Unraveling the Impacts of Preprocessing on the Metabolite Profile of Bitter Almond Using UPLC-MS/MS and GC–MS Analysis

Abstract

As a traditional Chinese medicine, bitter almonds contain amygdalin, which can be hydrolyzed by β-glucosidase to release highly toxic hydrogen cyanide. There are safety hazards in clinical use, and high-temperature pretreatment is required for use. This study is based on the differences in β-GC activity of bitter almonds treated with different pretreatment methods. SPME-GC-MS and UPLC-MS/MS combined techniques were used to analyze the metabolic profile differences before and after processing from both volatile and nonvolatile dimensions. A total of 30 compounds were identified through volatile component analysis, among which terpenoids such as d-limonene and linalool were significantly upregulated during steaming, while benzyl alcohol was significantly downregulated. A total of 199 metabolites were identified through nonvolatile component analysis, and 80 significantly different metabolites were screened. Among them, the number of downregulated substances in steaming (45) far exceeded that in boiling (19) and frying (23), showing a clear trend of substance degradation. Of particular importance is the significant degradation of the cyanide alkaloid prunasin during the processing, resulting in the formation of volatile products such as benzaldehyde and hydrogen cyanide, which confirms the detoxification mechanism of processing at the molecular level. KEGG pathway enrichment analysis revealed that differential metabolites mainly involve 56 pathways such as α linolenic acid metabolism and purine metabolism, constructing a complex metabolic network during the processing.

1. Introduction

Bitter almond is the dried mature seed of Prunus armeniaca L. (Rosaceae). It is a commonly used traditional Chinese medicine for relieving cough and asthma, and was first recorded in Shennong’s Herbal Classic. In traditional Chinese medicine, bitter almond is used to reduce qi, relieve cough and asthma, and moisten the intestines to promote bowel movements. Modern pharmacological studies have shown that bitter almond and its preparations possess a variety of biological activities, including anti-inflammatory, antitussive, expectorant, antiasthmatic, immunoregulatory, and antitumor effects. Bitter almonds are rich in nutrients and bioactive substances, including unsaturated fatty acids, proteins, sugars, minerals, vitamins, tocopherols, phytosterols, and secondary metabolites such as cyanidins, flavonoids, phenolic acids, and triterpenoids. Among these constituents, amygdalin is a characteristic cyanogenic glycoside that is not only a major pharmacologically active component responsible for the cough-relieving and antiasthmatic effects of bitter almond but also the source of its toxicity. Under the action of β-glucosidase, amygdalin can be hydrolyzed to produce benzaldehyde and hydrogen cyanide, the latter of which is a potent inhibitor of cellular respiration. Excessive intake may therefore lead to poisoning or even death. The coexistence of efficacy and toxicity makes the quality control and rational clinical use of bitter almond a persistent challenge in the modernization of traditional Chinese medicine.

Pretreatment is an essential component of traditional Chinese medicine processing. Through operations such as heating, addition of excipients, and comminution, pretreatment can modify drug properties, facilitate formulation and storage, and potentially enhance efficacy while reducing toxicity. Before clinical use, bitter almonds are commonly subjected to thermal pretreatments, such as frying and steaming. In addition to reducing cyanogenic risk, these two processes provide distinct physicochemical environments for metabolite transformation. Frying is a dry heat treatment with relatively low water activity and greater oxygen exposure, conditions that may favor Maillard/Strecker-type reactions and the accumulation of lipid-oxidation intermediates. By contrast, steaming is a moist-heat treatment characterized by rapid heat transfer and abundant water, which may facilitate enzyme denaturation, hydrothermal conversion, and broader remodeling of phenylpropanoid-, lipid-, and cyanogenic-related metabolites. Therefore, comparing these pretreatments at the metabolome level may help explain how different heat transfer environments reshape the chemical profile of bitter almond. Recent metabolomics studies have shown that roasting, one of the most common almond-processing methods, can improve color, flavor, and texture while also altering microstructure, physicochemical properties, unsaturated fatty acid composition, and volatile aroma compounds. However, most existing studies have focused on changes in one or only a few components, and a comprehensive understanding of pretreatment-dependent remodeling of the overall bitter almond metabolome remains lacking.

Metabolomics, as an important component of systems biology, qualitatively and quantitatively studies all small-molecule metabolites in biological systems through high-throughput detection and multivariate statistical analysis, revealing the overall metabolic response characteristics of organisms under physiological, pathological, or environmental interventions. In recent years, this technology has been widely used in the quality control, authenticity identification, processing mechanism interpretation, and pharmacological substance basis research of traditional Chinese medicine. Compared with traditional single-component analysis, metabolomics has the advantages of “holistic” and “systematic,” which can reflect the essential characteristics of multicomponent, multitarget, and synergistic effects of traditional Chinese medicine. In plant metabolomics research, ultrahigh-performance liquid chromatography–mass spectrometry (UPLC-MS/MS) and gas chromatography–mass spectrometry (GC–MS) are two mainstream analytical platforms. UPLC-MS/MS exhibits unique advantages in detecting polarity, thermal instability, and large molecular metabolites due to its high separation, sensitivity, and precision; GC–MS is good at analyzing volatile, semivolatile, and nonpolar small-molecule compounds, but usually requires derivatization treatment to improve detection sensitivity. The two technology platforms are highly complementary in terms of metabolite coverage, and their combined application can achieve comprehensive characterization of the plant metabolome. Similarly, GC–MS combined with LC-MS/MS-QTOF technology identified 13 and 30 metabolites, respectively, in the metabolic profile analysis of papaya seeds, revealing significant differences in chemical composition among different varieties.

