Isobacachalcone induces autophagy and improves the outcome of immunogenic chemotherapy

A number of natural plant products have a long-standing history in both traditional and modern medical applications. Some secondary metabolites induce autophagy and mediate autophagy-dependent healthspan- and lifespan-extending effects in suitable mouse models. Here, we identified isobacachalcone (ISO) as a non-toxic inducer of autophagic flux that acts on human and mouse cells in vitro, as well as mouse organs in vivo. Mechanistically, ISO inhibits AKT as well as, downstream of AKT, the mechanistic target of rapamycin complex 1 (mTORC1), coupled to the activation of the pro-autophagic transcription factors EB (TFEB) and E3 (TFE3). Cells equipped with a constitutively active AKT mutant failed to activate autophagy. ISO also stimulated the AKT-repressible activation of all three arms of the unfolded stress response (UPR), including the PERK-dependent phosphorylation of eukaryotic initiation factor 2α (eIF2α). Knockout of TFEB and/or TFE3 blunted the UPR, while knockout of PERK or replacement of eIF2α by a non-phosphorylable mutant reduced TFEB/TFE3 activation and autophagy induced by ISO. This points to crosstalk between the UPR and autophagy. Of note, the administration of ISO to mice improved the efficacy of immunogenic anticancer chemotherapy. This effect relied on an improved T lymphocyte-dependent anticancer immune response and was lost upon constitutive AKT activation in, or deletion of the essential autophagy gene Atg5 from, the malignant cells. In conclusion, ISO is a bioavailable autophagy inducer that warrants further preclinical characterization.


Introduction
Macroautophagy (to which we herein refer as "autophagy") is a unique cell biology phenomenon that leads to cytoplasmic vacuolization in response to nutrient deprivation as well as to a myriad of other cell stress-inducing conditions 1 . Portions of the cytoplasm are enveloped in two-membraned vesicles, the autophagosomes, which then fuse with lysosomes for the digestion of the autophagic cargo by hydrolases that operate at acidic pH 2,3 . Autophagy allows to mobilize the cell's energy reserves by digestion of cytoplasmic macromolecules and even entire organelles to recover their building blocks, including amino acids, simple sugars, and free fatty acids 4 . In addition, autophagy allows for the selective degradation of superficial, damaged, or aged cellular components, including dysfunctional organelles and potentially pathogenic protein aggregates. Genetic stimulation of autophagy has potent antiaging properties, reducing the manifestation of age-associated diseases, including arteriosclerosis, cancer, and neurodegeneration [5][6][7] . Pharmacological induction of autophagy has similar broad healthspan and lifespan-extending effects, as shown in model organisms including yeast, nematodes, flies, and mice [8][9][10][11] .
Obviously, there is much interest in identifying novel autophagy inducers that operate at low levels of toxicity and mediate broad antiaging and pro-health effects. Chalcones belong to the chemical class of flavonoids and are contained in multiple plants that are reputed for their dietary virtues. Based on these considerations, we have set out in the past to identify autophagy-inducing chalcones. Among a homemade library of chalcones, we identified two different agents, namely, 4,4′-dimethoxychalcone (4,4′DMC) 12 and its isomer 3,4-dimethoxychalcone (3,4-DMC) 13 as potent autophagy inducers. Of note, both chalcones differ in their mode of action. While 4,4′DMC inhibits autophagy-suppressive GATA transcription factors 12,14 , 3,4-DMC acts through the activation of the two related pro-autophagic transcription factors EB (TFEB) and E3 (TFE3) 13 . Irrespective of this difference, both 4,4′ DMC and 3,4-DMC reduce myocardial infarction in mice. Moreover, 4,4′DMC extended the lifespan of yeast, nematodes, and flies 12 , while 3,4-DMC enhanced anticancer immune responses in mice 13 . These preclinical data plead in favor of a potential medial utility for chalcones.
