Differential Function of the Two Atg4 Homologues in the Aggrephagy Pathway in Caenorhabditis elegans*

Background: The function of multiple homologues of Atg4 in higher eukaryotes remains largely unknown. Results: We demonstrated that the two Atg4 homologues in Caenorhabditis elegans have differential enzymatic activities in processing LGG-1/Atg8 and exhibit functional redundancy. Conclusion: The two Atg4s contain redundant yet differential processing activities in the autophagy pathway. Significance: We revealed physiological function of the two Atg4 homologues in the autophagy pathway. The presence of multiple homologues of the same yeast Atg protein endows an additional layer of complexity on the autophagy pathway in higher eukaryotes. The physiological function of the individual genes, however, remains largely unknown. Here we investigated the role of the two Caenorhabditis elegans homologues of the cysteine protease Atg4 in the pathway responsible for degradation of protein aggregates. Loss of atg-4.1 activity causes defective degradation of a variety of protein aggregates, whereas atg-4.2 mutants remove these substrates normally. LGG-1 precursors accumulate in atg-4.1 mutants, but not atg-4.2 mutants. LGG-1 puncta, formation of which depends on lipidation of LGG-1, are present in atg-4.1 and atg-4.2 single mutants, but are completely absent in atg-4.1; atg-4.2 double mutants. In vitro enzymatic analysis revealed that ATG-4.1 processes LGG-1 precursors about 100-fold more efficiently than ATG-4.2. Expression of a mutant form LGG-1, which mimics the processed precursor, rescues the defective autophagic degradation of protein aggregates in atg-4.1 mutants and, to a lesser extent, in atg-4.1; atg-4.2 double mutants. Our study reveals that ATG-4.1 and ATG-4.2 are functionally redundant yet display differential LGG-1 processing and deconjugating activity in the aggrephagy pathway in C. elegans.

essential for autophagy under normal conditions, but Atg4C Ϫ/Ϫ starved mice show decreased autophagic activity in the diaphragm (18). Mice deficient in Atg4B exhibit a reduction of basal-and starvation-induced autophagic flux in all tissues (19). The proteolytic cleavage of multiple Atg8 homologues is severely reduced but not totally impaired in Atg4B Ϫ/Ϫ cells (19,20). Atg4C and Atg4D display minimal processing activity toward Atg8 substrates (17). Although mice deficient in Atg3, Atg5, Atg7, and Atg16L show severe defects in the autophagy pathway and die during the neonatal starvation period, mice deficient in Atg4C and Atg4B do not display any obvious abnormalities and survive the early neonatal period, suggesting that there are functional divergences and redundancies between mammalian Atg4 homologues (18 -24). The physiological function of Atg4C and Atg4D and the contribution of each of the Atg4 paralogues during the autophagy process remains poorly understood.
Caenorhabditis elegans contains two Atg4 homologues, encoded by atg-4.1 and atg-4.2. Here we showed that genetic mutations in atg-4.1, but not in atg-4.2, cause defective autophagic degradation of a variety of protein aggregates during embryogenesis. In wild-type C. elegans, the nascent Atg8 homologue LGG-1 is rapidly processed by cleavage near the C terminus to give LGG-1-I. LGG-1-I can then be conjugated with PE to give LGG-1-II. Nascent LGG-1 precursors accumulate in atg-4.1 mutants but not in atg-4.2 mutants. The PEconjugated form of LGG-1 is present in atg-4.1 and atg-4.2 single mutants, but absent in atg-4.1; atg-4.2 double mutants. An in vitro cleavage assay revealed that ATG-4.1 has much higher processing activity than ATG-4.2 toward LGG-1/Atg8. Overexpression of the processed form of LGG-1 rescues the defective degradation of protein aggregates in atg-4.1 mutants and, to a lesser extent, in atg-4.1; atg-4.2 double mutants. Our results show that atg-4.1 and atg-4.2 have differential processing activities and contribute differentially in the aggrephagy pathway in C. elegans.
Starvation Assay-Embryos collected by the synchronization method were incubated in deionized distilled water at 20°C. An aliquot of each sample was placed on a plate seeded with OP50 at each time point. After 3 days, the number of animals surviving to L4 was counted. Differences in survival were analyzed with the Kaplan-Meier method and log-rank tests using SPSS version 20.0.