Although considerable progress has been made in studies of the chemical composition and pharmacological activity of bitter almond, the differential effects of traditional pretreatment methods on the overall metabolome have not yet been systematically compared. In addition, metabolite degradation, oxidation, and enzymatic transformation during sample preparation and storage may affect data reproducibility and comparability. Therefore, a standardized analytical workflow and an integrated multiplatform strategy are needed to investigate the relationship between pretreatment and metabolite remodeling in bitter almond. This study is based on the differences in β-glucosidase activity in bitter almond samples, and further combined with UPLC-MS/MS and GC–MS to systematically compare crude bitter almonds, boiled bitter almonds, fried bitter almonds, and steamed bitter almonds. The aims were not only to characterize differential metabolites among these pretreatment groups but also to interpret the observed changes in the context of enzyme inactivation, Maillard-related chemistry, lipid oxidation, and pathway-level remodeling induced by dry heat and moist-heat processing. This study therefore provides a metabolomics-based framework for understanding pretreatment-dependent changes in bitter almond and offers a useful reference for future optimization and quality evaluation of bitter almond pretreatment.

2. Materials and Methods

2.1. Instruments and Reagents

β-Glucosidase (β-GC) activity detection kit (Adamas life), Cary 60 UV Vis spectrophotometer (Agilent), 8890-7000D GC–MS/MS (Agilent), DB-5MS (30 m × 0.25 mm × 0.25 μm) column (Agilent), MM400 ball mill (Retsch), MS105DU electronic balance (METTLER TOLEDO), 120 μ m DVB/CWR/PDMS extraction head (Agilent), SPME Arrow solid-phase microextraction device (CTC Analytics AG), Fiber Conditioning Station aging device (CTC Analytics AG), agitator sample heating box (CTC Analytics AG), 5424R centrifuge (Eppendorf), and MU-G02-0448 constant temperature metal mixer (Hangzhou Mio Instrument Co., Ltd.) were used. Distilled water, methanol (Merck), acetonitrile (Shanghai Xingke), formic acid (Aladdin), sodium chloride (Sinopharm), and n-hexane (CNW) were all of chromatographic grade.

2.2. Experimental Samples and Pretreatment Methods

Bitter almond samples were collected from Longde County, Guyuan City, Ningxia Hui Autonomous Region, China, and authenticated by Ding Rui (Chief Pharmacist, Ningxia Hui Autonomous Region Drug Inspection and Research Institute) as Prunus armeniaca L. var. ansu Maxim. All processed bitter almond products were prepared in accordance with the Pharmacopoeia of the People’s Republic of China. Four treatment groups were included in this study: crude, boiled, fried, and steamed bitter almonds. For each group, 3 independent biological replicates were prepared, and each biological replicate represented an independently processed batch derived from the same raw material source. Crude bitter almonds were prepared by removing the flesh and shell, collecting the seeds, drying them, and peeling them. Boiled bitter almonds are prepared by boiling 10 times the amount of water in a boiling-water pot for 10 min until the seed coat of the bitter almonds shrinks and stretches. When the seed coat is easy to rub off, it is removed and placed in cold water to remove the seed coat. Fried bitter almonds were prepared by heating peeled, dry bitter almonds over low heat until the surface turned yellow. Steamed bitter almonds were prepared by steaming peeled dry bitter almonds over high heat in a boiling-water steamer for 30 min. The same set of biological replicates was used for both UPLC-MS/MS and GC–MS analyses. However, because the two analytical platforms required different sample preparation procedures, parallel aliquots from each biological replicate were analyzed separately by UPLC-MS/MS and GC–MS. The analytical injection order was randomized before instrumental analysis.

2.3. β-GC Activity Detection

Refer to the instructions of the β-GC activity detection kit for experimental methods.

2.4. UPLC-MS/MS Detection Conditions

For UPLC-MS/MS analysis, each biological replicate was freeze-dried and ground into powder at 30 Hz for 1.5 min. A 30 mg aliquot from each biological replicate was extracted with 1500 μL of precooled 70% methanol containing the internal standard. The mixture was vortexed for 30 s every 30 min for a total of six cycles and then centrifuged at 12,000 rpm for 3 min. The supernatant was filtered through a 0.22 μm membrane and transferred to an autosampler vial for LC-MS analysis.

To monitor analytical stability, a pooled QC sample was prepared by mixing equal aliquots of 20 μL from all LC-MS sample extracts. The pooled QC sample was injected 3 times at the beginning of the run to condition the system and subsequently once every 4 study samples throughout the analytical sequence.

The sample extracts were analyzed using a UPLC-ESI-MS/MS system. Chromatographic separation was performed on an Agilent SB-C18 column (1.8 μm, 2.1 mm × 100 mm). The mobile phase consisted of solvent A (water containing 0.1% formic acid) and solvent B (acetonitrile containing 0.1% formic acid). The gradient program was as follows: 95% A and 5% B at the initial condition; a linear gradient to 5% A and 95% B within 9 min; maintenance at 5% A and 95% B for 1 min; return to 95% A and 5% B within 1.1 min; and re-equilibration for 2.9 min. The flow rate was 0.35 mL/min, the column temperature was maintained at 40 °C, and the injection volume was 2 μL. The effluent was alternately introduced into an ESI-triple-quadrupole-linear ion trap (QTRAP)-MS system.