Driven by these considerations, we decided to identify additional pro-autophagic chalcones by screening another collection of agents. Here, we demonstrate that isobacachalcone (ISO) stimulates autophagic flux, delineate the molecular pathways involved in this effect, and suggest clinical utility for this chalcone as a stimulator of anticancer immunity in the context of immunogenic cell death (ICD)-inducing chemotherapy.

ISO induces autophagic puncta through the inhibition of AKT
ISO is known to inhibit protein kinase B (PKB, best known as AKT) 15,16 . Indeed, U2OS cells stably expressing a GFP-AKT fusion protein responded to stimulation with recombinant insulin growth factor-1 (rIGF1) by a partial translocation of the fluorescent signal to the plasma membrane, reflecting AKT activation. This effect was not detectable for a loss-of-function mutation of AKT consisting of an arginine-to-cysteine mutation in the pleckstrin homology domain of AKT (R25C) ( Fig. 2A, B). In addition, ISO inhibited the activating phosphorylation of AKT (Ser473) as well as, downstream of AKT, the phosphorylation of mechanistic target of rapamycin (mTOR) (Ser448), and the mTOR complex 1 (mTORC1)-dependent phosphorylation of S6K (Thr389) (Fig. 2C). Stable transfection of U2OS cells with a constitutively active AKT mutant (T308D/S473D) inhibited the formation of ISO-induced GFP-LC3 puncta (Fig. 2D, E) as well as the lipidation of LC3 (Fig. 2F, G). In conclusion, it appears that ISO stimulates autophagy through the inhibition of AKT.

ISO induces autophagic flux in vitro and in vivo
Next, we determined whether ISO induces actual autophagic flux by means of several fluorescent reporter-based assays. First, we took advantage of a cell line stably expressing an RFP-ATG4-GFP-LC3ΔG. When expressed in cells, the probe is cleaved into a stable/cytosolic part, RFP-LC3ΔG (that serves as an internal control) and a degradable/quenchable part, GFP-LC3 (which is destroyed by autophagy). Hence, a diminution of the GFP-to-RFP ratio indicates the occurrence of autophagy 17 . ISO consistently induced a decrease in the GFP-to-RFP ratio of cells expressing RFP-ATG4-GFP-LC3ΔG (Fig. 3A, C). We also used cells stably expressing a mCherry-GFP-p62 tandem fusion protein, in which the low pH-sensitive GFP-dependent fluorescence (but less so the pH-resistant mCherry fluorescence) was reduced upon the culture of the cells with ISO (Fig. 1B, D). Similarly, we used a rat adrenal gland (pheochromocytoma) PC12 cell line expressing a tetracycline-inducible variant of Q74-GFP, meaning that the GFP via a polyglutamine tail forms aggregates in the cytoplasm that can be degraded by macroautophagy 18 .
Official journal of the Cell Death Differentiation Association Again, we found that ISO reduced the number of Q74-GFP dots in this experimental system, supporting the idea that it indeed stimulates autophagic flux.
Encouraged by these findings, we determined whether ISO might inhibit the AKT pathway and induce autophagy in vivo. Multiple immunoblot experiments indicated that ISO reduces AKT, mTOR, and S6K phosphorylation while it enhances the abundance of LC3-II in the heart or liver of mice receiving intraperitoneal (i.p.) ISO injections. Thus, ISO can stimulate autophagy in vivo. Notably, the in vivo effects of ISO were not accompanied by measurable weight loss, suggesting that ISO is not toxic.
