Autophagy impairment mediated by S-nitrosation of ATG4B leads to neurotoxicity in response to hyperglycemia

ABSTRACT

The majority of diabetic patients develop neuropathy and there is an increasing prevalence of neurodegeneration in the central nervous system (CNS). However, the mechanism behind this is poorly understood. Here we first observed that macroautophagy/autophagy was suppressed in the hippocampus of diabetic GK rats with hyperglycemia, whereas it was unchanged in ob/ob mice without hyperglycemia. Autophagy could be directly inhibited by high glucose in mouse primary hippocampal neurons. Moreover, autophagy was protective in high-glucose-induced neurotoxicity. Further studies revealed that autophagic flux was suppressed by high glucose due to impaired autophagosome synthesis illustrated by mRFP-GFP-LC3 puncta analysis. We showed that decreased autophagy was dependent on NO produced under high glucose conditions. Therefore, (LC-MS/MS)-based quantitative proteomic analysis of protein S-nitrosation was performed and a core autophagy protein, ATG4B was found to be S-nitrosated in the hippocampus of GK rats. ATG4B was also verified to be S-nitrosated in neuronal cells cultured with high glucose. The activities of ATG4B in the processing of unmodified, precursor Atg8-family proteins and in the deconjugation of PE from lipidated Atg8-family proteins, which are essential for efficient autophagosome biogenesis were both compromised by S-nitrosation at Cys189 and Cys292 sites. In addition, ATG4B processing of the GABARAPL1 precursor was affected the least by S-nitrosation compared with other substrates. Finally, ATG4B S-nitrosation was verified to be responsible for decreased autophagy and neurotoxicity in response to high glucose. In conclusion, autophagy impairment mediated by S-nitrosation of ATG4B leads to neurotoxicity in response to hyperglycemia. Our research reveals a novel mechanism linking hyperglycemia with CNS neurotoxicity and shows that S-nitrosation is a novel post-transcriptional modification of the core autophagy machinery.

Introduction

As many as 60% to 70% of diabetic patients develop neuropathy,¹ and the prevalence of neurodegeneration in the central nervous system (CNS) is reported to be 40%.² T2D patients have a greater risk of developing vascular dementia and Alzheimer disease, with brain atrophy and cognitive impairments,³ and CNS damage is also involved in peripheral neuropathy.⁴ Cognitive deficits can significantly affect the quality of life and greatly increase the economic burden for diabetic patients.¹ Research on the underlying mechanisms of CNS degeneration in diabetes has been of pivotal importance in recent decades.⁵ The Goto-Kakizaki (GK) rat is a spontaneous type 2 diabetic animal model that exhibits decreased neuronal viability and learning and memory deficits. It is the most frequently studied animal model for analyzing the mechanism of diabetes-associated neurodegeneration in the CNS.⁶ The hippocampus is a critical neurogenic area associated with memory and learning processes and is a vulnerable target for diabetic alterations.^7,8

Hyperglycemia is the most prevalent and predominant metabolic disorder among diabetic patients and has been viewed as the leading risk factor for neuronal injury.⁹ oxidative stress,¹⁰ mitochondrial membrane hyperpolarization,¹¹ and apoptotic neuronal loss¹² are thought to contribute to high glucose-induced neurotoxicity. Among these factors, oxidative stress accounts for much of the toxicity and has been proposed as major factor for diabetic neuronal injury in the CNS.¹³ R. Noriega-Cisneros et al. report that the nitric oxide (NO) level is increased in the brain mitochondria of streptozotocin (STZ)-induced diabetic rats.¹⁴ Nitrated proteins accumulate, and increased NOS1/nNOS and NOS2/iNOS expression is observed in human Schwann cells exposed to high glucose concentrations.¹⁵ The NOS (nitric oxide synthase) inhibitor L-nitro-arginine methylester (L-NAME) can blunt glucose-induced PC12 cell death,¹⁶ and prevent the loss of hippocampal neurons and restore cognitive abilities in STZ-induced diabetic rats.¹⁷ Moreover, various antioxidants have been designed to prevent or slow the development of diabetic neuropathy.¹³ NO primarily executes its biological function via protein S-nitrosation, a reversible post-translational modification that adds an NO moiety to cysteine thiol (−SH) groups on target proteins.¹⁸ S-nitrosation can influence protein activity, localization, conformation, or interactions with other proteins.¹⁸ Cumulative evidence suggests that aberrant protein S-nitrosation plays a crucial role in the pathogenesis of many neurodegenerative diseases.¹⁹

Autophagy is a cellular degradation pathway that clears protein aggregates and dysfunctional organelles. The cargoes are first engulfed in autophagosomes and then degraded upon fusion with lysosomes.²⁰ In recent years, the importance of autophagy has been appreciated in the pathophysiology of diabetes. Emerging evidence indicates that autophagic activity could be modulated by hyperglycemia in many diabetic complications, and dysfunctional autophagy is responsible for glucotoxicity.^21-23 Autophagy dysfunction has been implicated in the pathogenesis of many neurodegenerative disorders,²⁴ and has been viewed as a link between diabetes and Alzheimer disease.²⁵ However, whether and how autophagy participates in diabetic neuropathy in the CNS remains largely unknown. Many studies have reported that NO is involved in regulating autophagy under different stress conditions,^26-28 but the mechanism is obscure. Giuseppe Filomeni et al. suspect that the core autophagy machinery, which contains many cysteine proteases with abundant reactive cysteine residues, will be the principal target for reactive oxygen species and reactive nitrogen species.²⁹

This study explores whether and how autophagy participates in high-glucose-induced neurotoxicity in the CNS and the mechanism by which high glucose regulates autophagy. We show that autophagy impairment mediated by S-nitrosation of ATG4B leads to neurotoxicity in response to hyperglycemia.

Results

Autophagy is suppressed by hyperglycemia in the hippocampus

Hippocampal dendritic spine density and cell viability are decreased in 28-wk-old GK rats.⁶ Autophagy was examined in the hippocampal tissues of 30-wk-old GK rats with hyperglycemia. The LC3B-II level was significantly decreased, whereas the level of SQSTM1, an LC3-binding protein that is degraded by autophagy, was increased (Fig. 1A). SQSTM1 and ubiquitin are 2 well-characterized autophagy substrates, and we also found increased accumulation of SQSTM1-positive aggregates in the hippocampus of GK rats (Fig. 1B), but without evident ubiquitin accumulation (Fig. S1). Proteasome activity was not altered in the hippocampus of GK rats (Fig. 1C). In contrast, in the hippocampus of ob/ob mice on the C57BL/6J genetic background, the plasma glucose level was similar to the lean control, and autophagy was not influenced, with unchanged levels of LC3B-II and SQSTM1 (Fig. 1D), indicating that autophagy may be mainly regulated by hyperglycemia in the brain. Indeed, high glucose decreased LC3B-II levels in mouse primary hippocampal neurons in the presence or absence of chloroquine (CQ), an inhibitor of lysosomal acidification and autophagosome degradation (Fig. 1E), suggesting that autophagic flux may be inhibited by high glucose. We also noticed that the increase in osmolarity induced by mannitol also increased autophagy (Fig. S2), as reported in plants³⁰ and mammalian cells.³¹ Therefore, high glucose may have a strong inhibitory effect on autophagy.

Figure 1.

