Isolation and Characterization of Rat Liver Amphisomes

Amphisomes, the autophagic vacuoles (AVs) formed upon fusion between autophagosomes and endosomes, have so far only been characterized in indirect, functional terms. To enable a physical distinction between autophagosomes and amphisomes, the latter were selectively density-shifted in sucrose gradients following fusion with AOM-gold-loaded endosomes (endosomes made dense by asialoorosomucoid-conjugated gold particles, endocytosed by isolated rat hepatocytes prior to subcellular fractionation). Whereas amphisomes, by this criterion, accounted for only a minor fraction of the AVs in control hepatocytes, treatment of the cells with leupeptin (an inhibitor of lysosomal protein degradation) caused an accumulation of amphisomes to about one-half of the AV population. A quantitative electron microscopic study confirmed that leupeptin induced a severalfold increase in the number of hepatocytic amphisomes (recognized by their gold particle contents; otherwise, their ultrastructure was quite similar to autophagosomes). Leupeptin caused, furthermore, a selective retention of endocytosed AOM-gold in the amphisomes at the expense of the lysosomes, consistent with an inhibition of amphisome-lysosome fusion. The electron micrographs suggested that autophagosomes could undergo multiple independent fusions, with multivesicular (late) endosomes to form amphisomes and with small lysosomes to form large autolysosomes. A biochemical comparison between autophagosomes and amphisomes, purified by a novel procedure, showed that the amphisomes were enriched in early endosome markers (the asialoglycoprotein receptor and the early endosome-associated protein 1) as well as in a late endosome marker (the cation-independent mannose 6-phosphate receptor). Amphisomes would thus seem to be capable of receiving inputs both from early and late endosomes.

Amphisomes, the autophagic vacuoles (AVs) formed upon fusion between autophagosomes and endosomes, have so far only been characterized in indirect, functional terms. To enable a physical distinction between autophagosomes and amphisomes, the latter were selectively density-shifted in sucrose gradients following fusion with AOM-gold-loaded endosomes (endosomes made dense by asialoorosomucoid-conjugated gold particles, endocytosed by isolated rat hepatocytes prior to subcellular fractionation). Whereas amphisomes, by this criterion, accounted for only a minor fraction of the AVs in control hepatocytes, treatment of the cells with leupeptin (an inhibitor of lysosomal protein degradation) caused an accumulation of amphisomes to about one-half of the AV population. A quantitative electron microscopic study confirmed that leupeptin induced a severalfold increase in the number of hepatocytic amphisomes (recognized by their gold particle contents; otherwise, their ultrastructure was quite similar to autophagosomes). Leupeptin caused, furthermore, a selective retention of endocytosed AOM-gold in the amphisomes at the expense of the lysosomes, consistent with an inhibition of amphisome-lysosome fusion. The electron micrographs suggested that autophagosomes could undergo multiple independent fusions, with multivesicular (late) endosomes to form amphisomes and with small lysosomes to form large autolysosomes. A biochemical comparison between autophagosomes and amphisomes, purified by a novel procedure, showed that the amphisomes were enriched in early endosome markers (the asialoglycoprotein receptor and the early endosome-associated protein 1) as well as in a late endosome marker (the cation-independent mannose 6-phosphate receptor). Amphisomes would thus seem to be capable of receiving inputs both from early and late endosomes.
Autophagy is a mechanism by which cells sequester and degrade parts of their own cytoplasm, including organelles (for a review, see Ref. 1). Autophagy is basically a nonselective, bulk process; a number of cytosolic enzymes with widely different half-lives have been shown to be autophagically sequestered at identical rates (2). However, a selective organelle autophagy can apparently take place during the regression of hypertrophied organelles like the peroxisomes (3) and, possi-bly, the smooth endoplasmic reticulum (4). In the first recognizable autophagic step, single or multiple membrane cisternae with a distinct osmiophilic morphology (5)(6)(7), called phagophores (8,9), wrap up a region of cytoplasm into a closed vacuolar organelle, an autophagosome. The phagophores, which form the autophagosome wall, may ultimately be derived from the endoplasmic reticulum (10 -13), perhaps representing highly modified endoplasmic reticulum cisternae that have acquired unique biochemical and structural properties (5,(13)(14)(15)(16).
There is good evidence that some of the autophagosomes may fuse directly with lysosomes (6,12,19,21,24,25). It has been suggested that autophagic delivery to the lysosomes may occur exclusively by autophagosome-lysosome fusion, i.e. without amphisome formation (24,25), but a recent morphometric study concluded that hepatocytic endosomes were 5-6 times more likely to enter the amphisome pathway than to fuse directly with lysosomes (21).
Differences in experimental approach, organelle markers, and organelle definitions make it difficult to resolve discrepant conclusions regarding the relative roles of the amphisomal and the direct autophagosomal pathway to the lysosome. Active lysosomes can be distinguished from the double-or multiplemembrane-enclosed autophagosomes by having only a single delimiting membrane, but amphisomes have not yet been unequivocally characterized in morphological terms. We therefore initiated the present study, using colloidal gold particles conjugated to asialoorosomucoid (AOM) both as an ultrastructural marker and as a density perturbant. AOM-gold is taken up by receptor-mediated endocytosis in isolated rat hepatocytes, and by means of different gold particle sizes and different loading times, a differential staining of lysosomes and prelysosomal endocytic vacuoles (which would include the amphisomes) can be obtained (26,27). In addition, the high density of the gold particles would be expected to change the behavior of goldladen endocytic vacuoles in density gradients, as has been shown in other cell types (28). By this approach, we have been able to identify the amphisomes both as a structural entities in sucrose density gradients and as distinct vacuoles in electron micrographs.

