several recent studies have described specific morphological alterations in subcellular structure in response to the accumulation of misfolded or unassembled proteins in the endoplasmic reticulum (ER). Raposo et al. (26) examined the accumulation of unassembled class I major histocompatibility complex (MHC) molecules in the ER of thymic epithelial cells, which overexpress heavy chains on a genetic background that is deficient for the peptide transporter TAP1. In these cells, class I MHC heavy-chain molecules accumulated in an expanded post-ER/pre-Golgi network, or ER-Golgi intermediate compartment, that consists of tubulated and fenestrated smooth membranes. Johnston et al. (18) examined the accumulation of misfolded cystic fibrosis transmembrane conductance regulator (CFTR) ΔF508 and presenilin-1 (PS1) molecules in human embryonic kidney 293 cells. Undegraded CFTR and PS1 molecules, modified by ubiquitination, were found in pericentriolar cagelike structures surrounded by the intermediate filament protein vimentin. These “aggresomes” were induced by high levels of expression of misfolded CFTR, PS1 molecules, or unassembled T cell receptor α-subunits or by chemical inhibition of proteasomal degradation in the presence of lower levels of expression of the misfolded membrane proteins. Aggresomes have also recently been described in cells in which there has been cytoplasmic accumulation of certain viral proteins (2).

It has been known for some years that the ER possesses complex machinery, called the quality control apparatus, by which it can recognize, retain, and degrade misfolded proteins (11,29). This includes proteins that are unable to fold properly, unable to undergo posttranslational modifications such as glycosylation or formation of intra- or intermolecular disulfide bonds, or unable to assemble into hetero- or homooligomers because of naturally occurring mutations associated with disease states/deficiency disorders or experimental conditions. Recent studies have shown that the ER degradative mechanism(s) involves retrotranslocation, or dislocation, of misfolded proteins across the ER membrane into the cytoplasm and that dislocation, for the most part, is coupled to covalent modification by polyubiquitin chains for proteolysis by the proteasome (11). Recent studies have also shown that the ER possesses signaling pathways, such as the unfolded protein response, that permit it to respond to the presence of retained misfolded proteins by altering expression of its chaperones and components of its membrane structure (29).

In the classic form of α₁-antitrypsin (α₁AT) deficiency (homozygous PIZZ α₁AT deficiency), the mutant α₁ATZ molecule is retained in the ER of liver cells rather than secreted into the blood and extracellular fluid, where it ordinarily functions as an inhibitor of neutrophil elastase. Carrell and Lomas (8) have shown that the mutation that characterizes the α₁ATZ molecule results in aberrant polymerization in the ER by a loop-sheet insertion mechanism. Retention of the misfolded mutant α₁AT protein in the ER is thought to cause severe liver injury and hepatocellular carcinoma in a subgroup of deficient individuals (31). In fact, this deficiency constitutes the most common genetic cause of liver disease in children. Studies in genetically engineered cell lines from patients with α₁AT deficiency have shown that there is a correlation between susceptibility to liver disease and delayed ER degradation of α₁ATZ (33). Deficient individuals who are “protected” from liver disease have more efficient ER degradation of α₁ATZ and a lesser net burden of mutant α₁AT molecules retained in the ER. Recent studies have shown that α₁ATZ is similar to other mutant proteins that are retained in the ER in that its degradation in the ER appears to involve the proteasome (5, 20, 24). Studies in intact genetically engineered cell lines and in the cell-free microsomal translocation system indicate that degradation involves at least several steps, including, first, stable binding to the transmembrane ER chaperone calnexin and then polyubiquitination of calnexin and proteolysis of the α₁ATZ-polyubiquitinated calnexin complex by the 26S proteasome (24). Fractionation of reticulocyte lysate used in the cell-free microsomal translocation system and reconstitution with purified ubiquitin proteins has shown that the ubiquitin-conjugating enzyme E2-F1 plays a role in the ubiquitin-dependent proteasomal mechanism for degradation of α₁ATZ (30). Moreover, studies in the cell-free system indicate that ubiquitin-independent proteasomal and nonproteasomal mechanisms may also contribute to intracellular α₁ATZ degradation.

In this study, we examined the effect of ER retention of α₁ATZ in human fibroblast cell lines to determine whether there are specific morphological changes that will ultimately provide more information about its hepatotoxic and oncogenic properties.

MATERIALS AND METHODS

Materials.

