Oral administration of Akkermansia muciniphila elevates systemic antiaging and anticancer metabolites

Experimental design

In this study, mice were subjected to a defined sequence of interventions involving sham gavage with phosphate buffered saline (PBS), oral administration of broad-spectrum antibiotics (ATB, a combination of ampicillin, streptomycin and colistin), FMT from cancer patients and gavage of a series of distinct bacterial species. These species were selected because they improve the anticancer effects of cycloheximide-based chemotherapy, as true for Burkholderia cepacia (Brc) and Enterococcus hirae (Hir) [21, 22]. Bacteroides fragilis (Frg) was chosen because it improves chemotherapy with oxaliplatin and immunotherapy with CTLA-4 blockade [23, 24]. In addition, we included Burkholderia sp. (Bur) and Catenibacterium mitsuokella (Mit) as controls. These species were administered after ATB conditioning without prior FMT, allowing for the spontaneous recovery of the gut microflora. In contrast, since Akk (a human gastrointestinal mucin-loving bacterium) requires the presence of other bacteria to achieve efficient colonization [25], Akk was gavaged to mice that had previously received a FMT from a cancer patient that had not responded to immunotherapy and whose stools had been screened for absence of Akk. In a first experiment, only live Akk was given to mice, comparing its effects to that of other bacteria or FMT alone over time on days 3, 7 and 14 after discontinuation of ATB (Figure 1A). In a second experiment (Figure 1B), live Akk was compared to Akk-past, based on the observation that pasteurization actually does not destroy the antiobesity and antidiabetic effects of Akk [18, 19]. This experiment was designed to characterize long-term effects of Akk versus Akk-past with one single time point (day 27).

Figure 1. Schematic view of the experimental design. (A) Several commensals in mono-associated mice previously decontaminated by broad spectrum antibiotics (ATB) were administered by oral gavage. Akkermansia muciniphila (Akk) was administered by oral gavage to mice previously decontaminated by broad spectrum ATB and transplanted with human fecal material (FMT). Continuous ATB was administrated in the drinking water of the animals. (B) Mice previously decontaminated by broad spectrum ATB and FMT-treated, received Akk or the pasteurized form of Akk (Akk-past) by oral gavage (versus continuous FMT or ATB). PBS was administered as control (in both, (A, B) designs). (C) General methodology for metabolomics. The impact of the different treatments on the local (ileal, colon) and distal (liver, plasma) metabolome was evaluated using gas and liquid chromatography coupled to mass spectrometry allowing to detect a broad range of targeted and untargeted metabolites. Metabolite extraction was performed on the different organs collected from mice at the experiment day marked with a red arrow in (A, B). Extracts were processed and analyzed by liquid- and gas-chromatography coupled to mass spectrometry. Data was merged and analyzed using the GRmeta package in R or a built-in software, Compound Discoverer, for untargeted metabolomics. ATB, antibiotics; PBS, phosphate buffer saline; Brc, Burkholderia cepacia; Hir, Enterococcus hirae; Frg, Bacteroides fragilis; Bur, Burkholderia sp.; Mit, Catenibacterium mitsuokella; FMT, fecal microbiota transplant; Akk, Akkermansia muciniphila; Akk-past, pasteurized Akkermansia muciniphila.

Both experiments, which involved a minimum of 6 mice per group and time point (total 245 mice) were followed by the recovery of ileal and colonic content as well as that of liver and plasma. These samples were then subjected to metabolite extraction [26], optional chemical derivatization, different types of (gas or liquid) chromatography and mass-spectrometric identification of metabolites, either in a targeted mode (in which each metabolite is identified based on its chromatographic characteristics coupled to its exact mass) or in an untargeted mode (in which each metabolite is identified as a discernible peak with a defined mass). After analysis of the samples on three different mass spectrometers, the data were subjected to R-based informatics treatments to combine results obtained by different methods into single files (one for targeted and one for untargeted results for each sample) and bioinformatics analysis (Figure 1C). In total, 1.637.227 mass spectrometric measurements were performed in this study.

