Untargeted Profiling and Differentiation of Volatiles in Varieties of Meat Using GC Orbitrap MS

Volatile compounds play vital roles in food sensory attributes and food quality. An analysis of volatile compounds could illustrate the sensory attributes at the microscale level. Here, untargeted profiling approaches for volatiles in five most-consumed meat species were established using headspace SPME-GC/high resolution Orbitrap MS. An extended high-resolution database of meat volatile compounds was established to enhance the qualification accuracy. Using sulfur-containing compounds, aldehydes, and ketones as the research model, the parameters including fiber coating types, extraction temperature, extraction time, and desorbing time were optimized. Principle component analysis, volcano analysis and partial least squares discriminant analysis were applied to run the classification and the selection of discriminant markers between meat varieties, respectively. Different varieties could be largely distinguished according to the volatiles’ profiles. 1-Octen-3-ol, 1-octen-3-one, 2-pentyl furan and some other furans degraded from n-6 fatty acids would contribute to distinguishing duck meat from other categories, while methyl esters mainly from oleic acid as well as dimethyl sulfoxide and carbon disulfide possibly produced from the sulfur-containing amino acids contributed to the discrimination of beef. Therefore, volatiles’ profiling not only could interpret the aroma style in meat but also could be another promising method for meat differentiation and authentication.


Introduction
Meat quality includes safety, nutrition, and sensory aspects, which may be largely affected by multiple factors such as heredity, breeding, and the environment [1]. With the improvement in living standards, consumers place more and more emphasis on the nutrition and flavor. Flavor consists of aroma and taste [2]. Aroma, an important part of sensory evaluation, has become one of the important determinants influencing consumers' purchasing choices.
Chemical composition is the basis of food aroma. With the maturation of analytical techniques and rapid development of instrumentation, more and more aroma compounds and their contributions have been discovered. Constituting the unique style of aroma, aroma compositions are complex, mainly including alcohols, esters, organic acids, ketones, aldehydes, furans, pyrazines, hydrocarbons, sulfur-containing compounds [3,4], and have very wide ranges of polarity, solubility, volatility, and thermal stability. This would bring a big challenge to the untargeted profiling of the aroma compounds. Basically, the analysis of aroma compounds includes three steps: extraction, separation, and identification.

Sample Handling and SPME Procedure
Pork, beef, mutton, chicken, and duck breast were purchased from the local market. Those meats were minced, packaged in the plastic bag, and cooked in a water bath at 80 • C for 30 min. In addition, the cooked meat was cooled and ground in liquid nitrogen for the following SPME procedure. SPME was directly performed in the TriPlus RSH autosampler (Thermo Fisher Scientific (Bremen, Germany)). The procedure was as follows: A 3 g minced sample was introduced in a 20 mL glass vial. The vials were immediately closed with a magnetic cap fitted with a polytetrafluoroethylene-silicone septum. The sample vial was incubated at 55 • C for 20 min and extracted at 55 • C for 40 min using a 50/30 µm Divinylbenzene/Carboxen/Polydimethylsiloxane (DVB/CAR/PDMS) fiber (Supelco, Inc., Bellefonte, PA, USA)). In order to ensure faster extraction, the vial was maintained in agitation during the extraction period. Once the extraction was finished, the fiber was automatically inserted into the injector and desorbed at 250 • C for 3 min. Between the consecutive analysis, the fiber was conditioned in the other injector port at 270 • C for 10 min.

Analysis of Volatile Compounds by GC-HRMS
All analyses were conducted on a Q-Exactive Orbitrap mass analyzer equipped with a TriPlus RSH autosampler and Trace 1310 GC (Thermo Fisher Scientific, Bremen, Germany). A VF-WAX ms column (60 m × 0.25 mm i.d. × 0.25 µm film thickness, Agilent, Santa Clara, CA, USA) was used. Helium (99.9999%) with a constant flow rate of 1 mL/min was used as the carrier gas. The column oven was temperature-programmed starting at 40 • C for 2 min, then increased to 230 • C at a rate of 4 • C/min and then maintained at 230 • C for 5 min. Both of the transfer line 1 and transfer line 2 were set at 250 • C. MS was performed using electron impact ionization (EI) at 70 eV, operating in full scan mode at a resolving power of 60,000 full width at half maximum (FWHM). The scan range was from 30 to 400 m/z with an automatic-gain-control target value of 1E6. Ion source and transfer line temperatures for MS were set at 280 • C and 250 • C, respectively.
GC-MS data were acquired and processed using the Xcalibur 4.1 and TraceFinder 4.0 softwares (Thermo Scientific), respectively. Volatile compounds were identified in accordance with mass spectra and linear retention indices (LRIs) from NIST17 (v2.3) and the domestic library. The domestic library, namely the homeflavor library, was established using authentic reference standards, integrated with high resolution mass spectra and linear retention indices. Moreover, the high-resolution filtering (HRF) tool from the Tracefinder software was utilized to annotate every measured m/z peak and evaluate the mass accuracy of those ions when the NIST library was used. A series of standard alkanes (C7-C40; Sigma-Aldrich, St. Louis, MO, USA) were run under the same chromatographic conditions to calculate LRIs.

