Dynamic involvement of ATG5 in cellular stress responses

Autophagy maintains cell and tissue homeostasis through catabolic degradation. To better delineate the in vivo function for autophagy in adaptive responses to tissue injury, we examined the impact of compromised autophagy in mouse submandibular glands (SMGs) subjected to main excretory duct ligation. Blocking outflow from exocrine glands causes glandular atrophy by increased ductal pressure. Atg5f/−;Aqp5-Cre mice with salivary acinar-specific knockout (KO) of autophagy essential gene Atg5 were generated. While duct ligation induced autophagy and the expression of inflammatory mediators, SMGs in Atg5f/−;Aqp5-Cre mice, before ligation, already expressed higher levels of proinflammatory cytokine and Cdkn1a/p21 messages. Extended ligation period resulted in the caspase-3 activation and acinar cell death, which was delayed by Atg5 knockout. Moreover, expression of a set of senescence-associated secretory phenotype (SASP) factors was elevated in the post-ligated glands. Dysregulation of cell-cycle inhibitor CDKN1A/p21 and activation of senescence-associated β-galactosidase were detected in the stressed SMG duct cells. These senescence markers peaked at day 3 after ligation and partially resolved by day 7 in post-ligated SMGs of wild-type (WT) mice, but not in KO mice. The role of autophagy-related 5 (ATG5)-dependent autophagy in regulating the tempo, duration and magnitude of cellular stress responses in vivo was corroborated by in vitro studies using MEFs lacking ATG5 or autophagy-related 7 (ATG7) and autophagy inhibitors. Collectively, our results highlight the role of ATG5 in the dynamic regulation of ligation-induced cellular senescence and apoptosis, and suggest the involvement of autophagy resolution in salivary repair.

Autophagy is a catabolic process that has an essential role in cellular adaptation to multiple types of stress by recycling of superfluous cellular material, safeguarding quality control in organelles, removing protein aggregates, and eliminating intracellular pathogens. 1 Conceptually, autophagy serves a pro-survival mechanism by providing sources of energy and biosynthetic building blocks during starvation, removing dysfunctional organelles and large aggregates toxic to cells to avoid unwarranted cell death. However, upon sustained stress conditions, cell death eventually takes place either by excessive autophagy or by the induction of apoptosis and/or necrosis pathways. 2 The ATG5, autophagy-related 5, has a pivotal role in autophagosome formation. Mouse neonates systemic deficient for ATG5 die within a day of birth, 3 whereas mice depleted of Atg5 in selected tissues have abnormalities ranging from neurodegeneration 4 and age-related cardiomyopathy 5 to liver tumors. 6 Autophagy and senescence are two distinct, however functionally intertwined, cellular responses to stress. 7 Cellular senescence is a state of stable growth arrest that is induced by telomere shortening, DNA-damage, oncogenes or other stresses. In general, senescence is a heterogeneous phenotype, which is characterized by a senescent-associated secretory phenotype (SASP), expression of senescenceassociated β-galactosidase (SA-β-gal) and other senescent markers, and increased cell size. 8 In culture system, inhibiting or enhancing autophagy leads to the opposite effect on premature senescence. [9][10][11][12] While premature senescence can be induced by a plethora of cell-extrinsic and cell-intrinsic stressors, 13 little is known about the possible role of autophagy in modulating injury-induced cellular senescence in vivo. Rodent salivary duct ligation has been used as an experimental model system to study salivary gland atrophy, which often occurs in patients with Sjögren's syndrome or receiving head and neck radiation therapy. Although autophagy induction has been implicated in the repair of rapamycin-treated, post-ligated salivary glands, 14,15 the roles played by autophagy in regulating the injury responses in submandibular glands (SMGs) have not been explored.
