Metabolomics Analysis Revealed the Characteristic Metabolites of Hemp Seeds Varieties and Metabolites Responsible for Antioxidant Properties

UPLC-ESI-Q TRAP-MS/MS

The sample extracts were analyzed using an UPLC-ESI-MS/MS system (UPLC, SHIMADZU Nexera X2, Shanghai, China; MS, Applied Biosystems 4,500 Q TRAP, Thermo Fisher, Shanghai, China) with an Agilent SB-C18 column (1.8 μm, 2.1 mm × 100 mm). The mobile phase consisted of solvent A (pure water with 0.1% formic acid) and solvent B (acetonitrile with 0.1% formic acid). Sample measurements were performed with the following gradient program: starting conditions of 95% A and 5% B; a linear gradient to 5% A and 95% B was adjusted within 9 min and kept for 1 min; and a composition of 95% A and 5.0% B was adjusted within 1.1 min and kept for 2.9 min. The flow velocity was set as 0.35 ml per min, and the column oven was set to 40°C. The injection volume was 4 μl. The effluent was alternatively connected to an ESI-triple quadrupole-linear ion trap (QTRAP)-MS.

ESI-Q TRAP-MS/MS

LIT and triple quadrupole (QQQ) scans were acquired on a triple quadrupole-linear ion trap mass spectrometer (Q TRAP; AB4500 Q TRAP UPLC/MS/MS System, Milwaukee, United States) equipped with an ESI Turbo Ion-Spray interface, operating in positive and negative ion mode, and controlled by Analyst 1.6.3 software (AB Sciex). The ESI source operation parameters were as follows: ion source, turbo spray; source temperature, 550°C; ion spray voltage (IS), 5,500 V (positive ion mode)/−4,500 V (negative ion mode); ion source, gas I (GSI), gas II (GSII), and curtain gas (CUR) set at 50, 60, and 25.0 psi, respectively; and collision-activated dissociation (CAD), high. Instrument tuning and mass calibration were performed with 10 and 100 μmol/L polypropylene glycol solutions in QQQ and LIT modes, respectively. QQQ scans were acquired as multiple reaction monitoring (MRM) experiments with collision gas (nitrogen) set to medium. Declustering potential (DP) and collision energy (CE) for individual MRM transitions were performed with further DP and CE optimization. A specific set of MRM transitions were monitored for each period according to the metabolites eluted within this period.

Qualitative and Quantitative Analysis of Metabolites

Based on the self-built database MWDB V2.0 (Metware Biotechnology Co., Ltd. Wuhan, China), primary and secondary mass spectrometry data were subjected to qualitative analysis. Isotopic signals, repeating signals containing K⁺ ions, Na⁺ ions, NH4⁺ ions, and repeating signals of fragment ions that are themselves other larger molecular weight species were removed from the analysis.

Metabolite quantification was performed using triple-quadrupole mass spectrometry in multiple reaction monitoring (MRM). In MRM mode, the quadrupole first screens the precursor ions (parent ions) of the target substance, and excludes the ions corresponding to other molecular weight substances to preliminarily eliminate interference; precursor ions are induced to ionize by the collision cell and then fragmented to form many fragment ions. The fragment ions are then filtered through triple quadrupole to select a desired characteristic fragment ion, eliminating the interference of non-target ions, making the quantification more accurate and repeatable. After obtaining the metabolite spectrum analysis data of different samples, perform peak area integration on all the mass spectrum peaks of the substances, and perform integration correction on the mass spectral peaks of the same metabolite in different samples (Fraga et al., 2010). In order to compare the differences in the content of each metabolite in different samples among all the detected metabolites, we corrected the mass spectral peaks detected in different samples for each metabolite according to the information of metabolite retention time and peak shape, in order to ensure the accuracy of qualitative and quantitative. The relative contents of metabolite in each sample were represented with chromatographic peak area integrals.

Multivariate Data Analysis and Statistical Analysis

Data were normalized with sum, log transformation, and auto-scaling before any statistical analysis. Unsupervised principal component analysis (PCA) was performed by the “prcomp” statistics function within R.¹ The data were unit variance scaled before unsupervised PCA. The hierarchical cluster analysis (HCA) results of samples and metabolites are presented as heatmaps with dendrograms, while Pearson correlation coefficients (PCC) between samples were calculated by the “cor” function in R and are presented as heatmaps. Both HCA and PCC were performed using the “pheatmap” package in R. For HCA, the normalized signal intensities of metabolites (unit variance scaling) are presented as a color spectrum. Significantly different accumulated metabolites (DAMs) between groups were determined by VIP (Variable importance in the projection) ≥ 1 and absolute log₂ (fold change) ≥ 1. VIP values were extracted from the OPLS-DA (orthogonal partial least squares discriminant analysis) results, which also contained score plots and permutation plots, and they were generated using the “MetaboAnalystR” package in R. The data were log transformed (log2) and mean centered before OPLS-DA. To avoid overfitting, a permutation test (200 permutations) was performed. Student’s t-test was carried out using IBM SPSS Statistics 25 Software to identify the significant differences at α level of 0.05 for all biochemical analyses.