Method Validation
Method validation was executed using 2-methyl-3-heptanone as the model compound. Linearity, sensitivity, accuracy, and precision were investigated. The calibration curve was plotted through the responses versus the concentrations. Sensitivity of the method was evaluated using the limit of detection (LOD, S/N = 3) and the limit of quantification (LOQ, S/N = 10) of 2-methyl-3-heptanone, calculated in light of the mutton with the lowest spiking level.

Statistical Analysis
Alignment of mass signals (signal/noise≥ 3) was performed using Tracefinder software with the deconvolution plugin. Mass signals present in ≤4 replicates were discarded.
For statistical analysis, the main tendencies of the generated data were compared and visualized using principal components analysis (PCA), volcano analysis, and partial least squares discriminant analysis (PLS-DA) after a log 10 transformation and Auto scaling of the samples using Metaboanalyst 5.0. Graphs were also produced using Microsoft Excel 16.30.

Optimization of Separation
Hydrocarbons, alcohols, aldehydes, ketones, furans, esters, sulfur-containing compounds, and other heterolytic compounds such as pyridines and pyrazines compose the main volatile compounds in different foods [23]. Complex compositions of volatiles would bring big challenges in their separation. Here, GC columns with stationary phases of nonpolar (5%-phenyl)-methylpolysiloxane and polar polyethylene glycol were compared. Apart from the boiling temperature, the polarity was another influencing factor in the separation mechanism for polar GC columns. This will lead to the decreased coelution and better separation of volatile compounds when the polar stationary phase of polyethylene glycol was used. In actuality, using mutton as the research model, the number of identified volatile compounds was increased by more than 50% when polar stationary phase polyethylene glycol was used, indicating that the polar stationary phase would be more suitable for volatiles' profiling.

Optimization of Extraction
SPME is one of the most widely used extraction methods for volatile compound analysis. Among the influencing parameters, the fiber type could be the most crucial one, which could significantly affect the profiling pattern [24]. The extracted compound types and their signal intensity would vary a lot according to the different fiber types. Totals of 85 µm Polyacrylate (PA), 50 µm/30 µm DVB/CAR/PDMS, 95 µm Carbon WR, and 30 µm PDMS were selected as the representative fibers for evaluating the extraction efficiency. Figure 1 shows the tendency of volatile compounds' intensities with four kinds of fibers. Obviously, DVB/CAR/PDMS and Carbon WR showed better performance with higher signal intensities for most of the compounds. However, Carbon WR showed poor performance for sulfur-containing compounds. Considering that sulfur-containing compounds play pivotal roles in the meat flavor [25], DVB/CAR/PDMS would be selected for the further analysis. The equilibriums of volatile compounds among the fibers, the sample matrices, and the gas phase would determine the actions of volatile compounds in the extraction [13]. As for the extraction efficiency, the extraction temperature and the extraction time contributed a lot with a specified fiber [26]. Those two parameters were optimized comprehensively in this research. Figure 2 shows the influences of the extraction The equilibriums of volatile compounds among the fibers, the sample matrices, and the gas phase would determine the actions of volatile compounds in the extraction [13]. As for the extraction efficiency, the extraction temperature and the extraction time contributed a lot with a specified fiber [26]. Those two parameters were optimized comprehensively in this research. Figure 2 shows the influences of the extraction temperature and the extraction time on the intensities of the main volatile compounds in meat, including hydrocarbons, aldehydes, ketones, esters, furans, sulfur-containing compounds, and other heterolytic compounds. Basically, with the extraction time prolonged, the intensities of the typical volatile compounds were increased, especially for the aldehydes, furans, and sulfur-containing compounds. However, different situations would happen to some other compounds, such as ketones, alcohols, and hydrocarbons. Extraction for 60 min would give rise to the enhanced desorbing of those compounds. Taken together, 50 min would be chosen as the optimum extraction time in the profiling of volatile compounds. As for the extraction temperature, the intensities of the typical volatile compounds would reach the highest at 55 • C when the extraction lasted for 50 min, possibly because the desorbing would be the dominant action with the temperature continually increased. The duration of desorbing in the injector would be also highly associated with the sensitivity and the reproducibility. Here, the desorbing time ranging from 1 min to 7 min was investigated. As shown in Figure 3, with the increase in the desorbing time, the signal intensities for most of the compounds were steadily increased at the beginning. However, the intensities for some of the compounds were decreased when the desorbing was further proceeding. Therefore, the SPME conditions of incubation at 55 °C for 10 min, extraction at 55 °C for 50 min, desorbing at 270 °C for 3 min would be considered as the optimum. As shown in Figure 3, with the increase in the desorbing time, the signal intensities for most of the compounds were steadily increased at the beginning. However, the intensities for some of the compounds were decreased when the desorbing was further proceeding. Therefore, the SPME conditions of incubation at 55 • C for 10 min, extraction at 55 • C for 50 min, desorbing at 270 • C for 3 min would be considered as the optimum.

