AKT-mediated phosphorylation of ATG4B impairs mitochondrial activity and enhances the Warburg effect in hepatocellular carcinoma cells

ABSTRACT

Phosphorylation is a major type of post-translational modification, which can influence the cellular physiological function. ATG4B, a key macroautophagy/autophagy-related protein, has a potential effect on the survival of tumor cells. However, the role of ATG4B phosphorylation in cancers is still unknown. In this study, we identified a novel phosphorylation site at Ser34 of ATG4B induced by AKT in HCC cells. The phosphorylation of ATG4B at Ser34 had little effect on autophagic flux, but promoted the Warburg effect including the increase of L-lactate production and glucose consumption, and the decrease of oxygen consumption in HCC cells. The Ser34 phosphorylation of ATG4B also contributed to the impairment of mitochondrial activity including the inhibition of F₁Fo-ATP synthase activity and the elevation of mitochondrial ROS in HCC cells. Moreover, the phosphorylation of ATG4B at Ser34 enhanced its mitochondrial location and the subsequent colocalization with F₁Fo-ATP synthase in HCC cells. Furthermore, recombinant human ATG4B protein suppressed the activity of F₁Fo-ATP synthase in MgATP submitochondrial particles from patient-derived HCC tissues in vitro. In brief, our results demonstrate for the first time that the phosphorylation of ATG4B at Ser34 participates in the metabolic reprogramming of HCC cells via repressing mitochondrial function, which possibly results from the Ser34 phosphorylation-induced mitochondrial enrichment of ATG4B and the subsequent inhibition of F₁Fo-ATP synthase activity. Our findings reveal a noncanonical working pattern of ATG4B under pathological conditions, which may provide a scientific basis for developing novel strategies for HCC treatment by targeting ATG4B and its Ser34 phosphorylation.

KEYWORDS:

• AKT
• ATG4B
• F₁Fo-ATP synthase
• HCC
• phosphorylation
• Warburg effect

Abbreviations

AKT1/PKB	=	AKT serine/threonine kinase 1
ATG4A	=	autophagy-related 4A cysteine peptidase
ATG4B	=	autophagy-related 4B cysteine peptidase
ATG4D	=	autophagy-related 4D cysteine peptidase
ATG5	=	autophagy-related 5
ATG7	=	autophagy-related 7
ATP5A1	=	ATP synthase, H+ transporting, mitochondrial F1 complex, alpha subunit 1, cardiac muscle
ATP5C1	=	ATP synthase, H+ transporting, mitochondrial F1 complex, gamma polypeptide 1
Baf A1	=	bafilomycin A1
CRISPR	=	clustered regularly interspaced short palindromic repeats
DGc	=	differential gradient centrifugation
GABARAPL1	=	GABA type A receptor associated protein like 1
GFP	=	green fluorescent protein
HCC	=	hepatocellular carcinoma
IP	=	immunoprecipitation
MAP1LC3/LC3	=	microtubule associated protein 1 light chain 3
MEF	=	mouse embryonic fibroblast
mitoROS	=	mitochondrial reactive oxygen species
MTP	=	mitochondrial targeting peptide
OMM	=	outer mitochondrial membrane
OXPHOS	=	oxidative phosphorylation
PK	=	proteinase K
PTEN	=	phosphatase and tensin homolog
siRNA	=	small interfering RNA
SMPs	=	submitochondrial particles
WCL	=	whole cell lysate

Introduction

Hepatocellular carcinoma (HCC), a primary malignancy of liver, has become a main cause of cancer-related death worldwide with an incidence of approximately 850,000 new cases per year [1]. HCC is a highly aggressive neoplasm characterized by frequent metastasis and therapeutic resistance. Although the treatments for HCC patients have improved in the past decades, there are still major needs for effective prevention at an early stage and new therapies at the intermediate and advanced stage. Therefore, it is necessary and urgent to understand the more detailed mechanisms about the occurrence and progression of HCC so as to discover novel targets for its management.

Mitochondria, as the powerhouses of the cells, produce the most energy-rich molecules, adenosine triphosphates (ATPs), via oxidative phosphorylation (OXPHOS) under normal physiological conditions. Previous reports have shown that the dysfunction of mitochondria often results in many disorders [2].^– [6] In 1926, Otto Warburg reported that tumor cells could produce ATPs mainly via glycolysis (i.e., not by OXPHOS) even under aerobic conditions, which may be due to the impairment of mitochondrial respiration. Lately, more and more studies have indicated that the impairment of mitochondrial metabolism can enhance glycolysis and play a key role in metabolic remodeling in many kinds of tumors [7–9]. This metabolic transformation of tumors from mitochondrial respiration to aerobic glycolysis is also named the Warburg effect, which can obviously promote occurrence, progression and chemoresistance of malignancies via different mechanisms [10]. It has been reported that the Warburg effect is also a typical characteristic in most HCC cases. PKM (pyruvate kinase M1/2), a key enzyme in regulating the Warburg effect, is upregulated in HCC cells and is associated with tumorigenesis, aggressive progression, and poor prognosis [11–13]. SLC2A1/glucose transporter 1, a rate-limiting transporter for glucose uptake, is also elevated in some HCC patients and affects tumorigenicity of HCC [14]. In addition, the microRNA-mediated inhibition of the Warburg effect can also influence HCC progression [15].^, [16] Moreover, Cuezva et al. have demonstrated that carcinogens in liver are related to a decrease of mitochondrial content, a selective downregulation of ATP5B/beta-F₁-ATP synthase level and an increase of the glycolytic GAPDH (glyceraldehyde-3-phosphate dehydrogenase), which supports the hypothesis of the Warburg effect [17]. Recently, targeting the Warburg effect has emerged as a promising option for HCC treatment [18]. Therefore, investigating the detailed mechanisms of the Warburg effect in HCC will be beneficial to develop novel related therapeutic strategies.

ATG4B (autophagy-related 4B cysteine peptidase), a member of C54 peptidases, plays a key role in the regulation of autophagy in mammalian cells. Previous studies have demonstrated that ATG4B regulates autophagy through either cleaving pro-MAP1LC3/LC3 for its further lipidation or delipidating LC3-II from the autophagosome, leading to a change of the autophagy level [19–28]. Growing evidence has shown that ATG4B participates in certain physiological and pathological processes including tumors through affecting the autophagic flux [29–39]. For example, ATG4B can promote tumor progression or chemoresistance via its canonical role of autophagy regulation. Meanwhile, a recent study in colon cancer cells has revealed that ATG4B can enhance cell proliferation independent of its role in autophagy [40], which opens new insight for the noncanonical roles of ATG4B. Besides the regulation of protein level, protein modifications such as phosphorylation, oxidative modification and O-GlcNAcylation are also important mechanisms to affect the function of ATG4B [41–44]. Previous research using high-resolution mass spectrometry-based proteomics has indicated that ATG4B could be phosphorylated at Thr7, Tyr8, Ser34, Ser44, Tyr122, Tyr123, Ser316, Ser383 and Ser392 (http://www.phosphosite.org). Moreover, recent studies have shown that the phosphorylation of ATG4B at Ser383 and Ser392 or Ser316 can influence ATG4B protease activity [42,45] suggesting that the phosphorylation of ATG4B would be an important aspect to regulate its function. However, it is still unclear as to the role of ATG4B phosphorylation in HCC cells.