Indirect Immunofluorescence-Permeabilization of embryos was performed by freeze-cracking methods. Freeze-cracked slides were fixed, blocked, and incubated with diluted antibody at room temperature for 2-4 h. The worms were then washed three times and incubated with secondary antibodies. Slides were viewed using a confocal microscope (Zeiss LSM 510 Meta).
Reporter Construction-A cDNA encoding ATG-4.1 was cloned into pPD95.79 to create a reporter with GFP fused at the C terminus. The reporter contained the unc-54 3Ј-UTR and was driven by the atg-4.1 promoter (ϳ2 kb). The plasmid was co-injected with pRF4(rol-6) into wild-type animals. The reporter was functional in rescuing defective degradation of SEPA-1 aggregates in the null allele of atg-4.1, bp501.
Plasmid Construction-cDNAs encoding ATG-4.1 and ATG-4.2 were cloned into pET-28a for His tagging. cDNAs encoding LGG-1 and LGG-1(G116A) were cloned into pET-21b for His tagging at the C terminus. Genomic DNA containing the promoter, coding sequence and 3Ј-UTR of lgg-1 (c32d5, nt 34,945-37,706) was cloned into pBluescript II SK (ϩ). This construct was further used to delete the C-terminal seven amino acids, directly exposing the conjugated glycine (Gly-116), and also to mutate glycine 116 to alanine.
Assay for in Vitro Cleavage-His-tagged ATG-4.1, ATG-4.2, and LGG-1 proteins were expressed in Escherichia coli BL21 strain and then purified using Ni-NTA beads (Qiagen). Purified His-ATG-4s (0.0125 mg/ml and 0.025 mg/ml) and LGG-1-His (0.5 mg/ml) were incubated together in reaction buffer (150 mM Nacl, 1 mM EDTA, 1 mM DTT, and 50 mM Tris-HCl, pH 7.2) in a volume of 20 l at 37°C for the given time. Laemmli sample buffer was added to stop the reactions. Proteins were then separated by 13% SDS-PAGE, followed by Coomassie Brilliant Blue (CBB) staining. The density of bands was measured by Image J (W. S. Rasband, ImageJ, U. S. National Institutes of Health, Bethesda, MD) to quantify the amount of products generated from the substrate. The percentage of substrate remaining was plotted against the reaction time and used to determine the optimal reaction time for further kinetic analysis of ATG-4.1 and ATG-4.2. Different concentrations of substrate (LGG-1-His 6 ) were incubated with ATG-4.1 and ATG-4.2 for the set optimal reaction time. The initial velocity (V, y axis) was determined from the change in LGG-1 product concentration per second (mM/s). It was then plotted against substrate concentration (S (mM), x axis) for each reaction and the resulting curves were fitted by the non-linear regression method (GraphPad Prism version 5.00, GraphPad Software, San Diego CA). The values of V max (mol liter Ϫ1 s Ϫ1 ) and K m (mol/liter) were derived from the curves. The catalytic constant, k cat (s Ϫ1 ), was calculated for each reaction by dividing V max by the enzyme concentration.
RNA Isolation and Real-time RT-PCR-atg-4.1, atg-4.2 and wild-type animals were collected and total RNA was extracted with Trizol reagent (Sigma). Total isolated RNA (2 g) from each sample was reverse transcribed using an Invitrogen Super-script III kit. Quantitative PCR reactions were carried out as described previously (25). atg-4.1 and atg-4.2 mRNA levels in mutants were normalized to the level of wild type worms, which was set to 1. Error bars indicate the standard deviation (S.D.) of three independent experiments.
Modeling-For constructing the HsAtg4B-LGG-1 complex model, Thr-118, Phe-119, and Thr-121 of LC3 were replaced with Val, Tyr, and Gly, respectively, using the program COOT while retaining the original configurations. These replacements did not cause any steric clash with the surrounding HsAtg4B residues. G258S and A263V mutations in HsAtg4B were also introduced in the same manner.

