Unveiling the spatial metabolome and anti-atherosclerosis effects of Allium macrostemon Bunge and Allium chinense G. Don

2.1 Materials and chemicals

The fresh bulbs of Allium macrostemon Bge. (AMB) and Allium chinense G. Don (ACGD) were collected from their genuine producing areas including Shenyang in Liaoning province and Xinjian in Jiangxi province of China, respectively. And they were taxonomically identified by Professor Xiaojian Zhang who works in department of pharmacy, first affiliated hospital of Zhengzhou University, China. Then, their surfaces were disinfected with 75% alcohol, and immediately stored at −80 °C before use.

The assay kits for testing triglyceride (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), lactic dehydrogenase (LDH), Creatine kinase-MB (CK-MB) were all obtained from Nanjing Jiancheng Bioengineering Institute (Nanjing, China). Rosuvastatin calcium tablets (RSV) were purchased from Chia-Tai Tianqing Pharmaceutical Group Co., Ltd (Nanjing, China). High fat and high cholesterol (HFHC) diets with 1.25% cholesterol (XT108C) and control diets (XT102C) were purchased from Jiangsu Synergetic Pharmaceutical Bioengineering Co., Ltd (Nanjing, China). 2,5-Dihydroxybenzoic acid (DHB), formic acid, and lithium trifluoroacetic acid (LiTFA) were purchased from Sigma-Aldrich (Shanghai, China). LC-MS grade acetonitrile, methanol and water were collected from Fisher Scientific (Fair Lawn, NJ). Other chemical reagents were of analytical grade.

2.2 Animal experiments

Male C57BL/6 mice (n = 6) and ApoE^−/− mice (n = 24) with 8–10 weeks old weighing (20 ± 2) g were purchased from Zhengzhou Chuangsheng Biological Engineering Co., Ltd (Zhengzhou, China), and fed in the Experimental Animals Center of Zhengzhou University. The mice were fed in an animal room with constant temperature (23 ± 2) °C and (60 ± 5)% relative humidity under a 12 h dark/light cycle. They had free access to diets and water ad libitum for a week before the tests. The animal protocols were approved and conducted based on the guidelines of Laboratory Animal Ethics Committee of Zhengzhou University. All experimental procedures were in accordance with the Guide for the Care and Use of Laboratory Animals (National Institute of Health).

Five groups including control group (wild-type C57BL/6 mice, control diets, n = 6), model group (ApoE^−/− mice, HFHC diets, n = 6), rosuvastatin group (ApoE^−/− mice, HFHC diets, n = 6), AMB group (ApoE^−/− mice, HFHC diets, n = 6) and ACGD group (ApoE^−/− mice, HFHC diets, n = 6) were set in this study. According to recommended doses for human and mice in previous reports (Chinese Pharmacopoeia Commission, 2020; Lee et al., 2010; Qin et al., 2016a), the dosages for AMB and ACGD extracts in this study were both 600.0 mg/kg/day. And the RSV dosage for intragastric administration was 2.0 mg/kg/day. During HFHC modeling for 12 weeks, the mice were treated with 0.3%CMC-Na, 0.3%CMC-Na, RSV solution, AMB extracts and ACGD extracts for control group, model group, RSV group, AMB group and ACGD group, respectively. All mice were fasted overnight before experiments. Blood samples were collected through the extraction of eyeball method. And serum samples were obtained after centrifugation at 3000 rpm for 10 min. The assay kits were used to test the serum levels of TC, TG, HDL-C, LDL-C, LDH and CK-MB. The heart and aortic valve tissues were fixed in 4% paraformaldehyde in phosphate buffered saline (PBS) and further embedded in paraffin. The paraffin sections were used to perform hematoxylin-eosin staining, Masson staining and Oil Red O staining analysis.

2.3 Sample preparation and extraction

The fresh bulbs of AMB and ACGD were steamed penetrably and dried in the sun, respectively. The dried samples were freeze-dried by vacuum freeze-dryer (Scientz-100F). Then, the freeze-dried AMB and ACGD samples were crushed using a mixer mill (MM400, Retsch) with a zirconia bead for 90 s (30 Hz). The obtained lyophilized powders (100 mg) were dissolved in 70% methanol solution (1.2 mL), and vertexing the mixture for 30 s every 30 min for 6 times in total. The samples were stood in a 4 °C refrigerator overnight, and centrifugated at 13800 g for 10 min. The supernatants were filtrated with 0.22 μm membrane (SCAA-104, ANPEL, Shanghai, China) and injected into UHPLC-MS/MS or UHPLC- triple-TOF/MS system to acquire data.