The ESI source parameters were as follows: source temperature, 500 °C; ion spray voltage (IS), 5500 V in positive ion mode and −4500 V in negative ion mode; ion source gas I (GSI), gas II (GSII), and curtain gas (CUR), 50, 60, and 25 psi, respectively; and collision-activated dissociation (CAD), high. Triple-quadrupole scans were acquired in multiple reaction monitoring (MRM) mode with nitrogen as the collision gas at the medium setting. The declustering potential (DP) and collision energy (CE) for each MRM transition were further optimized individually. A specific set of MRM transitions was monitored in each acquisition period according to the metabolites eluting within that period.

2.5. GC–MS Detection Conditions

For GC–MS analysis, a separate powdered aliquot prepared from the same biological replicate described in Section was used. Specifically, 0.2 g of sample powder was transferred to a 20 mL headspace vial containing 0.2 g NaCl to inhibit enzymatic reactions. 0.2 g of the sample was transferred immediately to a 20 mL headspace vial, containing 0.2 g NaCl powder to inhibit any enzyme reaction. The vials were sealed using crimp-top caps with TFE-silicone headspace septa (Agilent). At the time of SPME analysis, each vial was placed at 60 °C for 5 min, and then a SPME Arrow (Agilent) of 120 μm DVB/CWR/PDMS was exposed to the headspace of the sample for 15 min at 60 °C.

For SPME analysis, each vial was equilibrated at 60 °C for 5 min, after which a 120 μm DVB/CWR/PDMS SPME Arrow fiber was exposed to the headspace of the sample for 15 min at 60 °C. After extraction, VOCs were desorbed from the SPME Arrow coating in the GC injection port at 250 °C for 5 min.

VOCs were analyzed using an Agilent 8890 GC coupled to a 7000D mass spectrometer and equipped with a DB-5MS capillary column (30 m × 0.25 mm × 0.25 μm; 5% phenyl-polymethylsiloxane). Helium was used as the carrier gas at a flow rate of 1.2 mL/min. The injector temperature was maintained at 250 °C. The oven temperature program was as follows: 40 °C for 3.5 min; increased at 10 °C/min to 100 °C; increased at 7 °C/min to 180 °C; increased at 25 °C/min to 280 °C; and held for 5 min. Mass spectra were recorded in electron impact (EI) ionization mode at 70 eV. The quadrupole, ion-source, and transfer-line temperatures were set at 150, 230, and 280 °C, respectively. The mass spectrometer was operated in selected ion monitoring (SIM) mode for analyte identification and quantification.

2.6. Data Processing and Metabolomics Analysis

LC-MS data were processed using Analyst 1.6.3 software. Metabolites were qualitatively annotated and relatively quantified on the basis of the local MWDB database (Metware database). For the untargeted LC-MS data set, metabolite annotations were assigned by matching retention time, precursor ion, and MS/MS fragmentation patterns against the MWDB database and were reported as putatively annotated compounds.

Characteristic ions for each analyte were selected through triple-quadrupole screening, and the corresponding ion signal intensities (counts per second, cps) were recorded by the detector. The raw mass spectrometry files were then processed by using MultiQuant software for chromatographic peak integration and calibration. The peak area of each chromatographic peak was used to represent the relative abundance of the corresponding metabolite, and all integrated peak area data were exported for subsequent analysis.

For the GC–MS data set, chromatographic peaks were processed using MassHunter. VOCs were tentatively identified by comparison of EI mass spectra with the NIST mass spectral library, using a library match score of >70 together with consistent retention behavior [or retention index, if available]. Peaks that did not meet these criteria were excluded from the further analysis.

2.7. GC–MS Metabolite Relative Content Analysis

The internal standard semiquantitative method refers to a suitable compound standard added to the sample during the quantitative analysis process. In quantitative analysis, in order to ensure that the operation meets the requirements of the detection method, an appropriate compound standard is often selected as the accompanying reference material, and a known amount is added to the sample while detecting both the tested component and the reference material. According to the analysis of literature research results, isotopes are currently one of the commonly used internal standard reagents. Based on this, we selected deuterated isotope internal standards in n-hexane and calculated the relative content of volatile organic compounds in the sample using the following formula. The formula for calculating the relative content of compounds in solid samples is as follows:

$$X_i=\frac{V_{\mathrm{s}}\times C_{\mathrm{s}}}{M}\times \frac{I_i}{I_{\mathrm{s}}}\times {10}^{-3}$$

X_i is the content of compound i in the test sample (μg/g); V _s is the volume of the internal standard added (μL); C _s is the concentration of the internal standard substance (μg/mL); M is the amount of the sample to be tested (g); I _s is the peak area of the internal standard substance; and I _i is the peak area of compound i in the sample to be tested.

2.8. Statistical Analysis and Validation of Multivariate Models

Perform multivariate data analysis on the processed data, including principal component analysis (PCA), orthogonal partial least-squares discriminant analysis (OPLS-DA), and model validation analysis (permutation). VIP, P, and FC values were used to screen for differential metabolites between the two groups.

