Overlap in roles of human ATG4 enzyme types in cell recycling and protein processing

To investigate the overall effect of ATG4B loss on the processing of endogenous LC3B, we treated ATG4B KO and wild-type counterpart (control) HAP1 and HeLa cells with starvation medium (EBSS) or Torin1 in combination with baf A1 to assess autophagic flux, and analyzed cell lysates by western blotting. As shown in Figure 1(b), treatment of wild-type cells with baf A1 resulted in expected increased levels of the lower band of LC3B corresponding to the lipidated form, LC3B-II (LC3B–PE). There was a further increase in these lipidation levels upon co-treatment with Torin1, consistent with an increase in autophagic flux. Interestingly, starvation did not appear to stimulate further lipidation of LC3B, possibly since it was less efficient than Torin1 at causing de-phosphorylation of the upstream kinase ULK1 at its inhibitory phosphorylation site (Ser757) [22] (Figure 1(b)). Strikingly, in ATG4B KO cells we only observed a single specific band of LC3B that showed a similar migration to LC3B-II (Figure 1(b)). Note that the band marked by an asterisk at the height of LC3B-I in HeLa ATG4B KO cells was not detected using other LC3B antibodies and thus was deemed to be non-specific (see Figure S1B). The specific band seen in ATG4B KO cells did not accumulate further in response to Torin1 or baf A1 treatment, despite Torin1 treatment efficiently leading to dephosphorylation of ULK1. Cells lacking ATG4B also showed an accumulation of SQSTM1 and a reduced clearance of this protein upon Torin1 treatment, suggestive of an autophagy defect (Figure 1(b)). As human pro-LC3B was recently described to have a similar migration rate to lipidated LC3B by SDS-PAGE [23], we suspected that the single lower band of LC3B might correspond to pro-LC3B and thus potentially explain the impairment in autophagy.

In order to distinguish between endogenous pro-LC3B and LC3B-II, we devised an assay in which lysates of cells treated with Torin1 and baf A1 were subject to treatment with recombinant ATG4B or PLD (phospholipase D) prior to western blot analysis. Because PLD cleaves a bond that is only present in lipidated forms of LC3/GABARAP [24], a migration shift upon PLD treatment is indicative of the presence of the lipidated species (Figure 1(c)). As a control, recombinant ATG4B can cleave both pro- and lipidated LC3/GABARAP revealing any potential migration shift from both of these forms (Figure 1(c)). The predominant LC3B band in HeLa ATG4B KO cells was completely resistant to PLD treatment but underwent a migratory shift with ATG4B treatment, in contrast to the corresponding band that was sensitive to both PLD and ATG4B in wild-type cells (Figure 1(d)). This demonstrates that ATG4B KO cells accumulate pro-LC3B rather than LC3B-II, an observation that was also confirmed in HAP1 cells treated with baf A1 (Figure S1C). Therefore, ATG4B is essential for the priming of pro-LC3B and its subsequent lipidation, explaining the absence of LC3B puncta in ATG4B KO cells.

We could also determine the effect of ATG4B loss on GABARAP subfamily isoforms in the same set of experiments, using specific antibodies to detect endogenous GABARAP, GABARAPL1 and GABARAPL2. We observed that total levels of free GABARAP and GABARAPL1 were dramatically reduced in ATG4B KO cells compared to wild-type cells (Figure 1(b)), consistent with a recently discovered role of the C-terminal LC3 interacting region (LIR) of murine ATG4B in stabilizing the non-lipidated forms of these proteins and reducing their proteasomal degradation [25]. Our data suggest this function is conserved in human cells, since proteasome inhibition by treatment with the compound MG132 could also raise GABARAP protein levels in ATG4B KO HeLa cells (Figure S1D). In respect to lipidation, we could observe an accumulation of lower bands of all GABARAP isoforms in wild-type (control) HAP1 and HeLa cells in response to Torin1 and baf A1 treatment (Figure 1(b)), consistent with all of these proteins being involved in autophagy. Despite their levels generally being reduced in ATG4B KO cells, the lower bands of GABARAPL1 and GABARAPL2 still accumulated in response to treatment. To determine whether these lower bands corresponded to the lipidated forms of GABARAP proteins, we also detected endogenous GABARAP isoforms in PLD assay samples (Figure 1(d)). Indeed, the lower bands of GABARAPL1 and GABARAPL2 were sensitive to PLD treatment in both control and ATG4B KO HeLa cells, confirming that they corresponded to the lipidated forms of the proteins. This result could only be determined when the assay buffer contained the irreversible ATG4 inhibitor N-ethylmaleimide (NEM) [9] otherwise the lower band was absent or reduced by heat treatment alone presumably due to processing by endogenous ATG4 activity in the lysate. In fact, it proved essential to include NEM in the lysis buffer for all western blot experiments assessing GABARAP and GABARAPL1 lipidation, since the lipidated forms of these proteins were particularly unstable in wild-type cell lysates lacking NEM (Figure S1E).

Collectively, our data show that human cells lacking ATG4B fail to form lipidated LC3B due to a defect in the initial priming/activation of LC3B, and thus accumulate unprocessed pro-LC3B. However the presence of lipidated GABARAPL1 and GABARAPL2 in ATG4B KO cells suggests that these isoforms can be processed independently of ATG4B.