EXPERIMENTAL PROCEDURES
Biochemicals-Tyramine-cellobiose (TC) was a gift from professor Helge Tolleshaug (Nycomed Pharma A/S, Oslo, Norway). Na 125 I was from Amersham Pharmacia, Biotech (Little Chalfont, UK). 125 I-TC-AOM was synthesized as described previously (29), mixed with 20 times as much unlabeled AOM, and added to cell suspensions at a final concentration of 200 nM, i.e. 10 g/ml (specific activity, 20,000 dpm/g of protein). Leupeptin was purchased from Protein Research Foundation (Osaka, Japan); 3-methyladenine was from Fluka A.G. (Buchs, Switzerland); Metrizamide, Nycodenz, and iodixanol were from Nycomed Pharma A/S; and Percoll was from Amersham Pharmacia Biotech (Uppsala, Sweden). Kits for automated analysis of lactate dehydrogenase (LDH) and acid phosphatase were obtained from Boehringer (Mannheim, Germany). Antibodies against superoxide dismutase, endosome-associated protein 1, lysosomal glycoprotein Lgp120, and cathepsin B were kind gifts from Dr. Ling-Yi Chang (Duke University Medical Center, Durham, NC), Dr. Harald Stenmark (The Norwegian Radium Hospital, Oslo, Norway); Dr. William Dunn (University of Florida, Gainesville, FL), and Dr. David Buttle (University of Sheffield Medical School, Sheffield, UK), respectively. Antibodies against the asialoglycoprotein receptor (ASGPR), and the cation-independent mannose 6-phosphate receptor (MPR) were from Dr. Paul Weigel (University of Oklahoma Health Sciences Center, Oklahoma City, OK) and Dr. William Brown (Cornell University, Ithaca, NY), respectively. Other chemicals and biochemicals, including gold chloride (HAuCl 4 ), vinblastine, and collagenase, were purchased from Sigma.
Preparation and Uptake of AOM-gold-AOM-gold, i.e. AOM conjugated to colloidal gold particles (3-or 10-nm diameter), was prepared according to established procedures (31,32). Colloidal gold was made by mixing, at 60°C, 160 ml of 0.0125% gold chloride and 40 ml of tannic acid (0.125 or 0.0045%, for preparation of 3-and 10-nm gold particles, respectively) buffered with 0.2% trinatrium citrate dihydrate, keeping the mixture at 60°C until a stable wine-red color had developed (instantly with 3-nm gold; 1 h with 10-nm gold). The volume was then reduced to one-half by boiling/evaporation (yielding, for 10-nm particles, a 520-nm absorbance of approximately 1.6, corresponding to 1.3⅐10 13 gold particles/ml), and AOM or 125 I-TC-AOM was thoroughly mixed into the suspension (10 min at room temperature) at a final concentration of 10 g/ml. Bovine serum albumin (0.01%) was added as a stabilizer, and the AOM-coated gold particles were sedimented (45 min at 47,000 ϫ g for 10-nm particles; 2 h at 47,000 ϫ g for 3-nm particles) with a 95% recovery of gold and a 70% recovery of AOM, corresponding to about 3 molecules of AOM/gold particle. The pellet was resuspended with 0.9% NaCl to 1 ml and a 520-nm absorbance of 150 -200 (AOM, 600 -800 g/ml, or 14 -18 M) and dialyzed against 0.9% NaCl overnight.
Uptake of 125 I-TC-AOM or 10-nm 125 I-TC-AOM-gold was measured at "concentrations" of up to 200 -400 nM AOM (520-nm absorbance of 2-4), referring to either free or gold-bound AOM. After incubation, cells were washed three times at 0°C in 10% sucrose, and radioactivity in the cell pellets was measured by ␥-counting. Intracellular AOM or AOM-gold "concentrations" were calculated on the basis of the measured wet mass of cell pellets, assuming the intracellular fluid volume to be 50% of the wet mass (33). For density shift experiments, high concentrations of AOM-gold were used (520-nm absorbance of about 20, i.e. about 2 M AOM).
Cell Disruption-After incubation, the hepatocytes in each tube were washed twice in 10% (w/v) sucrose and finally suspended in 0.5 ml of 10% sucrose. Four samples were pooled and subjected to electrodisruption by a single high voltage pulse (34). The resulting preparation is referred to as the disruptate.
Subcellular Fractionation-The disruptate was diluted with an equal volume of buffered sucrose (0.25 M sucrose, 10 mM HEPES, 1 mM EDTA, pH 7.3) and homogenized by 10 strokes in a Dounce homogenizer with a tightly fitting pestle. The homogenate was centrifuged for 2 min at 4,000 rpm (2,000 ϫ g) in a Sorvall SS-34 rotor and washed once (resuspended in 4 ml of buffered sucrose and recentrifuged), and the combined postnuclear fractions were layered on top of a 25-ml metrizamide density cushion (8% (w/v) metrizamide, 1 mM dithiothreitol, 1 mM EDTA, 50 mM potassium phosphate, pH 7.5; adjusted to 300 mosM with sucrose) and centrifuged at 50,000 ϫ g for 1 h. The supernatant was removed by aspiration; the pellet (large granule fraction) was resuspended in 2 ml of buffered sucrose and homogenized (5 strokes); and 1.5 ml was layered on the top of a linear sucrose gradient. The sucrose gradients were prepared by mixing 63 and 15% sucrose solutions (5.5 ml of each) using a Biocomp Gradient Master (Nycomed Pharma A/S). The gradients were centrifuged at 285,000 ϫ g in a Beckman SW40 rotor at 4°C for 105 min. The gradients were then divided into 20 0.67-ml fractions by upward displacement with Maxidens fluid (Nycomed Pharma A/S). The densities of the fractions were calculated from the refractive indices. Homogenizations and centrifugations were carried out at 0 -4°C.
