Characterization and Discrimination of Marigold Oleoresin from Different Origins Based on UPLC-QTOF-MS Combined Molecular Networking and Multivariate Statistical Analysis

Marigold oleoresin is an oil-soluble natural colorant mainly extracted from marigold flowers. Xinjiang of China, India, and Zambia of Africa are the three main production areas of marigold flowers. Therefore, this study utilized ultra-performance liquid chromatography-quadrupole time-of-flight mass spectrometry (UPLC-QTOF-MS/MS) technology, combined with Global Natural Products Social Molecular Networking (GNPS) and multivariate statistical analysis, for the qualitative and discriminant analysis of marigold oleoresin obtained from three different regions. Firstly, 83 compounds were identified in these marigold oleoresin samples. Furthermore, the results of a principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) indicated significant differences in the chemical compositions of the marigold oleoresin samples from different regions. Finally, 12, 23, and 38 differential metabolites were, respectively, identified by comparing the marigold oleoresin from Africa with Xinjiang, Africa with India, and Xinjiang with India. In summary, these results can be used to distinguish marigold oleoresin samples from different regions, laying a solid foundation for further quality control and providing a theoretical basis for assessing its safety and nutritional aspects.


Introduction
Marigold oleoresin, also known as lutein oleoresin, is an oil-soluble natural colorant and antioxidant extracted from the flowers of Tagetes erecta L. It primarily contains lutein and lutein esters as its main active ingredients, which, together, constitute 70-79% of the total carotenoid content in marigold oleoresin [1].Incorporating marigold oleoresin into food and nutritional supplements can effectively enhance the products' antioxidant properties, protecting consumers against chronic diseases triggered by oxidative stress, such as cardiovascular disease, cancer, and neurodegenerative disorders.Additionally, lutein is particularly beneficial for eye health, helping to prevent macular degeneration and cataracts, thereby maintaining visual well-being [2].In skincare and cosmetic products, adding marigold oleoresin helps to improve their antioxidant, moisturizing, and reparative functions, meeting consumer demands for natural, safe, and efficacious skincare solutions.In animal nutrition, the inclusion of marigold oleoresin not only increases the commercial value of poultry products, but also improves animal health and enhances farming efficiency [3].In confectionery, beverages, baked goods, and seasonings, incorporating marigold oleoresin imparts attractive colors and supplementary nutrition.Furthermore, studies suggest that certain components in marigold oleoresin may exhibit anti-inflammatory, antimicrobial, antiviral, and anticancer bioactivities, demonstrating potential applications in the pharmaceutical field.
Marigold oleoresin, prepared by the hexane extraction of Tagetes erecta L., is the frontend raw material for producing lutein crystal and lutein ester.At present, the global marigold oleoresin market volume is about 10,000 tons, 80% of which is used as feed and 20% for health foods and pharmaceutical materials.The main producing areas of Tagetes erecta L. are China, India, and Africa, among which, about 600,000 mu is from China and about 300,000 mu is from India, while Zambia in Africa is now attempting to plant marigold due to its abundant planting area and suitable climate.However, the majority of existing studies have focused on lutein crystals and lutein ester, while there have been few studies on marigold oleoresin.Based on the above reasons, research on marigold oleoresin becomes very important and urgent.Therefore, marigold oleoresins from Xinjiang, India, and Africa were selected as research subjects in this study, with the aim of determining the chemical compositions and composition differences among marigold oleoresins from different regions.
Ultra-performance liquid chromatography coupled with time-of-flight mass spectrometry (UPLC-QTOF-MS/MS) is a highly sensitive and high-resolution platform extensively utilized for the identification of chemical components in herbal substances [4][5][6].Global Natural Products Social Molecular Networking (GNPS) is a publicly accessible web platform (https://gnps.ucsd.eduaccessed on 17 May 2023) that enables the systematic comparison and classification of numerous molecules based on the similarity of their MS/MS spectra.It also provides visual representations of the relationships among these molecules.GNPS is currently widely employed in metabolomics, the identification of chemical components in herbal medicines, and the discovery of novel compounds [7][8][9].Furthermore, multivariate statistical analysis methods such as principal component analysis (PCA) and orthogonal partial least-squares-discriminant analysis (OPLS-DA) offer reliable approaches for assessing overall variation and identifying chemical markers for distinguishing the origin, different parts, and processing methods of herbs [10][11][12][13][14].In summary, UPLC-QTOF-MS/MS, GNPS, and multivariate statistical analysis techniques play crucial roles in the identification and characterization of chemical components in herbal substances, facilitating the discovery of new compounds and the differentiation of herbal samples based on their origins and processing methods.
In this study, a comprehensive targeted metabolomic approach combining UPLC-QTOF-MS/MS with molecular networking (MN) was employed to perform a comparative analysis of metabolites in liposoluble extracts of marigold (Tagetes erecta L.) sourced from three different regions.The aim was to identify the differential metabolites that play significant roles in differentiating marigold oleoresins based on their geographical origins.This research not only contributes to the scientific evaluation and optimal utilization of natural resources, but also enhances product quality and market competitiveness, fostering related scientific advancements, technological innovations, and sustainable agricultural development.In general, this research has profound scientific value and socio-economic significance.

