Boosting Untargeted Benchtop Volatilomics to the Next Level: Chiral Trapped-Headspace GC-QMS-IMS

For the characterization of complex, volatile organic compound (VOC)-rich food products, the combination of gas chromatography–quadrupole mass spectrometry–drift tube ion mobility spectrometry (GC-QMS-IMS) is a powerful technique, offering simultaneous detection by QMS and IMS. This hyphenation pairs characteristic drift times and soft ionization from the IMS section with specific m/z values for database substance identification from the QMS detector. A challenge so far in this approach was the typically substantially lower sensitivity of the QMS system in comparison with the IMS system. This is a limitation, in particular with regard to low‑to‑no sample preparation, static headspace-based approaches, as the QMS section often delivered inferior signal‑to‑noise ratio mass spectra with low database hit rates. The present approach describes a new untargeted, trapped-headspace (THS)-GC-QMS-IMS setup for an exhaustive VOC analysis of complex food, beverage, and flavor products and allows the classification of mango cultivars in combination with chemometric data evaluation. This approach is fortified by the use of chiral gas chromatography, where the separation of enantiomeric compounds adds significantly more useful features for machine learning approaches.

Gas chromatography (GC) is a well-known and widely used technique for the separation of volatile organic compounds (VOC) in the domains of quality and authenticity control of flavorings, foods, or beverages. The very common hyphenation of quadrupole mass spectrometry (QMS) with gas chromatography (GC) has numerous applications in the field of food and beverage products. Within the last decade, ion mobility spectrometry (IMS) has become a popular technique for analysis of flavorings and food products (1). Its application is more practical at the point of care (POC), due to the robust and sensitive characteristics of IMS setups, as well as its operating principle at ambient pressure conditions (2).

In non-targeted workflows for food authentication, drift tube IMS (DTIMS) is a powerful tool for volatile profile fingerprinting. In contrast to electron ionization typically used in GC-QMS, DTIMS is most commonly based on a soft, ³H-based ionization, which forms proton-water clusters MH⁺[H₂O]of the analytes (3). In dependence of the gas phase concentration of analytes, protonated monomers MH⁺[H₂O]_n−x and dimers M₂H⁺[H₂O]_n−x are formed. In the drift tube, these are accelerated in an electrical field, counterflowing a buffer gas, such as nitrogen or synthetic air. Due to collisions with drift gas molecules, each analyte shows a drift time, depending on collision cross-section (CCS) (2,3).

The limited availability of libraries for substance identification remains one of the most relevant limitations in IMS, and typically requires time-intensive workflows for substance identification and the need of reference standards. Although there are existing hyphenations of GC with QMS and IMS for simultaneous analyte detection reported in the literature, which helps to overcome this issue, the substantially higher sensitivity of IMS typically outperforms the QMS in full scan mode, and as such, only adds limited mass-to-charge ratio (m/z) information to the most abundant analytes (4–6).

One reason for this limitation was the syringe-based static headspace sampling (SHS), where low-abundant analytes were not detected in the QMS detector. This is even more challenging in complex food and beverage matrices, where an increase in incubation temperature can easily lead to artifacts, as occurs in the Maillard reaction. In this context, an elegant alternative for a more exhaustive analysis of VOC profiles is trapped headspace sampling. In this approach, the sample vessel is pressurized in an oven, the gaseous VOCs are trapped using a sorbent system, and, finally, are thermally desorbed onto the column. An option here is multiple headspace extraction (MHE), where the pressurization and sorbent trapping steps are repeated multiple times. This preconcentration allows an increase in sensitivity by more than a factor of 20 in comparison to common SHS methods. Within this study, trapped HS-GC-QMS-IMS was applied to mango pulp samples, where we explicitly focused on increasing the signal in QMS to obtain maximum sensitivity in the m/z domain for the different analytes (7), but at the same time to avoid overloading of the IMS detector.

