Microbiota-derived 3-IAA influences chemotherapy efficacy in pancreatic cancer – Nature

Patients and human material

Patients who had been diagnosed with mPDAC and scheduled for treatment with GnP or FOLFIRINOX were recruited to the study (n = 30). Informed consent was obtained from all patients as approved by the ethics commission Hamburg (Ethikkommission der Ärztekammer Hamburg, Germany). Patients who received antibiotics during the first three months of treatment, did not deliver pre-treatment stool, did not start chemotherapy or suffered from COVID-19 were excluded from the analyses (n = 7). Stool was collected before the start of chemotherapy using a home sampling kit (OMNIgene, Gut OMR-200) for shotgun metagenomic sequencing. One sample was excluded after sequencing owing to overloading of the sample tube that led to unsuccessful fixation and limited representation of the microbiota. Therefore, the final cohort for microbiota analysis contained 22 patients. The classification of R and NR patients was primarily based on tumour shrinkage of at least 25%, which is calculated as a decrease of the primary tumour and largest metastasis in CT scans comparing the time point before treatment initiation to the time point that revealed the best response during first-line treatment as assessed by the local radiologist (20 to 30% of changes are usually considered significant in CT response evaluation³⁵). In cases of disease stabilization or missing CT scans, the criteria of a PFS of over 140 days (median PFS of real-world FOLFIRINOX and GnP cohorts³⁶) or a decline of at least 40% in serum tumour markers (prognostic cut-off in palliative PDAC treatment³⁷) were considered. Stool for functional experiments was collected in tubes without any additive, transported directly to the laboratory, subsequently diluted in 20% glycerol (Teknova, G1723), aliquoted and frozen at −80 °C. To preserve most of the donor microbiota for functional experiments, the maximum allowed turnaround time from toilet to freezer was 4 h. Blood was drawn at cycle three or four of chemotherapy treatment, mixed 1 to 1 with phosphate-buffered saline (PBS) and plasma was isolated using gradient centrifugation. Serum material of the Munich cohort was taken before the start of chemotherapy and processed according to local standards. Informed consent was obtained from all patients as approved by the ethics commission Munich (project number: 284-10).

Animal models

All mice used in this study were of a C57BL/6 background. All mice were used in accordance with the institutional review board ‘Behörde für Soziales, Familie, Gesundheit und Verbraucherschutz’ (Hamburg, Germany). Mice were kept under SPF or germ-free conditions, with an ambient temperature of 20 ± 2 °C, humidity of 55 ± 10% and a dark–light cycle of 12 h. Age- and sex-matched littermates between 4 and 16 weeks old were used for the most part. Mpo^–/– bone marrow used to establish bone marrow chimeras was provided by S. Baldus and M. Mollenhauer. Ahr^–/– bone marrow used to establish bone marrow chimeras was provided by C. Esser.

For colonization of human microbiota, stool was thawed, washed with brain heart infusion broth (BHI) and diluted in BHI. Two hundred microlitres of the suspension was gavaged once orally to gnotobiotic mice housed in isocages. Two to four weeks later, tumour experiments were initiated.

Sample sizes were calculated on the basis of small pilot experiments and mice were randomized before the beginning of treatment. The person treating the mice was not blinded because of the complex treatment schedules. For models of orthotopic PDAC, 5 × 10⁴–10 × 10⁴ KPC, 2 × 10⁵ Hy19636, 1 × 10⁵ mSt-ATG4B/mSt or 1 × 10⁵ knockdown KPC (AhR, GPX, scramble control) cells were injected orthotopically in a 1:1 mixture of PBS and Matrigel (Corning, 356231). Tumour weight was assessed at day 20 after tumour cell injection unless otherwise indicated. For subcutaneous tumour growth, 2.5 × 10⁵ LLC-GFP or 1 × 10⁶ MC38 cells were subcutaneously injected into the flank. Subcutaneous tumour growth was measured every other day using a caliper. Tumour weight was assessed at day 17 after tumour cell injection. The maximum allowed diameter of subcutaneous tumours for active experimental mice was 1.5 cm and limits were not exceeded.