Samples and Chemical Reagents
Raw materials of marigold granules were procured from the sales markets of their respective original production areas in March 2023: Zambia (FZ) in Africa, Xinjiang (XJ) in China, and India (YD).These materials were then utilized in subsequent oleoresin preparation experiments.For each origin, three batches were obtained and subjected to experimentation.

The Preparation of Marigold Oleoresin
After thorough mixing, 1.0 g of marigold granules was precisely weighed and transferred into a 100 mL volumetric flask.Next, n-hexane was added to the flask until it reached the calibration mark.Subsequently, the volumetric flask was immersed in a thermostatically controlled water bath maintained at 40 • C for extraction.At every 10 min interval, ultrasonication was performed for 5 min, and this process was repeated a total of four times.Following this, the flask was removed from the water bath and allowed to cool to room temperature.Then, once more, n-hexane was added up to the calibration mark.The resulting extract was then centrifuged at 12,000 rpm for 10 min at 4 • C. The supernatant was filtered through a 0.22 µm membrane into an autosampler vial and stored at 4 • C for subsequent mass spectrometry detection and analysis.This preparation procedure was identical for all batches of samples.

The Conditions of Chromatographic and Mass Spectrometry
A chromatographic analysis was performed using the ACQUITY UPLC I-class system, which is manufactured by Waters Corporation (Milford, MA, USA).This advanced analytical platform integrates a binary solvent management module, an automated sample injector, a vacuum degassing unit, and a temperature-controlled column chamber.For the chromatographic separations, an ACQUITY BEH C18 column (2.1 mm × 150 mm, 1.7 µm; Waters Corporation, Milford, MA, USA) was chosen.Employing a gradient elution strategy, the mobile phase was composed of two components: solvent A, containing 0.1% formic acid in deionized water, and solvent B, represented by a 1:1 volumetric blend of acetonitrile and isopropanol.The subsequent elution program was as follows: 0-2 min, 60%B; 2-50 min, 60-84%B; 50-120 min, 84-100%B; 120-121 min, 100%B; 121-122 min, 100-60%B; 122-124 min, 60%B.A sample injection volume of 3 µL was established, accompanied by a constant flow rate of 0.2 mL/min throughout the analysis.For mass spectrometric detection, a Waters Xevo G2-XS QTOF instrument, equipped with a Z Spray™ electrospray ionization (ESI) trap source, was employed.The desolvation gas flow rate was maintained at 800 L/h, while the cone gas flow rate was set to 50 L/h.The operating conditions for the ESI source included a source offset voltage of 80 V, a cone voltage of 40 V, and a source temperature of 100 • C. The capillary voltage was adjusted to 3.0 kV.Data acquisition proceeded under the Fast-DDA mode, with the dual dynamic collision energy ranging from 6 to 40 V for precursors up to 50 Da and from 40 to 120 V for those above 1500 Da.The MS scan rate was 0.04 s, whereas the MS/MS scan rate was 0.1 s.In the data-dependent acquisition, the top 3 most intense precursor ions underwent MS/MS fragmentation to generate the MN.Real-time calibration was achieved using leucine encephalin solution (m/z 556.2771 [M+H] + ) as an internal reference.MassLynx v4.1 software (Waters Corporation, Milford, MA, USA) facilitated data acquisition.