atg-4.1 Is Essential for Degradation of a Variety of Autophagy
Substrates-The germline P granule components PGL-1 and PGL-3 are selectively removed by autophagy in somatic cells during C. elegans embryogenesis (26,27). Degradation of PGL-1 and PGL-3 requires the self-oligomerization protein SEPA-1, which functions as a receptor in recruiting PGL-1 and PGL-3 into aggregates, known as PGL granules (28). SEPA-1 is also removed by autophagy during embryogenesis (28). In wild type animals, SEPA-1 aggregates are present only in early stage embryos and become undetectable from the comma stage onwards. In autophagy mutants, PGL granules (containing PGL-1/PGL-3/SEPA-1) dramatically accumulate and persist throughout embryogenesis. In genetic screens to identify mutants with defective degradation of GFP::PGL-1 and SEPA-1::GFP, we isolated 7 alleles of atg-4.1 (Fig. 1, A-C and data not shown). Endogenous PGL-3 and SEPA-1, detected by specific antibodies, also failed to be removed and accumulated into a large number of aggregates in atg-4.1 mutants (Fig. 1, E-G and data not shown).
We further investigated whether atg-4.1 is involved in other autophagy-regulated processes in addition to aggrephagy (26). Optimal survival of L1 larvae under starvation conditions requires autophagy activity. In the absence of food, atg-4.1 mutant L1 larvae had a significantly reduced survival rate compared with wild type L1 larvae (Fig. 1U). The defect in atg-4.1 mutants, however, was much milder than epg-6 mutants. Autophagy activity also contributes to the necrosis-like degeneration of six touch receptor neurons caused by a toxic ion channel variant encoded by mec-4(u231) (26). Loss of atg-4.1 activity ameliorated the neurodegeneration in mec-4(u231) animals (Fig. 1V). In summary, atg-4.1 mutants show defects in multiple autophagy-regulated processes.
We constructed a translational fusion reporter for atg-4.1 with gfp inserted at its C terminus to determine its expression pattern. The atg-4.1::gfp reporter was functional in rescuing defective degradation of SEPA-1 aggregates in atg-4.1 mutants (supplemental Fig. S1). ATG-4.1::GFP was expressed in the cytoplasm of most cells during embryogenesis (supplemental Fig. S1).
Molecular Lesions in atg-4.1 Mutant Alleles-Phylogenetic analysis reveals that ATG-4.1 is closely related to human Atg4A and Atg4B ( Fig. 2A). We identified molecular lesions in all seven alleles of atg-4.1 (Fig. 2B). In bp501, bp451, and bp410 mutants, glutamine residues are mutated to stop codons at positions 140, 160, and 385, respectively. The tryptophan at amino acid 347 is mutated to a stop codon in bp482. bp504 contains a histidine to proline mutation at codon 102. The glycine residue at amino acid 266 and the alanine residue at amino acid 271 are mutated to serine and valine in bp418 and bp321 mutants, respectively. atg-4.1(bp501) is likely to be null and was used in this study.
A previously published x-ray crystallography study of the HsAtg4B-LC3 complex showed that Cys-74, Asp-278, and His-280 of HsAtg4B form the catalytic triad (29). The catalytic cleft of HsAtg4B is masked by a regulatory loop (residues 259 to 262) that is lifted by Phe-119 of LC3 (29,30). HsAtg4B recognizes the Phe-119 and Gly-120 residues of LC3 using the regulatory loop and Trp-142, respectively, and this recognition is essential for the processing of LC3 by HsAtg4B (Fig. 2D). The catalytic triad, Trp-142 and the regulatory loop are conserved in ATG-4.1 (supplemental Fig. S2). His-102, Gly-266, and Ala-271 of ATG4.1, which are mutated in bp504, bp418, and bp321 mutants, correspond to H91, G258 and Ala-263 of HsAtg4B, respectively (Fig. 2D). Gly-258 and Ala-263 of HsAtg4B are located in the regulatory loop of Atg4, and both are involved in direct recognition of the side-chain of LC3 Phe-119 in the HsAtg4B-LC3 complex (29). Using the crystal structure of the HsAtg4B-LC3 complex (PDB ID 2ZZP), the C-terminal tail residues of LC3 were replaced with those of LGG-1. In this new version, ATG-4.1 will recognize Y115 (corresponding to LC3 Phe-119) and Gly-116 (corresponding to LC3 Gly-120) of LGG-1 using the regulatory loop and Trp-152 (corresponding to HsAtg4B W142) (Fig. 2E). The side-chains of Ser-266 in ATG-4.1(bp418) and Val-271 in ATG-4.1(bp321) interfere with the side-chain of Tyr-115 (Fig. 2F), destroying the proper interaction between ATG-4.1 and LGG-1 and resulting in severely impaired activity of ATG-4.1. The Pro-102 in ATG-4.1(bp504) is located in the same helix (orange) as the catalytic cysteine (Cys-74) (Fig. 2D). This mutation is likely to induce a large conformational change of the helix and thus affect the conformation of the catalytic cysteine. Alternatively, the H102P mutation may destroy the overall structure of ATG-4.1.

