Changes in volatile compounds of fermented minced pepper during natural and inoculated fermentation process based on headspace–gas chromatography–ion mobility spectrometry

Abstract Changes in volatile compounds of fermented minced pepper (FMP) during natural fermentation (NF) and inoculated fermentation (IF) process were analyzed by the headspace–gas chromatography–ion mobility spectrometry (HS‐GC‐IMS). A total of 53 volatile compounds were identified, including 12 esters, 17 aldehydes, 13 alcohols, four ketones, three furans, two acids, one pyrazine, and one ether. Generally, fermentation time played an important role in volatile compounds of FMP. It was found that most esters, aldehydes, and alcohols obviously decreased with the increase in fermentation time, including isoamyl hexanoate, methyl octanoate, gamma‐butyrolactone, phenylacetaldehyde, methional, and E‐2‐hexenol. Only a few volatile compounds increased, especially for 2‐methylbutanoic acid, 2‐methylpropionic acid, linalool, ethanol, and ethyl acetate. However, no significant difference in volatile compounds was found between NF and IF samples at the same fermentation time. In addition, the fermentation process in all samples was well discriminated as three stages (0 days; 6 day; and 12, 18, and 24 days), and all volatile compounds were divided into two categories (increase and decrease) based on principal component analysis and heat map.

| 3363 CHEN Et al. Rodríguez-Burruezo, 2018). Nowadays, pepper fruits are an important ingredient in several fermented food, including kimchi, fermented pepper paste, and fermented minced pepper (FMP;Wang, Wang, Xiao, Liu, Jiang, et al., 2019). FMP, as a traditional and local fermented vegetable in the southern regions of China, is widely consumed due to its nutritional and sensory properties (Li, Zhao, et al., 2016). It can be eaten directly or used as cooking ingredient. FMP can be prepared by natural fermentation (NF) and inoculated fermentation (IF). NF was a kind of fermentation methods in which microorganisms from natural environment were used for fermentation, and IF was a kind of fermentation methods in which inoculated microorganisms were used for fermentation. Whatever NF or IF, lactic acid bacteria (LAB) become the dominant microorganism when conditions are suitable for their growth (Sanlier, Gökcen, & Sezgin, 2017). Meanwhile, LAB fermentation involves the production of various metabolites in FMP such as alcohols, organic acids, and active metabolites that contribute to its nutrition, taste, flavor, and functionality Wang, Wang, Xiao, Liu, Jiang, et al., 2019).
Flavor plays a key important role in defining sensory and consumer acceptance of FMP. The flavor of FMP is a complex trait, including hot taste from peppers, umami taste from amino acids, and salty taste from NaCl. Previous researchers have focused on flavor characteristics of fermented vegetables products, including fermented peppers, sauerkraut, and kimchi (Sanlier et al., 2017).  showed that alcohols, esters, and ketones were the dominant volatile fractions in fermented/chopped pepper by solid-phase microextraction and gas chromatography-mass spectrometry, and the fermentation stage was mainly affected by esters, alcohols, aldehydes, and terpenes. Liu et al. (2019) found that the flavor profiles of Sichuan pickle fermented in glass jars (GL), porcelain jars (PO), and plastic jars (PL) were different. The compound with the highest concentration in both PO and GL was the alkanes, while the highest concentration of compound was ester in the PL. Wu et al. (2015) measured changes of flavor compounds in suan cai during NF, and found that there were 17 varieties of volatile flavor components in the early fermentation time, but increased to 57 in the middle fermentation time. In addition, the result also showed that esters and aldehydes were in the greatest diversity and abundance, contributed most to the aroma of suan cai. Kang and Baek (2014) found that 19 aroma-active compounds were detected by aroma extract dilution analysis in Korean fermented red pepper paste (gochujang), and 12 aroma-active compounds were detected by headspacesolid-phase microextraction-gas chromatography-olfactometry.
Hence, the variety and content of volatile compounds in fermented vegetables could be affected by multiple factors, such as raw material, container, and fermentation time. For FMP, the previous research and industrial production mainly focused on the optimization of process parameters, including raw material, NaCl and CaCl 2 content, LAB inoculum, fermentation temperature, and time based on the change of quality. However, knowledge about changes in flavor compounds of FMP at different stages with NF or IF is not available.
Ion mobility spectrometry (IMS), a rapid detection technique, was used to detect the gasified volatile compounds by ion separation based on their ion mobility velocity (Zhang et al., 2016). The detection technology presented many advantages, including easy operation, high analysis speed, high sensitivity, and no complex sample preparation steps. However, its analysis characteristics were often limited for complex samples, especially for complex systems in food and agricultural products (Arce et al., 2014).
Combining IMS with other instruments is a more suitable and effective way to make better use of its advantages. Recently, headspace-gas chromatography-ion mobility spectrometry (HS-GC-IMS) has been applied in detecting volatile compounds of fruits and vegetables, such as candied kumquats (Hu et al., 2019), jujube fruits (Yang et al., 2019), and dried peppers (Ge et al., 2020).
As a consequence, HS-GC-IMS is effective to identify the flavor characteristic of fruits and vegetables. However, the changes in characteristic compounds linked to the flavors of FMP during NF and IF process are still not available. Hence, HS-GC-IMS can be used to establish fingerprints of volatile compounds in FMP during NF and IF process.
In this study, the changes of volatile compounds in FMP during NF and IF process were analyzed, and several target volatile compounds in samples were detected using HS-GC-IMS, principal component analysis (PCA), and the heat map. The results confirmed the potential of HS-GC-IMS to identify the volatile compound characteristics and provided a rapid method to determine the flavor quality of FMP during fermentation process. stirred for 5 min. Above samples were divided into two groups: One group inoculated with 5% (w/w) W-4 LAB (10 7 CFU/ml) was regarded as IF, and another group without inoculated LAB was regarded as NF.
Prepared samples were put into pickle jars, and the jars were sealed with water to exclude air and fermented in a 30°C incubator. FMPs were obtained on 0, 6, 12, 18, and 24 days, respectively. The collected samples were detected immediately after freeze-drying and grinding.