A gradient of microbial effects from the intestine to the circulation

In the first experiment (Figure 1A), continuous antibiotic (ATB) treatment caused a massive depletion of ileal metabolites (as indicated by the green color in the heatmap) that was manifest on days 3, 7 and 14 post-ATB (Figure 2 and Supplementary Figure 1 and Supplementary Table 1), supporting the importance of the ileal microbiota for the breakdown of nutrients into small molecules. Thus, more than half of the metabolites in the small intestine were significantly reduced in their abundance after ATB treatment. This strong ATB effect was also observed for the colic content, in which many metabolites were reduced while others including several monosaccharides and amino acids were increased in their abundance (as indicated by the red color in the heatmap, Figure 2 and Supplementary Figure 2 and Supplementary Table 2). Gavage with individual bacteria (Brc, Bur, Frg, Hir, Mit) or FMT, alone or with Akk upon ATB discontinuation gradually corrected the ATB-induced depletion of ileal metabolites, as well as the shifts in colonic metabolism, over time, thus allowing the ileal and colic metabolomes to recover to a state that resembles the basal state (represented by the PBS control). Very similar tendencies were observed for the hepatic and plasma metabolomes, though with the important difference that the concentration of most metabolites remained close-to-unaltered and only a minority decreased (and rarely increased) upon ATB treatment (Supplementary Figure 3). Again, the ATB-induced changes in the hepatic and plasma metabolomes were reset to close-to-normal levels by reconstitution of the intestinal microbiota (Figure 2 and Supplementary Figures 4, 5 and Supplementary Tables 3, 4).

Figure 2. Targeted metabolomics analysis was performed on the extracts from ileum, colon, liver, and plasma samples from mice receiving oral gavages with several commensals or PBS (control), or continuous ATB, at days 3, 7 and 14 after the first oral gavage. Changes in metabolites relative abundance are illustrated. Ileum and colon showed the strongest treatment-dependent metabolites variations. Hierarchical clustering (euclidean distance, ward linkage method) of the metabolite abundance is shown. Note that Supplementary Figures 1, 2, 3 and 5 provide the names of each of the metabolites, for each of the different matrices. The purpose of this figure is to allow for a direct comparison of the amplitude of the metabolic effects of manipulation of the microbiota. ATB, antibiotics; PBS, phosphate buffer saline; FMT, fecal microbiota transplant; FC, fold change.

In conclusion, the composition of the hepatic and circulating metabolomes is less affected by ATB-mediated sterilization of the gut than the ileal and colic metabolome, confirming that the internal milieu is protected against external perturbations. The effects of individual bacteria on this system appear discrete.

Akkermansia muciniphila (Akk)-specific effects on metabolism

In the next step, we compared the effects of Akk plus FMT with those of FMT alone on the ileal, colic hepatic and circulating metabolomes, at 3, 7 and 14 days after discontinuation of the initial ATB treatment. In this comparison (Akk+FMT versus FMT) rather discrete effects were observed by targeted metabolomic analysis (Figure 3 and Supplementary Figures 6–9). Volcano plots revealed a few consistent early (3 days) changes in the ileal and colic metabolomes with an Akk-induced increase in N1,N12-diacetylspermine (in both ileum and colon) as well as an increase in colic short-chain fatty acids (day 3: propionate, day 7: butanoic acid, also called butyrate) (Figure 4 and Supplementary Figures 10, 11). A Akk-induced increase in N1-acetylspermidine and N1,N8-diacetylspermidine was observed in the ileum (Supplementary Figure 10) and in the colon on day 7 but not 14 (Supplementary Figure 11). Metabolites generated during the catabolism of branched chain amino acids were also found overabundant in the ileum (day 3: ketoisovaleric acid, ketoisocaproic acid, Figure 4) and the colon (day 7: ketoisocaproic acid, Supplementary Figure 11). Of note, Akk caused the levels of N1,N12-diacetylspermine to increase in the liver at a late time point (day 14, Supplementary Figure 12). Ferulic acid was also increased in Akk-treated mouse livers on days 3 and 14 (Supplementary Figure 12). In contrast, no consistent changes were observed in the plasma after treatment with Akk, perhaps with the exception of a significant but transient (on day 3) increase of 2−hydroxy−3−methylbutyric acid (Supplementary Figure 13). In conclusion, it appears that Akk has the capacity to increase the concentration of N1,N12-diacetylspermine in the gut as well as in the liver.