Mangos (Mangifera Indica L.) are widely cultivated fruits in tropical areas of Asia, Africa, and America. Due to their popularity, they are often referred to as the “king of fruits,” and are appreciated for their nutritional value and their fruity, sweet, and floral taste profile (8). In the field of mango fruits and products, literature on the use of GC-IMS published to date is scarce. Examples are application studies of GC-IMS to monitor the ripening process of Chinese fruit cultivars; however, substance identification was achieved using a separate GC–MS system (9,10). The aim of the present study focused on the differentiation of Mangifera indica L. cultivars, using a single-injection HS-GC-QMS-IMS non-targeted approach. For more detailed insights of the VOC profile acquired by GC-IMS, a number of relevant compounds were tentatively identified vs. the NIST database by EI-QMS simultaneously. Furthermore, this study demonstrates the potential of THS sampling in the analysis of foodstuffs and beverages, and describes the use of a chiral GC column to increase the number of usable features in chemometric data analysis.

Material and Methods

Sample Preparation

From each mango pulp or concentrate sample (30 samples), 2 g were transferred to a 20 mL headspace vial and closed with a butyl/polytetrafluoroethylene (PTFE). The samples involved 10 Indian cultivar “Totapuri,” including three concentrates, 10 samples of the Indian cultivar “Alphonso,” seven samples of the cultivar “Tommy” (from Mexico), including six concentrates, one sample of the cultivar “Kent” (from Peru), and two samples of the cultivar ‘Criollo’ (from Peru). All samples were provided kindly by the SGF International’s Voluntary Control System out of the years 2019 to 2023 (SGF International e.V.). Further, five reference standards were prepared in canola oil, including α-pinene, D-limonene, ethyl butyrate, α-terpineol (Sigma-Aldrich Chemie GmbH), and β-caryophyllene (Carl Roth GmbH + Co. KG).

Instrumental Analysis

A HS20 headspace sampler (Shimadzu Corporation) with an oven temperature of 50 °C was used, transfer and sample lines were operated at 150 °C. Trap cooling temperature was −10 °C, and desorption temperature was 250 °C. The trap material used was Tenax TA (Shimadzu Corporation). Multi-injection steps were set to 5. Chromatographic separation was performed using a chiral, β-cyclodextrin-based BGB 174 capillary column (30 m × 0.25 mm, 0.25-µm; BGB Analytik Vertrieb GmbH) and a Nexis GC-2030 (Shimadzu Corporation). Helium was the carrier gas in constant-pressure mode with 180 kPa and a splitter plate with an advanced pressure controller (APC) pressure of 38 kPa was used for feeding the QMS and IMS systems. The oven program started at 40 °C initial temperature, followed by a temperature ramp of 1 °C/min to 60 °C, 4 °C/min to 100 °C, and 6 °C/min to 160 °C, then held for 5 min, resulting in 45 min per run. The GC column gas flow was split by a SilFlow GC 4 port splitter plate (Trajan Scientific and Medical) into two retention gaps (IMS: 0.7 m length and 0.15 mm inner diameter, MS: 1.6 m length and 0.15 mm inner diameter). Transfer lines were set both to 220 °C for the QMS and the IMS (Hillesheim GmbH). Electron ionization energy was 70 eV, the emission current was 150 µA, ion source temperature of the QP2020 NX MSD (Shimadzu Corporation) was set to 200 °C, and the scan range was m/z 35 to m/z 400, with a duty cycle of 300 ms. A Focus-IMS cell with ³H ionization source (100 MBq β-emission) was used and the drift tube temperature was set to 100 °C. The drift tube had a diameter of 15.2 mm and a length of 98 mm. The IMS was operated in positive-ion mode at a constant voltage of 2.5 kV. Blocking voltage was set to 70 V, injection voltage was set to 2500 V. The drift gas was nitrogen with a purity of 99.9999%, controlled by a mass flow controller at 150 mL/min (Voegtlin Instruments AG). Each spectrum was averaged over six scans, using a repetition rate of 21 ms. The injection pulse was set to 100 µs with a sampling frequency of 228 kHz.

The package GC-IMS-tools version 0.1.7 and Python version 3.8.8 were used for preprocessing, multivariate analysis, and visualization of the IMS spectra. Each spectrum was treated by a “db3” wavelet compression, and all spectra were aligned in drift time dimensions and normalized to the reactant ion peak (RIP). The spectra were cropped to the region of 100 s to 2700 s in the retention time axis and 1.03 to 2.5 in drift time axis (ca. 6.5–16.25 ms), respectively. Further, a baseline correction by symmetric least squares was applied to the data set. To determine retention times, an automated 2D peak detection by persistent homology was used. Subsequently, Pareto scaling and mean centering was applied to the data set.

