The ATG conjugation systems are important for degradation of the inner autophagosomal membrane

Macroautophagy (hereafter, autophagy) is a highly inducible intracellular degradation system (1–3). First, a flat membrane sac, termed the isolation membrane or the phagophore, elongates, bends, and sequesters a part of the cytoplasm. Then, its edge is closed by membrane fission to form the double-membrane structure, the autophagosome (4). The autophagosome fuses with lysosomes and becomes the autolysosome. Lysosomal enzymes selectively degrade the inner autophagosomal membrane (IAM), but not the outer autophagosomal membrane (OAM), and then finally degrade the enclosed cytoplasmic contents.

To characterize autophagosome maturation in mammalian cells, we used the autophagosomal soluble N-ethylmaleimide–sensitive factor attachment protein receptor (SNARE) syntaxin 17 (STX17) as an autophagosome marker (5). Elongating isolation membranes were microtubule-associated protein light chain 3 (LC3) positive but STX17 negative (Fig. 1A, red arrows, and movie S1) (5). Later, STX17 was recruited to the whole circumference of ring-shaped structures (Fig. 1A, green arrows). The shape of elongating isolation membranes was elliptical but became almost completely spherical when STX17 was recruited (Fig. 1B). We hypothesize that fission between the OAM and IAM during the closure of the autophagosomal edge causes a morphological change into stable spherical bodies that occurs immediately before or after the recruitment of STX17.

• Download high-res image
• Open in new tab
• Download Powerpoint

After STX17 recruitment, several small LAMP1-positive lysosomes (or late endosomes) associated with the autophagosomes (Fig. 1C, 2 to 8 min, blue arrows). Then, the autophagosomal membrane became LAMP1-positive (Fig. 1C, 10 to 16 min, blue arrows). At almost the same time, these structures became positive for LysoTracker Red (LTR), a weak-base probe for acidic compartments. The LTR signals appeared also as ring-shaped structures on the STX17-positive structures (Fig. 1C, 4 to 10 min, red arrows), which suggests that the space between the OAM and IAM is acidified. Next, the ring-shaped LTR signal collapsed, and the lumen of the autolysosomes filled with LTR, which indicated the IAM degradation (Fig. 1C, 12 to 16 min). The STX17 signal gradually disappeared after the collapse of the LTR ring structure (Fig. 1C, 12 to 14 min). This STX17 dissociation was independent of luminal acidification because it was not affected by bafilomycin A₁ treatment (fig. S1). Thus, we detected four steps during autophagosome maturation: STX17 recruitment (step a), lysosomal fusion (step b), IAM degradation (step c), and STX17 release (step d) (Fig. 1C). The total lifetime of STX17 (steps a to d) and the durations (means ± SEM) between each step (a to b, b to c, and c to d) were 11.0 ± 0.6, 2.1 ± 0.3, 6.6 ± 0.6, and 2.2 ± 0.2 min, respectively (Fig. 1C).

Autophagosome formation requires the two ubiquitin-like systems: the ATG12 conjugation system and the ATG8 (LC3s and γ-aminobutyric acid receptor–associated proteins in mammals) conjugation system (6, 7). Ubiquitin-like ATG12 and ATG8 are covalently conjugated to ATG5 and phosphatidylethanolamine (PE), which are catalyzed by the common E1-like protein ATG7 and the specific E2-like proteins ATG10 and ATG3, respectively. Conjugation of ATG8 or LC3 with PE depends on ATG12–ATG5 (8, 9). ATG8 seems to be essential for an early stage of autophagosome formation in yeast (10, 11). By contrast, accumulation of elongated isolation membranes and even double-membrane autophagosome-like structures, but not autolysosomes, has been observed in mouse embryonic stem cells lacking ATG5 (9); mouse embryonic fibroblasts (MEFs) lacking ATG5 (12) or ATG3 (13, 14); and NIH-3T3 cells overexpressing a dominant-negative mutant of ATG4B (15). These findings led to the suggestion that the ATG conjugation systems are required for the closure step. However, the fate of these elongated isolation membranes is unknown, because the sole autophagosome marker LC3 and any other ATG proteins cannot localize to autophagosomes in these cells (16, 17).