Antibodies used against α₁AT included rabbit anti-human α₁AT from DAKO (Santa Barbara, CA), goat anti-human α₁AT from Cappel (Durham, NC), and monoclonal mouse anti-human α₁AT from Zymed (San Francisco, CA). Antibodies against human calnexin included rabbit anti-human calnexin (DP23 and DP33) generated in our laboratory (33), SPA-865 from StressGen (Victoria, BC, Canada), and mouse anti-human calnexin from Chemicon (Temecala, CA). Lyso-tracker Red (LTR) and ER-tracker Blue (ETB) were purchased from Molecular Probes (Eugene, OR). Anti-dinitrophenyl antibody, Cy3-conjugated anti-vimentin antibody, monodansylcadaverine (MDC), 3-methyladenine (3MA), and wortmannin were purchased from Sigma (St. Louis, MO). LY-294002 was purchased from Biomol Research Laboratories (Plymouth Meeting, PA). Liver tissue from three patients with α₁AT deficiency and one normal liver transplant donor was used for electron microscopy (EM). One of the deficient livers and the normal liver were processed for EM at the time of harvest. One of the deficient livers had been fixed in glutaraldehyde and stored until processed for EM for this study. The third deficient liver was reprocessed for EM from paraffin-embedded tissue. Liver tissue from four PiZ mice (7) and four wild-type mice of the same genetic background was freshly prepared for EM as described inEM.

Cell lines and labeling.

Human fibroblast cell lines engineered for stable expression of mutant α₁ATZ by transduction of amphotropic recombinant retroviral particles have been previously described (33). The same cell lines transduced with wild-type α₁ATM or with vector alone or not transduced at all were used as additional controls. Cell lines from three PIZZ α₁AT-deficient patients with liver disease (susceptible hosts) and two PIZZ individuals without liver disease (protected hosts) were used. For this study, the same human fibroblast cell lines as well as Chinese hamster ovary (CHO) cells were transduced with amphotropic recombinant retroviral particles bearing mutant CFTRΔF508 cDNA (kindly provided by Dr. Richard Gregory, Framingham, MA).

Two hepatoma cell lines were used. The murine hepatoma cell line Hepa1–6 engineered for stable constitutive expression of human α₁ATZ (Hepa1–6N2Z9) has previously been described (6). This study also used a rat hepatoma, H11 (kindly provided by K. Fournier, Seattle, WA), which has the properties of a well-differentiated hepatocyte but does not express endogenous α₁AT because hepatocyte nuclear factor-1α and -4 gene expression have been extinguished (27). The H11 cell line was engineered for stable constitutive expression of human α₁ATZ exactly as previously described for human fibroblast cell lines (33). Pulse radiolabeling experiments showed that human α₁AT gene expression had been conferred and that there was intracellular retention of α₁ATZ in this cell line, H11N2Z1 (data not shown).

This study also used a HeLa cell line engineered for inducible expression of human α₁ATZ, HTO/Z. The tet-off system of Bujard was used (14). Pulse radiolabeling showed that human α₁ATZ gene expression was absent in the presence of doxycycline at 1 ng/ml, that it could be induced in a time-dependent and concentration-dependent manner by removal of doxycycline from the cell culture fluid, and that, once induced, there was intracellular retention of α₁ATZ (unpublished observations).

Cell lines were subjected to pulse-chase radiolabeling, and samples were analyzed by immunoprecipitation and SDS-PAGE analysis of immunoprecipitates as previously described (33). Results were quantified by scanning of PhosphorImager plates (Storm System; Molecular Dynamics, Sunnyvale, CA) exposed to the radiolabeled gels. Values are reported as means ± SD.

For immunofluorescent staining of vimentin fibers, cells were fixed, permeabilized, and stained with Cy3-conjugated antivimentin antibody exactly as described by Johnston et al. (18). In experiments with proteasome inhibitors, the cells were incubated at 37°C for 18 h in N-acetyl-leu-leu-norleucinol (ALLN) (10 μg/ml for CHO cells, 50 μg/ml for fibroblasts) in normal growth media before fixation. All fluorescent micrographs and photomicrographs were obtained with standard techniques using a Zeiss Axioskop microscope.

Cell-free translation and translocation.

The pGEM-4Z vector (Promega) containing either α₁ATM cDNA or α₁ATZ cDNA was linearized beyond the 3′ end of the cDNA using Hind III. SP6 RNA polymerase was used for in vitro transcription in the presence of m⁷G(5′)ppp(5′)G (Pharmacia, Uppsala, Sweden) to generate ⁷mG-capped mRNAs following the protocol provided by Promega and previously described (24). α₁ATM and α₁ATZ polypeptides were synthesized in the reticulocyte lysate cell-free system according to the protocol provided by Promega. The cell-free reaction mixture (50 μl) contained 35 μl of micrococcal nuclease-treated rabbit reticulocyte lysate and was supplemented with the following final concentrations of additional components: 20 μM of 19-amino-acid mixture minus methionine, 0.8 U/μl of RNase inhibitor RNasin, 0.8 μCi/μl of [³⁵S]methionine, 4 A₂₆₀/ml of canine pancreatic microsomal vesicles, and 20 μg/ml of the appropriate mRNAs. The canine pancreatic microsomal vesicles were prepared by a previously described protocol (25) and kindly provided by Dr. R. Gilmore (Worcester, MA). The cell-free translation and translocation assay was performed for 1 h at 30°C. After the translation reaction, the microsomal vesicles that contained either α₁ATM or α₁ATZ polypeptide were isolated by centrifugation at 15,000 g for 15 min at 4°C. The pelleted microsomal vesicles were resuspended in fresh proteolysis-primed lysate contained in a final volume of 50 μl: 40 mM Tris · HCl, pH 7.5, 5 mM MgCl₂, 2 mM dithiothreitol, 0.5 mM ATP, 10 mM phosphocreatine, and 15 μg creatine phosphokinase (350 U/mg at 25°C; Boehringer Mannheim, Indianapolis, IN) and fresh reticulocyte lysate, followed by incubation at 37°C. After 20 min, the reaction mixture was fixed for EM as described.