2.4 UHPLC-MS conditions

The chromatographic separation was achieved using a UHPLC system (SHIMADZU Nexera X2) on an Agilent SB-C18 column (1.8 µm, 2.1 mm × 100 mm) with water and acetonitrile both including 0.1% formic acid. The gradient programs employed the starting conditions of 5% B. Within 9 min, a linear gradient to 95% B was programmed, and a composition of 95% B was kept for 1 min. Subsequently, a composition of 5.0% B was adjusted within 1.1 min and kept for 2.9 min. The flow rates and column temperature were 0.35 mL/min and 40 °C, respectively. The injection volume was 4 μL. The effluent was alternatively connected to an ESI-QTRAP/MS or an ESI-TOF/MS system.

The UHPLC system was coupled to a triple quadrupole-linear ion trap mass spectrometer (AB4500 Q TRAP system) equipped with an ESI turbo ion-spray interface, operating in both positive and negative ion modes, and controlled and processed by Analyst 1.6.3 software (AB SCIEX). The operation parameters were as follows: source temperature 550 °C; ion spray voltage 5500 V (ESI+)/−4500 V (ESI−); ion source gas I, gas II, curtain gas was set at 50, 60, and 25 psi, respectively; polypropylene glycol solutions with 10 and 100 μmol/L were used to carry out the instrument tuning and mass calibration in QQQ and LIT modes, respectively. And QQQ scans were performed under multiple reactions monitoring (MRM) modes with collision gas (nitrogen) set to medium. De-clustering potential (DP) and collision energy (CE) were optimized based on the characteristics of each metabolite. A specific set of optimized MRM transitions were monitored for each period according to the detected metabolites eluted within this period.

For the qualitative analysis, the UHPLC system was coupled to a triple time of flight (TOF) 6600 mass spectrometer (AB SCIEX) with ESI mode. The parameters were as follows: source temperature 650 °C; ion spray voltage 5.5 kV (ESI+) and 4.5 kV (ESI−); gas I 40 psi; gas II 80 psi; curtain gas pressure 25 psi. TOF/MS scan information dependent acquisition (IDA)-product ion scan mode was applied to collect data. The IDA principle for precursor ions filtering was that the top 15 most intense precursor ions with an intensity greater than 10 cps were selected after the dynamic background subtraction was opened. The m/z ranges and accumulation times for TOF-MS scan were 100–2000 Da and 250 ms, respectively, while these for TOF-MS/MS were 50–2000 Da and 50 ms, respectively. Meanwhile, a collision energy of 35 eV and a collision energy spread of 15 eV were set for TOF-MS/MS analysis.

2.5 Chemical profiling of primary and secondary metabolites

For the identification and characterization of metabolites in AMB and ACGD, a self-built database of primary and secondary metabolites was established and named Metware Database (Wuhan Metware Biotechnology Co., Ltd., Wuhan, China) (Chen et al., 2013; Li et al., 2022b). This database annotated the detailed retention times, primary and secondary MS data of metabolites by UHPLC-TOF/MS and UHPLC-MS/MS. During the analysis, the interference signals including repeated signals of Na⁺, K⁺, and NH4⁺ ions, the isotope signal, and repetitive signals of fragment ions, were first excluded to ensure the accuracy and specificity of metabolite annotations. Further, the structures of metabolites were elucidated in reference to Metware Database and several public databases including MassBank (https://www.massbank.jp), ChemSpider (https://www.chemspider.com/), and METLIN (https://metlin.scripps.edu/index.php) (Yan et al., 2019).

2.6 Multivariate data analysis

After the acquired experiment data was unit variance scaled, an unsupervised principal component analysis (PCA) between AMB and ACGD samples were carried out using the Metware Cloud, a free online platform for data analysis (https://cloud.metware.cn). Furthermore, after the data was processed with log transform (log₂) and mean centering analysis, orthogonal partial least squares discrimination analysis (OPLS-DA) model was obtained, and the p values and variable importance in the projection (VIP) values were derived from OPLS-DA model. Meanwhile, a total of 200 permutations fitting was performed to avoid overfitting. Traditionally, the well-accepted principles for identifying the differential metabolites were that VIP values ≥1.0, fold change ≥2 or ≤0.5, and p values <0.05. In addition, normalized signal intensities of differentially expressed metabolites (unit variance scaling) were visualized as a color spectrum. The radar chart, bar chart and correlation net plot were also used to more visually display the differential metabolites between AMB and ACGD samples.

2.7 KEGG annotation and enrichment analysis

Identified differential metabolites between AMB and ACGD were annotated using KEGG Compound database (https://www.kegg.jp/kegg/compound/), whereas the annotated metabolites were then mapped to KEGG Pathway database (https://www.kegg.jp/kegg/pathway.html). The mapped pathways with significantly regulated metabolites were then fed into MSEA (metabolite sets enrichment analysis), and their significances were obtained by hypergeometric test’s p values. In addition, differential abundance score plot was used to better explain the differential metabolic pathways between AMB and ACGD groups.

