence microscopy. Sections of liver from D were immunostained with anti-␣ 1 -AT using a red fluorophore together with anti-GFP (E and F). An area having hepatocytes with many ␣ 1 -AT globules is shown in E, and one with hepatocytes having few or no ␣ 1 -AT globules is shown in F. Several areas in which green autophagosomes adjacent to red globules are particularly apparent are indicated by arrowheads.
cifically enclosed by autophagosomes to be delivered and killed in lysosomes, further indicating that autophagic sequestration can occur in a substrate-specific manner (35). Previous morphometric studies demonstrated that administration of phenobarbital induced the proliferation of ER membranes in rat liver, and following the cessation of this drug excess ER membranes were removed in parallel with a specific increase in the volume and number of autophagosomes containing ER membranes (36). There has also been a report showing that ER membrane proteins, phenobarbital-inducible cytochrome P450, and NADPH-cytochrome P-450 reductase were segregated by autophagy to be degraded in lysosomes without leakage into the cytosol fraction (37). A similar mode of autophagic sequestration could facilitate the specificity by which ER-retained ␣ 1 -ATZ is degraded.
Previous studies have shown that autophagic activity is increased in the ␣ 1 -ATZ-expressing liver of the PiZ transgenic mouse model in the absence of starvation and that the increase reaches levels that are comparable with the levels of autophagic activity induced in the liver of wild-type mice by starvation (19,20). Moreover, in those studies starvation did not elicit any further increase in autophagic activity in the liver of the PiZ mouse (20), which could mean that autophagic activity is already at saturable levels or that starvation leads to increased clearance and well as induction of the formation of autophagosomes. Most interestingly, we found that autophagic activity was not increased in transiently transfected cell lines, as judged by LC3 modification (Fig. 2C), but was clearly induced in the Z ϫ GFP-LC3 mouse by 3 months of age (Fig. 5). The most likely explanation is that induction of autophagy requires a duration of accumulation of ␣ 1 -ATZ that is longer than the 3-day duration of a transiently transfected system. One attractive alternative possibility is that new autophagic activity is only induced when existing "constitutive" autophagy is saturated with mutant-aggregated protein. However, it is also possible that there are differences in the milieu in vivo or that retention of ␣ 1 -ATZ in the ER has complex effects on both formation and clearance of autophagosomes.
Another interesting result of this study was the induction of autophagosomes in globule-devoid as well as globule-containing hepatocytes. Previous studies have suggested that globule-devoid hepatocytes do express ␣ 1 -ATZ but to apparently lesser levels than globule-containing hepatocytes, presumably because they are progenitor or younger cells (27,33,34). Thus, the most likely explanation for the presence of autophagosomes in both of these cell populations is that the threshold for induction of autophagy is reached at the lower level of ␣ 1 -ATZ in the globule-devoid hepatocytes. If this explanation is correct, then it is possible that the induced autophagy could in turn be responsible for limiting the formation of globules in the globule-devoid hepatocytes. However, there are a number of other explanations for the presence of autophagosomes in both cell populations, including the possibility that the accumulation of ␣ 1 -ATZ to higher levels has an effect on the clearance of autophagosomes and the remote possibility that either cell population has a "trans" effect on induction of autophagy in the other cell population.
Immunofluorescence studies showed that many characteristic inclusion body-like structures were formed by ␣ 1 -ATZ in Atg5 Ϫ/Ϫ cells (Fig.  3A). These inclusions immunolabeled for ER membrane protein calnexin ( Fig. 3B) but not for KDEL-containing proteins. This labeling pattern could be explained by the formation of an ␣ 1 -ATZ-calnexin complex that is not dissociated during the formation of inclusions, whereas KDEL-possessing soluble proteins are not associated with, or become dissociated from, ␣ 1 -ATZ before evolution into inclusions within the ER lumen or during movement out of the ER. These structures share some characteristics with the recently described ER quality control compartment or concentric membranous body, including membranous morphology and colocalization of ER membrane chaperone calnexin and mutant substrate proteins (38,39). Because proper function of the ER appears to be maintained while mutant proteins accumulate within these structures, they may constitute reservoirs of aberrant proteins similar to what has been attributed to aggresomes (40,41). However, it should be noted that the overall morphology of ␣ 1 -ATZ inclusions and the lack of specific localization within the cytoplasm indicate that they are not structurally cognate with aggresomes, which converge at pericentriolar region by retrograde transport on microtubule network.
The tendency for formation of ␣ 1 -ATZ inclusions in autophagy-deficient cells may also bear on the apparent lack of additive or synergistic effects of proteasomal inhibition in Atg5 Ϫ/Ϫ cells (Fig. 1C). By using the proteasome sensor construct Ub-G76V-EGFP reported by Dantuma et al. (23), we did not detect differences in the proteasome activity in wild-type and Atg5 Ϫ/Ϫ cells, in the absence or presence of proteasome inhibitors (Fig. 1D). Proteasome activity was not altered following the expression of ␣ 1 -ATZ, and Atg5 Ϫ/Ϫ cells contained a normal amount of polyubiquitinated proteins. Another recent study has also reported that proteasome activity is unperturbed in the absence of autophagy (42). However, it could be hypothesized that the ␣ 1 -ATZ sequestered into inclusions cannot be a substrate for retrograde translocation out of the ER, if that is a necessary step in the proteasomal mechanism that characterizes the ER-associated degradation pathway for ␣ 1 -ATZ. Accordingly, the preferential formation of ␣ 1 -ATZ inclusions in Atg5 Ϫ/Ϫ cells might decrease the contribution of proteasomes to the overall cellular disposal of ␣ 1 -ATZ. If so, autophagy should be particularly important in the degradation of aggregated ␣ 1 -ATZ, and therefore it becomes increasingly important as the level of mutant protein accumulates in the ER.
In conclusion, this study demonstrates that autophagy is indispensable for efficient disposal of ER-retained ␣ 1 -ATZ. Although it is still not entirely clear how protein aggregates found in conformational diseases such as ␣ 1 -AT deficiency result in tissue pathology or whether the pathobiology reflects specific characteristics of the protein that aggregates in each of these diseases, there is a great deal of circumstantial evidence that the cellular accumulation of ␣ 1 -ATZ aggregates is an important determinant of hepatic injury in ␣ 1 -AT deficiency. In this regard, future studies of the detailed molecular mechanisms by which ER-retained ␣ 1 -ATZ induces autophagy and is sequestered in autophagosomes are likely to have clinical significance as well as important implications for the mechanism and function of the autophagic pathway itself.