Hepatoprotective Activity of Nature-Derived Polyphenols Studied by Mass Spectrometry Based Multi-OMICS Approach

2. RESULTS AND DISCUSSION

2.1. Overview of the Study Pipeline

To examine the hepatoprotective activity of natural and nature-derived polyphenols, we developed the following methodology (Figure 1). In the first step, cell cytotoxicity and hepatoprotective activity of investigated mixtures were determined on HepG2 hepatocyte cell culture. Secondly, samples with a pronounced hepatoprotective effect were included for in vivo testing in a subacute liver failure model in BALB/C mice. In vivo study included collection of urine and liver tissues for proteome analysis with LC-MS, and non-target analysis of metabolites with ultra-high resolution mass spectrometry. Study included n samples, extracted from natural sources and one synthetic mixture reported in our previous work [20].

2.2. Determination of the Hepatoprotective Effect In Vitro

MTT colorimetric assay was conducted to determine the hepatoprotective effect of nature-derived polyphenols (%) on average in 2 repetitions. Cells were incubated in a therapeutic mode - with the addition of a hepatoprotector to the culture medium after incubation of hepatocytes with CCl₄. Samples included: humic (HA) and fulvic acid (FA), LM (Legume musk), BP-Cx-1 and IFL. HA, IFL and BP-Cx-1 showed a pronounced hepatoprotective effect (Figure 2, Table S1) and were selected for further in vivo study. Cell viability was identified by measuring the optical density with a microplate reader at 630 nm. For all samples, the effective dose 50 (ED50) was calculated using linear-logarithmic interpolation using the AAT Bioquest online calculator: ED50 HA = 238.9997 μg/ml; ED50 BP-Cx- 1 = 236.9988 µg/ml; ED50 IFL = 87.1657 µg/ml (Figure S1).

2.3. Determination of the Hepatoprotective Effect In Vivo

The dynamics of body weight of mice in the model of subacute liver failure by CCl₄ presented in the Table S2. Statistically significant differences in the body weight of animals in the experimental groups compared to the negative control group were revealed in the CCl₄ group, as well as in the BP-Cx-1 group. On the 17th day after the first administration of CCl₄, a significant decrease in body weight was observed in the CCl₄ group to 20.01 ± 1.56 g (vs. 21.24 ± 1.47 g in the negative control group (p-value < 0.05). On 3, 7, 10, 14, 17, 21 days after the first administration of CCl₄, the body weight in group CCl₄ + BP-Cx-1 also decreased statistically significantly relative to the values of the negative control (p-value < 0.05). The data on the body weight dynamics of experimental animals indicate the presence of toxic effects in all groups where CCl₄ was administered, which corresponds to literature data and data from previous experiments conducted at our base using CCl₄ [25]. Also, significant changes compared to the CCl₄ group were revealed in the CCl₄ + BP-Cx-1 group on the 14th day. In connection with the data obtained on the dynamics of body weight of experimental animals, we can conclude that the weight gain was less pronounced in the CCl₄ + BP-Cx-1 group.

The liver mass index in the negative control group (5.4±0.11) was significantly lower than in all CCl₄ induced pathology groups except for the group IFL (5.6±0.07, p=0.7403) (Figure S2). Treatment with IFL significantly decreased the mass index of the liver in comparison with CCl₄ group (6.0±0.09, p=0.02). HA and BP-Cx-1 treatment slightly decreased liver mass index without statistical significance (5.96±0,11, p=0,9997 and 5,83±0,13, p=0,7118, respectively).

