STAT3 suppresses the AMPKα/ULK1‐dependent induction of autophagy in glioblastoma cells

Abstract Despite advances in molecular characterization, glioblastoma (GBM) remains the most common and lethal brain tumour with high mortality rates in both paediatric and adult patients. The signal transducer and activator of transcription 3 (STAT3) is an important oncogenic driver of GBM. Although STAT3 reportedly plays a role in autophagy of some cells, its role in cancer cell autophagy remains unclear. In this study, we found Serine‐727 and Tyrosine‐705 phosphorylation of STAT3 was constitutive in GBM cell lines. Tyrosine phosphorylation of STAT3 in GBM cells suppresses autophagy, whereas knockout (KO) of STAT3 increases ULK1 gene expression, increases TSC2‐AMPKα‐ULK1 signalling, and increases lysosomal Cathepsin D processing, leading to the stimulation of autophagy. Rescue of STAT3‐KO cells by the enforced expression of wild‐type (WT) STAT3 reverses these pathways and inhibits autophagy. Conversely, expression of Y705F‐ and S727A‐STAT3 phosphorylation deficient mutants in STAT3‐KO cells did not suppress autophagy. Inhibition of ULK1 activity (by treatment with MRT68921) or its expression (by siRNA knockdown) in STAT3‐KO cells inhibits autophagy and sensitizes cells to apoptosis. Taken together, our findings suggest that serine and tyrosine phosphorylation of STAT3 play critical roles in STAT3‐dependent autophagy in GBM, and thus are potential targets to treat GBM.

Akt/mTOR signalling axis plays a major role in GBM biology. 4,5 In addition, accumulating evidence has shown that signal transducer and activator of transcription 3 (STAT3) is an important oncogenic driver in many cancers including GBM. 6,7 Under normal physiological conditions, cytoplasmic STAT3 undergoes phosphorylation at Tyrosine (Y)-705 and Serine (S)-727 residues. STAT3 tyrosine phosphorylation induces homodimerization and/or heterodimerization with other STAT family proteins, nuclear translocation, and DNA binding, leading to the induction of cytokine responsive genes 8 and anti-apoptotic genes. 9 The role of S727 phosphorylation is less well understood, but studies suggest that it may be required for STAT3's maximum transcriptional activity. 10 STAT3 is constitutively phosphorylated in GBM cancer stem cells (GSCs) and inhibiting STAT3 phosphorylation attenuates GSC-driven tumour growth, 11,12 showing that STAT3 plays a critical role in GBM tumorigenesis.
Autophagy is a highly conserved cellular catabolic process that recycles damaged organelles, protein aggregates, and other toxic intracellular debris. Autophagy has a complex and context-dependent role in tumour development and cancer therapy. 13 Although autophagy suppresses primary tumour growth, it is required for advanced tumour growth with elevated metabolic demand and promotes multiple steps in tumorigenesis. 14 Constitutive activation of mTOR signalling impairs basal autophagy in GBM, which enhances proliferation and pluripotency of GSCs. 15 Conversely, restoration of autophagy through mTOR inhibition reduces the invasive potential of GSCs, suggesting that mTOR hyperactivation sustains GSC metabolism through suppressing autophagy. 16 Increased autophagy has been associated with both tumour survival and chemoresistance in GBM. 17 Although these studies underscore the relevance of autophagy in GBM, little is known about the function of STAT3 signalling in regulating autophagy in GBM. Pharmacologic inhibition of either JAK2 (using SAR317461) 18 or STAT3 (using AG490) 19 stimulates autophagy in GBM cells. Nuclear STAT3 inhibits autophagy by upregulating anti-autophagy genes and downregulating pro-autophagy genes. 20 An inverse correlation between phosphorylated STAT3 and a stimulator of autophagy, Beclin1, has also been observed in GBM. 21 Based on these observations, we sought to define the specific STAT3-dependent signalling mechanisms that modulate autophagy and thereby, impact GBM tumorigenesis and chemosensitivity.