For bone marrow chimeras, recipient mice were irradiated with 9.6 Gy (BioBEAM 2000) 24 h before bone marrow transfer. One day later, 1 × 10⁶–4 × 10⁶ bone marrow cells isolated from wild-type, Ahr^–/– or Mpo^–/– mice were transferred by intravenous injection. Three to four weeks after transfer, cancer cells were orthotopically injected and treated as described below. Successful engraftment was validated by quantitative PCR (qPCR) on tumour-infiltrating immune cells.

Treatment with chemotherapy was initiated intraperitoneally (i.p.) at the indicated time point (usually day 11 after cancer cell injection). Oxaliplatin (Accord) 5 mg kg⁻¹, irinotecan (Accord) 20 mg kg⁻¹ and 5-FU (Medac) 50 mg kg⁻¹ were used for FIRINOX treatment as described in other studies^38,39. Folinic acid was not used and gemcitabine (Hexal) 120 mg kg⁻¹ and nab-paclitaxel (Celgene) 30 mg kg⁻¹ were used for GnP treatment. Depletion of T cells was achieved by treatment with 200 μg anti-CD8 antibody clone 53-6.7 (BioXcell, BE0004) alone or in combination with 200 μg anti-CD4 antibody clone GK1.5 (BioXcell, BE0003) every third day i.p., beginning one day before chemotherapy treatment. Control mice were treated with the respective isotype control clone 2A3 (BioXcell, BE0089) at similar concentrations and intervals. Autophagy induction was achieved using two i.p. injections of trehalose (Sigma, T9449) at a concentration of 3 g kg⁻¹ one day before and on the day of chemotherapy treatment. Similarly, 60 mg kg⁻¹ chloroquine (Sigma, C6628) dissolved in PBS was injected i.p. one day before and on the day of chemotherapy treatment.

Treatment with 3-IAA (500 mg kg⁻¹) was applied by oral gavage using indole-acetic-acid sodium salt (Sigma, I5148) dissolved in PBS every day for five consecutive days (two days before chemotherapy until two days after chemotherapy), unless otherwise indicated. Indole-3-propionic acid (3-IPA; Sigma, 57400), GCA (Sigma, G2878), hippuric acid (Sigma, 112003) and DCA (Sigma, 30960) were dissolved in 1 M NaOH in PBS and pH-adjusted to 7.4 using 1 M HCl. The solution was gavaged for five consecutive days orally at a concentration of 500 mg kg⁻¹ (3-IPA, hippuric acid) or 250 mg kg⁻¹ each (GCA and DCA). NAC (Sigma, a7250) was applied for five days in the drinking water ad libitum at a concentration of 1 g l⁻¹.

Dietary tryptophan modulation was initiated three days before chemotherapy until one day after treatment. In one experiment, tryptophan modulation was applied for a total of 14 days. Standard diet (2.3 g kg⁻¹ tryptophan; Altromin, 1320) was changed to either synthetic crystalline AA tryptophan-free (0 g kg⁻¹; SSNIFF, S9336-E701) diet or crystalline AA tryptophan-high (12 g kg⁻¹; SSNIFF, S5714-E711) diet. Subsequently, the diet was changed back to the standard diet.

Dietary doxycycline (Sigma, D9891) was administered at a dosage of 625 mg kg⁻¹ through the diet starting at day five or eight after cancer cell injection for a total of up to seven consecutive days. Subsequently, the diet was changed back to the standard diet.

For measurement of serum metabolites, blood was drawn at the end of the experiment or at the indicated time point. Blood was allowed to clot for 30 min and was centrifugated (1,000g) for 10 min thereafter. Serum was diluted and used as described in the specific section.