All mice treated with CCl₄ developed severe microvesicular hepatic steatosis. In this liver injury model, the steatosis index 4-fold increased by CCl₄ as compared to negative control (1.93±0.22 and 0.46 ± 0.18, respectively, p=0,0002). IFL treatment showed a pronounced hepatoprotector activity significantly decreasing the steatosis index by 70% to 0.58 ± 0.23 (p=0,0008 vs CCl₄, and p=0,8300 vs Control) (Figure 3). Steatosis index was lower than in CCl₄ group in BP-Cx-1 group by 34% (1.27 ± 0.24, p=0,1574 vs CCl4, and p=0,0805 vs Control) and in HA group – by 11% (1.75±0.3, p=0,8300 vs CCl₄, and p=0,0014 vs Control). Thus, macro and micro studies of the liver showed subacute liver failure development by CCl₄ and hepatoprotective activity of selected agents. Moreover, results of in vitro and in vivo tests corroborated well with IFL demonstrating the highest activity.

2.4. Proteomic Analysis

Urine proteome Fraction Analysis

As the result of proteomic analysis of 20 urine pooled samples collected with metabolic cages 1882 protein groups were identified (FDR 1%) (Table S3). Evaluation of the proteomic composition in different groups was performed via hierarchical clustering based on the average levels of label-free quantification (LFQ) intensities. Semi-quantitative label-free proteomic analysis revealed 47 proteins (p-value < 0.01) which relative abundances were significantly changed in the CCl₄ group with respect to the negative control (Table S4). Enrichment analysis was conducted via the GO database using the STRING software [26]. Proteins significantly changed in CCL4 group are involved in extracellular matrix organization (Comp, Ctsb, Vcam1, Spp1, Ceacam1, Ctsl), as well as in abnormal wound healing (Scpep1, Comp, Tgfbr1, Spp1, Ceacam1).

Multisample ANOVA analysis of BP-Cx-1 and CCL4 groups with respect to the negative control group resulted in 117 significantly changed protein groups. Proteins are participated in extracellular matrix organization (Cdh1, Icam2, Comp, Nid1, Ctsb, Plg, Vtn, Madcam1, Vcam1, Pcolce, F11r, Fn1, Efemp2, Spp1, Ceacam1, Cd47, Ctsd), abnormal vascular wound healing (Comp, Lrp1, Spp1, Ceacam1) and in innate immune system pathways (Gm2a, Icam2, Ctsb, Vtn, Npc2, Ly86, Plau, Clu, Cfb, Dpp7, Dsc1, Gns, Hp, Cfi, Ggh, Ceacam1, Ear1, Cd47, Creg1, Ctsd, Ear6, Rnaset2b, Psap, Mme).

Liver Proteome Fraction Analysis

Liver tissues were collected on the 29th day. As the result of proteomic analysis of 62 extracts 4617 protein groups were identified (FDR 1%) (Table S5). Evaluation of the proteomic composition in different groups was performed via hierarchical clustering based on the average levels of label-free quantification (LFQ) intensities. ANOVA label-free proteomic analysis revealed 222 proteins (FDR < 0.01) which were significantly changed with respect to negative control (Table S6). T-SNE dimensionality reduction method shows the distribution of samples into major clusters (negative control, IFL, other groups) (Figure 4B). To assess the changes in the proteome composition between different treatment groups a Mann-Whitney test was performed. Proteomic composition of individual liver samples of BALB/c mice with subacute liver failure caused by CCl₄ allowed us to reliably distinguish intact and IFL samples. Heat map analysis shows that negative control and IFL groups could be clearly distinguished by hierarchical clustering as well (Figure 4A).

Pairwise comparison revealed 101 proteins that were significantly (p-value < 0.5, abs (Cohen’s d-size) < 0.05) changed between the CCL4 and negative control groups (Table S7).

Figure 4. A) Heat map of the significantly changed liver proteins (FDR <0.01) based on z-scores of the normalized LFQ values. The strength of the colors indicates the relative abundance of the protein in different groups. B) Clusterization by t-SNE of liver proteomic data for all 62 samples. Colored dots depict the liver samples: Control - green; CCL4- red; HA- orange; BPCX1 - blue; IFL – violet.