| HS-GC-IMS analysis
All the analyses were obtained by HS-GC-IMS instrument and related supplementary analysis software according to the method of Sun et al. (2019) with some modifications. Freeze-dried (0.5 g) FMP was weighted and then transferred into a 20-ml headspace bottle. The FMP was incubated at 80°C in the headspace bottle, with the speed of 500 rpm for 10 min. After incubation, 500 μl headspace was automatically injected using a heated syringe at 85°C into a FS-SE-54-CB-1 (15 m × 0.53 mm ID) capillary column. The carried gas during injection was nitrogen (99.99% purity), which was under the below programmed flow to carry samples: 2 ml/ min held for 2 min, flowed ramp from 2 ml/min to 100 ml/min in 18 min, and then maintained 100 ml/min for 10 min until stopping. The analyses were separated in the column at 60°C and then ionized in the IMS ionization chamber at 45°C. The constant flow of drift gas flow was set up to 150 ml/min. The instrument was standardized by linear retention index (RI) of n-ketones, for the reason that IMS had no response to alkanes. Calculation of RI of volatile compounds was based on the n-ketones C 4 -C 9 . Comparing the drift time and RI in the GC-IMS library, volatile compounds in NF and IF samples were well identified. The qualitative analysis of volatile compounds was conducted based on the IMS and NIST database built in GC × IMS Library Search. The quantitative analysis for volatile compounds was mainly based on the peak intensity in HS-GC-IMS, and the peak intensity was proportional to the content of volatile compounds.

| Data analysis
All the experiment was performed in triplicate. The spectra were analyzed with laboratory analytical viewer (LAV), and the difference profiles and fingerprints of volatile compounds were constructed with the Reporter and Gallery plug-ins. The NIST and IMS database were built into the software for qualitative analysis of samples.
The line and bar charts drawn by Origin 2018 were used to analyze change of volatile compounds in FMP during fermentation process.
PCA obtained by the original date was used for clustering analysis of principal compounds in samples. The heat map was generated using the heat map plug-in of origin 2018, and the methods used for clustering of original date were Ward minimum variance and Euclidean distance.

| Changes of HS-GC-IMS spectra in FMP
The volatile compounds of FMP during NF and IF were detected by HS-GC-IMS. The data were exhibited with the 3D spectrum, where the x-axis represented the ion migration time, the y-axis represented the retention time of the gas chromatograph, and the z-axis represented the peak intensity. As shown in Figure 1, there was the similar peak signal distribution in all samples during fermentation process.
This phenomenon suggested that all FMP posed the same volatile compounds during fermentation process. However, the peak signal intensity showed some differences in FMP during fermentation process. It indicated that content of volatile compounds changed (increasing or decreasing) with the prolonging of fermentation time. Yu et al. (2013) proved that the prominent microorganisms posed clear relationships with changes in flavor during fermentation. However, no obvious differences in volatile compounds were found between NF and IF samples. It was found that the signal intensities of almost all volatile compounds in IF samples were similar to those in NF ones.
This phenomenon showed that two fermentation methods played minor effects on changes in volatile compounds during fermentation process.
The volatile compounds in FMP were not easy for analysis by 3D spectra. Hence, the 2D spectra were used for further comparison, as shown in Figure  As shown in Figure 2a,b, the total number of dots in 2D spectra hardly changed. However, the number of blue or red dots changed (increase or decrease). These results indicated that the varieties of volatile compounds in samples were the same, but the content of volatile compounds changed with the prolonging of fermentation time. The reason was that metabolism and decomposition of microorganisms caused volatile compounds to increase or decrease during fermentation process. Wu et al. (2015) showed that various bacteria and fungi, especially for LAB and yeast, possessed obvious correlation with the changes in volatile compounds during fermentation process. In addition, some pathways of biosynthesis and degradation also existed in FMP, including EMP pathway, Strecker pathway, and decarboxylation pathway, which could increase volatile compound content (Kang & Baek, 2014;Li, Zhao, et al., 2016;Li, Dong, Huang, & Wang, 2016). However, the color of red or blue dots in NF and IF samples was similar to each other. Therefore, the content of volatile compounds in NF and IF samples was almost the same. This result was because of the fact that LAB was the prominent microorganism under suitable conditions whatever NF or IF.