Protective autophagy in neuronal cells is suppressed by high glucose, and it is due to impaired autophagosome synthesis. (A) Levels of LC3B-II and SQSTM1 in the hippocampus of 30-wk-old Wistar and GK rats, n = 6/group. Two independent cohorts of rats were examined. (B) Anti-SQSTM1 (red) and anti-RBFOX3 (green) costaining in the hippocampus of Wistar and GK rats. n = 6/group. Bar: 10 μm. (C) Proteasome activity in the hippocampus of Wistar and GK rats was measured using AMC-linked substrate peptides. A, Ac-Gly-Pro-Leu-Asp-AMC; B, Suc-Leu-Leu-Val-Tyr-AMC; C, Ac-Arg-Leu-Arg-AMC; D, Boc-Leu-Arg-Arg-AMC. n = 4–5/group. (D) LC3B-II and SQSTM1 levels in the hippocampus of 24-wk-old ob/ob mice and lean controls. n = 4 or 5/group. Two independent cohorts of mice were examined. (E) LC3B-II levels in DIV 7 mouse primary hippocampal neurons cultured with basal glucose (BG, 25 mM) or high glucose (HG, 100 mM) in the presence or absence of chloroquine (CQ; 50 μM) for the last 1 h. Mannitol (Man.) was used at a concentration of 100 mM as an osmotic control. (F) Cell death in SH-SY5Y cells cultured under BG or HG conditions with rapamycin (Rap) or CQ cotreatment determined by PI and Hoechst staining. Bar: 100 μm. (G) Cell death in SH-SY5Y cells cultured under BG or HG conditions with wortmannin (WM) or CQ cotreatment estimated by flow cytometry after ANXA5-PER and 7-AAD staining. (H) Levels of cleaved CASP3 in SH-SY5Y cells that were treated as described in (F) and (G). A representative result from 3 independent experiments is shown. (I) LC3B-II levels in SH-SY5Y cells that were cultured with BG or HG in the presence or absence of CQ and SQSTM1 levels in the absence of CQ. (J) LC3 puncta examined by immunofluorescence in SH-SY5Ycells treated with BG or HG in the presence of CQ. Bar: 10 μm. (K) Confocal images of LC3 puncta in SH-SY5Y cells expressing mRFP-GFP-LC3 that were treated with BG or HG. Bar: 20 μm. 100 to 150 cells from each group were counted and 3 independent experiments were performed. Graphical data denote mean ± SEM. *P < 0.05; **P < 0. 01; ***P < 0.001; n.s., not significant.

Autophagy protects against high-glucose-induced neurotoxicity

We manipulated autophagic activity using a pharmacological method to explore the role of autophagy in high-glucose-induced neurotoxicity. As expected, human neuroblastoma SH-SY5Y cells that were exposed to high glucose (100 mM) exhibited an increased number of propidium iodide (PI)-positive dead cells. Enhancing autophagy with rapamycin (Rap), an MTOR inhibitor, attenuated glucotoxicity in SH-SY5Y cells, with substantially decreased numbers of PI-positive cells. Conversely, inhibiting autophagy with CQ exacerbated cell death (Fig. 1F). Cell death was also monitored by ANXA5/annexin V-phycoerythrin (PER) and 7-aminoactinomycin D (7-AAD) staining. Flow cytometry analysis showed that high glucose induced apoptosis and necrosis in SH-SY5Ycells, and the proportion of late apoptotic and necrotic cells was further increased when autophagy was inhibited by the PtdIns 3-kinase inhibitor wortmannin (WM) and CQ (Fig. 1G). These results were confirmed by western blot analysis, which showed increased levels of cleaved CASP3 in response to high glucose, and CQ further significantly increased CASP3 activation, whereas Rap almost completely abolished CASP3 activation (Fig. 1H). In summary, these results indicated that autophagy serves as a protective mechanism to promote neuronal cell survival, and high glucose induces neurotoxicity in neuronal cells by inhibiting autophagy.

High glucose reduces autophagic flux in neuronal cells by impairing autophagosome synthesis

We monitored autophagic flux in SH-SY5Y cells cultured with a basal (25 mM) or high concentration (100 mM) of glucose to explore which step of autophagy was regulated by high glucose levels. High glucose decreased both the steady-state and CQ-accumulated LC3B-II levels and the number of LC3 puncta (Fig. 1I and J). The degradation of SQSTM1 was also significantly decreased by high glucose (Fig. 1I). These results suggested that autophagic activity is decreased by high glucose and that the decrease in the LC3B-II levels is caused by reduced autophagosome formation rather than enhanced clearance.

Autophagy involves phagophore formation, autophagosome synthesis, and the fusion of autophagosomes with lysosomes to form autolysosomes. We monitored the synthesis of both autophagosomes and autolysosomes using mRFP-GFP tandem fluorescent-tagged LC3 (tfLC3) to further examine the change in autophagic flux. The GFP signal is quenched in the acidic environment of lysosomes; thus, autophagosomes are marked with both GFP and RFP, which appear yellow in the merged images, whereas autolysosomes are only labeled with RFP.³² Compared with the basal glucose-treated cells, the GFP⁺ RFP⁺ LC3 puncta (autophagosome), GFP⁻ RFP⁺ LC3 puncta (autolysosomes) and total RFP puncta (total autophagic vacuoles) were all markedly decreased when the cells were challenged with high glucose, indicating an impaired formation of both autophagosomes and autolysosomes (Fig. 1K). Collectively, we conclude that high glucose reduces autophagic flux by impairing the formation of autophagosomes in neuronal cells, thus attenuating the protective effect of autophagy and leading to glucotoxicity.

The inhibitory effect of high glucose on autophagy and cell viability is mediated by NO

The expression levels of NOS1 and NOS2, which may function in neuronal cells, were examined to evaluate whether nitrosative stress is present in neuronal cells exposed to high glucose. The expression levels of NOS1 but not NOS2 were increased in the hippocampus of GK rats, and NOS3/eNOS levels were not changed (Fig. 2A). We also observed similar results that NOS1 levels were increased, whereas NOS2 levels were unchanged in SH-SY5Y cells (Fig. 2B). Intracellular NO levels were detected using the DAF-FM DA method and were increased approximately 2-fold (Fig. 2C). A relatively selective NOS1 inhibitor, Nω-nitro-L-arginine (L-NNA), was used to inhibit NO synthesis to explore whether NO is the casual factor for the high-glucose-induced autophagic flux impairment. L-NNA could significantly restore LC3B-II levels under high-glucose conditions (Fig. 2D), suggesting that the inhibitory effect of high glucose on autophagy in neuronal cells is largely attributed to NO. We also found that L-NNA could significantly decrease the number of early apoptotic (ANXA5-PER⁺ 7-AAD⁻ cells) and all forms of dead cells (ANXA5-PER⁺ cells) under HG (Fig. 2E). Meanwhile, the level of cleaved CASP3 was decreased by L-NNA treatment under HG (Fig. 2F). These results suggested that the toxic effect of high glucose on the viability of neuronal cells also rely on NO.

Figure 2.

The inhibitory effect of high glucose on autophagy and cell viability is mediated by NO, and ATG4B is S-nitrosated. (A) Levels of NOS1 and NOS3 in the hippcampus of GK rats. (B) Levels of NOS1 and NOS2 in SH-SY5Y cells treated with BG or HG. (C) NO production in SH-SY5Y cells treated with BG or HG was measured using the DAF-FM DA method. (D) LC3B levels in SH-SY5Y cells cultured with BG or HG in the presence of the NOS inhibitor L-NNA (0.2 mM) cotreatment. (E) Cell death in SH-SY5Y cells cultured as indicated in (D) estimated by flow cytometry after ANXA5-PER and 7-AAD staining. (F) Levels of cleaved CASP3 in SH-SY5Y cells that were treated as described in (D). (G) ATG4B S-nitrosation levels in the hippocampus of GK rats compared with Wistar controls were detected by IBP. n = 4 to 6/group. Two independent cohorts of rats were examined. (H) Levels of S-nitrosated ATG4B in the hippocampus of ob/ob mice and lean controls. n = 4 or 5/group. Two independent cohorts of rats were examined. (I) Levels of S-nitrosated ATG4B in mouse primary hippocampal neurons cultured with BG or HG. (J) ATG4B S-nitrosation levels in SH-SY5Y cells treated with BG or HG in the presence of L-NNA. (K) ATG4B S-nitrosation in mouse primary hippocampal neurons treated with SNOC (500 μM, 30 min). The data represent the results from 3 independent assays. Graphical data denote mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; n.s., not significant.

