Analysis of Caffeine, Chlorogenic Acid, Trigonelline, and Volatile Compounds in Cold Brew Coffee Using High-Performance Liquid Chromatography and Solid-Phase Microextraction—Gas Chromatography-Mass Spectrometry

This study investigated the non-volatile and volatile compounds in samples of cold brew (CB) coffee, coffee from a coffee shop (CS), ready-to-drink (RTD) coffee, and brewed coffee from a coffee maker (CM). The volatile compounds were identified using headspace solid-phase microextraction with gas chromatography-mass spectrometry, and the samples were treated with high-performance liquid chromatography for the quantification of caffeine, chlorogenic acid, and trigonelline. The results indicate that RTD coffee had the lowest amounts of non-volatile compounds. A total of 36 volatile compounds were semi-quantified; the contents of most volatile compounds in CS and Folgers samples were higher than those in CB and CM samples. The contents of 25 volatile compounds in the CM sample were higher than those in the CB sample. The consumer and instrumental data show that the bitterness intensity was correlated with pyrazines, pyrroles, and guaiacols, whereas the coffeeID intensity was correlated with phenols. Semi-quantification and principal component analysis results show that the extraction method and temperature could influence the volatile compound profiles.


Introduction
Coffee is one of the most complex beverages because it contains many non-volatile and volatile compounds generated during the brewing process. Non-volatile components of coffee are composed of carbohydrates, proteins, lipids, unknown melanoidins, minerals, and alkaloids such as caffeine, trigonelline, and chlorogenic acid [1,2]. Non-volatile compounds degrade when they undergo thermal processing involving roasting. Chlorogenic acid and trigonelline are rapidly degraded during roasting, and phenolic compounds and pyridines/pyrroles are produced, respectively [3]. These newly formed compounds, along with other volatile compounds, influence the coffee quality or flavor.
Recently, considerable research has been conducted on coffee volatiles, especially the volatile compounds in hot coffee from an espresso machine [4] and coffee that undergoes different degrees of roasting at varying temperature and time [5], using headspace solid-phase microextraction (HS-SPME) French  Ground coffee was first placed in a glass bottle (standard wide-mouth bottles with polyvinyl-lined caps, Fisher Scientific, Pittsburgh, PA, USA) for cold brewing, and bottled water (Samdasoo 2L, Kwang Dong Pharmaceutical. Co., Seoul, Korea) was poured over it. After 9 h of brewing at 4 • C, coffee was filtered through a stainless filter (stainless filter, 4 cups, Kinto, Hikone-shi, Shiga, Japan) and re-filtered through a paper filter (Mellitta paper filter 1 × 4, Mellitta, Minden, North Rhine-Westphalia, Germany) to remove the fine particles in it. The filtered brewed coffee was kept frozen until analysis.
A coffee maker (Phillips HD-7564, Amsterdam, The Netherlands) was used to make brewed coffee (CM); the coffee was brewed for 10 min, and a 1 × 4 paper filter was used for filtering (Mellitta paper filter, Mellitta, Minden, North Rhine-Westphalia, Germany). The CM brewed samples were cooled to room temperature (20 • C) and then kept frozen until the analysis.

Ready-to-Drink (RTD) Coffee
RTD coffee samples-French cafe CB coffee (Namyang, Seoul, Korea) and Barista Rules CB coffee (Maeil, Seoul, Korea)-were purchased from a convenience store (GS retail, Busan, Korea). The CB coffee was delivered by Babinski coffee (Korea Yacult, Seoul, Korea). The samples were kept frozen until the instrumental analyses, and the sample information is provided in Table 1.