| Qualitative analysis of volatile compounds in FMP
The qualitative analysis of volatile compounds in FMP during fermentation process was represented by numbers, as shown in Figure 3 and Table 1. Each dot represented a type of volatile compound. The marked dots were identified volatile compounds, but unmarked dots were nonidentified volatile compounds. All FMP showed 53 identified dots by GC × IMS library analysis. Therefore, there were the same volatile compounds in all samples. This result indicated that fermentation methods hardly affected varieties of volatile compounds during fermentation process. In Figure 3 and Table 1 Li et al. (2010) showed that the same volatile compounds with different concentrations also might produce multiple signals in tricholoma matsutake. In addition, new formed dimer showed higher molecular weight than the monomer, and further generated multiple signals.

| Fingerprint analysis of volatile compounds in FMP
Although the 3D and 2D spectrum presented the change tendency of volatile compounds during NF and IF, the specific volatile compounds were not accurately judged. Hence, all signal peaks were used to make a detailed fingerprint analysis, as shown in Figure 4. In fingerprints, each column represented a kind of volatile compounds, each row represented the FMP samples, and volatile compound content was determined by the brightness degree of color.
In the raw materials (fermented on 0 days), high content of esters, aldehydes, and alcohols were found, especially for isoamyl hexanoate, gamma-butyrolactone, phenylacetaldehyde, methional, E-2-hexanol, and 1-butanol. The change in volatile compounds during NF and IF was presented in different color frames. In the red frame, the content of volatile compounds, such as hexyl hexanoate,

