Sensomics-Assisted Aroma Decoding of Pea Protein Isolates (Pisum sativum L.)

The aroma of pea protein (Pisum sativum L.) was decrypted for knowledge-based flavor optimization of new food products containing pea protein. Sensomics helped to determine several volatiles via ultra-high performance liquid chromatography tandem mass spectrometry and 3-nitrophenylhydrazine derivatization. Among the investigated volatiles, representatives of aldehydes, ketones, and acids were reported in literature as especially important in pea and pea-related matrices. After validation of the method and quantitation of the corresponding analytes, sensory reconstitution as well as omission studies of a selected pea protein were performed and revealed nine odor-active compounds as key food odorants (3-methylbutanal, hexanal, acetaldehyde, (E,E)-2,4-nonadienal, (E)-2-octenal, benzaldehyde, heptanal, 2-methylbutanal, and nonanoic acid). Interestingly, eight out of nine compounds belonged to the chemical class of aldehydes. Statistical heatmap and cluster analysis of all odor activity values of different pea proteins confirmed the obtained sensory results and generalize these nine key food odorants in other pea proteins. The knowledge of key components gained shows potential for simplifying industrial flavor optimization of pea protein-based food.


Introduction
In the last several years, an increasing market for alternative foods based on plant ingredients such as pea, soy, oat, or hemp can be observed. While in 2018, the German sales for plant-based dairy substitutes were about €316 million, that number soared by 70% to a total revenue of €536 million in 2020 [1]. Furthermore, in the first quarter of 2020, the production of vegetarian or vegan meat substitutes increased by 37% compared to the previous quarter [2]. Even though these plant-based alternatives are more expensive compared to their original analogs, this does not prevent consumers from buying. This can be explained by an increasing tendency towards sustainability and animal welfare, e.g., as highlighted by a 2020 consumer barometer, saying that 69% of consumers are more willing to pay a higher price for food with sustainable origin [3]. As this current trend apparently results in good business and plant-based proteins while simultaneously showing a better footprint [4], it is indeed linked not only to start-ups but also to well-established companies concentrating their attention and their R&D on this highly topical issue.
Nevertheless, these new foods often suffer from off-flavors introduced by the plantbased ingredients. Aroma perception especially gets distorted by grassy, green, and beanlike changes, as it could be, e.g., shown for different functional milk desserts loaded with microparticulated pea protein as a fat replacer [5]. As Europe offers a large market for pea protein (the second largest in 2016, with 33%) [6], and global rising sales to $285 million through 2026 (compound annual growth rate of 2020-2026: 12%) are predicted [7], the focus of this study is dedicated to pea protein (Pisum sativum L.) and its characteristic flavor.
Driven by the aim to lose the distracting pea off-flavor, aroma and taste compositions have been investigated. Activity-guided fractionation in combination with taste dilution experiments could already reveal and explain bitter off-taste in pea protein isolates caused by different lipids and lipid oxidation products [8]. While hexanal and 3-methylbutanoic acid have been unequivocally identified as key food odorants (KFO) in raw peas [9], other aroma-active compounds such as methional, 2-undecanone, (E)-2-octenal, (E,E)-3,5octadien-2-one, (E,E)-2,4-decadienal, or phenylacetaldehyde seem to play a crucial role in pea proteins [10,11]. Nevertheless, a confirmative molecular-sensory aroma reconstitution of pea protein has been missing so far.
Over the last years, the Sensomics concept was successfully adopted to identify many KFO in different food matrices, such as in Chinese green tea [12], in Styrian pumpkin seed oil [13], or in raw licorice [14], for instance. By applying the recently described "unified flavor quantitation", we could advantageously supplement the Sensomics approach by a fast sample preparation, easy handling, and quick UHPLC-MS/MS measurements [15]. In addition, a recently published UHPLC-MS/MS method was also added to cover an increased variety of different odor-active sensometabolites in pea proteins [16].
Commonly reported odorants, identified by GC-O analyses in differently processed pea [9,17], pea proteins [10,11,18], and lupin flours [19], chosen because of their close biological affinity, were also included for targeted quantitation via accurate stable isotope dilution analysis. Therefore, the objective of the present investigation was to decode the aroma of a widely-used pea protein (Pisum sativum L.) within Europe, namely Nutralys ® S85F, by identifying KFO and providing the received knowledge for further flavor optimization, such as downregulation steps. Odor activity calculations followed by sensory reconstitution and omission experiments were performed to achieve an authentic aroma recombinant, as well as to highlight KFO in pea protein (Pisum sativum L.) for the first time ever.