DNA extraction and shotgun metagenomics

For DNA extraction, samples were isolated with the ZymoBIOMICS 96 MagBead DNA kit (Zymo Research, D4302) and purified with Zymo DNA Clean and Concentrator-5 (Zymo Research, D4004). Libraries were prepared with the NEBnext Ultra II DNA Library Prep Kit for Illumina (New England Biolabs, E7645) with 150 ng of total DNA, size selection of 400–500 bp and 4× PCR cycles.

For shotgun metagenomics, Illumina library preparation was performed using the NEBNext Ultra II FS DNA Library Prep Kit (New England Biolabs, E7805). The library preparation was performed according to the manufacturer’s instructions. The size selection was performed using AMPure XP beads (Beckman Coulter, A63882) and adaptor enrichment was performed using seven cycles of PCR using the NEBNext Multiplex oligos (New England Biolabs, E7335) from Illumina and then subjected to Illumina NovaSeq (2 × 150 bp) sequencing.

For bioinformatic analysis, raw reads were trimmed for low quality and filtered against the phix174 and human hg19 genome with bbduk (ref. ⁴⁰). For taxonomic species profiling, all libraries were mapped against the Unified Human Gastrointestinal Genome collection (n = 4,644) using BBMap (refs. ^41,42). Mapping rates were normalized into transcripts per million (TPM) and genomes with less than 10% genome coverage (genome breadth) were considered not prevalent in the sample. Data were summarized as metagenomics operational taxonomic units (OTUs) into biom format and analysed with phyloseq and LEfSe (refs. ^43,44). Species-level functional profiling was performed with HUMAnN3 and also using the 9.9 million gene integrated reference catalogue of the human microbiome⁴¹.

Bacterial strains and isolation

Bacteroides thetaiotaomicron (DSM 2079), B. fragilis (DSM 2151) and Prevotella copri (DSM 18205) were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ). For bacteria isolation, faecal samples frozen in glycerol were thawed and streaked out anaerobically in serial dilutions on BHI blood agar plates (5% defibrinated sheep’s blood & vitamin K₃) supplemented with vancomycin. After growing the agar plates inside an incubator at 37 °C for 2 days, single colonies were picked into BHI-S medium in a 96-well plate and were further incubated for 24 h. The resulting bacterial cultures were screened by PCR using specific primers for B. thetaiotaomicron (B. theta F 5′-GAGGGTGTCGTATTTCCGAAGG-3′ R 5′-GTTCCCTGATCCAGTGTGTTGG-3′) or B. fragilis (B. frag F 5′-AATGATTCCGCATGGTTTCA-3′ R 5′-ATTTTGGGATTAGCATACGG-3′). After passaging PCR-positive wells on agar plates to obtain pure cultures and additional confirmation of identity by Sanger sequencing, resulting strains were maintained in BHI-S until further use. All bacterial strains were cultured and maintained anaerobically in BHI broth supplemented with 10% fetal bovine serum (FBS) and vitamin K₃ (BHI-S).

Extraction of bacterial supernatant

BHI-S medium supplemented with 1% tryptophan was inoculated from a fully grown overnight bacterial culture (1:50 ratio) and was incubated anaerobically until the early exponential phase. Cultures were taken out of the anaerobic chamber and were centrifuged at 4,700 rpm (4,816g) for 5 min at room temperature. The supernatant was removed and immediately frozen at −20 °C.

16S rRNA sequencing

DNA from tumours was extracted using the DNeasy Blood & Tissue Kit (Qiagen, 69504). Approximately 10 mg tissue was digested with Proteinase K in ATL buffer at 56 °C for 1 h. Afterwards, samples were processed according to the manufacturer’s protocol. Blank extraction controls were included during extraction of samples.