Coffee from a Coffee Shop (CS)
CB coffee was purchased from nearby local Starbucks (Seattle, WA, USA) and Twosomeplace (CJ Foodville, Seoul, Korea) ( Table 1). Both CB coffee samples were purchased without ice to prevent the dilution and kept frozen until the instrumental analyses. Hitachi Koki Co. Ltd., Tokyo, Japan). The supernatant was then filtered through a 0.45 µm membrane filter (Acrodisc Syringe Filter, Gelman Sciences, Ann Arbor, MI, USA). The column for caffeine analysis was a XBridge™ Shield RP18 (5 µm, 4.6 mm × 250 mm, Waters, Milford, MA, USA), mobile phase A was 10 mM citric acid, and mobile phase B was methanol (HPLC-grade, Merck, Damstadt, Frankfurter, Germany). The gradient mode was initially set at an A/B ratio of 85/15 from 0 to 10 min, which linearly increased to 60/40 at the 10 min mark and ended at the 30 min mark. Of each sample, 20 mL were injected, and the flow rate was adjusted to 1.0 mL/min. A UV detector (Shimadzu SPD-20A, Kyoto, Japan) was set at 276 nm. The concentration of caffeine was calculated using the peak area of a commercial standard (Sigma-Aldrich Co., St. Louis, MO, USA).

Analysis of Caffeine, Chlorogenic Acid, and Trigonelline
For chlorogenic-acid analysis, the coffee samples were preprocessed and treated in the same way as caffeine, but the detector was set at 325 nm. The concentration of chlorogenic acid was calculated using the peak area of a commercial standard (5-caffeoylquinic acid, Sigma-Aldrich Co., St. Louis, MO, USA).
Each 10 mL coffee sample was clarified with 0.3 mL of Carrezz reagents I and II and 0.6 mL of absolute methanol (HPLC-grade, Merck, Damstadt, Frankfurter, Germany) for trigonelline analysis. After the treatment, the same HPLC process for caffeine and chlorogenic-acid analyses was performed with the following changes. As a mobile phase, 0.5% methanol (HPLC-grade, Merck, Damstadt, Frankfurter, Germany) was used, and a reverse-phase column (XBridge™ Shield RP18; 5 µm, 4.6 mm × 150 mm, Waters, Milford, MA, USA) was also used. The flow rate was 0.8 mL/min and detected at 264 nm. The concentration of trigonelline was calculated using the peak area of a commercial standard (Sigma-Aldrich Co., St. Louis, MO, USA). Each non-volatile compound was analyzed for 5 replications.

Headspace Solid-Phase Microextraction (HS-SPME)
In a 10 mL vial, 2 mL coffee samples were placed with a silicon/PTFE septum (18 mm (diameter) × 3.2 mm (thickness); Varian, Inc., Walnut Creek, CA, USA). Five microliters of 0.045 mg/mL 1,3-dichlorobezene (Sigma-Aldrich, St. Louis, MO, USA) solution (in methanol) was added to the sample as an internal standard for semi-quantification of the volatile compounds in the coffee samples. These coffee-sample vials were equilibrated for 10 min at 60 • C in an autosampler (Model GC Sampler 80, Agilent Technologies, Santa Clara, CA, USA) with agitation at 250 rpm. Next, a 50/30 µm 3-phase (DVB/CAR/PDMS) SPME fiber (Supelco, Bellefont, PA, USA) was used to collect the volatile compounds of the coffee samples at 60 • C for 45 min.

Gas Chromatography-Mass Spectrometry (GC-MS)
After sampling, the fiber was inserted into the injection port of the GC-MS (Model 7890A, Agilent Technologies, Santa Clara, CA, USA) connected to a spectrophotometer detector (Model 5977A, Agilent Technologies, Santa Clara, CA, USA), and HP-5 ms column (30 m × 250 µm × 0.25 µm thickness, Agilent Technologies, Santa Clara, CA, USA) was used to separate the volatile compounds. A split-less mode was applied at 250 • C for 5 min. The initial temperature of the column was 40 • C for 10 min, increased by 8 • C/min to 180 • C and then by 10 • C/min to 280 • C, and it was maintained at 280 • C for 10 min. The volatile compounds in each sample were analyzed in triplicate. Each volatile compound was quantified as the mean concentration (ng/mL).
Most compounds were identified and compared using 2 different analytical methods: Kovats indices and mass spectra comparison using the National Institute of Standards and Technology (NIST) library of compounds (NIST/EPA/NIH mass spectral library, version 2.2, 2014). Semi-quantification of the volatile compound concentrations in each coffee sample was calculated and reported based on the internal standard concentration.