To explore how autophagy contributes to salivary (patho) physiology, we established a transgenic mouse model deficient for ATG5 in the salivary acinar cells. Previously, we have identified a role for basal autophagy in salivary homeostatic mechanisms that restrict acinar cell size and the number of secretory granules. 16 Here, we report that ligation of the major SMG excretory duct triggers the glandular atrophy and the induction of autophagy. By comparing the acute and subacute stress responses from autophagy-impaired and -competent SMGs with duct obstruction, we established the intrinsic roles of ATG5-dependent autophagy in modulating salivary inflammatory responses, stress-induced senescence and cell death, which all occur sequentially in response to tissue injury. Our results provide in vivo evidence that stress-induced autophagic response is indispensable for resolving premature senescence in duct cells of the ligated glands, whereas ATG5 deficiency leads to delayed acinar cell death.

Results
Acinar-specific autophagy deficiency. To assess the contribution of autophagy to tissue injury, we impaired Atg5 expression in salivary acinar cells by crossing mice expressing aquaporin 5 (Aqp5)-driven Cre recombinase 17 with mice containing loxP sites that flank the third exon of Atg5 4 ( Figure 1a) and subjected these mice to ductal ligation. Expression of the Aqp5-Cre transgene mirrored endogenous Aqp5 in the salivary glands and was acinar cell specific. 17 Immunohistochemical (IHC) analyses revealed that ATG5 expression was not only abolished in AQP5-expressing acinar cells, but also decreased substantially in AQP5 non-expressors, mainly granular convoluted ducts (GCDs) and other duct cells in the SMGs of Atg5 f/ − ;Aqp5-Cre mice (Figure 1b). This is because offspring from Atg5 +/ − ;Aqp5-Cre and Atg5 f/f crossbreeds exhibited an ATG5 hypomorphic phenotype in SMGs (Figure 1c, +/+ versus F/F) introduced by the flox allele. 18 As expected, higher ATG5-ATG12 signals were coexisted with higher abundance of lipidated microtubuleassociated protein 1 light chain 3 (MAP1LC3-II) (and lower MAP1LC3-I) in the SMGs of Atg5 +/+ mice compared with those of Atg5 F/F , Atg5 F/ − or Atg5 F/ − ;Aqp5-Cre mice (Figure 1c). In the ensuing studies, Atg5 f/+ ;Aqp5-Cre mice (designated as Atg5 WT ) and acinar cell-specific Atg5-knockout mice, Atg5 f/ − ;Aqp5-Cre (designated as Atg5 KO ) were used to simultaneously explore the effects of ATG5 deficiency in acinar cells and reduced ATG5 level in duct cells of SMGs. To document the effect of compromised autophagy on basal gene expression in SMGs, quantitative RT-PCR analyses on message abundance from selected injury response and other genes were performed. Notably, the expression of the proinflammatory mediators such as Il6, Il1b, Tnf and Ifng was significantly elevated in SMGs of KO mice (Figure 1d). SMG duct ligation induces autophagy. Next, we investigated whether tissue injury by ligating SMG main excretory ducts from Atg5 WT and Atg5 KO mice for 1, 3 and 7 days induced autophagy. As expected, the MAP1LC3-interacting sequestosome 1/p62 (SQSTM1), a cargo-recognition protein for autophagy was readily detectable in control SMGs of Atg5 KO , but not Atg5 WT mice (Figure 2a, lane 5 versus lane 1). During ligation, MAP1LC3-II levels increased in Atg5 WT SMGs at day 1 and remained high throughout the time course (Figure 2a,Supplementary Figure S1A). Additionally, steadystate levels of SQSTM1 were undetectable in control and duct-ligated Atg5 WT SMGs at day 1, but increased at day 3 ( Figure 2a,Supplementary Figure S1B). Notably, SQSTM1 accumulated extensively in the ligated glands from Atg5 KO mice (Figure 2b,Supplementary Figure S1C). On the basis of comparable Sqstm1 mRNA levels between these two genotypes, both basal and post-ligation (Figure 1d, Supplementary Figure S2), and sustained accumulation of SQSTM1 protein in ligated SMGs of Atg5 KO mice, we conclude that duct ligation induced autophagy in SMGs.