EM.

For plastic-embedded EM, cells were grown to 85% confluence in 10-cm dishes and then trypsinized, pelleted, and washed. Cells were then fixed in 1% glutaraldehyde-0.1 M Na-cacodylate and embedded in polybed for ultrathin section transmission EM by standard techniques (26). For immune EM, a previously established protocol with fixation in paraformaldehyde glutaraldehyde and embedding in 10% gelatin for ultrathin sectioning was used (15). Labeling with the primary antibody was carried out for 2 h, and labeling with the appropriate species-specific secondary anti-IgG/Au gold conjugate (Jackson ImmunoResearch, West Grove, PA) was carried out for 1 h. Sections were stained with uranyl acetate and embedded in methyl cellulose. Specimens were viewed and photographed using a Zeiss 902 electron microscope. Where indicated, cells were incubated with 50 μMN-{3-[(2,4-dinitrophenylamino)propyl]}-N-(3-aminopropyl)- methylamine d-hydrochloride (DAMP) at 37°C for 30 min in growth media to label intracellular acidic compartments before fixation for EM (1, 9, 10). For each EM experiment, at least two samples of the cell lines were prepared, immunostained, and examined separately. All double-label immune EM experiments were performed at least twice with each primary antibody. To ensure specificity, separate experiments were also done with several different preparations of each primary antibody, including preparations made in rabbits, goats, and mice. For quantitation of autophagic vacuoles, we examined EM photomicrographs of 25 whole, intact cells containing a nucleus. Quantitative grids of the appropriate size were superimposed on the photomicrographs, and the grid area occupied by nascent (AVi) and degradative (AVd) autophagic vacuoles was compared with the grid area occupied by the cytoplasm.

RESULTS

Structural changes in human fibroblasts engineered for expression and ER retention of α₁ATZ.

First, we used transmission EM to determine whether there were specific morphological changes in human fibroblasts engineered for stable expression of mutant α₁ATZ compared with wild-type α₁ATM. Previous studies of these cell lines have shown that the wild-type molecule is rapidly and completely secreted, whereas the mutant is retained and ultimately degraded in the ER, recapitulating the known defect in α₁AT deficiency (24, 33). In the cell line expressing wild-type α₁ATM (Fig. 1), the ultrastructure was identical to that of the untransfected human fibroblasts, with thin, closely apposed rough ER (rER) cisternae in the perinuclear region interspersed among normal-appearing mitochondria. There were no alterations in any other organelles (data not shown). In the cell line expressing mutant α₁ATZ (Fig. 1) there was markedly dilated rER cisternae filled with granular material. Furthermore, the architecture of the rER cisternae was disrupted and widened by networks of multiple intervening electron-dense multilamellar vacuoles.

The network of electron-dense multilamellar vacuoles that surround and separate ER cisternae in cells expressing α₁ATZ is shown in greater detail in Fig. 2. In the cell line expressing mutant α₁ATZ, viewed under low magnification (Fig. 2 A), a striking network of electron-dense structures is visible in the perinuclear region. Higher magnification of the cell line expressing mutant α₁ATZ (Fig. 2 B) revealed that the electron-dense structures are multilamellar vacuoles that intervene and widen the spaces between rER cisternae. Many of the vacuoles are bound by a double, smooth membrane and surround debris and fragmented membranous structures. In many fields, the smooth surrounding double membranes are contiguous with, or budding from, the nearby rER membranes. This appearance is the hallmark of AVi (9, 10, 23). Other neighboring vacuoles enclose even more electron-dense, lamellar accumulations of membranes, characteristic of the maturation of AVi into AVd. These AVi and AVd also have unique asymmetric and elongated shapes and form complex nests adjacent to and in between rER cisternae (Fig. 2 B). The rER cisternae themselves are markedly dilated and filled with granular material. Dilation of rER cisternae, disruption of the rER architecture, and intense accumulation of networks of multiple intervening electron-dense multilamellar vacuoles were seen in cell lines from three PIZZ α₁AT-deficient patients with liver disease (susceptible hosts) and from two PIZZ α₁AT-deficient patients without liver disease (protected hosts; data not shown).

