Gut microbiota carcinogen metabolism causes distal tissue tumours – Nature

Animal experiments

Two-month-old C57BL/6J mice (Jackson Laboratory and Charles River) housed in the animal facility of the University of Split School of Medicine were used for all conventional mouse experiments. Two-month-old germ-free C57BL/6J mice housed in the animal facility of the European Molecular Biology Laboratory (EMBL) were used for all gnotobiotic mouse experiments. Animals were kept in individually ventilated cages with autoclaved bedding (Mucedola), with a 12-h light–dark cycle, controlled temperature (21–24 °C) and humidity (30%–70%), and ad libitum access to food (γ-irradiated, Ssniff-Spezialdiäten) and autoclaved water. Male mice were used in all experiments unless specified otherwise. All mouse experiments were approved by the local institutional animal care and use committee (The University of Split, School of Medicine, Animal Welfare Committee and EMBL Institutional Animal Care and Use Committee) and the national regulating authorities (Republic of Croatia Ministry of Agriculture, Veterinary and Food Safety Directorate; permit numbers 525-10/0255-14-4, 525-10/0255-15-5, 525-10/0543-21-8, 525-09/566-22-2 and 21-002_HD_MZ). All mice were randomly allocated into experimental groups; blinding was not carried out, except for the pathohistological analysis. No sample size calculation was carried out. The general status of the animals including behaviour, appearance and body weight was monitored. If signs of pain, suffering or weight loss of more than 20% were observed, humane end points were applied. Those limits were not exceeded in any of the experiments. For ex vivo BBN conversion experiments and 16S rRNA sequencing using conventional mice, contents from the three parts of the small intestine, caecum, large intestine and rectum were collected from five mice of each sex. One half of each sample was frozen at −80 °C in 20% glycerol, and the other half was frozen at −80 °C and used for 16S rRNA amplicon sequencing. For liquid chromatography-coupled mass spectrometry (LC–MS) quantification, tissues collected were directly frozen at −80 °C until further processing.

Chemicals

Chemicals used in this study are listed in Supplementary Table 49.

Long- and short-term nitrosamine treatments with conventional or germ-free mice

Male conventional mice were randomly assigned to the ABX–BBN or BBN group. The ABX–BBN group received 100 µg ml⁻¹ neomycin, 100 µg ml⁻¹ metronidazole, 50 µg ml⁻¹ streptomycin, 67.7 µg ml⁻¹ penicillin (all from Sigma-Aldrich) and 50 µg ml⁻¹ vancomycin (Pharma Swiss) in drinking water that was supplied fresh weekly. Following 2 months of treatment, the ABX mixture was changed to 1,000 µg ml⁻¹ streptomycin, 170 μg ml⁻¹ gentamicin (Krka), 125 μg ml⁻¹ ciprofloxacin (MCE) and 1,000 µg ml⁻¹ bacitracin (Sigma-Aldrich), to prevent development and overgrowth of resistant bacterial strains²³. The ABX–BBN group received ABX 2 weeks before starting the BBN treatment, and were kept on ABX for the remainder of the experiment.

For gnotobiotic mouse experiments, male germ-free mice were inoculated by oral gavage with 100–250 µl microbial community culture or single bacterial isolate culture for 4 days. The ABX–BBN group would receive the first ABX mixture for 2 weeks before starting BBN treatment for 10 days. During BBN treatment, the second ABX mixture was used for the ABX–BBN group.

Monitoring of the impact of ABX treatment on the gut microbiota was carried out using a combination of colony-forming unit (CFU) counting and 16S rRNA amplicon sequencing (microbiome composition). For CFU counting, faecal pellets from both groups were diluted (1 g in 10 ml) in PBS and homogenized by vortexing. Decimal dilutions (10⁻¹ to 10⁻⁹) were prepared and grown on blood agar (TSA, 10% blood) at 37 °C under aerobic and anaerobic conditions prepared with gas packs (AnaeroGen; Thermo Fisher). CFUs were calculated using two successive dilutions. For 16S rRNA sequencing, faecal pellets were collected from mice and DNA was extracted using the PowerSoil Kit (QIAGEN). Sequencing was carried out on an Illumina platform generating 250-base-pair (bp) paired-end sequences of the V4 region.

After checking the effects of ABX, we gave both groups of mice the nitrosamines (0.05% BBN and 2.87 mM EHBN, DBN and PBN) in drinking water. BBN was administered for 12 weeks, after which the chemical was removed from the drinking water for 8 weeks. PBN was administered for 17 weeks, whereas EHBN and DBN were administered for 20 weeks^6,24.