As depleting ATG5 affected basal abundance of proinflammatory response messages (Figure 1d), the expression of selected inflammatory mediators, including proinflammatory cytokines, Il6, Il1a, Il1b, Tnf and Ifng, and prostaglandinendoperoxide synthase 2/cyclooxygenase 2 (Ptgs2/Cox-2) in the post-ligated SMGs of Atg5 WT and Atg5 KO mice was assessed. We found that duct ligation caused robust increases in these mRNA abundances as early as day 1 post ligation in both genotypes, and the inactivation of ATG5 only affected relative Bcl2 modifying factor (Bmf) message induction in the ligated SMGs of Atg5 KO mice compared with Atg5 WT mice after 1 day of ligation (Supplementary Figure S2). There was an over 300-fold increase in Il6 mRNA abundance in ductligated SMGs of Atg5 KO mice at day 1, followed by a decline. Similar patterns of Il1a, Il1b and Ptgs2/Cox-2 mRNAs surges of different magnitudes were observed in ligated SMGs of Atg5 KO mice. To validate induction of proinflammatory cytokines, we measured TNF-α protein level by enzyme-linked immunosorbent assay (ELISA). TNF-α level peaked in L3 SMG from both genotypes at approximately 1.5 ng/mg total protein.
Autophagy alters duct ligation-induced morphological manifestations. Gross examination revealed that SMG size of Atg5 KO mice increased initially at day 1, which could be due to a combination of obstructed saliva outflow and an inflammatory response then leveled off at day 3 and atrophied at day 7 after duct ligation (Supplementary Figure  S3). In comparison, significant glandular weight loss was detected in ligated SMGs of Atg5 WT mice at day 3 post ligation (Supplementary Figure S3I). On average, we observed 19% decrease in gland weight compared with the contralateral ligation control SMG in the same mouse, and the degree of atrophy in terms of weight loss was indistinguishable between Atg5 WT and Atg5 KO mice at day 7 post ligation (Supplementary Figure S3I). However, gland weights of Atg5 KO mice were slightly higher than that of Atg5 WT mice at day 3 post ligation. Microscopically, GCD cells lost many secretory granules and exhibited a dilated phenotype after ligation (Figure 2c). Extensive dilation of ductal system, presumably overloaded with saliva as a result of the outflow obstruction, was particular pronounced in L1 and L3 SMGs of Atg5 WT mice and to a lesser extent in Atg5 KO mice (Figure 2c). The level of duct dilation reduced slowly toward baseline from day 3 to day 7 post ligation. In parallel, secretory acinar cells became enlarged and were staining pale from lacking eosinophilic cytoplasmic staining at day 1 in both genotypes (Figure 2c, L1 inset). By day 3, enlarged acinar cells remained in SMGs of Atg5 KO mice (Figure 2c, L3, arrowheads); however, they were relatively sparse in the L3 SMGs of Atg5 WT mice (Supplementary Figure S4A). By day 7, the atrophied glands from both Atg5 WT and Atg5 KO mice had residual GCD structures and scattered acinar cells (Figure 2c To confirm the loss of acinar cells in duct-ligated L3 SMGs of Atg5 WT mice, we examined red and green fluorescence patterns in duct-ligated SMGs of mT/mG;Aqp5-Cre reporter mice ( Figure 3a). As expected, acinar cells had a distinct pattern of green mG fluorescence due to excision of loxP-flanked red mT through acinar-specific Cre recombinase, whereas GCDs and other non-acinar cells were mostly marked with red fluorescence in the control gland (Figure 3a   (c) Correlation of decreased ATG5 expression and impaired MAP1LC3 lipidation in SMGs among different genotypes. Equal amounts of whole SMG lysates from two individual mouse of the indicated genotype were analyzed for ATG5 and MAP1LC3 levels by western blots. ATG5 expression was greatly reduced in the SMGs of Atg5 F/F than that in Atg5 +/+ mice. This hypomorphic phenotype in floxed mouse line has been reported previously in the loxP mouse gene targeting system. 