In the present study, we identified a novel phosphorylation site (Ser34) on ATG4B mediated by AKT in HCC cells. The phosphorylation of ATG4B at Ser34 presented little effect on basal autophagic flux of HCC cells, but dramatically contributed to the inhibition of F₁Fo-ATP synthase activity, the increase of mitochondrial reactive oxygen species (mitoROS) and the Warburg effect. Moreover, the phosphorylation of ATG4B at Ser34 promoted its mitochondrial enrichment and the subsequent colocalization with the F₁Fo-ATP synthase in HCC cells. Furthermore, recombinant human ATG4B protein suppressed the activity of the F₁Fo-ATP synthase in MgATP submitochondrial particles (SMPs) from patient-derived HCC tissues in vitro. In summary, our results demonstrate for the first time that the phosphorylation of ATG4B at Ser34 participates in the metabolic reprogramming of HCC cells via inhibiting mitochondrial function, which possibly results from the Ser34 phosphorylation-induced enrichment of mitochondrial ATG4B and the subsequent repression of F₁Fo-ATP synthase activity. Our findings reveals a noncanonical working model of ATG4B under pathological conditions and the role of ATG4B in HCC malignance, which may provide a scientific basis for developing novel strategies for HCC treatment by targeting ATG4B and its Ser34 phosphorylation.

Results

Activation of AKT induces the phosphorylation of ATG4B at Ser34 in HCC cells

It has been reported that overactivation of protein kinases including AKT is a typical phenotype of HCC, so we explored whether the activated AKT could influence the phosphorylation of ATG4B in HCC cells. As shown in Figure 1A, co-immunoprecipitation (Co-IP) assays using the ATG4B^CRISPR HepG2 cells revealed that the ectopically expressed AKT1/PKB and ATG4B appeared in 1 complex, suggesting the probability of interaction between the 2 proteins. Here, the ATG4B^CRISPR HepG2 cells were ATG4B hemizygous knockout cells generated by a CRISPR/Cas9-mediated genome editing system (Figure S1A and S1B). Then the effect of AKT1 on the phosphorylation of endogenous ATG4B was detected with Phos-tag technology. As shown in Figure 1B, overexpression of AKT1 in HepG2 cells significantly increased the phosphorylated ATG4B (p-ATG4B), which was reversed by phosphatase, suggesting that AKT1 may induce the phosphorylation of endogenous ATG4B in HCC cells. Meanwhile, we noticed that there were different positions of bands corresponding to ATG4B (i.e., phosphorylated bands of ATG4B) in the gel. As the migration rate of a protein in a Phos-tag gel can be affected by the number of phosphorylated sites, the different sites of gel shift may arise from different phosphorylation types of ATG4B in this condition.

Figure 1. Activation of AKT induces the phosphorylation of ATG4B at Ser34 in HCC cells. (A) ATG4B^CRISPR HepG2 cells (ATG4B hemizygous knockout cells) were transfected with the indicated expression plasmids. Then the whole cell lysates (WCL) were separately used for immunoblotting and immunoprecipitation assays with the corresponding antibodies. (B) HepG2 cells were transfected with MYC-AKT1 expression plasmid or control empty vector (EV). Then the cell lysates were obtained and treated with or without lambda phosphatase. Subsequently, the cell lysates were loaded onto SDS-PAGE gels with or without Phos-tag acrylamide and MnCl₂ for immunoblotting assays. (C) The potential phosphorylation sites in ATG4B were predicted with motifscan (http://scansite.mit.edu/motifscan_seq.phtml), and the potential AKT1 phosphorylation motif ₃₁RKYS₃₄ in human ATG4B is shown. The red label represents the location of Ser34 in the 3D structure of the ATG4B protein. (D) ATG4B^CRISPR HepG2 cells were transfected with Flag-ATG4B^WT or Flag-ATG4B^S34A expression plasmid in the presence or absence of MYC-AKT1^WT expression vector. Then the cell lysates were prepared and loaded onto SDS-PAGE gels with or without Phos-tag acrylamide and MnCl₂ for immunoblotting assays. The average ratio of F-p-ATG4B to t-ATG4B from 3 independent experiments is shown on the right. (E) ATG4B^CRISPR HepG2 cells were transfected with the Flag-ATG4B^WT expression plasmid in the presence or absence of MYC-AKT1^WT expression vector. Then the cells were treated with 3 μM MK2206 or vehicle control (DMSO). Subsequently, the cell lysates were obtained for immunoblotting assays. (F) HepG2 cells were transfected with control siRNA or PTEN siRNA, and then the cell lysates were prepared and loaded onto SDS-PAGE gels with or without Phos-tag acrylamide and MnCl₂ for immunoblotting assays. Data are mean ± SD from 3 independent experiments. *, P < 0.05; ns, no significance. MYC-AKT1^WT, 1 × MYC-tagged wild-type AKT1 expression plasmid; 3 × Flag-ATG4B^WT, 3 × Flag-tagged wild-type ATG4B expression plasmid; p-ATG4B, phosphorylated ATG4B; non-p-ATG4B, non phosphorylated ATG4B; t-ATG4B, total ATG4B; phos, SDS-PAGE gel containing Phos-tag acrylamide and MnCl₂; Flag-ATG4B^WT, 1 × Flag-tagged wild-type ATG4B expression plasmid; Flag-ATG4B^S34A, 1 × Flag-tagged mutant ATG4B expression plasmid (in which Ser34 of ATG4B was mutated to Ala); F-p-ATG4B, the first band of phosphorylated ATG4B; S-p-ATG4B, the second band of phosphorylated ATG4B; LE, long exposure; SE, short exposure; p-ATG4B (S34), Ser34-phosphorylated ATG4B; p-AKT (S473), Ser473-phosphorylated AKT.

Next, the potential phosphorylation sites in ATG4B were predicted with motifscan (http://scansite.mit.edu/motifscan_seq.phtml). As shown in Figure 1C, ATG4B (Homo sapiens) contains a potential AKT1 phosphorylation motif ₃₁RKYS₃₄ in its N terminus. Sequence homology alignment showed that several other species including S. cerevisiae, C. elegans, Gallus gallus, Mus musculus, Bos taurus, C. l. familiaris and O. r. rosmarus also contain this “R × × S” motif (Figure S1C). Moreover, the results from NCBI protein blast suggested that the Ser34 site in human ATG4B may be relatively specific compared to human ATG4A, ATG4C and ATG4D (Figure S1D). In addition, previous studies have demonstrated that the Ser34 of endogenous ATG4B can be phosphorylated in mouse liver [46], 3T3-L1 fibroblasts [47] and HEK293T cells [42]. Therefore, we further investigated whether the Ser34 of ATG4B was the phosphorylation target of AKT1 in HCC cells. As shown in Figure 1D, overexpresion of AKT1 dramatically elevated the level of F-p-ATG4B (the first presumably phosphorylated band of ATG4B from the bottom to the top of the gel) in the HCC cells transfected with a wild-type ATG4B expression plasmid (Flag-ATG4B^WT), but not in the cells transfected with the mutant ATG4B expression plasmid (Flag-ATG4B^S34A, in which the Ser34 of ATG4B was mutated to Ala). Meanwhile, the ratio of F-p-ATG4B to t-ATG4B (total ATG4B) was markedly increased in the cells co-overexpressing ATG4B^WT and AKT1, indicating that the phosphorylation of Ser34 was an important type of ATG4B phosphorylation in AKT1-expressing HCC cells. Co-IP assays also showed that overexpression of AKT1 significantly enhanced the Ser/Thr phosphorylation of ATG4B^WT (but not ATG4B^S34A)-containing immunoprecipitates using immunoblotting with the antibody against the AKT target motif (“RxRxxS/T” or “RxxS/T”), further suggesting that the Ser34 in the motif ₃₁RKYS₃₄ of ATG4B may be phosphorylated by AKT1 (Figure S2A).