Figure 3. Ileum, colon, liver, and plasma targeted metabolites relative abundance variations in mice treated with FMT or Akk at days 3, 7 and 14 after the first oral gavage. Hierarchical clustering (euclidean distance, ward linkage method) of the metabolite abundance is shown. Note that Supplementary Figures 6 to 9 provide the names of each of the metabolites, for each of the different matrices. The purpose of this figure is to allow for a direct comparison of the amplitude of the metabolic effects of manipulation of the microbiota. FMT, fecal microbiota transplant; Akk, Akkermansia muciniphila; FC, fold change.

Figure 4. Differential metabolite identification in samples from ileum (A), or colon (B) from mice receiving FMT or Akk at day 3 after the first oral gavage. The horizontal dashed gray line shows where p=0.05 with points above being metabolites with significantly different relative abundance (p<0.05). The vertical dashed gray lines correspond to FC=1. Targeted metabolites that display both large magnitude FC and higher statistical difference (-log10 of p value) are annotated. Families of metabolites are grouped by colors.

Metabolic effects of pasteurized (Akk-past) versus live Akkermansia muciniphila (Akk)

In a subsequent experiment, we specifically addressed the question as to whether Akk can induce long-term changes (day 27) using the aforementioned methodological approach (Figure 1B). We compared the metabolome of control mice (PBS), ATB-treated mice not receiving any bacteria (ATB group), mice receiving FMT alone (FMT group) or mice receiving FMT in combination with live Akk (Akk group) or Akk-past. The overview heatmaps (24000 features in the ileum, 22000 in the colon, 3000 in the liver, 1700 features in plasma) confirm the observations in the first experiment, with major changes in the ileal and colic metabolomes and comparably minor alterations in liver and plasma (Figure 5 and Supplementary Tables 5–8). Plotting the significant changes only (Figure 6), clearly revealed that live Akk has a stronger modulatory effect (as compared to FMT only) than Akk-past.

Figure 5. Metabolic signature obtained through untargeted metabolomic analysis, and illustrated by heatmaps showing the relative abundance changes in metabolites extracted from ileum, colon, liver, and plasma from mice treated with FMT, Akk or with Akk-past (versus PBS or continuous ATB). Untargeted analysis allowed the identification of more than 24000 and 22000 features in the ileum and the colon samples, respectively, while over 3000 and 1700 features were detected in liver and in plasma samples, respectively. Each row represents a detected feature. Hierarchical clustering (euclidean distance, ward linkage method) of the metabolite abundance is shown. PBS, phosphate buffer saline; ATB, antibiotics; FMT, fecal microbiota transplant; Akk, Akkermansia muciniphila; Akk-past, pasteurized Akkermansia muciniphila; FC, fold change.

Figure 6. Reduced picture of the untargeted metabolomics analysis illustrated in Figure 5, showing selected features which significantly increased after Akk or Akk-past oral gavage in the different extracted organs, compared to FMT, PBS, or continuous ATB. Each row represents a metabolite. Hierarchical clustering (euclidean distance, ward linkage method) of the metabolite abundance is shown. PBS, phosphate buffer saline; ATB, antibiotics; FMT, fecal microbiota transplant; Akk, Akkermansia muciniphila; Akk-past, pasteurized Akkermansia muciniphila; FC, fold change.

