Rapid Profiling of Soybean Aromatic Compounds Using Electronic Nose

Soybean (Glycine max (L.)) is the world’s most important seed legume, which contributes to 25% of global edible oil, and about two-thirds of the world’s protein concentrate for livestock feeding. One of the factors that limit soybean’s utilization as a major source of protein for humans is its characteristic soy flavor. This off-flavor can be attributed to the presence of various chemicals such as phenols, aldehydes, ketones, furans, alcohols, and amines. In addition, these flavor compounds interact with protein and cause the formation of new off-flavors. Hence, studying the chemical profile of soybean seeds is an important step in understanding how different chemical classes interact and contribute to the overall flavor profile of the crop. In our study, we utilized the HERCALES Fast Gas Chromatography (GC) electronic nose for identification and characterization of different volatile compounds in five high-yielding soybean varieties, and studied their association with off-flavors. With aroma profiling and chemical characterization, we aim to determine the quantity and quality of volatile compounds in these soybean varieties and understand their effect on the flavor profiles. The study could help to understand soybean flavor characteristics, which in turn could increase soybean use and enhance profitability.


Introduction
Soybean (Glycine max (L.)) seed protein content is 35-50% of its total dry weight and is a major source of protein in the human diet and for animal nutrition. Soybean protein also has a well-balanced amino acid profile and is rich in many essential amino acids. Soybean meal has been used extensively to make popular food products such as tofu, soy milk, soybean paste (miso), green soybeans (edamame), boiled beans (nimame), fermented soybeans (natto), soy sauce (shoyu), soybean sprouts (moyashi), and roasted soybean flour (kinako). Soybean consumption has been limited in the western world due to the beany flavor present in soy meal products. Enzymatic oxidation of linoleic acid and linolenic acid by lipoxygenase genes (Lox) is reported as a major cause of the beany flavor [1,2], and in soybeans there are three separate genes, Lox1, Lox2 and Lox3 controlling this trait [2]. Hexanal is commonly associated with the grassy flavor; hexanol; 1-octen-3-ol; 1-octen-3-one; trans,trans-2,4-decadienal; and trans,trans-2,4-nonadienal are other aromatic compounds linked with the beany taste in soy meal products [3]. Odor compounds of soybean products depend on the soybean cultivars and can change in each variety depending on growing season, storage conditions and processing technologies. Boiling the seeds at 100 • C deactivates the lipoxygenase enzymes and is the common method used for reducing the beany flavor. Breeding soybean lines with reduced beany flavor is another approach that can be used for minimizing the off-flavors in soybeans. In order to establish such breeding programs, establishing a reliable and fast screening method for testing beany flavor is necessary. Plant volatile Biosensors 2019, 9,66 2 of 13 compounds such as beany flavor are overlooked in plant phenotyping. Novel molecular techniques and marker assisted breeding can be used to map the QTLs controlling these traits and find the loci/genetic mechanisms regulating these compounds which can then be exploited for developing soybean lines with reduced off-flavor traits.
Various volatile compounds may serve as indicators of developmental maturity and as biochemical markers to evaluate seed quality. Several compound classes identified were alcohols, aldehydes, esters and lactones, ketones, and terpenoids. Many reports are available on the key volatiles of soybean [4][5][6][7][8]. The development of objectionable off-odor detection and classification methodology for use in grain grading has stimulated research on volatile components of soybeans and grains [9].
Sample preparation and scalability of GC-MS and similar instruments resulted in the development of cheaper, faster, and more user-friendly measurement instruments for routine use in analytical applications. Electronic nose (e-nose) devices are developed as versatile and low-cost alternatives to GC-MS instruments that minimize sample preparation and extraction, and offer many potential uses in biomedical and agriculture applications [12]. Volatile compounds can be measured from the sample headspace with minimal sample preparation time. The objective of this study was to use an e-nose instrument in measuring the volatile compounds among five different soybean cultivars and evaluate its potential as an alternative to GC-MS approach, and as a rapid screening tool for aromatic variations in soybean seeds.

