Culture media and supplements for cell culture were obtained from Life Technologies (Carlsbad, California, USA) and plasticware came from Greiner BioOne (Kremsmünster, Austria). PC12 cells stably expressing doxycycline-inducible Q74-GFP were cultured in Roswell Park Memorial Institute (RPMI)-1640 containing 5% fetal bovine serum and 10% horse serum.19 Human osteosarcoma U2OS TFE^-/-cells carrying a knockout for TFE3 and TFEB,20 MCA205 murine fibrosarcoma stably expressing shRNAs interfering with the expression of TFE3 and TFEB,20 and all the other cells were maintained in Dulbecco’s modified Eagle’s medium, supplemented with 10% fetal bovine serum and both containing 100 units/mL penicillin G sodium and 100 mg/mL streptomycin sulfate. All cells were cultured at 37°C under 5% CO₂. The Prestwick Chemical Library was obtained from Prestwick Chemical (Illkirch, France), flavonoids were purchased from Extrasynthese (Genay, France), fatty acids came from Larodan (Malmö, Sweden); known calorie restriction mimetics, cisplatin, dactinomycin, cycloheximide, staurosporine, thapsigargin, tunicamycin, bafilomycin A1, and cycloheximide were from Sigma-Aldrich (St. Louis, Missouri, USA).

Human osteosarcoma U2OS cells wild type or stably expressing GFP-LC3; GFP-TFEB; mRFP-GFP-LC3; mCherry-GFP-p62, HMGB1-GFP and CALR-RFP, or human glioma H4 cells stably expressing GFP-LC3 or PC12 cells expressing GFP-Q74 were seeded in 384-well black microplates. After treatment, cells were fixed with 4% paraformaldehyde (PFA, w/v in phosphate buffered saline (PBS)) overnight at 4°C and stained with 10 µg/mL Hoechst 33 342 in PBS. Images were acquired using an ImageXpress micro XL automated microscope (Molecular Devices, San Jose, California, USA) with a ×20 PlanApo objective (Nikon, Tokyo, Japan). A minimum of four view fields was acquired per well, and experiments involved at least triplicate assessments. Quantitation was done on 1000–2400 cells per condition. Following images were analyzed using the Custom Module Editor of the MetaXpress software (Molecular Devices). Briefly, cells were segmented and divided into nuclear and cytoplasmic regions based on the nuclear Hoechst staining and cytoplasmic GFP or RFP signal. GFP-LC3 dots were measured in the cytoplasmic compartment. For quinacrine assays, cells were incubated with Krebs-Ringer solution containing 5 µM quinacrine and 20 µg/mL Hoechst 33 342 at 37°C for 30 min, then cells were washed and imaged alive. Quinacrine dots were measured in the cytoplasmic compartment. GFP-TFEB or HMGB1 fluorescence intensities were measured in the nuclear compartment. CALR-RFP aggregation was measured perimembranous while excluding perinuclear signal. Data processing and statistical analyses were performed using the R software (http://www.r-project.org/).

All mice were kept in an environmental-controlled and pathogen-free facility with 12 hours light/dark cycles and food and water ad libitum. Female wild-type C57BL/6 mice and athymic female nude mice (nu/nu) at the age of 6–8 weeks were obtained from Envigo, France (Envigo, Huntingdon, UK). For immunoblotting assessment of LC3, wild-type C57BL/6 mice were treated with thiostrepton, liver were thereafter collected and processed. For the GFP-LC3 assessment, 8-week-old GFP-LC3 transgenic (Tg-GFP-LC3) mice, containing a rat LC3-EGFP fusion protein under the control of the chicken b-actin, were treated with thiostrepton with or without 40 mg/kg of the autophagic flux inhibitor leupeptin 2 hours before the organs were collected and processed. For tumor growth experiments, MCA205 or TC-1 tumors were established in C57BL/6 hosts by subcutaneously inoculating 3×10⁵ cells. Thiostrepton was dissolved in 3.34% dimethylsulfoxide (DMSO), 10.02 % PG400 (Sigma-Aldrich), 10.02% TWEEN80 (Sigma-Aldrich) and 76.6% corn oil (Sigma-Aldrich).