ISO induces TFEB/TFE3 activation and ER stress U2OS cells exposed to ISO exhibited the translocation of a TFEB-GFP fusion protein from the cytoplasm to the nucleus (Fig. 4A, B). Similarly, TFE3 detectable by immunofluorescence translocated to the nucleus upon culture with ISO (Fig. 4C, D). The nuclear translocation of TFEB and TFE3 could be confirmed by cellular fractionation and immunoblot detection of the two transcription factors in the cytoplasm and nuclei (Fig. 4E-G). Accordingly, knockout of TFEB alone ( In U2OS cells equipped with biosensors of endoplasmic reticulum (ER) stress, we found that ISO induced the upregulation of CHOP (measured by using a GFP gene inserted into the genome under the control of the CHOP promoter, Fig. 5A, B) and activated the IRE1/XBP1 axis (measured by means of an XBP1ΔDBD-venus fusion protein 19 that is only in-frame for venus, a variant of GFP, when XBP1 has been spliced by IRE1, Fig. 5C, D). Similar results were obtained when signs of ER stress were measured by immunofluorescence to detect the nuclear presence of CHOP (Fig. 5E, F) and ATF6 (Fig. 5G, H), the phosphorylation of eukaryotic initiation factor 2α (eIF2α) on serine 51 (Fig. 5I, J) and the expression of the spliced isoform of XBP1 (XBP1s) (Fig. 5K, L). In most of the cases, the ISO-induced signs of ER stress were comparable in magnitude to those induced by the positive controls thapsigargin (TG) and tunicamycin (TM) (Fig. 5A-L). Moreover, the expression of constitutively active AKT mutant blunted the signs of ER stress induced by ISO ( Fig. 5E-L).
Interestingly, a crosstalk between the pro-autophagic and the ER stress-inducing activities of ISO was observed. Thus, TFEB −/− TFE3 −/− cells exhibited a reduced activation of CHOP ( Supplementary Fig. S1A, B) and ATF4 ( Supplementary Fig. S1C, D). Such a reduced CHOP and ATF4 activation was also found for the single-gene knockout of TFEB or TFE3 ( Supplementary Fig. S2). Cells lacking the eIF2α kinase 3 (EIF2AK3, best known as PERK) exhibited reduced phosphorylation of eIF2α in response to ISO ( Supplementary Fig. S1E, F), coupled to reduced formation of autophagic RFP-LC3 puncta (Supplementary Fig. S1G, H). Both the knockout of PERK and a knock-in mutation of eIF2α rendering it nonphosphorylable (due to the substitution of serine 51 by an alanine residue: S51A) significantly reduced the activation of TFE3 by ISO ( Supplementary Fig. S1I-L). These findings suggest molecular crosstalk between the TFEB/ TFE3 and the PERK/eIF2α pathways triggered by ISO.

ISO improves the outcome of immunogenic chemotherapy
Although ISO alone had rather scarce cytotoxic activities, it was able to amplify the ATP release induced by treatment of U2OS cells with low doses of an ICD inducer (mitoxantrone, MTX), as determined by staining of cells with the ATP biosensor quinacrine (Fig. 6A, B) or by (see figure on previous page) Fig. 1 Isobacachalcone (ISO) is a candidate caloric restriction mimetic (CRM). A Human neuroglioma H4 cells stably expressing GFP-LC3 were treated with a selection of chalcones from the TargetMol library of flavonoids at the indicated concentrations. We compared the selected agents at different concentrations with the standard autophagy inducer torin 1 (300 nM), and identified conditions with significantly increased GFP-LC3 puncta formation (1.25 times of the vehicle control (DMSO)) and viability of at least 80% with respect to DMSO, as potent autophagy activation. B, C H4 cells stably expressing GFP-LC3 were treated with isobacachalcone (ISO) (10, 25, and 50 μM) for 6 h. Then the cells were fixed and imaged to assess the formation of GFP-LC3 puncta (C). Torin 1 (300 nM) was used as a prototypical