For tissue collection, mice treated with BBN and EHBN were euthanized by inhalation of CO₂, and intestinal contents, liver, kidney, urinary bladder and plasma were collected and frozen at −80 °C until further analyses. Bladder tissues were cut in half longitudinally, half of which was frozen for metabolomics quantification and the other half was fixed for histopathology analyses.

For monitoring urinary and faecal excretion, 24-h urine and faecal samples from mice treated with all chemicals were collected using metabolic cages (Techniplast). Mice were placed inside cages supplied with food and water (supplemented with nitrosamines with or without ABX for respective groups) for 24 h and faecal material and urine was collected from the collection tubes as well as from the surfaces of the cages. Samples were weighed and stored at −80 °C until further analyses.

Pharmacokinetics experiment

Fifty conventional mice of both sexes were randomly assigned to the ABX–BBN or BBN group. The ABX–BBN group was given a mix of ABX (neomycin, metronidazole, streptomycin, penicillin and vancomycin at the same concentrations as described above) for 2 weeks before receiving BBN. Both groups of mice were given a single dose of BBN (50 mg per kg body weight) by oral gavage and euthanized after 1, 3, 6 and 9 h (five mice per time point, per group) and intestinal contents, liver, gallbladder, kidneys and plasma were collected, weighed, snap frozen in liquid nitrogen and stored at −80 °C until further analysis.

Collection and preparation of human microbiome samples

Patients were recruited from the Gastroenterology Department at University Hospital of Split between March and August 2020. The study was approved by the ethics committee of the University Hospital of Split (permit number 2181-147-01/06/M.S.-20-4) and the University of Split School of Medicine (permit number: Ur. br. 2181-198-03-04-20-00400) and all participants gave informed consent before participation in the study.

Faecal samples were collected from 12 patients undergoing gastroenterological analysis for different medical conditions. Exclusion criteria for all patients included in the study were: usage of ABX one month before sample collection and presence of malignant pathology of the analysed tissues.

Half of the sample volume was mixed with glycerol (20% final concentration) and stored at −80 °C to conserve viable microorganisms for subsequent in vitro BBN conversion assays. The remaining sample volume was directly stored at −80 °C for DNA extraction (PowerSoil Kit, QIAGEN) and 16S rRNA sequencing.

Histological analysis

Following euthanization of mice, the bladders were collected and medially cut into equal halves using a scalpel. One half was immediately frozen in liquid nitrogen and kept for subsequent analysis. The other half was immersed in neutral-buffered formalin (10%) for 24 h. Following fixation, tissues were dehydrated using a series of ethanol dilutions (75%, 90%, 95% and three times 100%, for 1 h each), cleared with xylene (three series for 30 min each) and finally embedded in paraffin (immersed in two series of paraffin for 1 h each and embedded in a third series of paraffin). Embedded tissues were then cut into 5-μm sections with a microtome (RM2125 RTS, Leica) and stained with haematoxylin (Sigma) and eosin (Merck) for microscopical examination. Blinded histological assessment was carried out by researchers and by a trained pathologist.

HEP-G2 cell line experiment

HEP-G2 cells were treated at 70% confluence with either BBN (0.05%) or BBN (0.05%) plus ABX mixture (containing vancomycin, streptomycin, metronidazole, neomycin and penicillin (in the same concentrations as in the mouse experiment)) in DMEM medium (Sigma-Aldrich) with 10% FBS (PAA Laboratories) for 24 h. Cells were pelleted (300 rcf, 3 min) and supernatant was used for BBN and BCPN measurements by high-performance liquid chromatography (PerkinElmer series 200) as described previously²⁵. HEP-G2 (ACC-180) cells were obtained from the DSMZ (German Collection of Microorganisms and Cell Cultures).

Ex vivo and in vitro BBN conversion assays

CFU of ex vivo microbial communities were determined as described above. To isolate single ex vivo communities, bacterial cultures at different dilutions were streaked onto brain heart infusion (BHI) blood plates, and incubated under anaerobic, microaerobic and aerobic oxygen conditions and incubated at 37 °C for 24–48 h. Between 24 and 48 single colonies were then randomly picked from dilution plates when possible. The full-length DNA of the 16S rRNA gene of each isolate was PCR-amplified using the 27f (5′-AGAGTTTGATCATGGCTCA-3′) and 1492r (5′-TACGGTTACCTTGTTACGACTT-3′) primer pair and sent for Sanger sequencing (Eurofins; Supplementary Tables 50 and 51). Type strains from the German Culture Collection (DKFZ) or the American Type Culture Collection (ATCC) were streaked onto BHI blood plates and allowed to grow for 1–2 days before use.

