Chemicals and reagents. Water and methanol (MeOH) for liquid chromatography–high resolution mass spectrometry (LC–HRMS) were Optima grade from Fisher Scientific (Oslo, Norway), acetonitrile (MeCN) for LC–HRMS was LC–MS grade from Riedel-de Haën (Honeywell, Germany). RPMI 1640 medium with 2 mM L-glutamine, heat-inactivated foetal bovine serum (FBS), (−)-nicotine (≥ 99%), ammonium carbonate (analytical grade), ammonium acetate (LC–MS Ultra grade) and 25% ammonia solution (LiChropur grade) were from Merck (Merck KGaA, Darmstadt, Germany). Water, MeOH and MeCN for extraction and sample preparation, as well as targeted LC–MS/MS, was ULC/MS quality from Biosolve Chimie SARL (Dieuze, France). Methanol calibrant stock solutions (1 mg/mL) of S(−)-nicotine for targeted analyses as well as (−)-cotinine, (±)-nornicotine and 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (nicotine-derived nitrosamine ketone, NNK) were of Cerilliant® quality from Merck. Ammonium formate and formic acid (99%) were ULC/MS quality from Biosolve. HEPES buffer (1 M) in normal saline and sodium pyruvate (100 mM) were from BioWhittaker (Lonza, Basel, Switzerland).
Cell culture and exposure trial design. THP-1 monocytes were obtained from European Collection of Authenticated Cell Cultures (ECACC, Salisbury, UK) and maintained in RPMI 1640 medium with 2mM L-glutamine supplemented with 1 mM sodium pyruvate, 10 mM HEPES and 5% FBS. The monocytes were cultured and kept in an exponential growth phase at a density between 0.25–1 × 106 cells/mL in 20 mL medium in 75cm2 Falcon cell culture flasks (Corning®, VWR International, Radnor, PA, USA) at 37 °C in a humidified incubator in an atmosphere containing 5% CO2.
The THP-1 monocytes were seeded in Corning® Costar® TC-treated 6-well flat bottom plates (VWR International) at a density of 0.5 × 106 cells/mL in 2.5 ml medium per well. After resting for 24 h, the cells were exposed to 5 mM nicotine for 1 or 4 h. Stock solutions of nicotine were freshly made in cell culture medium prior to exposure. Quadruplicate samples were collected directly before exposure, as well as following 1 and 4 h of exposure. Control samples, to which only nicotine-free medium was added, were run in parallel. In addition, three solvent control samples and three medium control samples were included that went through the sample preparation.
Extraction and sample preparation. The extraction and sample preparation procedure was adapted and adjusted from Ser et al. 21. At the end of the exposure experiment, THP-1 cell suspensions were transferred to 15-mL polypropylene tubes and centrifuged to separate cells and medium (4 °C, 50 × g). Aliquots (40 µL) of the medium were transferred to 1.5-mL tubes (Sarstedt, Nümbrecht, Germany) and placed on ice, while the remaining medium was discarded. To the cells, 1 mL of 80% MeOH in water was added, which was pre-cooled at − 80 °C for at least 1 h. The tubes were vortexed for ca. 10 s and then incubated on ice for 15 min followed by centrifugation for 10 min (4 °C, 20,000 × g). The supernatants were transferred to conical glass tubes and evaporated to dryness at 35 °C using a gentle stream of nitrogen. The residue was dissolved in 15 µL of water by vortexing and diluted with 15 µL MeOH/MeCN (1:1). Samples were transferred to chromatography vials with fixed insert and stored at − 80 °C until instrumental analyses.
To the culture medium aliquots (40 µL), 200 µL cold 80% MeOH was added and vortexed for ca. 15 s. The samples were centrifuged for 10 min (4 °C, 20,000 × g). The supernatants (200 µL) were transferred to chromatography vials with fixed insert and evaporated to dryness at 35 °C using a gentle stream of nitrogen. The dry residue was dissolved in 30 µL of water by vortexing and then diluted with 30 µL of MeOH/MeCN (1:1). Samples were stored at − 80 °C until instrumental analyses.
Quality controls. Aliquots (5 µL) of each sample were pooled to obtain a Quality Control (QC) sample, which was used to monitor the system stability through the entire instrumental analysis as well as for data dependent acquisition (DDA).