2.8 Sample preparation for MALDI MSI assays

The fresh bulbs of AMB and ACGD were first crosscut. 0.5% CMC-Na solutions were used to fix the crosscut zone of fresh bulbs for cryosectioning. Leica CM1950 cryostat (Wetzlar, Germany) was used to crosscut the samples with 10 μm thickness at 20 °C. Afterwards, the cross sections were placed on electrically conductive slides coated with indium tin oxide (ITO), and the slides were dried in a vacuum desiccator for 0.5 h. Methanol-water solution (V/V, 9:1) containing 0.1 % LiTFA was used as the solvent to prepare DHB solution (15 mg/mL). Further, TM-Sprayer matrix spray apparatus was used to ensure that the DHB matrix solution was sprayed evenly on the ITO slide. The instrument parameters were as follows: temperature of 75 °C, flow rate of 0.1 mL/min, pressure of 8 psi; 24 cycles of sprays; and the drying time between each cycle was 10 s. The sprayed ITO slides were performed for data acquisition by MALDI timsTOF/MS.

2.9 MALDI MS imaging conditions

Bruker timsTOF flex MS system (Bruker Daltonics, Bremen, Germany) equipped with a smart beam 3D laser (10 kHz) was used to perform the MALDI MSI measurements. The laser power was set to 80%, and then it was fixed throughout the whole experiment. The mass ranges were m/z 200–1300 Da in positive ion mode. The imaging spatial resolution was set as 100 μm laser step size, and each MSI spectrum consisted of 500 laser shots. The MALDI mass spectra were first normalized with the Root Mean Square, and the signal intensity for each MSI image was displayed as the normalized intensity. Meanwhile, MS/MS fragmentations derived from the timsTOF flex MS system were used for further structural elucidation of the identified or tentatively characterized metabolites.

2.10 Statistical analysis

All data were presented as mean ± standard deviation (n = 6). The significant differences between treatment and control groups were obtained using Student’s t-test, and set at p < 0.05, p < 0.01 (**) and p < 0.001 (***).

3.1 Morphologies of AMB and ACGD bulbs

As shown in Fig. 1A, the morphologies of fresh AMB and ACGD bulbs are significantly different. AMB bulbs are externally yellowish-white or pale yellowish-brown, and irregularly circular or ovoid with 0.5–2.0 cm in diameter and 1.5–3.0 cm high (Fig. 1A). In addition, the fresh ACGD bulbs are externally pale yellowish-brown or brown, and lightly compressed long-ovate with 1.0–3.0 cm in diameter and 2.0–7.0 cm high (Fig. 1A). The cross cutting and longitudinal cutting graphs showed that the AMB and ACGD bulbs both consisted of tunics, leaf scales, and developing flower buds (Fig. 1B). After the fresh bulbs of AMB and ACGD were both steamed penetrably, the samples were dried in the sun. The final appearances of dried AMB and ACGD samples were displayed in Fig. 1C. The morphological sizes of ACGD were obviously larger than those of AMB.

Close

Fig. 1

The appearance of two Allium varieties. (A) the morphology of fresh AMB and ACGD bulbs; (B) the cross section and longitudinal section of two fresh bulbs; (C) dried bulbs of AMB and ACGD samples.

Export to PPT

3.2 Anti-atherosclerosis effects of AMB and ACGD extracts

In this study, the body, mesentery, epididymal fat, inguinal fat of HFHC-fed mice (with or without drug treatment) was weighed at the end of the experiment. As shown in Fig. 2A, mice in the model group exhibited a remarkable weight-gain compared to those in control group. Intervention with RSV, AMB or ACGD extracts for 12 weeks significantly reversed this increase of weight-gain in HFHC-fed mice as compared to those in model group. Consistently, a significant elevation in the weight of mesentery (Fig. 2B) and inguinal fat (Fig. 2D) was detected in the model mice with respect to the control mice, while no significant alterations of epididymal fat (Fig. 2C) were observed between model and control groups.

Close

Fig. 2

Effects of AMB and ACGD treatment on body weight, fat and metabolic lipid disorders in HFHC-induced AS development. (A) Representative mice pictures and final body weight; (B) appearance of mesentery and weight comparison; (C) comparison of epididymal fat; (D) comparison of inguinal fat; (E) Biochemical analyses including TG, TC, HDL-C and LDL-C; (F) levels of serum CK-MB and LDH in various experimental groups. Data are presented as mean ± SD (n = 6) (*p < 0.05, **p < 0.01, ***p < 0.001).

Export to PPT

As shown in Fig. 2E, HFHC feeding induced a significant elevation of serum TG, TC, LDL-C levels with respect to the control group. AMB or ACGD extracts treatment significantly lowered these levels, compared with untreated model group. And the HDL-C level did not alter after oral administration of AMB or ACGD extracts. In addition, HFHC-fed mice exhibited high-level indicators of myocardial injury including LDH and CK-MB than those in control group (Fig. 2F). And RSV, AMB or ACGD extracts treatment could reduce the release of these myocardial injury factors in treated model mice.