18 +/+, Atg5 +/+ ; +/ − , Atg5 +/ − ; +/ − ;C, Atg5 +/ − ;Aqp5-Cre; F/F, Atg5 F/F ; F/+, Atg5 F/+ ; F/+;C, Atg5 F/+ ; Aqp5-Cre (or Atg5 WT ); F/ − , Atg5 F/ − ; F/ − ;C, Atg5 F/ − ;Aqp5-Cre (or Atg5 KO ). Relative ATG5 levels were determined by setting average level of ATG5 in Atg5 +/+ mice as 1. ATG5 KO delays ligation-induced apoptosis in SMG acinar cells. We next evaluated the functional role of enhanced autophagy in ligation-induced cell death. In situ apoptosis assays revealed that the percentage of ApopTagpositive cells in duct-ligated SMGs of Atg5 WT mice markedly increased at day 1 to day 3, and then decreased at day 7 ( Figure 4a). Conversely, the percentage of ApopTag-positive cells peaked at day 3 and day 7 after ligation in Atg5 KO mice ( Figure 4a). Notably, most ApopTag-positive cells were localized outside the ductal system ( Figure 4a). In agreement with in situ apoptosis data, caspase-3 activation was comparable between L3 and L7 SMGs, whereas cleaved caspase-3 levels markedly increased between L1 and L3 SMGs (Figure 4b,Supplementary Figure S5A). In addition, the message abundances of both death executor caspase-3 (Casp3) and apoptotic mediators Bcl2l11/Bim and Bmf were significantly upregulated following duct ligation and fold induction of Bmf, compared with control, were higher in Atg5 WT than in Atg5 KO mice after 1 day of ligation (Supplementary Figure S2). A decrease in the abundance of acinar cell proline rich, lacrimal 1/mucin-10 (Prol1/Muc-10)  Figure S5B), was detected in SMGs from Atg5 WT at day 3 and day 7 after duct ligation. In contrast, the abundance of ductal marker kallikrein 1 (Klk1) message was comparable between the two genotypes, except that it significantly increased by approximately 3-fold over the control level in L7 SMGs of Atg5 WT mice only ( Figure 4c). Notably, the KLK1 protein abundance was higher in SMGs from Atg5 WT than that of Atg5 KO at day 1 and day 3 after duct ligation (Figure 4b,Supplementary Figure S5C). Since duct  19,20 the increased KLK1 messages in L7 SMGs of Atg5 WT mice may reflect the initiation of regeneration. 21 To validate the participation of ATG5 in cellular injury, H 2 O 2induced cell death was examined in Atg5 Tet-Off MEF m5-7 cells 22 as an in vitro model. ATG5 levels in m5-7 cells were regulated by concentration of doxycycline (Dox; Figure 4d) in culture medium. The decreased ATG5 level clearly compromised cellular autophagic capacity, evident by decreased MAP1LC3-II and increased p62 levels. Notably, an inverse ATG5-dose-dependent correlation was observed between ATG5 protein levels and cell viabilities upon H 2 O 2 treatment (Figure 4d). In addition to Atg5-KO cell model, we also tested sensitivity to H 2 O 2 -induced cell death affected by autophagy inhibition through Atg7-KO and autophagy inhibitors, chloroquine and bafilomycin A1 (BafA1), respectively, in Atg7-KO MEF and salivary Pa-4 cells. Both autophagy-compromised MEFs and salivary Pa-4 cells demonstrated reduced susceptibility to H 2 O 2 -induced cell death (Figure 4d). Taken together, we surmised that autophagy deficiency renders delayed cell death and/or greater apoptotic threshold upon stress.
