Raw materials: Fresh Cistanche deserticola samples were obtained from the Hetian region in Xinjiang Province of China. The samples were carefully selected with the same size (average length, diameter, and weight were 11.7 ± 1.1 cm, 7.0 ± 1.1 cm, and 360 ± 8.9 g, respectively). The samples were stored at room temperature in a dark environment with an initial moisture content of about 78.56% ± 3.47%. Prior to the experiments, Cistanche deserticola samples were washed with tap water to remove the dust on the surface. Excess water on its surface was removed by blotting paper.

Chemicals: Methanol, acetonitrile, and formic acid were liquid-chromatography mass spectrometry grade (LC-MS) and purchased from Merck (Sigma Aldrich, MO, USA). The other analytical standards presented a purity higher than 98% (Sigma Aldrich, MO, USA).

Previous studies have shown that the chemical compounds distribute unevenly in the longitudinal direction of Cistanche deserticola (1). Therefore, to obtain the same initial contents of chemical compounds in each sample, in the present research all selected Cistanche deserticola was cut into three equal parts for the fresh group (A), steamed without drying group (B), and dried after steaming group (C), respectively, by longitudinal segmentation with the longitudinal symmetry axis as the center (20).

For group B, the samples were successively steamed for 8 min according to preliminary experiments. A pulsed vacuum steaming equipment (self-developed by China Agricultural University, Beijing, China) was used for steaming treatment of fresh Cistanche deserticolas. Steamed samples were dried in a vacuum freeze-dryer (LGJ-25C, Si Huan Scientific Instrument Factory Co., Beijing, China). The heating plate and cold trap temperature were 30 and −60°C, respectively. For group C, the samples were successively steamed using pulsed vacuum equipment for 8 min and dried in the hot air impingement dryer (self-developed by China Agricultural University, Beijing, China) until the final moisture content of 10% (w.b.). The airflow rate and temperature were set at 6 m/s and 60°C, respectively, referring to the research results of Zou et al. (11). All samples were stored at −20°C no more than 7 days before further analysis.

The appearance color of the Cistanche deserticola before and after each thermal processing was measured using a colorimeter (SMY-2000SF, Shengming Yang Co., Beijing, China), and the blackness was characterized by L^* value.

The metabolite extraction was carried out according to the method reported previously by Chen et al. (21) with some minor modifications. In brief, the dried samples were crushed using a mixer mill (MM 400, Retsch Company, Haan, Germany) with a zirconia bead for 2 min at 60 Hz. Then 50 mg powder (sifted through a 65 mesh sieve) of each sample was precisely weighed, transferred to an Eppendorf tube, and extracted with 1 ml methanol/water mixture (v:v = 3:1). After 30 s vortex, the mixture was homogenized twice at 35 Hz for 4 min, sonicated for 15 min in an ice-water bath, and then shaken overnight at 4°C. After centrifugation at 12,000 rpm for 15 min at 4°C, the supernatant was collected and filtered through a 0.22-μm membrane, then the obtained extract was transferred to 2-ml glass vials and store at −80°C until the UHPLC-MS/MS analysis.

Qualitative and quantitative analyses of metabolites were performed according to the methods by Liu et al. (18). Primary and secondary mass spectrometry data were qualitatively analyzed based on the self-built human metabolome database (MWDB) (Metware Biotechnology Co., Ltd. Wuhan, China) and the public database. Meanwhile, to ensure the accuracy of the qualitative analysis of some substances, interferences from repeated signals of Na⁺, NH4+, K⁺ and ions, and repetitive signals of fragment ions derived from other relatively large molecules and isotope signals were removed during identification. Metabolite structural analysis was performed with reference to the public databases (Mass Bank, KNApSAcK, HMDB, MoTo DB, and METLIN).

Metabolite quantification was carried out using the MRM mode of the QQQ mass spectrometry. In the MRM mode, the precursor ions (parent ions) of the target substances and excluded ions corresponding to other substances with different molecular weights were screened first using the quadrupole rod to initially eliminate interference. The precursor ions then break through the collision chamber to form many fragment irons after ionization, which were filtered by QQQ to select single-fragment ions with the desired characteristics while eliminating interference from non-target ions. Finally, after obtaining the metabolite mass spectrometry data of different samples, the mass spectrum peaks of all substances were integrated, and the mass spectra peaks of the same metabolite in different samples were integrated and corrected using Multi Quant version 3.0.2 (ABSCIEX, Concord, Ontario, Canada). The corresponding relative metabolite contents were represented as chromatographic peak area integrals.