Pathological results of heart tissue could visually show the course of atherosclerosis. In control group, the mice myocardium was uniformly stained and the cardiomyocytes were neatly arranged (Fig. 3A). Meanwhile, the morphology and structure of myocardial cells were complete. And there were fewer cell swelling or necrosis (Fig. 3A). For model group, the arrangement and distribution of myocardial cells was obviously loose and the gap between muscle fibers significantly became larger as a whole (Fig. 3B). In addition, some nuclei of myocardial cell were blurred or even disappeared. And obvious defects and blank areas could be observed (Fig. 3B). Furthermore, after intragastric administration of RSV (Fig. 3C), AMB extracts (Fig. 3D) or ACGD extracts (Fig. 3E), the morphology and arrangement of cardiac myocytes remarkably improved, implying their myocardial protection effects.

Close

Fig. 3

The HE staining results of mice heart tissues. (A) control group; (B) model group; (C) RSV group; (D) AMB extracts group; (E) ACGD extracts group. The scales were 2.5 mm and 100 μM (magnification, 25×).

Export to PPT

Masson staining results demonstrated that model mice myocardial fibrosis was obviously aggravated along with some nuclei blurred or even disappeared (Fig. 4B) comparing with those in control group (Fig. 4A). In treatment group, the development of myocardial fibrosis got relief after oral administration of RSV (Fig. 4C), AMB extracts (Fig. 4D) and ACGD extracts (Fig. 4E). Moreover, visual changes of artery atherosclerotic plaques in heart aortic valve among multiple groups could be observed based on the oil red O staining results (Fig. 5). Obviously, fewer atherosclerotic plaques were detected in control mice (Fig. 5A), whereas massive plaques were deposited on the intima of the artery in model mice (Fig. 5B). RSV treatment (Fig. 5C) exhibited stronger anti-atherosclerotic plaque formation effects than those of AMB group (Fig. 5D) and ACGD group (Fig. 5E), following with the average lesion areas of plaques shown in Fig. 5F.

Close

Fig. 4

The Masson staining results of heart tissues in mice. (A) control group; (B) model group; (C) RSV group; (D) AMB extracts group; (E) ACGD extracts group. The scales were 2.5 mm and 100 μM (magnification, 25×).

Export to PPT

Close

Fig. 5

Photomicrographs of plaques at aortic valves stained with Oil red O. (A) control group; (B) model group; (C) RSV group; (D) AMB extracts group; (E) ACGD extracts group; (F) lesion area percentage of the total aorta. The scales were 2.5 mm and 200 μM (magnification, 12.5×). Data were shown as mean ± SD (n = 6). (*p < 0.05, ***p < 0.001).

Export to PPT

3.3 Chemical profiling of primary and secondary metabolites in AMB and ACGD samples by UHPLC-MS/MS

The AMB and ACGD samples were both acquired the original data under multiple reaction monitoring (MRM) modes in positive and negative ion modes. And the MRM overlay chromatograms of AMB samples were shown in Fig. S1A and B as well as those of ACGD samples in positive ion mode (Fig. S1C) and negative ion modes (Fig. S1D). Further, the obtained data were browsed and processed under Analyst platform (AB SCIX). Identification and tentatively characterization of primary and secondary metabolites in these two vegetables were performed based on the retention time behaviors and mass spectra properties comparing with the reference standards in Metware Data Library.

A total of 693 metabolites including 42 carbohydrates and derivatives, 101 lipids, 115 amino acids and derivatives, 62 nucleotides and derivatives, 87 organic acids, 16 vitamins and derivatives, 119 phenolic acids, 48 alkaloids, 54 steroidal saponins, 11 flavonoids, 5 quinones, 15 lignans and coumarins, and 18 other compounds were profiled in AMB and ACGD bulbs by UHPLC-MS/MS (Table S1). The match levels 1 and 2 meant that the match scores were over 0.7 and 0.5–0.7 after comparing the MS/MS spectra and retention times of detected metabolites with those of reference standards in Metware Database, respectively. Match level 3 meant that the characterized metabolites had the same or similar parameters of Q1, Q3, de-clustering potential (DP), and collision energy (CE) of reference standards in Metware Database. Obviously, massive primary metabolites were firstly detected and characterized in AMB and ACGD (Table S1). And they were also the important compositions for growth, development, nutrition accumulation, defense against pests, biosynthesis of secondary metabolites, and so on. In addition, steroidal saponins were both the main characteristic and active components for AMB and ACGD as described previously (Qin et al., 2016a, 2016b, 2018).

3.4 Elucidation of secondary metabolites firstly detected from AMB and ACGD samples by UHPLC-TOF/MS

Except steroidal saponins, another 32 secondary metabolites mainly including 13 flavonoids, and 10 lignans and coumarins were detected and reported in AMB and ACGD for the first time (Table 1). Herein, we would elucidate their structure based on the retention times and fragmentation behaviors comparing with those of reference standards in Metware Database.

