Comprehensive characterization of volatile and semi-volatile compounds in e-liquids for electronic cigarette using gas chromatography accurate mass spectrometry

Journal of Chromatography A 1703 (2023) 464114 Contents lists available at ScienceDirect Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma Comprehensive characterization of volatile and semi-volatile compounds in e-liquids for electronic cigarette using gas chromatography accurate mass spectrometry M. Cobo Golpe, M. Ramil, I. Rodríguez ∗ Department of Analytical Chemistry, Nutrition and Food Sciences. Research Institute on Chemical and Biological Analysis (IAQBUS). Universidade de Santiago de Compostela, 15782-Santiago de Compostela, Spain a r t i c l e i n f o Article history: Received 31 March 2023 Revised 16 May 2023 Accepted 28 May 2023 Available online 29 May 2023 Keywords: E-liquids Non-target screening Gas chromatography Time-of-flight mass spectrometry a b s t r a c t The consumption of electronic cigarettes is a habit with an increasing prevalence, particularly among youths. Knowing the composition of e-liquids used in these devices represents the first step to understand the potential impact of e-smoking in the health of consumers. Herein, a non-target screening methodology was applied to the identification of volatile and semi-volatile compounds in a set of eliquids from different suppliers, with different flavors, and containing different kinds of additives, such as nicotine or cannabidiol. To this end, samples were characterized by gas chromatography accurate mass spectrometry, using a time-of-flight mass analyzer. Combination of deconvoluted electronic ionization mass spectra with linear retention index values, obtained for two columns with different selectivity, permitted the identification of more than 250 chemicals with different confidence levels. Among them, respiratory pro-inflammatory compounds, acetals of propylene glycol and glycerin with aldehydes, nicotinerelated and non-related alkaloids, and psychoactive cannabinoids were confirmed as concerning compounds in e-liquid samples. Concentration ratios between propylene glycol acetals and parent aldehydes varied in the range from 2% (ethyl vanillin) to more than 80% (case of benzaldehyde). The ratios between the concentrations of delta-9-tetrahydrocannabinol and cannabidiol in e-liquids stayed in the range from 0.02% to 0.3%. ©2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ ) 1. Introduction Electronic cigarettes (e-cig) were marketed as a safe alternative to conventional cigarettes, avoiding, or at least limiting, human exposure to harmful compounds generated during combustion of tobacco, such as tar and polycyclic aromatic hydrocarbons [1] . The power supply unit in e-cig devices is activated by each puff, heating the stream of air which gets in contact with the e-liquid to produce an aerosol containing different flavours and aromatic compounds [2] . Given that temperatures in e-cig devices remain far below those reached during conventional smoking, the formation of toxic combustion by-products is avoided. Composition of liquid solutions vaporized during e-smoking is also different to that of tobacco. Major components are propylene glycol (PG) and vegetal glycerine (VG). In addition, e-liquids contain a large variety of ∗Corresponding author. E-mail address: [email protected] (I. Rodríguez) . flavouring agents [ 3 , 4 ], and some formulations incorporate other compounds, such as nicotine (NIC) [5] and/or cannabidiol (CBD) [6] . Even in these cases, exposure to NIC during e-smoking is reduced compared to smoking, CBD lacks psychotropic effects, and the employed flavours have been approved as food additives. Despite above assumptions, e-smoking rises some concerns, particularly after associations between this habit and outbreaks of lung injury [7] . Risks derived from e-cig consumption are related to several factors. On one hand, some common flavors in e-liquids, such as maltol, cinnamaldehyde and coumarin are known to produce oxidative stress in cultures of pleural cells [8] , through modulation of the formation of reactive oxygen species during their vaporization [ 3 , 9 ]. Thus, they might damage different tissues of the respiratory tract, including oral [10] and lung cells [11] . Furthermore, it was suggested that flavoring agents are transported from lungs to bloodstream, exerting cytotoxic effects in the cardiovascular and immune systems [12] . Another risk factor is the novo formation of concerning chemicals due to reactions among subhttps://doi.org/10.1016/j.chroma.2023.464114 0021-9673/© 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ ) M.C. Golpe, M. Ramil and I. Rodríguez Journal of Chromatography A 1703 (2023) 464114 stances mixed in e-liquids. In this vein, carbonyl species can react with alcohols (mainly with PG) to produce acetals, which are recognized as respiratory irritants. For very popular aldehyde flavors, such as benzaldehyde and vanillin, the rates of conversion to acetals have been estimated between 15% and 30% [13] . A third issue of concern is the generation of toxic species during the vaping process [14] . This latter group includes formaldehyde, acetaldehyde and glyoxal, which are produced from oxidation of base alcohols in e-liquids, as function of the voltage applied by the e-cig