To characterize the relative SDS-PAGE migration of each form (pro vs. -I vs. -II) of LC3/GABARAP isoform, and as an alternative approach to investigate LC3/GABARAP processing by ATG4B without relying on C-terminal tags, we also generated full-length LC3/GABARAP constructs tagged at the N terminus with 3xFLAG with their native C-termini (WT), uncleavable mutants (GA) and mutants corresponding to the primed form with their active glycine exposed (G) (Figure 2(c)). When these constructs were expressed in wild-type HeLa cells treated with Torin1 and baf A1, both WT and G mutants underwent similar levels of lipidation (formation of band with lower apparent molecular weight) suggesting that priming of WT constructs occurred normally prior to their lipidation and did not dramatically affect the level of lipidation (Figure 2(d), S2). However, in ATG4B KO cells, WT constructs were present as bands with similar motility to GA mutants suggesting they were mainly unprocessed (Figure 2(d), S2). In these cells, small amounts of LC3A, GABARAP and GABARAPL1 lipidation were observed at high contrast for WT constructs, but at reduced levels compared to G mutants (Figure 2(d)). In this particular experiment, it was difficult to resolve the different forms of 3xFLAG-GABARAPL2.

Altogether, these results suggest that most LC3 subfamily isoforms absolutely require ATG4B for priming (LC3B, LC3B2 and LC3C). Conversely, priming of other LC3/GABARAP isoforms (LC3A and GABARAP subfamily) can occur in ATG4B-deficient cells but is markedly reduced in efficiency, thus limiting the amount of lipidated LC3/GABARAP that can accumulate.

Next we decided to determine whether ATG4B plays an essential role in the autophagy pathway after executing pro-LC3B priming. To achieve this, we used a mutant form of LC3B that terminates at the C-terminal active glycine (G120) and thus bypasses the ATG4-dependent priming step and corresponds to primed/activated LC3B. Comparing the effects on autophagy of expression of LC3B-G120 in wild-type versus ATG4B KO cells should therefore reveal any autophagy defect caused specifically by the loss of ATG4B in steps subsequent to priming.

Because ATG4-mediated delipidation has previously been implicated in autophagosome closure, we next assessed whether LC3B-positive vesicles formed in ATG4B-deficient cells represent fully sealed structures using a protease protection assay (Figure S5C). In this assay, unpermeabilized cell homogenates and an equivalent permeabilized sample are subject to trypsin treatment followed by western blot analysis. Proteins that are specifically protected from degradation in the unpermeabilized sample but become sensitive to degradation in the presence of detergent are considered to be protected by membrane and thus localized within sealed vesicles or organelles. This assay was able to distinguish between proteins on the inside and outside of a sealed double-membrane organelle, since the mitochondrial matrix protein PDHA1 (pyruvate dehydrogenase E1 alpha 1 subunit) was only efficiently degraded by trypsin in the presence of detergent, while the cytosol-facing outer mitochondrial membrane protein MFN2 (mitofusin 2) was degraded completely in both presence and absence of detergent. The lysosome membrane protein LAMP1 served as a loading control in this assay since it was relatively resistant to trypsin. We found that a similar fraction of transiently transfected 3xFLAG-LC3B-G120 was resistant to trypsin degradation in the absence of detergent in both HAP1 control and ATG4B KO cells treated with Torin1 and baf A1 (Figure S5C). This suggests that closure of autophagosomes is not impaired in the absence of ATG4B-mediated LC3B delipidation.

Altogether, we find that bypassing pro-LC3B processing by high expression of pre-primed LC3B rescues the autophagy defect caused by loss of ATG4B and does not produce any detectable abnormalities in LC3B localization, SQSTM1 degradation or autophagosome closure. This suggests that ATG4B does not play an essential role in autophagy downstream of LC3/GABARAP priming.

We confirmed by immunocytochemistry that ATG4A/B DKO HeLa cells are impaired in SQSTM1 delivery to lysosomes (Figure S6B), showing the same defect as ATG4B KO cells (Figure 5(a)). Stable expression of both GFP-LC3B-G120 and the equivalent bypass mutant of GABARAPL1 (GFP-GABARAPL1-G116) in ATG4A/B DKO HeLa cells could rescue SQSTM1 delivery to LAMP1-positive structures (Figure 7(c)). Rescue of basal SQSTM1 turnover could also be observed in these cells, since by western blot we could detect reduced levels of SQSTM1 similar to those in control cells, in contrast to untransduced ATG4A/B DKO cells which exhibited strongly elevated basal levels of SQSTM1 (Figure 7(d)). This rescue of basal SQSTM1 turnover was specifically due to bypass construct expression, because knockdown of the bypass construct using siRNA targeting GFP resulted in an elevation of SQSTM1 expression levels approaching that seen in untransduced ATG4A/B DKO cells (Figure 7(d)). Further, rescue of SQSTM1 turnover by GFP-LC3B-G120 expression in ATG4A/B DKO cells was dependent on LC3/GABARAP lipidation and autophagy, since suppression of ATG3 and ATG7 in these cells led to a strong accumulation of SQSTM1 similar to the effect of GFP knockdown (Figure 7(e)). Collectively these data suggest that both ATG4A and ATG4B are dispensable for basal and Torin1-induced autophagy downstream of LC3/GABARAP priming.

We also separately generated HeLa ATG4A/B DKO cells stably expressing GFP-LC3B-G120 or GFP-GABARAPL1 under control of a weak (PGK) promoter, to determine whether expressing bypass mutants at lower levels would be sufficient to rescue autophagy (Figure S7A). These cells could form GFP puncta with the expected localization pattern (Figure S7B), however they were notably less efficient at reducing basal SQSTM1 level (Figure S7A) and rescuing SQSTM1 delivery to lysosomes (Figure S8) compared to equivalent ATG4A/B DKO cells expressing higher levels of GFP-LC3B-G120 under control of the CMV promoter. These results argue that the supply of primed LC3/GABARAP (i.e. through recycling) might be a stronger requirement for autophagy than ATG4 delipidation activity per se. Establishing whether delipidation by ATG4 isoforms is necessary for functional aspects of autophagy beyond regulating the level of primed LC3/GABARAP (e.g. autophagosome-lysosome fusion) requires a model system in which primed LC3/GABARAP supply is plentiful. Therefore, we continued characterizing ATG4-deficient cells expressing CMV promoter-driven levels of bypass mutants for the remainder of our study.