ATG4B is S-nitrosated in response to high glucose

We performed an LC-MS/MS-based quantitative proteomics analysis of S-nitrosation by isobaric tandem mass tag (TMT) labeling in the hippocampus of diabetic GK rats to explore whether NO regulated autophagy via protein S-nitrosation. The MS analysis detected a total of 962 proteins, of which 80.2% (772 out of 962) proteins exhibited increased levels of S-nitrosation in the GK rats compared with the Wistar rat controls (with increased fold changes in the signal of TMT-127 relative to TMT-126). We also detected the level of protein S-nitrosation in the hippocampus of diabetic rats using anti-biotin blot after the IBP assay, and found a global increase in protein S-nitrosation in the GK rats compared with Wistar control (Fig. S3). Moreover, we adopted KEGG (Kyoto Encyclopedia of Genes and Genomes) enrichment analysis to identify pathways regulated by S-nitrosation to provide biological insights. About 300 proteins whose S-nitrosation level were increased in GK rats with the ratio of TMT-127 to TMT-126 above 1.2 were chosen, and KEGG terms with P value smaller than 0.01 were shown. As shown in Table S1, 9 KEGG pathways were enriched and these pathways are mainly associated with amino acid and glucose metabolism. We also performed the GO (Gene Ontology) enrichment analysis, and as shown in Table S2, GO terms associated with glucose metabolism pathway were also significantly enriched. These analyses suggested that these biological processes performed by certain proteins with increased S-nitrosation level may have a relatively high influence on diabetes and deserve further study.

Among proteins with increased S-nitrosation levels in GK rats, 33 proteins were autophagy- related proteins or autophagy regulators. Table S3 lists these proteins with their molecular function and fold change in the modification. Among these proteins, only one protein, ATG4B, directly participates in the autophagy process, and we proposed that the S-nitrosation of ATG4B may exert a significant impact on autophagy regulation. We confirmed that ATG4B could be S-nitrosated in the hippocampus of hyperglycemic GK rats by IBP (Fig. 2G). The specificity of ATG4B S-nitrosation signal in the GK rats assessed by IBP assay was verified when 2 negative controls were used. Exposing the lysates to a strong UV light source (“pre-photolysis”) before performing the IBP assay or exclusion of ascorbate during the labeling step of the IBP could both completely abolish ATG4B S-nitrosation signal (Fig. S4B). Moreover, there was a global increase in protein S-nitrosation in the GK rats and the increase was diminished in 2 negative control groups (Fig. S4C). However, in nonhyperglycemic ob/ob mice, ATG4B was not S-nitrosated in the hippocampus (Fig. 2H). ATG4B also underwent S-nitrosation by high-glucose-induced NO in mouse primary hippocampal neurons (Fig. 2I). Furthermore, L-NNA could completely abolish ATG4B S-nitrosation in high-glucose-treated SH-SY5Y cells (Fig. 2J). There was a significant ATG4B S-nitrosation signal in mouse primary hippocampal neurons that were directly exposed to the NO donor S-nitrosocysteine (SNOC) (Fig. 2K).

ATG4B processing activity toward Atg8 family precursors is repressed by NO, and GABARAPL1 is the least affect substrate

In the autophagy process, ATG4B processes the Atg8-family precursor proteins to their cleaved form, and also deconjugates Atg8-family proteins from PE at the autophagosome outer membrane for recycling of Atg8 family members during and after autophagosome formation.^33,34 The normal function of both of these processes enables efficient autophagosome biogenesis.^35,36 We first studied whether the processing activity of ATG4B toward Atg8-family precursors could be modulated by S-nitrosoglutathione (GSNO), an endogenous NO donor in vitro. ATG4B processing activity was measured using a previously established method.³⁷ Briefly, Atg8-family precursors were fused with a GST tag at the C terminus, and the level of cleaved GST product was used to measure ATG4B activity according to a GST standard curve (Fig. 3A and Fig. S5). We observed that GSNO attenuated ATG4B processing activity toward LCB-GST (Fig. 3B) and other Atg8-family precursors (data not shown) in a dose-dependent manner. Other NO donors, such as SNP and spermine NONOate, were also used, and both effectively compromised ATG4B activity (Fig. S6).

Figure 3.

ATG4B enzyme activity is compromised by NO. (A) Schematic diagram showing the method used to measure the ability of ATG4B to process Atg8-family precursors. The amount of GST determined from a GST standard curve was used to measure ATG4B activity. (B) Cleavage efficiency of ATG4B toward LC3B-GST after GSNO treatment. (C) The effect of GSNO on the kinetics of ATG4B catalysis. The initial velocity [V (mM/s), y axis] was calculated as described in Fig S7 and plotted against each substrate concentration. The concentrations of vehicle- or GSNO-treated ATG4B were 0.07 μM and 0.25 μM, respectively. The curves were repeated at least 3 times and a representative curve is shown. ((D)and E) Processing activity of ATG4B toward the GABARAPL1 precursor in cells. SH-SY5Y cells were starved with EBSS for 1 h and then exposed to different concentrations of SNOC, as indicated, for 30 min. Cell lysates (20 μg) with (D) or without (E) the DTT (10 mM, 10 min) treatment were incubated with the GABARAPL1-GST substrate (5 μM). The levels of GST were determined by immunoblot analysis with a GST antibody. The first lane represents uncleaved GABARAPL1-GST. ((F)and G) Reconstitution of the ATG4B deconjugating system for PE conjugated to Atg8-family proteins in vitro. (F) Preparation of PE conjugation to Atg8-family proteins. The reaction mixtures containing purified C. elegans ATG-7 (1 μM), C. elegans ATG-3 (5 μM), cleaved LC3B (10 μM), liposomes (1 mM, consisting of 50% DOPE), and ATP-Mg²⁺ (1 mM). The asterisk may represent the remaining ATG-3-LC3B intermediate after boiling with DTT. (G) Deconjugation of LC3B–PE by ATG4B. After the LC3B conjugation reaction, ATP was depleted with apyrase (1 U/ml), and then ATG4B (1 μM) was added. The products were immunoblotted using a His antibody. (H) ATG4B was treated with increasing concentrations of GSNO, as indicated, and incubated with the PE-conjugated Atg8-family orthologs. The reaction time was 30 min for LC3A-II and 10 min for the other Atg8-family protein orthologs. The products were immunoblotted with an anti-His antibody, and the levels of the remaining substrates were quantified and are shown under each lane. The concentrations of the initial substrates were arbitrarily defined as 100%. The first lane represents undeconjugated orthologs of the Atg8-proteins conjugated to PE. (I) Deconjugation activity of ATG4B toward GABARAPL1–PE in cells. SH-SY5Y cell lysates (100 μg) treated as described in (D) were incubated with GABARAPL1-II; then, the products were immunoblotted with a His antibody. Graphical data denote mean ± SEM. **P < 0. 01; ***P < 0.001.

Considering the divergent tissue distribution and functional roles of Atg8-family ortholog proteins, we performed a kinetic analysis to explore whether NO exhibited different inhibitory potencies on ATG4B activity toward different Atg8-family precursors (LC3A, LC3B, LC3C, GABARAPL1, GABARAPL2, and GABARAP). Meanwhile, for GSNO-treated ATG4B, ATG4B cleaved the substrates substantially more slowly; therefore we increased the amount of ATG4B 2.5-fold compared with vehicle-treated ATG4B. The initial velocity for each reaction was determined (Fig. S7) and plotted against each substrate concentration (Fig. 3C). The catalytic efficiency (K_cat/K_m) of ATG4B toward all Atg8-family precursors was dramatically diminished by GSNO, but the affinity (1/K_m) was almost unchanged after the statistical analysis, with the exception of LC3A-GST (Table 1). Interestingly, GNSO decreased the V_max values for all substrates except GABARAPL1, whose fold change in K_cat/K_m (vehicle/GSNO) was the smallest among all the homologues, indicating that ATG4B processing activity toward GABARAPL1 is the most resistant to NO (Fig. 3C and Table 1).

Table 1.

The kinetic parameters of the action of vehicle- or GSNO-treated ATG4B toward GST-tagged Atg8-family precursors.