Purification of Autophagosomes, Amphisomes, and Lysosomes-We have recently developed methods for the purification of hepatocytic autophagosomes and lysosomes, described in detail elsewhere. 2 In the present study, certain modifications were introduced in order to enrich the autophagosome preparation with respect to amphisomes. Hepatocyte homogenates were prepared in buffered sucrose as described above. For purification of autophagosomes (with only 5% amphisome contamination), homogenates from vinblastine-treated cells (hepatocytes incubated for 3 h at 37°C with 50 M vinblastine) were incubated with 0.5 mM glycyl-L-phenylalanine 2-naphthylamide (GPN) and 0.33% Me 2 SO (GPN solvent) for 6 min at 37°C to disrupt the lysosomes (35). For preparation of an approximately 50:50 mixture of amphisomes and autophagosomes, homogenates were prepared from leupeptin-treated cells (3 h at 37°C with 0.3 mM leupeptin), and the GPN step was omitted (the enlarged lysosomes disappeared anyway, mostly in the nuclear fraction). For preparation of lysosomes, homogenates were prepared from cells in which autophagy had been blocked by incubation for 3 h at 37°C with 10 mM 3-methyladenine (3MA); moreover, the GPN step was omitted. In either case, the homogenate was cooled to 4°C, and further purification was performed at this temperature. The nuclei were sedimented by centrifugation for 2 min at 4,000 rpm (2,000 ϫ g) and washed once, and 14 ml of the combined postnuclear supernatants were placed on top of a discontinuous (two-step), isotonic Nycodenz gradient, containing 9.5% Nycodenz (in buffered sucrose diluted to isotonicity) in the 17-ml upper layer and 22.5% Nycodenz in the 7-ml bottom layer. After centrifugation for 1 h at 28,000 rpm (141,000 ϫ g) in a Beckman SW28 rotor to remove mitochondria and peroxisomes, the vacuole-enriched interface band between the Nycodenz layers (5 ml) was diluted with an equal volume of buffered sucrose and layered on top of a Percoll two-step gradient (21 ml of 33% Percoll in buffered sucrose in the upper layer; 7 ml of 22.5% Nycodenz in buffered sucrose in the lower layer) and centrifuged for 30 min at 20,000 rpm (72,000 ϫ g) to remove the endoplasmic reticulum. The autophagic vacuoles were recovered as a 5-ml fraction from the lower part of the Percoll gradient (near the Nycodenz interface) and diluted with 3.5 ml of isotonic 60% iodixanol, overlaid with 1.5 ml of 30% iodixanol and 2.5 ml of buffered sucrose, and centrifuged for 30 min in a Beckman SW40 rotor at 20,000 rpm (71,000 ϫ g) to sediment the Percoll particles. The purified vacuoles (autophagosomes, autophagosomes plus amphisomes, or lysosomes, depending on the pretreatment of the cells or the homogenate) floated to the iodixanol/sucrose interface, where they could be recovered for morphological or biochemical analysis.
Measurements of Radioactivity and Enzyme Activities-Acid-soluble and acid-insoluble 125 I radioactivity was measured by ␥-counting of gradient fractions after precipitation with 10% trichloroacetic acid (27). Acid phosphatase (␤-glycerophosphatase) was determined according to Ames (36), and LDH was determined according to Bergmeyer (37), using a Technicon RA-1000 autoanalyzer for both assays. In some particularly important experiments, the gradient resolution of LDH was improved by subtracting, from each fraction, the background of nonautophagocytosed LDH, measured in separate gradients as enzyme activity resistant to treatment of the cells with 10 mM of the autophagy inhibitor 3MA (19). Autophagic activity was measured in intact cells as the sequestration of cytosolic LDH into the sedimentable cell fraction ("cell corpses") following electrodisruption and expressed as a percentage of the total cell-associated LDH (2).
Protein Immunoblotting (Western Blotting)-Equivalent amounts of protein from the various organelle fractions (whole cell homogenates, lysosomes, autophagosomes, amphisomes ϩ autophagosomes) or from a "residual fraction" prepared as autophagosomes, but from cells incubated with 3MA and homogenates treated with GPN (and applied as an autophagosome-equivalent volume rather than as an equivalent amount of protein), were subjected to one-dimensional SDS-polyacrylamide gel electrophoresis, using the mini-Protean equipment from Bio-Rad. The residual fraction was essentially devoid of autophagosomes, amphisomes, or lysosomes, thus serving to define the background of contamination by endoplasmic reticulum, endosomes, etc. in the organelle fractions. The separated proteins were transferred onto Nitropure nitrocellulose supports (Micron Separations Inc., West Borough, MA) using a Bio-Rad Semidry Transblot apparatus. The blots were blocked overnight in 5% nonfat dried milk in Tris-buffered saline (TBS; 25 mM Tris, 0.8% NaCl, pH 7.4) and then incubated at room temperature for 2 h with primary antibody and for 30 min with horseradish peroxidase-conjugated secondary antibody (both in TBS with 2% dried milk). After each step, the blots were washed three times in TBS with 0.1% Tween 20. The antibodies were detected by chemiluminescence, using the SuperSignal peroxidase substrate (Pierce) and Kodak X-Omat LS films (Eastman Kodak Co.). The films were digitalized with a model 300A laser densitometer from Molecular Dynamics Inc. (Sunnyvale, CA).
Electron Microscopy-One ml of 3-nm AOM-gold suspension (absorbance value of 1.8 at 520 nm after 100-fold dilution) was injected intravenously into rats 24 h before cell isolation to prelabel hepatocytic lysosomes. To label vacuoles of the endocytic-lysosomal pathway in isolated hepatocytes, the cells were incubated for 3 h at 37°C with a low concentration of 10-nm AOM-gold (final absorbance of 2.0 at 520 nm, i.e. 1 ⁄10 of the concentration used for density shifting). The cells were washed once in washing buffer (30), fixed in 2% glutaraldehyde, 0.1 M cacodylate buffer overnight at 4°C, and postfixed for 60 min in 1% OsO 4 reduced with 1.5% potassium ferrocyanide, followed by en bloc staining with 1.5% uranylacetate. After serial dehydration in ethanol and propylene oxide, specimens were embedded in Epon and then sectioned and poststained with 0.2% lead citrate. Sections were examined in a Phillips CM10 electron microscope at 60 kV.
