2.1. Materials

P. cyrtonema Hua (wet) was obtained from Anhua (Hunan, China). Samples were steamed and dried nine times in the laboratory, and the processed samples were stored in Room 312 on the third floor of the laboratory building.

2.2. The Different Production Stages of Steaming Polygonatum cyrtonema Hua

The processing method of P. cyrtonema Hua refers to the previously published articles [22]. The whole dried rhizome of P. cyrtonema Hua was steamed for 2 h at 105°C and then placed for 22 h in a high-pressure steam sterilization pot (0.12 MPa) with 1–9 cycles. The fresh sample was marked as PF and the processed P. cyrtonema Hua samples from one, two, three, four, five, six, seven, eight, and nine cycles were collected and marked as P1, P2, P3, P4, P5, P6, P7, P8, and P9, respectively.

2.3. HS–GC–IMS Analysis

In this experiment, the IMS instrument (FlavourSpec, Gesellschaft für Analytische Sensorsysteme mbH, Dortmund) equipped with a CTC-PAL 3 static headspace automatic sampling device (CTC Analytics AG) that can be directly sampled from the headspace using a 1 mL air-tight heated syringe was used. The GC Agilent 490 (Agilent Technologies) was operated using the carrier gas nitrogen (99.999% purity) at programmed flow rates using a MXT-WAX capillary column (30 m, 0.53 mm ID, 1 μm FT) at a constant temperature of 60°C. Each sample was accurately weighed at 1.0 g and placed in a 20 mL headspace bottle, and incubated at 80°C for 20 min before sampling.

2.4. Statistical Analysis

The identification of VOCs was based on the retention index (RI) and drift time (reactive ion peak [RIP] relative) in the GC–IMS library. The analysis software used was VOCal and three plugins (Reporter, Gallery Plot, and Dynamic principal component analysis (PCA) plugins). VOCal software was used to view the analytical spectrum and qualitatively and quantitatively analyze the data. The Reporter plugin can directly compare the differences in spectra of the samples (3D spectrum, 2D top view, and difference spectrum). The Gallery Plot plugin for fingerprint comparison was used to compare the differences in VOCs intuitively and quantitatively among different samples. Chemometrics were performed by SIMCA 14.1 (MKS Umetrics, Sweden).

3.1. Color and Odor Assessments

The appearance of P. cyrtonema Hua during the nine-steam-nine-bask processing is shown in Figure 1. The fresh P. cyrtonema Hua (PF) sample was bright yellow, as expected. With the increase in processing time, the color gradually darkened, and the change was inconspicuous after three to four processing cycles. Since then, the color remained dark brown; meanwhile, the odor intensity changed proportionally during processing. The distribution of the fresh from the other steaming samples could be clearly observed. The essence was divided into four categories based on color and odor: PF, Step 1 (P1, P2, and P3), Step 2 (P4, P5, and P6), and Step 3 (P7, P8, and P9). In some reports, thermal processing often changed the color of the sample and affected the quality of obtained products [22]; the changes were associated with the Maillard reaction. During the nine-steam-nine-bask processing, the reducing sugar of P. cyrtonema Hua reacted with amino compounds to produce brown or even black macromolecular substances. However, the formation and conversion of characteristic volatile compounds in the processing step need further analysis. In this study, PF, P3, P6, and P9 were investigated via GC–IMS to find the relationship between the final flavor and the entire nine-steam-nine-bask processing.

3.2. Analysis of Different Production Stages of Steaming P. cyrtonema Hua by HS–GC–IMS

In this study, the VOCs in different production stages were determined by HS–GC–IMS. The HS-GS-IMS 3D spectrum of PF, P3, P6, and P9 are presented in Figure 2. The X-, Y-, and Z-axes represented ion drift time, retention of GC, and ion peak strength, respectively [23]. From this 3D drawing, we can intuitively observe the difference in different production stages of steaming P. cyrtonema Hua (PF, P3, P6, and P9). The background is blue, and each signal spot to the right of the RIP represents a volatile compound. For better observation, in this study, a top view was used for comparative analysis of differences.