WGCNA and Metabolite Network Visualization

The metabolites that detected in at least two replicates of any sample were used for WGCNA (version: WGCNA_1.70-3). The co-expression modules were obtained by using one-step network construction function with following parameters: soft threshold power: 20, TOMtype: unsigned, mergeCutHeight: 0.25 and minModuleSize: 30, other parameters: default. The initial clusters were merged on eigen-metabolites. Eigen-metabolites were calculated for each module, which were used to search the association with metabolite characteristics of varieties (Zhang and Horvath, 2005). The analysis results of WGCNA were exported to the cytoscape software (v3.7.0, United States) to generate metabolite interaction networks. Edges with weight less than 0.2 are removed due to excessive number of edges. The importance of metabolites was ranked based on the degree of metabolites to select the hub metabolites of modules.

Kyoto Encyclopedia of Genes and Genomes (KEGG) Annotation and Enrichment Analysis

Identified metabolites were annotated using the KEGG compound database,² and annotated metabolites were then mapped to the KEGG Pathway database.³ Pathways with significantly regulated metabolites were then subjected to metabolite set enrichment analysis (MSEA), and their significance was determined by hypergeometric test p-values.

Determination of Total Antioxidant Capacity and Antioxidant Analysis

Quantification of antioxidant activities was performed on a dry weight basis. Liquid nitrogen was added to the seven samples, which were then ground to pass a 60 mesh-sieve with an IKA all basic mill (IKA Works Inc., Wilmington, NC, USA). The powder was then freeze-dried to constant weight. A portion of the powder (0.2 g) was extracted in a capped centrifuge tube with 4 ml of solvents including 80% ethanol and 20% distilled water. The mixture was ultrasonically extracted at 50°C for 20 min at 38 kHz and 30 kW. The extracts were then centrifuged by an Allegra 21R Centrifuge (Beckman Coulter Ltd., Palo Alto, CA, USA) at 4500 rpm for 10 min, and the supernatants were removed into new tubes. Residues were extracted with the same solvents for the second time. Both extracts were combined and stored at 4°C in the dark until further analysis within 2 d. Extractions were performed in three replicates for each sample.

1,1-Diphenyl-2-Picrylhydrazyl Radical-Scavenging Activity

The 1,1-diphenyl-2-picrylhydrazyl (DPPH) radical-scavenging capacity of hemp seed extracts was evaluated according to the method of (Xie et al., 1993) with slight modifications. Briefly, 1 ml of the tested hemp seed extract was added to 3 ml of methanol solution containing DPPH radicals (final concentration of 0.1 mM). The mixture was shaken vigorously for 1 min by vortexing and left to stand at room temperature in the dark for 30 min. Thereafter, the absorbance of the sample (A_sample) was measured using an UV 160 spectrophotometer at 517 nm against an ethanol blank. The absorbance of a complete blank control (A_c-blank) was measured after adding 3 ml of methanol to 1 ml of ethanol. The absorbance of a blank control (A_blank) was measured after adding 3 ml of methanol to 1 ml of the sample solution. The absorbance of a negative control (A_control) was measured after adding DPPH solution to 1 ml of the extraction solvent. The percent of DPPH discoloration of the sample was calculated according to the following equation:

Percent discoloration={1−[(Asample−Ablank)/(Acontrol−Ac−blank}×100 .

Identification of Metabolites Responsible for Differences Among Seven Varieties

The difference of phenotypes among the hemp seed varieties is mainly in the contents of some metabolites. We use the supervised method, OPLS-DA and Student’s t (value of p<0.05) test, to find the metabolites responsible for differences among these seven varieties. In this study, the OPLS-DA model compared metabolite contents of the varieties in pairs to evaluate the differences. The Q² values of all comparison groups exceeded 0.8 except A vs. E, demonstrating that the models were stable (Li et al., 2021). The DAMs (|Log (FC)| > 1, VIP ≥ 1 and value of p < 0.05) between varieties were screened (Supplementary Table 3). Most of the metabolites were not highly accumulated in C variety, therefore, C variety was used to compared with other varieties. The number of DAMs in different comparison groups (C vs. A, B, D, E, F, and G) was shown in Figure 1D. The upset diagram showed that only two metabolites were shared among the comparison groups (C vs. A, B, D, E, F, and G). These results showed that the metabolites that caused the differences between C and A, B, D, E, F, and G were greatly different.