To observe the effect of loss of each individual ATG4 isoform on GABARAPL2 priming, we expressed 3xFLAG-GABARAPL2-GFP in each of the generated HeLa cell lines following transfection with siRNA ()). This revealed that loss of each additional ATG4 isoform led to a reduction in priming, but that a complete loss of priming could only be observed upon loss of all ATG4 isoforms. This effect could also be observed for endogenous GABARAPL1 in the same cell lysates, assuming that loss of lipidated GABARAPL1 in response to treatment was a result of impaired priming ()). Together these data show that all human ATG4 isoforms contribute to the priming of LC3/GABARAP isoforms, because complete loss of LC3/GABARAP priming and lipidation in cells could only be achieved upon loss of all ATG4 isoforms. Figure 8(c Figure 8(c

Because CASP3 promotes ATG4D activity in vitro [18], we performed knockdown of CASP3 by RNAi to determine whether the residual activity of ATG4C and ATG4D detected in HeLa cells lacking ATG4A/B was dependent on CASP3 expression (Figure 8(d)). Despite efficient knockdown, we could not detect a reduction in treatment-induced GABARAPL1 or GABARAPL2 lipidation caused by CASP3 suppression in ATG4A/B DKO and ATG4A/B/C TKO cells, in contrast to knockdown of ATG4D. Therefore, the residual priming activity of ATG4C and ATG4D detected in HeLa cells lacking ATG4A/B does not appear to be dependent on CASP3 expression.

Finally, we decided to determine if delipidation by any ATG4 isoform plays an important role in aspects of autophagy beyond LC3/GABARAP recycling. Since we already established that ATG4A/B-mediated delipidation is not required for autophagosome closure and lysosome fusion when primed LC3/GABARAP is expressed at high levels (and(c,d)), we decided to investigate the consequences of further depletion of ATG4C and ATG4D in the same scenario. Figures 6 7

To determine whether delipidation by any ATG4 isoform is important for basal autophagy, we monitored endogenous SQSTM1 protein level in ATG4A/B DKO HeLa cells stably expressing high levels of GFP-LC3B-G120 following siRNA-mediated knockdown of ATG4C and ATG4D (Figure 9(b,c)). As described above, stable expression of CMV promoter-driven GFP-LC3B-G120 in ATG4A/B DKO HeLa cells efficiently rescues the defect in endogenous SQSTM1 turnover. When quantified from 4 experiments (Figure 9(c), representative experiment shown in Figure 9(b)), loss of ATG4A/B resulted in a 5.23 ± 0.95 (mean ± SEM) fold increase in basal SQSTM1 protein level relative to wild-type control cells (P < 0.0001). Upon stable expression of GFP-LC3B-G120 in ATG4A/B DKO cells, SQSTM1 protein reverted to a level that was not significantly different to wild-type cells (1.39 ± 0.12, P = 0.890). Knockdown of the rescue construct using siRNA targeting GFP dramatically increased SQSTM1 protein level to 3.80 ± 0.16 (P < 0.0001, versus SQSTM1 level of 1.39 ± 0.12 in equivalent RLUC siRNA-transfected cells), signifying a disruption in autophagy. In contrast, levels of SQSTM1 upon knockdown of ATG4C (1.43 ± 0.16) or ATG4D (1.86 ± 0.13) alone were not significantly raised (P > 0.999 and P = 0.702 respectively, versus equivalent RLUC siRNA-transfected cells). Only upon combined knockdown of ATG4C and ATG4D was there a mild increase in SQSTM1 level (2.30 ± 0.16, P = 0.0337 versus equivalent RLUC siRNA-transfected cells) and this was substantially less that the raised level of SQSTM1 upon GFP knockdown (P = 0.0002, SQSTM1 level of 3.80 ± 0.16). From this, we conclude that delipidation by ATG4 isoforms is not essential for basal autophagy when primed LC3B is overexpressed. If the opposite were true, we would expect to observe a much more dramatic increase in SQSTM1 level upon loss of all ATG4 isoforms in cells stably expressing GFP-LC3B-G120, e.g. similar to the effect of GFP knockdown.

We also examined the localization of endogenous SQSTM1 in response to Torin1 and baf A1 treatment in HeLa ATG4A/B DKO cells stably expressing GFP-LC3B-G120 following knockdown of ATG4C and ATG4D (Figure 9(d)). This revealed that GFP- and SQSTM1 double-positive structures can accumulate in response to treatment and localize to LAMP1-positive lysosomes despite loss of all ATG4 delipidation activity, overall suggesting that delipidation by ATG4 isoforms is not functionally required for autophagosome-lysosome fusion.

Lastly, we decided to determine whether ATG4-mediated delipidation was necessary to prevent mislocalization of LC3/GABARAP, since this is known to be an important function of Atg4 in yeast [10–12]. We assessed the basal localization of GFP-LC3B-G120 and GFP-GABARAPL1-G116 stably expressed at high levels in wild type HeLa cells compared to cells lacking all ATG4 isoforms (Figure S10). Upon loss of ATG4-mediated delipidation, we were unable to detect any change in the localization of GFP-LC3B-G120 or GFP-GABARAPL1-G116 in relation to the endoplasmic reticulum, early endosomes, Golgi or lysosomes. We noted a partial Golgi localization of GFP-GABARAPL1-G116, but this was also observed in HeLa control cells and was therefore unrelated to delipidation.