Figure 4. A) Heat map of the significantly changed liver proteins (FDR <0.01) based on z-scores of the normalized LFQ values. The strength of the colors indicates the relative abundance of the protein in different groups. B) Clusterization by t-SNE of liver proteomic data for all 62 samples. Colored dots depict the liver samples: Control - green; CCL4- red; HA- orange; BPCX1 - blue; IFL – violet.

GO enrichment analysis revealed participation in regulation of cell death (21 protein), response on chemical stimulus (27 proteins), metabolic process (71 protein), in biosynthetic process (30 proteins) and etc.

2.5. Metabolomic Analysis

Molecular fingerprints of chloroform-methanol extracts from urine and liver samples were derived by HRMS analysis using FTICRMS.For each group the number of determined molecular formulae (MF) varied from 762 for HA group to 1246 MF for negative control (Figure S3). Molecular profiles of all experimental groups were characterized by significant differences (Figure S4). Only 350 MF were conservative, which was less than 50% of determined MF group-wise. Negative control, CCl4, HA, BP-Cx-1 and ISF groups contained 307, 112, 43m 119 and 82 unique MF groups, respectively. In addition, based constrained aromaticity index (AIcon) 31, 60, and 36 CHO-only MF were attributed to aromatic and condensed compounds for HA, BP-Cx-1 and ISF, respectively. Likely, these highly unsaturated CHO compounds derived directly from tissue penetration of hepatoprotective agents. Therefore, metabolite fraction also reflects the impact of administration of CCl4 or hepatoprotective agent. Based on intensity distribution, control group was distinct with higher similarity of all other groups (Figure S5).

3. DISCUSSION

Proteomic analysis revealed new insights on the molecular mechanism of the hepatoprotective activity of multicomponent mixtures of natural origin. The significant difference was observed in proteome profile of urine and liver of BALB/c mice with SALF between the negative control, CCl₄ control, and receiving potential hepatoprotectors of natural origin (BP-Cx-1, Ha, IFL). Molecular fingerprinting of metabolites with FTICR MS corroborated with proteome results. PCA analysis demonstrated a clear cluster of control samples, while samples from other groups showed a mixed picture (Figure S3). Nevertheless, the ISF group demonstrated higher similarity to control group with cosine similarity of 0.75 which also supports the proteomic results.

Proteins that were significantly changed in urine during CCL4 administration are involved in extracellular matrix organization (Comp, Ctsb, Vcam1, Spp1, Ceacam1, Ctsl), as well as in abnormal wound healing (Scpep1, Comp, Tgfbr1, Spp1, Ceacam1). A significant decrease in the level of Osteopontin (SPP1) was observed in the CCl₄ group. SSP1 is an extracellular matrix protein involved in immune system processes in particular type-1 immune responses through enhancing production of interferon-gamma, interleukin-10, interleukin-12 cytokine expression. [27]. Prominin-1 (PROM1) is another significantly increased protein involved in cell differentiation, proliferation and apoptosis [28]. In contrast, the level of Cartilage oligomeric matrix protein (COMP) was significantly increased, COMP is a suppressor of apoptosis by blocking the activation of caspase-3 and by inducing the IAP family of survival proteins [29]. Carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1) - plays the role of a co-inhibitory receptor in the immune response, affects the action of insulin, and also functions as an activator of angiogenesis [30,31,32]. Such changes can be associated acute-phase response and release of proteins from the liver, which act as inflammatory mediators and scavengers during tissue repair. The decreasing of apoptotic and immune proteins in treatment groups can be related with restoration of the inflamed tissue back to a normal status.