Consumer Evaluation
Including 1 blind duplicate sample (BaristaRTD (A) and (B)), 14 samples were provided to the participants. CB and CM samples were extracted using 7 g of ground coffee with 100 mL of water, and 50 mL of each sample was provided in a plastic cup. Unsalted crackers and water were also provided for cleansing the palate. The evaluations were held over 2 sessions, and 120 consumers participated in the evaluations; the sensory liking and intensity of coffee samples were investigated. This consumer study was approved by the institutional review board of Pusan National University (IRB 2016_49_HR). Details of the consumer data analyses were available [14], and only the consumers' acceptability data were used in this study to analyze their relationship with the constituents of coffee.

Statistical Analysis
Two-way analysis of variance (two-way ANOVA) was conducted to investigate the effects of the brewing methods, CB and CM, and coffee variety (brand), and their interaction, variety (brand) × brewing method. In addition, a one-way analysis of variance (one-way ANOVA) was conducted on RTD commercial coffee products, CS samples, and 8 samples of different brewing methods and coffee variety (brand). The significant difference (LSD) was also calculated post hoc. Pearson's correlation analysis was conducted to explore the relationship between volatile compounds and the consumers' perceived sensory intensity and overall liking. All these analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA). To understand the relationship between coffee samples and volatile compounds, principal component analysis (PCA) was used. Internal preference mapping was also used to examine the relationship between coffee samples and components based on the consumers' liking using XLSTAT 19 (Addinsoft, New York, NY, USA).

Caffeine, Chlorogenic Acid, and Trigonelline Compounds in CB Coffee Samples
Alkaloids, such as caffeine, chlorogenic acids, and trigonelline, are crucial to the coffee flavor [1]. Bitterness was one of the main sensory attributes influencing consumer acceptability [15]. Both caffeine and chlorogenic acid contributed to the bitterness, whereas the latter was also responsible for astringency [1,16].
p-values of caffeine, chlorogenic acid, and trigonelline contents in the CB and CM samples are listed in Table 2. The coffee variety (brand) and brewing method significantly affected these contents. The caffeine content was affected by both the coffee variety (brand) and brewing method, whereas the chlorogenic acid and trigonelline contents were significantly affected by the coffee variety. The amounts of the three non-volatile compounds are shown in Table 3. Generally, the RTD samples had the lowest amounts of these compounds. The caffeine contents descended in the following order: Those of FolgersCM, StarbucksCS, FolgersCB, and TwosomeplaceCS. The contents of chlorogenic acid and trigonelline were the highest in the StarbucksCS and FolgersCM samples. The CB and CM samples, except the Folgers samples, generally showed moderate contents: 0.74-0.8 mg/mL (caffeine), 0.18-0.23 mg/mL (chlorogenic acid), and 0.19-0.27 mg/mL (trigonelline) ( Table 3). Kim and Kim [17] studied the contents of caffeine, chlorogenic acid, and trigonelline using different extraction methods. They used Kenya AA coffee samples and steeped the coffee grounds for 9 h to obtain CB