atg-4.2 Mutants Exhibit No Defects in Degradation of
Autophagy Substrates-atg-4.2 encodes a second C. elegans Atg4 homologue, which is more closely related to human Atg4C and Atg4D ( Fig. 2A). ATG-4.1 and ATG-4.2 display 22% overall amino acid identity and 44% similarity. The catalytic sites and the regulatory loop are conserved in ATG-4.2 (supplemental Fig. S2). Mutations in atg-4.2 were not recovered in our genetic screens. We obtained a mutant allele of atg-4.2, tm3948, which has a deletion of 498 nt that affects exons 5 and 6 ( Fig. 2C). cDNA sequencing revealed the presence of an early stop codon, resulting in a truncated protein with 266 amino acids. In atg-4.2 mutants, there were no changes in the expression levels of GFP reporters for PGL-1, SEPA-1 and SQST-1 or the corresponding endogenous proteins (Fig. 1, D, H, L), indicating that atg-4.2 is not essential for autophagic degradation of protein aggregates during embryogenesis. Loss of function of atg-4.2 had no effect on the survival of L1 larvae under starvation conditions (Fig. 1U). The degeneration of touch neurons in mec-4(u231) mutants was also not affected by the atg-4.2 mutation (Fig. 1V).
We crossed the atg-4.1::gfp transgene into atg-4.2(tm3948) mutants and found that the fluorescent intensity remained unchanged (supplemental Fig. S1). Consistent with this, quantitative real-time PCR showed that the atg-4.1 mRNA level in atg-4.2 mutants was identical to that in wild-type worms. Similarly, the atg-4.2 mRNA level showed no difference to wild type in atg-4.1-null mutants. Thus, loss of function of one atg-4 paralogue does not cause compensatory up-regulation of the other.
LGG-1-II was present at a level similar to that in wild  Glutamine residues are mutated to stop codons at amino acids 140, 160, and 385 in bp501, bp451 and bp410 mutants, respectively. The tryptophan at amino acid 347 is mutated to a stop codon in bp482. bp504 contains a histidine to proline mutation at amino acid 102. The glycine residue at amino acid 266 and the alanine residue at amino acid 271 are mutated to serine and valine in bp418 and bp321 mutants, respectively. C, schematic structure of atg-4.2. tm3948 deletes 498 nucleotides, introducing an early stop codon and creating a truncated protein with 266 amino acids. Intron-exon boundaries were confirmed by cDNA sequencing. D, crystal structure of the HsAtg4B-LC3 complex. The residues that are mutated in bp504, bp418, and bp321 are labeled. Although HsAtg4B has C74S mutation, Ser-74 is labeled as "C74" in Fig. 2, D-F. E, modeled structure of the HsAtg4B-LGG-1 complex. The LGG-1 tail (green) is recognized by the regulatory loop (blue) and W142 (blue) of HsAtg4B. F, G258S and A263V mutations, found in bp418 and bp321, respectively, disrupt the interaction between Atg4B and LGG-1.
type animals, suggesting that the defective autophagic process in atg-4.1 mutants is not due to lack of lipidated LGG-1-II.
LGG-1-I, however, was very weakly detected in bp321, bp451, and bp504 mutants and absent in bp501 mutants. In atg-4.2 mutants, LGG-1 was properly processed and no LGG-1 precursors were found. Compared with wild type animals, we consistently detected more LGG-1-II than LGG-1-I in atg-4.2 mutants. These results indicate that atg-4.1, but not atg-4.2, plays an essential role in processing LGG-1 precursors.