Results

More than 80 compounds were detected in IMS and the majority of these also featured sufficient S/N ratios in the QMS detection. The detected species were in line with previously published VOC studies of mango fruits that applied solid-phase microextraction (SPME)-HS-GC–MS (11,12). While the reported VOCs were similar, the use of a chiral GC column exhibited additional information and as such is a promising approach for VOC analysis in general. This approach combined an EI-QMS system with highly reproducible, fragment-rich m/z information that is best suited for database comparison, whereas the IMS system can detect intact analyte ion species at ultra-low trace levels. A comparison of the retention times for IMS and QMS for selected compounds, including esters, aldehydes, terpenes, and sesquiterpenes, are listed in Table I.

For most of the analytes, MS and IMS retention times were in good accordance with only minor deviations. The match quality of the MS spectra was equal or above 95% for each compound. As already observed in prior studies, substances with a low boiling point featured better hit rates compared to those with higher boiling points (for example, sesquiterpenes). However, within this study, we were able to reduce the retention time difference for most of the analytes due to the use of optimized retention gaps for IMS and QMS. In comparison to previous studies, the retention time difference for β-caryophyllene was reduced from 0.19 to 0.15 min and, respectively, for α-terpineol from 0.18 to 0.05 min (4,7). While IMS featured a higher sensitivity for esters, aldehydes, and polar compounds in general, the QMS is beneficial for detection of high-boiling VOCs, such as sesquiterpenes. IMS cells are commonly used at temperatures of approximately 100 °C, which leads to tailing effects on high boiling VOCs and poor peak shapes for the detection of sesquiterpenes.

In Figure 1, a PCA scores plot of PC2 and PC3 is visualized, covering 19.0% and 14.1%, respectively. The arrow in Figure 1 indicates the direction of the PC2 loadings influence.

The scores plot of PC2 allows the differentiation of the cultivar “Alphonso,” while PC3 shows a separation of the other cultivars. The backwards-projected loadings (see Figure 2) enable the connection to the original data, where negative values of the loadings from PC2 are colored in blue and high positive loadings are colored in red. In combination with the simultaneous MS detection, 17 compounds were identified, which had a relevant impact on PC2 and, consequently, on the separation of the popular and higher priced cultivar “Alphonso.” While higher amounts of sesquiterpenes were detected in the cultivars “Criollo” and “Totapuri,” the compound 2,5-dimethyl-4-methoxy-3(2H)-furanone showed a highly negative value on the PC2 loading and is already reported to have a significant impact on the volatile profile of the cultivar “Alphonso” (13). In addition, the cultivar “Alphonso” featured γ-butyrolactone and γ-hexalactone, which were not detected for the cultivars “Totapuri” and “Tommy.” In Figure 3, a 3D-PCA scores plot of PC1, PC2, and PC3 is displayed. While PC2 and PC3 indicate a separation of the different mango cultivars, PC1 additionally depicts a separate cluster with positive scores values. These samples were concentrates, while the other samples were only minimally processed. With the help of the backwards-projected loadings plot, it was possible to identify several compounds using the simultaneously generated QMS spectra. 3-Carene, ethanol, and acetic acid featured positive loadings on PC1 and were more abundant in the concentrates, while ethyl acetate, 2,5-dimethyl-4-methoxy-3(2H)-furanone, and trans-β-ocimene indicated a negative correlation on PC1. Further, the analyzed mango concentrates of the cultivar “Totapuri” featured a lower abundance of a number of alcohols and esters in comparison to the minimal processed samples. These compounds were specifically isobutanol, isopentyl alcohol, ethyl propanoate, propyl acetate, and methyl butyrate.

For additional observation of the separation of minimal processed and concentrated samples, a hierarchical cluster analysis (HCA) was performed on the data set. The dendrogram (see Figure 4) indicates two clusters, one with minimal processed methods and one with the concentrated samples. In addition, the cultivar “Alphonso” also featured a separate cluster within the minimal processed samples. This is caused by the already mentioned composition of VOCs with a lower abundance of sesquiterpenes and a unique profile of several lactones (14). This underlines the benefits of the use of an enantiomeric GC-IMS fingerprint for non-targeted data analysis and enables a fast search for similarity between the processing methods and flavor profiles of different mango cultivars.