Variable regions V1 and V2 of the 16S rRNA gene were amplified using the primer pair 27F-338R in a dual-barcoding approach according to a previous report⁴⁵. For tumours, 3.5 µl DNA was used for amplification and PCR products were verified using agarose gel electrophoresis. Final PCR products were normalized using the SequalPrep Normalization Plate Kit (Thermo Fisher Scientific, A1051001), pooled equimolarly and sequenced on the Illumina MiSeq v3 2×300bp (Illumina). Demultiplexing after sequencing was based on 0 mismatches in the barcode sequences. Data processing was performed using the DADA2⁴⁶ workflow for big datasets (https://benjjneb.github.io/dada2/bigdata.html; the workflow adjusted for the V1–V2 region can be found here: https://github.com/mruehlemann/ikmb_amplicon_processing/blob/master/dada2_16S_workflow.R), resulting in abundance tables of amplicon sequence variants (ASVs). ASVs underwent taxonomic annotation using the Bayesian classifier provided in DADA2 and using the Ribosomal Database Project (RDP) version 16 release. Sequences that were not assignable to genus level were binned into the finest possible taxonomic classification.

Metabolomic screen

We used ultra-high-pressure liquid chromatography–tandem mass spectrometry (UPLC–MS/MS) to acquire data in both positive and negative ionization modes that allowed for identification and quantification of 630 metabolites. Plasma samples were processed using the MxP Quant 500 Kit (Biocrates) according to the manufacturer’s instructions. In brief, 10 µl of plasma sample, calibration standard and control sample were transferred onto a filter containing internal standards for internal standard calibration. Filters were dried under a stream of nitrogen using a pressure manifold (Waters). Samples were incubated with derivatization reagent phenyl isocyanate for 60 min. After drying under nitrogen, analytes were extracted with 5 mmol l⁻¹ ammonium acetate in methanol and the eluate was further diluted for the UPLC–MS/MS analysis. The targeted analysis covered 630 metabolites (https://biocrates.com/mxp-quant-500-kit/) detected by MS/MS after UPLC separation and flow injection analysis (FIA). Each measurement required two UPLC runs and three FIA runs to cover all metabolites. All analyses were performed on an ACQUITY UPLC I-Class system (Waters) coupled to a Xevo TQ-S mass spectrometer (Waters). Reversed-phase chromatographic separation was accomplished using a C18 LC-column (Biocrates) with 0.2% formic acid in water with 0.2% formic acid in acetonitrile as the eluent system. The FIA solvent was methanol, with a modifier provided by the kit manufacturer. Data analysis of the UPLC–MS/MS results was based on a seven-point curve or one-point calibration and internal standard normalization. Values below the lower threshold were set to zero. Concentration data were analysed using MetaboAnalyst v.5. Concentrations were log-transformed before analysis and raw P values and log₂-transformed fold change values are shown in the graphics of Fig. 1.

3-IAA and DCA CLIA

To quantify 3-IAA or DCA serum concentrations, mouse or human serum was diluted 1 to 10 with PBS and the chemiluminescence immune assay (CLIA) (Abbexa, 3-IAA abx190011; DCA 258844) was performed according to the manufacturer’s protocol. For detection of 3-IAA in cultures, supernatant was processed as described above and directly used for the assay. Chemiluminescence was detected using FLUOstar Omega (BMG Labtech) with 1-s sampling per well, and a gain of 3,400 to 4,000 was individually adjusted for each assay. The concentration was determined using supplied standards and interpolated using Prism 8.4.0.

3-IAA and MOI measurement in tumour tissue

Frozen tumour tissue was mixed with ice-cold methanol 1:1–3 and homogenized using 0.4-mm beads at 5,500 rpm for 2 × 30 s using a homogenizer (Precyllys 24 touch). Extracts were centrifuged (10,000g) at 4 °C and the supernatant was used for LC–MS/MS analysis. Reversed-phase chromatography was conducted using a biphenyl stationary phase (Raptor Biphenyl (Restek), dimensions: 50 mm × 2.1 mm ID; particle size: 2.7 µm) with eluent A: water + 0.1% formic acid + 5 mM ammonium acetate and eluent B: methanol + 0.1% formic acid + 5 mM ammonium acetate. The flow rate was set to 0.5 ml per min. Elution starts with 95% eluent A, which linearly decreases to 75% over 0.5 min. This composition was held for 4 min before returning to initial conditions. The injected volume was 2 µl and the column temperature was set to 55 °C. 3-IAA and MOI were detected in MRM (multiple reaction monitoring) mode. The following transitions were monitored after positive electrospray ionization: 3-IAA: m/z 176.2 > 103.0; m/z 176.2 > 130.2; MOI: m/z 148.1 > 120.2; m/z 148.1 > 130.2; m/z 148.1 > 133.1. Quantification was done according to a standard curve.