coffee. The contents of caffeine, chlorogenic acid, and trigonelline were 0.69 mg/mL, 0.26 mg/mL, and 0.41 mg/mL, respectively. Although they used 50 g of ground coffee from beans and 500 mL of water to obtain the coffee samples, the contents of non-volatile compounds (except trigonelline) were similar to those in this study. In addition, in Blending, Colombia, and Kenya coffee samples, BlendingCB and BlendingCM contained slightly lower amounts of chlorogenic acid. Fuller and Rao [18] analyzed the caffeine and 3-chlorogenic-acid (3-CGA) contents in coffee prepared from grounds with different degrees of roasting and granule sizes and reported that the roasting temperature could affect the 3-CGA content. However, there was no significant difference between cold and hot brewed coffee [18]. Given that the blending samples underwent a stronger roasting level (city roasting) than other samples (generally medium roasting), a higher roasting temperature might influence the degradation of chlorogenic acid, leading to a reduction in its amount in the final coffee samples. The loss in caffeine content due to general roasting is very low unless the coffee bean undergoes extreme-dark roasting [3]. The caffeine contents of CM samples (Colombia, Folgers) were higher than those of the CB samples, showing that the caffeine contents in these coffee samples were influenced by the brewing method (CB and CM). Considering that the caffeine content was higher in the 22 • C coffee samples than in the 5 • C samples [13], irrespective of the extraction method (brewing and dripping), this difference in caffeine content between the CB and CM samples was considered to be due to the extraction temperature. Furthermore, grinding and brewing methods can influence the caffeine content [19]. Thus, the low caffeine content in RTD samples in this study may be explained by the different grinding levels and brewing times of different manufacturing companies.
Blended CB and CM samples underwent a higher intensity of roasting than that for the Colombia and Kenya samples, which may explain the lower chlorogenic acid contents in the CM samples. This result agrees with those of other studies (e.g., Farah et al. [20] and Kim and Park [3]). According to their studies, chlorogenic acid can be degraded (>90%) by thermal processing such as dark roasting, suggesting that the roasting time and temperature influence the chlorogenic-acid content in coffee products [21,22].
Trigonelline contributes to the overall aroma perception of brewed coffee [23]. Usually, it has a very weak bitter taste, but it degrades rapidly during the roasting process and produces impactful volatile compounds such as pyridines and pyrroles [3]. The lower chlorogenic-acid and trigonelline Foods 2020, 9, 1746 7 of 20 contents in RTD samples can possibly be ascribed to several thermal processes during manufacturing because both chlorogenic and trigonelline were vulnerable to thermal processes such as roasting. Degradation during processing could also be considered a reason for the reduced amount of chlorogenic acid in the RTD samples. Additionally, according to a study conducted by Andueze et al. [24], caffeine, chlorogenic acid, and trigonelline contents increased with a decrease in particle size of the espresso coffee samples. Furthermore, when ground coffee was exposed to heat or high temperature, some flavor loss occurred [25].