The most pronounced changes were observed in the IFL group – 40 significantly changed proteins were measured with respect to the CCl₄ and negative control groups. Those proteins were participated in regulation of apoptosis (Ctsb, Plg, Cxadr, Ltf, Thbs1), innate immune system processes (Ctsb, Vtn, Hexb, Plau, Ltf, Ceacam1, Ear1, Creg1, Ctsl), as well as organization of the extracellular matrix (Comp, Ctsb, Plg, Vtn, Vcam1, F11r, Thbs1, Fn1, Ceacam1, Col15a1, Ctsl). Cathepsin B (CTSB) protein level was increased after CCl₄ administration and decreased after IFL treatment. CTSB is a thiol protease that is involved in intracellular protein and extracellular matrix degradation [33]. Plasmin (PLG) was another protein that decreased its level with respect to the CCL4 group. PLG acts as a proteolytic enzyme in a variety of processes including tissue remodeling and inflammation - in particular in neutrophil apoptosis, macrophage reprogramming, and efferocytosis [34].

Among 101 significantly changed in mouse liver proteins, 5 proteins (AOXC, NDRG1, ABCBB, DDX3Y, SEP11) (Table S7) can be suggested as potential targets for drug intervention, based on their average concentration between the CCL4 and control groups (Figure 5). It is worth noting that NDRG1 is a stress-responsive protein involved in hormone responses, cell growth, and differentiation and acts as a tumor suppressor in many cell types. NDRG1 also participates in p53/TP53-mediated caspase activation and apoptosis [35]. Another important protein target is septin 11 (SEP11) which is involved in the regulation of reactive oxygen species homeostasis and mediates 27-hydroxycholesterol activation of endothelial inflammation via estrogen receptor alpha [36].

According to the results of a proteomic analysis the greatest hepatoprotective effect is exerted by the use of IFL, which correlates well with the relative effectiveness of the drugs recorded in vitro studies. The results correlate with the relative effectiveness of the drugs recorded in previous in vitro and in vivo studies and may be due to the lower molecular weight of the active isoflavonoids compared to the components of the polyphenolic drug BP-Cx-1.

4. MATERIALS AND METHODS

3.1. Animal Studies

All animal studies were carried out at the center for preclinical research of the Federal State Institution “National Medical Research Center of Oncology named after N.N. Petrov” of the Ministry of Health of Russian Federation in accordance with the Study Protocol and standard operating procedures.

The study cohort comprised 62 female BALB/C mice obtained from the Stolbovaya branch of FSBSI “Scientific Center for Biomedical Technologies of the Federal Medical and Biological Agency” of Russia. Before the start of the experiment animals underwent quarantine and adaptation for 17 days. Mice were housed at a room temperature of 20–23°C, relative air humidity of 54–58% and an air exchange rate of 8 volumes per hour. All animals received standard complete briquetted chow (Laboratorkorm, Russia) and filtered water ad libitum. Mice were examined daily by the veterinarian.

Experimental Design

Mice were randomized by weight and the following groups of animals were formed:

• Control – negative control – intraperitoneal injection of sunflower oil (0.2 ml/mouse) (n=13);
• Control CCl4 – positive control – intraperitoneal injection of CCl4 (1 ml/kg) diluted ten times in sunflower oil 2 times per week (6 injections in total) (n=14);
• CCl4 + Fraction of humic acids from low-mineralized silt sulfide mud (peloids) of lake Molochka (MRC “Sergievskie Mineral Waters” FMBA of Russia) (HA) –intraperitoneal injection of CCl4 (1 ml/kg) diluted ten times in sunflower oil 2 times per week (6 injections in total) + intragastric administration of HA (60 mg/kg) 3 times a week, for 4 weeks (n=12);
• CCl4 + BP-Cx-1 –intraperitoneal injection of CCl4 (1 ml/kg) diluted ten times in sunflower oil 2 times per week (6 injections in total) + intragastric administration of BP-Cx-1 (60 mg/kg) 3 times a week, for 4 weeks (n=11);
• CCl4 + isoflavones isolated from the root of Pueralia lobata (IFL) –intraperitoneal injection of CCl4 (1 ml/kg) diluted ten times in sunflower oil 2 times per week (6 injections in total) + intragastric administration of IFL (60 mg/kg) 3 times a week, for 4 weeks (n=12).

