Helicobacter pylori VacA induces autophagic cell death in gastric epithelial cells via the endoplasmic reticulum stress pathway

The Helicobacter pylori vacuolating cytotoxin (VacA) can promote progressive vacuolation and gastric injury and may be associated with human gastric cancer. Increasing evidence indicates that autophagy is involved in the cell death induced by VacA, but the specific mechanisms need to be further elucidated. We show here that VacA could induce autophagy and increase cell death in human gastric cancer cell lines. Further investigations revealed that inhibition of autophagy could decrease the VacA-induced cell death in AGS cells. Furthermore, numerous dilated endoplasmic reticula (ER) were observed, and the phosphorylation of a subunit of eukaryotic translation initiation factor 2 subunit 1 also increased in the VacA-treated AGS cells, while repression of ER stress could reduce autophagy and cell death through knockdown of activating transcription factor 4 and DNA-damage-inducible transcript 3. In addition, the expression of pseudokinase tribbles homolog 3 (TRIB3) upon ER stress was triggered by VacA, and knockdown of TRIB3 could also decrease VacA-induced cell death. Finally, inhibition of autophagy could decrease VacAs1m1-induced cell death and apoptosis, and apoptosis inhibitor Z-VAD had no significant effect on autophagy induced by VacAs1m1. Thus, these results suggested that VacA causes autophagic cell death via ER stress in gastric epithelial cells.


Introduction
Helicobacter pylori (H. pylori) is a Gram-negative bacterium colonizing the human stomach, and the prevalence of its infection in adults is 50-90% in developing countries and between 30 and 50% in developed countries 1 . H. pylori-persistent infection can contribute to gastritis and peptic ulcer disease and is a definite pathogenic factor for gastric adenocarcinoma [2][3][4] . The interactions among the bacterium, gastric epithelium, and host innate defense responses are critical for the consequences of H. pylori infection. Vacuolating cytotoxin (VacA), a critical virulence factor of H. pylori, is considered a multifunctional toxin, responsible for eliciting several different effects on host cells, including vacuolization, necrosis, and apoptosis [5][6][7] . Numerous studies have shown that VacA can induce apoptosis through the mitochondrial pathway in gastric epithelial cells [7][8][9] . Alternatively, most VacA was localized to vacuoles rather than mitochondria using immunostaining and confocal microscopy with labeled Rab7-GFP 10 , indicating that VacA may indirectly lead to cytochrome c release from mitochondria, which suggests that VacA may involve other pathways leading to cell death.
The endoplasmic reticulum (ER) is a complex, multifunctional organelle that has a critical role in cellular biological effects by synthesizing proteins and monitoring protein folding and trafficking 11,12 . If the ER cannot resolve cell stress, it will cause unfolded or misfolded proteins to accumulate in the ER lumen, leading to ER stress, which is involved in signaling pathways, including inflammation and cell death 13 . To guard against or respond to ER stress, cells develop an integrated signaling mechanism to restore homeostasis and normal ER function 14 . ER stress activates a series of downstream transcriptional effectors, such as nuclear protein 1 (NUPR1), eukaryotic translation initiation factor 2 subunit 1 (EIF2S1), activating transcription factor 4 (ATF4), DNAdamage-inducible transcript 3 (DDIT3), and tribbles pseudokinase 3 (TRIB3), to regulate protein folding and protein quality control 15 . The coordination activity of the entire process determines the extent of endoplasmic reticulum stress and thus governs whether cells will reestablish an intracellular biological balance or activate cell death programs.
Macroautophagy (hereafter autophagy) is an intracellular quality-control and quantity-control process in which intracellular components are sequestered into double-membrane organelles and are delivered to lysosomes for degradation 16 . In addition to the protective role of cell homeostasis, including nutrient starvation and hypoxia stress, prolonged autophagy or overstimulated autophagy could contribute to autophagic cell death 17,18 . Recently, we showed that Shiga toxins purified from Escherichia coli O157:H7 result in autophagic cell death in Caco-2 cells through the ER stress signaling pathway 17 . In addition, gene products from other bacteria have been reported to participate in autophagic cell death 19,20 . The "enhanced intracellular survival" (eis) gene product of Mycobacterium tuberculosis can regulate inflammation and lead to autophagic cell death through redox-dependent signaling in macrophages 21 . Although some studies have reported that VacA of H. pylori can induce autophagy, the mechanism by which VacA induces cell death remains to be elucidated.
In this study, the relationships among VacA, ER stress, autophagy, and cell death were investigated in AGS cells. We provide evidence showing that H. pylori VacA induces autophagic cell death in gastric epithelial cells through the ER stress pathway.