Subsequently, we prepared an antibody against the phosphorylated Ser34 of ATG4B with a titer more than 1:32000 using KLH-C-LGRKY(p-S)IF-NH₂ as an immunogen (Figure S2B), and verified the specificity of this antibody with immunoblotting assays (Figure S2C and S2D). Using this antibody, we found that ectopic expression of AKT1 remarkably increased the Ser34 phosphorylation of ATG4B, which was dramatically attenuated by the AKT inhibitor MK2206 (Figure 1E). Furthermore, we evaluated the phosphorylation state of ATG4B by activation of endogenous AKT in the HCC cells with PTEN knockdown. As shown in Figure S2E, an optimal PTEN siRNA (PTEN siRNA-1) was identified for the subsequent experiments, which resulted in an efficient interference of PTEN and activation of AKT (S473). Figure 1F revealed that knockdown of PTEN obviously promoted the phosphorylation of ATG4B at Ser34.

Additionally, we also verified the specificity of ATG4B antibody so as to rule out the possibility of a wrong recognition against its homolog ATG4A (Figure S2F), as ATG4A and ATG4B showed an identity of 55% (NP_443168.2 vs NP_037457.3) in NCBI protein blast analysis. Moreover, motifscan showed that human ATG4A also contains a potential AKT1 phosphorylation motif “RKFS” (Figure S3A), so we examined the effect of AKT-mediated phosphorylation on ATG4A. As shown in Figure S3B, activation of AKT had little effect on ATG4A phosphorylation.

Taken together, the above results demonstrate that activation of AKT induces the phosphorylation of ATG4B at Ser34 in HCC cells.

The Ser34 phosphorylation of ATG4B has little effect on autophagic flux in ATG4B-overexpressing HCC cells

To investigate whether the phosphorylation of ATG4B at Ser34 can influence autophagic flux, autophagy-related markers were detected in the HCC cells overexpressing ATG4B^WT or ATG4B^S34A. As shown in Figure S4A, overexpression of ATG4B^WT or ATG4B^S34A dramatically decreased LC3-II and increased SQSTM1/p62. Moreover, overexpression of ATG4B^WT or ATG4B^S34A in GFP-LC3 stably expressing HepG2 cells markedly decreased the amount of GFP-LC3 puncta (Figure S4B). These results suggested that excess exogenous ATG4B may impair the lipidation of LC3 and negatively regulate autophagic flux in HCC cells, which was consistent with the previous studies in HeLa and MCF-7 cells [23,26]. Meanwhile, this effect was not influenced by the mutation of Ser34 under basal conditions (without extra stress, such as AKT overactivation).

As the phosphorylated degree of ATG4B under the basal condition would be limited, we further examined the effect of ATG4B Ser34 phosphorylation on autophagy in AKT-overactivated HCC cells. As shown in Figure 2A, co-overexpression of ATG4B^WT or ATG4B^S34A with AKT1^WT or AKT1 kinase-dead mutant (AKT1^AA, in which both Thr308 and Ser473 of AKT1 were mutated to Ala) significantly reduced LC3-II and elevated SQSTM1. Similarly, ectopic expression of ATG4B^WT or ATG4B^S34A combined with PTEN knockdown also dramatically decreased LC3-II and increased SQSTM1 (Figure 2B). As the decrease of LC3-II level may arise from the impairment of autophagy flux or the enhancement of autophagosome turnover [48], we further used bafilomycin A₁ (Baf A1, a compound that blocks the fusion of autophagosomes with lysosomes) in the subsequent experiments to detect the ATG4B-mediated effect on autophagy. As shown in Figure 2C, Baf A1 treatment significantly upregulated the level of LC3-II, which was largely reversed by overexpression of ATG4B^WT or ATG4B^S34A, whether or not there was AKT overactivation. In addition, Baf A1 treatment obviously increased the amount of GFP-LC3 puncta in GFP-LC3 stably expressing HepG2 cells, which was reversed by transfection with a plasmid expressing exogenous ATG4B^WT or ATG4B^S34A, whether or not there was AKT overexpression (Figure 2D). Similar results were also obtained in the cells with PTEN knockdown (Figure 2E). The assays with Baf A1 treatment suggested that the decrease of LC3-II level and GFP-LC3 puncta by excess exogenous ATG4B^WT or ATG4B^S34A may not be attributable to the enhancement of autophagosome turnover, whether or not there was AKT overactivation. In brief, the above data indicate that overexpression of ATG4B impairs autophagic flux in HCC cells, which is independent of ATG4B Ser34 phosphorylation.

Figure 2. The Ser34 phosphorylation of ATG4B has little effect on autophagic flux in ATG4B-overexpressing HCC cells. (A and B) ATG4B^CRISPR HepG2 cells were transfected with the indicated expression plasmids in the presence or absence of PTEN siRNA. Then the cell lysates were obtained for immunoblotting assays. The corresponding ratios of LC3-II to ACTB are listed below. (C) ATG4B^CRISPR HepG2 cells were transfected with the indicated expression plasmids. Then the cells were treated with 10 nM bafilomycin A₁ (Baf A1) or vehicle control for 3 h. Subsequently, the cell lysates were obtained for immunoblotting assays. The corresponding ratios of LC3-II (LE) to ACTB are listed below. (D and E) HepG2 cells stably expressing GFP-LC3 were transfected with the indicated expression plasmids or siRNAs, followed by treatment with 10 nM Baf A1 or vehicle control. Then the GFP-LC3 puncta in the cells were counted under a fluorescence microscope. The statistical diagrams represented the average numbers of GFP-LC3 puncta per cell (50 cells per sample were observed). Data are mean±SD from 3 independent experiments. *, P < 0.05; **, P < 0.01; ns, no significance. AKT1^AA, 1 × MYC-tagged mutant AKT1 expression plasmid (in which both Thr308 and Ser473 were mutated to Ala); AKT1^WT, 1 × MYC-tagged wild-type AKT1 expression plasmid; ATG4B^WT, 1 × Flag-tagged wild-type ATG4B expression plasmid; ATG4B^S34A, 1 × Flag-tagged mutant ATG4B expression plasmid (in which Ser34 of ATG4B was mutated to Ala); EV, empty vector; p-ATG4B (S34), Ser34-phosphorylated ATG4B; p-AKT (S473), Ser473-phosphorylated AKT; LC3-II (SE), LC3-II band with short exposure; LC3-II (LE), LC3-II band with long exposure.

The phosphorylation of ATG4B at Ser34 promotes the Warburg effect in HCC cells

Although the phosphorylation of ATG4B at Ser34 did not influence its role in the regulation of autophagic flux in HCC cells, we interestingly found that the medium color was different between ATG4B^WT- and ATG4B^S34A-expressing cells under the condition of AKT1 overexpression (Figure S5A). As the color of the medium could reflect the pH, the results indicated that the cells with ATG4B^WT and AKT1^WT expressing plasmid cotransfection produced more acidic material than that with ATG4B^S34A and AKT1^WT co-expression. It has been demonstrated that L-lactate is a main metabolic product of HCC cells, especially in the cells with a high Warburg effect [49], so we detected the level of L-lactate in the culture supernatant. As shown in Figure 3A, the ATG4B^WT- and AKT1^WT-co-expressing cells produced more L-lactate compared to the ATG4B^S34A- and AKT1^WT-co-expressing cells. This phenomenon was also found in PTEN-knockdown cells (Figure 3B). However, the difference of L-lactate levels between ATG4B^WT-expressing and ATG4B^S34A-expressing cells disappeared when AKT1^AA was co-expressed (Figure S5B). These results suggested that the phosphorylation of ATG4B at Ser34 may play an important role in AKT1-induced L-lactate production and glycolytic potential. As the increase of glycolysis will consume more glucose and less oxygen, we further examined the changes of glucose and oxygen consumption. As shown in Figures 3C and 3D, under the condition of AKT1^WT overexpression, the glucose consumption rate (Figure 3C) was higher but oxygen consumption rate (Figure 3D) was lower in ATG4B^WT-expressing cells compared to ATG4B^S34A-expressing cells, while these differences were abolished under the condition of AKT1^AA overexpression (Figure S5C and S5D). These data indicate that the phosphorylation of ATG4B at Ser34 by AKT promotes the Warburg effect in HCC cells.