When tumors became palpable, mice were treated with thiostrepton in 200 mg/kg 1 day before, on, and 1 day after the chemotherapy (10 mg/kg oxaliplatin, Sigma-Aldrich) by intraperitoneal injection. On the following days, thiostrepton were given 40 mg/kg alone three times per week, and mice well-being and tumor growth were monitored and documented. Animals were sacrificed when tumor size reached endpoint or signs of obvious discomfort were observed following the EU Directive 63/2010 and our Ethical Committee advice.

U2OS-GFP-LC3 stable cells were seeded in 384-well microplates for 24 hours. After experimental treatments, cells were fixed with 4% PFA for 20 min at room temperature. Thereafter, cells were incubated with an antibody specific for acetyl-alpha-tubulin (#5335, 1:500, Cell Signaling Technology) in 5% bovine serum albumin (BSA, w/v in PBS) for 1 hour to block non-specific binding sites and acetylated tubulins, followed by overnight incubation at 4°C with specific antibody to detect acetylated proteins at lysines (#623402, 1:400, BioLegend, San Diego, California, USA). After washing several times with PBS, cells were incubated with Alexa Fluor conjugates (Life Technologies) against the primary antibody for 2 hours at room temperature. Nuclei were stained by incubating the cells with 10 µg/mL Hoechst 33 342 in PBS. Fluorescent images were acquired and analyzed as described before.

Cell culture supernatants were collected and centrifuge to remove debris. The concentration of HMGB1 was detected by ELISA (HMGB1 ELISA kit II; Shino Test Corporation, Tokyo, Japan) according to the manufacturer's instructions. The positive control was from the same kit. Absorbance was analyzed by means of an Paradigm I3 multimode plate reader (Molecular Devices).

After treatment, cells were collected and lysed in cold RIPA lysis and extraction buffer (Thermo Fisher Scientific, Carlsbad, California, USA) containing Pierce Protease and Phosphatase Inhibitor Mini Tablet (Thermo Fisher Scientific) on ice for 40 min. After centrifugation at 12 000g for 15 min, supernatants were heated in sample buffer (Thermo Fisher Scientific) at 100°C for 10 min. Protein samples were separated on precast 4%–12% polyacrylamide NuPAGE Bis–Tris gels (Thermo Fisher Scientific) and electrotransferred to Polyvinylidene difluoride (PVDF) membranes (Millipore Corporation, Burlington, Massachusetts, USA). Membranes were probed overnight at 4°C with primary antibodies specific to LC3 (#2775, 1:1000, Cell Signaling Technology), p62 (#ab56416, 1:10 000, Abcam, Cambridge, UK), Atg5 (#A2859, 1:1000, Sigma-Aldrich, St. Louis, Missouri, USA), GAPDH (#ab8245, 1:10 000, Abcam), TFEB (#4240, 1:1000, Cell Signaling Technology, Danvers, Massachusetts, USA), TFE3 (ab93808, 1:2000, Abcam), and β-actin (#ab49900, 1:10 000, Abcam) followed by incubation with appropriate horseradish peroxidase–conjugated secondary antibodies (Southern Biotech, 1:5000, Birmingham, Alabama, USA). Immunoreactive bands were visualized with ECL prime western blotting detection reagent (Sigma-Aldrich). Equal loading was controlled for all blots and one representative loading control is depicted, as indicated in the figure legends.

Cells were incubated with 3.7% PFA for 20 min at room temperature. For TFE3 staining, fixed cells were then permeabilized with 0.1% Triton X100 on ice, and blocked with 5% bovine serum albumin (BSA, w/v in PBS) for 1 hour. For phospho-eIF2 alpha (Ser51) staining, fixed cells were directly blocked with 5% BSA for 1 hour. Then cells were incubated with antibodies specific to TFE3 (#PA5-54909, 1:500, Invitrogen) or phospho-eIF2 alpha (Ser51) (#ab32157, 1:500, Abcam) overnight (ON) incubation at 4°C. Thereafter, Alexa Fluor 647 nm conjugates (Thermo Fisher Scientific) against the primary antibody were applied for 2 hours at RT. Cells were then washed and imaged by high-content microscopy as described above. The nuclear intensity of TFE3 and cytoplasmic intensity of phospho-eIF2 alpha (Ser51) were measured and normalized on Ctrl.