This marked alteration of the ER and the intense accumulation of autophagic vacuoles was specific for α₁ATZ, as shown by the ultrastructural characteristics of the cell line expressing α₁ATM (Fig. 2, C–E). At low magnification (Fig. 2 C) there is no evidence for perinuclear accumulation of autophagosomes. At higher magnification, there are thin, normal-appearing rER cisternae and normal mitochondria near the nucleus (Fig. 2 D). Only a few spherical AVi and AVd can be found in the cytoplasm (Fig. 2 E). These AVi and AVd are not elongated, do not form nests, and are typical for cultured cells in general. Quantitative morphometry showed that AVi and AVd occupied 4 ± 2.5% of the cytoplasm in cells expressing wild-type α₁ATM compared with 17.5 ± 4.5% in cells expressing mutant α₁ATZ. There was no difference in percent cytoplasm occupied by autophagic vacuoles in the nontransduced parent fibroblast cell line or the fibroblast cell line transduced with the expression vector alone compared with the cell line expressing α₁ATM (data not shown).

Characterization of the electron-dense vacuoles in human fibroblast cell lines engineered for expression and ER retention of mutant α₁ATZ.

To provide further evidence that the electron-dense vacuoles were in fact autophagic, three approaches were used. First, we subjected fibroblasts expressing wild-type α₁ATM or mutant α₁ATZ to intravital staining with MDC, a fluorescent reagent known to specifically label autophagic vacuoles in vivo (3). The resulting fluorescent photomicrographs demonstrate intense staining of vacuolar structures in the center of the cytoplasm of the cells expressing the mutant protein (Fig.3 B). The staining is significantly increased in intensity compared with cells expressing the wild-type protein (Fig. 3 A). Moreover, the staining corresponds in location with the electron-dense vacuoles observed by EM in close proximity to, but distinct from, the ER. This is shown more clearly by comparing the MDC staining with that of ETB, a fluorescent vital dye that selectively stains ER membranes in living cells (32) (Fig. 3, C and D). ETB stains structures that are closer to the nucleus but overlapping with the structures stained by MDC. There is a marked increase in ETB staining of the cells expressing mutant α₁ATZ (Fig. 3 D) than in cells expressing α₁ATM (Fig. 3 C), corresponding to the expansion and dilation of the ER membrane observed at the ultrastructural level.

Second, we examined whether the structures stained by MDC were also stained by the fluorescent vital dye LTR, which accumulates and labels acidic intracellular compartments in living cells (Fig.4). Acidification is known to occur early in the formation of autophagic vacuoles (1, 9, 10). The results show that intense staining of vacuolar structures in the cytoplasm of cells expressing mutant α₁ATZ (Fig.4 D) is significantly increased over that in cells expressing α₁ATM (Fig. 4 A). This staining overlaps in location with that of rER membranes by ETB (Fig. 4, B andE), and dual-labeled images (Fig. 4, C andF) show that these acidic vacuoles are closely apposed to the ER.

Third, we examined the possibility that the electron-dense vacuoles were acidified by immune EM for DAMP (Fig.5, A and B). DAMP accumulates in acidic vesicles and can be immunolabeled with anti-dinitrophenyl antibodies and therefore can identify AVi and AVd, which acidify early in their biogenesis (1). In Fig. 5,A and B, there is intense accumulation of gold beads in double membrane-bound AVi containing debris. Elongated nests of multiple, DAMP-labeled AVd also containing electron-dense debris and lamellar arrangements of membranes are also identified (Fig.5 B). Because the vacuoles form a complex nest around the ER it sometimes looks like the beads labeling DAMP are outside the vacuoles, but in every case, when examined in another field of view, these beads were in fact in adjacent vacuoles. There was an occasional gold bead found free in the cytoplasm and in other nonacidic organelles, probably representing background staining (data not shown). These results indicate that the electron-dense vacuoles found in cells expressing the mutant α₁ATZ had the known acidic properties of autophagic vesicles and that they could be specifically labeled by immune EM for DAMP.

Next, we used immune EM to examine the possibility that the autophagic vesicles contained mutant α₁ATZ. The results show immunogold labeling of dilated ER membranes (Fig. 5 C) and tangled nests of AVi and AVd (Fig. 5 D). Beads were only rarely present in the nucleus, mitochondria, or other structures (data not shown). The same cells were then double labeled for DAMP (small, 12 nm beads) and α₁AT (large, 18 nm beads) to determine whether α₁ATZ is colocalized to autophagosomes. The result is illustrated by a high-magnification view within a cluster of autophagosomes that have engulfed multilamellar debris (Fig.5 E). These structures are labeled by both DAMP and α₁AT. As a control (Fig. 5 F), examination of cells transduced with the expression vector alone reveals only rare, round DAMP-positive AVi and acidified simple vacuoles but no complex nests of autophagosomes and no α₁AT labeling. These results demonstrate that the appearance of nests of DAMP-labeled autophagosomes containing α₁AT is specific for cells that express and retain the mutant α₁ATZ.