autophagy inducer. Representative images are shown in (B). Scale bar equals 10 μm. Data are means ± SD of quadruplicates ( ** P < 0.01; *** P < 0.001 vs. DMSO/Ctr, Student's t test). D, E U2OS cells were treated as described above, followed by the incubation with specific antibodies to block acetylated tubulin. Thereafter, immunofluorescence was conducted with antibodies against acetylated lysine residues and appropriate AlexaFluor-conjugated secondary antibodies. Representative images of lysine acetylation are shown in (D), and the decrease of acetylation in the cytoplasm was measured in (E). Scale bar equals 10 μm. Data are means ± SD of quadruplicates ( ** P < 0.01 vs. DMSO/Ctr, Student's t test). F, H U2OS cells transfected with a plasmid expressing p62 protein fused with an HA tag (HA-p62) were treated with ISO (25 μM) in the presence or absence of bafilomycin A1 (Baf A1, 100 nM) for 6 h. SDS-PAGE and immunoblot were performed, band intensities of HA-p62 and β-actin (ATCB) were assessed, and the ratio (HA/ATCB) was calculated (H). In parallel samples, band intensities of LC3-II and ATCB were assessed, and their ratio (LC3-II/ATCB) was calculated (G). Data are means ± SD of three independent experiments ( * P < 0.05, ** P < 0.01 vs. untreated control; ## P < 0.01, ### P < 0.001 vs. without Baf A1; Tukey's multiple comparisons test). I, K Human osteosarcoma U2OS cell stably expressing GFP-LC3 either wild-type (WT) or ATG5 knockout (I) were treated with ISO (25 μM) or torin 1 (300 nM) for 6 h. The cells were fixed, imaged, and GFP-LC3 dots were quantified (K). Scale bar equals 10 μm. Data are means ± SD of quadruplicates ( *** P < 0.001 vs. untreated control; ## P < 0.01, ### P < 0.001 vs. WT; Tukey's multiple comparisons test).
measuring ATP in the supernatant of the cells using a biochemical assay ( Fig. 6C-F). ATP is released from stressed cancer cells in an autophagy-dependent fashion 20,21 and acts in the extracellular space as an important chemotactic factor that attracts myeloid immune effectors into the tumor bed, thereby setting off the molecular cascade that permits anticancer immune responses in the context of ICD 22,23 . In contrast, ISO did not affect other autophagy-independent hallmarks of ICD 24 , including surface exposure of calreticulin or the release of high mobility group protein B1 from low-dose MTX-treated cells ( Supplementary Fig. S3). Of note, the knockouts of ATG5 (Fig. 6C) or PERK (Fig. 6D), the S51A mutation of eIF2α ( Fig. 6E) or the expression of a constitutively active AKT mutant (Fig. 6F) reduced the ATP release induced by the combination of low-dose MTX and ISO, supporting the idea that the aforementioned pathways are important for this phenomenon.
(OXA + ISO) allowed for tumor growth control in conditions in which ISO and OXA alone had no or little effect, respectively (Fig. 6H). The anticancer activity depended on the immune system since it was lost in mice lacking mature T cells due to the nu/nu mutation that causes athymia (Fig. 6I). Moreover, tumor cells engineered to lack Atg5 or to express constitutively active AKT failed to respond to the ISO/OXA combination treatment in the immunocompetent setting (Fig. 6J, K). Analysis of the immune infiltrates of the tumors treated with ISO, OXA, or ISO + OXA (Fig. 6L) revealed that the combination treatment was particularly efficient in reducing regulatory T cells (Tregs, defined as CD3 + CD4 + FoxP3 + cells), in improving the ratio of CD8 + cytotoxic T lymphocytes (CTLs) over Tregs and in reducing the expression of the exhaustion marker PD-1 on CTLs ( Fig. 6M-P). In conclusion, ISO stimulates anticancer immunity in the context of ICD-inducing chemotherapy.