Identification of Modules of Closely Related Metabolites

WGCNA is a powerful tool that can identify metabolite sets with highly synergistic changes. Compared with only focusing on differential metabolites, WGCNA uses the information of thousands or nearly tens of thousands of metabolites with the greatest changes or all metabolites to identify the metabolite set of interest and conducts significant association analysis with the phenotype. Thus, we performed WGCNA and constructed a network to further explore the relation between metabolites and hemp seed varieties. The metabolites have been clustered into six modules (comprised of 72–160 metabolites) and those who belong to none of these modules are indicated by grey (Figures 2A,​,B).B). Figure 2C shows the composition of the metabolites in the six modules. The details of the metabolites in each module were shown in Supplementary Table 2. The eigen-metabolite, which represented the accumulation profile of the metabolites contained in the module were calculated for each module. Further, we associated each of the co-expression modules with varieties of hemp seed via Pearson correlation coefficient analysis (Figure 2B).

Identification of Metabolite Characteristics in different Varieties

Based on the data of WGCNA, we further identified the hub metabolites of the modules and analyzed the correlation between modules and hemp seed varieties.

The brown module, which contained 114 metabolites, has a high correlation (r > 0.8, value of p < 0.05) with the D variety (Figure 2B). It indicated that there was a significant association between eigen-metabolite of brown and D variety trait. “Flavonoid biosynthesis,” “flavone and flavonol biosynthesis,” and “isoflavonoid biosynthesis” pathways related metabolites were highly accumulated in D variety (Figure 3A). In the D variety, 55 flavonoids, such as chrysoeriol, kaempferol, and luteolin, as well as their derivatives were detected at a high level. Two alkaloids (p-Coumaroylferuloylputrescine and N-p-Coumaroyl-N′-feruloylputrescine) and derivatives of tricin and diosmetin were hub metabolites in the brown module (Figure 3B). Betanin (mws1138), a natural pigment with powerful antioxidant activity, was also a hub metabolite detected at high levels in the D variety. Cannflavins, which are unique flavonoids in hemp, were also accumulated in the D variety (Figure 3C). The yellow module was highly correlated with the B variety (r > 0.8, value of p<0.05) and contained 99 metabolites including 47 flavonoids and 13 alkaloids (Figure 2B). Metabolites that involved in “anthocyanin biosynthesis” pathway were highly accumulated in B variety (Figure 3D). Lots of derivatives of quercetin and isorhamnetin are highly accumulated in the B variety and play as hub metabolites in the yellow module (Figure 3E). Meanwhile, several anthocyanins like Cyanidin-3-O-glucoside (Kuromanin) and Cyanidin-3-O-rutinoside (Keracyanin) are clustered into the yellow module (Figure 3F), which might be the reason for the dark color of the seed coat. The diet contains plant flavonoids is thought to have neuroprotective, antioxidant, and anticancer properties (Hostetler et al., 2017; Madunic et al., 2018). B and D varieties have a high level of flavonoids, indicating that both varieties may have high medicinal value.

The red module contained 72 metabolites and was correlated with the G variety (Figure 2B). Twelve free fatty acids, 15 organic acids, and 17 phenolic acids were clustered in the red module. And, “alpha-linolenic acid metabolism” pathways (value of p < 0.05) related metabolites are highly accumulated in G variety (Figure 4A). Several free fatty acids served as hub metabolites in the red module, including methyl-linolenate, undecanedioic acid, and 15(R)-hydroxylinoleic acid (Figures 4B,​,C).C). The green module correlated with the E variety, and it was accumulated in the following components: amino acids and derivatives; nucleotides and derivatives; organic acids; and saccharides and alcohols (others class; Figure 2B). “Starch and sucrose metabolism” and “Galactose metabolism” related metabolites are overrepresented in the green module (Figure 4D). Only few secondary metabolites were clustered in the green module. The hub metabolites in the green module were mainly composed of saccharides and alcohols, such as D-Mannose (pmf0138), D-Fructose (mws1164), D-Galactose (pmf0139), D-Glucose (mws4170; Figures 4E,​,F).F). E and G varieties are rich in saccharides and fatty acids, suggesting they are ideal varieties for nutrition.

The blue and turquoise modules were overrepresented in different classes of lipids. The lipids clustered in the blue module were lysophosphatidylcholines (LPCs), while the turquoise module contained free fatty acids and glycerol ester. Both modules also contained many amino acids and derivatives (Figure 2C). Although these modules were not significantly correlated with the G and F varieties, these data indicated that these two varieties may be the ideal varieties for oil extraction or consumption.