Molecular Networking
The raw data from UPLC-QTOF-MS/MS were converted into "mzXML" format using MS Convert Version 3.0.23052-0c85f26automated build (http://proteowizard.sourceforge.net,accessed on 17 May 2023) and subsequently uploaded to the GNPS online platform [15].The following parameters were configured for GNPS analysis: a tolerance of 0.02 Da was applied to both the precursor ion masses and the fragment ion masses.A cosine score exceeding 0.7 with a minimum of six matched peaks was considered for the analysis.A molecular network was constructed, and its visualization was accomplished utilizing the Cytoscape software version 3.9.1,accessible at http://www.cytoscape.org,accessed on 17 May 2023.

Multivariate Statistical Analysis
The synergistic application of UPLC-QTOF-MS/MS technology and multivariate statistical analysis offers a comprehensive approach to characterizing chemical constituents and efficiently identifying potential chemical markers [16,17].The raw mass spectrometry data were preprocessed using MSDIAL 4.92 (http://prime.psc.riken.jp/compms/msdial/main.html, accessed on 17 May 2023).Following preprocessing, the LC-MS data were subjected to multivariate statistical analysis within the SIMCA-P 14.1 software (MKS Umetrics AB, Sweden).To assess the relationships and differences among distinct experimental groups, a principal component analysis (PCA) and orthogonal partial least squares discriminant anal-ysis (OPLS-DA) were conducted.PCA, an unsupervised technique, transforms the initial variables into a set of orthogonal, non-correlated components, thereby condensing redundant information.In contrast, OPLS-DA, which employs a supervised modeling technique, aims to eliminate systematic noise and selectively extract informative variables.Moreover, S-plots were generated to identify chemical markers by examining the Variable Importance for the Projection (VIP) values.A volcano plot analysis was performed using the MetaboAnalyst 5.0 web platform (https://www.metaboanalyst.ca,accessed on 17 May 2023), while a heatmap analysis was conducted using Hiplot (https://hiplot.cn/basic/heatmap,accessed on 17 May 2023).

Identifying Chemical Constituents in Marigold Oleoresins of Diverse Origins
In this study, the comprehensive profiling of compounds in marigold oleoresin was achieved by integrating analytical data with a molecular networking chemical database and relevant literature resources.Consequently, a total of 83 compounds, including 6 Vitamin E and Vitamin E esters, 13 triterpenes, 7 lutein diesters, 4 lutein monoesters, 10 ceramide, 16 triglycerides, 13 diglycerides, 6 monogalactosyldiacylgylcerols, 3 steroids, and 2 others, were identified or tentatively characterized.The compounds were identified according to a well-established analytical methodology, which relied on comparing their retention times (Rt), fragment ions, and fragmentation pathways with corresponding data from the literature.A representative chromatogram of marigold oleoresin, displaying the base peak intensity in positive ion mode, can be seen in Figure 1.Additionally, Table 1 provides comprehensive details on the individual components, encompassing peak numbers, retention times (Rt), compound names, molecular formulas, classification, mass errors, and characteristic fragment ions.In this work, an integrative molecular network of marigold oleoresin samples from various regions was constructed based on the MS/MS spectral similarity of their constituents, as illustrated in Figure 2.This network comprised 2224 precursor ions, with 184 clusters (nodes consisting of at least 2 ions) and 991 single nodes.Further details regarding this molecular network can be accessed via the GNPS website (https://gnps-cytoscape.ucsd.edu/process?task=eafa6c21f8274b3390e634d651e1f8fd,accessed on 17 May 2023).In the MN, the relative content of a compound in the XJ, YD, and FZ samples is depicted by the proportions of the orange, purple, and green sectors, respectively, within each node.Molecular networking facilitates the analysis by structurally grouping related compounds into clusters, allowing for easier examination and comparison [18].Molecular networking reveals that triglycerides (I), diglycerides (II), lutein esters (III), ceramides (IV), triterpenes (V), monogalactosyldiacylglycerols (VI), and vitamin E esters (VII) each form distinct clusters, indicative of their structural similarities within these respective classes.The compounds were identified using a well-established analytical approach that combined retention time (Rt) data, the GNPS library, and fragmentation pathways referenced from the literature.