The distribution of LGG-1 was then examined in atg-4.1 and atg-4.2 mutants. In wild type embryos, LGG-1 forms distinct punctate structures, mostly around the ϳ100 to ϳ200 cell stage (Fig. 3, B and C). The number of LGG-1 puncta decreases as development proceeds and only a few could be detected during late embryonic stages. LGG-1-II (lipidated LGG-1) is required for the formation of LGG-1 puncta, as they are absent in atg-3 and atg-7 mutants (31). Consistent with the presence of the lipidated form of LGG-1, the temporal pattern of formation of LGG-1 puncta was identical to wild type in atg-4.1 and atg-4.2 mutant embryos (Fig. 3, D and E).

ATG-4.1 and ATG-4.2 Show Different Processing Activity toward
LGG-1-We further performed in vitro cleavage assays to determine the processing activity of ATG-4.1 and ATG-4.2 toward LGG-1. Wild type His-tagged ATG-4.1 and ATG-4.2 were expressed in bacteria and affinity-purified on Ni-NTA beads. The purified proteins were incubated with recombinant LGG-1-His 6 in reaction buffer. The cleaved form of LGG-1 (LGG-1-I), which lacks the final seven residues at its C terminus, was separated from the full-length LGG-1 precursor by SDS-PAGE. We found that both ATG-4.1 and ATG-4.2 were able to cleave full-length LGG-1, generating the shorter LGG-1-I in the assay (Fig. 4A). ATG-4.1 showed much higher LGG-1 processing activity than ATG-4.2. The LGG-1(G116A) mutant protein, in which the conserved glycine at amino acid 116 was mutated to alanine, failed to be cleaved by ATG-4.1 and ATG-4.2 (Fig. 4B).

ATG-4.1 Possesses Higher Catalytic Activity than ATG-4.2-
To further characterize the processing activity of ATG-4.1 and ATG-4.2, we determined their enzymatic kinetics by measuring levels of the Coomassie Brilliant Blue (CBB)-stained cleaved form of LGG-1.
LGG-1-His and the cleaved form of LGG-1 were separated by SDS-PAGE and then stained with CBB. The density of bands was measured to determine the amount of products generated from the substrate. This assay has been used previously to determine the kinetic differences among mammalian Atg4 homologues (17).
We first determined the optimal time for the kinetic analysis of ATG-4.1 and ATG-4.2. Under the given experimental conditions, about 50% of the LGG-1-His substrate was cleaved by ATG-4.1 in 45 min (Fig. 4C). The cleavage reaction became linear after 5 min, so this was the time point we selected for further experiments on ATG-4.1 (Fig. 4D). The catalytic protease reaction of ATG-4.2 was much slower than ATG-4.1. The substrate processing of ATG-4.2 occurred at very low levels, but persisted even up to 300 min (Fig. 4, E and F). For further kinetic analysis, we determined the activity of ATG-4.2 at 300 min.
We next incubated decreasing concentrations of the LGG-1-His substrate with fixed amounts of ATG-4.1 and ATG-4.2 at the chosen reaction time (Fig. 4, G and H). The amount of product generated from the substrate was measured, and the reaction velocity for each substrate concentration point was calculated. The values of K m and k cat /K m were then determined (Fig. 4, I and J). ATG-4.1 and ATG-4.2 bound to LGG-1/Atg8 with similar affinities (1/K m ) (Fig. 4K). However, ATG-4.1 displayed about 100-fold higher catalytic efficiency (k cat /K m ) in cleaving the LGG-1 precursor compared with ATG-4.2 (Fig.  4K). The catalytic efficiency of ATG-4.1 for LGG-1 was comparable with mammalian Atg4A toward its most highly preferred substrate GATE-16 (17). The catalytic efficiency of ATG-4.2 was in the same range as mammalian Atg4C and Atg4D (17). ATG-4.2 did not reduce the catalytic efficiency of ATG-4.1 in the mixture reaction (supplemental Fig. S3).