VacA induces cell death in human gastric cancer cells
Previous studies have indicated that VacA rapidly induces apoptosis and programmed cell necrosis of gastric cancer cells 6,22 . To determine whether VacA was associated with cell death, we employed an ANXA5/propidium iodide (PI) staining assay to detect AGS cells infected with cagA − /vacA s1m1 and cagA − /vacA s1m2 H. pylori. The cagA − / vacA s1m1 H. pylori infection markedly increased cell death compared with cagA − /vacA s1m2 H. pylori (Figs. 1a, b). To further investigate the level of cell death induced by VacA, we performed an MTT assay. Similar results were also obtained in AGS cells infected with cagA − /vacA s1m1 and cagA − /vacA s1m2 H. pylori (Fig. 1c). These data indicate that VacA has a critical role in H. pylori-induced cell death in human gastric cancer cells.

Cell death induced by VacA is dependent on autophagy
To further investigate the role of VacA in the H. pyloriinduced human gastric epithelial cell death, AGS cells were treated with VacA protein, which was purified from culture supernatants obtained from the cagA − /vacA s1m1 and cagA − /vacA s1m2 H. pylori clinical isolates using an affinity chromatography scheme. VacA s1m1 toxin could induce cell death with PI staining and MTT assay in a time-dependent manner, and VacA s1m2 toxin did not (Figs. 2a,b). Some studies reported that VacA can induce autophagy in human gastric cancer cells [23][24][25] . However, whether the activating autophagy promotes or inhibits cell death is unknown. To explore this problem, after pretreatment with a pharmacological inhibitor of autophagy (3-methyladenine; 3-MA) or an apoptosis inhibitor (Z-VAD), AGS cells were treated with VacA s1m1/ s1m2 toxin, and the level of cell death was subsequently detected by PI staining and MTT assay. 3-MA or Z-VAD could significantly reduce cell death induced by VacA s1m1 toxin in the AGS cells (Figs. 2c, d). These results suggest that autophagy may have been involved in the VacA-induced cell death.

VacA toxin induces autophagy in AGS cells
To elucidate whether VacA could induce autophagy in AGS cells, we applied flow cytometry to detect the formation of autolysosomes after staining with acridine orange (AO) or monodansylcadaverine (MDC). Intriguingly, VacA s1m1 toxin induced the percentage of autolysosome-accumulated cells up to more than 70% of AGS cells with the two staining assays (Figs. 3a, b). Compared with VacA s1m2 and control, VacA s1m1 significantly increased the number of autolysosomes in AGS cells (Figs. 3a, b). The VacA s1m1 toxin induced the ratio of microtubule-associated protein 1 light chain 3 beta-II (MAP1LC3B-II) to ACTA (Fig. 3c) and led to MAP1LC3B-II accumulation in time-dependent and dose-dependent manners (Figs. 3d, e). Pharmacological inhibitor of autophagic flux bafilomycin A1 (Baf A1) further promoted the accumulation of MAP1LC3B-II in the AGS cells (Fig. 3f). In addition, GFP-LC3 puncta formation assays and transmission electron microscopy (TEM) assays also showed that VacA s1m1 could increase the number of autophagosomes in AGS cells after 6 or 24 h (Figs. 3g, h). These data indicate that VacA could induce autophagy in AGS cells.

ER stress may be involved in VacA-induced cell death
With the exception of autophagosomes in TEM analysis, we also observed numerous cells with dilated ER in VacA s1m1 -treated cells (Fig. 4a). To further determine whether ER stress was involved in VacA s1m1induced cell death, an ER stress inducer thapsigargin was applied to confirm the role of ER stress. Western blot analysis showed that thapsigargin could increase the expression of MAP1LC3B-II (Fig. 4b), and the AO staining assay was also consistent with this result (Fig. 4c). Moreover, thapsigargin could increase VacA s1m1 -induced cell death (Fig. 4d). These data showed that ER stress may be involved in VacA-induced cell death.