For peak 1, it gave the [M + H]⁺ ion at m/z 595.1642 (−3.5 ppm) by TOF/MS and the [M + Na]⁺ ion at m/z 617.1482 (0 ppm) by MALDI/TOF scan. Its TOF-MS/MS scan showed the fragment ions at 577.1556, 559.1406, 457.1132, 427.1009 and 409.0916 were same as described previously (Fig. S2) (Picariello et al., 2017). Therefore, Peak 1 was identified as apigenin-6,8-di-C-glucoside. Similarly, peak 6 was regarded as vitexin because they had similar [M + H]⁺ ion at m/z 433.1108 (−6.2 ppm) and characteristic ions at m/z 415.1031, 397.0920 and 313.0710 with previous report (Fig. S2) (Avula et al., 2021). In addition, peaks 2 and 4 were considered as vitexin-2″-O-glucoside and apigenin-6-C-(2″-glucosyl)-arabinoside due to the retention times and fragmentation patterns, respectively (Table 1 & Fig. S2) (Li et al., 2022a, Picariello et al., 2017). For peak 9, the MALDI-TOF/MS scan gave the [M + K]⁺ ion at m/z 404.9691 (2.0 ppm) with a formula of C₁₅H₁₀O₉S. Kaempferol 3-O-sulfate exhibited the match score of 0.844 in the Metware Database (Fig. S2) (Meriçli, 1988). Similarly, peaks 10 and 11 were regarded as quercetin-3-O-(6″-acetyl)-glucosyl-(1 → 3)-galactoside and kaempferol-3-O-(6″″-malonyl)-sophorotrioside because they showed [M + H]⁺ and [M + H-H₂O]⁺ ions at m/z 669.1662 (−0.7 ppm) and 841.2012 (−3.2 ppm) with match scores of 0.997 and 0.885 in Metware Database (Table 1 & Fig. S2).

For peak 23, it displayed [M − H]⁻ ion at m/z 417.1538 (−2.6 ppm) by Q-TOF/MS and [M + K]⁺ ion at m/z 457.1250 (−3.3 ppm) by MALDI-TOF/MS (Table 1). The TOF-MS/MS scan gave the product ions at m/z 402.1271, 387.1046, and 181.0486 which kept in line with previous study (He et al., 2018a). Therefore, peak 23 was characterized as syringaresinol (Fig. S2). In the same way, peaks 21 and 22 were considered as syringaresinol-4′-O-glucopyranoside and syringaresinol-4′-O-(6″-acetyl)-glucoside based on the retention behaviors and diagnostic fragment ions (Table 1 & Fig. S2) (He et al., 2018a; Wang et al., 2017). Peak 18 showed [M + H]⁺ ion at m/z 193.0505 (2.1 ppm) and product ions at m/z 161.0769 and 105.0735 by MALDI-TOF/MS, while peak 26 gave the [M + H]⁺ quasi-molecular and fragment ions at 255.0657 (0 ppm), and 237.0641, 209.0702, 181.0756 and 154.0669 (Table 1 & Fig. S2). After matching with Metware Database, peaks 18 and 26 were tentatively characterized as 6,7-dihydroxy-4-methylcoumarin and 7,8-dihydroxy-4-phenylcoumarin, respectively (Kancheva et al., 2010).

Peaks 9–13 were mainly detected by MALDI-TOF/MS and not observed by TOF-MS (Table 1). Therefore, the TOF-MS/MS data of these five compounds were absent and they were deciphered based on MALDI-TOF/MS data and Metware Database. For other compounds, abundant TOF-MS/MS data and spectra were shown in Table 1 & Fig. S2, which benefited their structural elucidation.

3.5 Identification of differential primary and secondary metabolites between AMB and ACGD groups

First, the overlay total ion chromatograms (TICs) of QC samples in positive and negative ion modes (Fig. S3A and B) showed the high stability of the testing instrument. And the coefficients of variation in empirical cumulative distribution function also demonstrated the LC-MS data have a good stability (Fig. S3C). In addition, there were remarkable differences between AMB and ACGD samples because they could be significantly separated in unsupervised PCA plot with the first principal component (PC1) and second principal component (PC2) of 76.66% and 7.94% for the total variances (Fig. 6A). The fact that PC1 values of all samples were with ±2 × standard deviation also indicated that the instrument status was stable (Fig. S4A). And it was more reliable for the identification of differential metabolites if the correlation coefficients of intra-group samples relative to inter-group samples were higher (Fig. S4B).

Close

Fig. 6

Multivariate data analysis of AMB and ACGD samples. (A) the principal component analysis (PCA) between AMB and ACGD samples by UHPLC-MS/MS; (B) Volcano plots of differentially metabolites between AMB and ACGD samples; (C) the heat-map of multiple differentially metabolites between these two herbs. The green dots represent down-accumulated compounds, and the red dots represent up-accumulated metabolites between these two vegetables.