To overcome this limitation, we used STX17 as an autophagosome marker. Unexpectedly, a number of STX17-positive punctate or ring-shaped structures were observed under starvation conditions in ATG conjugation–deficient cells, such as ATG3 knockout (KO), ATG5 KO, and ATG7 KO MEFs (Fig. 2, A and B, and movie S2) and ATG3 KO HeLa cells (fig. S2A). These STX17-positive structures also accumulated even under growing (nonstarvation) conditions in ATG conjugation–deficient cells (Fig. 2, A and B, and fig. S2A). In ATG9A KO, FIP200 (also known as RB1CC1) KO, and ATG14 KO MEFs, STX17 did not form puncta, which is consistent with a known lack of isolation membranes in these cells (12). These STX17-positive rings did not represent mislocalization of STX17; immunoelectron microscopy showed the recruitment of STX17 on autophagosome-like structures in ATG3 KO MEFs (Fig. 2C, fig. S2D). Indeed, lactate dehydrogenase, a cytosolic enzyme, was sequestered in these structures (fig. S3). Furthermore, these STX17-positive structures in ATG3 KO cells were formed by the canonical pathway, because their formation was abolished by additional knockdown (fig. S2B) or knockout (fig. S2C) of FIP200 or ATG14 or treatment with wortmannin (Fig. 2, A and B).

• Download high-res image
• Open in new tab
• Download Powerpoint

These STX17-positive autophagosome-like structures in ATG3 KO cells were positive for SNAP29 (Fig. 2D) and became LTR positive (Fig. 2E). The STX17 structures were mostly colocalized with LAMP1 (Fig. 2F). The time between STX17 recruitment and acidification (i.e., between steps a and b in Fig. 1C) in ATG3 KO MEFs was only slightly longer than that in wild-type (WT) MEFs (Fig. 2E). Thus, the fusion between autophagosome-like structures and lysosomes occurs almost normally in ATG3 KO MEFs.

To reevaluate the autophagic activity in ATG KO cells, we generated ATG3 KO, ATG5 KO, ATG9A KO, and FIP200 KO HeLa cells by the CRISPR-Cas9 genome-editing method (fig. S4) and confirmed that autophagic flux was blocked in these cells by an LC3 conversion assay (9, 13, 18) (fig. S4), accumulation of p62 (also known as SQSTM1), and lysosomal long-lived protein degradation assay (Fig. 3A). Thus, both selective and nonselective autophagy are considerably suppressed in ATG conjugation–deficient cells.

• Download high-res image
• Open in new tab
• Download Powerpoint

The appearance rate of Venus-ATG5–positive structures was 1.27 ± 0.14 (n = 9) in WT and 1.03 ± 0.11/min per cell (n = 9) in ATG3 KO MEFs, suggesting that isolation membranes are generated at a normal rate in ATG3 KO cells. The lifetime of Venus-ATG5–positive structures in ATG3 KO MEFs appeared to be longer than that in WT MEFs, as previously reported (Fig. 3B) (13). The rate of the transition of Venus-ATG5–positive structures to structures positive for cyan fluorescent protein (CFP)–labeled transmembrane domains of STX17 (STX17TM), which represents the success rate of autophagosome formation, was only 30% in ATG3 KO MEFs compared with 93% in WT MEFs (Fig. 3B). Thus, the ATG conjugation systems are indeed important for the late step of autophagosome formation.