Nonetheless, as compared to live Akk, Akk-past induced higher levels of several polyamines (N1-acetylspermine, N1,N12-diacetylspermine, N1,N8-diacetylspermidine, spermidine) and their precursor ornithine, and the ornithine/arginine ratio (with no effect on the arginine level), short-chain fatty acids (acetate, propionate), as well as the ketone body 2-hydroxybutyrate (and its likely derivative 2-hydroxy-3-methylbutyrate) in the ileum (Supplementary Figures 14–16 and Supplementary Table 9). Many among these metabolites were also increased with Akk-past, but not with live Akk, in the colon, as observed for several polyamines (N1,N12-diacetylspermine, N1,N8-diacetylspermidine, spermidine, putrescine) an increase of ornithine with a relative depletion of its precursor arginine resulting in an increased ornithine/arginine ratio, as well as for 2-hydroxybutyrate and 2-hydroxy-3-methylbutyrate (Supplementary Figures 14, 15, 17 and Supplementary Table 10). We observed an augmentation of bile acids in response to Akk-past in the ileum (deoxycholic, hyodeoxycholic, murideoxycholic acid) and the colon (deoxycholic, hyocholic, lithocholic, omega−muricholic, murideoxycholic, taurodeoxycholic, tauro−muricholic, taurohyodeoxycholic, tauroursodeoxycholic acids) (Supplementary Figures 16, 17). Outside of the gut, the only consistent increase induced by Akk-past as compared to live Akk also concerned bile acids (such as chenodeoxycholic, deoxycholic, hyodeoxycholic, beta−muricholic, omega−muricholic taurodeoxycholic, ursodeoxycholic acids). This increase in bile acids was only observable in plasma but not in liver (Supplementary Figures 18, 19 and Supplementary Tables 11, 12).

In general, targeted analysis showed that Akk live or pasteurized, induced significant changes on the metabolome of the treated mice, being more remarkable in the gut (Supplementary Figure 20). Untargeted analyses of the plasma revealed numerous significant differences in the metabolome of mice receiving FMT plus live Akk versus FMT plus Akk-past (Figure 7A). One metabolite that was selectively induced in ileum, colon, liver and plasma by live Akk, not by Akk-past nor by any other condition, was annotated by its molecular weight (MW 115.0632) (Figure 7B).

Figure 7. Differential plasma metabolites identification between Akk- or Akk-past-treated mice (A). One metabolite (marked in blue in A) showed significant difference after comparison versus PBS, ATB and FMT groups, it was annotated by its molecular weight as MW 115.0632. (B) Plasma MW 115.0632 was identified in ileum, colon and liver, according to similar retention time and fragments, showing the same trend in the four extracted organs.

In summary, there are important differences in the metabolic effects of live Akk versus Akk-past that may be biologically relevant because they affect multiple bile acids, a ketone body, several polyamines, as well as short chain fatty acids.

Mice

Experiments were performed in accordance with Government and institutional guidelines and regulations (French and European laws and regulations). The institutional board of animal ethics and Ministère de la Recherche approved all mouse experimental settings (permission numbers: 2014-074-501 and APAFIS#13366-2018020510263031v3). Female C57BL/6 were purchased and used at 7 weeks of age from Envigo (France) and bred in specific pathogen-free conditions (unless otherwise specified) in the animal facility of Gustave Roussy Cancer Campus.

Antibiotic treatments

Intestine decontamination was achieved with an antibiotic solution (ATB) containing ampicillin (1 mg/mL), streptomycin (5 mg/mL) and colistin (1 mg/mL), added to the sterile drinking water of mice. Duration of ATB treatment was different based on the experimental settings (Figure 1A, 1B). Antibiotic activity was confirmed by cultivating fecal pellets resuspended in Brain Heart Infusion (BHI) + 15% glycerol at 0.1 g/mL on COS (Columbia Agar with 5% Sheep Blood, COS ref 43049, BioMérieux, France) plates for 48h at 37° C in aerobic and anaerobic conditions (GENbox Jar and GENbag, Biomérieux, France).