Discussion
Here, we identified ISO as an autophagy inducer that inhibits AKT and mTORC1 activity and activates the proautophagic transcription factors TFEB and TFE3, which both are known to be activated by mTORC1 inhibition 27,28 . We also found that ISO activates a broad ER stress response including the PERK-dependent phosphorylation of eIF2α, as a sign of the integrated stress response, which is known to be required for autophagy induction 29-31 as well as for the induction of ICD [32][33][34][35][36][37] . The two pathways, autophagy and ER stress induced by ISO exhibited crosstalk in thus far that (i) they both are inhibited by constitutively active AKT, (ii) TFEB/TFE3 knockout does not only reduce autophagy but also signs of ER stress, and (iii) PERK knockout or substitution of eIF2α by a non-phosphorylable mutant reduces TFEB/TFE3 activation and autophagy. Beyond these in vitro phenomena, ISO induced autophagy in vivo, in mouse tissues, and enhanced the immune response induced by immunogenic chemotherapy against established tumors, thus improving tumor growth control through mechanisms that rely on T cells as well as AKT inhibition and autophagy induction in the cancer cells.
ISO is a chalcone that was first isolated from the multipurpose medical plant Psoralea corylifolia. Reportedly, ISO possesses a wide spectrum of antibacterial 38,39 antifungal 40 50 , and phytoestrogene 51 activities. Hence, ISO has a very broad range of biological activities. In cell-free enzymatic assays, ISO inhibits beta-secretase 52 , acylcoenzyme A: cholesterol acyltransferase 53 , severe acute respiratory syndrome coronavirus (SARS-CoV) papainlike protease 54 , protein tyrosine phosphatase 1B (PTP1B) 55 , carboxylesterase 2 56 , and pancreatic lipase 57 , suggesting that ISO can act on multiple pharmacological targets, shedding doubts on its specificity. Based on its broad effects, it might be suspected that ISO has direct immunostimulatory effects that help to improve immunosurveillance in the context of ICD-inducing chemotherapies. Indeed, autophagy induction may stimulate dendritic and T-cell functions 23,58,59 .
With respect to its anticancer effects, ISO reportedly suppresses skin tumor promotion in an in vivo two-stage mouse skin carcinogenesis test using 7,12-dimethylbenz [a]anthracene (DMBA) as an initiator and 12-O-Tetradecanoylphorbol-13-acetate (TPA) as a promoter 65 . ISO has cytotoxic effects on neuroblastoma 66 , multiple myeloma cells 67,68 , leukemia 69 , as well as on chemoresistant carcinoma and glioblastoma cell lines 70 , enhances TRAILinduced apoptosis in prostate cancer and cervical carcinoma cells 71 , and reduces melanin production by B16 melanoma cells 72 . Here, we found that ISO failed to inhibit the growth of fibrosarcomas in mice when used as a standalone treatment, yet ameliorated the efficacy of ICD-inducing chemotherapy through an improved anticancer immune response. The absence of antitumor efficacy of ISO, when used as a standalone treatment, may (see figure on previous page) Fig. 3 ISO stimulates autophagic flux in vitro and in vivo. A-D Human osteosarcoma U2OS cells stably expressing the tandem reporter construct GFP-LC3-ATG4-RFP-LC3ΔG (A) or the tandem reporter mCherry-GFP-p62 (B) were treated with torin 1 (300 nM) or isobacachalcone (ISO; 25 μM) with or without bafilomycin A1 (Baf A1, 100 nM) for 6 h. After fixation, GFP and RFP fluorescence was measured by automated image analysis, and the ratio of RFP to GFP was calculated (C, D). Scale bar equals 10 μm. Data are means ± SD of quadruplicates ( * P < 0.05, ** P < 0.01, *** P < 0.001 vs. untreated control; ### P < 0.001 vs. without Baf A1; Tukey's multiple comparisons test). E, F Rat adrenal gland PC12 cells stably expressing an inducible variant of Q74-GFP were treated with doxycycline (1 μg/mL) for 8 h for the induction of Q74 expression. Then the medium was changed, and ISO (10, 25, 50 μM) was added for 24 h. Torin 1 (300 nM) was used as a positive control. Representative images are shown in (E), and GFP-Q74 levels were quantitated in (F). Scale bar equals 10 μm. Data are means ± SD of quadruplicates ( ** P < 0.01, *** P < 0.001 vs. DMSO/Ctr, Student's t test). G-M C57BL/6 mice received two intraperitoneal (i.p.) injections of 20 mg/kg/day ISO (n = 3 mice per condition, n = 2 experiments). Organs were collected, and representative immunoblots showing regulators and LC3I-to-LC3-II conversion in the heart (G-K) and in the liver (L-P). AKT, mTOR, and p70 abundance was evaluated, and parallel samples were probed with phosphoneoepitope-specific antibodies. β-actin (ACTB) or vinculin levels were monitored to ensure equal protein loading (H, J). Band intensities of pAKT and ACTB, pmTOR and Vinculin, pS6K and ACTB, as well as LC3-II and ACTB, were assessed, and their ratios were calculated (H-K, M-P). Data are means ± SD (n = 3; ( * P < 0.05, ** P < 0.01 vs. DMSO/Ctr, Student's t test).