Identify Elucidation of Triglycerides and Diglycerides
Cluster I and Cluster II predominantly comprised nodes representing triglycerides and diglycerides, as evident in Figure 3.The primary observed adduct ions included [M+NH 4 ] + , [M+Na] + , and [M+H] + .In the MS/MS spectra, a characteristic ion peak was generated by the loss of a single fatty acid fragment from triglycerides, represented as [M-FA+H] + (Figure 2) [19,20].The industrial production of marigold oleoresin yielded the identification of 16 triglycerides and 13 diglycerides.

Identify Elucidation of Triglycerides and Diglycerides
Cluster I and Cluster II predominantly comprised nodes representing triglycerides and diglycerides, as evident in Figure 3.The primary observed adduct ions included [M+NH4] + , [M+Na] + , and [M+H] + .In the MS/MS spectra, a characteristic ion peak was generated by the loss of a single fatty acid fragment from triglycerides, represented as [M-FA+H] + (Figure 2) [19,20].The industrial production of marigold oleoresin yielded the identification of 16 triglycerides and 13 diglycerides.

Identify Elucidation of Lutein Esters
As displayed in Figure 4, Cluster III predominantly comprised nodes signifying lutein esters.The main components identified in marigold oleoresin were lutein esters, including lutein dilaurate, lutein laurate-myristate, lutein dimyristate, lutein

Identify Elucidation of other Compounds
The constituents within each cluster were tentatively classified as follows: Cluster IV represents putatively identified ceramides, Cluster V represents putatively identified triterpenes, Cluster VI represents putatively identified monogalactosyldiacylglycerols, and Cluster VII represents putatively identified vitamin E and vitamin E esters.The identification of these clusters was based on matching the fragment ions with those reported in the literature [26][27][28][29][30][31][32][33].

PCA Analysis and Cluster Analysis
PCA analysis was performed on metabolites from three different origins of marigold oleoresin, as shown in Figure 5A.PC1 accounted for 43.9% of the total variance, while PC2 accounted for 20.6%.From the plot, it can be observed that samples from each group clustered together, indicating a clear separation trend among those from different origins.The PCA results overall reflected the differences in metabolites among the samples from different origins.A cluster analysis is illustrated in Figure 5B, which distinctly demonstrates distinct grouping patterns among the different origins.The metabolite types and contents of FZ and XJ were similar, clustering them together, while YD formed a separate cluster with significant differences from FZ and XJ.The combined PCA and cluster analysis indicated that the marigold oleoresin from the three different origins exhibited distinct metabolic characteristics.

Identify Elucidation of Other Compounds
The constituents within each cluster were tentatively classified as follows: Cluster IV represents putatively identified ceramides, Cluster V represents putatively identified triterpenes, Cluster VI represents putatively identified monogalactosyldiacylglycerols, and Cluster VII represents putatively identified vitamin E and vitamin E esters.The identification of these clusters was based on matching the fragment ions with those reported in the literature [26][27][28][29][30][31][32][33].