Fecal microbiota transplantation (FMT) experiments

Mice received 3 days of ATB before undergoing FMT the next day by oral gavage using animal feeding needles. FMT was performed by thawing fecal material. Two hundred μL of the suspension was then transferred by oral gavage into ATB pre-treated recipients. In addition, another 100μL was applied on the fur of each animal.

Bacterial preparations and gut colonization with dedicated bacterial species

Colonization of ATBs pre-treated or FMT-reconstituted C57BL/6 mice was performed by oral gavage with 100 μL of suspension containing 1×10⁸ bacteria. Enterococcus hirae were grown on COS plates for 24 h at 37° C in aerobic conditions while other bacteria, including A. muciniphila p2261, were grown on COS plates for 72 h at 37° C in anaerobic conditions. In some experiments, an identical quantity of grown A. muciniphila was inactivated by pasteurization for 30 min at 70° C, and then immediately frozen and stored at −80° C [19]. Colonization of 3-day ATB pre-treated or FMT-treated C57BL/6 mice was performed by oral gavage with 100 μL of suspension containing between 1×10⁸ and 1×10⁹ bacteria. Culture and identification of specific bacteria was accomplished through the combination of morphological tests and analysis by means of a Matrix-Assisted Laser Desorption/ Ionisation Time of Flight (MALDI-TOF) mass spectrometer (Andromas, Beckman Coulter, France or Microflex, Bruker). Five bacterial gavages were performed for each mouse, the first 24 h after stopping ATB administration. We next monitored in a kinetic manner (day 3, day 7, day 14) the impact of oral gavages with several commensals on the local (ileal, colonic) and distal (liver, plasma) concentrations of various metabolites using gas and liquid chromatography coupled to mass spectrometry as described below.

Sample preparation tissue (liver) and fecal content (ileum, colon)

About 30 mg of biological matrix for each condition were first weighted and solubilized into 1.5 mL polypropylene microtubes with ceramic beads with 1 mL of cold lysate buffer with internal standard (ISTD) (methanol/water/chloroform, 9/1/1, -20° C). They were then homogenized three times for 20 s at 5500 rpm using Precellys 24 tissue homogenizer (Bertin Technologies, Montigny-le-Bretonneux, France), followed by a centrifugation (10 min at 15000 g, 4° C). Then the upper phase of supernatant was split in three parts: first 300 μL were used for Gas Chromatography coupled to Mass Spectrometry (GC/MS) experiment in microtube. Next, two aliquots (50 μL and 300 μL, respectively) were used for Ultra High Pressure Liquid Chromatography coupled to Mass Spectrometry (UHPLC/MS) analysis. Concerning GC-MS aliquots, 300 μL were transferred from the microtube to glass tube and evaporated. 50 μL of methoxyamine (20 mg/mL in pyridine) were added on dried extracts, then stored at room temperature in dark, during 16 hours. The day after, 80 μL of MSTFA were added and final derivatization occurred at 40° C during 30 min. Samples were then transferred in vials and directly injected into GC-MS. Concerning UHPLC-MS aliquots (for short chain fatty acids, SCFA method), 50 μl were transferred from the injection vial, and mixed with 25μl of 3-Nitrophenylhydrazine (3-NPH, 200 mM) and 25 μl of N-(3-Dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC, 150 mM). The whole was heated at 40° C during 1h. Next, 100 μl of water were added and samples were injected into UHPLC-MS. Concerning the other UHPLC-MS aliquots, 300 μL were evaporated in microtubes at 40° C in a pneumatically-assisted concentrator (Techne DB3, Staffordshire, UK). UHPLC-MS dried extracts were solubilized with 200 μL of MilliQ water. Aliquots for analysis were transferred in LC vials and injected into UHPLC-MS or kept at -80° C until injection. Regarding the rest of the supernatant and the pellet, 200 μl of methanol with 2% of sulfosalicylic acid (SSA) were added before vortex and centrifugation (10 min at 15000 g, 4° C). Next, 200 μl of the supernatant were transferred in a microtube and evaporated. The dried samples were spiked with 200 μl of MilliQ water transferred in polypropylene vials before injection in UHPLC/MS for polyamines analysis.