Decreased ATG5 primes SMG duct cells for duct ligationinduced senescent phenotypes. Many upregulated proinflammatory mediator messages in ligated SMGs (Supplementary Figure S2) are also genes encoding SASP factors that are secreted by various cells undergoing premature senescence. 23 Therefore, we asked whether duct ligation induced cellular senescence. SA-β-gal activity, which is strongly associated with senescent cells, 24 was examined in control and ligated SMGs. While SMG SA-β-gal activity was detected sparingly at day 1 following duct ligation, SMGs from both Atg5 WT and Atg5 KO mice showed robust SA-β-gal staining after 3 days of ligation (Figure 5a, L3). Notably, SA-βgal activity was mainly detected in duct cells of GCDs of both genotypes. In addition, SA-β-gal activity was detected in limited duct cells of control SMGs from Atg5 KO mice, but not Atg5 WT mice. After 7 days of ligation, the SA-β-gal staining remained intense in the duct cells of post-ligated SMGs of Atg5 KO mice, but markedly decreased in the corresponding counterparts of Atg5 WT mice (Figure 5a, L7). We postulated that the attenuated autophagy resulting from reduced ATG5 abundance in SMG duct cells of Atg5 KO mice (Figure 1b) attributes to the lingering SA-β-gal activity from day 3 to day 7 in KO mice.
We then used a cell-based assay to confirm the role of ATG5 in premature senescence by examining H 2 O 2 -induced senescent phenotype in Atg5 Tet-Off MEF m5-7 and Atg5-KO MEF cells. 22 Treatment with sublethal concentrations of H 2 O 2 resulted in more SA-β-gal-positive Atg5-KO MEF cells than in WT MEF cells (Figure 5c). In addition, steady-state abundance of CDKN1A/p21 protein was substantially increased by sublethal concentrations of H 2 O 2 treatment in Atg5-KO MEF cells (Figure 5d). Time course analysis revealed that p21 level peaked in KO but not in WT cells, 1 day after H 2 O 2 treatment, while p38 was activated immediate (2 h) after the treatment in both cells (Figure 5d). In addition, there is a positive correlation between protein levels of p21 and p62 accumulation (Figure 5d).
Macrophage activation during duct ligation. Compare with Atg5 WT mice, duct ligation induced sustained senescent phenotypes in SMG duct cells (Figures 5a and b) and delayed apoptosis in acinar cells (Figures 3 and 4) of Atg5 KO mice. Macrophages are key components in tissue repair and remodeling during wound healing. To determine whether distinct programs for clearance of damaged cells may have arisen based on cellular autophagy proficiency, we examined recruitment of macrophages following duct ligation. IHC staining for F4/80 revealed that tissue injury induced by duct ligation was accompanied by macrophage recruitment throughout the progression of the tissue injury (Supplementary Figure S7). Notably, F4/80-positive macrophages massively accumulated in injured SMGs at day 3 and day 7 after ligation, however, indistinguishable between WT and KO mice. In addition, expression level of the  Figure S2), was also comparable between the two genotypes. We concluded that levels of ATG5 did not affect macrophage activation in the duct ligation model.

Discussion
It is now generally recognized that autophagic process has a pivotal role in cellular homeostasis and tissue adaptive responses to stress and injury. Herein, we used an injury model of SMG duct obstruction and conditional Atg5 KO to explore the simultaneous interplay among autophagy, inflammation, stress-induced premature senescence and apoptosis in vivo. Ligation of excretory duct of exocrine glands, such as SMG 26 and pancreas, 27 induces acinar cell injury and death as well as ductal cell proliferation when the passage of secretions is blocked. However, the mechanism dictating the fate of individual cell types, in response to stress, within the same tissue is still unclear. Our report represents the first of its kind using conditional ATG5-KO mice to delineate molecular mechanism of duct cell survival and acinar cell death in atrophied SMGs following duct ligation in vivo. One major finding of this study is that combined effect from reduced ATG5 abundance and disruption of SMG duct ligation-induced autophagy causes a phenotype of delayed apoptosis in acinar cells and sustained premature senescence in duct cells of post-ligated SMGs from Atg5 KO mice ( Figure 6). Conceivably, depending on stress intensity, autophagy could regulate cell death pathway with two distinct outcomes: (1) the induction of autophagy in responses to ductal obstruction may be initially beneficial for ligated SMGs by eliminating superfluous proteins accumulated within acinar cells due to blocked secretory outflow. This will allow cells to adapt to the hostile conditions. (2) The sustained autophagy may exacerbate stress condition and induce autophagic cell death (or 'autosis') accompanied by ER dilation and nuclear convolution. 