The metabolic data were processed using orthogonal partial least squares-discriminant analysis (OPLS-DA) and hierarchical cluster analysis (HCA). OPLS-DA was used to discriminate each group; it is more sensitive than other statistical methods to variables with low correlations (17). The OPLS-DA models were validated through a permutation analysis (200 times). The model was considered stable when the model parameters (R² and Q²) were both close to 1. The variable importance projection (VIP) values of metabolites were calculated. Any metabolite with VIP values greater than 1.0 and p-values less than 0.05 were selected as biomarkers for each paired comparison between different thermal processing stages of Cistanche deserticola. The screening of different metabolites was visualized in the form of the volcano plot. Metabolites accumulation among different samples was analyzed by using the R package (www.rproject.org/). The Venn diagram was built according to the program web-based smart diagram® (https://cloud.smartdraw.com/). The commercial databases, such as Kyoto Encyclopedia of Genes and Genomes (KEGG) (https://www.kegg.jp/kegg/), Pub Chem (https://pubchem.ncbi.nlm.nih.gov/), the Small Molecule Pathway Database (SMPDB) (https://smpdb.ca/), and HMDB (https://hmdb.ca/), were used for enrichment analysis of differential metabolites and finding metabolic pathways.

The difference in appearance color of Cistanche deserticola between the fresh, steamed and dried samples are representatively displayed in Figure 1. From fresh to dried sample with the going of the processing stage, the appearance color of the samples changed from yellow-brown to dark black, and the darkness in color became more and more obvious (the corresponding L^* value was reduced from 50.26 to 24.90). Obviously, the appearance changes of Cistanche deserticola mainly occurred in the steaming process. The Maillard reaction, during which sugars react with amino acids under thermal conditions (22), would be greatly responsible for the dark-colored appearance of processed rhizomes of Cistanche deserticola. Previous research studies have shown that precursors were converted into colorants and generated substances with dark color in the Maillard reaction (23). Similar findings were also observed in previous studies for steaming of Polygonum multiflorum (24) and rhizomes of Polygonatum cyrtonema (25). The darkness of the steamed samples was further deepened after drying. This phenomenon was probably due to the occurrence of a decrease in the pigment concentration during the drying process.

The total ion chromatogram (TIC) of quality control (QC) sample (a mixture of all the samples investigated) and a multi-peak detection plot of chemicals in the MRM mode of the same sample are illustrated in Supplementary Figure 1. Different colored peaks represented different components in the sample. As shown in Figure 2, a total amount of 776 metabolites were identified in the current study (Supplementary Table 1) in the fresh Cistanche deserticola samples, which were divided into 15 classes, including 40 amino acid and derivatives, 33 phenylpropanoids, 23 flavonoids, 68 flavone, 67 terpenes, 67 phenols, 87 alkaloids, 13 carbohydrates, 28 nucleotide and derivatives, 5 alcohols and polyols, 3 purine nucleosides, 15 carboxylic acids and derivatives, 14 organic acids and derivatives, 12 phytohormone, and 28 other chemicals. Among them, the largest group was amino acid and derivatives, the relative content of which accounted for 30.26% of the total metabolite composition. In addition, 10 kinds of phenylethanoid glycosides, such as echinacoside and verbascoside, were detected and classified in the phenylpropanoids group.

The accumulation pattern of metabolites among different treatment groups was analyzed by HCA. As shown in Figure 3, 107 identified metabolites of Cistanche deserticola were clustered in heat maps based on Euclidean distance arithmetic. Metabolites identified at different thermal processing stages were gathered into three clusters according to the dendrogram. The brighter color indicates the higher content of a particular metabolite in the respective sample. The heat map of HCA showed larger di?erences in abundance between the fresh and steamed samples than those between steamed and dried samples, indicating that metabolites in Cistanche deserticola may have different transformations during the steaming and drying stage, and the types and quantities of metabolites involved in the steaming process are more than those in the drying process.