Next we examined the possibility that calnexin was also present in autophagosomes by double labeling with antibodies to calnexin (small, 12 nm beads) and α₁AT (large, 18 nm beads). Previous studies have suggested that calnexin plays a critical role in the ER degradation of α₁ATZ and, moreover, that it is the α₁ATZ-calnexin complex that is attacked by the ER degradation pathway (24, 30). The result shows colocalization of calnexin and α₁ATZ in ER membranes (Fig. 5 G) and in AVi and AVd (Fig. 5 H). Many individual AVi and AVd had both small and large beads, indicating colocalization of calnexin and α₁ATZ.

To exclude the possibility that the autophagic response to ER retention of α₁ATZ is peculiar to fibroblasts, we used EM to examine model lines derived from several other cell types. First, we examined the mouse hepatoma cell line engineered for expression of α₁ATZ Hepa1–6N2Z9 (Fig.6). In cells from the parent untransfected Hepa1–6 cells, the nucleus is surrounded by normal cytoplasm and a few electron-dense structures representing simple lysosomes and an occasional multilamellar autophagic vacuole (Fig.6 A). In contrast, Fig. 6 B shows a similar region from the Hepa1–6N2Z9 cell line, engineered for expression of α₁ATZ; there are many electron-dense, multilamellar autophagic vacuoles. The typical multilamellar structure and close proximity to rER is particularly evident under higher magnification of one of the vacuoles in Fig. 6 C. There is also an increased number of autophagic vacuoles observed in the rat hepatoma cell line engineered for expression of α₁ATZ, H11N2Z1 (data not shown).

Second, we examined by EM a HeLa cell line engineered for inducible expression of α₁ATZ (Fig.7). When α₁ATZ expression is suppressed in the presence of doxycycline (Fig. 7 A), a few electron-dense structures are visible near the nucleus. In the absence of doxycycline for a period of time associated with induction of α₁ATZ expression and ER retention in Fig.7 B, there is a significantly increased focal, perinuclear accumulation of electron-dense, multilamellar vacuoles to the right of the nucleus. This intense, focal area of autophagic activity is shown under higher magnification in Fig. 7 C. Examination by fluorescence microscopy revealed that this perinuclear nest of vacuoles stained positively for acidity by LTR and stained positively by the autophagic marker MDC (data not shown). Together, these data provide evidence that vacuoles with the structural characteristics of autophagosomes are associated with expression of α₁ATZ in several cell types and include cells of hepatocyte lineage and that autophagosomes are specifically induced by expression of α₁ATZ in the inducible cell line.

Effects of chemical inhibitors of autophagy on the fate of mutant α₁ATZ.

To investigate the possible role of autophagy in the degradation of α₁ATZ, we examined the effect of 3MA, a chemical inhibitor of autophagy (28), on the fate of α₁ATZ in pulse-chase radiolabeling experiments (Fig.8 A). The results show that, in the absence of 3MA, α₁ATZ is synthesized as a 52-kDa polypeptide at time 0 and is retained for ∼1 h of the chase period. This polypeptide then progressively disappears between 2 and 6 h of the chase period. Only a trace amount of α₁ATZ is secreted into the extracellular fluid (data not shown). This result is consistent with our previous studies showing retention and degradation of α₁ATZ in the ER as a 52-kDa intermediate with high-mannose-type oligosaccharide side chains (24, 33). In the presence of 3MA, the α₁ATZ is also initially synthesized as a 52-kDa precursor polypeptide. However, the rate of disappearance is reduced. There is a greater amount of α₁ATZ remaining at 3 h and at each subsequent time point of the chase period. There is no increase or decrease in the trace amount of α₁AT secreted into the extracellular fluid (data not shown). Three identical experiments were used for quantification of the kinetics of disappearance by phosphorimaging analysis, as shown in Fig. 8 B. The results show a decrease in rate of degradation beginning at 3 h and particularly apparent at later time points. Experiments with wortmannin and LY-294002, two other chemical inhibitors of autophagy (4), had similar results, with a decrease in rate of degradation of α₁ATZ especially apparent after 3 h of the chase period (data not shown). Together with the observation that α₁ATZ can be detected in autophagic vacuoles by immune EM, these data provide evidence that autophagy plays a role in ER degradation of misfolded α₁ATZ molecule.

Morphological evidence for autophagic vacuoles in the liver of PiZ transgenic mice.