	LC3A	LC3B	LC3C	GABARAPL1	GABARAPL2	GABARAP
Kcat /Km (105 mol−1ls−1)
ATG4B (vehicle)	4.97 ± 1.43	4.67 ± 1.24	4.36 ± 1.01	4.86 ± 0.69	6.36 ± 0.15	3.09 ± 0.09
ATG4B (GSNO)	0.47 ± 0.19**	0.58 ± 0.08**	0.32 ± 0.06**	1.19 ± 0.46***	0.41 ± 0.02***	0.33 ± 0.13***
vehicle:GSNO	10.67 ± 3.06	8.01 ± 2.13	13.69 ± 3.17	4.07 ± 0.58	15.72 ± 0.36	9.43 ± 0.29
Km (10−6 mol/l)
ATG4B (vehicle)	3.54 ± 0.91	1.33 ± 0.62	4.72 ± 1.50	3.90 ± 0.29	4.35 ± 0.33	1.20 ± 0.16
ATG4B (GSNO)	8.87 ± 1.09###	2.21 ± 1.21	4.72 ± 1.23	3.40 ± 0.65	8.25 ± 1.54	2.44 ± 1.18
GSNO:vehicle	2.51 ± 0.31	1.66 ± 0.91	1.00 ± 0.26	0.87 ± 0.17	1.90 ± 0.35	2.03 ± 0.98

The V_max (mol l⁻¹ s⁻¹) and K_m (mol/l) values for each reaction were derived from the curve that was fitted (Fig. 3C and Fig. S7) using the nonlinear regression method. Then, K_cat /K_m (mol⁻¹ l s⁻¹) was derived by dividing V_max by the K_m value and the enzyme concentration. The K_cat /K_m (vehicle/GSNO) and K_m (GSNO/vehicle) are also shown. *, ^# Compared with vehicle-treated ATG4B for each substrate. All data are presented as the mean ± SD n = 3.

^*P < 0.05,

^**P < 0.01,

^***P < 0.001,

^###P < 0.001.

The catalytic capacity of cell lysates from SNOC-challenged SH-SY5Y cells was examined to determine whether ATG4B processing activity could also be influenced by NO in living cells. ATG4B processing activity was diminished by NO in a dose-dependent manner, even toward the most resistant substrate, GABARAPL1, and NO did not alter ATG4B expression (Fig. 3D). Meanwhile, DTT completely recovered the lost ATG4B activity (Fig. 3E), indicating that NO exerts its effect via a reversible modification.

NO also suppresses the deconjugating activity of ATG4B toward conjugates of Atg8-family proteins and PE

PE conjugated to proteins of the Atg8 family was prepared using an in vitro conjugation system in the presence of ATG-3, ATG-7 (these 2 from Caenorhabditis elegans), cleaved Atg8-family proteins (with a C-terminally exposed glycine residue), liposomes, ATP and Mg²⁺. Proteins of the Atg8 family could only be ligated to PE when all components listed above were present, as exemplified by LC3B (Fig. 3F). Although LC3B was conjugated to PE with relatively low efficiency due to the absence of ATG12–ATG5-ATG16L1 E3-like ligase,³⁸ as shown by Coomassie Brilliant Blue (CBB) staining, the signal intensity of LC3B–PE in immunoblots for the N-terminal His tag was substantially higher than LC3B-I, due to the conformational change in the N terminus of LC3B after lipidation³⁹ (Fig. 3F and G). Other Atg8-family orthologs could also be successfully ligated to PE, with the exception of LC3C (probably due to extremely low efficiency due to the absence of E3-like ligase) (Fig. 3H). Then, we verified that these Atg8-family proteins conjugated to PE could be successfully deconjugated by ATG4B after ATP inactivation (Fig. 3G and H). A quantitative analysis of the relative amount of the Atg8-family orthologs that remained ligated to PE after deconjugation suggested that the capacity of ATG4B to deconjugate Atg8-family proteins from PE was compromised by NO in a dose-dependent manner (Fig. 3H). In the same regard, we also observed that ATG4B deconjugating activity toward GABARAPL1–PE was decreased in SNOC-treated SH-SY5Y cells (Fig. 3I).

S-nitrosation at Cys189 and Cys292 is responsible for the decreased ATG4B activity

We sought to determine the S-nitrosation sites of ATG4B to investigate the functional significance of S-nitrosation. First, we identified the S-nitrosation sites by single site-directed mutagenesis and IBP. Plasmids for all 13 single-site cysteine mutants (Cys to Ser) of ATG4B were constructed and transfected into HEK293 cells, and we found that ATG4B S-nitrosation was attenuated by the Cys301Ser mutant (Fig. S8A). Surprisingly, the Cys301Ser mutant could not rescue the loss of ATG4B enzyme activity, caused by GSNO (Fig. S8B and C), suggesting that S-nitrosation at Cys301 is not responsible for the decreased activity. We then explored other S-nitrosation sites by an LC-MS/MS analysis of GSNO-treated recombinant human ATG4B protein. Cys189 and Cys292 were revealed as S-nitrosation sites with a biotin-maleimide modification (Fig. 4A). The 3D structure of the ATG4B protein shows that Cys189 and Cys292 are both surface-exposed. Cys292 is surrounded by an acid-base pair (Asp106 and Arg103) near a hydrophobic residue (Phe293), all of which favor S-nitrosation.¹⁸ However, we cannot clearly explain the characteristics of Cys189, because it is located on the N terminus of a flexible loop (residues 190 to 216), which lacks a defined electron density (Fig. 4B). The Cys189Ser or Cys292Ser single mutants could not attenuate S-nitrosation, whereas the Cys189Ser,Cys292Ser double mutant (C189S,C292S) exhibited significantly attenuated S-nitrosation. Moreover, the triple-cysteine mutant (C189S,C292S,C301S) completely abolished S-nitrosation, suggesting that Cys189, Cys292 and Cys301 are all S-nitrosation sites (Fig. 4C). All the 13 cysteine residues and S-nitrosated sites in the primary structure of ATG4B were illustrated (Fig. 4D).

Figure 4.

ATG4B S-nitrosation leads to autophagy impairment and neurotoxicity in response to high glucose. (A) Tandem mass spectrum of the peptides showing S-nitrosation of Cys189 and Cys292 in the GSNO-treated recombinant ATG4B protein. Summary of biotin-M-containing peptides is shown, including the calculated monoisotopic masses (mass), accuracy of the mass measurements in parts per million (ppm), and cross correlation score (Xcorr). Asterisks indicate the biotinylation sites. (B) Surface representations of the ATG4B S-nitrosation sites. The color is coded from hydrophobic (red) to hydrophilic (white) according to the normalized consensus hydrophobicity scale,⁶³ and the S-nitrosation sites are yellow. The images were prepared using PyMOL. (C) S-nitrosation status of ATG4B in HEK293 cells expressing WT or the cysteine mutations (Cys to Ser) ATG4B treated with SNOC. *indicates a significant difference compared with the SNOC-treated WT group. (D) Protein sequence of human ATG4B with all the 13 cysteine residues in bold and S-nitrosated sites highlighted in red. (E) The processing activity of WT or double mutant ATG4B^C189,292S after GSNO treatment. (F) The deconjugating activity of WT or ATG4B^C189,292S after GSNO treatment. The first lane represents undeconjugated LC3B–PE. (G) Cellular thermal shift assay (CETSA) curves of WT or ATG4B^C189,292S in HEK293 cells after SNOC treatment. The curves were repeated 3 times and a representative curve is shown. (H) LC3 puncta (detected by anti-LC3) in WT or ATG4B^C189,292S (with a GFP tag)-transfected SH-SY5Y cells cultured with BG or HG, with CQ (10 μM) added for the last 24 h. Bar: 10 μM. (I) SH-SY5Y cells expressing WT or ATG4B^C189,292S (with GFP tag) following a lentivirus infection were treated with BG or HG; then, the cytotoxicity was estimated by flow cytometry after ANXA5-PER and 7-AAD staining. 2CS: ATG4B^C189,292S. Graphical data denote mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; n.s., not significant.