Together we conclude that while all ATG4 isoforms contribute to the essential priming of LC3/GABARAP, delipidation by ATG4 isoforms (when primed LC3/GABARAP supply is plentiful) does not play a critical functional role in autophagosome-lysosome fusion and preventing LC3/GABARAP mislocalization in human cells.

HAP1 control (Horizon Genomics, C631) and HAP1 ATG4B KO cells (Horizon Genomics, HZGHC001241c011) were grown in IMDM containing 25 mM HEPES and L-Glutamine (ThermoFisher Scientific, 12440061) supplemented with Penicillin-Streptomycin (ThermoFisher Scientific, 15140122; 100 U/ml each) and 10% heat-inactivated FBS (ThermoFisher Scientific, 10500064). HeLa and HEK293T cells were grown in DMEM high glucose with GlutaMax and 1 mM pyruvate (ThermoFisher Scientific, 31966021), penicillin-streptomycin (100 U/ml) and 10% FBS (ThermoFisher Scientific, 10270106). All cells were maintained at 37°C, 5% CO₂ and passaged at a dilution of 1:5–1:30 every 2–3 days. Cell line authentication performed by Eurofins was used to confirm the HeLa identity of main cell lines used in this study (HeLa control, ATG4B KO, ATG4A/B DKO c25, ATG4A/B/C TKO c22).

Quantities listed are per well for cells grown in 12-well plate format, and were scaled according to volume of growth medium used in other formats. HAP1 cells were transfected using Polyethyleneimine ‘MAX’ (Polysciences, 24765; stock dissolved at 1 mg/ml in sterile water and filter sterilized) at 3 µg per 1 µg of DNA, and 1 µg total DNA per 1 ml of growth medium. Mixtures of DNA and transfection reagent were diluted in 100 µl of IMDM without supplements, incubated for 15 min at room temperature, and then added directly to cells grown in complete IMDM. Cells were returned to incubator and medium was replaced after 4 h. HeLa cells were transfected with 250 ng DNA using 1.5 µl jetPRIME reagent (Polyplus-transfection, 114–15) and 50 µl jetPRIME buffer per 1 ml of growth medium. Experiments were performed at 24 h post-transfection.

Gene-specific sgRNA sequences for human ATG4 isoforms were identified using Sanger CRISPR Finder tool (http://www.sanger.ac.uk/htgt/wge/find_crisprs↗) [35]. Sequences were chosen that targeted an early constitutive coding exon or residues close to the catalytic cysteine, and had the lowest possible probability of off-target cleavage in other gene regions. Oligonucleotides corresponding to desired guide sequences (listed in Table S1) were phosphorylated, annealed and cloned into BbsI-digested pSpCas9(BB)-2A-Puro (PX459) V2.0 (Addgene, 62988; Zhang lab) and delivered by transient transfection to generate knockout HeLa cell lines [36]. HeLa cells grown in 24-well plate format were transfected with 250–500 ng of DNA per well using PEI or jetPRIME. After 24 h, transfected cells were enriched by selection with 1 µg/ml puromycin for 2–3 days, following which puromycin was removed. Cells were then transferred to 10 cm dishes to recover for several days. For generation of HeLa ATG4A/B DKO and ATG4A/B/C TKO cells, the procedure was repeated 3 times in an attempt to increase the efficiency of gene disruption on multiple alleles. Single-cell clones were then isolated by limiting dilution in 96-well plates and grown for 2 weeks. Clones were screened for loss of target protein expression by western blot after expansion into 12-well plates. Positive clones were expanded and further validated by sequencing PCR-amplified genomic DNA of the target locus to confirm gene editing of all detected alleles. Loss of plasmid expression was demonstrated by confirming that clones had regained sensitivity to puromycin treatment. To generate HeLa control cells, unmodified HeLa cells were transfected with PX459 encoding a non-targeting sgRNA [37], selected with puromycin for 2 days, and then passaged as a pool for at least three weeks prior to use in experiments.

HEK293T cells grown in a 6-well plate format were co-transfected with 900 ng psPAX2 (Addgene, 12260; Trono lab), 100 ng pMD2.G (Addgene, 12259; Trono lab) and 1 µg of transfer plasmid per well using X-tremeGENE HP (Roche, 6366244001) at 2 µl per µg of total DNA. At 18 h post-transfection, medium was replaced with complete IMDM. Viral supernatant was harvested at 48 and 72 h post-transfection, filtered through a 0.22-µm low protein-binding syringe filter and frozen at −80°C. Target cells seeded at a medium density in 24-well plates were infected overnight with 300 µl of filtered viral supernatant containing polybrene at a final concentration of 8 µg/ml. Cells were selected at 48 h post-infection with 1 µg/ml puromycin, expanded into 10-cm dishes and passaged once before puromycin was excluded from growth medium. Experiments were performed without puromycin present using pooled populations of transduced cells.