Plant Materials
Five recent soybean releases were selected for these experiments including, UA5014C, UA5414RR, JTN-5503, JTN-5110, and JTN-5203. These lines were reported to have a higher yielding potential in a statewide comparison and resistance to common diseases in southern states of US ( Table 1). The UA5014C and UA5414RR lines were developed by the Arkansas Agricultural Experiment Station, while JTN-5503, JTN-5110, and JTN-5203 were developed at USDA-ARS Jackson Research Station. Parental information for these lines is provided in Table 1. These lines were grown at the Tennessee State University research farm in 2017. The experimental unit consisted of three replicates with two rows (20 feet deep) and a planting density of 5 seeds/ft.

Electronic Nose
The HERCALES GC Flash electronic nose (AlphaMos, Toulouse, France) was used to discriminate the odor patterns of different aroma models. For each variety, 20 gm of seed was weighed and grinded in a grinder (Waring WSG60 Grinder) at high speed for 2 min. The resulting soy flour was weighed (6 gm) and placed in a 20 mL glass vial. Following this, 7 mL of sterile distilled water was added to each tube. The sample was prepared in a septa-sealed screw cap vial and equilibrated for 200 s at 50 • C, separately. Subsequently, the aroma headspace above the sample was introduced into the electronic nose at the speed of 270 µL/s using automatic headspace sampler (PerkinElmer, MA, USA). The column temperature program used for the experiment was 40 • C (1 min)-2 • C/min-200 • C (3 min), and the injection temperature of the injector and detector were set at 180 • C and 220 • C, respectively. In addition, at the end of each column a FID detector was placed and the acquired signal was digitalized every 0.01 s. The Heracles electronic nose is equipped with two columns working in parallel mode. A non-polar column (MXT5: 5% diphenyl, 95% methylpolysiloxane, 10 m length, and 180 µm diameter), and a slightly polar column (MXT1701: 14% cyanopropylphenyl, 86% methylpolysiloxane, 10 m length, and 180 µm diameter). A single comprehensive chromatogram was generated by joining the chromatograms obtained with the two columns. This approach helps reduce incorrect identifications due to overlapping of chromatograms obtained with two different columns, and represents a useful tool for improved identification. For calibration of the instrument, an alkane solution (from n-hexane to n-hexadecane) was used to convert retention time in Kovats indices and to identify the volatile compounds using specific software (AromaChemBase). Each analysis was repeated a total of three times, and all of the response data was analyzed using Alpha Soft software (Version 3.0.0, Toulouse, France).