We studied the function of S-nitrosation at Cys189 and Cys292 and found that the C189S,C292S double mutant completely compromised the inhibitory effect of NO on ATG4B processing (Fig. 4E) and deconjugating activities (Fig. 4F). Therefore, ATG4B S-nitrosation at Cys189 and Cys292 was responsible for its decreased activity. Cellular thermal shift assay showed increased thermal stability for WT ATG4B after SNOC treatment, whereas there was nearly no change for the ATG4B^C189,292S double mutant (Fig. 4G). We also found that GSNO treatment caused a significant decrease in the intrinsic fluorescence intensity of ATG4B and a slight blue shift of maximum emission wavelength (from 333 nm to 331 nm; Fig. S9), suggesting that the tryptophan residues of ATG4B are embedded within a more hydrophobic microenvironment after S-nitrosation. These results suggest that S-nitrosation decreases ATG4B activity probably due to more rigid protein structure and increased stability.

ATG4B S-nitrosation leads to autophagy impairment and neurotoxicity in response to high glucose

Firstly, autophagy was examined in HeLa cells expressing WT and ATG4B^C189,292S exposed directly to NO donor to explore the role of ATG4B S-nitrosation in NO-mediated autophagy inhibition. SNOC severely decreased LC3B-II levels, but the C189S,C292S double mutant increased LC3B-II levels compared with WT ATG4B (Fig. S10A). Furthermore, we observed that the C189S,C292S,C301S triple mutant could not further rescue the autophagy level compared with the C189S,C292S mutant (Fig. S8D). We also noted that the overexpression of both WT and ATG4B^C189,292S could decrease LC3B-II levels (Fig. S10A), as previously reported.⁴⁰ SNOC also decreased the formation of ATG16L1-labeled phagophores in cells expressing WT ATG4B, whereas the C189S,C292S double mutant increased phagophore formation compared with WT ATG4B (Fig. S10B). ATG16L1 is located on the phagophores and is not found on the autophagosomes;⁴¹ therefore, these data strongly suggest that NO impairs autophagosome formation by halting phagophore expansion via ATG4B S-nitrosation.

Endogenous LC3 puncta were detected in WT and ATG4B^C189,292S-transfected SH-SY5Y cells to elucidate whether ATG4B S-nitrosation was responsible for autophagy impairment and neurotoxicity in response to high glucose. The number of LC3 puncta was significantly decreased by high glucose in the WT ATG4B group, whereas the C189S,C292S double mutant increased the number of LC3 puncta compared with WT ATG4B (Fig. 4H). Cell death in the WT and ATG4B^C189,292S-expressing SH-SY5Y cells that were transduced by lentiviral infection was measured by flow cytometry after ANXA5-PER and 7-AAD staining. As shown in Fig. 4I, high glucose induced late apoptotic and necrotic cell death in both types of cells, as shown by ANXA5-PER- and 7-AAD-positive staining. However, the proportion of dead cells was reduced in the C189S,C292S group compared with the WT group. Together, these results suggest that ATG4B S-nitrosation impairs autophagy and leads to neurotoxicity in response to high glucose.

Discussion

We showed that neuronal cells produce excess levels of NO in response to high glucose, which leads to S-nitrosation of ATG4B. ATG4B S-nitrosation could also be observed in the CNS of a diabetes model such as GK rats with hyperglycemia. S-nitrosation diminishes ATG4B activity in processing Atg8-family protein precursors and in deconjugating proteins of the Atg8 family from PE. The impairment of these 2 processes blocks the formation of phagophores. Ultimately, the number of autophagosomes and autolysosomes are not sufficient to ensure proper autophagic flux, resulting in neurotoxicity (Fig. 5). Our work revealed a novel mechanism linking hyperglycemia with neuronal cell damage and also indicated that S-nitrosation is a novel post-transcriptional regulation pathway for the core autophagy machinery.

Figure 5.

Model for the role of ATG4B S-nitrosation in high-glucose-induced neurotoxicity. In a hyperglycemic environment, such as the milieu in diabetic GK rats, the NO level in the hippocampus is increased, and ATG4B undergoes concomitant S-nitrosation. The S-nitrosation of ATG4B contributes to compromising the enzyme activity of both processing Atg8-family precursors and deconjugating Atg8-family members from PE, which prevents phagophore expansion and the subsequent formation of autophagosomes and autolysosomes. As a result, protective autophagy is suppressed, leading to neuronal cell death in the CNS.

Because autophagy promotes neuronal cell survival, the observed autophagy impairments in the hippocampal tissues of GK rats at 30 wk of age may lead to decreased hippocampal dendritic spine density and cell viability.⁶ Though autophagy was decreased in GK rats, it is worth noting that the modulation of autophagy in other diabetic models is distinct. For example, autophagy is induced in the hippocampus of young patients with poorly controlled type 1 diabetes.⁴² Autophagy is inhibited in the cortex of sucrose-induced diabetic mice⁴³ and cerebellar Purkinje neurons of STZ-induced diabetic rats.⁴⁴ Meanwhile, abnormal autophagosome accumulation is observed in the brains of db/db and high fat-diet (HFD)-induced diabetic mice.⁴⁵ We also verified that the LC3-II levels in HFD mice were increased at both 10 wk and 62 wk of age (Fig. S11). The inconsistency of these examples could be explained by the fact that the observed changes in autophagy may be the net effect of multiple factors. In addition to hyperglycemia, several coexistent metabolic alterations can also regulate autophagy, such as insulin deficiency or resistance, LEP (leptin), dyslipidemia, and obesity-related chronic inflammation. Insulin is well known to inhibit autophagy by phosphorylating and inactivating ULK1;⁴⁶ thus, defective insulin signaling would promote autophagy. Son et al. have shown that insulin resistance induces the accumulation of autophagosomes in SH-SY5Y cells.⁴⁵ LEP induces autophagy in many cell types;^47,48 thus, LEP deficiency is suspected to suppress autophagy. Palmitate has also been reported to inhibit autophagy in a neuronal cell line.⁴⁹ Furthermore, the brain regions and the stage of the disease examined are different in these models, which may also exert different influences on autophagy. GK rats exhibit both hyperglycemia and insulin resistance, but we observed that autophagy was decreased in the hippocampus. Meanwhile, in C57BL/6J ob/ob mice, whose blood glucose level is nearly the same as the lean controls, autophagy was not changed (Fig. 1D). These results indicate that hyperglycemia is a vital casual factor that inhibits autophagy in the CNS of diabetic subjects. In conclusion, we propose that autophagy may be directly downregulated by hyperglycemia in diabetes, and the mechanism we explored is vital for diabetic patients whose predominant metabolic disorder is hyperglycemia.

High glucose can modulate autophagy through the reactive oxygen species and reactive nitrogen species, AMPK and IKK-NFKB1 pathways,⁵⁰ and different pathways participate in regulating autophagy in each cell type and will result in different outcomes. Actually, high glucose differentially regulates autophagy in many cell types.^23,51 Our results illustrate that L-NNA could significantly restore autophagy suppressed by high-glucose, suggesting that NO may be the primary causal factor for glucotoxicity in neuronal cells. It has been reported that glucose-induced PC12 cell death is blunted by the NOS inhibitor L-NAME.¹⁶ L-NAME could also significantly reverse the cognitive deficits and associated biochemical alterations in STZ diabetic rats, whereas L-arginine, an NO donor, accentuates the behavioral and biochemical deficits.¹⁷ Although NO has a vital pathological role in many diseases, NOS inhibitor-therapies are not available in clinical trials because they can nonspecifically block the physiological function of NO as well. For this reason, a specific alteration of SNO-proteins would be a promising approach. The catalytic Cys74 of ATG4B is not an S-nitrosation site, so small molecules could be designed to specifically block Cys189 and Cys292, thus preventing their S-nitrosation without influencing ATG4B enzyme activity. Therefore, our research indicates that ATG4B S-nitrosation may be a good therapeutic target to alleviate vascular dementia.

At present, the comprehension of autophagic regulation by stress stimuli has only just begun, and its link to diseases awaits exploration. Many components and events of the core autophagy machinery are subjected to post-translational modification, such as the following: the phosphorylation of ATG4B, ATG5, and Atg8-family proteins; the ubiquitination of ATG4B; the acetylation of ATG5, ATG7, Atg8-family proteins, and ATG12; and the oxidation of ATG4A/B.^52-54 However, these events occur under basal or starvation-induced autophagy conditions. Our research indicates that a core autophagy protein, ATG4B, could be S-nitrosated in diabetes with hyperglycemia. Although the S-nitrosation of signaling molecules, such as MAPK8/JNK1-MAPK9/JNK2-MAPK10/JNK3 and IKBKB/IKKβ), has been reported to regulate autophagy,²⁶ our research is the first to illustrate that the core autophagy machinery could directly be regulated by S-nitrosation. Therefore, S-nitrosation is a new form of post-transcription modification of the core autophagy machinery.