| Western blotting
Cell lysates were prepared using mammalian protein extraction buffer (Cell Signalling Technology) and a protease inhibitor cocktail followed by SDS-PAGE. Proteins were transferred to Immobilon-P membranes (Millipore Bedford) and probed with primary antibodies overnight at 4°C in TBS buffer containing 0.1% Tween-20 and 5% nonfat dry milk (Bio-Rad). Membranes were subsequently incubated with horseradish peroxidase-conjugated secondary antibodies at room temperature for 1 h, and the immunocomplexes were visualized by the ECL detection system (Perkin Elmer) quantified on the Azure Biosystems C500. Membranes were stripped and re-probed for actin or GAPDH as loading controls. Representative Western blots from three experiments are shown. Densitometric analysis of all Western blots was performed using Image J software.

| Immunoprecipitation
Glioblastoma cells were rinsed with ice cold PBS and lysed using a cell lysis buffer (Cell Signalling Technology, Inc.) containing protease and phosphatase inhibitors (Thermo Fisher Scientific). The lysates were clarified by centrifugation at 14,300 g for 15 min at 4°C. The cell extracts containing equal amounts of protein were incubated with STAT3 antibody overnight at 4°C followed by addition of protein A/G agarose beads (Santa Cruz Biotechnology) with gentle rocking for 2 h. The beads were washed 3 times with lysis buffer and once with PBS, and the immunocomplexes were released by heating in Laemmli sample buffer and analysed by Western blotting using trimethyl-STAT3 antibody (EMD Biosciences/Millipore Corp.).

| Immunofluorescence and confocal microscopy
Cells were cultured in chamber slides (Millipore) to ~70% confluence and washed with PBS. Cells were fixed in 4% paraformaldehyde and methanol, and permeabilized with 1% Triton X-100. After blocking with 5% goat serum, cells were incubated with anti-rabbit LC3 and anti-mouse p62 antibodies and subsequently stained with Alexa Fluor 488 (goat anti-rabbit) and Alexa Fluor 633 (goat antimouse) secondary antibodies, as described previously. 24 DNA was counterstained with Vectashield mounting media with DAPI (Vectra Laboratories). Images were captured on a Zeiss LSM700 laser scanning confocal microscope.

| siRNA transfection
MT330 cells were grown to 60%-70% confluency in 6-well tissue culture plates, and siRNA transfection was performed using a protocol available from Santa Cruz Biotechnology. For ULK1 siRNA transfection, the cell monolayer was washed with siRNA transfection medium (Santa Cruz) and the siRNA/transfection reagent mixture was added dropwise on to the cell monolayer and incubated overnight at 37°C in a CO 2 incubator. The following day complete growth medium containing 2 times the normal serum and antibiotics was added without removing the transfection mixture. After an additional incubation for 18-24 h, the medium was aspirated and replaced with fresh 1X growth medium. After another 24 h of incubation, cells were treated with or without 100 nM Baf and assayed for autophagy and apoptosis markers. Efficiency of transfection was monitored using FITC-conjugated control siRNA.

| Apoptosis
The quantitative DNA fragmentation assay was performed using a cell death ELISA kit as described earlier. 25 Briefly, MT330 cells were either treated with MRT68921 or ULK1 siRNA and the attached cells were washed twice with Dulbecco's phosphate buffered saline (DPBS). Cells were lysed, and an aliquot of the nuclei-free supernatant was placed in streptavidin-coated plates and incubated with anti-histone biotin and anti-DNA peroxidase-conjugated antibodies for 2 h at room temperature. After incubation, the samples were aspirated, and the wells were washed 3 times with incubation buffer.
After the final wash, 100 μl of the substrate, 2,2′-azino-di[3-ethylbe nzthiazolin-sulfonate], was added in the wells for 3 min at room temperature. The absorbance was read at 405 nm using the SpectraMax iD3 microplate reader (Molecular Devices). Results were expressed as absorbance at 405 nm/mg protein/min.