Pea Protein Isolates
The following pea proteins (Table 1) were provided by our partners from the Industrial  Collective Research (IGF) branch of the FEI project under grant number AiF 20197 N and analyzed with the developed methods. All samples were stored in the dark at 4 • C.

Protein Content (PC)
Protein contents were determined using the Dumas method with the Vario MAX cube (Elementar Analysensysteme GmbH, Langenselbold, Germany) by the Chair of Food and Bioprocess Engineering at the Technical University of Munich. As proposed for pea proteins, a factor of 5.4 was used for conversion of nitrogen content into protein content [23].
Validation experiments: As no analyte-free pea-like matrix was available for recovery experiments, calibration curves were first analyzed with (standard addition calibration) and without (matrix-free calibration) the presence of pea protein (40 mg/mL, C) in acetonitrile/water (50:50, v/v), as described above. A comparison between both curves revealed either the same slope, just shifted by a certain amount for the analytes present in pea protein, e.g., for hexanal and hexanoic acid, or congruent curves for no or low abundance, such as for (E,Z)-2,6-nonadienal and 2,3-octanedione ( Figure 2). Hexanal, hexanoic acid, (E,Z)-2,6-nonadienal, and 2,3-octanedione represented one typical compound of each of the investigated compound classes; all analytes indicated the same behavior.
These experiments concluded that matrix effects during ionization were fully compensated by the selected internal standards and, therefore, recovery experiments were recorded to check the extraction and derivatization process in triplicate by spiking a constant volume (20 µL) of the diluted stock solution (1:120) with acetonitrile/water (50:50, v/v) and equilibrating with the IS solution (20 µL) for at least 20 h. The sample preparation was further conducted as detailed above. For the determination of the limit of detection (LOD) and the limit of quantitation (LOQ), the lowest calibration solution was further diluted, and the signal-to-noise ratio was measured using the MultiQuant software (AB Sciex, Darmstadt, Germany). The LOD was set to a signal-to-noise ratio of 3, and the LOQ was set to a signal-to-noise ratio of 10.

Sensory Analysis
General conditions and panel training: Orthonasal aroma experiments were performed by sixteen aroma panelists (nine women and seven men, age 22-58 years) from the Chair of Food Chemistry and Molecular Sensory Science and the Leibniz Institute for Food Systems Biology at the Technical University of Munich to characterize the aroma profiles of pea proteins (Pisum sativum L.). Each attendee trained weekly for a minimum of two years to be able to distinguish aroma qualities and quantities [16]. All panelists agreed to contribute and had no history of known anosmia. For aroma evaluation, aqueous solutions of the following reference odorants (20 mL; 10-fold odor thresholds) for the given odor qualities were used for quantitative descriptive analysis (QDA) training, according to Stone and Sidel [26]: hexanal (25.0 µg/L) for grassy, 3-isopropyl-2-methoxypyrazine (0.1 µg/L) for beans-like, acetic acid (60 mg/L) for sour, (E,E)-2,4-decadienal (0.32 µg/L) for fatty, phenylacetic acid (0.68 mg/L) for honey-like, 2-ethyl-5-methylpyrazine (1.0 mg/L) for nutty, 3-methylbutanal (5.0 µg/L) for malty, and 2,3,5-trimethylpyrazine (0.12 mg/L) for earthy. All sensory analyses took place in special sensory cabins, and temperature was regulated to 20-25 • C. Data were evaluated with Excel 2016 (Microsoft, Redmond, WA, USA) as well as Origin 2018b 9.55 (OriginLab Corporation, Northampton, MA, USA).
Aroma profile analysis (APA): For orthonasal aroma profile analysis, 10% pea protein was suspended in a mixture of water/triacetin (97.5:2.5, w/w) and presented in closed sensory vials (45 mL) to the panelists. By sniffing, the aroma intensities for grassy, beanslike, sour, fatty, honey-like, nutty, malty, and earthy notes were rated from 0 (not detectable) to 5 (very intense), and thus the aroma profiles were characterized.
Recombination studies and comparative APA (cAPA): For aroma recombination, deodorized pea protein powder with strongly minimized characteristic pea aroma was generated using the following steps: Pea protein C (500 g) was stirred in freshly distilled n-pentane (2 × 1.5 L), then in dichloromethane (2 × 1.5 L) at room temperature overnight, and was removed from solvent on the next day. The obtained pea protein powder was dried under a stream of nitrogen and could not be sensorially related to pea by the panelists. Moreover, recombination solutions, including all aroma-active compounds (odor activity value, OAV ≥ 1) or parts of it (minimal recombinant) in native pea protein concentrations, were prepared in triacetin. Aroma recombinants were prepared by suspending 10% deodorized pea protein powder in water and the recombination solution in triacetin (97.5:2.5, w/w) and assessed in comparison to C, as described for the APA.
Omission tests: Incomplete recombinants, each lacking one aroma-active compound or group (OAV ≥ 1), were evaluated against complete recombinants by means of 3-alternative forced choice tests (3-AFC). Based on the number of correctly identified samples and attended panelists (13)(14), p values were determined by binomial distribution [27].