Liver tissues were fixed with 10% neutral buffered formalin at overnight at 4°C and then transferred into 30% sucrose diluted in PBS for 24 hours at 4°C. Thereafter, the organs were embedded in optimal cutting temperature (OCT) solution, and consecutive tissue sections were cut by cryostat. Samples were stained with 4′,6-diamidino-2-phenylindole (DAPI) to detect nuclei, and images were acquired by confocal microscopy. GFP-LC3 dot area was measured by ImageJ.

Cellular supernatants were collected and processed with the ENLITEN ATP Assay System Bioluminescence Detection Kit (Promega, Madison, Michigan, USA; #FF2000) following the manufacturer’s methods. Fluorescence was detected by means of a Paradigm I3 multimode plate reader (Molecular Devices).

Cells were detached and collected and sequentially incubated with Zombie UV (#423114; Biolegend), anti-CALR rabbit monoclonal antibody (#ab2907; Abcam, Cambridge, UK) for 30 min at 4°C, followed by Alexa Fluor 488, 546 or 647 goat anti-rabbit IgGs (#A11034, Thermo Fisher Scientific) for 30 min at 4°C, then fixed with 4% PFA in PBS for flow cytometric assessment.

Cells were detached and collected and incubated with annexin V (Biolegende #640943) and DAPI according to the manufacturer’s advice. Following cells were assessed by flow cytometry.

Cells were treated as indicated and menadione was used for 1 hour at 100 µM as a positive control. CellROX deep red (#C10422, Thermo Fisher Scientific) was added for 30 min at a final concentration of 5 µM. Then cells were fixed and stained with 4% PFA containing Hoechst for 30 min at room temperature and images were acquired and analyzed by automated microscopy.

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 and a GentleMACS Octo Dissociator following the manufacturer’s instructions. Tumor homogenates were filtered through 70 µm MACS SmartStrainer (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 millilitre. Bulk tumor cell homogenates, each corresponding to 50 mg of the initial tumor sample, were stained with LIVE/DEAD Fixable Yellow dye (Thermo Fisher Scientific). Fc receptors were blocked with anti-mouse CD16/CD32 (clone 2.4G2, Mouse BD Fc Block, BD Pharmingen). Surface staining of murine immune cell populations infiltrating the tumor was performed with the following fluorochrome-conjugated antibodies: (1) ‘Neutrophils’ panel: anti-CD45 APC-Fire750 (clone 30 F-11, BioLegend), anti-Ly-6G PE (clone 1A8, BD Pharmingen), anti-Ly-6C FITC (clone AL-21, BD Pharmingen), anti-CD11b V450 (clone M1/70, BD Pharmingen); (2) ‘T-cell activation/exhaustion’ panel: anti-CD3 APC (clone 145–2 C11, BioLegend), anti-CD8 PE (clone 53–6.7, BD Pharmingen), anti-CD4 PerCP-Cy5.5 (clone RM4-5, Thermo Fisher Scientific), anti-CD25 PE-Cy7 (clone PC61.5, Thermo Fisher Scientific). Then, cells were fixed and permeabilized in eBioscience Foxp3/Transcription Factor Staining Buffer (Thermo Fisher Scientific). To complete the ‘T-cell activation/exhaustion ’ panel, an intranuclear staining was performed with anti-FoxP3 FITC (clone FJK-16s, Thermo Fisher Scientific). 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: weight of the harvested tumor and total volume of the dissociated tumor cell suspension (cell concentration typically set to 250 mg/mL in PBS), proportion of the whole cell suspension and proportion of the cell suspension used for cytometry. (Representative gating strategies are depicted in online supplementary figure S5 and S6.)