ATG4B processing of the GABARAPL1 precursor was affected the least by S-nitrosation compared with other substrates. We supposed that for neuronal cells under hyperglycemic conditions, this resistance is actually an attempt to support survival by maintaining the participation of GABARAPL1 in autophagy. The GABARAPL1 protein sequence is highly conserved through evolution,⁵⁵ and GABARAPL1 displays the highest expression levels in the CNS among the Atg8 protein family.⁵⁶ A study using a microarray analysis shows that GABARAPL1 expression, but not that of other Atg8 family proteins, is significantly decreased in the brains of patients with Parkinson disease.⁵⁷ These lines of evidence indicate the importance of GABARAPL1 in neurons. Sixin Jiang et al. report the involvement of GABARAPL1 in glycogen hemostasis.⁵⁸ STBD1/genethonin 1 acts as a receptor protein in autophagy for the vesicular trafficking of glycogen to lysosomes for hydrolysis through extensive interactions with GABARAPL1. Therefore, it is possible that the proper function of GABARAPL1 could eliminate the toxic effect of glycogen in diabetic conditions with excessive glycogen deposits in neurons.⁵⁹ The Atg8 family of proteins also determines the selectivity of autophagy by binding with different receptor proteins. BNIP3L/Nix-mediated mitophagy occurs through an exclusive interaction with GABARAPL1.⁶⁰ Thus, for organs with a high-energy demand, such as the brain, proper GABARAPL1 function in maintaining mitochondrial homeostasis is particularly important for neuronal survival. Nonselectivity of ATG4B toward each substrate is observed under normal conditions,³⁷ and our work shows that S-nitrosation regulates ATG4B processing activity in a substrate-specific manner in response to oxidative stress. The function of each ortholog of the Atg8 family is still unclear, and our findings may provide clues toward an understanding of the roles of different members of the Atg8 protein family under different pathological conditions.

In this study, we defined the ATG4B S-nitrosation-mediated autophagy impairment as a possible linkage between diabetes-related hyperglycemia and neuronal damage. We also show that S-nitrosation is a new form of post-transcription modification of the core autophagy machinery. Our work also indicates that the inhibition of ATG4B S-nitrosation may represent a novel approach to alleviate diabetic dementia. One limitation of our study is that further studies are needed to prove whether the S-nitrosation-resistant mutant ATG4B^C189,292S could prevent CNS damage in diabetes in vivo, optimally using an ATG4B^C189,292S knock-in diabetic animal model.

Materials and methods

Reagents

The following reagents were purchased from Sigma: sodium ascorbate (A7631); neocuproine (N1501); N-ethylmaleimide (NEM, E3876); CQ (C6628); methylmethanethiosulfonate (MMTS, 208795); biotin-maleimide (biotin-M, B1267); SNP (228710); PI (P4170); and Hoechst 33342 (B2261). Rap (13346) and WM (10010591) were purchased from Cayman. Neutravidin resin was obtained from Pierce (29201). NG-nitro-L-arginine (L-NNA, sc-3570) and spermine NONOate (sc-202816A) were obtained from Santa Cruz Biotechnology. ATP (P0756S), apyrase (M0393S), trypsin (P8101S), GluC (P8100S) and the restriction endonucleases were purchased from New England Biolabs. DAF-FM DA was obtained from Beyotime (S0025). DOPE (850725P), DOPC (850375P), and DOPA (840875P) were purchased from Avanti Polar Lipids. S-nitrosoglutathione (GSNO) and S-nitrosocysteine (SNOC) were synthesized from glutathione using acidified nitrite.

Antibodies

The following primary antibodies were used in the western blotting assays: mouse anti-ACTB/β-actin (Santa Cruz Biotechnology, sc-8432); mouse anti-NOS1/nNOS (Santa Cruz Biotechnology, sc-55521); mouse anti-NOS2/iNOS (Santa Cruz Biotechnology, sc-7271); rabbit anti-NOS3/eNOS (Cell Signaling Technology, 32027S); rabbit anti-GST (Epitomics, 3186–1); rabbit anti-SQSTM1 (Medical & Biological Laboratories, PM045); mouse anti-LC3B (Medical & Biological Laboratories, M186–3); mouse anti-His (CWBIO, CW0082); rabbit anti-ATG4B (Sigma, A2981); rabbit anti-cleaved CASP3 (Cell Signaling Technology, 9661S) and HRP-conjugated streptavidin (Thermo scientific, 89880D). The following primary antibodies were used in the immunofluorescence analysis: rabbit anti-ATG16L1 (MBL, PM040); rabbit anti-LC3 (Cell Signaling Technology, 2775S); rabbit anti-SQSTM1 (Medical & Biological Laboratories, PM045); rabbit anti-ubiquitin (Cell Signaling Technology, 3933); and mouse anti-RBFOX3/NeuN (Abcam, ab104224).

Plasmid construction and mutations

For expression in mammalian cells, the coding region of human ATG4B was subcloned into BamHI and XhoI sites of the mammalian expression vector pLVX-AcGFP1-N1 (Clontech, 632154) (WT ATG4B-GFP). Cys189Ser and Cys292Ser double-mutant ATG4B mammalian expression vector (ATG4B^C189,292S-GFP) was generated by overlap PCR based Mutagenesis (regular PCR using overlap primers containing mutant base-pairs). ATG4B^{C189,292,301S}-GFP mammalian expression vector was generated by the same methods. For prokaryotic protein expression, human ATG4B and the cleaved ATG8-family orthologs were cloned into the pET-28a (+) vector (Novagen, 69864–3) and fused with the N-terminal His6 tag. Plasmids with GST-tagged ATG8-family sequences were constructed by inserting the coding regions of the ATG8-family precursors into a modified pET-28a (+) vector with a GST tag at the C terminus. C. elegans atg-3 and atg-7 were cloned into the pGEX-6P-1 vector (GE Healthcare, 28–9546–48) with a GST tag. Site-directed mutations of ATG4B were generated by the overlap PCR method.

Protein expression and purification

Proteins were expressed in BL21 (DE3) cells after isopropyl β-D-1-thiogalactopyranoside (IPTG; Amresco, 0487) induction (500 μM at 16°C for 16 h). C. elegans ATG-3 and ATG-7 were purified with a GST-affinity column (GE Healthcare, 17–0756–01), and GST was cleaved by HRV3C protease (Sigma, H9916). Human WT and mutant ATG4B, the GST-tagged Atg8-family proteins, and cleaved forms of Atg8-family orthologs were purified by affinity chromatography using a Ni²⁺-NTA resin (GE Healthcare, 17–5318–01). After elution with imidazole (250 mM), the proteins were dialyzed against TBS buffer containing 25 μM DTT.

SDS-PAGE and western blotting

After SDS-PAGE, the gels were subjected to CBB staining or transferred to nitrocellulose (NC; Pall, 66485) or polyvinylidene fluoride (PVDF; Millipore, IPVH00010) membranes. After incubation with the appropriate primary antibodies and the corresponding HRP-conjugated secondary antibodies (Santa Cruz Biotechnology, sc-2004 and sc-2005), the signals were detected using the enhanced chemiluminescence ECL kit (Thermo Scientific, 89880) and the Molecular Image ChemiDoc XRS image system (Bio-Rad, Hercules, CA, USA). The band densities were quantified and analyzed by densitometry using ImageLab software (Bio-Rad).