28 Alternatively, autophagosomes could serve as platforms for caspase activation, 29 degrading anti-apoptotic factors, 30 and crosstalking with cell killing mechanisms such as p53 and JNK pathways. 31 Here we showed that acinar cells in post-ligated SMGs underwent programmed cell death after prolonged ligation. Although apoptosis occurred regardless of autophagy status, apoptotic cell death was delayed in ATG5deficient acinar cells, compared with autophagy-competent cells (Figures 3 and 4a). Furthermore, autophagy-impaired cells were more resistant to H 2 O 2 -induce cell lethality (Figure 4d), supporting our notion. ATG5, although an essential protein in autophagosome formation, has additional functions beyond autophagy. For instance, both ATG5dependent mitotic catastrophe 32 and cell death 33 in cells treated with cytotoxic agents have recently been reported. More importantly, calpain-cleaved ATG5 is known to activate caspases to enhance susceptibility toward apoptotic stimuli, thus ATG5 functions as a molecular link between autophagy and apoptosis. 34 Conceivably, the absence of ATG5dependent caspase activation could account for the initially delayed apoptosis in day-1 post-ligated SMGs of the Atg5 KO mice. Thus, the delayed Atg5-KO acinar cell death could be either autophagy dependent or ATG5 dependent or their combination.
Senescence is a state of stable cell-cycle arrest. 35,36 Accumulating evidence from several in vivo senescent mouse models have suggested that the early onset of senescent phenotypes results from the expression, activation or deletion of genes involved in cell-cycle progression. [37][38][39][40][41] Controversy over the relationship between autophagy and senescence exists because both a direct relationship and an inverse relationship were reported. 42 In agreement with inverse relationships, 9,43 the emergence of unstressed SA-β-gal activity in duct cells of SMGs was only detected from young (6-8 weeks old) Atg5 KO , but not from Atg5 WT , mice (Figure 5a, Ctrl). Moreover, stress-induced senescent phenotypes were activated transiently in duct cells of post-ligated SMGs from Atg5 WT mice, however, persisted through day 7 in Atg5 KO mice (Figures 5a and b). We hypothesize that senescent phenotypes were transient in duct-ligated SMGs from Atg5 WT mice because autophagy removes superfluous proteins resulted from blunted secretion, thus maintains cellular homeostasis. When the clearance mechanism is impaired, damaged organelles and other proteins accumulate, leading to unresolved stress phenotypes in the autophagy-impaired SMGs. These observations of increased p21 protein level and intensified SA-β-gal signals in sublethal H 2 O 2 -treated Atg5compromised MEF cells than the corresponding WT MEF cells (Figures 5c and d) supported the notion that lower ATG5 abundance in duct cells of Atg5 KO mice, compared with Atg5 WT mice (Figure 1b), primed these SMG duct cells susceptibility to persistent senescent-associated phenotypes (Figure 5a).
Extended outflow blockage led to extensive accumulation of p62 protein in SMGs of KO mice (Figures 2a and b). The polyubiquitin-binding protein p62, through interaction with LC3, is involved in delivering cargoes to the autophagy machinery or lysosomes for degradation. 44 In autophagydeficient Atg5-KO cells, excessive p62, instead, sequesters ubiquitinated proteins from proteasomal degradation, extending their half-life. 45 Conceivably, p21, a short-lived protein, could therefore prevail in day-7 post-ligated SMGs of Atg5 KO mice through reduced turnover (Figure 5b). However, we cannot rule out the possibility that the stressed acinar cells might modulate the fate of duct cells in a paracrine and autophagy-dependent manner. Even though cellular senescence is generally considered as irreversible, escape scenarios, such as p53 pathway inactivation, exist to allow cells to re-enter the cell cycle. 13,46 Duct ligation clearly leads to a transient senescent phenotype in post-ligated duct cells of Atg5 WT but not Atg5 KO mice. The exact underlying mechanism is still unclear. It is possible that autophagy keeps p62 levels in check, thus promotes resolution of senescent phenotype. Alternatively, autophagy could facilitate the removal of dead cells at the final stage of post-ligation period, minimizing the stress input from local environment.