Body weighing was carried out twice a week during experiment and prior euthanasia.

Urine samples were collected using metabolic cages (Tecniplast, Italy) for 18 hours a day before euthanasia. Urine samples were transferred in Eppendorf tubes and kept frozen at-80.

Euthanasia by CO2 was carried out on the 29th day after the first administration of CCL4. All animals were autopsied. Liver was excised and weighted, mass index (percentage ratio of liver mass to body weight) was calculated. Liver samples were transferred in Eppendorf tubes, immediately frozen in liquid nitrogen and kept frozen at-80 for further analysis. The rest was histologically examined after routine histological preparation (dehydration, impregnation with paraffin, cutting into sections, staining with hematoxylin and eosin). Evaluation of liver steatosis was carried out microscopically in 10 fields (x200) per slide by a semi-quantitative method (in points): 0 – no changes, 1 – up to 30% of hepatocytes contain lipid vacuoles in the cytoplasm, 2 – 30–60% of cells are changed, 3 – more than 60% of cells are changed. Mean steatosis index was calculated per animal and onward per group.

3.3. MTT Colorimetric Assay

The MTT colorimetric assay was used to determine cell cytotoxicity and hepatoprotective activity of multicomponent mixtures of polyphenolic compounds on HepG2 hepatocyte cell culture. Cells were cultured in DMEM (PanEco company, Russia) supplemented with 10% fetal bovine serum (Thermo Fisher Scientific, USA), 2 mM L-glutamine (Thermo Fisher Scientific, USA), 100 U/ml penicillin (Sigma, Germany), and 100 μg/ml streptomycin (Sigma, Germany). Incubation was carried out at 37°C in 5% CO2. In order to determine hepatoprotective activity of multicomponent mixtures, HepG2 cells were incubated (5×103 cells/well; 96-well plates, for 24-h) with 1 mmol CCl4 (LenReactiv, Russia) in 0.15% DMSO for inducing apoptosis [37]. After 24-h incubation cell cultures were washed twice with a PBS buffer. Next, potential hepatoprotectors were added at an initial concentration of each drug per well of 1 mg/ml with a dilution step of ½ and incubated for 24-h. Cell cultures were washed in a PBS buffer and 20 μl of MTT (Sigma, Germany) was added to each well and incubated for another 2 hours. After incubation medium was removed from the plates and 100 μl of DMSO was added to each well to dissolve the formed formazan crystals. Cell viability was identified by measuring the optical density with an ELISA reader (Tecan Infinite M200 PRO, Switzerland) at 630 nm.

3.4. Proteomic Analysis

Urine Samples Preparation

Urine samples were denatured and reduced by incubation with 8 M urea, 0.1 M dithiothreitol and 100 mM Tris × HCl (pH 8.0, +37 °C, 30 min). Next, the proteins were alkylated by a 30-min incubation in the dark with 55 mM iodoacetamide. Protein concentration was measured via

Bradford method. For trypsinolysis, the samples were diluted with 100 mM Tris × HCl (pH 8.0) until <1 M urea; L-(tosylamido-2-phenyl) ethyl chloromethyl ketone (TPCK)-treated trypsin (Worthington) was added at a 25:1 (protein:enzyme, w/w) ratio; and the samples were incubated for 16 h at 37 °C. The reaction was quenched by acidifying the samples with formic acid (FA) to a final concentration of 1.0% (pH ≤ 2). The resulting peptides were purified by solid-phase extraction on C18 cartridges (Oasis, Waters, USA), then lyophilized and dissolved for analysis in 0.1% formic acid.