3. Results and Discussion

3.1. β-GC Activity Detection Results

The standard curve results of β-glucosidase activity are shown in Figure A, with a regression equation of y = 0.008x – 0.0362 and a correlation coefficient of R ² = 0.9929. The enzyme activities in crude bitter almonds, boiling bitter almonds, fried bitter almonds, and steamed bitter almonds are 18,203 ± 343.53 U/g, 16,174.33 ± 287.02 U/g, 13,039 ± 361.21 U/g, and 6046.33 ± 156.67 U/g, respectively. According to the activity detection results, the β-GC enzyme activity is highest in crude bitter almonds. The activity decreases step by step along the processing method from boiled bitter almonds to fried bitter almonds to steamed bitter almonds, with steamed bitter almonds having the lowest activity. Due to the fact that β-glucosidase can hydrolyze amygdalin to produce HCN, the higher the activity, the stronger the toxicity. Therefore, it has been proven that crude bitter almonds have the highest toxicity, and the toxicity is significantly reduced after processing.

1.

(A) β-glucosidase activity standard curve. (B) Sample enzyme activity.

3.2. Identification and Analysis of Non-VOCS in Bitter Almonds

On the basis of the UPLC-MS/MS detection results, the total ion chromatogram (Figure A,B) and multipeak detection chromatogram (Figure C,D) of the quality control samples were overlaid and analyzed, demonstrating good repeatability and reliability of the data recorded in this study. By using Analyst 1.6.3 to identify the substance information on crude bitter almonds, boiled bitter almonds, fried bitter almonds, and steamed bitter almonds and retaining the substances with a secondary mass spectrometry, retention time, and database substance matching score of 0.7 or above, a total of 199 nonvolatile metabolites were identified in positive and negative ion mode. The specific results are shown in Figure , including 27 amino acids and their derivatives, 24 lipids, 20 phenolic acids, 19 organic acids, 15 terpenes, 15 nucleotides and their derivatives, 15 alkaloids, 12 flavonoids, 7 lignin and coumarin, 2 quinones, and 43 other metabolites.

2.

(A, B) Total ion chromatogram. (C, D) Multipeak detection of metabolites under multi-reaction monitoring mode

3.

Classification pie chart of nonvolatile substances in bitter almonds.

Hierarchical clustering analysis was used to display the variation patterns between different treatment groups, as shown in Figure . The heatmap was drawn based on row-standardized data, and the color depth represented the relative abundance level. The unsupervised clustering results of the rows and columns showed that there were clear distinctions between the four treatment groups of crude bitter almonds, boiled bitter almonds, fried bitter almonds, and steamed bitter almonds. The biological duplicate samples within each treatment group were closely clustered together, indicating the good repeatability of the experiment. However, there were significant differences in the characteristic spectra between crude and processed products, and some metabolites were significantly expressed in high concentrations in steamed bitter almonds but were lowly expressed in the other three bitter almonds. Some metabolites in crude bitter almonds are significantly overexpressed, while they are lowly expressed in the other three processed bitter almond products. Boiled bitter almonds are similar to those of fried bitter almonds. This figure simply and intuitively reflects the significant differences in the compound content among the four bitter almonds.

4.

Overall clustering heatmap of raw bitter almonds and different processed products of bitter almonds.

Based on UV standardized data, principal component analysis (PCA) was performed to display the distribution patterns between samples. The results are shown in Figure , where the confidence ellipse is used to reflect the dispersion within each group. It can be observed that there is significant separation of metabolic groups between groups, and the quality control samples are well clustered together, indicating good instrument stability. The first two principal components, PC1 and PC2, explained 44.77% and 22.54% of the total variation, respectively, indicating a clear clustering trend among different groups. Samples from different groups showed significant separation on the first and second principal components, indicating large differences between sample groups. The steamed bitter almond group was significantly separated from other groups on the PC1 axis, while there was partial overlap between the boiled bitter almond group and the fried bitter almond group. This difference trend corresponds to the results of β-GC activity detection, indicating that steaming has a significant impact on crude bitter almonds.

To further understand the evolution trajectory of nonvolatile metabolites during the processing, 199 nonvolatile metabolites were used as independent variables, and four treatment groups were used as dependent variables to establish OPLS-DA models for boiling bitter almonds-crude bitter almonds, frying bitter almonds-crude bitter almonds, and steamed bitter almonds-crude bitter almonds, respectively. The score plots are shown in A, C, and E of Figure . The principal component 1 and principal component 2 of the three models are 54.7% and 11.6%, 59.1% and 12.5%, and 61.1% and 10.4%, respectively, indicating that the established models can explain the data matrix information and have good predictive ability. The OPLS-DA model was validated using 200 random permutation tests, and the results are shown in A, C, and E of Figure . The rightmost point corresponds to R ²Y and Q ² of the original model, while the remaining points represent R ²Y’ and Q ²’ of the permutation-tested model. The results show that the points on the left do not exceed the rightmost point, indicating the significance of the OPLS-DA model. The VIP parameters obtained on the basis of the model are reliable and can be used for further analysis and screening.

5.

PCA principal component analysis score plot of nonvolatile metabolites in bitter almond samples.

Based on the OPLS-DA results, Welch’s t-test was used to calculate VIP parameters, and combined with FC values (Fold Change), metabolites with significant differences were screened. Threshold conditions were set as VIP ≥ 1, FC ≥ 2, or FC ≤ 0.5, with P < 0.05. By combining single-factor analysis and multivariate statistical analysis methods, 80 significantly different metabolites were screened among different groups of bitter almond samples, mainly including phenolic acids, free fatty acids, nucleotides and their derivatives, amino acids and their derivatives, and alkaloids. The specific differential metabolite information is shown in Table .