Estimation of Aroma Contribution
As only odorants with concentrations achieving their individual threshold contribute to the overall aroma, OAV were calculated by using Equation (1). Per definition, analytes with values ≥ 1, determined by the ratio of the quantified amount to the orthonasal odor threshold, had an impact on the food's authenticity and, thus, were considered in the recombination experiments [25].

Method Development and Validation Experiments
To guarantee fast, selective, and sensitive quantitation of 3-NPH tagged odorants by means of UHPLC-MS/MS, reference solutions were initially derivatized and used for software-assisted ramping of ion source and ion path parameters by syringe infusion [15,16].  Table S1). By establishing scheduled detection windows of ±30 sec, a suspension of 40 mg of protein powders in 1.0 mL acetonitrile/water derivatized with 3-NPH ( Figure 1) proved appropriately sensitive to detect all individual analytes [15,16].
Validation experiments revealed no crucial matrix effects for all analytes expressed as parallel upward shifted calibration curves (standard addition calibration), e.g., for hexanal and hexanoic acid ( Figure 2). Consequently, recovery rates were calculated by derivatizing known concentrations of each analyte in acetonitrile/water (50:50, v/v) and analyzing by means of 3-NPH-UHPLC-MS/MS. Determined recovery rates ranged between 80.5 and 106.9%. The LOD and LOQ were very low, varying from <0.1 to 5.8 nmol/L and <0.1 to 19.2 nmol/L, respectively ( Table 2). All LOQs showed higher sensitivity than specific aroma thresholds or could be counterbalanced by increasing sample loading, guaranteeing a detection of all analytes to the full extent.

Sensomics-Assisted Aroma Decoding
Using the Sensomics approach, the aroma composition of pea protein C should be quantitatively decoded and re-engineered. Therefore, in a first step, OAV were calculated via the ratio of concentration and corresponding odor threshold in water. This procedure resulted in a list of 27 analytes with OAV ≥ 1 (Table 3) (Table S4).  Thus far, odorants in pea protein have only once been quantified by means of GC-MS, and relative quantities of several analytes were calculated using hexanal-d 12 as an internal standard, yielding in concentrations of 83 mg/kg of hexanal and 2.1 mg/kg of phenylacetaldehyde. (E,E)-2,4-decadienal could not be quantitated because of coeluted peaks [10].
Based on the results by means of accelerated targeted UHPLC-MS/MS analysis, and on a SIDA using various IS with different functional groups matched on the examined analytes and added before sample preparation, the present quantitation approach provides reliable and precise insights into the quantitative aroma composition of different pea proteins (Pisum sativum L.).