The top view plot (Figure S1-1) of HS–GC–IMS was obtained by the normalization of the ion drift time and RIP position. The whole spectrum represents the total VOCs, and each color indicates the concentration of individual compounds. White indicates a lower concentration and red indicates a higher concentration; the darker the color, the greater the concentration. Most signals appeared at a 100–1500(s) retention time and 1.0–1.75 drift time.

To clearly compare different samples, different comparison models were used. In this model, the PF spectrum was selected as the reference, and the spectra of other samples were subtracted (Figure S1-2). If the VOCs were the same, the background after deduction was white, while red indicated a higher concentration of the compound than the reference, and blue indicated a lower concentration of the compound than the reference. The red area is mainly distributed between 1100–1500(s) and 1.0–1.25, while the blue area is distributed between 500–1000(s) and 1.5–1.75.

3.3. Fingerprint Analysis of Different Production Stages of Steaming P. cyrtonema Hua

Fingerprints were used to further identify specific differences in different production stages. In the fingerprint, each row represents the signal peak of all compounds in one kind of sample, while each column represents the signal peak of the same volatile compound in different samples. The brightness of each square roughly represents the content of the volatile compound. Ninety-two compounds were produced in different processing stages of steaming P. cyrtonema Hua, of which 57 were identified (Table 1). As shown in Figure S2, the components included 17 aldehydes, 15 alcohols, 15 ketones, 5 esters, 3 furans, and 2 acids. The percentage of each type of compound is shown in Figure S3. Here, the contents of alcohol and ester were the highest in PF samples than in others and decreased with the progress in the production stages. Among them, 1-hexanol and 1-pentanol were present in three forms, as monomers, dimers, and polymers. There were 15 kinds of compounds with two existing forms, of monomers and dimers, mainly including (E)-2-octenal, (E)-2-heptenal, (E)-2-hexenal, 1-(2-furanyl)ethenone, 1-butanol, 1-hydroxy-2-propanone, 2-methyl-1-propanol, 2-methyltetrahydrofuran-3-one, 3-hydroxy-2-butanone, 3-methylbutan-1-ol, 4-methyl-2-pentanone, acetic acid, cyclopentanone, ethyl hexanoate, and heptaldehyde.

Further comparison of the changes before (PF) and after (P3, P6, and P9) steaming revealed common regions in the samples and exhibited their characteristic peaks. As shown in Figure 3, in PF samples, compounds in Region A had the highest concentration and were mainly isoamyl 3-methylbutyrate, 2-pentanone, 1-propanol, ethyl (E)-2-butenoate, (E)-2-octenal, (E)-2-hexenal, 1-pentene-3-ol, 1-pentene-3-one, heptaldehyde, 4-methyl-2-pentanone, 2-heptanone, 1-hexanol, (E)-2-heptenal, 1-pentanol, acetic acid ethyl ester, 3-methylbutan-1-ol, and 1-hexanol (Figure 4(a)). In addition, with the increase in the number of processing times, the compounds whose concentration decreased gradually included 1-hexanol dimer(12), 1-hexanol monomer(13), and 3-methylbutan-1-ol dimer(38) (p < 0.05). On the other hand, in Region B, the concentrations of VOCs after steaming were higher than before steaming and included benzaldehyde, 2-butanone, 1-hydroxy-2-propanone, acetic acid, butanal, 3-methyl butanal, 2-methyl propanal, 3-methyl-2-butenal, 2-methyltetrahydrofuran-3-one, 2-furaldehyde, and 1-(2-furanyl)ethenone (Figure 4(b)). With the increase in steaming times, the compounds whose concentration increased significantly included 1-(2-furanyl)ethanone monomer(8), 2-furaldehyde(25), and 3-methyl butanal(36) (p < 0.05).