VacA induces autophagy through ER stress
To further research the role of ER stress in VacAinduced cell death, the phosphorylation of EIF2S1 (eukaryotic translation initiation factor 2, subunit 1 alpha, 35 kDa) was investigated after VacA s1m1 treatment. VacA s1m1 could increase the phosphorylation of EIF2S1 in a time-dependent manner in AGS cells (Fig. 5a). The VacA-induced autophagy and cell death also had the same trend with EIF2S1 phosphorylation (Figs. 5c, e). In addition, knockdown of the ER stress effector DDIT3 or ATF4, two downstream targets of EIF2S1, could drastically decrease the formation of autolysosomes and cell death in the VacA s1m1 -treated AGS cells (Figs. 5b, d, f). To further elucidate the relationship of ER stress to the effects of VacA, mitochondrial dysfunction and apoptosis, we also detected the mitochondrial dysfunction and apoptosis with knockdown of DDIT3 and ATF4 in the VacA-treated AGS cells. VacA could decrease the mitochondrial ATP level and increase apoptosis in a time-dependent manner in AGS cells ( Supplementary Figs. 1a, c). Moreover, as shown in Supplementary Figs. 1b, d, knockdown of DDIT3 or TRIB3 could also decrease VacA-induced mitochondrial dysfunction and apoptosis. These results suggest that VacA-induced ER stress is antecedent to autophagy generation.
TRIB3, a target of DDIT3-ATF4, plays an important role in ER stress-induced cell death. To evaluate this role, TRIB3 was knocked down by TRIB3 siRNA in AGS cells (Fig. 6a). Knockdown of DDIT3 or TRIB3 could prevent VacA s1m1 -induced autophagy in AGS cells (Figs. 6b-d). Moreover, knockdown of DDIT3 or TRIB3 could also decrease VacA-induced cell death (Figs. 6e, f). Furthermore, we generated TRIB3 shRNA lentivirus to knockdown the expression of TRIB3 in the gastric tissue of mice. After the tissue was treated with VacA s1m1 toxin, cell death and autophagy were detected with hematoxylin and eosin staining or western blot analysis. The damage degree of the VacA s1m1 groups was significantly increased compared to the control, and the damage degree of the shRNA-TRIB3 lentivirus group was lower than the shRNA-NC lentivirus group (Fig. 6g). As shown in Fig. 6h, shRNA-TRIB3 lentivirus could also decrease the ratio of MAP1LC3B-II to ACTA. These data indicated that VacA induces autophagy and cell death in the AGS cells by triggering ER stress, which is involved in the stimulation of EIF2S1 phosphorylation and the upregulation of DDIT3 and TRIB3.

Autophagy is upstream of apoptosis in VacA-induced cell death
VacA s1m1 -induced autophagy can be suppressed by 3-MA (Figs. 7a, b). To investigate the relationship of VacAinduced apoptosis and autophagy, autophagy inhibitor 3-MA was applied to analyze active CASP3/caspase 3 in VacA-treated AGS cells. Active CASP3/caspase 3 was decreased by 3-MA (Fig. 7c). To further assess the role of autophagy in VacA-induced cell death, we examined the silencing effect of ATG12 or BECN1 siRNA in AGS cells (Fig. 7d). Knockdown of ATG12 or BECN1 could decrease VacA s1m1 -induced cell death and PARP1 cleavage (Figs. 7e, f). In addition, similar results were also present in Az-521 cells (Supplementary Figs. 2a-c). These results revealed that autophagy generation preceded the appearance of apoptosis by VacA. In addition, to further verify the relationship of autophagy and apoptosis under VacA stress conditions, AGS cells were pretreated with apoptosis inhibitor Z-VAD for 24 h, and autophagy induced by VacA s1m1 was evaluated with MDC, AO, and western blot assays. Z-VAD had no significant effect on autophagy induced by VacA s1m1 (Figs. 7g-i). These results suggest that autophagy may be upstream of apoptosis in VacA-induced cell death.