Collectively, our results explicitly show that ATG5 deficiency primes SMG cells with increased expression of proinflammatory cytokine and Cdkn1a/p21 messages, and the lack of proper duct ligation-induced autophagy dysregulates the tempo and duration of sequential and overlapping elements of tissue injury, including inflammatory responses, stressinduced premature senescence and apoptosis. We therefore conclude that duct ligation-induced autophagy has a dynamic role in preserving the structural integrity of duct and acinar cells in duct-obstructed SMGs. One potentially important mechanism of regulating autophagy in post-ligated SMGs is through the mammalian target of rapamycin (mTOR).  Bozorgi et al. 14 have recently reported while that mTOR is switched off in normal salivary glands, it gets switched on during the course of duct ligation. As mTOR inhibition is instrumental for autophagy induction, 1 the activation of mTORC1 appears to be sufficient to suppress autophagy by preventing the formation of ATG complex. 47 One possible explanation to reconcile our data with theirs is that mTOR activation could represent a feedback mechanism to escape from autosis or autophagy-induced cell death. Nonetheless, these discordances underlie the complexity of in vivo role ATG5 or autophagy partakes in post-ligated SMGs and hint at their respective interactions with other stress-triggered signaling pathways.
Generation of reporter mice. Reporter mT/mG;Aqp5-Cre mice were generated by crossing mT/mG 49 and Aqp5-Cre mice. 17 Cell line, H 2 O 2 treatment and cell viability assay. WT, Atg5-KO and Atg7-KO MEF cells (kind gift from Dr. Chengyu Liang, USC) were cultured in DMEM with 10% FBS. Salivary Pa-4 cells were cultured as we described previously. 50 The Atg5 Tet-Off MEF m5-7 cells 22 were maintained in DMEM containing 10% FBS with doxycycline (Dox, 20 ng/ml) to fully suppress Atg5 expression, as we described previously. 51 To restore Atg5 expression, cells were cultured either in Dox-free medium for full restoration or in medium with 1 ng/ml Dox for partial restoration. These cells were treated with up to 1 mM H 2 O 2 alone or in the presence of chloroquine (20 μM) or BafA1 (10 nM) as indicated, and cell viability was assessed 48 h later based on acid phosphatase assay described by Yang et al. 52 In brief, cells in a 96-well plate were washed with PBS and incubated with 100 μl p-nitrophenyl phosphate (Sigma-Aldrich, St. Louis, MO, USA; N4645) in 5 mM in 0.1 M sodium acetate and 0.1% Triton X-100, pH 5.0, for 1 h at 37°C. The reaction was stopped with the addition of NaOH (10 μl, 1 N) and absorbance measured at 405 nm using a microplate reader. For senescence induction, m5-7 cells were maintained in growth media containing different concentrations of doxycycline (0, 1 and 20 ng/ml). At day 1 after seeding, cells were treated with sublethal dose of H 2 O 2 (75 μM) for 2 h, washed and cultured in growth medium for 5 days. The H 2 O 2 treatment and recovery were repeated once before cell harvesting. WT and Atg5-KO MEF cells were treated with 150 μM H 2 O 2 using the same protocol as for m5-7 cells, and the cells were analyzed for SA-β-gal activity as below.
Fluorescent microscopy. Fluorescent images of frozen sections from SMG of mT/mG;Aqp5-Cre mice were acquired using an Olympus AX70 microscope (Olympus, Tokyo, Japan). All images were compiled using Image-Pro (version 6.3, Media Cybernetics, Rockville, MD, USA). Submandibular duct ligation. Due to gender dimorphism in mouse SMGs, 53 male mice (6-8 weeks old) were used exclusively in all ligation studies reported here. Mice were maintained under 12-h light/dark cycles with unlimited access to food and water. Unilateral ligation of the main excretory ducts of right SMGs, for a period from 1 to 7 days, was performed using surgical sutures, according to Turner et al. 54 The contralateral left SMGs, which served as ligation controls, from the same mice were also harvested for comparison. Animals were euthanized at indicated times following ligation. Both control and ligated SMGs were removed and either snap-frozen in liquid nitrogen for western and real-time RT-PCR analyses or fixed in 4% paraformaldehyde for Hematoxylin and Eosin (H&E) staining and IHC analyses. City of Hope Institutional Animal Care and Use Committee approved all surgical procedures reported here.