Figure 3. The phosphorylation of ATG4B at Ser34 promotes the Warburg effect in HCC cells. (A) ATG4B^CRISPR HepG2 cells were transfected with the indicated expression plasmids. Then the concentrations of L-lactate in the culture media were measured. (B) ATG4B^CRISPR HepG2 cells were transfected with the indicated expression plasmids and siRNAs, and the concentrations of L-lactate in the culture media were determined. (C) ATG4B^CRISPR HepG2 cells were transfected as in (A). Next, the concentrations of glucose in the culture media at different time points were detected separately. The time after 6 h of plasmids transfection was considered as the time “0 h” in the diagram. (D) ATG4B^CRISPR HepG2 cells were seeded into 24-well microplates and treated as in (A). Then the oligomycin-sensitive respiration (OSR) was assayed. Data are mean ± SD from 3 independent experiments. *, P < 0.05; ns, no significance. AKT1^WT, ATG4B^WT and ATG4B^S34A were the corresponding expression plasmids as described in Figure 2.

The Ser34 phosphorylation of ATG4B contributes to the inhibition of F₁Fo-ATP synthase activity and the elevation of mitoROS in HCC cells

As the oxygen consumption rate was decreased in ATG4B^WT- and AKT1^WT-co-expressing cells, we deduced that the mitochondrial function may be suppressed or impaired in this condition. Previous reports have shown that the F₁Fo-ATP synthase plays a key role in mitochondrial OXPHOS, and downregulating the expression of ATP synthase or inhibition of ATP synthase activity can induce dysfunction of mitochondrial respiration, which finally contributes to the Warburg effect [50,51]. Therefore, we investigated the influence of Ser34 phosphorylation of ATG4B on F₁Fo-ATP synthase activity in HCC cells. As shown in Figure 4A, the activity of the F₁Fo-ATP synthase was inhibited in ATG4B^WT- and AKT1^WT-co-expressing cells, but not in ATG4B^S34A- and AKT1^WT (or ATG4B^WT- and AKT1^AA or ATG4B^S34A- and AKT1^AA)-co-expressing cells. Similarly, the F₁Fo-ATP synthase activity was also repressed in ATG4B^WT (but not ATG4B^S34A)-expressing cells with PTEN knockdown (Figure 4B). These results indicated that the AKT-mediated phosphorylation of ATG4B at Ser34 contributed to the inhibition of F₁Fo-ATP synthase activity in ATG4B-overexpressing HCC cells.

Figure 4. The Ser34 phosphorylation of ATG4B contributes to the inhibition of F₁Fo-ATP synthase activity and the elevation of mitoROS in HCC cells. (A-C) ATG4B^CRISPR HepG2 cells were transfected with the indicated expression plasmids and siRNAs. Then the activity of the F₁Fo-ATP synthase was examined. (D-F) ATG4B^CRISPR HepG2 cells were transfected with the indicated expression plasmids and siRNAs. Then the level of mitoROS was detected with a fluorescence microplate reader. Data are mean ± SD from 3 independent experiments. *, P < 0.05; ns, no significance. a.u., arbitrary units; ATG4B^C74A, 1 × Flag-tagged mutant ATG4B expression plasmid (in which Cys74 of ATG4B was mutated to Ala); ATG4B^S34A,C74A, 1 × Flag-tagged mutant ATG4B expression plasmid (in which both Ser34 and Cys74 were mutated to Ala); AKT1^AA, AKT1^WT, ATG4B^WT and ATG4B^S34A were the corresponding expression plasmids as described in Figure 2.

To further determine whether this effect is involved in the proteolytic activity of ATG4B on LC3, we constructed ATG4B^C74A (Cys74 was mutated to Ala) and ATG4B^S34A,C74A (both Ser34 and Cys74 were mutated to Ala) expression plasmids. As shown in Figure 4C, the activity of the F₁Fo-ATP synthase was suppressed in ATG4B^C74A (but not ATG4B^S34A,C74A)-expressing HCC cells with PTEN knockdown, indicating that the inhibitory effect of ATG4B on F₁Fo-ATP synthase activity was dependent on its Ser34 phosphorylation, but independent of its proteolytic activity on LC3. Additionally, as mitochondria are the major source of cellular ROS, the dysfunction of mitochondria usually results in the increase or decrease of mitoROS. Accordingly, we detected the role of ATG4B Ser34 phosphorylation on mitoROS generation in HCC cells. As shown in Figure 4D, the level of mitoROS was dramatically upregulated in the cells cotransfected with plasmids expressing ATG4B^WT and AKT1^WT, but not in the cells co-expressing ATG4B^S34A and AKT1^WT (or either ATG4B^WT and AKT1^AA or ATG4B^S34A and AKT1^AA). Moreover, the mitoROS levels had no significant difference between ATG4B^WT- and ATG4B^S34A-expressing cells, or between ATG4B^C74A- and ATG4B^S34A,C74A-transfected cells in the presence of control siRNA (Figure 4E). However, upon PTEN knockdown, the level of mitoROS in ATG4B^WT- or ATG4B^C74A-expressing cells was higher compared to ATG4B^S34A- or ATG4B^S34A,C74A-expressing cells (Figure 4F), indicating that the Ser34 phosphorylation of ATG4B was important for ATG4B-induced mitoROS generation. Taken together, the above results indicate that the Ser34 phosphorylation of ATG4B contributes to the ATG4B-mediated inhibition of F₁Fo-ATP synthase activity and elevation of mitoROS, which may repress or impair mitochondrial function in AKT-activated HCC cells.

The Ser34 phosphorylation of ATG4B enhances its mitochondrial translocation in HCC cells

A previous report has shown that a small amount of ATG4B can be detected in mitochondrial-enriched fractions of RNF5^−/− macrophages [52]. Moreover, ATG4D, a homolog of ATG4B, can translocate into the interior of mitochondria after removing 63 amino acid residues in the N terminus [53], and influence mitochondrial structure and function. Additionally, some proteins that do not locate in mitochondria under a normal physiological state can translocate into mitochondria in pathological condition, such as FOXP3 and TERT in HCC cells [54–58]. Therefore, we explored whether the mitochondrial dysfunction mediated by ATG4B Ser34 phosphorylation was involved in the mitochondrial translocation of ATG4B in HCC cells.

First, the mitochondrial fractions from 3 HCC patient-derived tumor samples (#1 to #3, Table S1) were separately isolated using a differential gradient centrifugation (DGc) strategy (#1 and #2) and anti-TOMM22 magnetic beads technique (#3) according to the methods described previously [59,60] As shown in Figures S6A and S6B, ATG4B could be detected in the DGc-enriched and magnetic beads-sorted mitochondrial fractions from HCC samples. Furthermore, ATG4B (but not ATG4A) could be detected in the magnetic beads-sorted mitochondrial fractions from HepG2 cells (Figure 5A). Meanwhile, we noticed that LC3-II was also contained in the mitochondrial fractions in this condition, which was consistent with the previous finding in RNF5^−/− macrophages [52]. These data suggested that ATG4B in mitochondrial fractions of HCC cells may result from (1) ATG4B directly located in mitochondria, or (2) ATG4B copurified via mitochondria that contain LC3-II.