Misfolded α₁ATZ molecules are retained in the ER of liver cells in the PiZ mouse model transgenic for the human α₁ATZ gene (7). Here we examined the livers of four PiZ mice and four wild-type mice of the same genetic background by EM for the presence within hepatocytes of multilamellar, electron-dense structures indicative of autophagic vacuoles (Fig.9). The results showed that in some hepatocytes there were nearly normal areas of cytoplasm with normal-appearing rER and other organelles (Fig. 9 A). However, in most hepatocytes there were areas of dilated rER membranes filled with granular deposits as previously described (data not shown), as well as focal nests of electron-dense, multilamellar vacuoles (Fig.9 B). When these vacuoles were examined under higher magnification (Fig. 9 C), they were clearly multilamellar structures containing electron-dense debris that were closely apposed to rER membranes. Interestingly, nests of these vacuoles were most commonly found within hepatocytes containing dilated rER membranes. Examination of the livers from the wild-type mice revealed occasional multilamellar vacuoles within the cytoplasm of hepatocytes, but they were much fewer in number than seen in the PiZ mice and were not localized around the rER (data not shown).

Morphological changes in liver tissue from patients with α₁AT deficiency.

It is well known from many clinical studies that the ER in liver cells of α₁AT-deficient patients is dilated by granular α₁AT deposits. Early clinical studies also noted areas of ribosome-free ER membrane and surrounding vacuoles (12, 16,34), but it is not entirely clear from these studies whether these vacuoles were truly autophagic, whether the autophagic response was prominent, and whether it was only present in liver cells with dilated ER. Here we examined by transmission EM liver tissue from three patients with liver disease caused by α₁AT deficiency (Fig. 10) to determine whether an autophagic response is also present in vivo. The results showed normal-appearing rER, mitochondria, Golgi, nucleus, and other structures in some cells (Fig. 10 A). However, in most cells there were areas of markedly dilated rER membranes filled with granular, proteinaceous material (Fig. 10 B). In many of these cells, multilamellar autophagic vacuoles budding from, and still contiguous with, ER membranes were observed (Fig. 10 C). Numerous, fully formed, double membrane-bound AVi containing debris could be seen that had freshly budded from nearby ER membranes (Fig.10 D). In other areas, nests of multilamellar, electron-dense structures with the appearance of AVd could be easily identified (Fig.10 E). These autophagic structures are remarkably similar to those seen in fibroblast cell lines that express α₁ATZ. Moreover, autophagic vacuoles were almost always seen in the cells that had dilated ER cisternae. This type of intense autophagic response was seen in the livers of all 3 α₁AT-deficient patients but not in the liver from the normal individual (data not shown). The autophagic response in the α₁AT-deficient liver could not be attributed to the method by which the liver tissue was stored or processed because each of these specimens was stored and/or processed differently. Moreover, one α₁AT-deficient liver and one normal liver were processed immediately after harvest by an identical protocol. Thus the results indicate that autophagosomes are also induced by α₁ATZ in vivo.

Morphological changes in microsomes that specifically degrade mutant α₁ATZ in a cell-free system.

Our previous studies have shown that α₁ATZ is specifically degraded in a cell-free microsomal translocation system and that the biochemical characteristics of its degradation in the cell-free system recapitulate those that occur in intact cells (24). Here we examined the possibility that degradation of α₁ATZ in this cell-free system was associated with morphological changes in the microsomal vesicles (Fig.11). The vesicles were first subjected to cell-free translation/translocation reaction for 1 h at 30°C. These conditions were associated with translocation of wild-type α₁ATM and mutant α₁ATZ in similar amounts (data not shown) (24, 30). Moreover, similar amounts of wild-type α₁ATM and mutant α₁ATZ were protected from protease digestion under these conditions (24). The vesicles were then pelleted, resuspended in proteolysis-primed reticulocyte lysate, and incubated at 37°C for 20 min. By this time, mutant α₁ATZ, but not wild-type α₁ATM, had begun to undergo degradation (data not shown) (24, 30). The results show that native microsomal vesicles (Fig. 11 A) and microsomal vesicles that had translocated wild-type α₁ATM (Fig. 11 B) contained intact round vesicular structures studded with ribosomes and surrounded by proteinaceous reticulocyte lysate. Microsomes that had translocated α₁ATZ were only rarely round and spherical. In almost every field, there was budding and elongation of these microsomes (Fig.11, C–F). Some of the microsomal membrane was devoid of ribosomes, and there was formation of many nests of ribosome-free membrane ghosts (Fig. 11 E). The budding, elongation, loss of ribosomes, and nesting of vesicles were rarely, if ever, seen in the controls. These data indicate that there are indeed morphological changes in microsomal vesicles associated specifically with the translocation and degradation of α₁ATZ in vitro.

Expression and ER retention of α₁ATZ does not induce aggresome formation.

A recent study by Johnston et al. (18) has shown that when mutant CFTRΔF508 or PS1 A246E molecules accumulate in the ER of transfected CHO cells there is formation of non-membrane-bound cagelike structures adjacent to the pericentriolar region of the nucleus, called aggresomes (18). Aggresome formation was induced by expression of the mutant proteins at high levels or by expression of the mutant proteins at lower levels but in the presence of inhibitors of their degradation by the proteasome. Aggresome formation also occurred at a low rate in cells that did not express a mutant protein but were treated with chemical proteasomal inhibitors. Here we examined the possibility that expression of α₁ATZ was also associated with aggresome formation.