Method Validation
Method validation was processed in the optimal conditions in order to investigat the applicability of the developed method. Linearity with the matrix-matched calibratio curve for 2-methyl-3-heptanone was investigated and proved very well in the range o 0.33 to 33.33 μg/g with the linear regression correlation coefficient higher than 0.999 which is good for the quantitative analysis. LOD and LOQ were 0.01 and 0.03 μg/g respectively. The precision was evaluated through spiking 2-methyl-3-heptanone (5 μL 5 μg/mL) in mutton. The RSD of six replications was less than 15%, indicating that th developed method was capable to be applied in the routine analysis.

Identification of Volatile Compounds in Different Meats
The identification was processed through the MS pattern in NIST and homeflavo libraries, the retention index, and the HRF. When using the traditional NIST spectra matching at unit resolution, the HRF assigned tentative identification was based on th use of high-resolution mass spectra. The HRF scores obtained for the candidates shoul be higher than 95. The match factor based on the MS pattern should be higher than 750 Additionally, the difference in the retention index should be less than 20 for th homeflavor library while within 50 for the NIST library. It should be emphasized that th home-built library is necessary since the MS pattern could be changed according to th different MS analyzer. The fragments in the low-mass range would be lost for orbitra MS. This would lead to the penalized score for the identification if the Nist library wa used. With the optimum analysis conditions, five different kinds of boiled meats wer analyzed, and 148 volatile compounds in total were identified as shown in Table 1 base on the above identification rules. Based on this simple cooking method at a low temperature for reflecting the original flavor of meats, the number of the detected volatile was certainly large. Additionally, the final list of volatile compounds was used to run th statistical analysis for the assessment of the meat grouping and to determine differentia volatile compounds.

Method Validation
Method validation was processed in the optimal conditions in order to investigate the applicability of the developed method. Linearity with the matrix-matched calibration curve for 2-methyl-3-heptanone was investigated and proved very well in the range of 0.33 to 33.33 µg/g with the linear regression correlation coefficient higher than 0.999, which is good for the quantitative analysis. LOD and LOQ were 0.01 and 0.03 µg/g, respectively. The precision was evaluated through spiking 2-methyl-3-heptanone (5 µL 50 µg/mL) in mutton. The RSD of six replications was less than 15%, indicating that the developed method was capable to be applied in the routine analysis.