For quantitation of organelle numbers and gold contents, three blocks from each treatment group within each experiment were sectioned, and 2-4 cell profiles (with a nucleus) were examined on each grid. The number of endocytic and autophagic vacuoles per cell profile was counted in the microscope at ϫ 13,500 magnification. For measurements of organelle diameters and volume fractions, 10 random micrographs for each treatment group were taken from four separate experiments (i.e. a total of 40 micrographs/treatment) and magnified to ϫ 34,500, and the relative cytoplasmic volume fraction occupied by each type of organelle was determined morphometrically using a double lattice. Organelle profile diameters were generally recorded as the average of the shortest and longest diameter, except for small endocytic tubules and vesicles, where the smallest diameter was used.

Receptor-mediated Uptake of AOM-gold by Isolated Rat
Hepatocytes-AOM-coated 10-nm gold particles (AOM-gold), used in the present experiments both as an ultrastructural vacuole marker and as a perturbant of vacuole density, were taken up into isolated hepatocytes by a saturable process, like free AOM (Fig. 1A). The uptake of radiolabeled AOM-gold was competed out by unlabeled AOM with the same efficiency as was radiolabeled, free AOM ( Fig. 1B), indicating that AOMgold at the concentrations used (200 -400 nM AOM) was taken up virtually exclusively by receptor-mediated endocytosis. The higher uptake efficiency of gold-bound AOM relative to free AOM in Fig. 1A probably relates to the fact that each gold particle carries about three molecules of AOM, which may be internalized by a single receptor engagement, as opposed to only one molecule of free AOM per receptor engagement.
Density Shifting of Hepatocytic Lysosomes and Autophagic Vacuoles by Endocytosed AOM-gold-Isolated rat hepatocytes were incubated at 37°C under conditions of maximal autophagy (amino acid-free medium), with 125 I-TC-AOM (200,000 dpm; 10 g of protein/ml) included as a marker of endocyticlysosomal vacuoles (26,27). To shift the density of these vacuoles, colloidal 10-nm AOM-gold was added, replacing unlabeled AOM so as to leave the uptake of 125 I-TC-AOM unaltered. After 3 h of endocytosis in the absence of AOM-gold, virtually all of the 125 I-TC-AOM banded in a sucrose density gradient as a single peak at 1.19 g/ml, mostly in acid-soluble (i.e. degraded) form ( Fig.  2A), at the same position as the lysosomal marker enzyme, acid phosphatase (Fig. 2B). When 10-nm AOM-gold was endocytosed along with the 125 I-TC-AOM, nearly all of the acid-soluble and acid-insoluble radioactivity, as well as most of the acid phosphatase, sedimented to the bottom of the gradient tube (Ͼ1.20 g/ml), indicating that the lysosomes had become extensively densityshifted by fusion with AOM-gold-loaded endosomes.
Prelysosomal autophagic vacuoles (autophagosomes and amphisomes) can be recognized in density gradients by their contents of 3MA-sensitive, i.e. autophagocytosed, LDH (35), which becomes degraded once these vacuoles fuse with lysosomes (2). As seen in Fig. 2C, the peak of 3MA-sensitive LDH at 1.14 g/ml was only slightly shifted toward higher densities by endocytosed AOM-gold. This would suggest that in control cells, autophagocytosed LDH resides largely in autophagosomes rather than in amphisomes.
Elevation of Autophagocytosed LDH Levels by Late Stage Inhibitors of Autophagic Flux-Since so little prelysosomal LDH could be detected in control cells, an attempt was made to induce prelysosomal autophagic vacuole accumulation by incubating the cells with various inhibitors of late stages in the autophagic-lysosomal pathway (18,35,38). As shown in Fig. 3, the proteinase inhibitor leupeptin and the microtubule inhibitor vinblastine (Fig. 3A) as well as the amino acid asparagine and the lysosomotropic amines propylamine and ammonium chloride ( dependent vacuole translocations, thereby preventing fusion between endosomes and lysosomes (27,39,40), between endosomes and autophagosomes (17), and between prelysosomal autophagic vacuoles and lysosomes (19,41). In the presence of this drug, endocytosed 125 I-TC-AOM could be seen to have accumulated in light (1.11 g/ml) and dense endosomes (1.14 g/ml) at 3 h, whereas very little radioactivity had reached the lysosomes to become acid-soluble (Fig. 4A, open symbols). The endosomes were extensively density-shifted upon co-endocytosis of AOM-gold (Fig. 4A, closed symbols), but the shift was less complete than in control cells, probably because vinblastine inhibits the overall endocytic uptake of AOM-gold (27) .
Whereas the lysosomes from control cells banded as a sharp peak at 1.19 g/ml (Fig. 2B), the lysosomes from vinblastinetreated cells showed a broad density distribution between 1.13 and 1.19 g/ml (Fig. 4B). This distribution was virtually unaffected by AOM-gold, confirming the efficiency of vinblastine as an inhibitor of endosome-lysosome fusion (9,40). Furthermore, the prelysosomal, LDH-containing autophagic vacuoles, accumulating to high levels in the presence of vinblastine (35), were also essentially unaffected by endocytosed AOM-gold under these conditions (Fig. 4C). The density shift experiment thus confirms the ability of vinblastine to prevent endosome-autophagosome fusion (9) and indicates that the autophagic vacuoles accumulating in the presence of vinblastine are autophagosomes rather than amphisomes.