Figure 5. The phosphorylation of ATG4B at Ser34 promotes its mitochondrial translocation in HCC cells. (A) Cytosol and mitochondria fractions were separately isolated from HepG2 cells using a differential centrifugation strategy and an anti-TOMM22 magnetic beads method. Then the proteins in the cytosol and mitochondria were analyzed by immunoblotting assays, taking the total cellular proteins from the whole cell lysate (WCL) as a control. (B) HepG2 cells were transfected with the control siRNA or ATG7 siRNA. Then the WCL and mitochondrial proteins were detected by immunoblotting assays. (C) The mitochondrial fractions obtained from HepG2 cells using the differential gradient centrifugation strategy were treated with or without proteinase K (PK), and then subjected to immunoblotting assays. (D) The mitoplasts separated from HepG2 cells were treated with or without PK, and then subjected to immunoblotting assays. (E) The distributions of ATG4B in ATG4B^CRISPR HepG2 cells and wild-type HepG2 cells were examined with immunogold-labeling electron microscopy assays (bar: 200 nm), and the ATG4B-positive dots were counted. Sixty random fields were observed from 3 samples (20 fields per sample), and the statistical diagram represents the average number of ATG4B-positive dots per field. (F and G) HepG2 cells were transfected with the indicated siRNAs or expression plasmids. Then the distributions of ATG4B in the cytosol and mitochondria were determined with immunogold-labeling electron microscopy assays, and the average ATG4B-positive dots in one field were calculated and statistically analyzed. (H) ATG4B^CRISPR HepG2 cells were transfected with the indicated expression plasmids in the presence or absence of PTEN siRNA. Then the distributions of ATG4B in mitochondria were detected with immunogold-labeling electron microscopy assays, and the average ATG4B-positive dots in one field were calculated and statistically analyzed. (I) ATG4B^CRISPR HepG2 cells were transfected with the indicated expression plasmids. Then the cells were stained with MitoTracker Red (MT-Red) and observed under a fluorescence confocal microscope (bar: 5 μm). The colocalization of GFP-puncta with mitochondria was evaluated. The data in the statistical diagram were from 30 random fields of 3 samples (10 fields per sample). Data are mean±SD from 3 independent experiments. *, P < 0.05; **, P < 0.01; ns, no significance. Mito, mitochondria; OMM, outer mitochondrial membrane; IMM, inner mitochondrial membrane; GFP-ATG4B^WT, 1 × GFP-tagged wild-type ATG4B expression plasmid; GFP-ATG4B^S34A, 1 × GFP-tagged mutant ATG4B expression plasmid (in which Ser34 of ATG4B was mutated to Ala); AKT1^AA, AKT1^WT, ATG4B^WT and ATG4B^S34A were the corresponding expression plasmids as described in Figure 2.

To further investigate the source of ATG4B in the mitochondrial fraction in HCC cells, ATG7 siRNA was used to inhibit the formation of LC3-II, and the level of ATG4B in magnetic beads-sorted mitochondrial fractions was examined. As shown in Figure 5B, in the presence of ATG7 siRNA, the LC3-II level in the whole cell lysate (WCL) was significantly decreased, and little LC3-II was detected in the mitochondrial fractions, but ATG4B was still present in the mitochondrial fractions with a mild decrease compared to control siRNA treatment, which indicated that most ATG4B could be enriched in mitochondria of HCC cells in a manner that was independent of LC3-II binding.

Second, the mitochondrial fractions and mitoplast fractions (including mitoplast, inner mitochondrial membrane and mitochondrial matrix) were isolated for proteinase K (PK) protection assays. As shown in Figure 5C, treatment of the mitochondrial fractions with PK effectively degraded the outer mitochondrial membrane (OMM) protein TOMM22, but not ATG4B and matrix-resident HSPD1/HSP60, suggesting that ATG4B may locate within the mitochondria. Figure 5D showed that ATG4B could still be detected in the PK-treated mitoplast fractions, indicating that ATG4B could (at least partially) be transloacted into the mitochondrial matrix. Moreover, the subcellular localization of ATG4B was observed using immunogold-labeling electron microscopy. As shown in Figure 5E, the ATG4B-positive dots in wild-type HepG2 cells were much more evident than in ATG4B^CRISPR HepG2 cells. Most of the labeled ATG4B distributed in the cytoplasm including cytosol, mitochondria, autophagosomes and endoplasmic reticulum (Figure 5E). The majority of mitochondrial ATG4B was located inside of mitochondria, while a small part localized at the circumference (Figure 5E). These results indicated that a small portion of ATG4B could probably be imported into the mitochondrial matrix in HCC cells.

Third, the effect of ATG4B Ser34 phosphorylation on its mitochondrial import was examined. Using immunogold-labeling electron microscopy, we counted the average ATG4B-positive dots per visual field in the cytosol and mitochondria. As shown in Figures 5F and 5G, activation of AKT (overexpression of AKT1 or knockdown of PTEN) markedly promoted mitochondrial translocation of ATG4B. Meanwhile, the amount of overexpressed Flag-ATG4B^WT in mitochondria was much more than that of overexpressed Flag-ATG4B^S34A under the condition of PTEN knockdown (Figure 5H). In order to observe the cellular localization of ATG4B via fluorescent tracing, the GFP-ATG4B^WT and GFP-ATG4B^S34A expression plasmids were constructed. As shown in Figure S6C, the green fluorescence signals distributed throughout the cell with small aggregates of GFP-ATG4B (hereafter referred to as ATG4B-puncta) in the cytoplasm, while the percentage of ATG4B-puncta in mitochondria was quite low in GFP-ATG4B^WT- or GFP-ATG4B^S34A-expressing cells and presented no statistical difference between the 2 groups. However, the amount of ATG4B puncta and the percentage of ATG4B puncta in mitochondria were dramatically increased in AKT1^WT- and GFP-ATG4B^WT- co-expressing cells, but not in the cells co-expressing AKT1^WT and GFP-ATG4B^S34A or AKT1^AA and GFP-ATG4B^WT (Figure 5I). Additionally, activation of AKT significantly increased the amount of ATG4B in DGc-enriched mitochondrial fractions in ATG4B^WT (but not ATG4B^S34A)-expressing cells (Figure S6D).

In brief, the above data demonstrate that a small portion of ATG4B can be imported to mitochondria in HCC cells, whereas overactivation of AKT further enhances the enrichment of ATG4B in mitochondria via phosphorylation of Ser34.

ATG4B is present in one complex with the subunits of the mitochondrial F₁Fo-ATP synthase in HCC cells

As mitophagy is an important mechanism for maintaining mitochondrial homeostasis, and ATG4B can interact with mitophagy-related proteins such as LC3-II, we investigated whether the increased mitochondrial translocation of ATG4B in AKT-activated cells could influence mitochondrial function via mitophagy. As shown in Figures 6A and S7A, overexpression of Flag-ATG4B^WT or Flag-ATG4B^S34A significantly inhibited the formation of GFP-LC3 puncta and its colocalization with mitochondria, whether or not there was AKT activation, suggesting that the effect of ATG4B Ser34 phosphorylation on mitochondrial function may not attribute to mitophagy in the condition of ATG4B overexpression. Moreover, an immunofluorescence assay was used to examine the colocalization of ATG4B-puncta and LC3 or GABARAPL1 in ATG4B-overexpressing HCC cells. As shown in Figure S7B, knockdown of ATG7 remarkably decreased the amount of LC3 or GABARAPL1 puncta, demonstrating the utility of LC3 or GABARAPL1 antibody for immunofluorescence.