Intestinal and faecal samples obtained from mice and humans were grown as ex vivo microbial communities and single bacterial isolates and strains were grown as axenic cultures at 37 °C under 0% (anaerobic, Coy chamber), 10% (microaerobic, Coy chamber) and 21% (aerobic, incubator) oxygen conditions in BHI-S medium (37.0 g BHI broth, 5 g yeast extract supplemented with l-cysteine HCl, haemin and vitamin K1 in 1 l distilled water) or modified GAM (MGAM) broth (HyServe, catalogue number 5433) for 2 h for communities to prevent overgrowth of strains in microbial communities or overnight for axenic cultures. Cultures were then treated with 10 μM BBN after mixing with an equal volume of twofold-diluted BHI-S medium to allow further exponential growth. A 20 µl volume of the treated culture was collected after 0, 4, 8, 12 and 24 h of BBN incubation and immediately frozen in a 96-well V-bottom storage plate (Fisher Scientific, catalogue number 10304513), and sealed with aluminium foil at −80 °C for storage until further processing. To ensure even oxygen distribution in each well under microaerobic and aerobic conditions, the assays were carried out on an orbital plate shaker (Thermo Shaker PHMP) at 600 r.p.m.

16S rRNA sequencing

DNA was isolated from mouse and human samples using the QIAGEN PowerSoil kit and quantified on Qubit (Thermo Fisher Scientific). DNA library preparation and sequencing were carried out as described before²⁶. In brief, for the first PCR reaction, 5–20 ng of DNA was amplified for 25 to 30 cycles with KAPA or Q5 (BioLabs) master mix with primers for the variable region V4 of the 16S rRNA gene (forward: 5′-GTGCCAGCMGCCGCGGTAA-3′; reverse: 5′-GGACTACHVGGGTWTCTAA-3′)²⁷. For the second PCR reaction, 1 μl or 2 μl, depending on the PCR product band width on the gel, of the first PCR product was mixed with primers (NEXTFLEX 16S V1-V3 Amplicon-Seq Kit, PerkinElmer), and KAPA or Q5 (Biolabs) master mix and amplified for 25 cycles. After pooling, the library was cleaned with the AMPure XS kit (Beckman) with 0.75× ratio according to the manufacturer’s protocol. All samples were paired-end sequenced with 250-bp read lengths on the Illumina MiSeq platform at the Genomics Core Facility of EMBL Heidelberg.

Raw data quality was assessed using FastQC v0.11.5 and raw data were imported in QIIME2 v2020.8 for downstream analysis²⁸. Samples were checked for adapter contamination with cutadapt²⁹ and passed to dada2³⁰ for denoising, dereplication and chimera filtering; a feature table describing the distribution of reads in each sample among the identified ASVs was created, together with a representative sequence for each of the ASVs. The representative sequences identified for each ASV were used to create a fragment insertion tree using sepp³¹. The feature table and the built insertion tree were used to compute α- and β-diversity metrics, separately for mice and humans.

DNA sequences of the whole 16S rRNA gene (V1–V9) from isolated bacteria obtained by Sanger sequencing were also imported into QIIME v2020.8, together with their corresponding taxonomy, as derived from the ACT service³². When possible, ACT taxonomy was further detailed at species level (if not already present) by directly aligning complete sequences to the 16S–/ITS reference database using BLAST³³, only if a percentage identity of >99% could be detected over an alignment length of minimum 700 bp. ASV representative sequences were aligned to these full-length 16S rRNA sequences using vsearch³⁴ with the following parameters: –p-maxaccepts ‘all’, –p-perc-identity 0.98, –p-query-cov 0.98, –p-top-hits-only True. Representative sequences not aligned to any Sanger sequence were assigned a taxonomy with a pre-fitted sklearn-based classifier³⁵, trained over the Greengenes 13_8 99% operational taxonomic unit full-length sequences. The feature table and taxonomy thus obtained were exported to plain tsv files, to be imported into R 4.0.0 (https://www.R-project.org/).

All ASVs found in mouse and human samples were respectively aligned with mafft³⁶ and used to construct a phylogeny with fasttree2³⁷. The phylogenetic tree was exported in Newick format and imported into iTOL³⁸ for visualization and tree annotation.

Whole-genome sequencing

Bacterial isolates were grown in liquid culture overnight, and DNA was extracted with PowerSoil kit (QIAGEN). Libraries for whole-genome sequencing were prepared using the MagicPrep NGS system with 40 μl of 100 ng DNA. High-throughput sequencing was carried out with MiSeq Reagent Kit v2 Micro, resulting in 155-bp-long paired-ends reads. Sample quality was assessed using FastQC v0.11.9³⁹ and MultiQC 1.12⁴⁰; sample reads were trimmed for both quality and length using Trimmomatic 0.39⁴¹ with the following options: removal of TruSeq adapters sequences; sliding window trimming, clipping the read once the average quality within the window (4 bp) falls below 20; finally, drop the read if it is shorter than 38 bp (Supplementary Table 35). Subsequently, reads were de novo-assembled using Spades v3.15.3⁴², and the quality of assembly was assessed using QUAST v5.0.2⁴³.