| Statistical analysis
All data were analysed by GraphPad Prism 9 program (GraphPad Software Inc.), and an unpaired 2-tailed Student's t-test was used to assess statistical significance. Data are expressed as mean ± SE.

| STAT3 deletion increases autophagy
We previously established STAT3-KO MT330 GBM cells by CRISPR/ Cas9 gene editing; as a control, cells were transduced with empty vector (EV). Absence of STAT3 protein in MT330 cells was validated in whole-cell extracts by immunoblotting with antibodies to STAT3 ( Figure 1A). STAT3 expression was restored in the STAT3 knockout cells by transduction with lentiviral vectors encoding either the WT-STAT3 or the STAT3 mutants (Y705F and S727A). In EVtransduced MT330 cells, STAT3 was phosphorylated on both Y705, and Ser727 ( Figure 1A  STAT3 deletion in MT330 cells markedly increased total ULK1 protein expression and phosphorylation at all sites ( Figure 1E). Our results demonstrate significantly higher levels of phosphorylated ULK1 S555 in STAT3-KO cells compared to EV ( Figure 1F). Given that ULK1 acts as a direct target for both mTORC1 and AMPKα, 28  ( Figure 1F) suggesting that S727-STAT3 phosphorylation is involved in AMPKα-dependent pathway but does not involve ULK1 to suppress autophagy. Reconstitution with WT-STAT3 in KO cells completely reversed both pathways. Bafilomycin (Baf) treatment had no effect on AMPKα and ULK1 phosphorylation in MT330 cells, indicating that these upstream pro-autophagy pathways are unaltered by disruption of autophagy flux.
Since AMPKα inhibits mTORC1 by phosphorylating and activating TSC2, 29 we examined TSC2 phosphorylation in STAT3-KO MT330 cells. We found that KO of STAT3 significantly increased TSC2 S1387 and Thr1462 phosphorylation, which was reversed with WT-STAT3 expression ( Figure 1G,H). Expression of Y705F mutant also increased TSC2 S1387 and T1462 phosphorylation. In contrast, cells expressing the S727A mutant showed significantly lower levels of phosphorylated TSC2 T1462 but high levels of S1387 phosphorylation, demonstrating that STAT3 S727 phosphorylation selectively regulates TSC2 T1462 phosphorylation ( Figure 1G,H). Thus, AMPKα activation in STAT3-KO cells acts through more than one pathway to activate autophagy, by directly activating ULK1 and by indirectly impairing inhibition of ULK1 through activation of TSC2 S1387 phosphorylation. Since Akt inhibits TSC2 via phosphorylation on T1462, resulting in basal mTORC1 activation, it is possible that mTORC1 could phosphorylate and inhibit ULK1. This may be counteracted by ULK1 phosphorylation at S555 leading to its sustained activity, which is required for maintaining autophagy flux in STAT3-KO cells.