Epistasis Analysis between atg-4.1 and Other Autophagy Mutants-By analyzing a series of genetic phenotypes, including the formation, morphology and distribution of LGG-1 The cleaved form of LGG-1 lacks the final seven residues as well as the His-tag at its C terminus, resulting in a lower molecular weight than LGG-1-His 6 when separated by SDS-PAGE. Bacterially expressed His-tagged LGG-1, ATG-4.1, and ATG-4.2 were purified with Ni-NTA. LGG-1-His 6 substrate (0.5 mg/ml) was incubated with vehicle control (first lane), ATG-4.1 (0.0125 mg/ml, second lane), or ATG-4.2 (0.0125 mg/ml and 0.025 mg/ml, third and fourth lanes), in a volume of 20 l for 1.5 h at 37°C. The mixture was resolved using SDS-PAGE followed by CBB staining after the reaction was stopped. B, LGG-1(G116tA)-His 6 substrate (0.5 mg/ml), in which the conserved glycine is mutated to alanine, was incubated with vehicle control (first lane), ATG-4.1 (0.0125 mg/ml, second lane), or ATG-4.2 (0.025 mg/ml, third lane), under the same reaction conditions as in Fig. 4A. C, ATG-4.1 (0.0125 mg/ml) was incubated with the substrate LGG-1-His 6 (0.5 mg/ml) in a volume of 20 l for the indicated time. The mixture was resolved using SDS-PAGE followed by CBB staining after the reaction was stopped. D, gel was analyzed by Image J and the densities of both substrate and product were measured. About 50% LGG-1-His substrate was cleaved by ATG-4.1 in 45 min. E, ATG-4.2 (0.025 mg/ml) was incubated with the substrate LGG-1-His 6 (0.5 mg/ml) in a volume of 20 l for the indicated time. The mixture was resolved using SDS-PAGE followed by CBB staining. F, gel was analyzed by Image J and the densities of both substrate and product were measured. G-J, decreasing concentrations of the substrate LGG-1-His 6 were incubated with fixed amounts of ATG-4.1 (0.0125 mg/ml) (G) and ATG-4.2 (0.025 mg/ml) (H) for 5 min and 300 min, respectively. The mixture was resolved using SDS-PAGE followed by CBB staining. The initial velocity (V, y axis (mM/s), i.e. the change in LGG-1 product concentration/s) was plotted against each substrate concentration (S (mM), x axis) for each reaction. The curves were fitted by the non-linear regression method (GraphPad Prism 5.00). K, kinetic parameters of ATG-4.1 and ATG-4.2 toward LGG-1. The values of V max (mol liter Ϫ1 s Ϫ1 ) and K m (mol/liter) were derived from curves fitted by the non-linear regression method (GraphPad Prism 5.00). The catalytic constant, k cat (s Ϫ1 ), was calculated by dividing V max by the enzyme concentration of each reaction. puncta and SEPA-1 aggregates, autophagy genes have been placed in a hierarchical order in the aggrephagy pathway in C. elegans (31,32). We thus performed epistasis analysis between atg-4.1 and other autophagy mutants. In epg-1/Atg13 mutants, LGG-1 forms large aggregates in a few cells but is absent in most embryonic cells (Fig. 6, A and B) (31,33). We constructed atg-4.1; epg-1 double mutants and found that LGG-1 accumulated into large puncta, resembling epg-1 single mutants (Fig. 6, C and D). We also examined phenotypes in atg-4.1; epg-6 double mutants. epg-6 is involved in the progression of omegasomes to autophagosomes (32). In epg-6 mutants, SEPA-1 aggregates and LGG-1 puncta accumulate and form clusters and are largely colocalized (Fig. 6, E and F, supplemental Fig. S4). In atg-4.1 mutants, SEPA-1 aggregates and LGG-1 puncta are small and spherical shaped and do not colocalize (Fig. 6G, supplemental Fig. S4). In atg-4.1; epg-6 double mutants, SEPA-1 formed small, spherical aggregates that were separable from LGG-1 puncta, resembling those in atg-4.1 mutants (Fig. 6H, supplemental Fig. S4). These results indicate that atg-4.1 acts upstream of epg-6 in the autophagic pathway by which PGL granules are degraded.