Human fibroblast cell lines transduced to express moderately high levels of wild-type α₁ATM, mutant α₁ATZ, or mutant CFTRΔF508 were incubated in the absence or presence of proteasomal inhibitor ALLN and then fixed and stained with fluorescent antibodies to vimentin (Fig. 12). In the absence of proteasome blockade by ALLN, all of the cells maintained a spindle or fan-shaped typical fibroblast morphology, with networks of vimentin fibers visible throughout the periphery of the cells (Fig. 12,A–C). In the presence of ALLN, there was no collapse of peripheral vimentin fibers in the majority (∼65%) of cells expressing wild-type α₁ATM or mutant α₁ATZ (Fig. 12, D and E). The remaining 35% of the cells had variable levels of vimentin fiber collapse, with a few cells demonstrating a compact aggresome. There was no difference between cells expressing mutant α₁ATZ, wild-type α₁ATM, and the nontransduced parent fibroblast cell lines (data not shown). However, in the fibroblast cell lines expressing CFTRΔF508 and incubated with ALLN (Fig. 12 F), the majority of the cells (70%) demonstrated complete collapse of vimentin fibers from the periphery of the cell into a basketlike, perinuclear structure indicative of aggresome formation.

A similar analysis was also performed in CHO cells transduced to express moderately high levels of α₁ATZ or CFTRΔF508. In the absence of ALLN, <1% of CHO cells expressing α₁ATZ demonstrated the vimentin fiber collapse typical of aggresomes (Fig. 13). This was similar to the rate of aggresome formation observed in nontransduced parent CHO cells and in CHO cells transduced with wild-type α₁ATM (data not shown). In the absence of ALLN, ∼5% of cells expressing CFTRΔF508 spontaneously formed an aggresome with a perinuclear, basketlike structure of collapsed vimentin fibers (Fig. 13). In the presence of ALLN, ∼50% of the CHO cells expressing mutant α₁ATZ demonstrated the collapse of vimentin fibers into aggresomes (Fig. 10), which is identical to the rate of aggresome formation for nontransduced parent CHO cells (data not shown). However, in the presence of ALLN, nearly 100% of the CFTRΔF508-expressing CHO cells demonstrated the collapse of cytoplasmic vimentin fibers into tightly packed, perinuclear aggresomes (Fig. 13). Examination of these cell lines by EM and fluorescent microscopy after intravital staining with MDC showed a marked autophagic response only in the cell line expressing mutant α₁ATZ (data not shown). These results indicate that the accumulation of α₁ATZ does not induce the formation of aggresomes and suggest that the morphological changes that characterize the response of cells to the accumulation of mutant proteins in the ER are, at least in part, substrate specific.

DISCUSSION

These results indicate that retention of mutant α₁ATZ in the ER is associated with a marked autophagic response. There is mention of electron-dense vacuoles within the expanded ER-Golgi intermediate compartment of thymic epithelial cells that accumulate unassembled MHC class I molecules (26) and of double-membrane vesicular structures in the immediate vicinity of aggresomes in cells that accumulate CFTRΔF508, suggesting an autophagic response in these cases (18). However, it is not clear at this time whether the autophagic response in those cases is as intense or as generalized as the one seen here in cells in which there is ER retention of α₁ATZ. There was a marked difference in the degree of autophagy seen here in human fibroblasts transduced with the CFTRΔF508 gene compared with those transduced with the α₁ATZ gene.

Even though α₁ATZ and CFTRΔF508 are both mutant proteins that are degraded in the ER, there are differences in the properties of the two proteins that could explain the differences in cellular responses. CFTRΔF508 is a membrane protein with multiple membrane-spanning domains expressed at lower concentrations than α₁ATZ. The proteasome is involved in degradation of both proteins, but there is evidence that other mechanisms play a contributory role or, at least, that the proteasomal mechanism cannot fully account for degradation of either protein (17, 30). Because CFTRΔF508 is degraded more rapidly than α₁ATZ, it is likely that there are differences in how CFTRΔF508 and α₁ATZ reach the proteasomal machinery in the cytoplasm or differences in the nonproteasomal mechanisms for degradation. It is also possible that differences in the absolute concentration of mutant protein, the duration of time it has accumulated at a certain concentration, and the rate or mechanism by which it is extruded into the cytoplasm account for the differences in morphology and response. Moreover, it is possible that intrinsic properties of the mutant protein determine whether the autophagic or aggresomal responses are invoked; i.e., aggregated α₁ATZ molecules are so toxic when free in the cytoplasm that cells only survive when capable of engulfing them within a membranous subcompartment. This is a particularly important issue for α₁AT deficiency because a subgroup of individuals affected by this deficiency develop severe liver injury that is thought to be caused by the hepatotoxic effects of the retained α₁ATZ molecule. There are now many other naturally occurring human deficiency syndromes in which abnormal proteins are retained in the ER (22). In some of these cases, misfolded secretory proteins such as mutant fibrinogen, coagulation, or complement proteins are retained in the ER of liver cells without apparent hepatotoxic effects, implying that there is something intrinsically different about the retention of α₁ATZ, or the response to it, that is associated with hepatotoxicity and carcinogenesis.