| STAT3-deletion increases autophagic flux in LN229 cells
We F I G U R E 1 STAT3 deletion activates autophagy through AMPKα-ULK1-TSC2 signalling pathways in MT330 cells. (A) Confluent EV MT330 cells, STAT3-knockout cell line # 2 (KO2), STAT3-knockout cell line # 3 (KO3), STAT3-KO3 rescued with wild-type (WT), and STAT3-KO3 cells expressing Y705F-STAT3 and S727A-STAT3 mutants were exposed to Bafilomycin (Baf, 100 nM) for 3 h. Untreated (UT) cells served as controls. Total cell lysates were prepared and immunoblotted with indicated antibodies. (B) Quantification of the ratio of phospho-STAT3,, and total-STAT3 from three independent experiments. (C) Cell lysates were analysed for p-AMPKα T172. Blots were stripped and probed for total-AMPKα. (D) Quantification of the ratio of phosphorylated and total AMPKα shown in C. (E) Cell lysates were analysed for p-ULK1 S555 and S638. Blots were stripped and probed for total-ULK1. (F) Quantification of the ratio of phospho-ULK1 S555 and total-ULK1. (G) Western blotting of cell lysates with phospho-TSC2 antibodies. Blots were stripped and probed with total-TSC2 antibody. (H) Quantification of the ratio of phospho-T1462-and total-TSC2 and phospho-S1387 and total-TSC2  We also examined the role of STAT3 in mTOR phosphorylation in LN229 cells. The mTOR protein and its basal phosphorylation status at S2481 were unaltered in the KO and lines expressing STAT3 phosphorylation-inactive mutants ( Figure 2G), demonstrating that basal mTOR activity is unaltered. Furthermore, expression of S6 ribosomal protein (S6Rbp) and its phosphorylation at S235/S236 were also unaltered in KO cells and cells expressing STAT3 mutants ( Figure 2G). Therefore, autophagy induction in LN229 cells either lacking STAT3 or expressing the phosphorylation-defective STAT3 mutants involves AMPKα but is independent of the mTOR pathway.
We examined whether STAT3-KO activates ULK1 signalling in LN229 cells. Our data show that ULK1 protein was detectable in LN229 cells. Although STAT3-KO decreased ULK1 protein levels ( Figure 2H), it significantly increased ULK1 activity ( Figure 2I).
Rescue with WT-STAT3 decreased ULK1 activity ( Figure 2H,I), but expression of the Y705 mutant increased ULK1 activity in LN229 cells, which is consistent with our findings in MT330 cells. However, expression of the S727A mutant in STAT3-KO cells had no effect on ULK1 activity, also suggesting that the Y705 mutant is the key critical regulator of AMPKα and ULK1 signalling in LN229 cells.

| The dependence of autophagy on STAT3
To further assess the impact of STAT3 deletion and rescue of KO deletion with mTOR inhibition increases autophagy ( Figure 3D,G).

| The role of STAT3 on autophagy-associated gene expression
BNIP3 has been shown to play a role in regulating the autophagy pathway. 35 In previous studies, we showed that STAT3-KO in MT330 cells inhibited the expression of classical STAT3 genes such as Cyclin D1 and vascular endothelial growth factor. 23 We found that STAT3-KO decreases BNIP3 protein levels in MT330 cells ( Figure 4A,B). Expression of Y705F and S727A mutants decreased BNIP3 protein levels ( Figure 4A) and WT-STAT3 reversed these changes. Since we found that STAT3-KO increased ULK1 protein levels and decreased p62 levels in MT330 cells (Figures 1 and 3), we next performed qPCR on these autophagy-related genes. In agreement with these observations, our data show that STAT3 increases BNIP3 gene expression in MT330 cells (Figure 4C), showing a correlation between mRNA and protein levels. Although BNIP3 is a potential target of ULK1, decreased BNIP3 protein levels in STAT3-KO and mutant expressing cells indicate that BNIP3 is degraded by ULK1-dependent autophagy, a mechanism found to decrease BNIP3 protein levels in tumour cells. 36 Consistent with data in Figure 3, STAT3-KO reduced p62 protein and transcript levels ( Figure 4D,E), which is reversed by WT-STAT3 expression. In addition, both Y705 and S727 residues are necessary for expression of p62 transcript and protein levels since cells expressing STAT3 mutants showed similar expression of p62 to STAT3-KO cells. In contrast, STAT3 reduces expression of ULK1 gene ( Figure 4F) and STAT3 phosphorylation on both Y705 and S727 residues is necessary for ULK1 repression ( Figure 4D,F). Together, these results demonstrate that STAT3 may regulate autophagy in MT330 cells in part through the transcriptional regulation of several autophagy-related genes.
We next examined p62 and ULK1 transcript levels in LN229 cells. Our data demonstrate that STAT3-KO and expression of STAT3 mutants significantly reduced p62 protein levels ( Figure 4G,H). To demonstrate whether mRNA and protein levels of p62 and ULK1 were correlated, we examined their transcript levels in LN229 cells.
We found that p62 and ULK1 gene expressions were unaltered in STAT3-KO LN229 cells ( Figure 4I,J), which differed from their expression in STAT3-KO MT330 cells. In addition, expression of STAT3 mutants in STAT3-KO cells had no effect on p62 and ULK1 genes.
These results demonstrate that reduction of p62 protein levels in LN229 cells is due to enhanced degradation during autophagy, not due to changes in p62 gene transcription.