DISCUSSION
The physiological function of the four mammalian homologues of the single yeast Atg4 protein remains poorly understood. The multiple homologues might act redundantly, function in a tissue-specific fashion, or be expressed at different developmental stages; alternatively they may have evolved to accomplish other functions distinct from autophagy. The substrate preferences of four mammalian Atg4s toward different Atg8s have been characterized in vitro (11,13,14,17). HsAtg4B cleaves all Atg8s (13)(14)(15). HsAtg4A efficiently cleaves GATE-16, but not other Atg8s, while HsAtg4C and HsAtg4D both have minimal activity (11,17). HsAtg4B is likely to be the principle Atg4 homologue in processing and deconjugation of Atg8s (15,28). Targeted deletion of Atg3, Atg5, Atg7, and Atg16L in mice causes lethal phenotypes during the neonatal starvation period (21)(22)(23)(24). Tissue-specific knock-out of Atg5 and Atg7 has been shown to lead to neurodegeneration, hepatomegaly, and many other defects (22,23). Atg4C-and Atg4B-deficient mice exhibit largely normal development, indicating that Atg4s act redundantly in processing and deconjugation of Atg8s (18 -20). Although HsAtg4D is a poor processing and deconjugation enzyme, truncation of the N-terminal 63 amino acids by caspase-3 increases its activity against GABARAP-L1, thus linking mitochondrial dysfunction with apoptosis (34).
C. elegans has two Atg4 orthologues, atg-4.1 and atg-4.2. ATG-4.1 is closer to HsAtg4B and HsAtg4A, while ATG-4.2 is more related to HsAtg4C and HsAtg4B. Here we showed that mutations in atg-4.1, but not atg-4.2, cause defects in degradation of various autophagy substrates.
LGG-1 precursors are absent in wild type and atg-4.2 embryos but dramatically accumulate in atg-4.1 mutants. In vitro cleavage assays showed that ATG-4.1 has ϳ100-fold higher LGG-1 processing activity than ATG-4.2. The LGG-1 puncta formed by the conjugated form of LGG-1 are detected in atg-4.1 and atg-4.2 single mutants, but are absent in the atg-4.1; atg-4.2 double mutants. In atg-4.1 single mutants, the lipidated form of LGG-1 (LGG-1-II) accumulates at a level comparable to that in wild type embryos, indicating that ATG-4.2 processes LGG-1 precursors in vivo. However, in atg-4.1 mutants, unlipidated LGG-1 (LGG-1-I) is present at very low levels, suggesting that ATG-4.2 possesses a very weak deconjugating activity toward LGG-1-PE, or that ATG-4.2 is sequestrated by nascent LGG-1 precursors because it processes LGG-1 inefficiently. The normal lipidated LGG-1 level in atg-4.1 mutants could also be resulted from the accumulation of nonfunctional autophagosomes that fail to mature, subsequently affecting the LGG-1 turnover. We consistently detected more LGG-1-II in atg-4.2 mutants. One explanation for this could that in wild type embryos, LGG-1-I is bound by ATG-4.2 and is unavailable for lipidation.
Both the processing and deconjugating activities of Atg4 are required for normal progression of autophagy in yeast (4). A recent study showed that yeast Atg4 also recycles Atg8-PE pro- duced on inappropriate membranes to maintain a cytoplasmic pool of unlipidated Atg8 that can undergo lipidation and participate in autophagosome formation at the pre-autophagosomal structure (PAS) (35). The mutant form of LGG-1, in which the conjugated glycine residue is directly exposed, rescues defective degradation of protein aggregates in atg-4.1 mutants. atg-4.1 mutant animals expressing this mutant form of LGG-1 show no significant changes in levels of LGG-1-II, but accumulate unlipidated LGG-1, which is absent in atg-4.1 mutants, implying that maintenance of an unlipidated pool of LGG-1 is essential for autophagy flux. The LGG-1(G116 end) transgene also rescues, although not completely, defective degradation of protein aggregates in atg-4.1; atg-4.2 double mutants. In ⌬atg4⌬atg8 yeast cells, expressing Atg8G116 also partially suppresses the defect in autophagy (35). Thus, the deconjugation activity is not completely required for autophagic flux. Our results demonstrate that ATG-4.1 and ATG-4.2 act redundantly, but ATG-4.1 plays a principle role in processing and deconjugation of LGG-1 in the autophagy pathway.