battery [15] . Finally, inclusion of addictive species in e-liquids, such as nicotine and cannabinoids, is regarded as an additional risk factor. In line with above concerns, the characterization of flavors existing in e-liquids, and those by-products formed during generation of the aerosol have become recurrent research topics [14] . Particularly, knowing the composition of e-liquids is the first step to identify concerning compounds attending to their toxicity, their addictive potential, and/or their role as precursors of toxic by-products, either generated directly in the e-liquid solutions, or produced during e-smoking. Nuclear magnetic resonance (NMR) spectroscopy and liquid chromatography mass spectrometry (LC-MS) have been proposed for the quantification of major flavors [16] , and the determination of specific families of compounds, such as volatile carbonyl species (after a derivatization reaction), and/or nicotine-related alkaloids in e-liquids [ 17 , 18 ]. However, gas chromatography (GC) MS is regarded as the gold standard for the analysis of e-liquids [ 2 , 19 ], either during quantitation of pre-selected compounds [20] , or for the qualitative characterization of these mixtures [ 2 , 19 , 21 ]. So far, more than 140 volatile compounds have been reported in e-liquids [2] ; however, considering the number of suppliers and the different available flavors, this number is likely below the figures of volatile and semi-volatile organic species included in these formulations. This statement is supported by several facts. First, most GC–MS studies rely on low-polarity, limited selectivity, siloxane-type columns, which fail to separate major components of e-liquids from minor flavors. Second, quadrupole-type MS instruments, employed in most of previous studies [ 2 , 11 , 21 , 22 ], offer a limited sensitivity when operated in the scan mode, so minor compounds might remain undetected, particularly if they co-elute with major e-liquid solvents. Finally, nominal resolution EI-MS spectra might lead to misidentification problems. Despite the improved performance provided by accurate MS platforms, such as GC-TOF-MS and GC-Orbitrap-MS in non-target screening studies, as far as we could trace, they have not been used for the comprehensive characterization of e-liquids. Thus, the aim of the current study is to build a database of accurate EI-MS spectra for volatile and semi-volatile compounds often present in this matrix. To this end, a non-target data mining strategy was combined with the use of GC-TOF-MS, as main determination technique, under different chromatographic conditions. Particular attention was paid to (1) the identification of species related to wellknown additives in e-liquids, such as minor alkaloids structurally related to NIC, and CBD-like compounds; and (2) the assessment of the presence of acetals in e-liquids and/or their potential formation as artefacts during GC–MS analysis. In addition to GC-TOF- MS, LC-MS was employed as complementary analytical technique to confirm those compounds prone to formation in the injector of the GC-TOF-MS instrument. 2. Material and methods 2.1. Samples and sample preparation E-liquids (39 samples) were produced by nine different companies and purchased either from local retail markets, authorized for distribution of conventional cigarettes in Spain, or from on-line suppliers. Fourteen e-liquids contained nicotine (concentrations in the range from 3 to 12 mg mL −1 ), and 6 samples declared the presence of cannabidiol (CBD). The range of flavors stated in e-liquids was relatively large, including tobacco, beverages, fruity and mint as main aroma descriptors. Table S1 summarizes the main features of the considered samples, including the ratio between PG and VG solvents. After acquisition, e-liquid samples were maintained at room temperature for a maximum of 1 month before opening. Thereafter, they were diluted with a suitable organic solvent (ethyl acetate and methanol for GC-TOF-MS and LC-MS analysis, respectively), and stored at −20 ºC until analysis. Unless otherwise stated, the dilution factor employed for GC-TOF-MS analysis was 1:100, whilst a 1:10 sample to solvent ratio was considered in LCMS studies. 2.2. Solvents and standards Ethyl acetate (trace analysis grade quality) was acquired from Merck (Darmstadt, Germany). Methanol (MeOH) and/or acetonitrile (ACN) for LC-MS, were provided by the same supplier. Ultrapure deionized water (18.2 M cm −1 ) was obtained from a Geni U system (Rephile, Shanghai, China). Standards of some tentatively identified flavors, and other volatile compounds detected in e-liquids were provided by Supelco (Bellefonte, PA, USA) and Sigma-Aldrich (St. Louis, Missouri USA). Mixtures of allergen fragrances were supplied by Restek (Bellefonte, PA, USA). Concentrated stock solutions of several cannabinoids, in MeOH, were also provided by Supelco. Individual stock standard solutions were prepared in MeOH, further dilutions and mixtures were made either in ethyl acetate, for GC-TOF-MS analysis, or in MeOH, when using LC-MS as determination technique. A mixture of n-alkanes (C 7 to C 40 , 10 0 0 μg mL −1 in dichloromethane) was provided by Supelco. Benzaldehyde-d 5 (deuterated in the aromatic ring) was purchased from Sigma-Aldrich. 