HeLa cells grown in a 12-well format were transfected with a total of 500 ng MISSION esiRNA (Sigma) per well diluted in 100 µl jetPRIME buffer using 3 µl of jetPRIME reagent. This mixture was added directly to 1 ml of growth medium present in the well. In experiments where 2 targets were depleted simultaneously, the amount of esiRNA against each target was halved to keep the total amount of esiRNA constant. In these experiments, negative control esiRNA targeting RLUC was used to keep the esiRNA dosage constant in samples where only one target was depleted. Experiments were performed at 48 h after siRNA transfection. In siRNA experiments involving cDNA transfection, medium was changed at 24 h following siRNA transfection, and DNA transfection was subsequently performed as described. Catalog numbers and targets of esiRNAs are as follows: RLUC (Renilla luciferase; Sigma, EHURLUC), human ATG4C (Sigma, EHU060781), human ATG4D (Sigma, EHU035721), EGFP (enhanced GFP; Sigma, EHUEGFP), human CASP3/caspase-3 (Sigma, EHU085021), human ATG3 (Sigma, EHU015141), human ATG7 (Sigma, EHU092581).

Primary antibodies and dilutions used for western blot in this study were as follows: ATG3 (Abcam, ab108251; 1:1,000), ATG4A (Cell Signaling Technology, 7613; 1:1,000), ATG4B C terminus (Cell Signaling Technology, 5299; 1:1,000), ATG4B N terminus (Sigma, A2981; 1:1,000), ATG4C (Cell Signaling Technology, 5262; 1:200–1:500), ATG4D (Proteintech, 16924–1-AP; 1:1,000), ATG7 (Abcam, ab52472; 1:2,000), Beta-actin (Sigma, A1978; 1:4,000), CASP3/caspase-3 (Santa Cruz Biotechnology, sc-56053; 1:200), FLAG biotin-conjugated (Sigma, F9291; 1:1,000), GABARAP (Abgent, AP1821a; 1:1,000), GABARAPL1 (Proteintech, 11010–1-AP; 1:1,000), GABARAPL2 (Abcam, ab126607; 1:500), GFP (Clontech, 632381; 1:2,000), LAMP1 (BD Biosciences, 611043; 1:500), LC3A/B (Cell Signaling Technology, 12741; 1:500), LC3B (Sigma, L7542; 1:1,000 was used for LC3B western blotting unless otherwise stated) and (Cell Signaling Technology, 3868; 1:1,000 was used for western blotting in Figure S1B only), MFN2/mitofusin 2 (Abcam, ab56889; 1:500), SQSTM1/p62 (Sigma, P0067; 1:1,000), PDHA1/pyruvate dehydrogenase E1 alpha 1 subunit (Abcam, ab110330; 1:1,000), ubiquitin (Santa Cruz Biotechnology, sc-8017; 1:500), ULK1 (Cell Signaling Technology, 4733; 1:500), p-ULK1 S757/phospho-ULK1 Ser747 (Cell Signaling Technology, 6888; 1:1,000), VCL/vinculin (Abcam, ab129002; 1:1,000). Secondary antibodies used were: goat anti-mouse/rabbit IgG HRP (Cell Signaling Technology, 7076/7074; 1:5,000), IRDye 800CW goat anti-rabbit (LI-COR Biosciences, 926–32211; 1:15,000 with 0.02% SDS) and IRDye 680LT goat anti-mouse (LI-COR Biosciences, 926–68020; 1:25,000 with 0.02% SDS).

For immunocytochemistry, the following antibodies and dilutions were used: CANX/calnexin (BD Biosciences, 610523; 1:50), EEA1 (BD Biosciences, 610456; 1:200), GABARAPL1 (Proteintech, 11010–1-AP; 1:200), GFP (Abcam, ab290; 1:500), GOLGA2/GM130 (BD Biosciences, 610822; 1:500), LAMP1 (BD Biosciences, 555798; 1:1,000), LC3B (Cell Signaling Technology, 3868; 1:200), SQSTM1/p62 (Sigma, P0067; 1:1,000). Secondary fluorescent antibodies used were: Alexa Fluor 488/568 goat anti-rabbit IgG (H + L; ThermoFisher Scientific, A-11008/11011; 1:400), Alexa Fluor 647 goat anti-mouse IgG (H + L; ThermoFisher Scientific, A-21235; 1:400).

Torin1 (Merck-Millipore, 475991) was dissolved in DMSO at a final stock concentration of 250 µM. Bafilomycin A₁ from Streptomyces griseus (Sigma, B1793) was dissolved in DMSO at a final stock concentration of 10 µM. Recombinant GST-ATG4B WT was described previously [38]. Other reagents were as follows: PLD/phospholipase D from Streptomyces chromofuscus (Sigma, P0065), EBSS containing calcium and magnesium (ThermoFisher Scientific, 24010–043). For the protease protection assay, Trypsin from porcine pancreas was used (Sigma, T4799).

Expression constructs of tagged WT and mutant human LC3/GABARAP isoforms were generated using Gateway cloning technology (Life Technologies). Entry clones were first generated by PCR using complementary or mutagenic primers with attB sequence overhangs (listed in Table S2) followed by recombination with pDONR223 using BP clonase II enzyme mix (ThermoFisher Scientific, 11789020) [39]. To generate N-terminal 3xFLAG-tagged constructs, entry clones containing stop codons were recombined with the destination vector pCMV-tripleFLAG-Gateway [40] using LR clonase II enzyme mix (ThermoFisher Scientific, 11791020). To make transient expression constructs of GFP-LC3/GABARAP, pDEST-CMV-N-EGFP was used as the destination vector. For generating 3xFLAG-ATG8-EGFP double-tagged constructs, entry clones lacking stop codons were recombined with pDEST-CMV-3xFLAG-gateway-EGFP. To generate lentiviral transfer vectors for the production of cell lines stably expressing CMV promoter-driven GFP-tagged ATG8 isoforms, entry clones containing stop codons were recombined with pLVpuro-CMV-N-EGFP.

