Marleen Vetter, Steffen Bräkling, Joël Fritschi, and Sonja Klee
TOFWERK, Thun, Switzerland
Introduction
The field of metabolomics has witnessed an exponential growth over the last decades, driven by advances in instrumentation and resulting in new applications in a wide range of areas. Yet, a comprehensive analysis of all metabolites in cells, tissue, or body fluids is still challenging due to the complexity of the metabolome. In metabolomic research, both targeted and non-targeted approaches are usually employed. Whereas the targeted approach aims for a quantitative measurement of selected groups of metabolites and requires prior knowledge of the substances of interest, the non-targeted approach is a global profiling of the metabolome [1,2].
The identification of disease biomarkers is one of the most important goals of metabolomics. As a non-invasive and a cost-effective tool for screening, diagnosis and monitoring of diseases, urine analysis is the second most frequently performed medical analysis used for diagnostic purposes after the analysis of blood [3,4]. Urinary metabolites provide indications on diseases and tissue injuries, which can be used to determine the type of pathology, environmental influences, toxins, nutritional problems, cancer, or diabetes [5].
Mass spectrometry (MS) in combination with chromatographic separation is one of the major techniques for the analysis of metabolites in complex biological matrices. Gas chromatography mass spectrometry (GC-MS) plays an increasingly significant role in the field of metabolomic analysis [6]. This study highlights use of the ecTOF coupled with a gas chromatograph (GC-ecTOF) for metabolomic research. The unique feature of the ecTOF to acquire both chemical ionization (CI) and 70 eV electron ionization (EI) mass spectra in parallel in one chromatographic experiment provides additional information needed for complex non-targeted analysis such as metabolomic profiling [7].
Method
A standard of 40 endogenous steroids was purchased from Steraloids, Newport RI, USA. An anonymized urine sample of a patient was also provided by the Inselspital (Insel Gruppe AG, Bern, CH). Sample preparation was performed in the Department of Nephrology and Hypertension of the Bern University Hospital (Inselspital) (c.f. Bileck et al. (2018) [8]). The instrumental conditions for the GC, as routinely employed at Inselspital, and the ecTOF conditions can be found in Table 1. Medical-ethical guidelines from the Swiss Academy of Medical Sciences as implemented at the Inselspital were followed during sample collection and analysis.
| Injection | 2 µL (direct splitless) |
| Inlet Temperature | 260 °C |
| Carrier Gas Flow | 1.2 mL/min sccm He |
| Purge Flow | 10.0 mL/min |
| Column | HP5-1MS (15 m, 0.25 mm, 25 µm; Agilent Technologies) |
| Septum Purge | 3.0 mL/min |
| Temperature Program | 50 °C for 2 mins, 30 °C/min to 210 °C, 2 °C/min to 265 °C |
| Flow Split | 1:1 CI/EI |
| Heated Transfer Line Temperature | 280 °C |
| Source Temperature EI | 280 °C |
| Source Temperature CI | 300 °C |
| Ionization Sources | StarBeam 70 eV EI source HRP CI source reagent: NH2+ (N2) [9] |
| Mass Range | 1-700 m/Q (Th) |
Employing the ecTOF for Known and Unknown Identification
Routine Target Analysis
Figure 1 shows the EI and CI traces of both the reference standard (a) and the patient sample (b). For ease of visualization, the EI extracted ion chromatograms of m/Q 73.0472 are shown in Figure 1 (c,d). The fragment ion arises from the derivatization reagent (N-trimethylsilylimidazole (TMSI)). Compounds containing at least one trimethylsilyl (TMS) group show a prominent m/Q 73.0472 (Si(CH3)3+). The specific formation mechanism of the fragments is described in the literature [10,11]. The 40 target steroids are highlighted as green. As can be seen in Figure 1, the ecTOF was able to identify and detect all 40 steroids, demonstrating that the ecTOF can be used for routine targeted analysis (Table 2, Figure 1).