Liver Samples Preparation

Liver tissue (10-20 mg) was ground into a fine powder in liquid nitrogen. Cell lysis was conducted by adding 50 mM Tris-HCl pH 8.0.150 m M NaCl, 0.1% SDS, 0.5% Na deoxycholate, 1% NP-40 lysis buffer containing a cocktail of protease inhibitors (Roche). Tissues were incubated for 30 minutes and sonicated in an ultrasonic bath twice for 3 minutes. Samples were centrifuged at 40C at 10000 g for 10 min, the supernatant was collected, and the sediment was re-extracted with denaturing buffer (8 M urea, 2 M thiourea, 50 mM Tris-HCl pH8.0, 0.5% NP-40). The supernatant was collected and pooled after the centrifugation at 40C at 10000g for 10 minutes. Protein concentration was measured by the Bradford method. 100 μg of protein was taken and precipitated with ice-cold acetone. Samples were denatured and reduced by incubation with 8 M urea, 0.1 M dithiothreitol and 100 mM Tris × HCl (pH 8.0, +37 °C, 30 min). Next, the proteins were alkylated by a 30-min incubation in the dark with 20 mM iodoacetamide. For trypsinolysis, the samples were diluted with 100 mM Tris × HCl (pH 8.0) until <1 M urea; L-(tosylamido-2-phenyl) ethyl chloromethyl ketone (TPCK)-treated trypsin (Worthington) was added at a 25:1 (protein:enzyme, w/w) ratio; and the samples were incubated for 16 h at 37 °C. The reaction was quenched by acidifying the samples with formic acid (FA) to a final concentration of 1.0% (pH ≤ 2). The resulting peptides were purified by solid-phase extraction on C18 cartridges (Oasis, Waters, USA), then lyophilized and dissolved for analysis in 0.1% formic acid.

LC-MS/MS Analysis

Urine tryptic peptide fractions were analyzed on a nano-HPLC (high-performance liquid chromatography) Dionex Ultimate 3000 system (Thermo Fisher Scientific, USA) coupled to a timsTOF Pro (Bruker Daltonics, USA) mass spectrometer. HPLC separation was carried out using a packed emitter column (C18, 25 cm × 75 μm 1.6 μm) (Ion Optics, Parkville, Australia), with gradient elution. Mobile phase A was 0.1% formic acid in water; mobile phase B was 0.1% formic acid in acetonitrile. LC separations were performed at a flow of 400 nL/min using a 40 min linear gradient from 2% to 37% solvent B, followed by an LC column wash step (12 min isocratic with 90% solvent B) and equilibration (15 min, isocratic, with 2% solvent B). The MS data were acquired using the ddaPASEF method. Electrospray source (ESI) settings were as follows: capillary voltage 1500 V, dry gas flow - 3.0 L/min at a temperature of 180⁰C. The MS and MS/MS spectra were acquired from 100 to 1700 m, an ion mobility range of 0.6–1.6 1/K0 (V s/cm2). The ion mobility was scanned from 0.6 to 1.6 Vs/cm2. The ramp time was set to 100 ms. The collision energy was ramped linearly as a function of the mobility from 59 eV at 1/K0 = 1.6 Vs/cm2 to 20 eV at 1/K0 = 0.6 Vs/cm2.

Tissue tryptic peptide LC separations were performed at a flow of 400 nL/min using a 90 min linear gradient from 2% to 37% solvent B, followed by an LC column wash step (10 min isocratic with 90% solvent B) and equilibration (15 min, isocratic, with 2% solvent B). The MS data were acquired using the diaPASEF method. Electrospray source (ESI) settings were as follows: capillary voltage 1400 V, dry gas flow - 3.0 L/min at a temperature of 180⁰C. The MS and MS/MS spectra were acquired from 100 to 1700 m, an ion mobility range of 0.6–1.6 1/K0 (V s/cm2). The ion mobility was scanned from 0.6 to 1.6 Vs/cm2. The ramp time was set to 100 ms. The collision energy was ramped linearly as a function of the mobility from 59 eV at 1/K0 = 1.6 Vs/cm2 to 20 eV at 1/K0 = 0.6 Vs/cm2.