| Immunolocalization of LC3 and p62
To further validate the role of STAT3 in autophagy, we performed immunolocalization studies to detect LC3 and p62 puncta formation. Under basal conditions, we observed diffuse LC3 (green) and p62 (red) staining (green) predominantly in the cytoplasm of control (EV) MT330 cells with some nuclear staining ( Figure 5). Treatment of EV cells with a low dose of Baf (1μΜ) for 48 h had no effect on LC3 and p62 staining. These observations are consistent with our data presented in Figure 3 and demonstrate impaired autophagy in

| Targeting of ULK1 blocks autophagy and induces apoptosis in STAT3-KO MT330 cells
Because STAT3-KO in MT330 cells induces autophagy and concomitantly activates AMPKα/ULK1 signalling, we tested the effects   cells. MRT68921 did not significantly block mTOR S2448 phosphorylation ( Figure 6A). Since mTORC1 is phosphorylated predominantly on S2448, these observations suggest that ULK1 inhibition has no significant effect on mTORC1.
MRT68921 also significantly increased LC3-II levels in STAT3-KO cells and in STAT3-KO cells transduced with the phosphorylationdefective Y705F or S727A mutants ( Figure 6A,E). Given that ULK1 inhibition has been shown to disrupt autophagosome maturation downstream of LC3 conjugation, 37   To further confirm the role of ULK1 in autophagy regulation, we knocked down ULK1 with siRNA in STAT3-deleted MT330 cells and STAT3-mutant expressing lines that contained high levels of ULK1 ( Figure 1E). Control scrambled siRNA had no effect on ULK1 levels, but ULK1-specific siRNA significantly decreased ULK1 levels in STAT3-KO cells and cells expressing both Y705F and S727A mutants ( Figure 7A,B). LC3-II levels did not increase in ULK1