The same chromatographic run of the ecTOF can also be used to provide information on additional steroids and other compounds present in patient urine samples. Conventionally, unknowns are identified using EI information, particularly comparison to standards and database matches. However, when no standard is available and/or the library match is insufficient, the unique feature of the ecTOF to also provide CI information is invaluable. This is exemplified using two example compounds from the patient’s urine sample. Potential steroids were initially identified using exclusion factors (signal higher than m/Q 400 in the CI data, presence of the m/Q 73.0474 (Si(CH3)3) fragment as well as a neutral loss of m/Q 90.0500 (SiC3H9OH) in the EI as an indication of derivatisation). In Figure 1b, these potential steroids are highlighted in orange, two of which are discussed in more detail (highlighted in red as “1” and “2”).
| No. | Trivial Name | Abbreviation | Comment |
| 1 | Androsterone | ANDRO | |
| 2 | Etiocholanolone | ETIO | Coelution with Dihydroandrosterone |
| 3 | Dihydroandrosterone, Androstanediol | DH-ANDRO | Coelution with Etiocholanolone |
| 4 | Dehydroepiandrosterone | DHEA | |
| 5 | Androstenediol | ANDRO-DIOL | |
| 6 | 5α-Dihydrotestosterone | 5a-DHTST | Coelution with 11-oxo-Etiocholanolone |
| 7 | 11-oxo-Etiocholanolone | 11-OXO-ETIO | Coelution with 5α-Dihydrotestosterone |
| 8 | 17β-Estradiol | 17b-ESTRADIOL | |
| 9 | Testosterone | TESTOSTERONE | |
| 10 | 11β-Hydroxyandrosterone | 11b-OH-ANDRO | |
| 11 | 11β-Hydroxyetiocholanolone | 11b-OH-ETIO | Coelution with 11β-Hydroxyetiocholanolone |
| 12 | 17-Hydroxypregnanolone | 17-HP | Coelution with 11β-Hydroxyetiocholanolone |
| 13 | 16α-Hydroxy-DHEA | 16a-OH-DHA | |
| 14 | Pregnanediol | PD | |
| 15 | Pregnanetriol | PT | |
| 16 | Androstenetriol | 5-AT | |
| 17 | Tetrahydro-11-deoxycortisol | THS | |
| 18 | Tetrahydrodeoxycorticosterone | THDOC | Coelution with Estriol |
| 19 | Estriol | ESTRIOL | Coelution with Tetrahydrodeoxycorticosterone |
| 20 | Pregnanetriolone | PTONE | |
| 21 | Pregnenetriol | 5-PT | Coelution with Tetrahydrocortisone |
| 22 | Tetrahydrocortisone | THE | Coelution with Pregnenetriol |
| 23 | Tetrahydro-11-dehydrocorticosterone | THA | |
| 24 | Tetrahydrocorticosterone | THB | |
| 25 | 5α-Tetrahydrocorticosterone | 5a-THB | |
| 26 | Tetrahydrocortisol | THF | |
| 27 | 5α-Tetrahydrocortisol | 5a-THF | |
| 28 | α-Cortolone | a-CORTOLONE | |
| 29 | Tetrahydroaldosterone | THALDO | |
| 30 | β-Cortol | b-CORTOL | |
| 31 | β-Cortolone | b-CORTOLONE | |
| 32 | α-Cortol | a-CORTOL | |
| 33 | Cortisone | CORTISONE | |
| 34 | Cortisol | CORTISOL | |
| 35 | 20β-Dihydrocortisone | 20b-DHE | |
| 36 | 20α-Dihydrocortisone | 20a-DHE | |
| 37 | 20β-Dihydrocortisol | 20b-DHF | |
| 38 | 6β-Hydroxycortisol | 6b-OH-F | Coelution with 18-Hydroxycortisol |
| 39 | 18-Hydroxycortisol | 18-OH-F | Coelution with 6β-Hydroxycortisol |
| 40 | 20α-Dihydrocortisol | 20a-DHF |
Increasing Confidence in Library Matches
The first peak (Peak “1” in Figure 1d) elutes at 15.2 minutes. Figure 2a shows the EI and corresponding CI mass spectra for this peak. The NIST hit provides a fair match (match factor 765, reverse match factor 788), and the probability of 95.9 % for enterodiol (2,3-bis(3-hydroxybenzyl) butane-1,4-diol tetratms) (Figure 2). Using the NIST hybrid similarity search without additional CI information, the derivatized enterodiol is still the best match, however match factors do not considerably improve (Match 748, reverse Match 762, hybrid search match factor (o. Match) 688) with a delta mass of m/Q – 90.0500.
The main issue with the match results is the missing M-CH3 peak at m/Q 575.2821 in the EI mass spectrum, and the loss of molecular mass information using 70 eV ionization for derivatized enterodiol. Including the additional CI information in the NIST hybrid similarity search, the match factor for derivatized enterodiol can be improved to 830, reverse match factor to 847 and o. Match to 768 and both the [M-CH3]+ as well as the [M+H]+ can be identified. In conjunction with the accurate mass and isotopic ratios information (Figure 2c,d), this information can be used to support and improve identification confidence in the NIST library search results.