Data are reported as means±SD of triplicate determinations and experiments were repeated at least twice, yielding similar results if not otherwise reported. Statistical significance was assessed by Student’s t-test. TumGrowth and GraphPad were used to analyze in vivo data arising from murine models.21

To identify bona fide autophagy inducers, we screened a total of 1560 compounds contained in the Prestwick library, as well as home-made libraries containing flavonoids18 20 and a range of saturated and unsaturated fatty acids22 for their capacity to induce the formation of autophagic puncta in human H4 neuroblastoma cells that express a green fluorescent protein (GFP) fused to microtubule-associated proteins 1A/1B light chain 3B (best known as LC3) (figure 1A,B; online supplementary table S1). The 175 compounds that were most efficient at one of the chosen concentrations were re-tested on another cell line, namely, human U2OS osteosarcoma cells stably expressing the monomeric red fluorescent protein (mRFP)-GFP-LC3 tandem reporter23 for their capacity to induce autophagic flux (meaning that they increase the ratio of RFP-positive over GFP-positive puncta) (figure 1C,D; online supplementary table S2). The 34 most effective candidate compounds (online supplementary table 1) were finally investigated for their capacity to reduce intracellular ATP concentrations measured with the fluorochrome quinacrine in combination with the prototypic ICD inducers mitoxantrone, using 3,4-dimethoxychalcone20 as a positive control (figure 1E,F; online supplementary table S3). This approach led to the identification of thiostrepton, an antibiotic produced by Streptomyces azureus and Streptomyces laurentii,24–26 as a potential ICD enhancer.

When used at micromolar concentrations for 6 hours, thiostrepton had little or no cytotoxic activity on cultured cells and reduced cytoplasmic protein acetylation, as measured by immunofluorescence staining, as it induced cytoplasmic GFP-LC3 puncta in human and murine biosensor cells (figure 2A–D, online supplementary figure S1A,B). In the presence of bafilomycin A1, thiostrepton continued to increase the number of cytoplasmic GFP-LC3 puncta per cell (figure 2E,F, online supplementary figure S1A,B). Moreover, the autophagy-associated lipidation of LC3 was enhanced by thiostrepton, more so in the presence of bafilomycin A1 (figure 2G,H, online supplementary figure S1C,D), indicating that thiostrepton truly induced autophagy. Accordingly, thiostrepton acted on another biosensor that measures autophagic flux, namely, a mCherry-GFP-sequestosome-1 (SQSTM1, best known as p62) to preferentially reduce the GFP-dependent fluorescence, indicating an increase in autophagy (figure 2I,J).27 Moreover, thiostrepton stimulated the degradation of a Q74-huntingtin-GFP fusion protein indicative of an increase of autophagic flux28 (figure 2K,L). When systemically injected into mice, thiostrepton also caused increased LC3 lipidation in the liver, suggesting that it is active in vivo (figure 2M–P). Of note, thiostrepton lost its capacity to induce GFP-LC3 puncta in U2OS cells that lack the essential autophagy gene ATG5 or the two proautophagic transcription factors TFEB and TFE3 (figure 3A–D). Indeed, thiostrepton induced the translocation of TFEB and TFE3 from the cytoplasm to the nucleus, as did the positive control torin (figure 3E,F). Moreover, thiostrepton induced the phosphorylation of eIF2α, an event that is often associated with autophagy.29 However, thiostrepton did inhibit transcription and translation in human cells cultured in vitro (see online supplementary figure S2A–D), indicating that it can interfere with protein synthesis in eukaryotes as it does in bacteria.30 Thiostrepton did not lead to an overproduction of reactive oxygen species (see online supplementary figure S2E,F), yet had (limited) cytotoxic effects when cells were exposed to the drug for an extended amount of time (24 hours) (see online supplementary figure S2G–J). This effect was further increased when thiostrepton was combined with low-dose oxaliplatin. Altogether, these results confirm that thiostrepton is a potent autophagy-inducing and cell stress-inducing agent.