1. Information on Metabolites with Significant Difference.

compounds	classification	boiled vs crude	fried vs crude	steamed vs crude
9-hydroxy-12-oxo-15(Z)-octadecenoic acid	Alkaloids	Down	Down	Down
vanillate	Alkaloids		Down	Down
9(10)-EpOME	Alkaloids	Up	Up	Down
glutathione disulfide	Alkaloids			Down
scopoletin	Alkaloids			Down
isoferulic acid	Amino acids and derivatives	Down	Down	Down
cinchonain 1a	Amino acids and derivatives			Down
crataegolic acid	Amino acids and derivatives			Down
zizybeoside I	Amino acids and derivatives			Down
chlorogenate	Amino acids and derivatives	Up	Up	Up
oleoylethanolamide	Amino acids and derivatives	Up	Up	Up
S-adenosyl-l-homocysteine	Amino acids and derivatives	Up	Up	Up
embelin	Amino acids and derivatives		Down
9,10-dihydroxystearate	Amino acids and derivatives		Up
lumichrome	Anthraquinone			Down
3,4-dihydroxybenzoate	Chalcones	Down	Down
uridine	Coumarins	Down	Down	Down
histidylleucine	Coumarins			Down
iminodiacetate	Coumarins			Down
sinapate	Coumarins			Down
isocitric acid	Flavanols	Down	Down	Down
l-arginine	Flavanols	Up	Up	Up
O-acetyl-l-serine	Flavanones			Down
xanthosine	Flavanones	Up	Up
3′,5′-cyclic GMP	Flavanonols	Up
caffeate	Flavonols	Down	Down	Up
ADP	Free fatty acids	Up	Up	Down
alphitolic acid	Free fatty acids	Up		Down
(+)-catechin	Free fatty acids			Down
(+)-syringaresinol	Free fatty acids			Down
α-d-glucose1,6-bisphosphate	Free fatty acids			Down
adenosine	Free fatty acids	Down	Down	Up
AMP	Free fatty acids	Down	Down	Up
aloe emodin anthrone	Free fatty acids	Up	Up
guanosine	Lignans			Down
FAD	Nucleotides and derivatives	Down	Down	Down
phloretate	Nucleotides and derivatives		Down	Down
isoguanosine	Nucleotides and derivatives			Down
neochlorogenic acid	Nucleotides and derivatives			Down
nicotinamide	Nucleotides and derivatives			Down
pinobanksin	Nucleotides and derivatives			Down
isofraxidin	Nucleotides and derivatives	Down	Down	Up
isoscopoletin	Nucleotides and derivatives	Down	Down	Up
naringenin	Nucleotides and derivatives	Down	Down	Up
phenyllactate	Nucleotides and derivatives	Up	Up	Up
myo-inositol 4-phosphate	Nucleotides and derivatives	Up	Up
2,3-dihydroxybenzoate	Organic acids	Down	Down	Down
12(13)-EpOME	Organic acids			Down
soyasapogenol E	Organic acids			Up
9-hydroxy-12-oxo-10(E),15(Z)-octadecadienoic acid	Organic acids	Down	Down
4-hydroxyaniline	Organic acids	Up
guanine	Others	Down
3′,5′-cyclic AMP	Phenolic acids	Down	Down	Down
indolelactate	Phenolic acids	Down	Down	Down
methyloxaloacetate	Phenolic acids	Down	Down	Down
eriodictyol chalcone	Phenolic acids		Down	Down
(7S,8S)-DiHODE	Phenolic acids			Down
(S)-mandelonitrile β-d-glucoside	Phenolic acids			Down
13(S)-HOT	Phenolic acids			Down
Kaempferol 3-O-rhamnoside-7-O-glucoside	Phenolic acids			Down
pinocembrin	Phenolic acids			Down
prunasin	Phenolic acids			Down
scopolin	Phenolic acids			Down
ferulate	Phenolic acids	Up	Up	Up
l-Aspartate	Phenolic acids	Up	Up	Up
5-Oxoproline	Phenolic acids		Up	Up
9,10,13-TriHOME	Phenolic acids			Up
2,5-dihydroxybenzoate	Phenolic acids	Up		
ursolic acid	Plumerane	Up		Up
oleanolic acid	Quinones			Down
betulinic acid	Saccharides	Up	Up	Down
benzoate	Saccharides		Up	Up
5-acetamidopentanoate	Triterpene		Down	Down
(R)-2-hydroxyisocaproate	Triterpene			Up
4-p-coumaroylquinic acid	Triterpene			Up
4′-O-β-d-glucosyl-cis-p-coumarate	Triterpene		Down
(R)-mevalonate	Triterpene	Up	Up
3-(4-hydroxyphenyl)lactate	Triterpene	Up	Up
glycyl-leucine	Vitamin			Down
pyridoxine	Vitamin	Down

Through in-depth analysis of the changing patterns of specific substance categories, phenolic acid metabolites showed significant changes in all three processing methods, indicating that heat treatment affects the bioavailability of polyphenols. Heat can damage cell walls, mobilize phenolic compounds, improve their availability, enhance their oxidation process, and degrade them based on their thermal stability. Amino acid and its derivative metabolism showed upregulation advantages in the three methods, which is closely related to protein denaturation and hydrolysis caused by heat treatment. When the temperature increases, the spatial structure of the protein is disrupted, releasing more free amino acids, in the initial stage of heat treatment, enzymatic reactions are significantly enhanced, leading to the release of sugars and amino acids, and simultaneously initiating the Maillard reaction. Three methods involve differential regulation of 15 nucleotide metabolites, and metabolic changes reveal the deep intervention of heat treatment on the genetic information molecules of bitter almonds. Studies have shown that heat treatment above 70 °C significantly promotes the enzymatic release of nucleotides, but sustained high temperatures lead to further degradation into free bases. Free fatty acids show a significant downward trend after processing, which is consistent with literature reports that increasing baking temperature can lead to a decrease in oleic acid and linoleic acid content. The downregulation trend of alkaloid metabolites after processing is due to the hydrolysis and degradation of cyanogenic alkaloids such as prunasin and phenylacetonitrile glucoside, which generate volatile products such as benzaldehyde and HCN at high temperatures, thereby achieving a detoxification effect.