Tentative Compound Structure without Library Match
Furthermore, the combination of EI and CI information can be used to provide sum formulas and structures for compounds not in the NIST library. As an example, we will discuss the peak at 12.5 minutes (Peak “2”, Figure 1d, Retention Indexof 2646). The EI and CI mass spectra for this compound are found in Figure 3a.The best EI NIST library hit is thioinosine, 3 TMS (Figure 3b). However, this identification is not supported by the CI information which indicates [M+H]+ of m/Q 484.2311 and [M-CH3]+ of m/Q 468.2005 (a typical fragment in CI for TMS derivatized compounds). The EI does not yield any molecular ion. Since no EI library hit corresponds to the CI derived sum formula, the target compound is probably not listed in the library. Including the information of the molecular mass to the hybrid similarity search option of the NIST [Version 2.4] search software, thioinosine with 3 TMS groups (due to derivatization) still provides the best hit but does not explain the delta mass of -17 Da observed (Table 2). The results of the hybrid similarity search are given in Table 2. Applying the nitrogen rule the molecular ion suggests an uneven number of nitrogen atoms in the molecule, which is not the case for thioinosine, 3 TMS [12]. Using the accurate mass information of the CI spectrum, we can deduce that the SH group was replaced by a NH2 group, thus explaining the mass difference of -17 Da to the library hit (Table 3). Since the purine derived group (C5H3N4S m/Q 151.0078) does not greatly contribute to the fragmentation pattern of thioinosine, 3 TMS, these adenosine derivates show very similar EI mass spectra.
In addition, the combination of EI and CI information can be used to provide sum formulas and structures for compounds not in the NIST library. As an example, we will discuss the peak at 12.5 minutes (Peak “2”, Figure 1d, Retention Indexof 2646). The EI and CI mass spectra for this compound are found in Figure 3a.The best EI NIST library hit is thioinosine, 3 TMS (Figure 3b). However, this identification is not supported by the CI information which indicates [M+H]+ of m/Q 484.2311 and [M-CH3]+ of m/Q 468.2005 (a typical fragment in CI for TMS derivatized compounds). The EI does not yield any molecular ion. Since no EI library hit corresponds to the CI derived sum formula, the target compound is probably not listed in the library. Including the information of the molecular mass to the hybrid similarity search option of the NIST [Version 2.4] search software, thioinosine with 3 TMS groups (due to derivatization) still provides the best hit but does not explain the delta mass of -17 Da observed (Table 2). The results of the hybrid similarity search are given in Table 2.
Applying the nitrogen rule the molecular ion suggests an uneven number of nitrogen atoms in the molecule, which is not the case for thioinosine, 3 TMS [12]. Using the accurate mass information of the CI spectrum, we can deduce that the SH group was replaced by a NH2 group, thus explaining the mass difference of -17 Da to the library hit (Table 3). Since the purine derived group (C5H3N4S m/Q 151.0078) does not greatly contribute to the fragmentation pattern of thioinosine, 3 TMS, these adenosine derivates show very similar EI mass spectra.
| Reverse NIST Hit | Name and Retention Index | Match Factor/ o. Match Factor | Delta mass |
| 1 | Thioinosine, 3TMS derivative (RI 2760) | 918/ 808 | -17 (-SH replaced with -NH2) |
| 2 | 1-Methyladenosine, tris(trimethylsilyl) ether (RI 2669) | 887/ 808 | -14 (-CH3 replaced with -H) |
| 12 | 5′-Methylthio- adenosine, 2 TMS derivative (Not available) | 829/ 564 | 42 (-SCH3 replaced with -OSi(CH3)3) |
The same procedure can be applied to other hits of the hybrid search result. 1-methyladenosine, tris(trimethylsilyl) ether, the second suggested NIST hit in the hybrid search show a discrepancy between the peak at m/Q 178 (C7H8N5O) of the 1-Methyladenosine, tris(trimethylsilyl) ether and the m/Q 164 found in the compound spectra.
Here, delta mass of -14 can be explained by the -N-CH3 group being reduced to a -NH on the purine structure. This NIST hit also has a very similar retention time index to that obtained for the compound of interest (Table 2, Figure 3c). Furthermore, the base structure of adenine can be corroborated by match hit 12 5′-methylthioadenosine, 2 TMS derivative. Here, the delta mass of 42 can be explained by the -SCH3 group being replaced with a -OSi(CH3)3 on the ribose part of the molecule (Table 2, Figure 3d). Using this approach, the mass differences in most initial NIST hits can be explained.