| Changes in volatile compounds of FMP
Seven kinds of volatile compounds, including eight esters, eight aldehydes, eight alcohols, and 11 other volatile compounds (four ketones, three furans, two acids, one ether, and one pyrazine), were used to explore the variation during NF and IF. Vegetable fermentation could be divided into aerobic fermentation and anaerobic fermentation (Gobbetti, Di Cagno, & De Angelis, 2010).
The production of FMP was mainly anaerobic fermentation, such as alcohol and lactic acid. As shown in Figure 5a-d and Table 2, most of volatile compound content decreased with prolonging of fermentation time, only several volatile compounds increased.
The reason for this phenomenon was that the growth of many microorganisms in FMP was inhibited by high concentration of salt, low pH, and oxygen-deficient environment, so reduced the development of many volatile compounds. However, LAB posed a strong ability to resist adverse environment and produced volatile compounds by fermentation (homo-and heterofermentative LAB fermentation), especially for organic acids and alcohols (Esteban-Torres et al., 2015;Yu et al., 2013). The followed analysis was used to further explore the specific change in various types of volatile compounds.
Esters, closely related to the fruity and sweet flavor, played an important role in FMP flavor. The change of eight esters during fermentation process is shown in Figure 5a and Table 2. Most of esters decreased with the increasing in fermentation time, including hexyl hexanoate, ethyl nonanoate, methyl octanoate, isoamyl hexanoate, and gamma-butyrolactone-M. High-acid environment might promote the hydrolysis of above esters (Tomita et al., 2018).
In the early fermentation, isoamyl hexanoate-M sharply deceased, but the isoamyl hexanoate-D sharply increased. This phenomenon indicated that high content of isoamyl hexanoate-M turned into the isoamyl hexanoate-D, which was in accordance with the result of fingerprint analysis. In addition, isoamyl hexanoate posed an acid stability and was not easy to hydrolyze at low pH. Hence, isoamyl hexanoate content was the highest in all the eaters during fermentation process.  found that ethyl salicylate, ethyl palmitate, methyl linoleate, ethyl linoleate, and ethyl stearate were the main esters in FMP during fermentation process, but no many isoamyl hexanoate was produced. The reason for the difference in two experiments was in accordance with fermentation conditions, salt content, and pepper varieties. According to these results, it was speculated that isoamyl hexanoate played an important role in the aroma quality of FMP. Ethyl acetate, with pineapple flavor, showed an increased tendency during fermentation process. Microorganisms turned the fermented carbohydrates into the end products during fermentation process, including organic acids and alcohols (Kim et al., 2016). The organic acids and alcohols could combine each other to form several esters, such as ethyl acetate . It was found that more ethyl acetate was produced during liquor fermentation, and became a key volatile compound of liquor during liquor brewing process (Cai et al., 2019   the pungency of FMP at some extent. Zhao et al. (2015) found that almost all aldehydes, such as butanal, octanal, and others, decreased with the prolonging of pumpkin juice fermentation time.
For 2-methylbutanal, the content increased in the first 6 days, then decreased in the late 18 days. Li, Dong, et al. (2016) explored the relationship between bacterial communities and volatile compounds in red pepper paste, and found that Pseudomonas posed an obvious correlation with 2-methylbutanal. Hence, it was speculated that the change in 2-methylbutanal content might be related to Pseudomonas during fermentation process. Hazelwood, Daran, van Maris, Pronk, and Dickinson (2008) found that some aldehydes, such as methylbutanal, could be generated by Strecker degradation of methionine and leucine or the Ehrlich pathway.
Although octanal content slightly increased during fermentation process, the content was still lower than most other aldehydes.
At the end of fermentation, the content of all aldehydes was low, which weakened the pungency of FMP.
Alcohols, with high threshold values, were mainly generated by alcohol and lactic acid fermentation . However, the threshold of unsaturated aldehyde was low, which made many contributions to food flavor (Lorenzo, Carballo, & Franco, 2013). As shown in Figure 5c and Table 2  Note: The significance analysis results were based on the peak volume value of FMP during NF and IF process, and different letters indicated significant differences (p < .05). McFeeters (1998) found that the content of linalool in fermented cucumber increased to several times than its odor threshold during fermentation.
The content of 2-methylpropionic acid and 2-methylbutanoic acid increased after fermentation. In particular for 2-methylpropionic acid, the content sharply increased to a high level. Whatever NF or IF, lactic acid fermentation was the main fermentation pathway, and involved the oxidation of carbohydrates to other compounds, which caused the production of lots of acids. Li et al. (2020)  were about 2 times than those of the initial ones. It indicated that some furans were generated during fermentation process. Furans were hardly detected in many fermented foods (Kim et al., 2019). It is indicated that 2-ethylfuran, 2-pentyfuran, and 2-acetylfuran were probably the characteristic volatile compounds in FMP. Only 1 pyrazine and ether were detected. Hence, it was difficult to analyze the change in pyrazine and ether during fermentation process.
As shown in Figure 5a 1 -d 1 , it was found that there was similar content in volatile compounds between NF and IF samples. Under suitable fermentation conditions, including temperature, pH, O 2 , the LAB fermentation was the prominent pathway whatever NF or IF (Sanlier et al., 2017). Many other microorganisms were inhibited by LAB. Hence, two fermentation methods produced the similar volatile compounds, but IF might shorten fermentation period and promote FMP quality. It was proved that the use of inoculated microorganism starters guaranteed the agreeable sensory properties, including volatile compounds (Woutets et al., 2013;Zhao et al., 2015).
In summary, main volatile compounds, including esters, aldehydes, alcohols, and acids, notably changed during fermentation.
Almost all esters, aldehydes, and some alcohols posed an obvious decrease, and some alcohols and all acids posed an obvious increase during fermentation process. These results were related to microbial

| Analysis based on PCA results
Principal component analysis, a multivariate statistical analysis, was used to turn original correlated variables into linearly uncorrelated variables by several related transformation. These uncorrelated variables obtained by PCA could reflect the relationships among the observed variables (Cirlini et al., 2019;Yao et al., 2015). The correlated variables could be distinguished by the positive or negative score values in PC1 and PC2. Yilmaztekin and Sislioglu (2015) found that most important volatile compounds in fermented and raw European cranberrybush were divided into four uncorrelated parts by PCA, and PCA discriminated the fermentation stage as three groups. In our research, PCA obtained by original date was used to compare the difference in principal compounds ( Figure 6). An obvious separation of two principal compounds was found during fermentation process and two fermentation methods (NF and IF). Two principal compounds (PC1 and PC2) were up to 87% in variation of originate date. PC1 was up to 74%, but PC2 was only up to 13% in variation of originate date.
As shown in Figure 6, the samples showed a well separation degree each other, and three parts were presented in PCA based on previous analysis and heat map.

| Clustering analysis based on heat map
To further analyze the differences in volatile compounds of FMP during fermentation process, the cluster analysis was obtained based on the heat map, as shown in Figure 7. In the heat map, the samples were clustered in the horizon, and volatile compounds were clustered in the vertical. There was the high correlation degree in

| CON CLUS ION
The

CO N FLI C T O F I NTE R E S T
The authors declare that they do not have any conflict of interest.

H U M A N A N D A N I M A L R I G HT S
No animals or humans were used for studies that are the basis of this manuscript.