Figure 6. ATG4B is present in one complex with the subunits of the mitochondrial F₁Fo-ATP synthase in HCC cells. (A) HepG2 cells stably expressing GFP-LC3 were transfected with the indicated siRNAs and expression plasmids. Then the cells were labeled with the indicated antibodies using immunofluorescence assays and observed under a fluorescence confocal microscope. Representative images are shown (bar: 10 μm). (B) ATG4B^CRISPR HepG2 cells were transfected with the 3 × Flag-ATG4B^WT expression plasmid. Then the whole cell lysate (WCL) was obtained for immunoprecipitation assays. Subsequently, the ATG4B antibody-precipitated complex was further subjected to mass spectrometry, taking the IgG-precipitated complex as a control. Three independently prepared lysates were used for immunoprecipitation assays. Only the proteins pulled down in all 3 independent IP-MS assays were selected as candidates, and the mitochondrial candidates in the ATG4B antibody-precipitated complex are listed. (C) HepG2 cells were transfected with the 3 × Flag-ATG4B^WT expression plasmid, and then the cell lysates were subjected to immunoprecipitation assays. (D) ATG4B^CRISPR HepG2 cells were transfected with the 3 × Flag-ATG4B^WT expression plasmid. Then the cell lysates were obtained for immunoprecipitation and immunoblotting assays. 3 × Flag-ATG4B^S34A, 3 × Flag-tagged mutant ATG4B expression plasmid (in which Ser34 of ATG4B was mutated to Ala); Flag-ATG4B^WT, Flag-ATG4B^S34A and 3 × Flag-ATG4B^WT were the corresponding expression plasmids as described in Figure 1.

Figure S7C showed that the ATG4B puncta displayed minor colocalization with LC3 or GABARAPL1 puncta in GFP-ATG4B-overexpressing cells, suggesting that most of the GFP puncta may not be autophagy-protein positive dots under this condition. To further explore the function of mitochondrial ATG4B, we screened the potential mitochondrial binding partners of ATG4B in HCC cells using the combined methods of IP and mass spectroscopy (IP-MS). As shown in Figure 6B, 81 proteins from 3 independent experiments were screened out (only the proteins pulled down in all 3 IP-MS assays were selected), including nuclear proteins and cytoplasmic proteins (mitochondrial proteins, ribosomal proteins, cytosolic proteins, etc.). Some known ATG4B-binding proteins including GABARAPL2, MAP1LC3B and MAP1LC3A were among the candidates, further indicating the efficiency of the affinity isolation in this experiment. Interestingly, we found that ATP5C1 and ATP5A1 (2 components of the mitochondrial F₁Fo-ATP synthase) were also candidates with a high score. The Co-IP assay using the lysate of 3 × Flag-ATG4B^WT-expressing HCC cells showed that ATP5A1 and ATP5C1 were present in the immunoprecipitates obtained with anti-Flag antibody (Figure 6C). Furthermore, the Co-IP assay in Figure S7D showed that ATG4B present in the immunoprecipitates resulted from the anti-ATP synthase immunocapture antibody that could recognize multiple subunits of the F₁Fo-ATP synthase, indicating that ATG4B could be present in 1 complex with the F₁Fo-ATP synthase in HCC cells.

Subsequently, the effect of ATG4B on the activity of the F₁Fo-ATP synthase in vitro was evaluated. As shown in Figure S7E, recombinant human ATG4B (but not ATG4A) suppressed the activity of the F₁Fo-ATP synthase in MgATP SMPs from patient-derived HCC tissue (#4, Table S1). Additionally, the Co-IP assay in Figure 6D showed that the level of ATP5C1 was much higher in ATG4B^WT-immunoprecipitates than that in ATG4B^S34A-immunoprecipitates under the condition of PTEN knockdown, indicating that the enhanced interaction of ATP5C1 and the ATG4B-containing complex was dependent on ATG4B Ser34 phosphorylation. The data above, taken together with the results in Figures 4A-4C and Figures 5F-5I, indicate that the AKT-induced Ser34 phosphorylation of ATG4B can promote its mitochondrial import and the subsequent interaction with ATP5C1-containing complex, leading to the inhibition of F₁Fo-ATP synthase activity in HCC cells.

Discussion

To eliminate the influence of endogenous ATG4B, we intended to generate a homozygous ATG4B^−/− HCC cell line using the CRISPR/Cas9 system at the beginning of this study. However, ATG4B^−/− clones were not obtained in 4 consecutive rounds of CRISPR/Cas9 screening in HepG2 cells. We ultimately got a hemizygous knockout cell line, and its ATG4B level was approximately 50% of that in wild-type HepG2 cells. It has been reported that atg4b^−/− mice [61], atg4b^−/− mouse embryonic fibroblast (MEF) cells [42] and immortalized atg4b^−/− 3T3 cells [40] have been successfully generated. Nevertheless, to our knowledge, no ATG4B^−/− cancer cells have been generated until now. Meanwhile, Liu et al. have revealed that only knockdown of ATG4B presents a significant inhibitory effect on colorectal cancer cell growth in vitro and in vivo [40]. Therefore, we deduce that a certain level of ATG4B is required for the survival and/or growth of some cancer cells including HCC cells. In this study, we used the assays of overexpressing ATG4B^WT or ATG4B^S34A in this hemizygous knockout HCC cell line to investigate the Ser34 phosphorylation of ATG4B in order to reduce the influence of endogenous ATG4B as much as possible.

Previous reports have shown that AKT can either improve or impair OXPHOS through multiple mechanisms, and the final effect may be determined by different situations [62–64] Here, we found that the AKT-mediated phosphorylation of ATG4B at Ser34 dramatically enhanced the mitochondrial location of ATG4B and inhibition of mitochondrial function, which may favor the initiation of the Warburg effect. Therefore, suppressing the AKT-mediated Ser34 phosphorylation of ATG4B may be beneficial to attenuate cellular metabolic remodeling and delay HCC progression. Additionally, Yang et al. have revealed that inhibition of the AKT-downstream target MTOR in MEF or HEK293T cells by rapamycin or incubation in Hank's balanced salt solution induces autophagy via phosphorylating ATG4B at Ser383 and Ser392 [42]. It is interesting that the phosphorylation pattern of ATG4B can be changed according to the different states of the phosphoinositide 3-kinase-AKT-MTOR pathway or diverse cell backgrounds. In future work, we will thoroughly investigate the patterns of ATG4B phosphorylation in HCC cells under different situations.