2.3. GC-TOF-MS equipment Identification of volatile compounds in e-liquids diluted solutions was carried out by GC–MS. The employed system was comprised of a 7200 quadrupole time-of-flight (QTOF) instrument (Agilent, Wilmington, DE), equipped with an EI source, and connected to a 7890 GC system from the same supplier. The ionization source and the quadrupole were maintained at 230 °C and 150 °C, respectively. The system was operated in the single MS mode (only the TOF analyzer was operative for separation of ions formed at the EI source), at 2.5 GHz, offering a typical mass resolution of 60 0 0 (calculated as FWHM) at m/z 131. EI-TOF-MS spectra were recorded in the profile mode, within the range of masses comprised between 40 and 600 Da. The e-liquids were analyzed using two capillary columns with different polarities. They were a semi-polar siloxane-type HP5-MS column (30 m x 0.25 mm, 0.25 μm), and a Carbowax, polyethylene glycol-type (DB-WAXetr), polar column with the same dimensions as the HP5-MS one, but a film thickness of 0.50 μm. Both were purchased from Agilent. They were operated using He as carrier gas, at a constant flow of 1 mL min −1 . The temperature of the injector was set at 280 ºC, and the transfer line with the MS instrument maintained at 270 ºC. For the HP5-MS column, a linear gradient at 5 ºC min −1 , from 65 ºC (1 min) to 260 ºC (8 min) was used. With the Carbowax one a first rate of 5 ºC min −1 , from 60 ºC (1 min) to 220 ºC, followed by a second one at 20 ºC min −1 to 240 ºC, were employed. In both cases, the final column temperature was held for different times to adjust the duration of chromatographic runs to 48 min. 2 M.C. Golpe, M. Ramil and I. Rodríguez Journal of Chromatography A 1703 (2023) 464114 2.4. Screening of volatile compounds in e-liquids from GC-TOF-MS data GC-TOF-MS raw data were deconvoluted using the Unknowns Analysis (UA) tool, integrated in the Mass-Hunter software (version 10.0), which controlled all acquisition and data processing parameters of the GC-TOF-MS instrument. The workflow employed during data mining and compounds identification was similar to that described in previous articles [ 23 , 24 ]. In brief, chromatographic peaks with responses (peak heights) above a threshold of 10,0 0 0 counts were deconvoluted, and their accurate EI-MS spectra were compared to those compiled in NIST17 EI-MS library. Identifications were based on the following requirements: (1) spectral match above 75% between experimental and database spectra (spectral comparison was carried out using a mix of forward and reverse modes, with normalized contributions of 0.3 and 0.7, respectively); (2) differences between experimental and database linear retention index (LRI) values below 50 units for each chromatographic column; and (3) mass bias between deconvoluted spectra and calculated m/z, for known fragments in the NIST17 database, lower than 50 ppm, for a minimum of two intense ions. It is worth noting that, LRI values for Carbowax-type columns are relatively scarce compared to those existing for HP5 ones. In the first case, most LRI values were directly obtained from publications dealing with analysis of aroma and volatile compounds in vegetable origin foods [25] and beverages [26] , or using the information accessible through the Chemspider database. In some cases, proposed identities were confirmed against standards. Based on above statements, species in e-liquids were identified with different confidence levels. The highest one (level 1) was assigned to compounds fulfilling the three requirements (spectral, LRI and accurate mass match) and further confirmed against authentic standards with, at least, one of the columns. Level 2 corresponded to compounds not verified against standards, but matching spectral and retention time requirements in both columns. Finally, level 3 was assigned to compounds tentatively identified using just one of the columns. The latter situation mostly corresponds to, highly volatile species, poorly retained by the HP5-MS column; as well as to compounds detected in both columns, but without available reference LRI in one of them (usually the Carbowax one). The above confidence level identification scale was inspired in that proposed by Schymanski et al. [27] , keeping in mind that, the latter has been proposed for MS and product ion scan accurate spectra obtained using soft ionization sources, such as ESI. 2.5. LC-MS confirmatory experiments For certain compounds (identified by GC-TOF-MS at level 1), it was considered relevant to discriminate between their presence in the e-liquids, and their potential formation in the hot injector of the GC-TOF-MS system, through re-arrangements in the chemical structures of precursor species. In these cases, LC-MS was used as additional technique. Separations were carried out using a CORTECS Phenyl column (150 mm x 2.1 mm, 2.7 μm), acquired from Waters (Milford, MA, USA), and operated under reversedphase conditions, with ultrapure water (A) and ACN (B), both 0.1% in FA, as mobile phases. The column was maintained at 40 ºC and the mobile phase flowrate was 0.3 mL min −1 . Different chromatographic gradients, and MS instruments, were considered depending on the investigated species: acetals of aldehydes, or cannabinoids. Further details on specific conditions for identification and/or determination of these compounds are given as supplementary information, Text S1 and S2. The complete data mining workflow, integrating data provided by GC-TOF-MS and LC-MS approaches, is provided as supplementary information, Fig. S1. 