be linked to suboptimal dosing as well as to its pharmacokinetics, knowing that ISO has a half-life of~6 h in rats 73 . However, we have observed as a general pattern that autophagy induction with non-toxic agents is not sufficient to inhibit tumor growth of established tumors in mice. Thus, the biological activity of ISO is reminiscent of other autophagy inducers including 3,4-DMC 13 , hydroxycitrate, resveratrol, spermidine 74,75 , and thiostreptone 76 , all of which can ameliorate the therapeutic activity of ICD inducers in suitable mouse models but lack intrinsic anticancer properties.
Although ISO has multiple pharmacological effects and targets, several of the in vitro effects of ISO correlated with the inhibition of the AKT/mTORC1 pathway, and expression of a constitutively active AKT mutant largely reversed the ISO-induced signs of cellular stress including autophagy (with its upstream events, mTORC1 inhibition and TFEB/TFE3 activation) and ER stress (at all levels of the unfolded stress response, including its PERK/eIF2α/ ATG4/CHOP, ATG6, and IRE1α/XBP1 arms), as shown in human U2OS cells. Moreover, mouse cancer cells stably expressing a constitutively active AKT enzyme (or lacking the essential autophagy gene Atg5) became resistant against the anticancer activity of ISO combined with ICD induction, suggesting some sort of 'specificity' for the ISO effect. However, at this point, it is not clear whether ISO may directly inhibit AKT or an enzyme upstream of AKT (such as phosphatidylinositol 3-kinases). Reportedly, ISO inhibits PTP1B 55 , which would result in the activation, not the inhibition of the AKT pathway. Hence, the precise molecular target of ISO remains elusive.
ISO was initially isolated from Psoralea corylifolia, but has also been identified in other plants, including in Angelica keiskei 50 , Artocarpus species 46 , Cullen corylifolium 77 , Dorstena barteri 38 , Erythrena fusca 78 , Fatoua pilosa 44 , Morus alba 79 , and Piper longum 72 . This suggests that ISO is rather prevalent in plants, perhaps contributing to the broad pro-health effects of plant-enriched diets 80,81 . However, additional studies are required to confirm this conjecture.
In summary, here we identified a particular chalcone, ISO, as a potent autophagy inducer that acts in vitro and in vivo, on human cell lines and mouse organs, respectively. Through the induction of autophagy, ISO is able to stimulate anticancer immune responses in the context of immunogenic chemotherapy.

Cell culture and chemicals
Culture media and supplements for cell culture were obtained from Life Technologies (Carlsbad, California, USA) and plastic materials came from Greiner Bio-One (Kremsmünster, Austria) and Corning (Corning, NY, USA).