Isolation of immune cells and flow cytometry

Tumours were taken and cut into similar-sized pieces. Tumours were rinsed with cold PBS and digested in RPMI (Sigma, 61870044) supplemented with 10% FBS (Gibco, 10500064), 2.5 mg ml⁻¹ collagenase D (Roche, 11088866001) and 0.2 mg ml⁻¹ DNase l (Roche, 11284932001) for 35 min at 37 °C with continuous shaking. Afterwards, the suspension was strained through a 40-µm cell strainer and quenched with cold PBS. Subsequently, immune or tumour cells were stained with Fc block and live/dead staining (Thermo Fisher Scientific L34957 and L10119) for 30 min in the dark. Afterwards cells were washed, stained with the indicated flow cytometry antibodies and incubated for 30 min in the dark again. Flow cytometry was performed on a Fortessa flow cytometer (BD). To assess the cytokine profile of immune cells, restimulation of T cells with 50 ng ml⁻¹ PMA, 500 ng ml⁻¹ ionomycin and 1 µg ml⁻¹ brefeldin A was performed for 3 h at 37 °C. After surface staining, cells were fixed and permeabilized using the eBioscience Foxp3 intracellular staining kit (00-5523-00). The following intracellular antibodies were used: IFNγ (1:200 dilution) and TNF (1:400 dilution). The following surface antibodies were used to classify lymphocytes (CD3 (1:300 dilution), CD4 (1:500 dilution), CD8 (1:400 dilution), CD19 (1:100 dilution), CD44 (1:500 dilution), PD-1 (1:400 dilution) and NK1.1 (1:400 dilution)) or myeloid cells (CD11b (1:800 dilution), CD11c (1:300 dilution), CD45 (1:800 dilution), Ly6G (1:800 dilution), Ly6C (1:600 dilution), Ly6B (1:200 dilution), CD115 (1:300 dilution), F4/80 (1:600 dilution) and MHCll (1:600 dilution)). Epithelial cancer cells were stained with EPCAM (1:200 dilution) antibody. For Ly6B staining, the goat anti-rat IgG antibody (1:400 dilution) was used. Cells were counted by flow cytometry using 2,500 to 5,000 beads (Spherotec, ACBP 100-10). Viability of neutrophils in vitro was assessed using PI/Annexin V staining according to the manufacturer’s protocol. ROS expression was determined in immune cells in vivo after tissue digestion as described above. Afterwards, the ROS dye CellROX (Thermo Fisher Scientific, C10422) was stained in parallel with flow cytometry antibodies at a dilution of 1:1,000. Software analysis and histogram generation was carried out using FlowJo v.10.

Cell culture

KPC cells were obtained from Ximbio under catalogue number 153474; Hy19636_GLRM reporter cells and mSt-ATG4B/mSt cells were provided by A. Kimmelman³¹; MC38 and LLC-GFP (ATCC) cells were provided by A. Giannou; and MIA PaCa-2, BxPC-3 and T3M-4 cells (all ATCC) were provided by C. Güngör. All cells tested negative for mycoplasma contamination. Cells were maintained under standard conditions at 37 °C and 5% CO₂ and regularly visually inspected. Cells were grown in DMEM GlutaMAX (Thermo Fisher Scientific, 10566016) supplemented with penicillin and streptomycin (Gibco, 15140122) and 10% FBS (Gibco, 10500064). For in vitro experiments, cancer cells were plated at 5,000–10,000 cells per well in a 96-well plate. Cells were allowed to seed overnight, except in experiments with neutrophil co-culture, treatment of GPX-knockdown cells or co-treatment with MPO. Subsequently, treatment with indicated compounds and treatment duration was initiated. 3-IAA or 3-IPA (3-IPA, Sigma, 57400; 3-IAA, Sigma, I3750) were dissolved in 1 M NaOH in PBS and PH-adjusted to 7.4 using 1 M HCl or DMSO when indicated. NAC was diluted in PBS and used at a concentration of 1 mM. H₂O₂ (Sigma, H1009) was used at a concentration of 400 µM.