Differential Metabolomic Analysis of Marigold Oleoresin from Different Origins PCA Analysis and Cluster Analysis
PCA analysis was performed on metabolites from three different origins of marigold oleoresin, as shown in Figure 5A.PC1 accounted for 43.9% of the total variance, while PC2 accounted for 20.6%.From the plot, it can be observed that samples from each group clustered together, indicating a clear separation trend among those from different origins.The PCA results overall reflected the differences in metabolites among the samples from different origins.A cluster analysis is illustrated in Figure 5B, which distinctly demonstrates distinct grouping patterns among the different origins.The metabolite types and contents of FZ and XJ were similar, clustering them together, while YD formed a separate cluster with significant differences from FZ and XJ.The combined PCA and cluster analysis indicated that the marigold oleoresin from the three different origins exhibited distinct metabolic characteristics.
In conducting a pairwise comparison between FZ and XJ utilizing the statistical method OPLS-DA, several parameters were evaluated to assess the model's performance.The R 2 X (cum) value, which indicates the explained variation in X (predictor) variables, was found to be 0.845.The R 2 Y (cum) value, representing the explained variation in Y (response) variables, was 1, indicating a good fit of the model (Figure 6A).The Q 2 (cum) value, which estimates the model's predictive ability, was determined to be 0.939, suggesting a high predictability.To validate the model, a permutation test was conducted with 200 iterations.The permutation plot showed that both the Q 2 and R 2 values obtained from the permutations (R 2 = [0.0,0.984]; Q 2 = [0.0,0.375]) were significantly lower than the original values (Figure 6B).This indicates that the model did not exhibit randomness and overfitting, further supporting its reliability.S-plots and volcano plots (Figure 6C,D) were generated under the OPLS-DA model to facilitate the rapid and visual identification of key markers.In S-plots, red indicates a value of VIP > 2, while green signifies VIP < 2. In the vol-cano plots, red spots represent variables with significant differences and a high abundance, while green spots represent variables with significant differences but a low abundance.Chemical markers that met certain criteria (VIP > 2, FC > 2 or <0.5, and p < 0.05) in the univariate statistical analysis were considered to be significant.Twelve markers displaying notable differences between FZ and XJ were identified and are listed in Table 2.These potential markers were further visualized using a heatmap (Figure 6E).In the heatmap, rows correspond to individual metabolites, while columns represent distinct samples.The coloration of each cell within the heatmap reflects the expression level of the respective metabolite in the corresponding sample.In the heatmap, high metabolite concentrations are denoted by red, whereas low concentrations are represented by blue.From the heatmap, it can be observed that one chemical marker (peak 4) exhibited higher relative contents in XJ compared to FZ.On the other hand, two chemical markers (peak 15,16) were found to be higher in FZ than in XJ.These markers represented metabolites that are potentially important in distinguishing between FZ and XJ.In conducting a pairwise comparison between FZ and XJ utilizing the statistical method OPLS-DA, several parameters were evaluated to assess the model's performance.The R 2 X (cum) value, which indicates the explained variation in X (predictor) variables, was found to be 0.845.The R 2 Y (cum) value, representing the explained variation in Y (response) variables, was 1, indicating a good fit of the model (Figure 6A).The Q 2 (cum) value, which estimates the model's predictive ability, was determined to be 0.939, suggesting a high predictability.To validate the model, a permutation test was conducted with 200 iterations.The permutation plot showed that both the Q 2 and R 2 values obtained from the permutations (R 2 = [0.0,0.984]; Q 2 = [0.0,0.375]) were significantly lower than the original values (Figure 6B).This indicates that the model did not exhibit randomness and overfitting, further supporting its reliability.S-plots and volcano plots (Figure 6C,D) were generated under the OPLS-DA model to facilitate the rapid and visual identification of key markers.In S-plots, red indicates a value of VIP > 2, while green signifies VIP < 2. In the volcano plots, red spots represent variables with significant differences and a high abundance, while green spots represent variables with significant differences but a low abundance.Chemical markers that met certain criteria (VIP > 2, FC > 2 or <0.5, and p < 0.05) in the univariate statistical analysis were considered to be significant.Twelve markers displaying notable differences between FZ and XJ were identified and are listed in Table 2.These potential markers were further visualized using a heatmap (Figure 6E).In the heatmap, rows correspond to individual metabolites, while columns represent distinct samples.The coloration of each cell within the heatmap reflects the expression level of the respective metabolite in the corresponding sample.In the heatmap, high metabolite concentrations are denoted by red, whereas low concentrations are represented by blue.From the heatmap, it