High-content microscopy
Human osteosarcoma U2OS and neuroglioma H4 cells stably expressing GFP-LC3 or RFP-LC3 and rat adrenal gland PC12 cells stably expressing doxycycline-inducible Q74-GFP were seeded in 384-well black imaging plates at a density of 2000 cells per well and allowed to adapt for overnight. Cells were treated with the indicated agents for 6 h, subsequently, cells were fixed with 3.7% paraformaldehyde (PFA, w/v in PBS) (F8775, Sigma-Aldrich) at 4°C overnight and stained with 1 µg/ml Hoechst 33342 in PBS. Moreover, 2000 U2OS cells either wild-type or stably expressing HMGB1-GFP/CALR-RFP, GFP-ATF6, CHOP:: GFP, GFP-TFEB, or XBP1-DDBD-venus were seeded in 384-well black imaging plates (Greiner Bio-One) and let adhere overnight. Cells were then treated for 6 h to detect TFEB translocation, 16 h to assess ATF6 translocation and spliced XBP1 (XBP1s) levels, or 24 h to measure CHOP promoter activity. For CALR redistribution and HMGB1 release, cells were incubated for 8 h or 24 h respectively. Next, cells were fixed with 3.7% formaldehyde supplemented with 1 μg/ml Hoechst 33342 (H3570, Thermo Fisher Scientific) at 4°C overnight. Subsequently, the fixative was exchanged to PBS, and the plates were analyzed by automated microscopy. Image acquisition was performed using an ImageXpress Micro XL automated microscope (Molecular Devices, Sunnyvale, CA, USA) equipped with a ×20 PlanApo objective (Nikon, Tokyo, Japan), followed by automated image processing with the custom module editor within the MetaXpress software (Molecular Devices). At least four view fields were acquired per well, and experiments involved at least triplicate assessment. Cellular regions of interest, cytoplasm and nucleus, were defined and segmented by using the MetaXpress software (Molecular Devices). After exclusion of cellular debris and dead cells from the dataset, parameters of interest were normalized, statistically evaluated, and graphically depicted with R software. Using R, images were extracted and pixel intensities scaled to be visible (in the same extent for all images of a given experiment).

Nuclear extraction experiment
U2OS-GFP-LC3 cells were collected and processed with the Nuclear Extraction Kit (#ab113474, Abcam) following the manufacturer's methods. The GAPDH antibody (#2118, 1:1000, Cell Signaling Technology) was used as the cytoplasmic control, and H3 (#9715, 1:1000, Cell Signaling Technology) was selected as the nuclear control.

Detection of protein deacetylation
U2OS cells stably expressing GFP-LC3 (~2000 cells/ well) were seeded in 384-well microplates overnight. After experimental treatments, cells were fixed with 3.7% PFA containing 10 μg/ml Hoechst 33342 overnight at 4°C. Thereafter, cells were incubated with an antibody specific for acetyl-alpha-tubulin (#5335, 1:500, Cell Signaling Technology) in 5% BSA (w/v in PBS) for 1 h to block nonspecific binding sites and acetylated tubulins, followed by overnight incubation at 4°C with an antibody specific to acetylated lysine residues (#623402, 1:400, BioLegend, San Diego, California, USA). After washing three times with PBS, cells were incubated in AlexaFluor-568-conjugated secondary antibodies (Life Technologies) for 2 h at room temperature. Fluorescent images were acquired and analyzed as described before.

ATP release assays
Intracellular ATP levels were detected by quinacrine stain assay (Calbiochem) kits, subsequently, the images of quinacrine were obtained by high-content microscopy and the cytoplasmic intensity of quinacrine was quantitated described above. Extracellular ATP levels were measured by the ENLITEN ATP Assay System Bioluminescence Detection Kit (Promega, Madison, Michigan, USA; #FF2000) following the manufacturer's methods. Luminescence was detected by means of a Paradigm I3 multimode plate reader (Molecular Devices).

Animal experimentation
The animal experiments were approved by the Gustave Roussy ethical committee with project number 24771-2020032413235413, and all procedures were performed under the governmental and institutional guidelines and regulations. All mice were maintained in a temperature-controlled and pathogen-free environment with 12-h light/dark cycles, with food and water ad libitum. Animal experiments were conducted in compliance with the EU Directive 63/2010 and protocols 2019_030_20590 and were approved by the Ethical Committee of the Gustave Roussy Campus Cancer (CEEA IRCIV/IGR no. 26, registered at the French Ministry of Research).