Conclusion

This study used trapped HS-GC-IMS-MS to detect over 80 compounds in a simultaneous data acquisition with highly comparable retention times. The complementary data from one single injection renders obsolete the need for two separate GC–MS and GC-IMS systems, which typically requires a potentially complex retention index based correlation. In analytics of foodstuffs and beverages, this approach allows for an exhaustive VOC analysis at comparatively low incubation temperatures, which lowers the possibility of artifact formation. The acquired data allowed for the differentiation of cultivars and processing methods, using HCA, PCA scores, and backwards-projected loadings plots. The use of a column with chiral selectivity substantially increased the number of usable features for the subsequent data analysis. The presented approach may be beneficial for authenticity and quality control of beverages, foods, and flavorings. With the simultaneous IMS and MS detection, the establishment of databases for GC-IMS can be simplified and verified, which is to date one of the most challenging points.

Acknowledgments

This work was supported by the Project FH Kooperativ “Deep Authent” (grant number 13FH138KX0) of the Federal Ministry of Education and Research (BMBF), Berlin, Germany.

References

(1) Wang, S.; Chen, H.; Sun, B. Recent Progress in Food Flavor Analysis Using Gas Chromatography-Ion Mobility Spectrometry (GC-IMS). Food Chem. 2020, 315, 126158. DOI: 10.1016/j.foodchem.2019.126158

(2) Capitain, C.; Weller, P. Non-Targeted Screening Approaches for Profiling of Volatile Organic Compounds Based on Gas Chromatography-Ion Mobility Spectroscopy (GC-IMS) and Machine Learning. Molecules 2021, 26 (18), 5457. DOI: 10.3390/molecules26185457

(3) Borsdorf, H.; Eiceman, G. A. Ion Mobility Spectrometry: Principles and Applications. App. Spectrosc. Rev. 2006, 41 (4), 323–375. DOI: 10.1080/05704920600663469

(4) Brendel, R.; Schwolow, S.; Rohn, S.; Weller, P. Volatilomic Profiling of Citrus Juices by Dual-Detection HS-GC-MS-IMS and Machine Learning-An Alternative Authentication Approach. J. Agric. Food Chem. 2021, 69 (5), 1727–1738. DOI: 10.1021/acs.jafc.0c07447

(5) Brendel, R.; Schwolow, S.; Rohn, S.; Weller, P. Gas-Phase Volatilomic Approaches for Quality Control of Brewing Hops Based on Simultaneous GC-MS-IMS and Machine Learning. Anal. Bioanal. Chem. 2020, 412 (26), 7085–7097. DOI: 10.1007/s00216-020-02842-y

(6) Schanzmann, H.; Ruzsanyi, V.; Ahmad-Nejad, P.; Telgheder, U.; Sielemann, S. A Novel Coupling Technique Based on Thermal Desorption Gas Chromatography with Mass Spectrometry and Ion Mobility Spectrometry for Breath Analysis. J. Breath Res. 2023, 18 (1), 016009. DOI: 10.1088/1752-7163/ad1615

(7) Bodenbender, L.; Rohn, S.; Sauer, S.; Jungen, M.; Weller, P. Chiral Trapped-Headspace GC-QMS-IMS: Boosting Untargeted Benchtop Volatilomics to the Next Level. Chemosensors 2024, 12 (8), 165. DOI: 10.3390/chemosensors12080165

(8) Lauricella, M.; Emanuele, S.; Calvaruso, G.; Giuliano, M.; D’Anneo, A. Multifaceted Health Benefits of Mangifera indica L. (Mango): The Inestimable Value of Orchards Recently Planted in Sicilian Rural Areas. Nutrients 2017, 9 (5), 525. DOI: 10.3390/nu9050525

(9) Xie, H.; Meng, L.; Guo, Y.;et al. Effects of Volatile Flavour Compound Variations on the Varying Aroma of Mangoes ‘Tainong’ and ‘Hongyu’ During Storage. Molecules 2023, 28 (9), 3693. DOI: 10.3390/molecules28093693