Results and Discussion
The volatile profiles were generated using the e-nose and were subjected to PCA analysis. The PCA plot ( Figure 1) shows the distinct clusters formed for different soy varieties indicating that the volatile profiles of soy varieties are distinctly different from each other. It also demonstrates the potential use of this system in rapid profiling of volatile compounds in different soybean cultivars. UA5414RR and UA5014C were comparable in their volatile profiles while other samples namely JTN5203, JTN5503, and JTN5110 were distantly diverse different from one another. The different clusters formed for different samples are due to their differential volatile compounds and their composition.
More than 90% of the volatile compounds were identified with Kovats index and Arochembase software in UA5014C ( Figure 2). The total volatile composition is distributed between acids, aldehydes, alcohols, esters, pyrazines (Table 2 and Figure 3). However, the major volatile composition was contributed by Ethyl-2-Methyl Butyrate (22.72%), 2-Methyl Propanal (18.21%) and 2-Propanol (16.45%). These three volatile components nearly contribute 50% of the total volatile composition in this cultivar. In UA5414RR (Table 3, Figures 2 and 3), the contribution of Ethyl-2-Methyl Butyrate (24.07%) and 2-Methyl Propanal (19.42%) is still high but instead of 2-Propanol, contribution of Ethyl 2-Methylbutanoate (16.01%) was higher in the total volatile composition. From Figure 3, it is clear that esters were the major contributor of the volatiles followed by aldehydes and alcohols in both UA5414RR and UA5014C. Acids and monoterpenes were not detected in UA5414RR. Alcohol was significantly higher in UA5014C compared to UA5414RR. Biosensors 2019, 9, x FOR PEER REVIEW 4 of 14 More than 90% of the volatile compounds were identified with Kovats index and Arochembase software in UA5014C ( Figure 2). The total volatile composition is distributed between acids, aldehydes, alcohols, esters, pyrazines (Table 2 and Figure 3). However, the major volatile composition was contributed by Ethyl-2-Methyl Butyrate (22.72%), 2-Methyl Propanal (18.21%) and 2-Propanol (16.45%). These three volatile components nearly contribute 50% of the total volatile composition in this cultivar. In UA5414RR (Table 3, Figures 2 and 3), the contribution of Ethyl-2-Methyl Butyrate (24.07%) and 2-Methyl Propanal (19.42%) is still high but instead of 2-Propanol, contribution of Ethyl 2-Methylbutanoate (16.01%) was higher in the total volatile composition. From Figure 3, it is clear that esters were the major contributor of the volatiles followed by aldehydes and alcohols in both UA5414RR and UA5014C. Acids and monoterpenes were not detected in UA5414RR. Alcohol was significantly higher in UA5014C compared to UA5414RR.  More than 90% of the volatile compounds were identified with Kovats index and Arochembase software in UA5014C (Figure 2). The total volatile composition is distributed between acids, aldehydes, alcohols, esters, pyrazines (Table 2 and Figure 3). However, the major volatile composition was contributed by Ethyl-2-Methyl Butyrate (22.72%), 2-Methyl Propanal (18.21%) and 2-Propanol (16.45%). These three volatile components nearly contribute 50% of the total volatile composition in this cultivar. In UA5414RR (Table 3, Figures 2 and 3), the contribution of Ethyl-2-Methyl Butyrate (24.07%) and 2-Methyl Propanal (19.42%) is still high but instead of 2-Propanol, contribution of Ethyl 2-Methylbutanoate (16.01%) was higher in the total volatile composition. From Figure 3, it is clear that esters were the major contributor of the volatiles followed by aldehydes and alcohols in both UA5414RR and UA5014C. Acids and monoterpenes were not detected in UA5414RR. Alcohol was significantly higher in UA5014C compared to UA5414RR.      In JTN5503 (  Figure 4), Dimethyl sulphide alone contributed to over 64% of the total volatile composition. A visual comparison of the peaks in Figure 4 clearly indicates the differences between JTN cultivars in peaks 2, 5, and 17. From Figure 5, it is clear that esters were the major contributor of the volatiles followed by aldehydes and ketones in JTN5203, JTN5503 and JTN5110. The sulfur containing compounds were a major volatile contributor in JTN5110, and JTN5203 but not in JTN5503. Furans were detected only in JTN5110 and were absent in the other two varieties. In general, acids, furans and pyrazines were low in all the samples.
PCA analysis indicated that UA5414RR and UA5014C were comparable in their volatile profiles while other samples namely JTN5203, JTN5503, and JTN5110 were distantly different from each other (Figure 1). Different clusters formed in different samples according to their differential volatile compounds and their compositions (Tables 2-6). It should be noted that beany flavor is caused by a combination of different compounds and assigning specific flavor to a cultivar should be carried out using sensory analysis with a panel of trained evaluators.  Dimethyl sulphide alone contributed to over 64% of the total volatile composition. A visual comparison of the peaks in Figure 4 clearly indicates the differences between JTN cultivars in peaks 2, 5, and 17. From Figure 5, it is clear that esters were the major contributor of the volatiles followed by aldehydes and ketones in JTN5203, JTN5503 and JTN5110. The sulfur containing compounds were a major volatile contributor in JTN5110, and JTN5203 but not in JTN5503. Furans were detected only in JTN5110 and were absent in the other two varieties. In general, acids, furans and pyrazines were low in all the samples.

Conclusions
E-nose has been used in a wide range of applications including odor analysis, quality control in food products, and biomedical applications. This study illustrates the use of e-nose as a versatile analysis tool and alternative method for measuring volatile compounds in soybean seeds with minimal sample preparation time. This approach can be used in a high-throughput phenotyping system and for screening different soybean lines. This system can be used as a rapid screening tool in breeding programs, in the selection of soybean mutants/varieties with different volatile profiles, and also for mapping the QTLs and loci responsible for these traits. This platform can also be used to link the beany flavor to seed volatile compounds, ultimately developing varieties with reduced off-flavor taste and better acceptance by the consumer.