Generally, the proteins that can locate within the mitochondria often contain a typical mitochondrial targeting peptide (MTP) in the N terminus. The MTP (usually 20∼30 amino acid residues) is cleaved by mitochondrial processing peptidases after the proteins are imported into mitochondria, leading to a decrease of approximately 2∼3 kDa of the protein molecular mass, accordingly. Of course, some proteins such as TST/rhodanese, HADHAP1 and BCS1L can be imported to mitochondria via a noncleavable mechanism [65–67], Moreover, Ruan et al. have reported that the cytosolic proteins prone to aggregation are imported into the mitochondrial intermembrane space and matrix in yeast after heat shock [68]. Meanwhile, Berendzen et al. have revealed that the aggregation-prone proteins expressed in the neurons of C. elegans bind to mitochondria, eliciting a global induction of mitochondrial stress and affecting whole-animal physiology [69]. In this study, we found that ATG4B did not contain a typical MTP in the N terminus, and the molecular weight of mitochondrial ATG4B was almost equivalent to that of cytoplasmic ATG4B, indicating a nonclassical pattern of ATG4B mitochondrial import. In addition, the results from GFP-ATG4B-expressing cells (Figure 5I) suggested that the Ser34 phosphorylation of ATG4B may enhance the formation of ATG4B aggregates, and the aggregates that presented minor colocalization with LC3- or GABARAPL1-puncta were prone to mitochondrial import in the AKT-activated cells. Here, we guess that the Ser34 phosphorylation of ATG4B may alter its conformation or the binding partners, which facilitates the aggregation of ATG4B and the subsequent mitochondrial import. Of course, more studies are needed to explore the detailed mechanisms.

It has been reported that some autophagy-related proteins (such as BECN1/Beclin 1 [70], ATG5 [71,72] ATG7 [73] and ATG4D [53,74]) can function in an autophagy-independent manner. As a key protease in autophagy process, ATG4B has aroused great interesting in the field of autophagic research. However, a recent study has revealed that ATG4B can promote the proliferation of colorectal cancer cells independent of autophagy [40]. In this study, we demonstrated that ATG4B could influence metabolic actions in HCC cells in a macroautophagy-independent manner, opening a new door for the research on the relationship between autophagy and mitochondrial function under some pathological conditions. Additionally, the ATG4B that modulates autophagy and the ATG4B that targets mitochondria may compete with each other because of the limitation of endogenous ATG4B in the cells. In the present study, maybe a portion of exogenous ATG4B is enough to regulate macroautophagy under basal conditions, or other ATG4s could compensate for the autophagy-regulating role of ATG4B, so the mitochondria translocation of ATG4B presented little effect on the macroautophagic signaling. The balance between the canonical and noncanonical roles of ATG4B in some physiological and pathological contexts is really interesting, and needs to be further studied.

As to the mechanism by which ATG4B regulates mitochondrial activity, we propose the following 3 preferential possibilities based on the previous reports and our data. (1) ATG4B regulates macroautophagy via cleaving LC3/GABARAP proteins and influences the clearance of mitochondria. (2) ATG4B is translocated into mitochondria and functions as a mitophagy receptor, which participates in the mitophagy-mediated mitochondrial degradation. (3) ATG4B is imported to mitochondria and regulates mitochondrial activity via directly affecting the function of mitochondrial proteins. In this study, we found that macroautophagy/mitophagy had no significant difference between ATG4B^WT- and ATG4B^S34A-overexpressing HCC cells whether or not AKT was activated, whereas the mitochondrial activity was impaired in ATG4B^WT-overexpressing cells with AKT activation. Therefore, we favor that the third possibility above by which ATG4B Ser34 phosphorylation regulates mitochondrial activity. Of course, the detailed mechanisms need to be further studied. It has been reported that chemotherapy and irradiation can upregulate ATG4B protein in several tumor cells including HCC cells [29,32,75] In addition, our unpublished data also indicated that the protein level of ATG4B was dramatically increased in HCC tissues (even more than a 5-fold increase in some samples), and the Ser34 phosphorylation of ATG4B could be detected in some clinical HCC samples. Therefore, we used the strategy of ATG4B overexpression so as to relate to the relevant biology in this study. Moreover, compared to the site directed mutation strategy for targeting endogenous ATG4B at Ser34, overexpression of ATG4B (WT and S34A) would induce little off-target effect, making it relatively specific for investigating the role of ATG4B Ser34 phosphorylation. Of course, the mutation of endogenous ATG4B at Ser34 will be more in line with the actual physiological condition compared to the overexpression strategy, and we will perform the corresponding experiments in the future.

Due to the key role of the ATP synthase in mitochondrial OXPHOS, the inhibition of ATP synthase activity or downregulation of ATP synthase subunits can obviously induce dysfunction of mitochondrial respiration, which finally contributes to the Warburg effect [17,50,51,76] Moreover, ATP synthase can also affect the production of mitochondrial superoxide radicals through regulating mitochondrial transmembrane potential hyperpolarization mediated by H⁺-translocation [77–80]. According to the findings in this study, we deduced that ATG4B may bind to the mitochondrial F₁Fo-ATP synthase and suppress its activity, leading to an increase of mitoROS in AKT-activated HCC cells. However, we also noticed that the percentage of mitochondria that colocalized with ATG4B was not high in our model, so we conjectured that the effect of mitochondrial ATG4B on a portion of F₁Fo-ATP synthase in HCC cells may be very sensitive, strong and cumulative, leading to a mild increase (even not detectable) of ROS production. This mild alternation of ROS may further influence mitochondrial function persistently and ultimately drives a detectable change in ROS level. Of course, more studies are needed to verify this hypothesis. Meanwhile, we cannot exclude F₁Fo-ATP synthase-independent mechanisms by which ATG4B affects mitochondrial activity, so more research is required to clarify the details. In addition, the mitochondria-targeting ATG4D can also lead to disruption of mitochondrial cristae architecture and sensitize cells to ROS-associated damage in differentiating erythroid cells [53]. These results indicate that ATG4s can influence ROS level via mitochondrial import. Furthermore, due to the diverse roles of ROS on cell biology, the mitochondrial ATG4s-mediated increase of ROS may present distinct functions under different situations, which needs to be further investigated.

In summary, we identify a novel phosphorylation site (Ser34) on ATG4B by AKT. Moreover, the ATG4B Ser34 phosphorylation participates in the metabolic reprogramming of HCC cells via suppressing mitochondrial function, which is possibly due to the Ser34 phosphorylation-mediated increase of ATG4B mitochondrial location and the subsequent inhibition of F₁Fo-ATP synthase activity. Our findings reveals a noncanonical working model of ATG4B under pathological conditions, which may provide a scientific basis for developing novel strategies for HCC treatment by targeting ATG4B and its Ser34 phosphorylation.

Materials and methods

Antibodies

The information of antibodies is as follows: TOMM20 (Santa Cruz Biotechnology, sc-17764); P4HB (Proteintech, 11245-1-AP); PEX1 (Proteintech, 13669-1-AP); ACTB (Proteintech, 60008-1-Ig); COX4I1 (Cell Signaling Technology, 6014); ATG4B (Abcam, ab70867); ATP synthase (Abcam, ab109867); HSPD1 (Abcam, ab46798); TOMM22 (Abcam, ab57523); ATG4A (Abcam, ab108322); Flag-tag (Proteintech, 66008-2-Ig, 20543-1-AP); ATP5A1 (Proteintech, 14676-1-AP, 66037-1-Ig); ATP5C1 (Proteintech, 10910-1-AP, 60284-1-Ig); Alexa Fluor 647 anti-rabbit IgG (Invitrogen, A21245); Alexa Fluor 594 anti-mouse IgG (Invitrogen, A21203); SQSTM1 (Cell Signaling Technology, 8025); LC3B-I/II (Cell Signaling Technology, 2775); GABARAPL1 (Abcam, ab86497); p-AKT (Cell Signaling Technology, 4060); MYC-tag (Cwbiotech, CW0089); phospho-AKT substrate (RXXS*/T*) (Cell Signaling Technology, 9614).