Three different approaches were used to assess species assignation: a k-mer-based approach, using kraken v2.0.7⁴⁴; a marker gene approach, using gtdbtk v2.1.1⁴⁵; a 16S-based approach, using mTAGs v1.0.4⁴⁶. All three approaches agreed in classifying all isolates as Escherichia or Escherichia–Shigella.

All assemblies were therefore annotated using Prokka v1.13⁴⁷ with the genus Escherichia as Organism details. Annotation results were passed to Roary v3.7.0⁴⁸ for pangenomic analysis. The 7 isolates shared 3,277 core genes over a total of 7,606 genes (Supplementary Table 36). Pangenome results were visualized with Phandango v1.3.0⁴⁹, and the tree built from accessory genome presence absence was visualized using iTOL v6⁵⁰.

Sample preparation for LC–MS measurements

Sample preparation for LC–MS measurements was carried out as previously described⁵. In brief, for liquid tissues (that is, urine and plasma), 20 µl of the samples was subjected to a freeze–thaw cycle before the addition of 100 µl of acetonitrile/methanol (1:1), and 5 µl of 8 µM of internal standard, warfarin, was added to each sample. Samples were incubated at −20 °C for 30 min before centrifugation at 4 °C at 4,500 r.p.m. for 15 min. A 10 µl volume of the supernatant was diluted with 10 µl of water before LC–MS measurements. For solid tissues (that is, intestine, kidney, liver and bladder), samples were homogenized with 3-mm tungsten carbide beads (QIAGEN, catalogue number 69989) in 300–400 µl of acetonitrile/methanol (1:1) containing 320 nM internal standard by 5 min bead-beating at 30 Hz on a QIAGEN TissueLyser II. After 30 min incubation at −20 °C and 15 min centrifugation at 4 °C and 10,000 rcf, 10 µl of the supernatant was diluted with water at 1:1 or more before LC–MS measurements.

Quantification of BBN and BBN metabolites by LC–MS

Chromatographic separation was carried out on a 3.0 mm × 10 cm Poroshell120 HPHC18 column with 1.9 µM particle size (Agilent Technologies) mounted on an Agilent 1290 Infinity II LC system coupled to a 6550 iFunnel qToF mass spectrometer. The column temperature was maintained at 45 °C. The mobile phase was composed of A: water with 0.1% formic acid; and B: methanol with 0.1% formic acid. A 5 μl volume of each sample was injected at 0.6 ml min⁻¹ flow rate starting from 5% mobile phase B followed by a linear gradient to 95% over 5.5 min. The column was allowed to re-equilibrate to starting conditions for 1.1 min before each sample injection. The mass spectrometer was operated in both negative and positive scanning mode (50–1,700 m/z) with the following source parameters: VCap, 3,500 V; nozzle voltage, 2,000 V; gas temperature, 275 °C; drying gas, 13 l min⁻¹; nebulizer, 45 psi; sheath gas temperature, 275 °C; sheath gas flow, 12 l min⁻¹, fragmentor, 365 V. Online mass calibration was carried out using a second ionization source and a constant flow (2 ml min⁻¹) of reference mass solvent (purine m/z = 121.0509 and hexakis m/z = 922.0098 for positive mode, betaine m/z = 112.9856 and hexakis m/z = 1033.9881 for negative mode). Standard curves for BBN and BCPN were obtained by serial-diluting each compound in water at twofold from 10 µM to 4.9 nM.

The MassHunter Qualitative Analysis software (Agilent Technologies, v10.0) was used to determine the retention time for all compounds to enable targeted analysis and quantification. Peak integration was carried out in MassHunter Quantitative Analysis software (Agilent Technologies, v10.0) with the following settings: mass tolerance = 20 ppm; peak filter at signal-to-noise ratio = 2; and retention time tolerance of 0.5 min. All statistical analyses and plotting were carried out in RStudio (v4.2.0) after exporting the data files from MassHunter Quantitative Analysis software. BBN and BCPN concentrations were calculated using the recorded calibration curves and the signals of the internal standard. In brief, the area of the internal standard warfarin in each sample was divided by the median area of warfarin across all samples in the same tissue to obtain the correction factor for each sample, which was then used to normalize the areas for the targeted compounds in each sample. Total drug amounts in each intestinal compartment were calculated using the corresponding total sample weight.

Reporting summary

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