| DISCUSS ION
A key finding of these studies is that constitutive STAT3 phosphorylation suppresses autophagy in GBM cells. KO of STAT3 and the The role of autophagy in cancer is controversial because it has been reported to both promote and inhibit tumorigenesis. 40 Autophagy confers drug resistance to radiotherapy and chemotherapy but also slows tumour progression. 41 In contrast, rapid tumour growth in GBM and insufficient nutrient supply from the tumour vasculature contribute to activating autophagy and desensitizing tumour cells to chemotherapy. 42 44 Other STAT3 PTMs include acetylation on K685 by CBP/p300, S-glutathionylation by intracellular oxidative stress and trimethylation by EZH2. 45 We found that STAT3 is constitutively phosphorylated on Y705 and S727 residues, acetylated on K685 and trimethylated on K180 ( Figure S1), but these modifications were interdependent in GBM cells. STAT3 has been shown to regulate autophagy through several mechanisms. 46 Nuclear STAT3 regulates autophagy through the transcriptional regulation of pro-autophagy genes such as Beclin1 (BECN1) and anti-or pro-autophagy modulating microRNAs. 46,47 Our qPCR analyses of autophagy genes demonstrate that STAT3 F I G U R E 7 Knockdown of ULK1 expression blocks autophagy and induces apoptosis in STAT3-KO and STAT3-mutant expressing lines. (A) MT330 EV, STAT3-KO, WT, Y705F and S727A mutant expressing cells were grown to confluence and treated with control or ULK1 siRNA followed by treatment with or without Baf for 3 h. ULK1 knockdown was verified by Western blotting with ULK1 and phospho-ULK1, and cell lysates were analysed for LC3, cleaved caspase-3 with β-Actin as a loading control. Quantification of the ratio of (B) total-ULK1 and Actin; (C) LC3-II and Actin; (D) active caspase-3 and Actin. (E) Apoptosis was measured by ELISA as in Figure  including Beclin1 and ATG101. 50  were blocked by expression of WT-STAT3. This is consistent with earlier studies showing STAT1 as a transcriptional suppressor of autophagy through inhibition of ULK1 protein and mRNA levels. 51 In LN229 cells, STAT3-KO had no effect on ULK1 mRNA and protein levels but increased ULK1 activity, demonstrating that enhancement of ULK1 activity is transcription independent. Our studies highlight the function of various phosphorylation-sites in STAT3 mutants, and our data unambiguously demonstrate that Y705F and S727A mutants differentially regulate AMPKα and ULK1 signalling to activate autophagy in GBM lines. In this context, our data also support the cellular and molecular heterogeneity seen between LN229 and MT330 GBM lines. This may account for the observed variability in AMPKα signalling in response to STAT3 deletion and expression of STAT3 mutants.
AMPKα activation inhibits mTORC1, which leads to autophagy. In addition to regulating mTORC1, AMPKα activation inhibits tumour cell growth by phosphorylating TSC2 on S1387, which in turn inhibits mTORC1 leading to autophagy activation. 52 Our data demonstrate that cells lacking STAT3 and cells expressing STAT3 mutants have higher amounts of TSC2 T1462 and S1387 phosphorylation. The regulation of TSC2 and mTORC1 by AMPKα has special implications in autophagy regulation. While TSC2 T1462 phosphorylation inhibits its activity leading to mTORC1 activation, 53 AMPKα inhibits mTORC1 in part by phosphorylating and activating TSC2 on S1387. 54 ULK1 S555 phosphorylation is mediated through AMPKα and indicates autophagy activation, whereas other ULK1 sites are targeted by mTOR to inhibit autophagy. 28 Conversely, ULK1 inhibits the kinase activity of mTORC1 to stimulate autophagy. 55 We find that AMPKα is activated in STAT3-KO and STAT3-KO GBM cells expressing phosphorylation-defective STAT3 mutants, which suggests that AMPKα directly phosphorylates ULK1 on several sites required to sustain ULK1 activation. This mechanism of ULK1 activation is sufficient to inhibit mTORC1 and activate autophagy. Future studies are needed to further define how STAT3 regulates AMPKα signalling, but our data suggest that activated AMPKα induces autophagy in STAT3-KO cells through TSC2 to directly activate ULK1.
Consistent with STAT3 suppressing autophagy through inhibition of AMPKα/ULKI signalling, we found that inhibiting ULK1 activity by MRT68921 or ULK1 protein knockdown by siRNA decreased autophagy, and sensitized STAT3-KO and STAT3-phosphodeficient mutant expressing lines to caspase-3-dependent apoptosis. These results suggest that autophagy is cytoprotective in STAT3-KO cells.
Thus, approaches directed at inhibiting ULK1 may inhibit autophagy through downstream blockade of ULK1-Atg14-Beclin1 signalling and consequently lead to GBM cell death. To our knowledge, this is the first example of the involvement of ULK1 signalling in the regulation of STAT3-dependent autophagy/apoptosis and highlight the targeting of both STAT3 and ULK1 as a potential novel therapeutic approach for GBM treatment.

ACK N OWLED G EM ENTS
The authors thank Weihong Huo for technical assistance.

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study and the STAT3-KO and mutant expressing GBM cells generated during and/or analysed during the current study are available from the corresponding author on reasonable request.