Untargeted LC–HRMS and LC–HRMS/MS metabolomics. Samples were randomly placed into the autosampler tray of the LC–HRMS instrument and kept at 8 °C. The pooled QC sample was run six times at the beginning of the sequence, and every eighth sample throughout the entire LC–HRMS experiment. LC–HRMS analyses were performed using a Q Exactive™ Hybrid Quadrupole-Orbitrap mass spectrometer equipped with a heated electrospray ion source (HESI-II) and coupled to an ultrahigh-performance liquid chromatography (UHPLC) Vanquish Horizon system (Thermo Fisher Scientific, San Jose, CA, USA). Chromatographic separation was achieved by hydrophilic interaction chromatography (HILIC) using a zwitterionic SeQuant ZIC-pHILIC column (Merck; 150 × 4.6 mm, 5 µm). The column was eluted using a mobile phase consisting of 20 mM ammonium carbonate (pH 8.3) and MeCN (B). The elution proceeded isocratically at a constant flow rate of 0.3 mL/min for 1 min with 80% B, followed by a linear gradient to 20% B over 29 min. Finally, the column was flushed with 8% B for 5 min, returned to the initial conditions, and equilibrated for 9 min. The HRMS instrument was run in full-scan positive and negative ion modes using fast polarity switching in the mass-to-charge (m/z) range 58 to 870. The HESI-II interface was operated at 300°C. The spray voltage was 2.8 and 3.2 kV (positive and negative mode, respectively), the ion transfer capillary temperature was 280°C, the sheath and auxiliary gas flow rates were 35 and 10 units, respectively, and the S-lens RF level was 55%. The automated gain control (AGC) target was set to 5 × 105, and the maximum injection time (IT) was set to 250 ms. A mass resolution of 75,000 full width half maximum (FWHM) at m/z 200 was used. All analyses were performed without a lock mass. Xcalibur software (version 4.6) was used for instrument control and LC–HRMS data acquisition.
Samples for LC–HRMS/MS data acquisition. In addition to the full-scan analyses, a set of LC–HRMS/MS data was acquired for three different samples: the QC sample, the mixture of nicotine-related metabolites, and a pooled sample of nicotine-exposed samples using DDA. The conditions were as follows: full-scan MS/MS product ion spectra were acquired of the top-five most intense MS ions in the mass range m/z 58 to m/z 870 with a mass resolution of 17,500 for product ion detection. Fragmentation was performed by applying three different collision energies (NCE 15, 35, and 65) in separate runs independently for each ionization mode. An inclusion list targeting the m/z and retention time (± 1 min) for each compound in the in-house library was attached to the method. An exclusion list was generated from the most intense ions present in the first blank and attached to the method to eliminate potential contaminants and increase the number of relevant MS/MS spectra.
Processing and quality control of metabolomics data. The raw data from both ionization modes were processed using Compound Discoverer (CD, version 3.3 SP2, Thermo Scientific). First, a retention time alignment was performed with the ChromAlign algorithm 22 followed by peak picking considering a 5 ppm mass tolerance, a 100,000 minimum peak intensity, and including the following types of ions: [M + H]+, [M + NH4]+, [M + Na]+, [M + K]+, [M + MeCN + H]+, [M + Cl]−, [M + MeOH + H]+, [M + H-H2O]+, [M + H − NH3]+, [M − H] −. In the compound-grouping step, the peak area integration was done in extracted ion chromatograms based on the most common ion, with [M + H]+ and [M − H]− as preferred ions, and setting the mass tolerance to 5 ppm and retention time tolerance to 0.2 min. Only peaks with an Original Peak Rating above 5 (in a scale of up to 10) in at least 5 samples were considered for further processing. Missing value imputation was performed using the Random Forest Algorithm 23 considering 100 trees and a maximum number of 10 iterations.
QC correction was performed using the SERRF method 24, considering 200 trees and samples as one batch, and values were interpolated using non-linear regression. The maximum QC RSD allowed before correction was set to 50%, and the maximum after correction was 25%. The minimum percentage of usable hits (actual signals found) per compound in at least one group was 80%. Samples were normalised to the maximum peak area median of all samples and then was log transformed and Pareto scaled for further statistical processing.
Statistical analyses. Univariate statistical analyses. Paired t-tests were performed between each group at each time point using R 25.