Volatile Compounds in CB Coffee Samples
A total of 36 volatile compounds, including phenolic compounds such as guaiacol, pyrazines, pyridines, and pyrroles, were identified ( Table 4). The relative standard deviation (RSD, %) for each coffee sample and threshold of each volatile compound are listed in Tables 5 and 6, respectively. Pyridine (V1) is known to have bitter, burnt, roasted, and astringent characteristics [26,27], and it can give a sharp burnt taste at concentrations as low as those on the ppm scale [28]. In this aspect, only StarbucksCS (1229 ng/mL) and TwosomeplaceCS (1695 ng/mL) might have a burnt taste. Pyridine is produced by trigonelline degradation and Maillard reactions [29], especially in coffee that is subjected to high temperatures or strong roasting processes [5]. According to Bhumiratana et al. [10], the degree of roasting and preparation process contributed to aroma profiles rather than the coffee variety. Corresponding to this research, Blending CB and CM samples showed relatively higher contents of pyridine (694 and 643 ng/mL) than those in Colombia and Kenya samples of CB (521 and 508 ng/mL) and CM (268 and 309 ng/mL), respectively. As indicated by the sample sellers, the blending sample was roasted to the city level (about medium~dark roasting), and Colombia and Kenya beans were roasted to a medium level. Commercially ground coffee, Folgers, had a higher pyridine content, although it was indicated as a medium-roasted, and BaristaRTD had a higher pyridine content than those of other RTD samples. The CS samples had the highest amount of pyridine among all the evaluated coffee samples. In the case of 3-ethylpyridine (V6), which was correlated with a roasted, burnt aroma [27] and tobacco flavor [30], was found in TwosomeplaceCS and BaristaRTD at higher concentrations. N-acetyl-4-(H)-pyridine (V12) showed higher concentrations in CS and Folgers samples. The threshold of furfural (V3) was 3 ppm [31], which was significantly higher than the furfural contents of all the coffee samples in this study.
Maltol (V23) is associated with caramel-like odor and chocolate flavor [30]. When in solution, it exhibits fruity characteristics, especially when the concentration is below 20 ppm [28]. StarbucksCS had a higher amount of maltol, followed by TwosomeplaceCS, ColombiaCM, and KenyaCM showed lower contents than those of ColombiaCB and KenyaCB. These volatile compound patterns suggested that the Colombia and Kenya samples were more influenced by the extraction method than the origin of coffee. Fisk et al. [33] compared the contents of aroma compounds in brewed coffee. Among the key aroma compounds that they reported, guaiacol (V20) and maltol (V23) were identified in our coffee samples. In their study, guaiacol (V20) and maltol (V23) contents detected by LLE-GC-MS were higher in the Kenya brewed coffee sample than in the Colombia brewed coffee sample. Similarly, Kenya samples had higher amounts of guaiacol (V20) and maltol (V23) than those in Colombia samples, regardless of the extraction method (CB and CM) in this study. This result might be due to the difference in coffee bean components that affect volatile compounds rather than the difference in extraction methods.
Nonanal (V22) is related to citrus, fat, and green attributes [32]. KenyaCB (2 ng/mL), ColombiaCB and CM (0 and 3 ng/mL, respectively), FrenchCafeRTD (3 ng/mL), and BabinskiRTD (2 ng/mL) are more likely to have a refreshing, citrusy, and waxy taste because n-nonanal showed these characteristics when it was below a concentration of 5 ppm [28]. Nonanoic acid (V28) is also related to green, fatty, and bitterness, and this compound was found in higher concentrations in FolgersCM and StarbucksCS. Decanoic acid (V32), which had rancid fatty notes [28] were present in slightly higher amounts in FolgerCM and StarbucksCS samples. Although there were no peppermint-like attributes, ColombiaCB and FolgersCB showed slightly higher amounts of methyl salicylate (V27), which is associated with peppermint characteristics [32].
The results of volatile compounds showed that CS and Folgers samples had higher amounts of volatiles than those in RTD samples. Specifically, RTD samples, except BaristaRTD, generally had lower contents of most volatile compounds, which corresponded with non-volatile compounds, showing the lowest contents in RTD samples. Kumazawa and Masuda [34] suggested that heat treatment for sterilization could affect the sensory aspect of canned coffee. The results of sensory evaluation showed significant differences between heated coffee and non-heated coffee samples. Furthermore, after heat treatment, 3-methylbutanoic acid, and another "unknown" compound related to sour odor were identified by gas chromatography-olfactometry (GC-O). Although coffee volatiles that were affected by heat processing might vary according to coffee variety or specific brewing preparation, some volatile compounds of RTD could change. Amanpour and Selli [35] compared the concentrations of volatile compounds in Turkish coffee and French press coffee. The results showed that volatile profiles of both coffee samples were similar, but there were still a few differences in the volatile compounds and concentrations. This research suggests that the brewing method could influence volatile profiles. In a study of volatile compounds in brewed green teas [36], the authors mentioned that there was a possibility of temperature influencing the threshold of volatile compounds.