Discussion
According to the present results, we found that VacA induces autophagic cell death through endoplasmic reticulum stress in human gastric epithelial cells. This novel mechanism is derived from the following results: (i) VacA s1m1 could significantly increase autophagy and cell death in AGS cells, (ii) the autophagy by VacA s1m1 is dependent on ER stress, which participates in VacA s1m1induced cell death, and (iii) autophagy may be upstream of apoptosis in VacA s1m1 -induced cell death.
VacA exerts pleiotropic actions on gastric cells, including membrane channel formation, vacuolization, immunomodulation, disruption of endosomal/lysosomal function, and apoptosis 26 . Among these cellular activities, cell death induced by VacA plays a critical role in the origination and development of peptic ulcers in H. pylori infection. Until now, apoptosis was believed to be a crucial factor in cell death induced by VacA. VacA induces apoptosis by connexin 43 (Cx43) accumulation through a Rac1/ERK-dependent pathway 27 . Although VacA did not affect Cx43 expression, Cx43 accumulation resulted in the apoptotic cell death of AZ-521 cells. Here, we investigated the role of VacA in gastric epithelial cell death through autophagy and ER stress and its underlying molecular mechanism. We demonstrate that VacA induces both autophagy and apoptosis and leads to cell death. Interestingly, autophagy generation appears to occur earlier than apoptosis in AGS cells. Pharmacological inhibitor of autophagy 3-MA or knockdown of ATG12 or BECN1  (Figs. 7c, e, f). These data indicate that autophagy promotes gastric epithelial cell death in VacA stress conditions and mediates the generation of apoptosis. Although mitochondria are well-known targets of VacA, how VacA affects the mitochondria during apoptosis remains to be elucidated. Recently, one study reported that most VacA is localized to vacuoles, marked by Rab7-GFP, rather than mitochondria using immunostaining and confocal microscopy 10 . Our study reported that VacA could induce ER stress in AGS cells, which is consistent with a previous report in AZ-521 cells 7 . In our study, VacA could increase the phosphorylation of EIF2S1, and knockdown of the ER stress effector DDIT3 or ATF4 could drastically decrease autolysosomes and cell death in the VacA-treated AGS cells. Akazawa et al. also reported that VacA transcriptionally increased C/EBP homologous protein (CHOP) and phosphorylation of EIF2S1 in gastric epithelial cells, and knockdown of CHOP could lead to inhibition of VacA-induced apoptosis 7 . Consistent with the report, our study also indicated that the activation of ER stress mediated by VacA contributes to apoptosis in gastric epithelial cells.
VacA, a critical virulence factor of H. pylori, was sufficient to induce autophagy, but prolonged exposure to VacA could also disrupt autophagy in human gastric epithelial cells and lead to infection and cell death. The different effects of VacA on autophagy are dependent on the duration/level of cell exposure to the toxin, but its timing and related molecular mechanisms deserve further investigation. In the current study, our results demonstrated that VacA could induce autophagy in AGS cells. As shown in Fig. 3, VacA significantly induced the ratio of MAP1LC3B-II and the formation of autolysosomes in AGS cells, and an increase in the number of autophagosomes was observed by TEM analysis in the VacA-treated AGS cells. Recently, Yahiro reported that VacA induced Cx43 accumulation and autophagy, and Cx43 colocalized with autophagosomal marker MAP1LC3B-II 28 . Terebiznik et al. found that H. pylori triggered autophagy with the appearance of autophagosomes, different in size, and morphology from typical VacA-induced vacuoles 24 . These data indicated that there is a close relationship between VacA and autophagy, and its specific regulatory mechanism will continue to be a fascinating and rewarding subject for future studies. In summary, these results provide novel evidence that VacA triggers the ER stress response to activate autophagy and induces cell death of AGS cells. Future research involving mouse models may lead to a better understanding of the role of ER stress in the pathogenesis of VacA-induced gastric mucosal injury, and experiments should be designed to establish the dynamics and specific mode of action of VacA in affecting autophagy.