Data Analysis

The obtained LC-MS/MS data were analyzed using PEAKS Studio 11 software (BSI, North Waterloo, ON, Canada), using the following parameters: parent ion mass measurement error – 20 ppm; fragment mass error - 0.05 Da. An unspecific restriction search for peptides 6 – 45 amino acids long was conducted across UniprotKB Mus Musculus database with carbamidomethylation (C), oxidation (M) as possible modifications. FDR thresholds were set to 0.1% at the PSM level and 1% at the protein group level, with the requirement of at least one unique and one significant peptide identification. Peaks library (.tsv) for diaPASEF date search was generated by ddaPASEF acquisition of mouse liver pooled samples.

The statistical analysis and data visualization were performed by Python (3.7.3) with the following packages: SciPy [38], Seaborn [39], Matplotlib [40] and Pandas [41]. Significant differences in protein concentrations in the mouse groups were estimated using the Mann–Whitney U-test. We considered concentration values to be statistically significant if the p-value was less than 0.05 and the absolute value of Cohen's d effect size was greater than 0.5. Heat-map hierarchical clustering and principal component analysis (PCA) with t-distributed stochastic neighbor embedding (t-SNE) were used for preliminary estimation of the differences between studied groups and particular samples. For visualizing the clustering of mice in a single group, we used the UMAP dimensionality reduction algorithm [42]. For Lasso regression, the Scikit-Learn [43] package was used, and all data were preprocessed with z-scoring. The default value was chosen for the parameter C, and all possible pairwise comparisons were conducted between all groups. Training was performed on the entire dataset.

3.5. Metabolomic Analysis

Extraction of Metabolites

For metabolites analysis a chloroform-methanol extraction method was used [44]. Samples were mixed with chloroform-methanol (1:2) mixture and incubated for 30 minutes with constant stirring, after that samples were sonicated twice in an ultrasonic bath. Next, water and chloroform were added up to ratio 1/2/0.8 and centrifuged at 3000 rpm. Upper water-methanol phase was evaporated via centrifugal vacuum evaporator and purified using PPL cartridges. Metabolites were eluted with acidic methanol (0.5% FA/MeOH) and evaporated.

FTICR MS Analysis

Mass spectrometric analysis of metabolites was carried out using a FT-ICR MS Apex Ultra device (Bruker Daltonics), with harmonized cell equipped with a 7 T superconducting magnet and electrospray (ESI) ionization source. All mass spectra were acquired in negative ionization mode by direct infusion at flow rate 90 mL/h. The electrospray source (ESI) settings were as follows: capillary voltage - 3500 V, dry gas flow -3 (N2) l/min at a temperature of 200⁰C. Measurements were carried out in the mass/charge (m/z) range from 200 to 1000. Spectra were acquired at 150,000 resolutions (at m/z 400), with 250 scans accumulated for each spectrum.

Data Analysis

Raw FTICR MS data were treated using open-source software and lab-written Python scripts. Visualization of data has been performed with Python library Matplotlib [40]. Statistical analysis has been performed with Python libraries NumPy, Pandas and Seaborn. All *.raw files were converted to *.mzML format using msconvert with continuous wavelet transform algorithm to extract peaks with S/N >3 [45]. Peak lists were extracted using Python script based on pyteomics library [46]. Formulae assignment for all samples was conducted with python script based on open-source NOMspectra python library [47] with a denoising step adapted from Zielinski et al [48]. using the following constraints: O/C ratio ≤ 1, 0.3 < H/C ratio < 2.5; element counts [1 < C ≤ 60, 2< H ≤ 100, 0 < O ≤ 60, N ≤ 3, S ≤ 1]; z = -1 and mass accuracy window <1 ppm after build-in internal calibration based on the construction of the probability density of assignments [49].

On the next step molecular assignments which presented in at least three sample of each group were concatenated in a new group-representative table giving aggregated data. This step aimed to decrease the uncertainty of qualitative comparison raised from the unknown mass balance. Further statistical analysis and evaluation was conducted similar to proteomic data.