3. Results and discussion 3.1. Comprehensive overview of volatile organic compounds in e-liquids Fig. 1A shows the total ion current (TIC) chromatogram for a diluted (1:100) e-liquid (sample code S3, Table S1) using the HP5- MS column. Several chromatographic items, of different intensities, are apparent. Moreover, the region below 11 min presents a broad chromatographic band (corresponding to VG) overlapping Fig. 1. GC-TOF-MS chromatograms for an e-liquid (sample code S3) under different conditions. A, Direct injection of a 100-times diluted sample using the HP5-MS column. B, Liquid-liquid extraction with ethyl acetate after water dilution (solvent to sample ratio 10:1) of same sample using the HP5-MS column. C, Direct injection of the diluted e-liquid (100-folds in ethyl acetate) using the Carbowax-type column. 3 M.C. Golpe, M. Ramil and I. Rodríguez Journal of Chromatography A 1703 (2023) 464114 the peaks of several compounds. Despite the data mining capabilities of the Unknowns Analysis deconvolution algorithm, this band might disturb the identification of minor species; moreover, it prevents the direct injection of e-liquids using lower dilution factors. Thus, different approaches were explored to allow a more comprehensive characterization of e-liquids. In first term, liquid-liquid extraction (LLE) of commercial samples with ethyl acetate (0.1 mL of e-liquids, diluted with 5 mL of ultrapure water and extracted with 1 mL of organic solvent) was assessed [2] . Fig. 1B shows the chromatogram obtained for same e-liquid as that in Fig. 1A after this process. LLE extraction removed the chromatographic band of VG and increased the number of noticeable peaks due to the use of a lower dilution factor; however, polar species present in some eliquids, such as nicotine related alkaloids, remained mostly in the water phase, Fig. 1A and 1B . Table S2 summarizes the efficiency of the LLE process for a selection of compounds with different polarities. In addition to nicotine and nicotyrine, which were hardly recovered, extraction yields below 60% were observed for the acetals of vanillin with PG, several alcohols, such as maltol and benzyl alcohol, and the alkaloid acetyl pyrazine. The second approach involved the testing of an alternative column for compound separation. Fig. 1C shows the chromatogram for same e-liquid (sample code S3, diluted 1:100 in ethyl acetate) with the Carbowax-type column. PG and VG increased their retention times and the latter rendered a narrower band than using the HP5-MS column. Moreover, chromatographic peaks for eliquid components were distributed in a wider interval of retention times. Obviously, certain compounds could still be hindered by those of solvents, as illustrated for nicotyrine, Fig. 1C . In order to obtain a comprehensive characterization of e-liquids, retaining as much as possible of their original composition, LLE was not considered. Samples were just diluted 1:100 with ethyl acetate, injected by duplicate in both columns and the obtained chromatograms processed using the data mining workflow described in Material and Methods section. Table 1 summarizes a list of 82 compounds observed in the chromatograms for the set of 39 e-liquid samples, using both columns, including their detection frequency ranges. These species were identified with a minimum confidence level of 2. Compounds are sorted attending to their increasing LRI in the HP5-MS column. In addition to CAS numbers and empirical formulae, the m/z ratios for the most intense peaks in their EI-MS spectra are provided. The largest number of species in Table 1 fit within the class of terpenes (including terpenes, sesquiterpenes and their hydroxylated terpenoids and sesquiterpenoids), followed by aromatic species and esters of carboxylic acids, Fig. 2A . Regarding their detection frequencies, vanillin and ethyl maltol were the compounds showing the highest prevalence in the set of processed e-liquids, Table 1 . Table S3 and S4 compile a selection of compounds mostly identified with a confidence level 3, and some that were confirmed with standards (confidence level 1). A total of 133 compounds fulfilled the criteria of mass accuracy, spectral match and agreement between experimental and database LRI values using the HP5-MS column, Table S3. In a few cases (i.e. quinolinamine), it was hard to discriminate between positional isomers, since the database LRI values for aminoquinoline analogues, substituted in different positions, differed in just 5–10 units. Most of these compounds were also apparent in the chromatograms corresponding to the Carbowax column; however, reference LRI values were not available for this stationary phase. On the other hand, limited volatility compounds, as it is the case of the cannabinoids and diterpenes, were not observed in chromatograms recorded with the Carbowax column. Likely, they were too strongly retained by this column, with thicker stationary phase and a lower maximum operational temperature than the HP5-MS one. As regards compounds identified only with the polar column, most of them are high volatility alcohols, aromatic compounds (including some solvents