Antibodies and reagents
The adenovirus of the GFP-LC3B fusion protein (C3007) was obtained from Beyotime Institute of Biotechnology. Some chemical reagents were purchased from Sigma,  Collection (Manassas, VA, USA). The two cell lines were cultured in F12 cell culture medium (Gibco, Grand Island, NY, USA, #11765-054) or Earle's minimal essential medium (Sigma, M0275) containing 10% fetal bovine serum (Gibco, #10099-141) and 100 U/mL penicillin/ streptomycin (Thermo Scientific, #15140148) in a 5% CO 2 incubator at 37°C. The gastric tissue culture was performed with a modification of the method that was described previously 29 . Briefly, the gastric tissues of C57BL/6 mice were sliced into fragments, followed by soaking in ice-cold PBS (Sigma, P5493) containing gentamicin (0.5 mg/mL; Sigma, G9654), with gentle shaking at 4°C for 10 min, and then washed three times. The washed gastric tissues were cultured in DMEM medium (Gibco, #11965) containing 10% fetal bovine serum (Gibco, #10099-141) and 100 U/ mL penicillin/streptomycin (Thermo Scientific, #15140148) in a 5% CO 2 incubator at 37°C. Finally, the tissues were infected with pGCSIL-GFP-shRNA-TRIB3 or pGCSIL-GFP-shRNA-NC (2 × 10 8 TU/mL) for 24 h and then challenged with VacA s1m1 toxin or the control for 24 h. Histological assessment was performed according to the Sydney classification by two pathologists who were blinded to the other experimental results. The study was approved by the ethics review board at Third Military Medical University.

Purification of VacA
VacA was purified from culture supernatants obtained from the cagA − /vacA s1m1 and cagA − /vacA s1m2 H. pylori clinical isolates using an affinity chromatography scheme. The immunoaffinity chromatography column was prepared by VacA subunit-specific monoclonal antibody 5E4 (Santa Cruz, 32746) coupled to CNBr-activated sepharose 4B matrix (GE Healthcare, 17043001). Then, an affinity chromatography column (GE Healthcare, 17040401) was used to purify the samples containing VacA. The relative molecular weight of VacA was 88 kDa, confirmed by SDS-PAGE.
The pGCSIL-GFP vector was digested by AgeI and EcoRI and purified by Qiagen (Shanghai, China), a quick gel extraction kit. The pairs of complementary hairpin DNA oligos that were mentioned above were synthesized, annealed, and ligated into linearized pGCSIL-GFP vector. The ligated DNA solution was transformed into E. coli DH5α and incubated on a Luria Bertani plate (50 ng/ml ampicillin) at 37°C for 16 h. Positive clones were identified by DNA sequence analysis, and the resulting plasmids were called pGCSIL-GFP-shRNA-TRIB3 and pGCSIL-GFP-shRNA-NC. Lentiviruses were generated in 293T cells by co-transfection of pGCSIL-GFP-shRNA-TRIB3 or pGCSIL-GFP-shRNA-NC with pHelper1.0 and pHelper2.0 plasmids. These plasmids were transfected into 70% confluent 293T cells using Lipofectamine 2000  Detection of MDC and AO staining of cells treated with 50 mM Z-VAD or 50 ng/ml VacA using flow cytometry analysis. (i) The MAP1LC3B-II conversion was detected using western blot analysis. AGS cells were treated as above 1% OsO 4 for 1 h and a half, and then stained in 3% aqueous uranyl acetate for 1 h. These cells were dehydrated with graded alcohol (50%, 70%, 80%, 90%, 95%, 100%) and embedded in Epon-Araldite resin (Canemco, #034). Ultrathin sections were prepared on a Leica EM UC7 ultramicrotome, counterstained with 0.3% lead citrate, and observed under a Zeiss EM 902A electron microscope. The autophagosome counting method was followed as described previously by Yla-Anttila et al 31 .