Identification of Volatile Compounds in Different Meats
The identification was processed through the MS pattern in NIST and homeflavor libraries, the retention index, and the HRF. When using the traditional NIST spectral matching at unit resolution, the HRF assigned tentative identification was based on the use of high-resolution mass spectra. The HRF scores obtained for the candidates should be higher than 95. The match factor based on the MS pattern should be higher than 750. Additionally, the difference in the retention index should be less than 20 for the homeflavor library while within 50 for the NIST library. It should be emphasized that the home-built library is necessary since the MS pattern could be changed according to the different MS analyzer. The fragments in the low-mass range would be lost for orbitrap MS. This would lead to the penalized score for the identification if the Nist library was used. With the optimum analysis conditions, five different kinds of boiled meats were analyzed, and 148 volatile compounds in total were identified as shown in Table 1 based on the above identification rules. Based on this simple cooking method at a low temperature for reflecting the original flavor of meats, the number of the detected volatiles was certainly large. Additionally, the final list of volatile compounds was used to run the statistical analysis for the assessment of the meat grouping and to determine differential volatile compounds.
It could be found that hydrocarbons, aldehydes, ketones, esters, acids, furans, sulfurcontaining compounds, and lactones composed the main components of volatile compounds in meat from Table 1. Aromatic hydrocarbons including toluene, xylenes, and ethylbenzene were the dominant component, mainly coming from the feedstuffs [27]. The alkanals and alkenals such as hexanal, octanal, 2-decenal, and (2E,4E)-heptadienal were mainly formed through the autoxidation of the lipids [28]. The acetaldehyde and the aromatic aldehydes were produced through the Strecker degradation of amino acids and their derivatives, such as glycine, phenyl alanine, and so on [29]. Furans, alcohols, acids, and ketones were also from the autoxidation of the lipids with unsaturated fatty acids [30]. As for the sulfur-containing compounds, thiamine degradation, Maillard reaction, Strecker degradation, and its further oxidation could be their main formation routes [5,31,32]. Furaldehyde and their derivatives would be formed through the Maillard reaction [29]. Lactones are a group of cyclic esters and are formed by the intramolecular condensation of hydroxy fatty acids [33,34].

Statistical Analysis
Volatilomics, utilizing the metabolomic pipelines, aims to characterize comprehensively a wide range of volatile molecules with the masses less than 550 Da, helps to compare the global profile of volatile compounds between groups of samples accurately, as well as identifies discriminatory compounds. PCA was performed to visualize how the volatile profiling differed between the five different meats (Figure 4). Based on all identified volatile compounds, the first two PCs, explaining over 60% of the total variance between the samples, showed a clear separation of the profiles of the five different meats. The clustering of the five replicate samples indicated that the applied analytical methods of GC-Q/orbitrap high resolution MS coupled with SPME were of good reproducibility and robustness, which are very important in untargeted studies for differentiation.  Another aim of this study was to define the specific compounds which are pivotal for the discrimination between the different meat samples. With the unpaired t-test, compounds responsible for the differences between the sample groups were found, as shown in Table 2. With the comparison of every two kinds of meats, dozens of compounds were present at abundances two times higher or lower (fold change > 2 or fold change < 0.5). In the comparison of mutton and beef meats, there were only 38 differentiating compounds found. In the formation pathways of meat aroma, lipid oxidation was one of the primary pathways. Saturated fatty acids (SFA) and monosaturated fatty acids (MUSFA) were the dominant fatty acids, and their levels in mutton and beef meats were similar [35]. In addition, the levels of polyunsaturated fatty acids (PUSFA) in both mutton and beef were relatively low. Therefore, this would lead to the similar pattern of volatile compounds produced through lipid autoxidation, resulting in the fewest differentiating compounds between mutton and beef. However, when comparing pork, duck, and chicken meats, the number of the differentiating volatile compounds would be elevated. It was convinced that PUSFAs [35], including linoleic acid, linolenic acid, and arachidonic acid, were the most easily autoxidized fatty acids. These fatty acids could produce complex patterns and large volumes of volatile compounds during heating procedures.

