We found that overexpression of Atg4B has an inhibitory effect on autophagy independent of its delipidation activity in mammalian cells. The inactive Atg4BC74A
mutant specifically blocks the lipidation of LC3 paralogues by sequestering LC3 paralogues in stable complexes, resulting in blockade of the Atg7-LC3 reaction. It is interesting that the Atg4B enzyme has high affinity for its product, processed LC3 paralogues. This is in sharp contrast to overexpression of inactive ubiquitin-deconjugating enzymes, which act in a dominant-negative manner, elevating levels of their substrate, ubiquitin-conjugated proteins (Hang and Dasso, 2002
; Li et al., 2002
). In the course of the enzymatic reaction, Atg4 and LC3 paralogues should associate, but once cleavage has occurred, the proteins should dissociate for the next step, the Atg7-LC3 reaction. It is possible that an unknown factor is necessary for dissociating LC3 from Atg4, and Atg4 overexpression might result in limitation of this factor.
In cells overexpressing Atg4BC74A
, a large number of the autophagic structures were not closed, although the length of these membranes was comparable to the length of autophagosomal membranes in control cells (M). This observation fits with the results that Atg5-positive membrane structures accumulated and had prolonged lifetimes (), based on the previous report that the Atg16L complex detach once the autophagosome formation is completed (Mizushima et al., 2001
). In Atg5 knockout cells expressing the GFP-Atg5K130R
mutant, in which Atg12-Atg5 conjugation does not occur, GFP-Atg5K130R
signals also remain longer in membranous structures. These cells have incomplete Atg16L complexes, which lack Atg12, whereas in Atg4BC74A
overexpressing cells, the Atg16L complex is intact. Therefore, it seems that the trigger that liberates the Atg16L complex from the membrane upon completion of autophagosome formation is not within the Atg16L complex.
It is interesting that autophagosome formation proceeded to a relatively late stage in Atg4BC74A
-overexpressing cells; formation of autophagosome-like structures takes place. In addition, ULK1 is recruited to the structure in a manner similar to that observed in the isolation membranes of control cells (Supplemental Figure S5). The apparent defect in autophagosome completion is closure of the end of each elongating membrane. This does not necessarily exclude the proposal that Atg8 functions in expansion of autophagosomal membranes in yeast (Xie et al., 2008
), because we observe only terminal phenotype and cannot exclude the possibility that elongation speed is slower. Atg8-PE can cause hemifusion of vesicles in vitro (Nakatogawa et al., 2007
). One possibility is that LC3 paralogues function to complete autophagosome formation by fusing membranes in mammalian cells. Interestingly, in atg8Δ yeast, autophagosome-like structures were also detected by electron microscopy, however, at low frequency (Kirisako et al., 1999
). There remains a possibility that expansion of autophagosomal membrane is not completely hampered by the deletion of Atg8.
We do not believe the phenotypes we observed are due to incomplete inhibition of LC3 paralogues function, because autophagosome-like structures were also observed in Atg3 (a specific E2 enzyme for Atg8 homologues) knockout MEF cells, where PE conjugation of LC3 paralogues is defective (Dr. Keiji Tanaka and Dr. Masaaki Komatsu, personal communication). By utilizing an inactive mutant of Atg4B, we could exclude secondary effects that might be brought about by overexpression of wild-type Atg4B, such as hyperdelipidation. Therefore, the phenotypes we observed likely reflect the physical sequestration and deficiency of the LC3 paralogues.
One strategy that may provide important information is artificial inhibition of autophagy. Indeed, treatment with drugs such as wortmannin or 3-methyladenine is widely used in studies of autophagy, but these drugs have side effects. RNA interference–mediated gene knockdown is a potential approach; however, nearly complete suppression of the ATG
genes is needed to fully inhibit autophagy, and this is often difficult to achieve (Hosokawa et al., 2006
; Yoshimura et al., 2006
). The use of genetic knockouts of ATG
genes is an alternative option for complete inhibition, but available cell types are restricted. In the case of Atg4BC74A
overexpression, it seems possible to fully inhibit autophagy in any type of cell. Several studies have suggested that the Atg12-Atg5 conjugate has other roles in addition to autophagy (Pyo et al., 2005
; Yousefi et al., 2006
; Takeshita et al., 2007
). However, currently reported Atg5- or Atg7-deficient cells do not distinguish between autophagic and nonautophagic function of the Atg12-Atg5 conjugate, as it is lacking in both cell types. As overexpression of the Atg4B mutant inhibits formation of autophagosomes, but not generation of the Atg12-Atg5 conjugate, such problems can be avoided. We believe that the inactive Atg4B mutant will provide a useful tool for a broad range of studies analyzing autophagy.