ATP measurements
After treatment with 100 nM siRNA or 50 ng/ml VacA s1m1 toxin, cells were collected, centrifuged, and washed with phosphate-buffered saline (PBS) three times. Next, 200 μL lysis buffer from an ATP Assay kit (Beyotime, S0026) was added to each tube and then ultrasonicated. The lysates were centrifuged at 12,000 r.p.m. for 5 min at 4°C. The supernatant was transferred to a new 1.5-mL tube for the ATP assay with the ATP detection kit. The protein level of the supernatant was measured at 562 nm with Bicinchoninic Acid assay (Beyotime). The relative ATP level was calculated according to the following formula: relative ATP level = ATP value/protein value.

GFP-LC3 puncta formation assays
AGS cells were infected with GFP-LC3B adenovirus (MOI = 100:1) for 24 h, then cultured with 50 ng/ml VacA s1m1 toxin for 6 h, and then fixed in 4% paraformaldehyde for 10 min at 37°C. Confocal microscopy was performed with a Radiance 2000 laser scanning confocal microscope (Bio-Rad, San Francisco, CA), followed by image analysis with LaserSharp 2000 software (Bio-Rad). Images were acquired in a sequential scanning mode. According to methods for monitoring GFP-LC3 puncta formation assays, the average number of MAP1LC3B puncta per cell in GFP-MAP1LC3B-positive cells was determined.

Western blot analysis
Western blotting was performed to determine the protein level of ATG12, MAP1LC3B, p-EIF2S1, ATF4, TRIB3, PARP1, DDIT3, BECN1, and ACTA in AGS cells as described previously 17 . AGS cells were washed and then lysed by the M-PER Mammalian Protein Extraction Reagent (Pierce, 78501, Thermo Scientific, Waltham, MA, USA). After centrifugation, the total protein concentration was determined with a BCA protein assay kit (Pierce, 23227, Thermo Scientific). Cell lysis solutions were separated using SDS-PAGE and transferred to polyvinylidene difluoride (PVDF) membranes. Primary antibodies were diluted 1:1000. Membranes were developed using Supersignal® West Dura Duration substrate reagent (Thermo Scientific, 34080). Densitometric analysis on the western blot was performed by Image Gauge software (Fujifilm, MD, USA). ACTA was used as an internal control, and the ratio of the intensity of the protein of interest to ACTA was calculated.

Cell viability
According to the manufacturer's instructions (Sigma), AGS cell viability was determined with an MTT assay. Following treatment, AGS cells were incubated with MTT at a final concentration of 5 mg/L for 2 h and then dissolved in the MTT solubilization solution. The cell survival rate was measured with an absorbance at 590 nm (A590) by a microplate reader (Bio-Rad).

MDC staining and AO staining assays
Following treatment with H. pylori or transfection with plasmids/siRNAs, AGS cells were stained with 10 mM MDC at 37°C for 10 min and fixed in 3% paraformaldehyde for 30 min. A Radiance 2000 laser scanning confocal microscope was used to take photographs. Other cells with the same treatment were quantified by flow cytometry (BD FACScan Flow cytometer).
In AO staining assays, AGS cells were stained with 1 mg/mL AO solution for 15 min at 37°C. After incubation, cells were washed and immediately analyzed with a Radiance 2000 laser scanning confocal microscope. Then, other cells with the same treatment were quantified by flow cytometry.
Cell death assay AGS cell death was assessed using a PI staining assay as previously described 17 . The cells were trypsinized, collected, and resuspended in 2 ml of PBS, and then incubated with the PI staining solution at 37°C for 30 min in the dark before being finally measured with flow cytometry.

Apoptosis
The ratio of apoptotic cells was evaluated by staining 5 × 10 5 cells with an ANXA5/annexin V-FITC/PI Detection Kit (Invitrogen, V13242) according to the manufacturer's protocol 17 . The samples were analyzed by flow cytometry (BD FACScan Flow cytometer, USA) to determine the percentage of cells displaying annexin V + / PI − (early apoptosis) or annexin V + /PI + staining (late apoptosis). For each sample, we report the percentage values corresponding to annexin V-FITC-positive cells. Three independent experiments were performed for each assay condition.

Statistical analysis
The results are expressed as the mean ± SEM. Two group data were analyzed by Student's t-test, and multiple group data were analyzed using one-way ANOVA. Statistically significant differences are indicated by asterisks (*P < 0.05, **P < 0.01).