With this, a sum formula of C19H37N5O4Si3 is highly likely. The accurate mass (5.5 ppm error) and isotope ratio (1.7 % error) of the [M+H]+ ion further supports this (Figure 4). Combining all the information from above, a tentative identification for this peak is adenosine 3TMS. Only the 4TMS derivative of adenosine can be found within the NIST library. Indeed, adenosine 4TMS was also found in within the urine sample data at 13.6 minutes (accurate mass error -6.2 ppm, isotope ratio error 4.3 %), suggesting that the derivatization process for adenosine was incomplete (Figure 5).
Conclusion
These examples demonstrate identification with the ecTOF of additional compounds of interest within a patient’s urine sample. The ecTOF increases identification confidence in potential NIST hits and enables tentative identification of compounds not within the library. This can be done in parallel to routine analysis, or this additional information can later be investigated. With this comprehensive data, biomedical research laboratories can obtain a complete overview of the compounds present in a sample, which in turn enables the search for novel disease biomarkers.
References
[1] O. Fiehn, “Metabolomics by Gas Chromatography-Mass Spectrometry: Combined Targeted and Untargeted Profiling,” Curr Protoc Mol Biol, vol. 114, Apr. 2016, doi:10.1002/0471142727.MB3004S114.
[2] M.M. Koek et al., “Semi-automated non-target processing in GC × GC–MS metabolomics analysis: applicability for biomedical studies” Metabolomics, vol. 7, pp. 1-14, Mar. 2011,
doi: 10.1007/s11306-010-0219-6
[3] K. K. Pasikanti et al., “Noninvasive Urinary Metabonomic Diagnosis of Human Bladder Cancer,” Journal of Proteome Research, vol. 9, no. 6, pp. 2988–2995, Jun. 2010,
doi: 10.1021/pr901173v.
[4] S. S. Dinges et al., “Cancer metabolomic markers in urine: evidence, techniques and recommendations,” Nature Reviews Urology 2019 16:6, vol. 16, no. 6, pp. 339–362, May 2019,
doi: 10.1038/s41585-019-0185-3.
[5] A. Zhang et al., “Urine metabolomics” Clinica Chimica Acta, vol. 414, pp. 65-69, Dec. 2012, doi: 10.1016/j.cca.2012.08.016.
[6] D. J. Beale et al., “Review of recent developments in GC-MS approaches to metabolomics based research,” Metabolomics, vol. 14, no. 11, Nov. 2018, doi: 10.1007/S11306-018-1449-2.
[7] S. Bräkling et al., “Parallel Operation of Electron Ionization and Chemical Ionization for GC-MS Using a Single TOF Mass Analyzer,” Anal Chem., vol. 94, no. 15, pp. 6057-6064, April 2022,
doi: 10.1021/acs.analchem.2c00933
[8] A. Bileck et al., “A comprehensive urinary steroid analysis strategy using two-dimensional gas chromatography-time of flight mass spectrometry,” Analyst, vol. 143, no. 18, pp. 4484–4494, 2018, doi: 10.1039/c7an01990d.
[9] S. Bräkling et al., “Hydrogen Plasma-Based Medium Pressure Chemical Ionization Source for GC-TOFMS,” J Am Soc Mass Spectrom, vol. 33, no. 3, pp. 499–509, Mar. 2022,
doi: 10.1021/JASMS.1C00329.
[10] J. M. Halket and V. G. Zaikin, “Derivatization in Mass Spectrometry—1. Silylation,” European Journal of Mass Spectrometry, vol. 9, no. 1, pp. 1–21, Feb. 2003,
doi: 10.1255/ejms.527.
[11] A. G. Smith, S. J. Gaskell, and C. J. W. Brooks, “Trimethylsilyl group migration during electron impact and chemical ionization mass spectrometry of the trimethylsilyl ethers of 20- hydroxy-5α-pregnan-3-ones and 20-hydroxy-4-pregnen-3-ones,” Biological Mass Spectrometry, vol. 3, no. 4, pp. 161–165, Aug. 1976, doi: 10.1002/bms.1200030404.
[12] T. Kind and O. Fiehn, “Seven Golden Rules for heuristic filtering of molecular formulas obtained by accurate mass spectrometry,” BMC Bioinformatics, vol. 8, no. 1, pp. 1–20, Mar. 2007, doi: 10.1186/1471-2105-8 105/TABLES/6.