such as toluene and styrene), short chain carboxylic acids and some terpenes, Table S4. The latter ones likely co-eluted with more intense species, with several common ions in their spectra, when using the HP5-MS column, hampering their identification. It is worth noting that accuracy of EI-MS spectra was lost in case of saturated peaks; thus, the data mining strategy failed to identify some evident compounds, such as NIC and CBD, unless e-liquids were either injected in the split mode, or diluted up to 10,0 0 0 times. The Venn diagram provided as supplementary information illustrates the advantage of combining the information provided by both columns for the comprehensive characterization of volatile compounds in e-liquids, Fig. S2. Globally, 253 compounds were identified in the processed e-liquids with confidence levels between 1 and 3. A recent review described the identification of 60 different com pounds, including or ganic and inor ganic species, in e-liquids [19] . Using a target GC–MS/MS method, 54 compounds, from a selection of 90 candidates, were confirmed in a set of 25 e-liquids from 5 different brands [20] . The pioneer study by Hutzler et al. [2] reported the identification of 140 organic substances in e-liquids. In summary, the number of compounds found using the non-target data-mining methodology reported in this research is larger than those so far compiled in the existing literature [ 2 , 19 , 20 ]. The accurate EI-MS spectra for compounds identified with confidence levels 1 and 2 have been made available through Zenodo repository ( https://zendo.org ; doi: 10.5281/zenodo. 7738079 ). It is worth noting that, the processed e-liquids showed a high grade of heterogeneity, as regards their GC–MS profiles, see Fig. S3. This variability suggests that the number of species included in the formulation of e-liquids is likely even larger than those identified in this research. Within the set of tested samples, the detection frequency of a selection of aroma compounds suggested as respiratory pro-inflammatories, proved to display cytotoxic effects during in-vitro studies [12] , and identified with a minimum confidence level of 2, are shown in Fig. 2B . Observed values varied from 13% for coumarin up to 68% for ethyl maltol. 3.2. Heterocyclic nitrogenated compounds The main families of heterocyclic nitrogenated compounds identified in e-liquids were 3-pyrrol-pyridine compounds, 3- pyridine substituted species, and pyrazines. Occasionally, aminoquinolines, imidazoles, thiazoles and pyrimidine derivatives were also found. Fig. 3 shows their relative detection frequencies for the groups of nicotine (14 samples) and nicotine-free (25 samples) e-liquids. In addition to pyrrol-3-pyridine type compounds (cotinine, nicotine, myosmine, nicotyrine and nicotine-N-oxide) other heterocyclic species were mostly, or exclusively, found in nicotine-containing e-liquids, as it is the case of 3-pyridine carboxyaldehyde and 3-vinyl-pyridine, quinolinamine (the exact position of the amino group could not be identified), trimethylated-1,8- napthyridine (two isomers were detected) and pyrazino-pyrazine. Other alkaloids with a similar structure, such as anabasine and anatabine, previously reported in nicotine-containing e-liquids [17] , were not found in this research. Minor nicotine related alkaloids might have either a natural origin, as it is the case of myosmine and cotinine [28] , which are probably co-extracted together with nicotine from vegetable sources; or, they could be the result of the oxidation reactions occurring in e-liquid solutions, as reported for nicotyrine [22] . The rest of nitrogenated compounds included in Fig. 3 were less often detected, without a clear prevalence between nicotine-free and containing e-liquids. In the case of pyrazines, their presence was mostly associated to tobacco-flavored e-liquids, rather than to 4 M.C. Golpe, M. Ramil and I. Rodríguez Journal of Chromatography A 1703 (2023) 464114 Table 1 Summary of compounds identified using both columns. Compounds in bold were confirmed against authentic standards. CAS a LRI, HP5-MS column a LRI, Carbowax column m/z ratios of relevant ions b Detection Compound Chemical class number Formula Exp. NIST17 database Exp. Database Ion 1 Ion 2 Ion 3 Frequency Camphene Terpene 79-92-5 C 10 H 16 960 952 1092 1092 93.0691 91.0532 121.0997 + Benzaldehyde Aldehyde 100-52-7 C 7 H 6 O 971 962 1551 1555 105.0329 77.0381 51.0224 + Pyrazine, 2-methoxy-3-methyl- Alkaloid 2847-30-5 C 6 H 8 N 2 O 978 970 1394 1339 124.0618 106.048 95.0292 + Bicyclo[3.1.0]hexane, 4-methylene-1-(1-methylethyl)- ( Terpene 3387-41-5 C 10 H 16 984 974 1125 1120 93.0692 91.0533 77.0364 ++ Hexanoic acid, ethyl ester Ester 123-66-0 C 8 H 16 O 2 1001 1000 1247 1253 88.0497 73.027 99.0784 + Thiazole, 4-methyl-2-(1-methylethyl)- Alkaloid 15679-13-7 C 7 H 11 NS 1021 1022 1369 1339 126.0362 141.0599 70.9939 + 2-Acetylpyrazine Alkaloid 22047-25-2 C 6 H 6 N 2 O 1025 1023 1657 1646 94.0519 122.0476 80.0364 + p-Cymene Aromatic 99-87-6 C 10 H 14 1027 1025 1291 1292 119.0845 134.1071 91.0538 + D-Limonene Terpene 5989-27-5 C 10 H 16 1031 1018 1217 1219 67.0544 93.0693 121.1003 ++ Eucalyptol Terpenoid 470-82-6 C 10 H 18 O 1035 1032 1227 1228 93.0692 139.1113 154.1342 + 2-acetylpyridine Alkaloid 1122-62-9 C 7 H 7 NO 1036 1035 1632 1603 79.0412 121.0516 