Pea Protein Aroma Simulation and Omission Experiments
By means of cAPA, a complete aroma recombinant, consisting of all 27 analytes with OAV > 1 (Table 3) in native concentrations in a deodorized pea protein solution, was evaluated sensorially in comparison to C (Figure 4). The results proved the quantified aroma compounds in their correct ratio and the successful aroma reconstitution, respectively. . Aroma profile analysis of 10% C in water/triacetin (97.5:2.5, w/w) and its complete as well as minimal recombinant (10% deodorized protein in water/triacetin, 97.5:2.5, w/w). Odorants specified in brackets were used as reference attributes for the corresponding odor quality and perceived intensities were rated from 0 (not detectable) to 5 (very intense) by the panelists.

OAV Mapping of Commercially Available Pea Proteins
As aroma re-engineering and discovery of KFO by means of the Sensomics approach is complex and time-consuming, the question arose whether the impact of revealed KFO in pea protein C was also transferrable to the overall aroma of pea proteins (Pisum sativum L.) in general. Therefore, further pea protein isolates (A, B, D, E, F, G, H, I, J) were also analyzed with the described UHPLC-MS/MS methods and checked for molecular sensory differences. To be able to draw reliable conclusions, OAVs were calculated (Table S5) and logarithmically visualized based on the quantified concentrations and odor thresholds of each analyte in the examined pea protein samples ( Figure 5). The first insights indicated that mainly aldehydes with relatively low odor thresholds automatically clustered together, underlining the importance on the overall aroma among all pea proteins. Moreover, eight out of nine KFO could be spotted within this cluster, namely 3-methylbutanal, hexanal, acetaldehyde, (E,E)-2,4-nonadienal, 2-methylbutanal, benzaldehyde, (E)-2-octenal, and heptanal.
Except for nonanoic acid, also determined as KFO in C, and diacetyl, further acids, ketones, and γ-octalactone seemed to play a minor role for the aroma. Thus, statistical cluster analysis additionally substantiated the sensory results obtained by omission experiments.
Furthermore, cluster analysis also revealed intra-pea protein variations expressed by two different protein groups: A, B, J, C, G (cluster 1) and D, E, F, H, I (cluster 2). While, e.g., (E)-2-dodecenal was highly present in cluster 2, it did not exceed the OAV > 1 for pea proteins within cluster 1. Additionally, the individual contribution of KFO aldehydes slightly differed among the examined pea proteins.
In summary, OAV heatmapping highlighted comparable influences on the overall aroma of pea proteins and emphasized the group of aldehydes. Moreover, there were strong indications that the examined KFO of pea protein C were transferable as general KFO among different pea proteins (Pisum sativum L.).

Conclusions
Pea protein (Pisum sativum L.) could be a promising ingredient for the development of new functional foods such as meat substitutes. Unfortunately, the characteristic grassy, green, and bean-like off-flavor often distracts consumers, whereby these foods suffer from a lower acceptance and consequently need to be optimized regarding flavor. By means of high-throughput UHPLC-MS/MS analysis including 3-NPH derivatization, the presence and concentration of selected odorants, described to be important in different pea or related matrices, could be evaluated within a few minutes. OAV of the quantified analytes revealed 27 odor-active compounds, which resulted in their distinctive concentration composition in a confirmative aroma recombination. Finally, recombination and omission experiments as well as OAV heatmapping highlighted nine general key food odorants in pea protein, namely, 3-methylbutanal, hexanal, acetaldehyde, (E,E)-2,4-nonadienal, (E)-2-octenal, benzaldehyde, heptanal, 2-methylbutanal, and nonanoic acid ( Figure 6). The present investigation provides new insights into the aroma of different pea proteins (Pisum sativum L.) by KFO identification as well as cluster analysis and, therefore, may be helpful for knowledge-based flavor optimization of foods using pea protein as a (main) ingredient.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/foods11030412/s1, Table S1: MRM transitions of analyzed 3-NPH tagged odorants, Table S2: MRM transitions of pyrazines, Table S3: Used IS, IS calibration curves, and R2 of the quantified odorants, Table S4: Concentrations of the quantified odorants in different pea protein samples, Table S5: OAV of the quantified odorants in different pea protein samples and R script of OAV heatmap.