For tumor growth experiments, 7-week-old female wildtype C57BL/6 mice or athymic female nude mice (nu/nu) were obtained from Envigo, France (Envigo, Huntingdon, UK). MCA205 wild-type (WT), or continuous activation of AKT T308D/S473D cells (4 × 105), MCA205 cells carrying an ATG5 knockdown (WT, 6 × 105) were subcutaneously injected into C57BL/6 hosts. When tumors became palpable, mice were treated with 20 mg/kg ISO dissolved in corn oil (Sigma-Aldrich) or an equivalent volume of vehicle alone or in combination with 10 mg/kg oxaliplatin (OXA, Sigma-Aldrich) by intraperitoneal injection. On the following days, mice well-being and tumor growth were monitored and documented. Animals were sacrificed when tumor size reached the ethical endpoint or signs of obvious discomfort were observed following the EU Directive 63/2010 and our Ethical Committee advice.

Ex vivo phenotyping of the tumor immune infiltrate
Tumors were harvested, weighed, and transferred on ice into gentleMACS C tubes (Miltenyi Biotec, Bergisch Gladbach, Germany) containing 1 mL of RPMI medium.
Tumors were dissociated first mechanically with scissors, then enzymatically using Miltenyi Biotec mouse tumor dissociation kit (Miltenyi Biotec) and a GentleMACS Octo Dissociator according to the manufacturer's instructions. The dissociated bulk tumor cell suspension was resuspended in RPMI-1640, sequentially passed through 70-μm MACS Smart-Strainer (Miltenyi Biotec), and washed twice with PBS. Finally, bulk tumor cells were homogenized in PBS at a concentration corresponding to 250 mg of the initial tumor weight per milliliter. Prior to staining of tumor-infiltrating lymphocytes (TILs) for flow cytometry analysis, samples (~50 mg) were incubated with LIVE/ DEAD ® Fixable Yellow Dead Cell dye (Thermo Fisher Scientific) to discriminate viable cells from damaged cells. Fc receptors were blocked with anti-mouse CD16/CD32 (clone 2.4G2, Mouse BD Fc Block, BD Pharmingen) before staining with fluorescent-labeled antibodies targeting Tcell surface markers. Surface staining of murine immune cell populations infiltrating the tumor was performed with the following fluorochrome-conjugated antibodies: anti-CD45-AF700, anti-CD3-BV421, anti-CD8-PE, anti-CD4-Percp.Cy5.5, anti-CD25-PE/Cy7, and anti-PD-1-APC/Cy7 (BioLegend). Then, cells were fixed and permeabilized in eBioscience Foxp3/Transcription Factor Staining Buffer (Thermo Fisher Scientific) and stained for intracellular Foxp3. Finally, stained samples were run through a BD LSR II flow cytometer. Data were acquired using BD FACSDiva software (BD Biosciences) and analyzed using FlowJo software (TreeStar). Absolute counts of leukocytes and tumor cells were normalized considering the following parameters: the weight of the harvested tumor and total volume of the dissociated tumor cell suspension (cell concentration typically set to 250 mg/mL in PBS), the proportion of the whole-cell suspension, and proportion of the cell suspension used for cytometry.

Statistical analysis
Unless otherwise mentioned, data are reported as means ± SD of triplicate determinations, and experiments were repeated at least three times yielding similar results. Statistical significance was assessed by Student's t test. TumGrowth and GraphPad were used to analyze in vivo data arising from murine models 84 . TumGrowth is available at Github/Kroemerlab. P values of 0.05 or less were considered to denote significance ( * P < 0.05; ** P < 0.01; *** P < 0.001; ns, not significant).