93.0572 + 2-Cyclopenten-1-one, 2-hydroxy-3-methyl- Ketone 80-71-7 C 6 H 8 O 2 1037 1034 1852 1839 112.052 84.0564 69.0335 ++ Benzyl alcohol Aromatic 100-51-6 C 7 H 8 O 1038 1036 1899 1917 79.0533 108.056 107.056 ++ Butanoic acid, 3-methyl-, butyl ester Ester 109-19-3 C 9 H 18 O 2 1047 1047 1261 1271 103.0748 57.0695 85.0642 + Butanoic acid, 3-methylbutyl ester Ester 106-27-4 C 9 H 18 O 2 1056 1056 1279 1256 70.0771 89.0587 115.074 + Furaneol Ketone 3658-77-3 C 6 H 8 O 3 1059 1070 2052 2050 128.0466 57.0331 85.0279 + Gamma-Terpinene Terpene 99-85-4 C 10 H 16 1060 1060 1302 1263 93.0688 121.0998 136.1232 + Pentanoic acid, 4-oxo-, ethyl ester Ester 539-88-8 C 7 H 12 O 3 1063 1045 1636 1610 99.0434 129.0545 71.0481 + Diethyl malonate Ester 105-53-3 C 7 H 12 O 4 1072 1069 1597 1603 115.0381 133.0491 87.0069 + Hexanoic acid, 2-propenyl ester Ester 123-68-2 C 9 H 16 O 2 1082 1080 1387 1370 71.0849 99.0801 113.0597 + Pyrazine, 2,3,5,6-tetramethyl- Alkaloid 1124-11-4 C 8 H 12 N 2 1089 1089 1494 1466 136.0999 54.0466 94.0653 + Terpinolene Terpene 586-62-9 C 10 H 16 1091 1088 1302 1293 93.0662 121.0986 136.1223 + Butanoic acid, 2-methyl-, 3-methylbutyl ester Ester 27625-35-0 C 10 H 20 O 2 1102 1101 1291 1276 103.0729 70.077 57.0695 + Linalool Terpenoid 78-70-6 C 10 H 18 O 1103 1099 1561 1565 93.0698 121.1002 71.0482 ++ Butanoic acid, 3-methyl-, 3-methylbutyl ester Ester 659-70-1 C 10 H 20 O 2 1107 1104 1308 1312 103.0736 85.0648 70.0772 + (2R,4R)-4-Methyl-2-(2-methylprop-1-en-1-yl)tetrahydro-2H-pyran Terpenoid 5258-11-7 C 10 H 18 O 1113 1127 1350 1371 139.1116 69.0326 83.0481 + Maltol ketone 118-71-8 C 6 H 6 O 3 1114 1110 1991 1991 126.0304 71.0117 97.0275 ++ Phenyl ethyl alcohol Aromatic 60-12-8 C 8 H 10 O 1120 1116 1935 1941 91.0541 92.0613 122.0731 + Isophorone Aldehyde 78-59-1 C 9 H 14 O 1125 1124 1623 1623 82.0413 54.0468 138.1045 + Benzoic acid Aromatic 65-85-0 C 7 H 6 O 2 1160 1170 2460 2475 105.0338 122.0371 77.0377 + Benzene, 1,3-dimethoxy- Aromatic 151-10-0 C 8 H 10 O 2 1170 1166 1775 1730 138.0677 109.0642 78.0453 + endo-Borneol Terpenoid 507-70-0 C 10 H 18 O 1171 1167 1723 1726 95.0851 67.054 110.1081 + Benzenemethanol, .alpha.,.alpha.,4-trimethyl- Aromatic 1197-01-9 C 10 H 14 O 1189 1183 1869 1851 135.0802 91.0534 115.0537 + Butanoic acid, hexyl ester Ester 2639-63-6 C 10 H 20 O 2 1193 1192 1430 1410 93.0691 121.1009 136.1245 + Menthol Terpenoid 89-78-1 C 10 H 20 O 1194 1174 1635 1626 81.0701 95.0859 123.1174 + alpha.-Terpineol Terpenoid 98-55-5 C 10 H 18 O 1195 1190 1718 1719 93.0672 121.1005 136.1247 ++ Ethyl maltol Ketone 4940-11-8 C 7 H 8 O 3 1200 1199 2041 2047 140.0463 139.0387 97.0282 +++ Gamma-Terpineol Terpenoid 586-81-2 C 10 H 18 O 1202 1197 1690 1690 121.1006 93.0694 136.1244 + 2-Cyclohexen-1-ol, 2-methyl-5-(1-methylethenyl)-, cis- Terpenoid 1197-06-4 C 10 H 16 O 1223 1229 1853 1845 109.0634 91.0525 84.0553 + Citronellol Terpenoid 106-22-9 C 10 H 20 O 1230 1228 1780 1776 67.0542 81.0699 123.116 + Nerol Terpenoid 106-25-2 C 10 H 18 O 1232 1228 1862 1818 69.0708 93.0705 121.1025 + Pulegone Terpenoid 89-82-7 C 10 H 16 O 1244 1237 1676 1665 81.0696 152.1199 137.0961 + Benzeneacetic acid, ethyl ester Aromatic 101-97-3 C 10 H 12 O 2 1248 1246 1783 1784 91.0536 164.0836 136.0538 + Linalyl acetate Ester 115-95-7 C 12 H 20 O 2 1257 1257 1561 1573 93.0703 121.102 80.0624 + 2-Cyclohexen-1-one, 3-methyl-6-(1-methylethyl)- Terpenoid 89-81-6 C 10 H 16 O 1259 1253 1757 1841 110.0733 95.0495 137.097 + Gamma-Octalactone Lactone 104-50-7 C 8 H 14 O 2 1262 1261 1948 1946 85.0289 100.0526 57.0337 + 1,3-Dioxolane, 4-methyl-2-phenyl- Acetal 2568-25-4 C 10 H 12 O 2 1271 1272 1840 1871 163.0752 105.0333 119.0464 ++ Cinnamaldehyde, (E)- Aldehyde 104-55-2 C 9 H 8 O 1274 1274 2043 2037 131.0507 103.0518 77.0381 + 5-Thiazoleethanol, 4-methyl Alkaloid 137-00-8 C 6 H 9 NOS 1281 1288 2332 2311 112.0226 143.0417 85.0111 + Anethole Aromatic 4180-23-8 C 10 H 12 O 1289 1284 1854 1817 148.0897 147.0822 117.0708 + Nonanoic acid, ethyl ester Ester 123-29-5 C 11 H 22 O 2 1297 1296 1520 1526 88.0508 73.0272 61.0281 + Thymol Terpenoid 89-83-8 C 10 H 14 O 1303 1291 2234 2236 135.0798 150.1025 107.0836 + ( continued on next page ) 5 M.C. Golpe, M. Ramil and I. Rodríguez Journal of Chromatography A 1703 (2023) 464114 Table 1 ( continued ) CAS a LRI, HP5-MS column a LRI, Carbowax column m/z ratios of relevant ions b Detection Compound Chemical class number Formula Exp. NIST17 database Exp. Database Ion 1 Ion 2 Ion 3 Frequency Piperonal Aldehyde 120-57-0 C 8 H 6 O 3 1337 1330 2268 2279 149.0224 150.0298 121.0274 + Methyl anthranilate Aromatic 134-20-3 C 8 H 9 NO 2 1345 1343 2271 2283 119.0371 151.0636 92.0268 + Benzyl butyrate Ester 103-37-7 C 11 H 14 O 2 1349 1345 1894 1866 108.0568 91.0538 178.087 + 4-Acetylanisole Ketone 100-06-1 C 9 H 10 O 2 1356 1348 2173 2173 135.0441 150.0673 107.0489 + Eugenol Aromatic 97-53-0 C 10 H 12 O 2 1361 1357 2193 2207 164.0829 149.0582 131.0486 + Nerol acetate Ester 141-12-8 C 12 H 20 O 2 1385 1364 1774 1776 69.0698 93.0697 121.1008 + Hydrocoumarin Aromatic 119-84-6 C 9 H 8 O 2 1388 1387 2269 2269 148.0524 120.056 91.0537 + 2-Buten-1-one, 1-(2,6,6-trimethyl-1,3-cyclohexadien-1-yl)-, (E)- Ketone 23726-93-4 C 13 H 18 O 1389 1386 1823 1847 121.1007 175.1108 105.069 + beta.