Novel destination vector constructs were generated as follows. EGFP was PCR amplified from pEAK13-EGFP using the primers 5ʹ-GFP-Kpn1 (5ʹ-acgtaGGTACCGCCACCATGGTGAGCAAGGGCGAGGAGCTGTTC-3ʹ) and 3ʹ-GFP-HindIII (5ʹ-acgtaAAGCTTCTTGTACAGCTCGTCCATGCCGAGAGTGATC-3ʹ), cut and ligated into pCMV-tripleFLAG-Gateway using KpnI and HindIII enzymes to generate pCMV-EGFP-gateway. A portion of the CMV promoter and EGFP was then PCR amplified using the primers 5ʹ-CMV-SacI (5ʹ-gacGAGCTCGTTTAGTGAACCGTC-3ʹ) and 3ʹ-EGFP-SacI (5ʹ-gacGAGCTCGTCCATGCCGTGAGTGATC-3ʹ), cut and ligated into Gateway® pcDNA™-DEST53 (ThermoFisher Scientific, 12288015) using SacI digestion to generate pDEST-CMV-N-EGFP. The EGFP tag and gateway cassette portion of pDEST-CMV-N-EGFP was PCR amplified using the primers 5ʹ-EGFP-NheIPacIXbaI (5ʹ-GACGCTAGCTTAATTAAGACTCTAGAGCCACCATGGTGAGCAAG-3ʹ) and 3ʹ-attR2-BstBI (5ʹ-gacTTCGAATCACACCACTTTGTACAAGAAAG-3ʹ) and cloned into pLV-CMV-y/hNubI-tripleFLAG-linker-Gateway- PuroR using NheI and BstBI digestion to generate pLVpuro-CMV-N-EGFP. The gateway cassette and 3ʹ linker region of Gateway® pcDNA™-DEST47 (ThermoFisher Scientific, 12281010) was PCR amplified using the primers 5ʹ-gateway-KpnIHpaINheI (5ʹ-GACGGTACCGGCGTTAACGACGCTAGCGGagttaagcttgatcaaacaagtttg-3ʹ) and 3ʹ-linker-XhoIPmeIPacI (5ʹ-GACTTAATTAAGCCGGTTTAAACACCGCTCGAGtctagatcgaaccactttgtaca-3ʹ) and ligated into pDEST-CMV-N-EGFP following digestion with KpnI and PacI to generate pDEST-CMV-polysite. EGFP was PCR amplified from pEAK13-EGFP using the primers 5ʹ-EGFP-XhoI (5ʹ-GACCTCGAGATGGTGAGCAAGGGCGAG-3ʹ) and 3ʹ-EGFP-stop-PacI (5ʹ-GACTTAATTAATTACTTGTACAGCTCGTCCATG-3ʹ) and ligated into pDEST-CMV-polysite after digestion with XhoI and PacI to generate pDEST-CMV-C-EGFP. 3xFLAG-attR1 fragment was digested from pCMV-tripleFLAG-Gateway and subcloned into pDEST-CMV-C-EGFP following digestion with KpnI and NotI to generate pDEST-CMV-3xFLAG-gateway-EGFP.

To generate lentiviral transfer plasmids for the production of stable cell lines expressing PGK promoter-driven LC3B or GABARAPL1 mutants, the GFP-LC3/GABARAP coding sequence was PCR amplified from the respective variant of pDEST-CMV-N-EGFP and subcloned by restriction digests between the BamHI and SalI sites of pLenti PGK GFP Puro (w509-5) (Addgene, 19070; Campeau and Kaufman labs) [41]. These PCRs were performed using the common forward primer 5ʹ-EGFP-BamHI (5ʹ- GACGGATCCGCCACCATGGTGAGCAAG-3ʹ) and one of the following reverse primers: 3ʹ-LC3B-G120-SalI (5ʹ- GACgtcgacTTACCCGAACGTCTCCTGG-3ʹ), 3ʹ-LC3B-GA-SalI (5ʹ-GACgtcgacTTACACTGACAATTTCATCGCG-3ʹ), 3ʹ-GABARAPL1-G116-SalI (5ʹ-GACgtcgacTTACCCATAGACACTCTCATCA-3ʹ) or 3ʹ-GABARAPL1-GA-SalI (5ʹ-GACgtcgacTTATTTCGCATAGACACTCTCATC-3ʹ). All cloning was confirmed by diagnostic restriction enzyme digests and Sanger sequencing of coding sequences.

Genomic DNA was isolated using NucleoSpin Tissue genomic DNA purification kit (Macherey-Nagel, 740952) according to the manufacturer’s protocol for cultured cells. The genomic locus of interest was PCR-amplified using specific primers (listed in Table S1) designed using NCBI Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast↗) [42] against relevant regions of reference genome sequence retrieved using the ‘Table Browser’ function of UCSC Genome Browser (https://genome.ucsc.edu/cgi-bin/hgTables↗) [43]. Genomic primers incorporated attB sequence overhangs to allow subsequent gateway cloning of gel-purified PCR product into pDONR223 using BP Clonase II enzyme mix. The reaction mix was transformed into bacteria and plasmid DNA was extracted from overnight cultures of multiple individual clones and sequenced by Sanger sequencing using M13 F and R primers. Genotyping results of main CRISPR-modified clones used in this study are provided in Table S3.