For a better understanding of the impact of each processing on the metabolites of Cistanche deserticolas, the OPLS-DA scatter scores of pairwise comparison groups are shown in Figure 4A, showing that the fresh, steamed, and dried after steaming Cistanche deserticolas were significantly different. Moreover, the R²Y and Q² (as shown in Supplementary Figure 2) with high-test values indicated that this model was highly reliable without overfitting.

To screen the expression level of metabolites between the fresh, steamed, and dried after steaming Cistanche deserticola, the analysis of volcano plot was further applied among all 776 metabolites identified according to the fold-change, combined with VIP values to screen the differentially expressed metabolites. Significant differential metabolites were selected according to the criterion that a fold change score of ≥2 or ≤ 0.5 with a VIP ≥ 1. The screening results are illustrated in Figure 4B. In the volcanic map, each point represents a metabolite and the color of the scattered dots represents the final screening result. Red represents metabolites that are significantly upregulated (UR), green represents those significantly downregulated (DR), and gray represents those insignificantly different. As shown in Figure 4B, 47 metabolites in the fresh vs. steamed group (17 UR and 30 DR), 30 metabolites in steamed vs. dried group (22 UR and 8 DR), and 65 metabolites in the fresh vs. dried group (29 UR and 36 DR) were selected to be significantly differential. The number of significantly different metabolites in the fresh vs. steamed group was higher than those in the steamed vs. dried group, indicating that the influence on metabolites in the steaming process is higher than that of the drying process. The differential metabolites produced during thermal processing of Cistanche deserticola were further classified and compared. These differentially expressed metabolites were classified into 21 classes, mainly amino acids and their derivatives, flavonoids and their derivatives, phenylpropanoids, alkaloids, terpenes, phenols, and nucleotide and their derivatives (Table 1). In fresh vs. steamed group, it can be found that flavonoids (such as isoquercitrin, troxerutin, cyanidin, and fisetin), phenylpropanoids (such as chlorogenic acid and 3-(3,4-Dihydroxy-5-methoxy)-2-propenoic acid), and nucleotide and their derivatives (uracil and beta-Nicotinamide mononucleotide) were significantly DR, while amino acids and their derivatives (such as N6-Acetyl-L-lysine, 1-Methy-L-histidine, and L-Phenylalanine) were significantly UR. However, in the steamed vs. dried group, the expression trends of these types of differential metabolites were opposite. Some amino and their derivatives (such as N, N-Dimethylglycine), nucleotide and their derivatives (such as 2′-Deoxyuridine; Deoxyuridine) were significantly DR, while most of the phenols (such as methyl gallate and 4′-Prenyloxyresveratrol), flavonoids (such as isoquercitrin and cyanidin), phenylpropanoids (verbascoside), and terpenes (such as terpinolene and furanodiene) were significantly UR.

These results showed that the chemical composition of Cistanche deserticola has undergone conversion during thermal processing, which is mainly reflected in the conversion of flavonoids, phenylpropanoids, and amino acids, and the conversion mechanism of these components is different in different processing stages. The use of a high temperature during the steaming and drying processes was previously found to promote hydrolysis, redox, isomerization, substitution, and other thermophysical and chemical reactions of metabolites (26). In this study, it was found that the metabolites, such as flavonoids and phenylpropanoids, were significantly accumulated in the steamed Cistanche deserticola compared to their corresponding fresh one, indicating that some key physiological and metabolic activities leading to the synthesis of flavonoids and phenylpropanoids might be activated under high temperature and humidity. This result can also be supported by the report from Peng et al. (10) who found that the content of PhGs (belonging to phenylpropanoids) increased after steaming. However, the accumulation of these components in the dried sample after steaming showed a significant decrease, which may be attributed to the thermal degradation of these heat-sensitive components during the long-term drying process. Previous studies have shown that flavonoid glycosides can be decomposed into sugar bodies and flavonoid aglycones under thermal conditions, and flavonoids loss during the drying process was synthetically affected by temperature and drying time (26, 27). The upregulation of amino acids and their derivatives (N6-Acetyl-L-lysine, 1-Methy-L-histidine, and L-Phenylalanine) is attributed to the high-temperature-promoting protein degradation during steaming processing. In addition, it was also observed that some other amino acids and their derivatives of N, N-Dimethylglycine, L- Kynurenine, glycine, serine, and threonine were DR. The decrease in the content of these amino acids might be associated with thermal-induced Maillard reaction during which reducing sugars react with amino acids to generate 5-HMF, contributing to the production of black appearance in Cistanche deserticola (22). The results of Section appearance color changes of cistanche deserticola during thermal processing further verified this hypothesis. Therefore, the blackening of Cistanche deserticola during steaming was probably related to the metabolism of amino acids.