Determination of the kinetic parameters of ATG4B processing activity

The kinetic parameters were calculated using a previously established method.³⁷ Purified ATG4B and GST-tagged Atg8-family precursors (5 μM) were incubated together at 37°C for different time periods, and then the reaction mixtures were stopped and resolved on SDS-PAGE gels. After CBB staining, the densities of the substrates (GST-tagged Atg8-family proteins) and cleaved products (GST and Atg8-family proteins) were measured to determine the percentage of remaining substrates using the equation OD_{GST-tagged Atg8-family proteins}/(OD_{GST-tagged Atg8-family proteins} + OD_GST + OD_{Atg8-family proteins}) × 100%, and the results were then plotted against the reaction times. The optimal reaction times were determined from this curve for further kinetic analysis, at which the reaction is linear. Subsequently, decreasing concentrations of GST-tagged Atg8-family precursors (from 1.25 to 10 μM) were incubated with a fixed concentration of ATG4B for the optimal time period, as described above. The initial velocity (V, y axis) for each substrate concentration was determined from the change in GST product concentration per second (mM/s) using a GST standard curve. The initial velocity was then plotted against the substrate concentration (S (μM), x axis) for each reaction and the resulting curves were fitted by the nonlinear regression method (GraphPad 6.00, San Diego CA), from which the values of V_max (mol liter⁻¹ s⁻¹) and K_m (mol liter⁻¹) were derived. Then, the catalytic efficiency, K_cat/K_m (mol^−1 L s⁻¹), was derived by dividing V_max by the K_m value and the enzyme concentration.

Liposome preparation

Phospholipids (1.25 mM) composed of 50% dioleoylphosphatidylethanolamine (DOPE), 40% 1-palmitoyl-2-oleoylphoshatidylcholine (POPC) and 10% dioleoylphosphatidic acid (DOPA) dissolved in chloroform were placed in a glass tube. Chloroform was removed with a desiccator under vacuum overnight. Then, the resulting lipid film was suspended in a buffer (25 mM Tris-HCl, pH 7.5, 137 mM NaCl, 2.7 mM KCl) at a final concentration of 2.5 mM. After hydration at room temperature for 1 h, the lipid film was suspended by vigorous vortexing. Subsequently, the lipid suspension was subjected to repeated freeze thaw cycles at −70°C and 42°C and sonication to obtain small unilamellar liposomes. Finally, the lipid suspension was centrifuged at 20,000 × g for 10 min at 4°C, and the supernatant was collected.

Preparation of PE-conjugated Atg8-family proteins and ATG4B deconjugation assay

The PE-conjugated forms of proteins of the Atg8-family were produced in a reaction mixture containing the following components: 1 μM ATG-7; 5 μM ATG-3; 10 μM cleaved Atg8-family proteins (with an exposed glycine at the C terminus); 1 mM liposomes (the final PE concentration was 500 μM); 1 mM ATP; and 1 mM Mg²⁺ in a volume of 20 μl. After incubation at 30°C for 3 h, the reaction was stopped with Laemmli sample buffer, and the conversion of form I to form II of protein orthologs of the Atg8 family was detected by immunoblotting with the His antibody. After Atg8-family protein conjugation, apyrase (1 U/ml) was added to the mixture and incubated for an additional 45 min to stop the conjugation reaction by depleting ATP. Then, the ATG4B protein or cell lysates were incubated with this reaction mixture at 37°C for the indicated times. The percentage of the remaining form II of proteins of the Atg8 family was detected by immunoblotting with the His antibody.

Detection of protein S-nitrosation using the irreversible biotinylation procedure (IBP)

Protein S-nitrosation was detected using the irreversible biotinylation procedure (IBP), as described previously,⁶¹ which is an improved method based on the original biotin-switch assay.⁶² Briefly, cells or hippocampal tissues were homogenized with HEN buffer (250 mM HEPES, pH 7.7, 1 mM EDTA, 0.1 mM neocuproine) with 1% NP40 (Abcam, ab142227) and complete protease inhibitor cocktail (Roche, 000000004693132001). Free thiols were blocked with 20 mM methyl methanethiosulfonate (MMTS) at 50°C for 40 min, then the proteins were precipitated with acetone and washed with 80% acetone twice to remove excess MMTS. Nitrosothiols were reduced and biotinylated with 10 mM ascorbate and 0.2 mM biotin-maleimidein at room temperature for 2 hr. After precipitation and washing with acetone, the protein pellet was resuspended in HENS buffer containing 200 mM DTT and boiled for 10 min to reduce potential intermolecular disulfide bonds. Biotinylated proteins were pulled down with neutravidin-agarose beads, eluted by Laemmli sample buffer, and resolved through western blot analysis with an anti-ATG4B antibody.

LC-MS/MS determination of the S-nitrosation sites

Purified His-ATG4B protein (10 μM, 0.2 mg) was treated with 200 μM GSNO at RT for 1 h; then, the free thiols were blocked with N-ethylmaleimide (NEM), and the S-nitrosated proteins were labeled with biotin-maleimide, as described for IBP. After protein digestion with trypsin or GluC, the S-nitrosated peptides were purified with neutravidin agarose (Pierce, 29201) and subjected to LC-MS/MS analysis. The peptides were separated on C18-AQ trapping columns (i.d., 150 μm; resin, 5 μm) and C18-AQ analytical columns (i.d., 75 μm; resin, 3 μm), eluted, and then ionized and analyzed by a LTQ Orbitrap Velos mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA). A survey full scan followed by 10 CID events were used, and dynamic exclusion for selected precursor ions was set at 120 s. The MS/MS spectra were searched against the UniProt human proteome database (Release 2013_06) using SEQUEST with the following parameters: 10 ppm mass tolerance for precursor ions; 0.8 Da mass tolerance for product ions; and 2 missed cleavage sites for trypsinized peptides. The variable modifications included oxidation on methionine (M, +15.9949 Da), N-ethylmaleimide (C, +125.0477 Da) and biotin-maleimide (C, +451.1889 Da) modification on cysteine.

Cell culture, transfection and treatment

Primary hippocampal neurons from E18 C57BL/6J embryos were prepared as described previously.⁶³ The dissociated cells were cultured in Neurobasal medium (Invitrogen, 21103–049) containing B27 supplements (Invitrogen, 17504–044). HEK293, HeLa and undifferentiated SH-SY5Y cells were cultured in Dulbecco's modified Eagle's medium (Hyclone, SH30243.01) supplemented with 10% fetal bovine serum (GIBCO, 10019–141), 2 mM L-glutamine (Invitrogen, 35050–061), 100 U/ml penicillin, and 100 mg/ml streptomycin (Hyclone, SV30010). The cells were transfected with Lipofectamine 2000 (Life Technologies, 11668–019) or lentivirus according to the manufacturer's instructions.

When neuronal cells were subjected to high-glucose treatment, days in vitro (DIV) 7 mouse primary hippocampal neurons were cultured in Neurobasal medium with a basal (BG, 25 mM) or high dose (HG, 100 mM) of glucose for 72 h. SH-SY5Y cells were cultured in serum-free DMEM with a basal (BG, 25 mM) or high dose (HG, 100 mM) of glucose for 48 h. Chemical compounds used for the treatment in cells were 100 nM rapamycin for 48 h, 0.5 μM wortmannin for 48 h, 0.5 mM SNOC for 30 min, 50 μM chloroquine for 1 h for autophagic flux analysis and 10 μM chloroquine for 24 h for cell death analysis.

Cellular thermal shift assay

HEK293 cells treated as indicated were lysed in buffer (250 mM HEPES, pH 7.7, 1 mM EDTA, 1% NP40, 1x complete inhibitor cocktails and 1 mM PMSF). After centrifugation, cell lysates were aliquoted and individually heated to the indicated temperatures for 3 min, and then cool at room temperature for 3 min. Insoluble proteins were separated by centrifugation, and the soluble proteins were used for SDS-PAGE followed by western blotting analysis.

Cell death measurement by propidium iodide and Hoechst staining

High-glucose-induced neuronal cell death was measured by PI and Hoechst staining. SH-SY5Y cells were stained with Hoechst 33342 followed by PI (5 μg/ml) for 10 min; then, the PI was removed, and the cells were immediately examined and imaged with a confocal microscope (LSM 710 Meta confocal microscope, Carl Zeiss, Jena, Germany). More than 200 cells were counted in each experiment.