(10) Li, L.; Yi, P.; Sun, J.; et al. Genome-Wide Transcriptome Analysis Uncovers Gene Networks Regulating Fruit Quality and Volatile Compounds in Mango Cultivar ‘Tainong’ During Postharvest. Food Res. Intl. 2023, 165, 112531. DOI: 10.1016/j.foodres.2023.112531

(11) Farag, M. A.; Dokalahy, E. U.; Eissa, T. F.; Kamal, I. M.; Zayed, A. Chemometrics-Based Aroma Discrimination of 14 Egyptian Mango Fruits of Different Cultivars and Origins, and Their Response to Probiotics Analyzed via SPME Coupled to GC-MS. ACS Omega 2022, 7 (2), 2377–2390. DOI: 10.1021/acsomega.1c06341

(12) Shimizu, K.; Matsukawa, T.; Kanematsu, R.; et al. Volatile Profiling of Fruits of 17 Mango Cultivars by HS-SPME-GC/MS Combined with Principal Component Analysis. Biosci., Biotechnol., Biochem. 2021, 85 (8), 1789–1797. DOI: 10.1093/bbb/zbab097

(13) Kallio, H. P. Historical Review on the Identification of Mesifurane, 2,5-Dimethyl-4-methoxy-3(2 H)-furanone, and Its Occurrence in Berries and Fruits. J. Agric. Food Chem. 2018, 66 (11), 2553–2560. DOI: 10.1021/acs.jafc.8b00519

(14) Deshpande, A. B.; Chidley, H. G.; Oak, P. S.; et al. Data on Changes in the Fatty Acid Composition During Fruit Development and Ripening of Three Mango Cultivars (Alphonso, Pairi and Kent) Varying in Lactone Content. Data in brief [Online] 2016, 9, 480–491. DOI: 10.1016/j.dib.2016.09.018

Lukas Bodenbender is currently a PhD student in the research group of Philipp Weller at the Institute for Instrumental Analytics and Bioanalytics, Mannheim University of Applied Sciences. He received his BS and MS in biotechnology, and his current research is focused on volatile profiling of complex food matrices, such as citrus fruit juices and oils, using GC-IMS and GC–MS in combination with machine learning.

Sascha Rohn is a full professor for food chemistry at the Technische Universität Berlin, Germany, since October 2020. He graduated from the University of Frankfurt/Main, Germany, with the first and second state examination in food chemistry. In 2002, he obtained his Ph.D. from the Institute of Nutritional Science, University of Potsdam, Germany. His group is focused on the analysis of bioactive food compounds and their reactivity and stability. The aim is to identify degradation products that serve as quality parameters, as process markers during food/feed processing, or as biomarkers in nutritional physiology.

Simeon Sauer is a professor for maths, physics, and statistics at the Faculty of Biotechnology at Mannheim University of Applied Sciences. He had previously worked at BASF SE as a research scientist in the Department for Digitalization of R&D. There, he dealt with data analysis and modeling of biotechnological procedures and processes, in which he supplemented classical kinetic modeling approaches with data-driven methods from machine learning (hybrid models/greybox models). He also worked in the field of predicting degradation phenomena in large-scale chemical plants using time series analysis and machine learning.

Markus Jungen received an engineering degree (Dipl.-Ing.) after studies in food technology and biotechnology at the University of Bonn with a focus on flavor chemistry and statistical design of experiments. Since 2009 he has been technical manager at SGF International e.V., Saulheim, Germany. He is expert in questions of fruit juice authentication, performing analytical evaluation of fruit juices authenticity throughout the supply chain, and is involved in the organization of plant inspections and audits worldwide in the framework of SGF’s voluntary control systems.

Philipp Weller is a full professor at the Institute of Instrumental Analytics and Bioanalytics at the Mannheim University of Applied Sciences. Furthermore, he also heads the competency center for chemometrics and material analysis CHARISMA (www.charisma.hs-mannheim.de). He holds a PhD in food chemistry (2004) from the University of Hohenheim. His research is mainly focused on data-driven omics approaches based on the multimodal analysis of complex food matrices, fermentation processes, and cell cultures. His research group develops specialized machine learning/deep learning toolboxes for high-dimensional data and data fusion.