Through in-depth analysis of the changing patterns of specific substance categories, phenolic acid metabolites showed significant changes in all three processing methods, indicating that heat treatment affects the bioavailability of polyphenols. Heat can damage cell walls, mobilize phenolic compounds, improve their availability, enhance their oxidation process, and degrade them based on their thermal stability. Amino acid and its derivatives metabolism showed upregulation advantages in the three methods, which is closely related to protein denaturation and hydrolysis caused by heat treatment. When the temperature increases, the spatial structure of the protein is disrupted, releasing more free amino acids, In the initial stage of heat treatment, enzymatic reactions are significantly enhanced, leading to the release of sugars and amino acids, and simultaneously initiating the Maillard reaction. Three methods involve differential regulation of 15 nucleotide metabolites, and metabolic changes reveal the deep intervention of heat treatment on the genetic information molecules of bitter almonds. Studies have shown that heat treatment above 70 °C significantly promotes the enzymatic release of nucleotides, but sustained high temperatures lead to further degradation into free bases. Free fatty acids show a significant downward trend after processing, which is consistent with literature reports that increasing baking temperature can lead to a decrease in oleic acid and linoleic acid content. The downregulation trend of alkaloid metabolites after processing is due to the hydrolysis and degradation of cyanogenic alkaloids such as naringenin and phenylacetonitrile glucoside, which generate volatile products such as benzaldehyde and hydrocyanic acid at high temperatures, thereby achieving a detoxification effect.

According to metabolomics statistics, three thermal processing methods, namely, boiling, frying, and steaming, have a systematic impact on nonvolatile metabolites of bitter almonds, involving 19 metabolic categories, including amino acids and their derivatives, phenolic acids, nucleotides and their derivatives, flavonoids, triterpenoids, organic acids, free fatty acids, and sugars. These changes do not occur in isolation but are interconnected through mechanisms such as thermal degradation, Maillard reaction, enzymatic conversion, and oxidative stress, collectively forming a complex metabolic network during the processing.

3.3. Analysis of Volatile Organic Compounds in Bitter Almonds

The total ion chromatogram of volatile component quality control samples is shown in Figure . By using MassHunter software to retrieve relevant mass spectrometry information, a total of 30 compounds with a matching degree greater than 70% and an absolute peak area greater than 3000 were identified, including 8 alcohols, 6 terpenes, 5 esters, 4 ketones, 3 aldehydes, 3 phenols, and 1 nitrogen-containing compound. By calculating the relative percentage content of each substance based on peak area, specific information on the metabolic profiles of four bitter almonds was obtained. The results are shown in Table . Among them, the relative content of benzyl alcohol is the highest, which is 339.9 ± 25.24 μg/g, 99.36 ± 8.64 μg/g, 80.28 ± 4.63 μg/g, and 34.43 ± 1.87 μg/g in crude bitter almonds, boiled bitter almonds, fried bitter almonds, and steamed bitter almonds, respectively. According to the literature, it is speculated that the benzaldehyde generated during the degradation of amygdalin is converted into benzyl alcohol under the catalytic action of alcohol dehydrogenase.

6.

Orthogonal partial least-squares discriminant analysis and permutation test. (A, C, E) The OPLS-DA model score plots for crude bitter almonds-boiled bitter almonds, crude bitter almonds-fried bitter almonds, and crude bitter almonds-steamed bitter almonds, respectively. (B, D, F) The corresponding model validation diagrams.