3.4. Resemblance Analysis of Different Production Stages of Steaming P. cyrtonema Hua by PCA and Partial Least Squares Discrimination Analysis (PLS-DA)

PCA is a common statistical method often used to show the differences between samples [24]. PCA of different production stages of steaming P. cyrtonema Hua is shown in Figure S4. The parallel samples of different processed products appear closely spaced, indicating a good parallelism. The distance between fresh products and steamed samples is far, indicating a great difference in volatile components before and after steaming. The distance between samples with different times of steaming was shorter, indicating a little difference between different times of steaming. To sum up, the conclusion of PCA is consistent with that of Gallery Plot, and PCA analysis could distinguish the four samples. In addition, the contribution rates of principal components PC1 and PC2 were 59.8% and 29.3%, respectively.

In order to further understand the changes of VOCs during processing, 57 VOCs were used as independent variables and their different steaming times’ samples were used as dependent variables to perform PLS-DA. As shown in Figure S5 (X [1] = 0.658, X [2] = 0.256), the PF samples were on the rightmost of the Figure S5 and P9 were on the leftmost. The results indicate that steaming is the key factor in VOC change.

3.5. Differences of VOCs at the Different Production Stages of Steaming P. cyrtonema Hua

To clarify the differences of VOCs at the different production stages of steaming P. cyrtonema Hua, orthogonal PLS-DA (OPLS-DA) of PF vs. P3 and P3 vs. P6 and P6 vs. P9 was performed. OPLS-DA is a supervised statistical method for discriminant analysis that could better access information from group differences. Through analysis, a projected variable importance (VIP) value can be obtained for every metabolite, and the larger the VIP value, the greater the contribution to distinguish the two groups [25]. The OPLS-DA of PF vs. P3 are shown in Figure S6-1a (R₂X [cum] = 0.981, R₂Y [cum] = 1.0, Q2 = 1), Figure S6-2a (R₂X [cum] = 0.842, R₂Y [cum] = 0.999, Q2 = 0.990), and Figure S6-3a (R₂X [cum] = 0.884, R₂Y [cum] = 0.998, Q2 = 0.988), which showed that different production stages of steaming P. cyrtonema Hua might be classified through OPLS-DA. In order to prevent overestimation, the accuracy of the OPLS-DA model was confirmed through a rearrangement test, and the results of 200 transverifications showed that R2 and Q2 of all rearrangement trials were lesser than the raw data, and Q₂Y is all worth less than 0.05, indicating that the simulation equation was not overadapting (Figures S6-1b, S6-2b, and S6-3b). Therefore, the established OPLS-DA simulation results are consistent and reliable.

In addition, according to the principle of VIP > 1 and fold change ≥ 2 or ≤ 0.5 for the screening of the different VOCs, 11 differences were observed between PF and P3 (Figure 5(a)); 4 VOCs of 2-methyl propanal, 2-methyl-1-propanol, 2-methyltetrahydrofuran-3-one, and 4-methyl-2-pentanone showed upregulation; 7 VOCs of (E)-2-octenal, (E)-2-heptenal, (E)-2-hexenal, 1-penten-3-one, 4-methyl-2-pentanone, ethyl (E)-2-butenoate, and isoamyl 3-methylbutyrate showed downregulation. Five differences were observed between P3 and P6 (Figure 5(b)); (E)-2-octenal and (E)-2-heptenal showed upregulation; 1-hexanol polymer, 1-pentanol polymer, and ethyl hexanoate dimer showed downregulation. There are 7 differences between (E)-2-heptenal, (E)-2-hexenal, 1-butanol, 1-hexanol polymer, 1-pentanol, 1-pentanol polymer, and 3-methylbutan-1-ol showing downregulation between P6 and P9 (Figure 5(c)). The outcomes of gathering heatmap showed that the selected 18 VOCs were good for distinguishing the differences between fresh samples and different production stages of steaming P. cyrtonema Hua (Figure 6).

3.6. Nearest Neighbor Fingerprint Analysis of Different Production Stages of Steaming P. cyrtonema Hua

The NNFA can quickly compare the samples according to the intensity of the compounds in the selected evaluation area. It works by calculating the Euclidean distance between every two samples, and finding the “nearest neighbor” by searching for the minimum distance [26]. As shown in Figure S7 and Table S1, the Euclidean distance before and after steaming was far, indicating that the volatile components changed greatly before and after steaming, while the processed products with different steaming times were relatively concentrated, and a relatively small difference was observed between them.