Although the results of this study provide evidence that autophagy itself contributes to the ER degradation pathway for α₁ATZ, it is still difficult to ascertain the relative significance of its contribution. Previous studies have shown that there is a ubiquitin-dependent proteasomal mechanism that targets the α₁ATZ-calnexin complex for degradation as well as a ubiquitin-independent proteasomal mechanism and nonproteasomal mechanism, or mechanisms, for degradation of α₁ATZ retained in the ER (24, 30). The results reported here show that α₁ATZ and calnexin molecules are present in autophagic vacuoles. Alone, these data do not prove that α₁ATZ is degraded in autophagosomes and certainly do not address the relative significance of autophagy in degradation of α₁ATZ. In fact, because there is substantial evidence that autophagosomes form at least in part from ER (9, 10), it is possible that a certain number of α₁ATZ molecules that are retained in the ER and calnexin molecules that are integral to the ER are nonspecifically carried into the autophagosomes. The results of the current study do show that degradation of α₁ATZ is partially abrogated by 3MA, wortmannin, and LY-294002. However, the effect of these chemical inhibitors of autophagy was significantly lower in magnitude than that of the proteasomal inhibitors lactacystin and MG132 previously reported (24, 30). Neither lactacystin nor MG132 completely inhibit degradation of α₁ATZ, and the presumed nonproteasomal component of ER α₁ATZ degradation is particularly apparent at later time points (24, 30). The inhibitory effects of 3MA, wortmannin, and LY-294002 in this study were, in fact, observed at later time points, raising the possibility that proteasomal and autophagic pathways constitute independent mechanisms for ER degradation of α₁ATZ, perhaps acting at different stages and/or on different pools of retained α₁ATZ molecules. Nevertheless, because it is difficult to know the relative inhibitory efficacy in our current model cell culture systems of 3MA, wortmannin, or LY-294002 on autophagy compared with that of lactacystin on proteasomal activity, it is not yet possible to determine whether autophagy plays a relatively minor role in the overall degradation of α₁ATZ retained in the ER. Sophisticated genetic techniques for abrogating autophagy may be required to more definitively address this issue in the future.

In addition to playing a protective role by contributing to the degradation of mutant α₁ATZ molecules, the autophagic response may represent a mechanism for preventing carcinogenesis. Recent studies have suggested that autophagic activity/proteins are decreased in tumors and that reconstitution of autophagic activity inhibits tumorigenesis in vivo (19, 21). In this regard, it is of some interest that autophagic vacuoles were most commonly seen in liver cells with dilated ER in the α₁AT-deficient patient. Moreover, in one genetically engineered mouse model of α₁AT deficiency, hepatocarcinogenesis appeared to evolve in nodular aggregates of hepatocytes that were negative for α₁AT expression by immunohistochemical staining (13).

A striking alteration in morphology was also induced by translocation and degradation of the α₁ATZ polypeptide in isolated microsomal vesicles in vitro. Our previous studies have shown that α₁ATZ is rapidly and specifically degraded after it is translocated into microsomal vesicles in vitro and that the biochemical characteristics of its degradation in vitro are very similar to those that occur in intact cells (24). Moreover, the efficiency of translocation of α₁ATZ is identical to that of wild-type α₁ATM, which undergoes two distinct endoproteolytic cleavages but is not degraded over many hours in isolated microsomal vesicles in vitro. Electron microscopic analysis of vesicles from these microsomal translocation reactions shows that translocation and degradation of α₁ATZ is specifically accompanied by a marked expansion and distortion of the vesicles that in some ways resembles the alterations of ER seen in intact cells and in human liver. There were even areas of ribosome-free microsomal membranes and invaginations that clearly resemble the initial stages of autophagosome formation in vivo. These results suggest that at least some of the profound morphological alterations that accompany accumulation of the mutant secretory protein α₁ATZ in the ER are intrinsic to the ER or the combination of ER membrane vesicles and proteolysis-primed reticulocyte lysate.

FOOTNOTES

• These studies were supported in part by National Institutes of Health grants HL-37784, DK-52526, P01-DK-56783, P30-DK-52574, and DK-02379 (J. H. Teckman).

• Address for reprint requests and other correspondence: J. Teckman, Dept. of Pediatrics, Washington Univ. School of Medicine at St. Louis Children's Hospital, 1 Children's Place, St. Louis, MO 63110 (E-mail: teckman@kids.wustl.edu).

• The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.