Effect of Lysosomotropic Amines on AOM-gold-induced Density Shifts-Ammonia, propylamine, and other lysosomotropic, pH-elevating amines inhibit intralysosomal protein degradation and cause lysosomal swelling (42)(43)(44)(45), possibly also accumulation of prelysosomal autophagic vacuoles (41). Both propylamine and ammonia (NH 4 Cl) prevented the formation of acid-soluble degradation products from endocytosed 125 I-TC-AOM (Fig. 4, D and G) and made the lysosomes light, causing them to band broadly in the sucrose gradients with a peak around 1.15 g/ml (Fig. 4, E and H). Whereas the endocytosed 125 I-TC-AOM was extensively shifted (Fig. 4, D and G), AOMgold induced only an insignificant lysosomal density shift (Fig.  4, E and H), suggesting that the amines had effectively prevented fusion between endosomes and lysosomes. Both amines caused extensive accumulation of autophagocytosed LDH (Fig.  4, F and I; compare with the control cells in Fig. 2C), presumably reflecting an inhibition of intralysosomal LDH degradation (18). Like the lysosomal marker enzyme acid phosphatase, the LDH peak was only slightly density-shifted by endocytosed AOM-gold (Fig. 4, F and I), compatible with an intralysosomal localization of the accumulated LDH.
Density Shifting of Endosomes, Lysosomes, and Autophagic Vacuoles in the Presence of Asparagine-Asparagine, an inhibitor of autophagic-lysosomal fusion that does not affect endocytic-lysosomal fusion (17)(18)(19), allowed endocytosed 125 I-TC-AOM to be density-shifted by endocytosed AOM-gold almost to the same extent as in control cells (Fig. 5A). Unlike vinblastine and the lysosomotropic amines, asparagine also allowed extensive density shifting of lysosomes (Fig. 5B) and of the LDH-containing (prelysosomal) autophagic vacuoles that had accumulated in its presence (Fig. 5C); i.e. a significant fraction of latter were apparently amphisomes.   (46) and of intralysosomal protein degradation (47,48), reduces the ability of lysosomes to fuse with endosomes (49,50) and with prelysosomal autophagic vacuoles (18), thereby causing accumulation of the latter (41). In cells incubated with leupeptin, endocytosed 125 I-TC-AOM accumulated mainly in dense endosomes at 1.14 -1.15 g/ml and was strongly density-shifted by AOM-gold (Fig. 5D), whereas the lysosomes at 1.19 g/ml were moderately density-shifted (Fig. 5E), suggesting a partial inhibition of endosome-lysosome fusion. Remarkably, the autophagocytosed LDH distributed as two distinct peaks: one dense peak coinciding with the lysosomal marker enzyme acid phosphatase at 1.19 g/ml and another, light peak at 1.15 g/ml in a region with very little acid phosphatase activity (Fig. 5F). These two peaks would probably represent LDH inside degradation-suppressed lysosomes, and LDH in prelysosomal autophagic vacuoles, respectively (41). Endocytosed AOM-gold altered the LDH distribution dramatically, causing an extensive density shift of the light peak, with some of the LDH sedimenting to the bottom of the tube (Fig. 5F). A major fraction of the light autophagic vacuoles accumulating in the presence of leupeptin would thus appear to be amphisomes, acquiring a high density upon fusion between autophagosomes and heavy, AOM-gold-loaded endosomes.
One 10-nm gold-labeled endocytic vacuole category was represented by small endocytic vesicles and tubules, about 70 nm in diameter, usually found in the periphery of the cell (Fig. 6A). These are thought to be extensively interconnecting elements of an early, tubulovesicular endocytic compartment (26,51). The vesicle profiles contained only one or two gold particles each, implying that a significant fraction of them probably remained unlabeled.
Another morphologically recognizable vacuole category was the multivesicular endosomes, 300 -400 nm in diameter (Fig.  6B). The smaller of these may still be connected to the periph-eral tubulovesicular network, whereas the larger ones are mostly detached, late endosomes en route to the lysosomes (26,27), from which they could be distinguished by the absence of 3-nm gold. The multivesicular endosomes contained, on average, 20 -30 10-nm gold particles/vacuole profile, which was sufficient to label about 90% of them (as indicated by a counting of both gold-positive and gold-negative endosomes).
Autophagosomes were recognized by their contents of undegraded cytoplasm, delimited by double or multiple membranes, and by the absence of any gold marker (Fig. 6C).
Amphisomes (Fig. 6, E-J) resembled autophagosomes in having unaltered or only slightly altered cytoplasm, but in addition they contained 10-nm gold, indicating that fusion with endosomes had occurred. In leupeptin-treated cells, where the fusion apparently was retarded, the endosomal fusion partner could sometimes be recognized as a distinct vacuole, tightly apposed to its autophagosomal partner in a sort of prefusion complex (Fig. 6F). Amphisomes were often composite, containing several discrete packages of cytoplasm or of 10-nm gold, indicative of repeated fusions with autophagosomes and endosomes (Fig. 6G). Amphisomes contained on average about 30 gold particles/vacuole profile, which would indicate a detectability of at least 90%. A gold particle distribution histogram,

FIG. 6. Morphology of endocytic and autophagic vacuoles.