Tumour cell viability and proliferation was assessed using a MTT or MTS assay (Abcam, ab201191 and ab197010) according to the manufacturer’s protocol. Absorbance was assessed using the FLUOstar Omega (BMG Labtech). In other experiments, viability was assessed by flow cytometry using SYTOX (Thermo Fisher Scientific, S34857), PI (Biolegend, 421301) and 2,500–5,000 counting beads (Spherotec, ACBP 100-10) as a reference. In parallel, intracellular ROS was measured using CellROX (Thermo Fisher Scientific, C10422) according to the manufacturer’s protocol and assessed by flow cytometry.

Neutrophils or T cells were isolated from bone marrow or spleens of healthy, untreated mice and sorted using fluorescence-activated cell sorting (FACS) as LY6G⁺CD11b⁺ and TCRβ⁺CD4⁺ or CD8⁺, respectively. Cells were seeded at a concentration of 20,000–50,000 per well in a 96-well plate. Viability of neutrophils or T cells was assessed by flow cytometry using SYTOX (Thermo Fisher Scientific, S34857) or PI/Annexin V staining (Biolegend, 640914) at the indicated time of the experiment. In some experiments, 200 mU ml⁻¹ MPO or 400 mU ml⁻¹ MPO (Merck, 475911) or indicated concentrations of MOI (Sigma, 493397) were added. Degranulation of neutrophils was assessed by the release of MPO. A total of 1 × 10⁶ neutrophils were sorted using FACS as described above. Neutrophils were incubated in HBSS (Gibco, 14065-56) and the indicated treatment, or N-formylmethionyl-leucyl-phenylalanine (fMLP; Merck, F3506), as a positive control, was added. After 30 min of incubation, MPO activity was measured using an MPO activity assay kit (Abcam, ab105136) according to the manufacturer’s protocol. NET formation of neutrophils was determined using SYTOX stain after 3 h of incubation with the indicated compounds or 100 nM phorbol-12-myristate-13-acetate (PMA; Merck, 524400) as a positive control. NETs were measured as SYTOX-positive cells using flow cytometry.

Pre-neutrophils were sorted by FACS (lineage-negative (CD3,NK1.1,CD19,B220)⁻, CD115⁻, Ly6B⁺, Ly6G^int-low) as previously described²⁶ and cultured at a density of 20,000–50,000 per well in a 96-well plate. Indicated treatments were applied and viability was assessed using PI staining in flow cytometry.

MPO activity

Intratumoural or bone-marrow-derived neutrophils (LY6G⁺CD11b⁺) or pre-neutrophils (lineage-negative (CD3,NK1.1,CD19,B220)⁻, CD115⁻, Ly6B⁺, Ly6G^int-low) were sorted by FACS. A total of 50,000 cells were processed for MPO activity measurement using an MPO activity assay kit according to the manufacturer’s protocol (Abcam, ab219925). Fluorescence was analysed using FLUOstar Omega (BMG Labtech). MPO activity was calculated based on a standard curve as suggested by the manufacturer’s protocol.