Multivariate statistical analyses. Normalised peak areas acquired in CD were imported into SIMCA (version 16.0.2., Sartorius Stedim Biotech, Umea, Sweden), log-transformed and Pareto scaled prior to multivariate analyses. Metabolite patterns and discriminant features were investigated using Orthogonal Partial Least-Squares Discriminant Analysis (OPLS-DA) models. Pair-wise models were created for each group of cell extracts and medium extracts, and for each time point independently. Cross-validation of variance (CV-ANOVA) was performed to evaluate the reliability of the models, and p-values ≤ 0.05 were considered significant. The selection of the most significant features for each significant model was determined combining three different cut-off criteria as follows: a p(corr) value ≥ |0.50|, a variable importance in the projection (VIP value) ≥ 1.5, as well as a p-value of ≤ 0.05 from the univariate test. The p(corr) value represents the correlation and reliability of the data in Y (exposed or non-exposed for the present study).
Molecular Network Analysis. Molecular networking analysis was included within the CD workflow. The parameters were as follows: mass tolerance 2.5 mmu, minimum MSn coverage: 70, minimum fragments: 3, and minimum MSn score: 50. In order to be considered, nodes (compounds) were required to have MSn data and transformation. For exploration of specific nodes, settings were modified dynamically and should be explicitly mentioned, otherwise standard parameters were used.
Annotation and identification. Putative annotations using spectral libraries and in-silico tools. The annotation process of the most relevant metabolites began during the pre-processing in CD with matching the information of the HRMS/MS data obtained from DDA of three different samples, i.e., a pooled QC of cell extract samples and a QC of medium extract samples, a mixture of nicotine metabolite standards containing nicotine, cotinine, nornicotine and nicotine-derived nitrosamine ketone (NKK). Annotations were attained automatically by matching the measured mass spectra to our in-house library (Supplementary Material), comparing mass accuracies, retention time, and isotope patterns 26. Spectral similarity scores between measured features and reference metabolites were determined in CD using the HighChem HighRes algorithm with a precursor mass tolerance of 10 ppm. For the identification of the metabolites, we set a spectral similarity cut-off of 85%. The use of public spectral libraries contained in CD was included in the annotation process to provide tentative annotations.
For the manual annotation of the remaining features with unknown identity, the raw data was transformed into the open-source format,*.msp. Then, the data was imported into the automated class assignment and ontology prediction tool CANOPUS 27 that is an integrative part of the SIRIUS software (v.5.8.3) 28. Using CANOPUS allowed assigning the chemical fingerprint and the chemical class following the ClassyFire ontology 29.
Level of identification. Distinct levels of annotation confidence were assigned to the compounds, following the current guidelines for annotation or identification for metabolomic studies 30. Level 1 corresponds to the unambiguous identification of a metabolite by matching it to a reference compound with at least two independent and orthogonal properties (e.g., retention time and accurate m/z). Level 2 describes the putative identification of metabolites based on the spectral similarities between the HRMS/MS fragmentation data and spectral libraries. Level 3 refers to metabolite classes tentatively characterized by spectral similarity to published HRMS/MS fragmentation data. Level 4 indicates unidentified or unclassified metabolites that are differentiable because of their specific spectral data.
Solid-phase extraction (SPE) of nicotine-related metabolites from cell extracts using cation exchange. For the semi-targeted extraction of nicotine-related metabolites, THP-1 cell lysates were extracted using cation exchange SPE. THP-1 cell suspensions were transferred to 15-mL polypropylene tubes and centrifuged to separate cells and medium (4 °C, 50 × g). The cells were washed with 4 mL PBS, centrifuged and the supernatant discarded. Another 300 µL of PBS were added to the cells and the suspension transferred to a 1.5-mL Eppendorf tube. The cells were lysed by application of a sonication probe using an amplitude of 30% (Sonics & Materials, Inc., Newton, CT, USA). An Oasis MCX mixed-mode cation exchange cartridge (60 mg, Waters Corporation, Milford, MA, USA) was conditioned with 2 mL methanol, followed by 2 mL water and 2 mL of 20 mM ammonium formate (pH adjusted to 2.5 using formic acid). An aliquot of the cell lysate (200 µL) was diluted with 800 µL water, and 100 µL of formic acid added. The diluted extract was applied to the column, which was washed with 20 mM ammonium formate in water, followed by 20 mM ammonium formate/methanol (1:4, v/v). The column was dried under vacuum, and nicotine-related metabolites eluted with 1 mL 5% aqueous ammonia solution/methanol (1:9, v/v).