Hepatocytes, isolated from a rat injected intravenously with 3-nm AOM-gold 24 h before sacrifice (to label the lysosomes), were incubated for 3 h with 10-nm AOM-gold (to label the endocytic pathway) and processed for conventional electron microscopy. Small endocytic tubules and vesicles near the cell surface (A) and larger, multivesicular endosomes (B) are seen to contain 10-nm gold. Autophagosomes (C) are gold-negative vacuoles with a content of undegraded cytoplasm, delimited by double or multiple membranes. Lysosomes (D) are labeled with 3-nm gold (small arrowhead) and often with 10-nm gold (large arrowhead) as well and contain more or less degraded material of autophagic or endocytic origin. Amphisomes, from control (E) or leupeptin-treated cells (F, G), are autophagic-endocytic fusion vacuoles that contain both endocytosed 10-nm gold (large arrowhead) and undegraded cytoplasm but no 3-nm gold particles. In the amphisome in F, the two endocytic fusion partners can be recognized as multivesicular endosomes. Amphisomes recovered from sucrose gradients (H) or from a preparation of purified autophagic vacuoles (I and J) have the same morphology as in intact cells, containing autophagocytosed material and endocytosed 10-nm AOM-gold (large arrowhead). Magnification, ϫ 32,000 (bar, 100 nm). extrapolated to 0, confirmed that the expected fraction of false negative amphisomes was only about 6% (results not shown).
The hepatocytic lysosomes included a few small dense bodies (presumably resting lysosomes), but under the present conditions, with maximal autophagy, the majority were large, electron-lucent vacuoles that contained autophagocytosed cytoplasm at various stages of degradation (Fig. 6D). Most of the lysosomes were labeled with recently endocytosed 10-nm gold, demonstrating their simultaneous engagement in both endocytic and autophagic degradation (9). Leupeptin treatment had little effect on the numbers of endosomes and 3-nm gold-labeled lysosomes but induced a significant accumulation of autophagosomes and 3-nm gold-negative, 10-nm gold-positive autophagic vacuoles ( Table I). The latter represent a mixture of amphisomes and 3-nm gold-negative lysosomes, only about onehalf of the lysosomes being labeled by 3-nm AOM-gold overnight in vivo (26). Assuming that 3-nm gold-positive and -negative lysosomes are similarly unaffected by leupeptin, the increase within the 3-nm gold-negative vacuole class would all be due to amphisome accumulation. Volume fraction changes closely followed the changes in vacuole numbers (according to a morphometric analysis of 40 micrographs from four independent experiments; results not shown), e.g. with a highly significant 130% leupeptin-induced increase in the volume fraction of the 3-nm gold-negative amphisomes/lysosomes. Furthermore, a 10-nm gold particle count, indicative of endocytic flux, revealed that whereas leupeptin had no effect on gold flux through the endocytic compartments, it caused an accumulation of gold particles in the mixed amphisome/lysosome class at the expense of the pure (3-nm gold-positive) lysosomes (Table  I). Leupeptin thus clearly interfered with the flux of endocytosed AOM-gold from amphisomes to lysosomes.
Purification and Characterization of Amphisomes and Autophagosomes-For further structural and biochemical characterization of amphisomes, we modified our recently developed method for autophagosome purification. 2 by using leupeptintreated rather than vinblastine-treated hepatocytes as starting material, thereby obtaining an autophagic vacuole preparation enriched in amphisomes (40 -50% amphisomes and 50 -60% autophagosomes, as compared with 95% autophagosomes and 5% amphisomes in the purified autophagosome preparation). The amphisomes could easily be distinguished from the autophagosomes by their contents of recently endocytosed 10-nm AOM-gold particles (Fig. 6, I and J). Fig. 7 shows the presence of various protein markers (as detected by Western immunoblotting) in the amphisome-enriched fraction as well as in purified autophagosomes and lysosomes and in a residual fraction from autophagy-suppressed cells. The stable cytosolic marker, superoxide dismutase, is delivered to autophagic vacuoles through autophagy (52) and was, accordingly, present in all fractions (except in the residual fraction). The lysosomal markers, cathepsin B and the membrane glycoprotein Lgp120, were present virtually exclusively in the lysosomes. Two early endosome markers, the ASGPR and the early endosome-associated protein 1 (53), were clearly present in the amphisome-enriched vacuole preparation but hardly detectable in autophagosomes or lysosomes. The same was true for a late endosome marker, the MPR. Both early and late endosomes would thus seem able to fuse with autophagosomes to form amphisomes. The endosomal contamination of these fractions was negligible, as indicated by the lack of mark-TABLE I Leupeptin-induced alterations in the numbers of, and contents of endocytosed AOM-gold in, endocytic and autophagic vacuoles of isolated rat hepatocytes Isolated rat hepatocytes, containing lysosomes prelabeled with 3-nm AOM-gold by an intravenous injection 24 h before cell isolation, were incubated for 3 h at 37°C with 10-nm AOM-gold in the presence or absence of leupeptin (0.3 mM). After incubation, the cells were fixed and processed for conventional electron microscopy. The number of autophagic and endocytic organelle profiles and the number of gold particles in each of the categories listed were counted in four or five independent experiments, representing 40 -50 cell profiles within each treatment group. Among the endocytic vacuoles (multivesicular endosomes or small tubules/vesicles), only those containing 10-nm gold were counted. Among the autophagic vacuoles (including the lysosomes), both vacuoles with and without gold particles were counted. Each value represents the mean Ϯ S.E. of four or five independent experiments (specified in parentheses).  7. Protein immunoblotting (Western blotting) of organelle markers. Hepatocytes were incubated for 3 h at 37°C with 0.3 mM leupeptin (for preparation of amphisome-enriched autophagic vacuoles, Amphisomes ϩ auto), or with 50 M vinblastine (for preparation of all other fractions). Subcellular fractions were prepared as described under "Experimental Procedures," and equivalent amounts of protein from the various organelle fractions were subjected to SDS-polyacrylamide gel electrophoresis and blotted onto nitrocellulose membranes. The cytosolic marker protein superoxide dismutase (SOD), the lysosomal glycoprotein Lgp120, the lysosomal proteinase cathepsin B, the early endosome-associated ASGPR, the early endosome-associated protein 1 (EEA1), and the late endosome-associated MPR were detected with specific antibodies and a peroxidase/chemiluminescence staining method. ers in the autophagy-suppressed residual fraction (results from vinblastine-treated and leupeptin-treated cells being similar). The lack of endosomal markers in the lysosomes indicates, furthermore, that the markers are either rapidly degraded there or (more likely, since e.g. ASGPR is metabolically stable) recycle rapidly from amphisomes or lysosomes. DISCUSSION The present results demonstrate that amphisomes can be distinguished physically (in sucrose density gradients and by purification) as well as morphologically (in the electron microscope) from other, closely related autophagic and endocytic vacuoles. Sucrose density gradients have previously been shown to be suitable for the separation of hepatocytic endosomes from lysosomes (54), whereas prelysosomal autophagic vacuoles, accumulating in the presence of vinblastine, have been found to band in these gradients at the same position as the lysosomes (35). In the present study, the two types of vacuole could be seen to have overlapping gradient distributions not only after vinblastine treatment but also after treatment with other autophagic flux inhibitors such as propylamine, ammonia, or asparagine, all of which induced autophagic vacuole accumulation as previously observed (2,18). Prelysosomal autophagic vacuoles from control cells were lighter than lysosomes, but these vacuoles are rather short lived (55), and their number at steady state is low even when autophagy is running at its maximal rate (2,18). The small peak of LDH-containing autophagic vacuoles from control cells was only slightly densityshifted in the sucrose gradients by endocytosed AOM-gold, suggesting that few amphisomes were present. However, the vacuole numbers recovered were too small to permit meaningful quantitative estimates on the basis of the density shift.