Sample preparation plasma (lithium heparin)

A volume of 25 μL of plasma was mixed with 250 μL of cold solvent mixture with ISTD (methanol/water/chloroform, 9/1/1, -20° C), into 1.5 mL microtube, vortexed and centrifugated (10 min at 15000 g, 4° C). Then the sample preparation followed the same step as for the tissues and fecal contents: 50 μL were used for GC-MS experiment in glass tubes, 50 μL for UHPLC-MS analysis, and 40 μL were used for short chain fatty acids detection. Regarding the rest of the supernatant and the pellet, 85 μl of methanol with 2% of sulfosalicylic acid (SSA) were added before vortex and centrifugation (10 min at 15000g, 4° C). 200 μl of the supernatant were transferred into a microtube and evaporated. The dried samples were spiked with 200 μl of MilliQ water transferred in polypropylene vials before injection in UHPLC/MS for polyamines analysis.

Quality control policy

A daily qualification of the instrumentation was set up with automatic tune and calibration processes. These qualifications were completed with double injections of standard mix, at the beginning and end of the run, as for a blank extracted sample to control background impurities. Standard mixes were adapted for each chromatographic method. After extraction, fractions of each biological sample were pooled to create a quality control (QC) sample, use to passivate the column before analysis with proper biological matrix and re-injected throughout the batch to monitor and correct analytical bias occurring during batch (m/z, retention time and sensitivity drifts) during post acquisition treatment signal.

Widely-targeted analysis of intracellular metabolites by gas chromatography (GC) coupled to a triple quadrupole (QQQ) mass spectrometer

GC-MS/MS method was performed on a 7890A gas chromatography (Agilent Technologies, Santa Clara, CA, USA) coupled to a triple quadrupole 7000C (Agilent Technologies,) equipped with a high sensitivity electronic impact source (EI) operating in positive mode. Detailed analytical methods are described in [26].

Targeted analysis of bile acids by ion pairing ultra-high performance liquid chromatography (UHPLC) coupled to a triple quadrupole (QQQ) mass spectrometer

Targeted analysis was performed on a RRLC 1260 system (Agilent Technologies) coupled to a Triple Quadrupole 6410 (Agilent Technologies) equipped with an electrospray source operating in negative mode. Gas temperature was set to 325° C with a gas flow of 12 L/min. Capillary voltage was set to 4.5 kV [26]. Peak detection and integration of analytes were performed using the Agilent Mass Hunter quantitative software (B.07.01), exported as tables and processed within R software.

Targeted analysis of polyamines by ion pairing ultra-high performance liquid chromatography (UHPLC) coupled to a triple quadrupole (QQQ) mass spectrometer

Targeted analysis was performed on a RRLC 1260 system (Agilent Technologies) coupled to a Triple Quadrupole 6410 (Agilent Technologies) equipped with an electrospray source operating in positive mode. The gas temperature was set to 350° C with a gas flow of 12 l/min. The capillary voltage was set to 3.5 kV [26]. Peak detection and integration of analytes were performed using the Agilent Mass Hunter quantitative software (B.07.01), exported as tables and processed within R software.

Targeted analysis of short chain fatty acids by ion pairing ultra-high performance liquid chromatography (UHPLC) coupled to a triple quadrupole (QQQ) mass spectrometer

Targeted analysis was performed on a RRLC 1260 system (Agilent Technologies) coupled to a Triple Quadrupole 6410 (Agilent Technologies) equipped with an electrospray source operating in negative mode. Gas temperature was set to 350° C with a gas flow of 12 L/min. Capillary voltage was set to 4.0 kV [26]. Peak detection and integration of analytes were performed using the Agilent Mass Hunter quantitative software (B.07.01), exported as tables and processed within R software.