can be observed that one chemical marker (peak 4) exhibited higher relative contents in XJ compared to FZ.On the other hand, two chemical Similarly, a pairwise comparison was conducted between the FZ and YD groups.The OPLS-DA score plot further confirmed a significant distinction between the two groups, indicating a highly significant model (R 2 X = 0.889, R 2 Y = 1, Q 2 = 0.993) (Figure 7A).The established OPLS-DA model (Figure 7B) demonstrated reliability and reproducibility (R 2 = [0.0,0.948], Q 2 = [0.0,0.119]).Additionally, S-plots and volcano plots (depicted in Figure 7C,D) were generated in order to pinpoint potential chemical markers.By applying criteria such as VIP > 2, FC > 2 or <0.5, and p < 0.05, a total of 23 potential markers were identified, as presented in Table 3.The results were visually represented in a heatmap format (Figure 7E) for intuitive interpretation.Furthermore, it was observed that FZ exhibited significantly higher levels of glycerolipids, including TG 54:     heatmap format (Figure 7E) for intuitive interpretation.Furthermore, it was observed that FZ exhibited significantly higher levels of glycerolipids, including TG 54:8 (65), TG 54:6 (73), TG 54:9 (56), TG 54:7 (69), TG 54:5 (77), and TG 54:4 (84), compared to YD.Conversely, the steroids 1α,25-dihydroxy-19-nor-22-oxacholecalciferol ( 13) and 1α,25-dihydroxy-2β-hydroxymethyl-19-norvitamin D3 ( 16) had higher contents in YD compared to FZ.Likewise, a comparative study was executed between the XJ and YD samples.The OPLS-DA score plot confirmed a significant distinction between the two groups, indicating a highly significant model (R 2 X = 0.905, R 2 Y = 1, Q 2 = 0.997) (Figure 8A).The established OPLS-DA model (Figure 8B) demonstrated reliability and reproducibility (R 2 = [0.0,0.965], Q 2 = [0.0,0.27]).S-plots and volcano plots (Figure 8C,D) were generated to identify potential chemical markers.By applying criteria such as VIP > 2, FC > 2 or < 0.5, and p < 0.05, a total of 39 potential markers were identified and are presented in Table 4.The results were visually represented in a heatmap format (Figure 7E) for intuitive interpretation.Moreover, XJ exhibited significantly higher levels of glycerolipids, including TG 54:8 (65), TG 54:6 (73), TG 54:9 (56), TG 54:7 (69), TG 54:5 (77), and TG 54:9 (84), compared to YD.Conversely, YD showed a higher sterol content, specifically in 1α,25-dihydroxy-19-nor-22-oxacholecalciferol (13) and 1α,25-dihydroxy-2β-hydroxymethyl-19-norvitamin D3 ( 16), compared to XJ.In conclusion, the secondary metabolite composition of marigold flowers is significantly influenced by variations in soil conditions, climatic factors, and growth environments specific to their regional origins.As evidenced by the classification of the differential metabolites identified in samples from three distinct production areas, these variations predominantly involve non-pigment components.The relatively small differences observed in pigment content among the samples from the three regions indicate that pigments are not only influenced by environmental factors, but also have inherent associations with factors such as variety, cultivation management, harvesting, processing, and quality control.In industrial applications, marigold flowers, serving as a source of pigments, demonstrate a limited impact of production region on their pigment content, thus contributing to the stability of supply and consistency in product quality.Moreover, non-pigment components, particularly glycerides, sterols, and steroid compounds, are significantly affected by regional environmental factors.From a scientific perspective, this diversity facilitates the elucidation of connections between non-pigment constituents in marigold flowers and ecological factors, genetic determinants, and so on, spurring advancements in fields such as plant chemistry, ecology, and genetic breeding.Concurrently, this enables the development of diversified products tailored to market demands and customer preferences.Additionally, the analytical methodology established in this study is not only applicable to the comparative analysis of marigold oil resin samples from different production regions, but also possesses the versatility to be extended to the investigation of variations among samples derived from various plant species and distinct processing techniques.In conclusion, the secondary metabolite composition of marigold flowers is significantly influenced by variations in soil conditions, climatic factors, and growth environments specific to their regional origins.As evidenced by the classification of the differential metabolites identified in samples from three distinct production areas, these variations predominantly involve non-pigment components.The relatively small differences observed in pigment content among the samples from the three regions indicate that pigments are not only influenced by environmental factors, but also have inherent associations with factors such as variety, cultivation management, harvesting, processing, and quality control.In industrial applications, marigold flowers, serving as a source of pigments, demonstrate a limited impact of production region on their pigment content, thus contributing to the stability of supply and consistency in product quality.Moreover, non-pigment components, particularly glycerides, sterols, and steroid compounds, are