Leupeptin, an inhibitor of intralysosomal proteolysis (47,48), has previously been shown to induce the development of congested, protein-filled lysosomes as well as a secondary accumulation of prelysosomal autophagic vacuoles (41,56), reflecting a transient inhibition of autophagic flux into the lysosome (18). In isotonic density gradients, the congested lysosomes were found to have a higher density than normal lysosomes (57), but since leupeptin increases the density of prelysosomal autophagic vacuoles as well, the latter could not be separated from lysosomes in isotonic metrizamide density gradients (58). The presently observed banding of prelysosomal autophagic vacuoles from leupeptin-treated cells as a distinct, light peak in hypertonic sucrose density gradients was, therefore, a pleasant surprise. One possible explanation is that autophagosomes and amphisomes (and multivesicular endosomes as well) contain vesicles with intact membranes that may, under hypertonic conditions, better resist the osmotic extraction of water than do the lipolytically degraded membrane structures inside the lysosomes. Although the criteria used to define prelysosomal autophagic vacuoles in the gradients (presence of LDH, lack of acid phosphatase) could not distinguish amphisomes from autophagosomes, the extensive density shifting of the prelysosomal peak by endocytosed AOMgold demonstrates unequivocally that the majority of these vacuoles are amphisomes, containing both autophagocytosed (LDH) and endocytosed (AOM-gold) material. AOM-gold-induced density shifting is thus a powerful method, which may provide a valuable supplement to the established diaminobenzene density shift technique (59).
The present ultrastructural, morphometric analysis, using endocytosed AOM-gold to mark the amphisomes, confirms our gradient data in suggesting that the majority of the prelysosomal autophagic vacuoles accumulating in the presence of leupeptin were indeed amphisomes. Furthermore, our quantitative analysis of the endocytic flux of AOM-gold clearly showed a leupeptin-induced accumulation of gold particles in the amphisomes at the expense of the lysosomes, thus providing additional evidence for an inhibition of amphisome-lysosome fusion by leupeptin. Somewhat surprisingly, leupeptin did not induce any detectable retention of AOM-gold in the late, multivesicular endosomes, despite the documented ability of the drug to inhibit endosome-lysosome fusion (49). A likely explanation would be that the "endosomes" that accumulate undegraded 125 I-TC-AOM in the presence of leupeptin are actually amphisomes, banding at the same sucrose gradient density (1.14 g/ml) as the late endosomes (26). Perhaps leupeptin inhibits only that part of the endocytic flux that passes via the amphisomes, leaving any direct endosome-lysosome fusion unaffected. The fusion activity of the autophagosomes, on the other hand, is probably inhibited by leupeptin, as suggested by the 2-fold accumulation of autophagosomes seen in the presence of the drug, but it is not clear whether this represents a primary inhibition of autophagosome-lysosome fusion or reduced autophagosome-amphisome fusion, e.g. as a secondary consequence of impaired amphisome-lysosome fusion (cf. Fig. 8 for overview).
Our electron micrographs indicate that amphisomes may undergo multiple independent fusions with multivesicular endosomes, at different places on the amphisome surface. Multivesicular endosomes can be either early (ligand-recycling, attached to the peripheral tubulovesicular endocytic network) or late (free endosomes without ligand recycling) (26,27). The fact that amphisomes (but not autophagosomes or lysosomes) contain the MPR, generally regarded as a marker of late endo- somes (60), suggests that they have undergone fusion with late endosomes. However, the amphisomal localization of typical early endosome markers like the asialoglycoprotein receptor (21) and the early endosome-associated protein 1 (53) indicates that even relatively early (sorting) endosomes may fuse with autophagosomes and amphisomes (Fig. 8). The possibility should probably be considered, however, that the transition from early to late endosomes may represent a very gradual maturation process (60) and that, in hepatocytes at least, the endosomes fusing with autophagosomes/amphisomes may be free multivesicular endosomes carrying a variable mixture of early and late endosome markers. The total absence of endosome markers in the lysosomes indicates that the markers are either rapidly degraded in the lysosomes or (more likely, since e.g. ASGPR is metabolically stable) recycle rapidly from amphisomes or lysosomes. Hepatocytes thus present a clean cut distinction between MPR-positive, Lgp120-negative prelysosomal vacuoles (endosomes and amphisomes) and MPR-negative, Lgp120-positive lysosomes (9). In other cell types, early lysosomes may contain MPR (61), and under some conditions all of the lysosomes may be MPR-positive (22).