-Bourbonene Sesquiterpene 5208-59-3 C 15 H 24 1394 1384 1540 1539 81.0687 123.1158 161.1311 + Vanillin Aldehyde 121-33-5 C 8 H 8 O 3 1401 1404 2602 2623 151.0393 152.0467 123.044 +++ Diphenyl ether Ether 101-84-8 C 12 H 10 O 1405 1404 2040 2017 170.0723 141.07 115.0541 + 2-Buten-1-one, 1-(2,6,6-trimethyl-1-cyclohexen-1-yl)- Ketone 35044-68-9 C 13 H 20 O 1419 1417 1841 1836 177.1274 192.151 123.1158 + Alpha-cedrene Sesquiterpene 469-61-4 C 15 H 24 1419 1411 1592 1599 119.0847 161.1306 204.1864 + Myosmine Alkaloid 532-12-7 C 9 H 10 N 2 1426 1427 2231 2236 145.0762 118.0536 146.0836 ++ Cyclohexanepropanoic acid, 2-propenyl ester Ester 2705-87-5 C 12 H 20 O 2 1426 1435 1820 1800 121.101 95.0851 113.0594 + cis-Thujopsene Sesquiterpene 470-40-6 C 15 H 24 1438 1429 1627 1592 119.084 105.0685 91.0524 + Coumarin Aromatic 91-64-5 C 9 H 6 O 2 1441 1441 2501 2519 146.0354 118.0407 90.0456 + Ethyl Vanillin Aldehyde 121-32-4 C 9 H 10 O 3 1459 1453 2556 2570 137.0235 166.0626 109.028 + Humulene Sesquiterpene 6753-98-6 C 15 H 24 1460 1454 1693 1682 93.0691 147.1155 121.0998 + Gamma-decalactone Lactone 706-14-9 C 10 H 18 O 2 1471 1470 2176 2179 85.029 57.0333 128.0824 ++ (1 ´ s, 2 ´ s)-Nicotine-N ´ -oxide Alkaloid 29419-55-4 C 10 H 14 N 2 O 1479 n.a. 2203 n.a. 118.0652 119.0722 60.0439 ++ trans-.beta.-Ionone Ketone 79-77-6 C 13 H 20 O 1491 1486 1841 1863 177.1264 91.053 135.0787 + Delta-decalactone Lactone 705-86-2 C 10 H 18 O 2 1499 1496 2228 2201 71.0484 99.0441 43.0542 + Isoeugenol methyl ether Aromatic 93-16-3 C 11 H 14 O 2 1500 1492 2118 2176 178.0985 107.0536 163.0752 + 2H-1-Benzopyran-2-one, 6-methyl- Aromatic 92-48-8 C 10 H 8 O 2 1563 1574 2551 2629 160.0524 132.0561 131.0489 + Gamma-undecalactone Lactone 104-67-6 C 11 H 20 O 2 1576 1576 2289 2259 85.028 95.0482 57.033 + Hedione Ester 24851-98-7 C 13 H 22 O 3 1658 1649 2307 2229 83.0488 153.1275 156.0782 + Delta.-Dodecalactone Lactone 713-95-1 C 12 H 22 O 2 1712 1719 2465 2467 71.0481 99.0433 55.0533 + Benzyl Benzoate Ester 120-51-4 C 14 H 12 O 2 1770 1762 2672 2693 105.0329 194.0721 167.0833 + Phytol Diterpenoid 150-86-7 C 20 H 40 O 2090 2114 2621 2633 123.115 95.0845 81.0689 + a LRI, linear retention index. b Frequency of detection: + ( < 30% of samples), ++ (30%- 60% samples), +++ ( > 60% samples). 6 M.C. Golpe, M. Ramil and I. Rodríguez Journal of Chromatography A 1703 (2023) 464114 Fig. 2. A, Distribution of compounds identified with both column in chemical classes. B, Detection frequencies of respiratory pro-inflammatory compounds. nicotine containing samples. Despite its similar chemical structure, methyl nicotinate appears to have a different origin than the rest of nicotine-like compounds, since it was identified in a limited number of both kinds of e-liquids. This compound has been correlated with dermatitis problems when included in topic use pharmaceutical preparations [29] ; thus, it might behave also as an irritant of the respiratory system. 3.3. Acetals Acetals are respiratory irritant compounds. Their presence in eliquids is the result of reactions between aldehydes and alcohols. The data mining strategy described in the Material and Methods section, combined with manual inspection of GC-TOF-MS records, showed the frequent presence of two clusters of signals corresponding to the diastereomers of acetals formed between main aldehydes in e-liquids (i.e., vanillin and ethylvanillin) with PG (2 peaks) and VG (4 peaks). In the latter case, two different acetals corresponding to 5 and 6 member rings are possible. Fig. 4 plots the extracted ion chromatograms (obtained with the HP-5 MS column) and the accurate scan spectra for two representative species. Whatever their retention times, the congeners in both groups of vanillin acetals (with PG, or VG) showed identical spectra. They share the same fragment at m/z 151 (nominal mass) correspond- 7 M.C. Golpe, M. Ramil and I. Rodríguez Journal of Chromatography A 1703 (2023) 464114 Fig. 3. Detection frequencies of heterocyclic nitrogenated compounds in nicotine and nicotine-free e-liquids. Fig. 4. A, EIC chromatogram at m/z 151.0390 (mass window 50 ppm) for sample S19. Accurate EIMS spectra for the acetals of vanillin with PG (B) and VG (C). 8 M.C. Golpe, M. Ramil and I. Rodríguez Journal of Chromatography A 1703 (2023) 464114 Table 2 Concentrations of vanillin, ethyl vanillin, benzaldehyde and their acetals with PG in a selection of e-liquid samples. Average values for duplicate injections. Data in μg mL −1 . Sample code Brand Vanillin Ethyl Vanillin Benzaldehyde Vanillin Acetal Ethyl Vanillin Acetal Benzaldehyde acetal 1 A 754.3 33.4 2 A 937.6 37.8 3 A 1922.9 2447.0 77 121.1 4 A 5365.0 1050.6 112.7 26.4 5 B 467.8 15.6 10 A 672.6 46.8 30.4 24.2 12 D 3752.9 625.5 217.1 29.6 19 F 3923.0 485.4 225.1 23.4 20 G 701.4 21.6 21 G 4736.7 1051.1 283.8 58.2 25 G 595.3 22.9 30 G 487.8 289.5 13.5 9.6 31 G 277.1 98.5 14.25 81.0 34 H 489.9 15.7 ing to the deprotonated parent aldehyde ([M-H] + ion of vanillin); moreover, other ions, with calculated exact m/z values of 87.0441 and 103.0393, are compatible with the non-aromatic rings (C 4 H 7 O 2 and C 4 H 7 O 3 moieties) in the structures of vanillin-PG and vanillin- VG acetals. Same chromatographic and spectral behavior was observed for ethyl vanillin acetals. In case of benzaldehyde, a single peak corresponding to its acetal with PG was detected in the chromatograms for e-liquids, figure not shown. The identities of