Export to PPT

Furthermore, orthogonal partial least squares discriminant analysis (OPLS-DA) model with T score of 76.5% and orthogonal T score of 8.6% was developed and applied to identify the most important contributors to grouping through deleting unrelated variances (Fig. S5A). The permutation validation results of R²X (0.851), Q² (0.994, p < 0.005) and R² (1, p < 0.005) showed the OPLS-DA model had a high separating capacity, and the permutations (n = 200) for this model did not overfit (Fig. S5B). Then, OPLS-DA S-plot was introduced to obtain the variable influence on projection (VIP) values (Fig. S5C). The data including the VIP values, p-values and fold-change values of tested metabolites was performed the volcano plot (Fig. 6B). A total of 365 metabolites (127 up-regulation and 238 down-regulation) were appraised as the differential variables between AMB and ACGD samples based on the three principles: VIP ≥ 1, p-value < 0.05, fold change ≥ 2 or ≤ 0.5 (Table S2).

As shown in Fig. 6C, a visual heat-map including 365 differential variables was emerged to better investigate the differences of primary and secondary metabolites between AMB and ACGD samples. These findings could understand the overall trends of each type of metabolites. To better present the contribution of differential metabolites, the radar chart and bar chart of Top40 fold changes differential metabolites were displayed in Fig. S6A and B, respectively. To better display the intra-group differences, histogram comparisons of Top40 fold change differential metabolites between AMB and ACGD samples were shown in Fig. S7. Obviously, primary metabolites (carbohydrates, amino acids, lipids, etc.) and secondary metabolites (steroidal saponins, flavonoids, alkaloids, etc.) fluctuated the most between AMB and ACGD groups. In addition, correlation net plot (Fig. S6C) also proved that these main secondary compounds were closely related to several primary metabolites. Therefore, it was necessary to interpret the biosynthesis pathways of primary and secondary metabolites in these two vegetables.

3.6 KEGG pathways of differential metabolites in AMB and ACGD samples

To better decipher the basic roles of these differentially regulated metabolites, we analyzed the KEGG metabolic pathways disturbed by these metabolites. As shown in Fig. S8A and B, the enrichment metabolic pathway analysis illustrated these differential metabolites were involved in many metabolic pathways including the biosynthesis of secondary metabolites, linoleic acid metabolism, purine metabolism, pyrimidine metabolism, tyrosine metabolism, and so on. And the differential abundance score (Fig. S8C) also showed the overall changes of the metabolites in linoleic acid metabolism, purine metabolism, tyrosine metabolism, α-linoleic acid metabolism and ascorbate and aldarate metabolism pathways were down-regulated in AMB than those in ACGD. On the contrary, the general trends of the primary metabolites derived from AMB for the vitamin B6 metabolism, glutathione metabolism, lysine metabolism were up-regulated than those in ACGD.

Actually, there were no significant differences between the levels of D-glucose, D-galactose, D-fructose, D-sucrose and raffinose in AMB and ACGD samples (Fig. 7A). In contrast, stachyose and manninotriose were both obviously up-regulated in ACGD along with the increased activity of β-fructofuranosidase. The activity of L-threonate 3-dehydrogenase could also be increased due to the remarkable elevation of 3-Dehydro-L-threonate. Similarly, the function of rhamnulose-1-phosphate aldolase may be improved in ACGD. In addition, the levels of D-glucuronate and D-glucurono-lactone decreased with the reduction of uronolactonase. The catalytic capacity of L-gulonolactone oxidase and glucuronolactone reductase may be partially inhibited because the levels of D-glucurono-lactone and D-ascorbate were both decreased.

Close

Fig. 7

The main KEGG enrichment pathways of differential metabolites between AMB and ACGD samples. (A) carbohydrate and aldarate metabolism; (B) purine metabolism; (C) linoleic acid metabolism; (D) α-linolenic acid metabolism; (E) vitamin B6 metabolism; (F) phenylalanine & tyrosine metabolism; (G) lysine metabolism. Red and green arrows represent up-accumulated and down accumulated metabolites, respectively.

Export to PPT

Purine metabolism was another important differential metabolic pathway between AMB and ACGD groups. As shown in Fig. 7B, most metabolites in purine metabolism pathways were down-regulated in AMB, such as guanosine, deoxyguanosine, 3,5′-cyclic-GMP, xanthosine, adenosine, adenine, deoxyadenosine, deoxyinosine, and so on. This trend kept in line with the total levels of nucleotides and their derivatives (Fig. 6C). These may be attributed to the decreased activities of 5′-nucleotidase, purine-nucleoside phosphorylase and AMP deaminase. On the contrary, several main precursors for purine synthesis including succinate semialdehyde, L-glutamate, and L-glutamine were markedly higher in AMB samples. And the more efficient catalytic activities of adenylate kinase, nucleoside-triphosphate-adenylate kinase, and AMP deaminase may result in higher levels of IMP, AMP and ADP.