Mitoxantrone (MTX) and oxaliplatin (OXA) are potent ICD inducers.31 As such, they elicit the ICD hallmarks in vitro, in cultured cancer cells, including the autophagy-dependent ATP release into the supernatant, calreticulin (CALR) exposure on the cell surface, as well as nuclear exodus of HMGB1.32–34 We determined the capacity of thiostrepton to enhance these stigmata of ICD in a series of in vitro assays. The importance of thiostrepton-induced ATP release needs to be further validated in vivo. Nevertheless, it appears clear that thiostrepton reduced the abundance of quinacrine-detectable ATP in the cytoplasm of cultured cells, more so if combined with low concentrations of MTX or OXA, suggesting additive interactions (figure 4A,B). Accordingly, ATP concentrations in the supernatant increased with the thiostrepton/MTX or thiostrepton/OXA concentration, and this ATP release was inhibited in cells in which ATG5 or both TFEB and TFE3 had been knocked down (figure 4C–I) even after extended incubation for 24 hours (see online supplementary figure S3A,B). Thiostrepton was able to cause the relocation of an RFP-CALR fusion protein to the periphery of the cells either alone or in the presence of MTX or OXA (figure 5A,B), and this effect did not require TFEB plus TFE3 (see online supplementary figure S4A). The capacity of thiostrepton to cause CALR exposure was validated by surface immunofluorescence staining (figure 5C,D). Finally, protracted incubation (48–72 hours) with thiostrepton could cause HMGB1 release, more so when it was combined with low-dose MTX or OXA (figure 5E–G), which was reduced when TFEB plus TFE3 was knocked down (see online supplementary figure S4B). Hence, thiostrepton can amplify ICD surrogate markers in the context of low-dose MTX or OXA.

Although thiostrepton alone had no tumor growth reducing effect against TC1 non-small cell lung cancers or MCA205 fibrosarcomas established in immunocompetent C57/BL6 mice, it did increase the anticancer activity of the ICD inducer OXA (but not cis-diamminedichloridoplatinum(II), abbreviated CDDP, best known as cisplatin, which is unable to trigger ICD see online supplementary figure S5),7 further diminishing tumor size and extending the survival of the animals (figure 6A,B). This thiostrepton/OXA combination effect depended on T lymphocytes because tumor developing in athymic nu/nu mice failed to reduce their growth in response to the treatment (figure 6C). Moreover, tumors formed by cancer cells lacking TFEB and TFE3 expression that were implanted in immunocompetent C57/BL6 mice, failed to exhibit the combination effect as well (although they responded to OXA) (figure 6D), supporting the idea that TFEB/TFE3-dependent autophagy induction is critical for the anticancer action of thiostrepton. Analysis of the leukocyte infiltrate by immunophenotyping and cytofluorimetry (see online supplementary figure S6) revealed that thiostrepton had no significant effect on the abundance of CD8⁺ CTLs (figure 7A) but reduced the number of CD4⁺ CD25⁺ Foxp3 regulatory T cells (Tregs) if combined with OXA (figure 7B), thus improving the CTL/Treg ratio (figure 7C). Moreover, the combination treatment was particularly efficient in increasing the number of tumor infiltrating neutrophils (see online supplementary figure S7 and S8). In conclusion, thiostrepton can synergize with OXA to improve tumor control in an immune-dependent fashion, correlating with reprogramming of the tumor microenvironment.

Thiostrepton is a fully synthetic oligothiopeptide antibiotic that has been approved for veterinary but not human use, for the treatment of superficial infections including dermatitis and mastitis.35 Preclinical experiment suggest that it may inhibit human breast cancer growth and overcome cisplatin resistance.36 Here, we identified thiostrepton in an unbiased screen designed to identify autophagy inducers, and subsequently discovered that this agent can enhance the stigmata of ICD in cultured cells, thus enhancing ATP release, CALR exposure and HMGB1 exodus. Although thiostrepton had some activities on these ICD hallmarks when used alone, it was particularly active when combined with low doses of established ICD inducers such as MTX and OXA. Moreover, thiostrepton enhanced the anticancer activity of OXA in vivo. This adjuvant effect required the expression of proautophagic transcription factors by the cancer cells and was fully lost in the absence of T lymphocytes. Indeed, thiostrepton combined with OXA stimulated a major change in the tumor immune infiltrate, enhancing the density of CTL, but reducing that of Tregs and G-MDSC. Thiostrepton is an antibiotic,24 and the intestinal microbiota has been shown to contribute to anticancer immune responses elicited by some immunogenic chemotherapies37 38 or immunotherapy, meaning that broad-spectrum antibiotics causing sterilization of the intestinal tract reduce the efficacy of anticancer treatments.39 40 For this reason, it appears unlikely that the immune-dependent anticancer effects of thiostrepton would be related to its antibiotic activity. Thus, thiostrepton could be repurposed as an enhancer of immunogenic chemotherapies, independent of its antibiotic activity.