Finally, we found that STX17-positive LTR ring-shaped structures were maintained for a much longer period (>30 min) in ATG3 KO, ATG5 KO, and ATG7 KO MEFs compared with WT and rescued ATG3 KO MEFs (Fig. 4, A and B). Thus, degradation of the IAM is significantly delayed in ATG conjugation–deficient cells. Indeed, by immunoelectron microscopy, intact IAMs could be detected in LAMP1-positive autophagosome-like structures in ATG3 KO MEFs (Fig. 4C, fig. S5). This was not due to a lysosomal defect, because degradation of epidermal growth factor receptor and activity of cathepsin B were normal in ATG3 KO cells (fig. S6). Thus, in ATG3 KO cells, the IAM becomes resistant to degradation by lysosomal enzymes.

• Download high-res image
• Open in new tab
• Download Powerpoint

Nonetheless, collapse of the IAM was occasionally observed after a certain time even in ATG3 KO MEFs (Fig. 4D, red arrows). Consistently, autolysosomes containing digested cytoplasmic materials were observed—but only rarely (fig. S7) (13). STX17 remained on the autophagosome-like structures as long as the IAM was maintained (although the intermembrane space was acidified) but immediately dissociated from the membrane after the collapse of the IAM (Fig. 4D, green arrows). Thus, dissociation of STX17 is triggered by IAM degradation rather than acidification. Accordingly, the lifetime of STX17-positive structures was longer in ATG3 KO MEFs than that in WT MEFs (Fig. 4E).

Many of the STX17-positive autophagosome-like structures in ATG3 KO MEFs were not spherical but elliptical compared with those in WT MEFs (Fig. 4F). If the fission between the OAM and IAM causes the spherical change (Fig. 1, A and B), the elliptical shape of the STX17 structures suggests that the final fission process is impaired in ATG3 KO MEFs. Indeed, in a few cases, the edges of the STX17-positive structures were not closed (Fig. 4G). In extreme cases, the edge became wide open, followed by disappearance of the STX17 signal. These data suggest that, even though STX17 is recruited, fission of the edge of the autophagosomes is likely not completed in ATG3 KO cells.

Our observation does not represent so-called “Atg5- or Atg7-independent alternative autophagy,” which was proposed to initiate from the trans-Golgi, not from the endoplasmic reticulum (ER), and typically was induced by etoposide treatment (19). We observed, however, that ATG5- or STX17-positive punctate structures were generated in a canonical manner in close proximity to the ER throughout the cytoplasm, not exclusively in the Golgi region in ATG3 KO cells (movie S2 and S3, fig. S8, A and B).

Here, we showed that ATG3 is important for two critical steps: (i) efficient transition from the isolation membrane to the autophagosome and (ii) efficient degradation of the IAM (Fig. 4H). The former finding is consistent with the previously proposed hypothesis that the ATG conjugation systems are required for closure and/or fission of the edge (11). We hypothesize that ATG conjugation–dependent closure is important for the subsequent processes, including spherical morphological change and efficient IAM degradation, but not absolutely essential for STX17 recruitment and lysosomal fusion (Fig. 4H). The inward membrane fission could be a common mechanism for efficient degradation by vacuolar or lysosomal enzymes. For example, during formation of the multivesicular body and vacuolar or lysosomal microautophagy, invaginated membranes become sensitive to lysosomal or vacuolar degradation only after separation from the outer membranes, which are resistant to degradation (fig. S9).

It is not expected that autophagosome-like structures can mature into autolysosomes after degradation of the IAM in ATG conjugation–deficient cells, which suggests that autophagic activity remains at a very low level. It may explain the phenotypic difference among ATG KO mice: Embryonic lethality is observed in mice lacking upstream ATGs—such as FIP200, ATG9A, and ATG13 (20–22), whereas mice lacking ATG conjugation components, such as ATG3, ATG5, ATG7, ATG12, and ATG16L1, can survive the embryonic period (but die shortly after birth) (13, 18, 23–25). So far, the function of ATG proteins has been investigated only in the context of autophagosome formation, but our findings also revealed a function of ATGs with regard to the maturation steps.