Almost the metabolites in linoleic acid metabolism and α-linoleic acid metabolism were all down-regulated in AMB than those in ACGD (Fig. 7C and D). These findings also indicated that the most key enzymes including cytochrome P450 families, lipoxygenase, linoleate 8R-lipoxygenase, 12(13)-EpOME hydrolase, 13-HODE oxidoreductase, and arachidonate 15-lipoxygenase for linoleic acid metabolism were down-regulated in AMB. Similarly, the reduced functions of hydroperoxide lyase, allene oxide cyclase, and chloroplastic oxoene reductase may bring the decreased levels of primary metabolites for α-linoleic acid metabolism. In total, these phenomena were consistent with the overall trends of lipids expression in AMB and ACGD samples as shown in Fig. 6C.

For the metabolism of phenylalanine and tyrosine, the levels of most metabolites were lower in AMB than those in ACGD (Fig. 7F). For example, the metabolites of phenylalanine containing phenylethyl alcohol, 2-hydroxy-3-phenylpropanoate, trans-cinnamate, cinnamic acid, salicylate, and the main metabolites containing p-coumaryl alcohol, caffeic acid, gentisate aldehyde and sinapyl alcohol were all down-regulated in AMB. This may be contributed to the reduced activities of phenylalanine ammonia-lyase, tyrosine ammonia-lyase, benzoic acid 2-hydroxylase, and aryl-alcohol dehydrogenase in phenylalanine metabolism, and p-coumarate 3-hydroxylase, dopamine beta-monooxygenase, aldehyde dehydrogenase and monoamine oxidase in tyrosine metabolism. For the lysine metabolism (Fig. 7G), the possible elevated activity of lysine N-acetyltransferase resulted in increase of N-Acetyl-lysine. In addition, the reduction of 5-acetamido-pentanoate, L-saccharopine, and 2-oxoadipate were attributed to the lower catalysis function of 5-acetamidopentanoate oxidoreductase, saccharopine dehydrogenase and 2-aminoadipate aminotransferase, respectively. In addition, the inhibited function of pyridoxine 4-dehydrogenase and pyridoxamine 5′-phosphate oxidase may bring a drop of pyridoxine (vitamin B6), while aldehyde oxidase and pyridoxal oxidase may result in a rise of 4-pyridoxate for the metabolism of vitamin B6 (Fig. 7E).

3.7 Visualization and spatial distribution of biosynthesis pathways of several main secondary metabolites in AMB and ACGD

As shown in Fig. 1B, the fresh bulbs of AMB and ACGD both consisted of tunics, leaf scales and developing flower buds. Each leaf scale orderly and alternately arranged between tunics and the developing flower buds. Over 90% of the whole transverse section were occupied by the leaf scales both in AMB and ACGD. In addition, the biosynthesis of secondary metabolites (especially steroidal saponins and flavonoids) was one of the most important differential metabolic pathways in KEGG analysis (Fig. S8A). Herein, we would decipher the spatial distribution of several main secondary metabolites in AMB and ACGD combination with MALDI MSI visual images for the first time.

AMB and ACGD were both particularly rich in steroidal saponins. The basic building block of steroidal saponins were sapogenins which could be further glycosylated with different hexose and/or pentose groups. And their sapogenins and sugar substituents in AMB and ACGD samples were significantly different. In addition, the proposed biosynthetic pathway and visual ion images (including formula, adduct ions, errors, etc.) of saponins in fresh bulb sections of AMB and ACGD were shown in Fig. 8. The errors of these corresponding adduct ions were almost <10.0 ppm. Obviously, two smaller sections which reflect the typical structural patterns of the produced intermediates involved in these conversions were displayed in the biosynthetic pathway of steroidal saponins. A sequence started the reactions from acetyl-CoA to cholesterol by series of active enzymes (Fig. 8A), and further formed different types of sapogenins (Fig. 8B). The sapogenins were primarily to bind different glycosyl substituents to produce the representative steroidal saponins in AMB and ACGD bulbs. Hecogenin and laxogenin (C₂₇H₄₂O₄, [M + H]⁺, m/z 431.3136, −5.8 ppm) were the characteristic sapogenins for AMB and ACGD, respectively (Fig. 8B). However, they had the same formula, and carbonyl substitution at different positions of sapogenins, the MALDI MS spectra gave their visual MSI images. A similar isomer situation also occurred for other steroidal saponins. In total, the MSI ion images demonstrated that most quasi-molecular ions were distributed in the tunic and outside leaf scale regions of AMB, while these same or similar ions were observed in tunic and the whole leaf scales of ACGD. These ions in AMB and ACGD were all rarely distributed in developing flower buds.

Close

Fig. 8

MALDI MSI-oriented visual distribution for steroidal saponins. (A) the spatial distribution of primary metabolites for the biosynthesis of steroidal saponins in AMB and ACGD samples; (B) the visual distribution of several sapogenins for steroidal saponins in these two herbs. The scale was 1 mm.