Cells were washed in ice-cold PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, pH 7.4) prior to lysis in NP-40 lysis buffer (150 mM NaCl, 1% IGEPAL® CA-630/NP-40 substitute [Sigma, I8896], 50 mM Tris-HCl, pH 8.0) containing protease inhibitors (c0mplete, EDTA-free, Roche, 11873580001) and 20 mM N-ethylmaleimide (unless otherwise stated) on ice. Lysates were cleared by centrifugation at 15,200 x g at 4°C and the resulting pellet was discarded. Protein levels were quantified using Pierce BCA Protein Assay Kit (ThermoFisher Scientific, 23,225) and lysates were diluted to approximately equal concentrations before heating in SDS sample buffer (with final concentration in the sample of 50 mM Tris-Cl, pH 6.8, 2% SDS, 10% glycerol, 5% beta-mercaptoethanol, 0.01% bromophenol blue) at 95°C for 5 min. Equal amounts of sample (typically 25 µg total protein per lane) were loaded into lanes of a 15-well 4–20% Mini-PROTEAN TGX Precast Gel or 26-well 4–20% Criterion TGX gel (Bio-Rad) and run at 115 V constant for 75 min or 200 V for 35 min. Protein was transferred to Immobilon-FL PVDF membrane (Merck-Millipore, IPFL00010) before blocking with 5% skimmed milk in PBS-T (PBS containing 0.05% Tween® 20 [Sigma, P1379]). The membrane was probed with primary antibody in blocking buffer at room temperature for 1 h or overnight at 4°C. After washing with PBS-T, the membrane was incubated in secondary antibody in blocking buffer for 1 h at room temperature. Further washes were carried out before image acquisition on an Odyssey infrared imaging system (LI-COR Biosciences) or development using EZ-ECL chemiluminescence detection kit for HRP (Geneflow Ltd, K1-0172) followed by exposure on an ImageQuant chemilumenscent imaging system (GE Healthcare).

After treatment with compounds, cells were lysed on ice in PLD assay buffer (150 mM NaCl, 50 mM Tris-HCl, pH 8.0, 5 mM CaCl₂, 1% Triton X-100 [Sigma, X100]) without EDTA or protease inhibitors, because PLD requires Ca²⁺ for its activity [44]. Lysates were cleared by centrifugation and then divided into separate tubes of equal volumes. On ice, PLD was added to a final concentration of 2.5 units/µl, or GST-ATG4B was added to a final concentration of 0.1 µg/µl. For negative control samples, an equivalent volume of PLD assay buffer was added. Samples were then heated to 37°C for 1 h while an unheated sample was kept on ice for this duration as a control. The reaction was stopped by the addition of 5X SDS sample buffer (250 mM Tris-Cl, pH 6.8, 10% SDS, 50% glycerol, 25% beta-mercaptoethanol, 0.05% bromophenol blue) to the tube with immediate boiling at 95°C for 5 min. Equal volumes of sample were then loaded on a 4–20% gel and assessed by western blot.

Our protease-protection assay was adapted and optimized from a combination of 2 published protocols [,]. Cells were seeded 24 h prior to transfection in one 10-cm dish per sample. The day after transfection, cells were treated with Torin1 and baf A1 for 3 h before harvesting by trypsinization (ThermoFisher Scientific, 25200056) at 37°C. Trypsin for harvesting was inactivated by resuspending intact cells in pre-warmed complete growth medium. Cells were then gently centrifuged and washed once in ice-cold PBS to remove traces of trypsin. Cells were again centrifuged and resuspended in 0.7 ml of ice-cold homogenization buffer (HB: 10 mM HEPES-KOH, pH 7.5, 0.22 M D-mannitol, 0.07 M sucrose [Sigma, S0389], 1 mM EDTA, 1 mM DTT). All subsequent steps were performed on ice or at 4°C. Cells were ruptured by passing the suspension 10 times through a 27G needle attached to a 1 ml syringe. Following this, cells were centrifuged twice at 300 x g and the nuclear pellet was discarded. Each sample was then divided into 2 tubes and permeabilized or not permeabilized by addition of 5% Triton X-100 in HB to a final concentration of 0.5%, or an equivalent volume of HB. These mixtures were then further subdivided into 3 tubes, one of which had 1 mg/ml trypsin (dissolved fresh in HB) added to a final concentration of 100 µg/ml, while the other 2 untreated tubes had an equivalent volume of HB added. Two were then incubated at 37°C for 30 min, while one untreated tube was kept on ice as a control. To stop the reactions, all tubes were placed on ice and 5X SDS sample buffer was added, before immediate boiling at 95°C for 10 min. Equal volumes of sample were then immediately loaded on 4–20% polyacrylamide gels and analyzed by western blotting. 45 46

Cells seeded on borosilicate glass coverslips (13-mm diameter, thickness no. 1.5; VWR, 631–0150) were washed in PBS and fixed for 12 min at room temperature in 4% paraformaldehyde in PBS. For imaging fluorescent proteins alone, the coverslips were then washed, stained with 1 µg/ml Hoechst 33342 (ThermoFisher Scientific, H3570) for 10 min, washed and mounted onto glass slides using ProLong Diamond mounting medium (ThermoFisher Scientific, P36970↗) and cured overnight at room temperature in the dark. For immunocytochemistry, cells were permeabilized and fixed a second time in ice cold methanol at −20°C for 12 min, washed in PBS, quenched in 50 mM NH₄Cl in PBS for 20 min, washed, and incubated in blocking buffer (3% goat serum [ThermoFisher Scientific, 16210064] in PBS) for 1 h. Coverslips were transferred to humidified staining boxes and in incubated overnight at 4°C with 45 µl primary antibody diluted in blocking buffer. Coverslips were then washed 3 times with PBS for a total of 10 min, and then incubated with 45 µl secondary antibody in blocking buffer containing 1 µg/ml Hoechst 33342 for 1 h at room temperature, prior to washing and mounting onto slides. For co-staining experiments, antibodies raised in different species against different target proteins were incubated simultaneously. Fixed cells were imaged on a Leica SPE, inverted SPE, SP5 or SP8 confocal laser-scanning microscope using a 63x oil immersion objective with a numerical aperture (N.A.) of 1.3 (SPE, inverted SPE) or 1.4 (SP5, SP8) at 400 Hz with sequential channel acquisition. Laser and gain settings were optimized for the brightest sample to avoid signal saturation, and the same settings were used to image all samples within a single experiment. All images shown are of a single confocal Z-slice. For some confocal microscopy figures, the contrast level was increased in a linear manner using the ‘Levels’ tool in Adobe Photoshop to improve clarity. When performed, this adjustment was applied equally and uniformly to all experimental conditions and controls.