Protein screen

Tumour tissue from three individual mice per group was pooled and proteins were extracted using scioExtract buffer (Sciomics). The samples were labelled for 2 h with scioDye 2 (Sciomics) at an adjusted protein concentration. The reference sample was labelled with scioDye 1 (Sciomics). After 2 h, the reaction was stopped and the buffer was exchanged with PBS. All labelled protein samples were stored at −20 °C until use. The samples were analysed in a dual-colour approach using a reference-based design on scioDiscover antibody microarrays (Sciomics). The arrays were blocked with scioBlock (Sciomics) on a Hybstation 4800 (Tecan) and afterwards the samples were incubated competitively with the reference sample using a dual-colour approach. After incubation for 3 h, the slides were thoroughly washed with 1× PBSTT, rinsed with 0.1× PBS as well as with water and subsequently dried with nitrogen. Slide scanning was conducted using a Powerscanner (Tecan) with constant instrument laser power and PMT settings. Spot segmentation was performed with GenePix Pro 6.0 (Molecular Devices). Acquired raw data were analysed using the (LIMMA) package of R-Bioconductor after uploading the median signal intensities. For normalization, a specialized invariant Lowess method was applied. Downregulated proteins (M-value of < −0.35) or upregulated proteins (>0.35) in the 3-IAA and FIRINOX sample were uploaded to the STING database (https://string-db.org/cgi/input.pl). The standard pipeline was used to perform KEGG enrichment analysis. Only pathways with a false discovery rate (FDR) < 0.05 were considered statistically significant.

RNA extraction and mRNA sequencing

Fresh frozen tumour tissue was lysed in TRIzol reagent (Thermo Fisher Scientific, 15596018) and RNA was extracted using the chloroform-isopropanol method. mRNA was purified from total RNA using poly-T oligo-attached magnetic beads. After fragmentation, the first-strand cDNA was synthesized using random hexamer primers followed by second-strand cDNA synthesis. The library was ready after end repair, A-tailing, adapter ligation, size selection, amplification and purification. The library was checked with Qubit and qPCR for quantification, and Bioanalyzer for size distribution detection. Quantified libraries were sequenced on the Illumina platform with at least 60 million reads per sample.

Sequence reads were processed with fastp (v0.20.1) to remove sequences of sequencing adapters and low-quality (Phred quality score below 15) sequences from the 3′ end of the sequence reads. Afterwards, reads were aligned to the mouse reference assembly (GRCm39.104) using STAR (v.2.7.9a)⁴⁷. Differential expression was assessed with DESeq2 (ref. ⁴⁸). A gene was considered significantly differentially expressed if the corresponding absolute log₂-transformed fold change (log2FC) was no less than 0.6 and, in addition, the FDR did not exceed a value of 0.1. The gene set enrichment analysis (GSEA) of the Reactome pathway Autophagy (R-HSA-9612973) was performed using fgsea (v.4.1)⁴⁹.

qPCR

Total RNA was extracted from cell lines using Trizol Reagent (Invitrogen,15596018) and the total RNA extraction kit (Qiagen, 74004/74104) according to the manufacturer’s protocol. The High-Capacity cDNA Synthesis Kit (Thermo Fisher Scientific, 4368813) was used for cDNA synthesis. Primers and probes were purchased from Applied Biosystems. Mouse primers and probes are: Gpx3 (Mm00492427_m1), Gpx7 (Mm00481133), Ahr (Mm00478932_m1). qPCR was performed using the TaqMan Master Mix (Thermo Fisher Scientific, 4369016) on the StepOne Plus system (Applied Biosystems). Forty to 44 cycles were applied for every assay and technical doublets or triplicates were used. If at least two out of three values of the technical replicates were undetectable, the expression was considered non-detectable. Relative expression was normalized to Gapdh (Mm99999915_g1).