Export to PPT

Besides, although several assays had been carried out to explore the biosynthetic pathway and localize synthetic enzymes of flavonoids, the information about the holistic spatial distribution of these products and intermediates in AMB and ACGD was limited. In the present study, we illustrated that MALDI MSI images could be used to visualize the accumulation sites of products and major intermediates in the biosynthetic pathway of flavonoids (Fig. 9). A total of 11 metabolites involved in the flavonoids biosynthetic pathway were visualized in AMB and ACGD bulbs, including apigenin-6,8-di-C-glucoside, vitexin-2″-O-glucoside, chrysoeriol-6,8-di-C-glucoside, and so on. This was the first time for the detection and report of these flavonoids from AMB and ACGD. As shown in Fig. 9, the distribution features of target metabolites in these two vegetables shared similarities and differences. For example, luteolin 7-O-[β-D-glucuronosyl-(1 → 2)-β-D-glucuronide]-4′-O-β-D-glucuronide ([M + H]⁺, m/z 815.1573, 6.8 ppm), vitexin 2″-glucoside ([M + Na]⁺, m/z 617.1482, 0 ppm) and quercetin-3-O-(6″-acetyl)-glucosyl-(1 → 3)-galactoside ([M + H]⁺, m/z 669.1662, −0.7 ppm) were all widely distributed in tunics and developing flower buds of AMB and ACGD bulbs. And several metabolites like kaempferol 3-O-sulfate ([M + K]⁺, m/z 404.9691, 2.0 ppm), naringenin-7-O-rutinoside-4′-O-glucoside ([M + Na]⁺, m/z 765.2182, −4.7 ppm) and chrysoeriol-6,8-C-glucoside ([M + H]⁺, m/z 625.1769, 0 ppm) were localized mainly tunics of AMB, and tunics and the whole leaf scales of ACGD bulbs.

Close

Fig. 9

MALDI MSI images of several primary and secondary metabolites for the biosynthesis of flavonoids in AMB and ACGD fresh bulbs. The scale was 1 mm.

Export to PPT

Except steroidal saponins and flavonoids, lignans and coumarins were another two important structural types of compounds in AMB and ACGD. And they had similar biosynthetic pathways with sinapic acid ([M + Na]⁺, m/z 247.0558, −9.7 ppm) and sinapyl alcohol ([M + K]⁺, m/z 249.0526, −1.2 ppm) as important intermediates mainly distributed in tunics and leaf scales of AMB and ACGD bulbs (Fig. S9). Then, a series of lignans including syringaresinol ([M + K]⁺, m/z 457.1250, −3.3 ppm), pinoresinol/matairesinol ([M + H]⁺, m/z 359.1487, −2.2 ppm), and their glycosides were represented in MALDI MSI images (Fig. S9A). They were mainly distributed in tunics and outside leaf scale of AMB, and located in tunics, the holistic leaf scale and rarely in developing flower buds of ACGD. The distribution characteristics of coumarins in AMB and ACGD was basically consistent with the spatial distribution of lignans. However, phellopterin ([M + K]⁺, m/z 339.0649, 4.1 ppm) was widely observed in the whole AMB and ACGD bulbs (Fig. S9B). And the spatial distribution of phellopterin displayed a clear descending trend from the developing flower buds to tunics and leaf scales in these two vegetables.

Moreover, the proposed biosynthetic pathway of rosmarinic acid in AMB and ACGD bulbs was visually observed in Fig. S10A. Tunics and outside leaf scales were the richest regions of rosmarinic acid ([M + Na]⁺, m/z 383.0771, 7.3 ppm) in these two bulbs, and it was also partially presented in developing flower buds. Meanwhile, the intermediates including salvianic acid A ([M + K]⁺, m/z 237.0147, −7.6 ppm) and isorinic acid ([M + H]⁺, m/z 345.1009, 9.6 ppm) were also mainly distributed in tunics and leaf scales. In addition, the intensity of alkaloids in AMB and ACGD were relatively poor. Therefore, ebeinone ([M + Na]⁺, m/z 434.2999, −8.3 ppm) and puqiedinone ([M + K]⁺, m/z 452.2952, 4.6 ppm) were sporadically spaced in the whole bulbs (Fig. S10B). Besides, several glycosylation products of phenolic acids were detected in MALDI MSI images (Fig. S10C). For instance, salicylic acid-2-O-glucoside ([M + K]⁺, m/z 339.0481, −0.3 ppm) was mainly observed in tunics and developing flower buds of AMB and ACGD bulbs. For 4-O-β-D-glucosyl-4-coumaric acid ([M + Na]⁺, m/z 349.0925, 7.4 ppm), the inside leaf scales were the most abundant areas of ACGD bulbs. The other metabolites including arbutin ([M + Na]⁺, m/z 295.0801, 2.4 ppm), 1-O-feruloyl-D-glucose ([M + H]⁺, m/z 357.1164, −6.2 ppm), sibiricose A5 ([M + Na]⁺, m/z 541.1533, 0 ppm) were mainly distributed in tunics and leaf scales of these two crops.