Pseudo-targeted analysis of intracellular metabolites by ultra-high performance liquid chromatography (UHPLC) coupled to a Q-Exactive mass spectrometer. Reversed phase acetonitrile method

The profiling experiment was performed with a Dionex Ultimate 3000 UHPLC system (Thermo Scientific) coupled to a Q-Exactive (Thermo Scientific, Waltham, MA, USA) equipped with an electrospray source operating in both positive and negative mode and full scan mode from 100 to 1200 m/z. The Q-Exactive parameters were: sheath gas flow rate 55 au, auxiliary gas flow rate 15 au, spray voltage 3.3 kV, capillary temperature 300° C, S-Lens RF level 55 V. The mass spectrometer was calibrated with sodium acetate solution dedicated to low mass calibration. 10 μL of sample were injected on a SB-Aq column (100 mm × 2.1 mm particle size 1.8 μm, Agilent Technologies), protected by a guard column XDB-C18 (5 mm × 2.1 mm particle size 1.8 μm) and heated at 40° C by a pelletier oven. The gradient mobile phase consists of water with 0.2% of acetic acid (A) and acetonitrile (B). The flow rate was set to 0.3 mL/min. Initial condition is 98% phase A and 2% phase B. Molecules were then eluted using a gradient from 2% to 95% phase B in 22 min. The column was washed using 95% mobile phase B for 2 minutes and equilibrated using 2% mobile phase B for 4 min. The autosampler was kept at 4° C. Peak detection and integration were performed using the Thermo Xcalibur quantitative software (version 2.2).

Untargeted analysis of intracellular metabolites by ultra-high performance liquid chromatography (UHPLC) coupled to a Q-Exactive mass spectrometer. Reversed phase acetonitrile method

Orbitrap mass spectrometer (q-Exactive, Thermo Scientific) operating in both negative and positive ion, was used coupled to an UHPLC system (Ultimate 3000 UHPLC, Dionex) and acquired samples in full scan analysis mode. LC separation was performed on reversed phase (Zorbax Sb-Aq 100 x 2.1mm x 1.8μm, Agilent; A: water 0.2% acetic acid, B: ACN). The details about the analytical method are described above. Data acquisition was performed with the Thermo Scientific Xcalibur software (version 2.2). Raw data files were used to perform unbiased profiling analysis, with Thermo Scientific Compound Discoverer small molecule identification software (version 3.1).

Data analysis using compound discoverer

Raw data files obtained by the previously described pseudo-targeted analysis were also used to perform unbiased profiling analysis, using the Thermo Compound Discoverer (3.1.). After sample injection and data acquisition, raw data files were processed with Compound Discoverer software following a customized node-based workflow for identifying unknown compounds in metabolomics. First, spectra selection and retention time alignment were performed, followed by removal of background noise and baseline correction. Next, the processing workflow found chromatographic peaks for unknown compounds (molecular weight, MW, x retention time, RT) extracting all relevant spectral and chromatographic information, to predict the elemental composition of the unknowns. The possible identity of the unknown compounds was then searched against selected MS databases, such as ChemSpider (from MS1 scans by using MW or predicted composition when available), mZcloud (MS/MS spectral library), built-in databases (custom, local libraries), and Metabolika or KEGG databases (metabolic pathway search). Annotations are assigned to the detected compounds, to rank putative database results. Finally, the software performed statistical analysis using a multivariate method approach, e. g. PCA (unsupervised), and data visualization, e.g. volcano plots.

Statistical analysis

All targeted and pseudo-targeted treated data were merged and cleaned with a dedicated R (version 3.4) package (@Github/Kroemerlab/GRMeta). Calculations and statistical tests were performed using R v3.4. Wilcoxon-Mann-Whitney test was used to assess differences in concentration between two different groups. Data representation was performed with softwares R v3.6 and Rstudio v1.2.1335 using tidyverse, dplyr, ggplot2, ggpubr, complexheatmap and corrplot packages.