Comparison Type
Number of Differentiating Compounds Another aim of this study was to define the specific compounds which are pivotal for the discrimination between the different meat samples. With the unpaired t-test, compounds responsible for the differences between the sample groups were found, as shown in Table 2. With the comparison of every two kinds of meats, dozens of compounds were present at abundances two times higher or lower (fold change > 2 or fold change < 0.5). In the comparison of mutton and beef meats, there were only 38 differentiating compounds found. In the formation pathways of meat aroma, lipid oxidation was one of the primary pathways. Saturated fatty acids (SFA) and monosaturated fatty acids (MUSFA) were the dominant fatty acids, and their levels in mutton and beef meats were similar [35]. In addition, the levels of polyunsaturated fatty acids (PUSFA) in both mutton and beef were relatively low. Therefore, this would lead to the similar pattern of volatile compounds produced through lipid autoxidation, resulting in the fewest differentiating compounds between mutton and beef. However, when comparing pork, duck, and chicken meats, the number of the differentiating volatile compounds would be elevated. It was convinced that PUSFAs [35], including linoleic acid, linolenic acid, and arachidonic acid, were the most easily autoxidized fatty acids. These fatty acids could produce complex patterns and large volumes of volatile compounds during heating procedures. From the results based on the PLS-DA analysis for the five meat categories, complete separation between every two meat categories was achieved. The model showed good interpretability, predictability, and non-overfitting due to the high value of R2 (cum, 98.6%) and Q2 (cum, 96.2%). Each meat category showed compact groups, indicating the authenticity of different meat species. In addition, 58 discriminant compounds were screened (VIP ≥ 1). Figure 5 shows 15 most important discriminant compounds. Interestingly, the levels of 2-pentyl furan, 2-butyl furan, 2-hexyl furan, 1-octen-3-ol, and 1-octen-3-one in duck meat, produced from the oxidation of n-6 fatty acids, were higher than those in the other four categories of meats. The highest levels of linoleic acid and arachidonic acid in duck meat [35] would contribute to the discrimination. In addition, carbon disulfide, dimethyl sulfoxide, fatty acid methyl esters including methyl butanoate, methyl octanoate, and methyl nonanoate were the most important discriminants for beef. Methyl esters would be produced through lipid degradation [36]. Carbon disulfide and dimethyl sulfoxide should come from the sulfur-containing amino acids such as methionine and cysteine. From the results based on the PLS-DA analysis for the five meat categories, complete separation between every two meat categories was achieved. The model showed good interpretability, predictability, and non-overfitting due to the high value of R2 (cum, 98.6%) and Q2 (cum, 96.2%). Each meat category showed compact groups, indicating the authenticity of different meat species. In addition, 58 discriminant compounds were screened (VIP ≥ 1). Figure 5 shows 15 most important discriminant compounds. Interestingly, the levels of 2-pentyl furan, 2-butyl furan, 2-hexyl furan, 1-octen-3-ol, and 1-octen-3-one in duck meat, produced from the oxidation of n-6 fatty acids, were higher than those in the other four categories of meats. The highest levels of linoleic acid and arachidonic acid in duck meat [35] would contribute to the discrimination. In addition, carbon disulfide, dimethyl sulfoxide, fatty acid methyl esters including methyl butanoate, methyl octanoate, and methyl nonanoate were the most important discriminants for beef. Methyl esters would be produced through lipid degradation [36]. Carbon disulfide and dimethyl sulfoxide should come from the sulfur-containing amino acids such as methionine and cysteine.
Taken together, five categories of meat could be apparently distinguished according to the volatiles' profiling. The profile pattern would vary a lot with specific volatiles generated from certain aroma-precursors such as lipids or amino acids. In addition, nowadays, volatile profiling combined with chemometric models show great potentials in protecting from food fraud, such as virgin olive oil [37] and honey [38]. Here, this study also gave a solid foundation to the meat differentiation as well as the meat authentication.  Taken together, five categories of meat could be apparently distinguished according to the volatiles' profiling. The profile pattern would vary a lot with specific volatiles generated from certain aroma-precursors such as lipids or amino acids. In addition, nowadays, volatile profiling combined with chemometric models show great potentials in protecting from food fraud, such as virgin olive oil [37] and honey [38]. Here, this study also gave a solid foundation to the meat differentiation as well as the meat authentication.

Conclusions
The present study developed an analytical strategy for the untargeted profiling of volatile compounds in boiled meat. The method based on the HS-SPME-GC-MS was investigated in detail, and the influencing parameters including extraction fibers, extraction temperatures, extraction time, and desorbing time were optimized comprehensively. High sensitivity and good reproducibility were achieved. For the confirmation of the volatiles, the combination of retention indices, the home-built high-resolution database (home flavor), NIST database matching using low-resolution mass spectrometry, and HRF scores based on the accurate masses was utilized. The home-built high-resolution database could significantly enhance the identification accuracy. The optimum untargeted approach in tandem with multivariate data analysis techniques was promising for meat categories' discrimination and meat authentication. Duck, chicken, beef, mutton, and pork meats were easily distinguished based on their volatiles' profiles. 2-Pentyl furan, 2-butyl furan, 2-hexyl furan, 1-octen-3-ol, and 1-octen-3-one were positively correlated with duck meat. Methyl esters would be the main discriminant biomarkers for beef. This phenomenon was caused by the different fatty acids' pattern in different meat species.

Data Availability Statement:
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Conflicts of Interest:
All of the authors have no conflict of interest to declare. As for Jiangtao Xing from Thermo Fisher Scientific, his work mainly focused on the instrumentation and the method optimization. His participation would not affect the authenticity and objectivity of the experimental results. In addition, Thermo Fisher Scientific did not provided funding or equipment in the experimental design.