IHC analysis. IHC staining was performed on 5-μm thick tissue sections prepared from formalin fixed, paraffin-embedded (FFPE) tissues. FFPE tissue sections were deparaffinized in xylene then hydrated through graded ethanol and distilled water. Samples were then quenched in 1% hydrogen peroxide and pretreated with steam from citrate buffer (Vector Laboratories, Burlingame, CA, USA; H-3300) to promote antigen retrieval. After antigen retrieval, slides were incubated in Protein Block for 1 h, then incubated with primary antibodies ATG5 After three more washes in Dako buffer, slides were incubated with chromogen 3,3′diaminobenzidine tetrahydrochloride (DAB), counterstained with hematoxylin, and mounted.
Real-time reverse-transcription PCR. Total RNA was extracted from SMGs with TRIzol (Life Technologies, Grand Island, NY, USA), according to the manufacturer's instructions. Synthesis of cDNA was performed using the iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA; 170-8891). The cDNA was amplified using IQ SYBR Green Supermix and specific primer pairs in a My IQ Real-Time PCR Detection System (Bio-Rad, 170-8882). Relative mRNA expression levels were calculated using the ΔC t method, as previously described, 50 against Gclc, a message that we found stable throughout the duct ligation process. The primer pairs used are shown in Supplementary Table S1. , and GAPDH (Santa Cruz Biotechnology, sc-25778). Blots were visualized using an enhanced chemiluminescence detection kit (ECL-Plus, GE Healthcare, Pittsburgh, PA, USA; RPN2132) and a VersaDoc 5000 Imaging System (Bio-Rad) as we described previously. 50 Signal intensities of the captured images were analyzed with the Quantity One Software (Bio-Rad). Results of western blots shown are representatives of two to four independent experiments.
Quantitation of TNF-α. Tissue lysates were diluted ten-fold with 1X Assay Diluent B and levels of TNF-α were determined by the TNF alpha Mouse ELISA Kit (Abcam, ab100747) following the manufacturer's protocol.
SA-β-gal Staining. Control and duct-ligated SMGs were embedded in OCT compound for cryosection and stained for SA-β-gal activity according to the manufacturer's protocol (Cell Signaling Technology, 9860). Tissue slides were counterstained with hematoxylin before mounting medium was applied. The H 2 O 2treated m5-7 and MEF cells were fixed in 2% fomaldehyde, 0.2% glutaraldehyde in PBS for 10 min at room temperature and stained for SA-β-gal activity as above.
In situ apoptosis detection. Apoptosis was detected in FFPE tissue slides (prepared as described above) using an ApopTag Peroxidase In Situ Apoptosis Detection Kit (EMD Millipore, S7100) following the manufacturer's instructions. The percentage of apoptotic cells was calculated by dividing the number of ApopTagpositive cells over total number of nuclei enumerated from randomly selected ten microscopic fields (10X) of individual samples using ImagePro.
Statistical analysis. Gene expression data from quantitative RT-PCR were processed by Microsoft Excel, and analyzed by using GraphPad Prism 6.0 (GraphPad Software, La Jolla, CA, USA). The non-parametric Mann-Whitney test was employed to determine the significant differences between the ligated groups and control group of each genotype as well as the significant differences between Atg5 WT and Atg5 KO mice. Statistic significance of in situ apoptosis assays between Atg5 WT and Atg5 KO groups was determined using Student's t-test. For cell viability assay, data were analyzed with ANOVA followed by Bonferroni t-test.