Plasmid construction

3 × Flag-ATG4B^WT, Flag-ATG4B^WT, Flag-ATG4B^S34A, Flag-ATG4B^C74A, Flag-ATG4B^S34A,C74A, MYC-AKT1^WT, MYC-AKT1^AA, GFP-ATG4B^WT, GFP-ATG4B^S34A and Flag-ATG4A^WT expression plasmids were prepared in our lab. To obtain the cDNAs of ATG4B and AKT1, total RNA was extracted from HepG2 cells and reverse-transcribed to cDNA by M-MLV Reverse Transcriptase (Invitrogen, 2641A) with the primer oligo(dT)18 (Takara, 3806). The human ATG4A cDNA was purchased from Guangzhou Genecopoeia Technology Co. Ltd in China. The cDNA fragments were separately cloned into pCMV5 (YouBio,VT2114) or pcDNA3.1 (YouBio,VT1001) vector. The mutant constructs were obtained from site-directed mutagenesis by overlap extension PCR, taking the corresponding wild-type constructs as templates.

siRNA and plasmid transfections

siRNAs were purchased from Genepharma Technology Co. Ltd in China. Transfections were performed using Lipofectamine 2000 (Invitrogen, 11668019) according to the manufacturer's protocol. Briefly, cells were seeded in culture plates and cultured overnight. Then siRNA or plasmid and Lipofectamine 2000 were separately diluted in Opti-MEM (Gibco, 31985070). After 5-min incubation, the diluted siRNA or plasmid and the diluted Lipofectamine 2000 were mixed gently and the mixtures were incubated for 20 min at room temperature. Subsequently, the cell culture medium was removed and the corresponding mixture was added into culture plates. After incubation for 6 h, the mixture was removed and the cells were cultured in fresh culture medium.

Purification of recombinant human ATG4B and ATG4A

The recombinant His-tagged human ATG4B (rhATG4B) and ATG4A (rhATG4A) proteins were prepared as previously described [81].

Isolation of subcellular fractions

The isolation of cytosol was performed as previously described [82]. The isolation of mitochondria was separately performed using a differential gradient centrifugation strategy [60] and anti-TOMM22 magnetic beads method [59] with minor revision. The corresponding detailed protocols were described in supplementary Materials and methods. The isolation of mitochondrial subfractions was performed according to the methods described previously [83].

Measurement of L-lactate and glucose

Cells were seeded into 6-well plates and transfected with different plasmids. Then the medium was subjected to the detection of L-lactate and glucose separately using the L-lactate assay kit (Eton Bioscience, 1200051002) and glucose colorimetric/fluorometic assay kit (BioVision, K606).

Determination of oligomycin-sensitive respiration

The detection of oligomycin-sensitive respiration (OSR) was performed as previously described [50]. Briefly, cells were seeded into XF24-well microplates and given different treatments. Then the cellular oxygen consumption was detected using an extracellular flux analyzer (Seahorse Bioscience, Seahorse XFe24) with the injected substances including oligomycin (1 μM; Seahorse Bioscience, 101706-100), FCCP (1 μM; Seahorse Bioscience, 101706-100), rotenone (0.5 μM; Seahorse Bioscience, 101706-100) and antimycin (0.5 μM; Seahorse Bioscience, 101706-100).

SMPs isolation

The isolation of MgATP SMPs was performed according to the previous method with minor modification [84]. Briefly, mitochondria separated from human HCC tissues were exposed to ultrasonics for 60 sec in the buffer containing 1 mM ATP, 15 mM MgSO₄ and 0.25 M sucrose (pH 7.6). Then the sonicate was centrifuged at 12,000 g for 15 min. Subsequently, the supernatant was centrifuged at 105,000 g for 30 min, and the pellet was washed and stored using 0.25 M sucrose (Santa Cruz Biotechnology, sc-204311). To activate the SMPs, a buffer containing 125 mM sucrose, 25 mM Tris-acetate and 2 mM EDTA (pH 7.8) was added and incubated at room temperature for 30 min. Then the SMPs were used to detect the activity of ATP synthase.

MitoROS detection

MitoROS was measured using the mitoROS-sensitive probe MitoTracker Red (Invitrogen, M7512) as described previously [85]. The fluorescence values of the cells stained with MitoTracker Red were detected at excitation/emission wavelength 575 nm/595 nm using a fluorescence microplate reader.

Co-IP assays

The cell extracts were diluted in lysis buffer (Beyotime, P0013) and incubated with anti-ATP synthase, anti-Flag or the corresponding mouse/rat monoclonal antibodies overnight at 4°C. Then protein-A/G agarose (Santa Cruz Biotechnology, sc-2003) was added to the mixtures and adequately resuspended. After incubation at room temperature for 90 min, the mixtures were centrifuged at 1,100 g for 3 min and washed with lysis buffer 7 times. Next, the immunoprecipitates were dissolved in lysis buffer containing loading buffer and heated at 100°C for 8 min, followed by SDS-PAGE for immunoblotting assays. In addition, for the IP-MS experiment, the immunoprecipitate was washed with 0.2 M glycine-HCl and the supernatant was subjected to MALDI TOF mass spectrometry for protein identification. The mass spectrometry was performed by Shanghai Applied Protein Technology Co. Ltd in China.

Immuno-electron microscopy

The samples of cultured HEK293 cells were prepared using a high-pressure freezing technique. Then the samples were stained with 1% uranyl acetate and dehydrated, infiltrated and embedded with the polymerization of Lowicryl resins (Polyscience, 14330). Subsequently, the ultra-thin sections of embedded cells with 80-nm thickness were prepared and incubated with specific antibody against ATG4B or a negative IgG control, followed by the incubation with gold colloid-conjugated goat anti-rabbit IgG (Sigma-Aldrich, G7402). Next, the immuno-protein complexes were stabilized with gluteraldehyde, and then the sections were imaged using a JEOL 1200 electron microscope.

Detection of ATP synthase activity

ATP synthase activity was detected with an F₁Fo-ATPase activity spectrophotometric assay kit (JianCheng, S50083) according to the manufacturer's instructions. The absorbance at 340 nm was measured continuously for 5 min. Additionally, the protein concentration was quantified by Bradford to normalize the reaction system. The unit of ATP synthase activity was defined as that required for oxidation of 1 μmol NADH/min/mg.

Immunoblotting assays

Immunoblotting assays were performed as described previously [81]. Briefly, the protein extracts were subjected to SDS-PAGE at 30∼80 µg/lane and transferred to PVDF membranes. Then the membranes were incubated with 5% milk in PBST (Boster, AR0194) or 3% BSA (Solarbio, A8020) in TBST (Boster, AR0195; for detection of phosphorylated proteins) at 37°C for 60 min, followed by incubation with different primary antibodies at 4°C overnight. After washing with PBST or TBST, the membranes were incubated with the corresponding secondary antibodies at 37°C for 60 min. Finally, the immunoreactive bands on the PVDF membranes were detected using enhanced chemiluminescence reagents (Thermo Scientific, 34076). Additionally, in order to separate the phosphorylated and nonphosphorylated proteins, the SDS polyacrylamide gel was prepared according to the manufacturer's instructions (Wako, 300–93523).

Statistical analysis

All data were displayed as mean ± SD (standard deviation) from at least 3 independent experiments. The difference between 2 groups was analyzed by Student t test, while the analysis of difference among 3 or more groups was performed by one-way ANOVA with the post hoc Tukey test. P < 0.05 was considered statistically significant.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (31671464, 81572375 and 31501061), National Postdoctoral Program for Innovative Talents (BX201600023), and Natural Science Foundation Project of Chongqing (cstc2017jcyjBX0079 and cstc2014jcyjA10110).