the acetals of vanillin, ethyl vanillin and benzaldehyde with PG were confirmed with commercial standards. Verification of their presence in e-liquids, instead of formation at the injection of the GC–MS system, was achieved by LC-TOF-MS. Considering vanillin as potential precursor aldehyde, the acetals of this compound with PG and VG showed single peaks, at retention times slightly higher and lower than that of vanillin, respectively. The ESI ( + ) product ion (MS/MS) spectra of both acetals showed a product ion at m/z 153.0547, matching the [ M + H ] + ion of vanillin, Fig. S4. Further research to detect partial transformation of aldehydes to acetals, in the injector of the GC–MS instrument was carried out by adding benzaldehyde-d 5 (addition level 100 μg mL −1 ) to ethyl acetate diluted e-liquids, just before GC– MS analysis. The obtained GC-TOF-MS chromatograms did not contain any extra peaks, neither at nominal m/z 168 not at m/z 184, corresponding to the acetals of benzaldehyde-d 5 with PG and VG, respectively. So, we concluded that acetals of the above aldehydes existed in e-liquids and that, they were not artifacts associated to thermally induced reactions in the injector of the GC-TOF-MS system. Table 2 summarizes the concentrations of vanillin, ethyl vanillin and benzaldehyde together to those for their acetals with PG in the set of processed e-liquids. Concentrations were determined against solvent-based standards, considering split and splitless injection of diluted samples (1:100), for aldehydes and acetals respectively, and using benzaldehyde-d 5 as internal standard. Levels of vanillin and ethyl vanillin in positive samples varied between 0.3 mg mL −1 and 5.4 mg mL −1 , within the range of values reported in previous studies developed in Europe [20] , and below maximum levels (33 mg mL −1 for vanillin) reported for samples purchased in USA [11] . The concentrations of their acetals with PG (as sum of diastereomers) represented between 2% and 6% of the content of parent aldehydes. These percentages are significantly lower than those reported in laboratory experiments, combining aldehydes with e-liquids containing different percentages of PG [13] . Benzaldehyde was found just in two out of 39 samples; however, the ratio between the concentration of its acetal with PG and the free aldehyde (values from 0.52 to 0.83) were higher than those observed for vanillin and ethyl vanillin. 3.4. Cannabinoids Among commercially available e-liquids, those containing cannabidiol (CBD) are gaining popularity. This fact raises concerns in relation to consumers exposure to other cannabinoids, with potential psychotropic effects, either existing in cannabis extracts, or formed from transformation of CBD during consumption of ecigarettes [30] . Fig. 5 shows the EI-MS spectra and the chemical structures assigned to cannabinoids identified in the GC-TOF-MS chromatograms for this type of e-liquids. Cannabinoids were detected only using the HP-5MS column due to their limited volatility. Fig. S5 shows the chromatographic profiles for the major congeners within this group of compounds. Two bicyclic species, with the same chemical structure of CBD, but shorter alkyl chains were identified as cannabidivarin (CBDV), and the homologue species with a butyl group bonded to carbon number 5 in the aromatic ring, Fig. 5A and 5B . Their structures were not confirmed with authentic standards; however little doubt exists on their identities considering differences between ions existing in their EI-MS spectra and that of CBD. To sum up, the radical cation [(M . + )] and the base peak in the spectra of these compounds are shifted in multiples of 14.0156 Da, versus those in the spectrum of CBD, Fig. 5A to 5C . The presence of cannabinol (CBN, C 21 H 26 O 2 ) was first identified, and then confirmed in e-liquids. Its spectrum is shown in Fig. 5D . A second group of cannabinoids, with the same empirical formula of CBD (C 21 H 30 O 2 ), were cannabichromene (CBC) Fig. 5E , with a similar spectrum but a most intense fragment at m/z 174, and a slightly higher retention time than CBD (Fig. S4), and delta- 9-tetrahydrocannabinol ( 9-THC) Fig. 5H . In addition, two relatively intense peaks, at longer retention times than that of CBD (Fig. S5), corresponding to three oxygenated species (C 21 H 30 O 3 ) with identical EI-MS spectra, were also detected, Fig. 5F and 5G . Identifications derived from NIST database search pointed out to cannabielsoin (CBE) as the most probable candidate for both peaks, whilst the retention time for a commercial CBE standard matched that obtained for the first species. CBE was already described by Tsujikawa et al. [31] in CBD-type e-liquids. Taking into account reports describing the partial, and variable, decomposition of CBD in the injector of GC instruments (the percentage of CBD conversion to 9-THC varies depending on temperature and injection mode) [31] , and the very low intensity of the peak assigned to 9-THC compared to that of CBD (Fig. S5), it was hard to determine whether the former compound (the only psychoactive cannabinol identified in this research) is present in e-liquids, or if it was an artifact generated in the injector of the GC-TOF-MS system. Thus, LC-MS/MS was used as complementary technique. Under conditions given in the supplementary informa- 9