Lentiviral transfer of shRNAs

Lentiviral vectors expressing short hairpin RNAs (shRNAs) under control of the human U6 promoter (MISSION pLKO.1-puro) directed against mouse Gpx3 (TRCN000076539), mouse Gpx7 (TRCN0000076563) and mouse Ahr (TRCN0000218025) as well as a nontargeted control shRNA (SHC002, scrambled) were obtained from Sigma-Aldrich. The production of lentiviral particles has been described in detail elsewhere⁵⁰, and protocols are available online (http://www.LentiGO-Vectors.de). For the transduction of KPC cells with the HIV-1-derived lentiviral vectors, 2.5 × 10⁴ cells were plated in 0.5 ml medium with 8 μg ml⁻¹ polybrene per well of a 24-well plate. After plating, the addition of 10 μl VSV-G pseudotyped, non-concentrated lentiviral particles led to the stable integration of shRNAs and the puromycin resistance gene into the cell’s genome. To increase the transduction rate by spin-inoculation, the plate was centrifuged at 1,000g and 25 °C for 1 h. The selection of successfully transduced cells with 1 μg ml⁻¹ puromycin in the culture medium was started four days after transduction.

Immunohistochemistry and quantification

Tissues were fixed in 4% formalin in PBS and embedded in paraffin using an ASP300S dehydration machine (Leica) and an EG1160 tissue embedding system (Leica). Paraffin sections (2 µm) were cut and stained with H&E or processed for immunohistochemistry as follows: after dewaxing and inactivation of endogenous peroxidases (3% hydrogen peroxide in PBS), antibody-specific heat-mediated antigen retrieval was performed using the Ventana Benchmark XT machine (Ventana). Sections were blocked (10% FCS in PBS) and then incubated with anti-LC3B antibody (1:400 dilution, Thermo Fisher Scientific, PA1-46286); anti-nitrotyrosine antibody (1:100 dilution, Thermo Fisher Scientific, A-21285); anti-CC3 antibody (1:100 dilution, Cell Signaling, 9661); p62/SQSTM1 (1:500 dilution, Thermo Fisher Scientific, PA5-20839); and anti-Ki67 antibody (1:100 dilution, Abcam, 15580). For detection of specific binding, the Ultra View Universal DAB Detection Kit (Ventana, Roche) was used, which contains secondary antibodies, DAB stain and haematoxylin counterstaining reagent. Slides were scanned using NanoZoomer 2.0-HT (Hamamatsu Photonics) and representative images were taken using Fiji.

Quantification for LC3B, Ki67, nitrotyrosine and CC3 was done in a blinded manner. Positive cells were determined using ImageJ v.2.1.0/1.53c and the threshold was adjusted according to the staining intensity of the respective antibody and maintained for all tumours analysed with the same staining. The number of positive cells was counted per 250–500 × 500-μm field (10× to 40× magnification) in three to five fields per sample.

Widefield microscopy

To visualize the GFP-LC3-RFP reporter signal expression of Hy19636_GLRM PDAC cells, tissues were fixed in 2% PFA solution at 4 °C overnight, incubated in PBS containing 30% sucrose and embedded in Tissue-Tek OCT compound (Sakura Finetek) on dry ice. For further analysis, 7-μm sections were used. Widefield imaging was performed using the THUNDER Imager 3D Live Cell and 3D Cell Culture (Leica Microsystems) equipped with a 40× 1.10 NA water immersion objective. LED power and exposure time and other system-specific settings were first optimized using positive control tissue and not changed between image acquisitions of the different groups to provide optimal comparability. For each tissue section, an average of five regions of interest were randomly selected and imaged for subsequent quantitative analysis. ImageJ imaging software was used for file navigation, adjustment of colour balance and image analysis. GFP and RFP quantification was determined using ImageJ v.2.1.0/1.53c. The mean fluorescence intensity for each respective signal was determined per slide in at least five areas per tumour.

Graphical abstracts

Figure 1a, Extended Data Fig. 2a,d and small icons in Figs. 1–4 were created with BioRender.com.

Statistical analyses

All statistical analyses, unless otherwise indicated, were performed using GraphPad Prism 9.3.1. Normality and log-normality were tested using Shapiro–Wilk or Kolmogorov–Smirnov tests. If normality was not given, non-parametric testing was performed. Tests were performed two-sided if not otherwise indicated and resulting significant (P < 0.05) P values are shown.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.