Lentivirus production

The cDNAs for WT and mutant ATG4B^C189,292S were amplified by PCR and cloned into the pLE4 lentiviral vector (a gift from Dr. Tomoaki Hishida) to generate lentiviral vectors that express WT and mutant ATG4B^C189,292S. The lentiviral vector expressing mRFP-GFP-LC3 was purchased from Addgene (84573; deposited by Noboru Mizushima). HEK293T cells were transfected with these lentiviral vectors together with packaging plasmids pMD2.G (Addgene, 12259; deposited by Didier Trono) and psPAX2 (Addgene, 12260; deposited by Didier Trono). Supernatants containing the virus were harvested at 48 and 72 h after transfection, filtered with a 0.45-μm PVDF membrane (Millipore, SLHV033RB), and then concentrated by ultracentrifugation.

Measurement of the intracellular NO levels with DAF-FM DA fluorescence

After incubation with the indicated treatments, the cells were washed 2 times with phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 2 mM KH₂PO₄) and incubated with DAF-FM DA (10 μM) for 30 min at 37°C. After removing the excess probe by extensive washing, the cells were trypsinized, suspended in PBS, and then plated in a 96-well black assay plate (Corning, 1817940). DAF fluorescence was measured using a fluorescence microplate reader (Thermo Scientific, Varioskan, USA) at excitation and emission wavelengths of 488 and 515 nm, respectively. The relative fluorescence values were corrected by the protein concentration for each treatment.

Immunofluorescence staining

Cells grown on a glass-bottom dish were fixed with 4% paraformaldehyde for 10 min, permeabilized with Triton X-100 (Amresco, M143–1L) or ice-cold methanol, and then blocked with goat serum (Invitrogen, 16210064) for 1 h at room temperature. The cells were incubated with primary antibodies diluted in goat serum overnight at 4°C. After extensive washing with PBS, the cells were incubated with fluorescently-labeled secondary antibodies (Jackson ImmunoResearch Laboratories, 115–545–003 and 111–585–003) at room temperature for 1 h. After 3 washes with PBS, fluorescence was examined under a confocal microscope (LSM 710 Meta confocal microscope, Carl Zeiss, Jena, Germany) with 63 × 1.4NA plan-apochromat oil immersion lens. More than 100 cells were counted in each experiment.

Assessment of cytotoxicity by flow cytometry

After treatment, SH-SY5Y cells were collected, resuspended in binding buffer, and 100 μl of cells were stained with ANXA5-PER and 7-aminoactinomycin D (7-AAD) (BD Biosciences, 559763) according to the manufacturer's instructions. The stained cells were then subjected to flow cytometry analysis (BD Biosciences, FACSCalibur, USA). EGFP-positive cells were gated and analyzed for late apoptotic and necrotic cells (both ANXA5-PER- and 7-AAD-positive). A minimum of 30,000 events were collected and analyzed.

Proteasome activity assay

Rat hippocampal tissues were homogenized in lysis buffer (50 mM HEPES, pH 7.5, 5 mM EDTA, 150 mM NaCl, 1% NP40, 1 mM DTT, 2 mM ATP [New England Biolabs, P0756S]). Lysate (250 µl) containing 3 μg of protein were incubated with 40 nM aminomethylcoumarin (AMC)-linked synthetic peptide substrates for 30 min at 37°C. Ac-Gly-Pro-Leu-Asp-AMC, Suc-Leu-Leu-Val-Tyr-AMC, and Ac-Arg-Leu-Arg-AMC (and Boc-Leu-Arg-Arg-AMC) were used to detect caspase-like, chymotrypsin-like, and trypsin-like activity, respectively (Proteasome Substrate Pack, Enzo Life Sciences, PW9905–0001). The reaction was stopped by adding 250 μl of a cold 96% ethanol solution. Proteasome activity was measured by detecting the fluorescence resulting from AMC hydrolysis (380 nm excitation and 460 nm emission).

Immunohistochemistry

After transcardial perfusion with 4% formaldehyde, the brains of rats were post-fixed for 48 h at 4°C, then dehydrated, embedded in paraffin, and cut in 5-μm coronal sections. After deparaffinization and rehydration, antigen retrieval was performed by heating in 10 mM boiling sodium citrate buffer (pH 6.0) for 5 min twice. Nonspecific binding was blocked with normal goat serum, then the sections were incubated with primary antibodies at 4°C overnight. After washing with PBS 3 times, the sections were incubated with fluorescent-labeled secondary antibodies at room temperature for 1 h. After washing, sections were counterstained with Hoechst 33342, mounted and examined under a confocal microscope (LSM 710 Meta confocal microscope, Carl Zeiss, Jena, Germany) with 63 × 1.4NA plan-apochromat oil immersion lens.

Animals

Male Wistar rats weighing 450 to 600 g and male GK rats weighing 370 to 400 g were used at 30 wks age. Male ob/ob mice and lean controls were used at 24 wks age. Male C57BL/6J mice with HFD- or HFD/STZ-diabetes were used at 10 or 62 wk age. Rats and mice were kept in a 12-h light/dark cycle at a controlled room temperature and had free access to standard chow and tap water. Rats and mice were anesthetized with Nembutal, and the hippocampal tissues were immediately harvested and resolved for in vitro analyses. The Institute of Biophysics Administrative Panel on Laboratory Animal Care approved all procedures involving animals.

Statistics

The results are expressed as the mean ± SEM from at least 3 independent experiments. Differences between 2 groups were analyzed by the Student t test, and multiple comparisons were evaluated by one-way ANOVA, or 2-way ANOVA followed by the Bonferroni post hoc test. The statistical analysis was performed by Prism software (GraphPad). For all analysis, P < 0.05 were considered statistically significant. ***, P < 0.001; **, P < 0.01; *, P < 0.05; NS, nonsignificant.

Abbreviations

ANXA5-PER

ANXA5/annexin V-phycoerythrin (PER)

ATG4B

autophagy-related 4B

Atg8-family proteins

autophagy-related 8-family proteins

BG

basal glucose

CASP3

caspase 3, apoptosis-related cysteine peptidase

CNS

central nervous system

CQ

chloroquine

EBSS

Earle's balanced salt solution

GABARAP

gamma-aminobutyric acid receptor associated protein

GK

Goto-Kakizaki

GSNO

S-nitrosoglutathione

HFD

high-fat diet

HG

high glucose

IBP

irreversible biotinylation procedure

LC-MS/MS

liquid chromatography-tandem mass spectrometry

L-NAME

L-nitro-arginine methylester

L-NNA

Nω-nitro-L-arginine

MAP1LC3B/LC3B

microtubule-associated protein 1 light chain 3 β

NO

nitric oxide

NOS

nitric oxide synthase

PE

phosphatidylethanolamine

PI

propidium iodide

Rap

rapamycin

SNOC

S-nitrosocysteine

SQSTM1

sequestosome 1

STZ

streptozotocin

TEM

transmission electron microscopy

WM

wortmannin

WT

wild type

7-AAD

7-aminoactinomycin D

Disclosure of potential conflicts of interest

No potential conflicts of interest needed to be disclosed.

Acknowledgments

We thank Prof. Hong Zhang from the Institute of Biophysics for providing the plasmids for prokaryotic expression of C. elegans ATG-3 and ATG-7 and valuable discussions. We thank Prof. Yongsheng Chang at the Institute of Basic Medical Science, Chinese Academy of Medical Science and Peking Union Medical College for valuable discussion. We thank Prof. Yong Liu at Wuhan university for providing the HFD mice. We thank Jifeng Wang and Peng Xue from the Institute of Biophysics Core Facility for their assistance with the Q Exactive mass spectrometer and LTQ Orbitrap mass spectrometer analyses.

Funding

This research was supported by the National Basic Research Program of China (973 Program) (2012CB911000, 2011CB503900, 2011CB910900, 2013CB530602), the National Natural Sciences Foundation of China (31225012, 31030023, 31500693, 31300696, 31127901, 31421002, 31571163), Personalized Medicines-Molecular Signature-based Drug Discovery and Development, the Strategic Priority Research Program of the Chinese Academy of Sciences (XDA12020316, NNCAS-2012–2), National Key Research and Development Program (2016YFC0903100 and 2017YFA0504000), and Youth Innovation Promotion Association CAS to L. W.