2. Relative Content Table of Volatile Components in 4 Different Bitter Almond Samples.

	relative content (μg/g)
compounds	crude	boiled	fried	steamed
furan, 3-(4-methyl-3-pentenyl)-	0.05 ± 0	0.05 ± 0	0.05 ± 0.01	0.16 ± 0.02
phenol, 2-methoxy-	0.07 ± 0	0.07 ± 0.01	0.07 ± 0	0.14 ± 0.02
acetic acid, phenylmethyl ester	1.04 ± 0.13	0.11 ± 0.01	0.13 ± 0.01	0.58 ± 0.5
2H-pyran-2-one, tetrahydro-	0.09 ± 0.01	0.12 ± 0.01	0.1 ± 0.01	0.09 ± 0
nonanal	0.03 ± 0.01	0.02 ± 0	0.03 ± 0	0.05 ± 0.01
linalool	0.05 ± 0.01	0.03 ± 0	0.03 ± 0	0.3 ± 0.01
benzeneacetic acid, methyl ester	1.1 ± 0.19	0.07 ± 0	0.06 ± 0	0.77 ± 0.06
benzyl alcohol	339.9 ± 25.24	99.36 ± 8.64	80.28 ± 4.63	34.43 ± 1.87
phenol	0.27 ± 0	0.29 ± 0.03	0.32 ± 0.02	0.28 ± 0.03
hydrazinecarbothioamide	0.48 ± 0.2	0.16 ± 0.11	0.3 ± 0.03	0.29 ± 0.01
.α.-pinene	0.09 ± 0	0.07 ± 0	0.04 ± 0	0.07 ± 0
benzenemethanol, .α.-methyl-	0.96 ± 0.11	0.45 ± 0.03	0.53 ± 0.04	0.43 ± 0.03
3-oxatricyclo[4.1.1.0(2,4)]octane, 2,7,7-trimethyl-	0.17 ± 0	0.14 ± 0.02	0.15 ± 0.01	0.23 ± 0.01
endofenchol	0.05 ± 0.01	0.04 ± 0.01	0.04 ± 0.01	0.62 ± 0.04
>d-limonene	0.04 ± 0	0.06 ± 0.01	0.06 ± 0	1.01 ± 0.03
3-heptanol, 6-methyl-	0.02 ± 0	0.02 ± 0	0.02 ± 0	0.02 ± 0
2-heptanol, 6-methyl-	0.03 ± 0	0.03 ± 0	0.06 ± 0.04	0.2 ± 0.12
3-heptanone, 5-methyl-	0.11 ± 0	0.21 ± 0.01	0.02 ± 0	0.02 ± 0
2(5H)-furanone, 5-methyl-	0.17 ± 0	0.31 ± 0.02	0.26 ± 0.01	0.03 ± 0
1-hexanol, 4-methyl-	0.15 ± 0	0.14 ± 0.02	0.12 ± 0.01	0.09 ± 0.01
3-methylthiobutyraldehyde	0.03 ± 0	0.04 ± 0	0.03 ± 0	0.04 ± 0
pentanoic acid, 3-methyl-, ethyl ester	0.13 ± 0.01	0.26 ± 0.01	0.23 ± 0.01	0.22 ± 0.01
2-cyclohexen-1-one, 3-methyl-	1.78 ± 0.03	1.42 ± 0.24	1.64 ± 0.1	1.76 ± 0.09
propanoic acid, 2-hydroxy-, ethyl ester	0.03 ± 0	0.04 ± 0	0.19 ± 0.02	0.03 ± 0
butylated hydroxytoluene	0.12 ± 0.01	0.24 ± 0.04	0.18 ± 0.02	0.21 ± 0.02
1-octen-3-yl-acetate	0.08 ± 0	0.08 ± 0.01	0.08 ± 0.01	0.23 ± 0.01
1-undecyn-3-ol	0.27 ± 0.01	0.28 ± 0.01	0.28 ± 0.01	0.55 ± 0.11
1,3-pentanediol	0.03 ± 0	0.03 ± 0.01	0.03 ± 0.01	0.02 ± 0
(Z,Z)-3,5-nonadienal	0.06 ± 0.01	0.09 ± 0.02	0.11 ± 0.03	0.15 ± 0.02
7-octen-4-ol, 2-methyl-5-methylene-, (S)-	0.04 ± 0	0.04 ± 0	0.04 ± 0	0.1 ± 0.02

From the perspective of substance categories, ester compounds exhibit differential responses in three processing methods. Pentanoic acid, 3-methyl-, ethyl ester shows a significant upregulation in both boiling and frying, while aromatic esters such as benzeneacetic acid, methyl ester, acetic acid, and phenylmethyl ester are generally downregulated. This difference may be due to the competitive relationship between esterification and hydrolysis reactions under different heat-treatment conditions. Studies have shown that esters are prone to hydrolysis reactions in high-temperature steam environments to produce corresponding alcohols and acids, while frying and boiling under dry heat conditions are more conducive to retaining or generating specific ester compounds. High-temperature treatment can also promote the esterification reaction between fatty acids and alcohols, generating ethyl ester compounds with fruity characteristics. Ketone compounds (3-heptanone, 5-methyl-, 2­(5H)-furanone, 5-methyl-) are significantly enriched during boiling, but their changes vary during frying and steaming, indicating that ketone generation is highly sensitive to temperature and processing methods. The generation of these compounds is closely related to lipid oxidation and the Maillard reaction. During heat treatment, unsaturated fatty acids undergo oxidative degradation, and methyl ketone compounds are generated through free radical chain reactions, boiling and frying, as dry heat treatment methods, provide suitable temperature conditions for lipid oxidation and Maillard reaction, while steaming, due to its high moisture content, may inhibit the progress of the partial oxidation reaction. Terpenes exhibit the most significant processing specificity, with d-limonene sharply upregulated during steaming, linalool and perilla significantly increased, and 7-octen-4-ol, 2-methyl-5-methylene-, and (S)- significantly downregulated during steaming, indicating that steaming has a unique retention and enrichment effect on thermosensitive terpenoids. Studies have shown that unstable monoterpenes can undergo thermal rearrangement or cracking reactions under high-temperature steam, limonene serves as a central intermediate to participate in the subsequent transformation of complex terpene structures, and can further dehydrogenate to generate stable derivatives such as p-cymene. Among phenolic compounds, butylated hydroxytoluene is significantly upregulated in fermentation, possibly due to the responsive accumulation of lipid oxidative stress, and phenol, 2-methoxy-, as a characteristic product of lignin thermal degradation and ferulic acid decarboxylation, was slightly downregulated during frying. Alcohol substances showed an overall downward trend, but endofenchol was significantly upregulated during steaming, consistent with the changes in terpenes. Among them, the main component of bitter almonds, benzaldehyde, can be oxidized to form benzoic acid or reduced to produce benzyl alcohol, and benzyl alcohol evaporates after high-temperature treatment. Nitrogen-containing hydrazinecarbothioamide showed a downward trend in all three treatments. Aldehyde (Z,Z)-3,5-nonadienal is significantly upregulated during steaming, and this compound belongs to lipid oxidation products, mainly derived from the oxidative degradation of linoleic acid.