Consumers' Perceptions of Coffee Samples
p-Values for the CB and CM samples are listed in Table 7. Table 7 and results of mean intensity of consumers' overall liking, bitterness intensity, and coffeeID intensity (Table 8) showed that the variety × brewing method effect and variety (brand) effect were significant in all categories (overall liking, bitterness intensity, and coffeeID intensity). FolgersCB was higher than FolgersCM in overall liking. Coffee variety (brand) significantly affected coffeeID intensity more than the brewing method. Table 7. p-Values of coffee variety (brand) and brewing method effects for consumers' overall liking, bitterness intensity, and coffeeID intensity of CB and CM samples.

p-Value
Variety (Brand) 1 Brewing Method

Relationship between Coffee Acceptability and Volatile Compounds
Volatile compounds in the coffee samples were investigated by PCA (Figure 1a,b). StarbucksCS, TwosomeplaceCS, and Folgers samples were positioned in quadrants 1 or 4, and most volatile compounds were also placed close to CS and Folgers samples. The CB and CM samples were positioned close together by extraction method rather than coffee variety. ColombiaCB and KenyaCB were related to 3,5-octadien-2-one (V15), 5-methyl-2-furaldehyde (V7), furfural (V3), and tetradecanoic acid (V36). ColombiaCM and KenyaCM were not related to any particular volatile compounds. This suggests that different extractions could influence volatile compound profiles markedly. However, blended coffees, BlendingCB and CM, were placed close to each other, showing that they were not different regardless of the extraction method. RTD samples were arranged in quadrant 3, and only BaristaRTD samples were slightly related to 3-methylphenol (V13) and nonanal (V22). The overall result of the confidence ellipse explained that sample discrimination by the brewing method was in the Colombia and Kenya samples. Furthermore, CS samples and Folgers samples were clearly discriminated between Colombia, Kenya, and RTD samples by volatile compounds. that they were not different regardless of the extraction method. RTD samples were arranged in quadrant 3, and only BaristaRTD samples were slightly related to 3-methylphenol (V13) and nonanal (V22). The overall result of the confidence ellipse explained that sample discrimination by the brewing method was in the Colombia and Kenya samples. Furthermore, CS samples and Folgers samples were clearly discriminated between Colombia, Kenya, and RTD samples by volatile compounds.  Table 4.  (Table 9). This result seemed to be due to the absence of strong flavors induced by volatile compounds, affecting the overall liking of consumers. Bitterness intensity was correlated with many volatile compounds such as pyrazines (methylpyrazine, V2, r = 0.81; 2,6-dimethylpyrazine, V5, r = 0.80; 2-ethyl-3-methylpyrazine, V9, r =  Table 4. 3.5. Relationship between Sensory Data, Volatiles, and Caffeine, Chlorogenic Acid, and Trigonelline Compounds in Coffee

Internal Preference Mapping
Coffee quality and its volatile compounds can be greatly influenced by coffee species, growing altitude, presence or absence of defects, postharvest processing such as washing and drying, roasting, and even storage conditions [72]. Moreover, synergistic effects among volatile compounds should also be considered when exploring volatile compound profiles. For example, maltol can be enhanced by vanillin and heliotropine [28]. Meanwhile, it appeared that the synergistic effect of volatile compounds existed according to the study of Ito and Kubota [73]. 4-Hexanolide was added to jasmine tea samples, and even though it was under the threshold level, it had an impact on sweet and astringent odor perception in the descriptive analysis. Thus, further studies could be conducted linking volatile compounds and their synergistic impact on consumers or a descriptive panel's perception of coffee characteristics.  2. Internal preference map using consumers' liking, coffee volatile and caffeine, chlorogenic acid, trigonelline, and coffee samples. V1-V36 stands for each volatile compound numbering in Table 4.

Conclusions
This study revealed the caffeine, chlorogenic acid, trigonelline, and volatile compound profiles of CB coffee. Typically, the amounts of both non-volatile and volatile compounds in CS and Folgers samples were higher than those in the RTD and CB samples. A comparison between CB and CM samples indicated that some volatiles were influenced by the brewing method. A comparison between the CB and CM samples did not reveal any consistent patterns of caffeine, chlorogenic acid, or trigonelline compounds. However, the caffeine content was higher in the CM samples than in the CB samples owing to the different brewing methods and extraction temperatures. The results of the correlation analysis suggested that particular pyrazines, pyrroles, and guaiacols were related to the bitterness intensity, and phenols were correlated with the coffeeID intensity. Overall, the RTD samples showed little correlation with volatile compounds and caffeine, chlorogenic acid, and trigonelline contents, whereas the CS and some CM samples (Colombia and Folgers) showed correlations with the volatile compounds and caffeine, chlorogenic acid, and trigonelline contents. Further studies on more RTD and CS samples should be conducted to gain a comprehensive understanding of CB coffee.
Author Contributions: Conducting sample preparation and writing the manuscript, J.H.; analyzing the caffeine, chlorogenic acid, and trigonelline compounds, K.S.C.; analyzing volatile compounds and revising the manuscript, K.A.; supervising and revising the manuscript, J.L. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.