Endosomes are apparently capable of fusing directly with lysosomes as well as with autophagosomes/amphisomes (6,12,19,21,24,25), but they seem to have a 5-6-fold higher propensity to fuse with autophagosomes or amphisomes (21). The inability of some authors to observe fusion between endosomes and prelysosomal autophagic vacuoles (24,25) may relate to the definitions, markers, and methods used. Amphisome formation has been demonstrated in several nonhepatic cells and tissues (22,23) and is likely to be a general biological phenomenon.
Amphisomes will eventually fuse with small lysosomes, probably independently of endosomes, as indicated by the separate attachments of lysosomes and endosomes to the surface of amphisomes from leupeptin-treated cells. Several small lysosomes can apparently attach to, and fuse with, the same autophagic vacuole to form large, active lysosomes, accounting for lysosomal heterogeneity and for the gradual increase in lysosomal hydrolytic capacity (52). Secondary lysosomes with autophagic contents are often described as "autolysosomes," but since all lysosomes are capable of simultaneous engagement in autophagy and endocytosis (9,62), the term "active lysosomes" may be more appropriate (to be contrasted with the "passive," unengaged lysosomes of the dense body type; cf. Refs. 58 and 63). It is also possible that amphisomes may receive some lysosomal enzymes (albeit apparently not cathepsin B) through the endocytic pathway. Certain lysosomal enzymes, such as acid phosphatase, are synthesized as transmembrane proteins that enter the endocytic pathway (via the plasma membrane) due to specific sorting motifs in the cytoplasmic tail of the enzyme (51,64,65). Other lysosomal enzymes are bound to mannose 6-phosphate receptors in the trans-Golgi network (66,67) and brought to a late endosomal compartment, directly or via the plasma membrane, by the help of similar endocytic sorting motifs in the receptor molecule (51). Since amphisomes are acidic (12,20), probably due to a proton pump brought in by the endosomal fusion partner (68), the possibility cannot be excluded that limited proteolytic activity may take place in the amphisomes.
Morphological evidence indicates that autophagic inputs to amphisomes and lysosomes can also be multiple (Ref. 52 and present data). Biochemical studies have suggested that the amphisomes exhibit a certain permanence and are able to receive a continued autophagic influx even if the endocytic influx is blocked (19). Conversely, endocytic influx to amphisomes as well as to lysosomes can apparently proceed in the absence of autophagy (19), and direct fusion between lysosomes and late endosomes has been observed under cell-free conditions (69). There is, furthermore, strong morphological evidence for direct fusion of autophagosomes with lysosomes (6,12,24,25). In addition to these various heterotypic fusions, the lysosomes are known to undergo frequent homotypic fusions and fissions, with extensive exchange of contents (70,71), and late endosomes have been shown to be capable of homotypic fusion under cell-free conditions (72). Autophagosomes, multivesicular endosomes, amphisomes, and lysosomes seem, therefore, to be rather promiscuous in their choice of fusion partners, although some preferences can be discerned (21).
In the original biochemical studies that identified the amphisome, asparagine was used, as a selective inhibitor of amphisome-lysosome fusion, to induce accumulation of LDH-containing amphisomes (17). The present study confirmed the ability of asparagine to induce accumulation of autophagocytosed LDH and of LDH-containing vacuoles. Although the latter could not be resolved from lysosomes in sucrose density gradients, they were extensively density-shifted by endocytosed AOM-gold, unlike the LDH-containing lysosomes from ammonia-or propylamine-treated cells, thus discounting the possibility that the asparagine effect could be mediated by its deamination to ammonia, a lysosomotropic agent (43,44). Asparagine (but not ammonia or propylamine) likewise allowed density shifting of endosomes and lysosomes, supporting the notion that the latter were not detectably affected by asparagine-generated ammonia. The gradient data are thus consistent with the hypothesis that LDH accumulates in amphisomes after asparagine treatment and in lysosomes after ammonia or propylamine treatment (18).
Microtubule inhibitors like vinblastine and colchicine block both the endocytic and the autophagic flux in isolated rat hepatocytes (39,73) and cause accumulation of endosomes and prelysosomal autophagic vacuoles (9,41,74). We have recently demonstrated that vinblastine inhibits the transfer of endocytosed AOM to late endosomes by preventing the microtubule-dependent maturation of multivesicular endosomes from the early tubulovesicular network (26,27). The present results show that the density of the endosomes could still be shifted by AOM-gold in the presence of vinblastine, but the density distributions of autophagic vacuoles and lysosomes remained unaffected. Vinblastine thus appears to effectively block endocytic influx to the autophagic pathway. Its effect within the autophagic pathway is apparently to inhibit the fusion of autophagosomes and amphisomes with lysosomes, probably without affecting autophagosome-amphisome fusion (19). However, since no new amphisomes are formed, the net result of vinblastine treatment is an extensive accumulation of autophagosomes, which has been utilized in attempts to purify these organelles (35).
In conclusion, the present study shows that amphisomes are distinct organelles that can be separated from other autophagic vacuoles, using leupeptin to prevent their fusion with lysosomes and using endocytosed AOM-gold to density-shift them away from autophagosomes. Amphisomes form by single or multiple autophagosome-endosome fusions and can be recognized morphologically by their mixed autophagic-endocytic contents. Novel procedures for purification of autophagosomes and partial purification of amphisomes have allowed a preliminary biochemical characterization of both organelles. Amphisomes lack lysosomal marker enzymes (acid phosphatase and cathepsin B) and lysosomal membrane proteins (Lgp120) but may carry both early and late endosomal markers like the asialoglycoprotein receptor, the early endosome-associated protein 1, and the cation-independent mannose 6-phosphate receptor.