Cells seeded on glass coverslips were first fixed by addition of an equal volume of double-concentrate fixative solution (3% formaldehyde, 0.2% glutaraldehyde in PBS) for 15 min at room temperature. Cells were then fixed a second time in 2% glutaraldehyde in 0.1 M sodium cacodylate for 15 min. Coverslips were then rinsed twice in 0.1 M sodium cacodylate and fixed in 1% osmium tetroxide 1.5% potassium ferrocyanide for 1 h at 4°C. After rinsing 3 times in 0.1 M sodium cacodylate, samples were stained with 1% (w:v) tannic acid in 0.05 M sodium cacodylate for 45 min at room temperature in the dark, before incubation with 1% (w:v) sodium sulfate in 0.05 M sodium cacodylate. Samples were then dehydrated by sequential 5-min incubations in distilled water, followed by 70% ethanol (x2), 90% ethanol (x2) and 100% dry ethanol (2x). Coverslips were then transferred to a 1:1 (v:v) mix of propylene oxide:TAAB 812 resin (TAAB Laboratories Equipment Ltd, T024) for 1 h. Samples were transferred again twice to fresh resin for 1.5 h each time. Coverslips were then mounted onto pre-polymerized blocks with a drop of fresh resin and baked overnight at 60°C. Glass was removed by cooling in liquid nitrogen and 70-nm ultrathin sections were cut parallel to the growth surface using a diamond knife and collected on formvar-coated copper slot grids (Agar Scientific, AGG2525C). Sections were stained with lead citrate, washed 6 times with distilled water, and air dried before imaging on a Tecnai Spirit (FEI) with a Morada digital camera (EMSIS) at 120 kV.

Stably transduced cells were seeded on a gridded glass coverslip in a 35-mm cell culture dish (MatTek, P35G-2–14-C-GRID). The next day, cells were treated with compounds, and MitoTracker Red cmxROS (Lonza, PA-3017) was added to the growth medium at a concentration of 100 nM for the final 45 min of the experiment. Cells were fixed by addition of an equal volume of double-concentrate fixative solution (3% formaldehyde, 0.2% glutaraldehyde, 2 µg/ml Hoechst 33342 in PBS) for 15 min at room temperature. Cells were then washed twice in PBS and left in PBS for imaging. A cell near the center of the coverslip was imaged on a Leica inverted SP5 confocal laser scanning microscope with DIC using a 63x numerical aperture 1.4 oil immersion objective. Z-stacks were acquired at a slice separation of 300 nm and channels were acquired sequentially for each slice in turn. A DIC image showing the morphology of the cell and its grid position was acquired using a 40x objective. After imaging, samples were fixed a second time with 2% glutaraldehyde 0.1 M sodium cacodylate for 15 min at room temperature. After washing, samples were osmium fixed, stained, dehydrated and embedded as described above for conventional TEM. The same cell imaged by confocal microscopy was then re-located on the block face by referring to its grid position and morphology from the DIC image. The block was trimmed to this region and serial sections were collected, stained, and imaged by TEM as above. Light microscopy data were adjusted for clarity in a linear manner using brightness and contrast tools, prior to being manually aligned with electron microscopy data using the transform and layers tools in Adobe Photoshop. DIC images and TEM low power overview, together with MitoTracker signal and mitochondria ultrastructure, were used as unbiased fiducial markers.

Olympus iTEM software was used. For each cell, cytoplasm area was first calculated using a low magnification image by outlining the entire cell using the interpolated polygon tool, measuring total area, and subtracting the nuclear area. Autophagosomes were then identified, using higher magnification images, based on their distinctive morphology of cytoplasmic contents within membrane-limited compartments. To measure autophagosome size, the outer limiting membrane of each structure was traced using the freehand selection tool in 2D.

Protein levels were quantified using the Analyze>Gel function in FIJI/ImageJ (NIH). The raw unprocessed western blot image was opened and a box was drawn surrounding the band of interest. After background subtraction, the histogram area of the signal of each band was determined and divided by the corresponding signal of a loading control band (ACTIN or VCL) from the same gel lane.

Puncta of endogenous LC3B and GABARAPL1 were counted in an unbiased semi-automated manner using the ‘Find Maxima’ tool in FIJI/ImageJ (NIH), with a noise tolerance value of 150 or 140, respectively. The nuclear region was first excluded from analysis by inverting the selection of a binary image generated using the Hoechst channel. Automatically identified puncta were manually counted and assigned to a parent cell to generate each data point. Colocalization between SQSTM1 and LAMP1 was assessed using the Coloc 2 plugin in FIJI. An individual cell was manually selected as an ROI using the polygon selection tool to generate each data point.

All statistical tests were done using GraphPad Prism. Unpaired two-tailed t-tests were performed for experiments comparing two sets of data. For experiments involving the comparison of multiple sets of data, an ordinary one-way ANOVA was performed using Sidak’s multiple comparison post-test to compare relevant pre-defined dataset pairs.