The changes in these volatile components not only affect the odor characteristics of bitter almonds but may also be related to changes in the pharmacological substance basis of bitter almonds, such as the anti-inflammatory and analgesic activity of terpenoids, the antioxidant activity of phenolic compounds, etc. Therefore, in the optimization of bitter almond-processing technology, appropriate processing methods should be selected according to clinical medication needs to achieve the best balance between the efficacy and safety.

3.4. KEGG Pathway Enrichment Analysis

The Kyoto Encyclopedia of Genes and Genomes (KEGG) database is the main publicly available database for metabolic pathways, which can be used for general network research on gene-expression information and metabolite accumulation. In this study, a total of 99 differential metabolites from three control groups obtained by GC/MS and HPLC-MS/MS were summarized and classified into different pathways. The biosynthetic processes and biochemical reactions of multiple plant metabolites were obtained, as shown in Figure .

7.

GC–MS total ion chromatogram (TIC) of bitter almond quality control samples.

Based on the P value corrected by the BH method (P < 0.05), KEGG pathway enrichment analysis was performed on the data, and a total of 56 significantly enriched pathways were identified. The results showed that the vast majority of enriched pathways belonged to the metabolism category, accounting for 69.64% of the total number of pathways. Among them, metabolic pathways were the most significant entry (39, 69.64%), followed by biosynthesis of secondary metabolites (27, 48.21%) and biosynthesis of cofactors (9, 16.07%). In addition, specific pathways, such as flavonoid biosynthesis and purine metabolism, also showed significant enrichment. In contrast, pathways for genetic information processing, environmental information processing, and cellular process categories account for a relatively small proportion of the total. These metabolic pathways are closely interconnected and interwoven in the biosynthesis and metabolism of compounds, forming a complex metabolic network. The comprehensive utilization of metabolites is achieved through shared intermediates and precursors, providing diverse possibilities for the biosynthesis. This series of related metabolic pathways reveals the interconversion between organic compounds during the high-temperature processing of bitter almonds, indicating that the high-temperature pretreatment of bitter almonds may exert pharmacological effects by regulating the key signaling pathways mentioned above, fully reflecting the complexity and coordination of organic compound metabolism (Figure ).

8.

Metabolic pathway analysis of differential metabolites.

4. Conclusions

This study is based on the differences in β-GC activity among different pretreatment methods of bitter almonds. Metabolomics technology was used to systematically analyze the chemical composition differences of bitter almonds and their three pretreatment methods (boiling, frying, steaming). SPME-GC-MS and UPLC-MS/MS were used to analyze crude bitter almonds and different processed products from two dimensions: volatile and nonvolatile components.

The activity of β-glucosidase decreases step by step from crude products to boiled products, fried products, and steamed products, confirming that high temperature can significantly reduce toxicity. In the analysis of volatile components, a total of 30 compounds were identified. PCA and OPLS-DA analyses showed that the steaming group had the most significant differences from other groups. Steaming treatment led to a sharp upregulation of terpenoids such as d-limonene and linalool, while alcohols such as benzyl alcohol were significantly downregulated, indicating that steaming has a unique retention and enrichment effect on thermosensitive terpenoids. A total of 199 metabolites were identified through nonvolatile component analysis, and 80 significantly different metabolites were screened. The results showed that steaming had the most profound impact on the metabolite spectrum of bitter almonds, with a much greater number of downregulated substances than boiling and frying, showing a clear trend of substance degradation. Of particular concern is that alkaloid metabolites were generally downregulated after processing, among which cyanogenic alkaloids such as prunasin undergo hydrolysis degradation during steaming, producing volatile products such as benzaldehyde and hydrogen cyanide, thus achieving detoxification effects. This discovery provides a molecular-level scientific basis for the detoxification mechanism of bitter almonds pretreatment. KEGG pathway enrichment analysis revealed that the differential metabolites between volatile and nonvolatile metabolites mainly involve 56 pathways, including metabolic pathways, secondary metabolite biosynthesis, flavonoid biosynthesis, and purine metabolism. A complex metabolic network was constructed during the pretreatment process.

This study confirms from the perspective of metabolomics that steaming has the most profound impact on the chemical composition of bitter almonds, followed by boiling and frying; different processing methods may selectively regulate the volatile and nonvolatile component spectra of bitter almonds through mechanisms such as thermal degradation, enzymatic conversion, and oxidative stress, providing a molecular basis for elucidating the scientific connotation and process optimization of detoxification and efficiency enhancement of bitter almonds after pretreatment. Subsequently, the dynamic degradation process of key components such as amygdalin will be tracked, and research methods will be added, including modern processing techniques such as microwave, baking, and frosting, as well as processing parameters (temperature, time, water addition). An in-depth analysis will be conducted on the changes in metabolic flow during the processing.

Liwei Zhao was mainly responsible for the experimental design, execution, and writing of this study; Xuebin Li completed data analysis and assisted in experimental operations; Qingqi Feng is responsible for experimental operations and sample collection; Jianlong Ma is responsible for the outline design, review, and revision of the paper; and Fengxia Hao is responsible for providing fund support.

The authors declare no competing financial interest.