Conclusions
A comprehensive approach combining a multi-dimensional sampling strategy and diverse data analysis techniques was devised to characterize and classify marigold oleoresin samples originating from three different regions.Using GNPS, a total of 83 compounds were identified or tentatively characterized in the marigold oleoresin samples.These constituents included 6 vitamin E and vitamin E esters, 13 triterpenes, 7 lutein diesters, 4 lutein monoesters, 10 ceramides, 16 triglycerides, 13 diglycerides, 6 monogalactosyldiacylglycerols, 3 steroids, and 2 others.Subsequently, a multivariate statistical analysis coupled with a heatmap revealed clear distinctions among the marigold oleoresin samples from the three different regions.Specifically, a total of 12, 23, and 38 significantly different metabolites were identified between FZ and XJ, FZ and YD, and XJ and YD, respectively.This finding indicates significant variations in the metabolite composition of marigold oleoresin based on its origin.The results of this study can be used to distinguish marigold oleoresin samples from different regions, laying a solid foundation for further quality control and providing a theoretical basis for assessing safety and nutritional aspects.

Figure 1 .
Figure 1.Representative base peak intensity chromatogram of marigold extract in the positive ion mode.Figure 1. Representative base peak intensity chromatogram of marigold extract in the positive ion mode.

Figure 1 .
Figure 1.Representative base peak intensity chromatogram of marigold extract in the positive ion mode.Figure 1. Representative base peak intensity chromatogram of marigold extract in the positive ion mode.

Figure 4 .
Figure 4.The enlargement of cluster III molecular network (A), MS spectra (B) and possible fragmentation pathway of lutein dimyristate and lutein 3-O-palmitate from marigold oleoresin in positive ion mode.

Figure 4 .
Figure 4.The enlargement of cluster III molecular network (A), MS spectra (B) and possible fragmentation pathway of lutein dimyristate and lutein 3-O-palmitate from marigold oleoresin in positive ion mode.

Figure 6 .
Figure 6.Multivariate statistical analysis result of FZ and XJ groups: OPLS-DA (A) score plots and OPLS-DA model permutation test (B), S-plot (C) and volcano plots (D), (C,D) were used to explore markers in FZ and XJ, the heatmap of potential chemical markers (E).

Figure 6 .
Figure 6.Multivariate statistical analysis result of FZ and XJ groups: OPLS-DA (A) score plots and OPLS-DA model permutation test (B), S-plot (C) and volcano plots (D), (C,D) were used to explore markers in FZ and XJ, the heatmap of potential chemical markers (E).

Figure 7 .
Figure 7. Multivariate statistical analysis result of FZ and YD groups: OPLS-DA (A) score plots and OPLS-DA model permutation test (B), S-plot (C) and volcano plots (D), (C,D) were used to explore markers in FZ and YD, the heatmap of potential chemical markers (E).

Figure 8 .
Figure 8. Multivariate statistical analysis result of XJ and YD groups: OPLS-DA (A) score plots and OPLS-DA model permutation test (B), S-plot (C) and volcano plots (D), (C,D) were used to explore markers in XJ and YD, the heatmap of potential chemical markers (E).

Figure 8 .
Figure 8. Multivariate statistical analysis result of XJ and YD groups: OPLS-DA (A) score plots and OPLS-DA model permutation test (B), S-plot (C) and volcano plots (D), (C,D) were used to explore markers in XJ and YD, the heatmap of potential chemical markers (E).

Table 1 .
Identification of the chemical compounds of marigold oleoresin by UPLC-QTOF-MS.

Table 2 .
Statistical table of specific differential metabolites of FZ vs. XJ.

Table 2 .
Statistical table of specific differential metabolites of FZ vs. XJ.

Table 3 .
Statistical table of specific differential metabolites of FZ vs. YD.

Table 4 .
Statistical table of specific differential metabolites of XJ vs. YD.