Venn diagram was used to differentiate the common and exclusive metabolites of Cistanche deserticola during different thermal processing stages. As shown in Figure 4C, both common and unique metabolites exist between the different comparison groups. Twenty-one common metabolites were observed between the fresh and steamed group, while only 5 and 10 metabolites were found common between the fresh and dried group, steamed and dried group, respectively. Thus, a total of 23 and 17 exclusive metabolites (p < 0.05) were observed in Cistanche deserticola during the thermal processing stage of steaming and drying, respectively. This result further confirmed that steaming was particularly critical for the conversion of metabolites during Cistanche deserticola processing.

The differential metabolites (p < 0.05) in fresh and processed samples were mapped to the KEGG, HMDB, and PubChem online databases, which contain knowledge of the molecular interaction, reaction, and relation networks, and the enrichment results and detailed metabolic pathways are shown in Supplementary Table 2 and Figure 5. As shown in Figures 5a1,a2, pathway impact revealed the enrichment of phenylpropanoid biosynthesis, flavonoid biosynthesis, alanine metabolism, riboflavin metabolism, taurine and hypotaurine metabolism, and nicotinate and nicotinamide metabolism during steaming of Cistanche deserticola. Whereas, during the drying process after steaming, the metabolic pathways of the differential metabolites mainly contained glycine, serine and threonine metabolism, thiamine metabolism, pyrimidine metabolism, and unsaturated fatty acids biosynthesis. Furthermore, some metabolic pathways between these two pairwise comparisons overlapped, such as nicotinate and nicotinamide metabolism, phenylpropanoid biosynthesis, and flavonoid biosynthesis, but their enrichment levels were very different in two pairwise comparisons. These results suggested that the conversion pathways of metabolites between the steaming and drying processes of Cistanche deserticola were different, and the differences in metabolic pathways could explain the differences in the presence of differentially exclusive metabolites during thermal processing. These biochemical alterations might be used to comprehend the impact of thermal processing stages on Cistanche deserticola composition.

Based on the KEGG annotation and enrichment analysis, four metabolic pathways (phenylpropanoid biosynthesis, flavonoid biosynthesis, alanine metabolism, and glycine, serine, and threonine metabolism) were chosen as key metabolites to characterize the conversion of the main active components of Cistanche deserticola during thermal processing (Figures 5b1,b2). The current study indicated that phenylpropanoids and flavonoids were accumulated but amino acids were degraded in steamed Cistanche deserticola compared to fresh and dried samples. The phenylpropanoid biosynthetic pathway is upstream of biosynthetic pathway of flavonoid. Similar conclusions were published by Liu et al. (18) who reported that the accumulation level of phenylpropanoids in the process of rice yellowing has increased significantly, compared with normal rice. Phenylpropanoids are derived from cinnamic acid, and their precursor is phenylalanine, which can be synthesized by activating the activity of phenylalanine ammonia-lyase (PAL) when heated (28). Previous studies reported that the phenylpropanoid pathway led to the biosynthesis of coumarins, flavones, isoflavones, and flavanols, which are the important weapons for plant defense (29), and to prevent cell death caused by the strong heat stress in the steaming process, the phenylpropanoid pathway may be enhanced due to the biological stress caused by high temperature (30, 31). Flavonoids are the main secondary metabolites derived from phenylpropanoids (32), and their accumulation could protect plants from oxidative damage by scavenging-free radicals (33). Compared to the fresh and dried Cistanche deserticola, the higher biosynthesis of flavonoids in the steamed Cistanche deserticola may be associated with enhanced heat stress during the steaming process providing protection against reactive oxygen species (ROS) (34, 35). As shown in Figures 5b3,b4, amino acid metabolism played an important role in the thermal processing of Cistanche deserticola. Content changes of alanine, glycine, serine, and threonine after steaming found in medicinal herbs have been used to indicate the occurrence of the Maillard reaction (36). Nevertheless, due to the complicated process of the Cistanche deserticola steaming, a comprehensive evaluation of the Cistanche deserticola steaming, such as blackening in appearance, active compounds, and metabolic biomarkers, should be further investigated.