Effect of Roasting Degree on Major Coffee Compounds: A Comparative Study between Coffee Beans with and without Supercritical CO 2 Decaffeination Treatment

between coffee consumption and hepatocellular carcinoma mortality in a cohort comprising 46,399 males and 64,289 females aged 40–79 years and showed that the hazard ratio of death caused by hepatocellular carcinoma for drinkers of ≥ 1 cup of coffee per day was 0.50 （ 95 ％ confidence in-terval 0.31–0.79 ） compared with non-drinkers 3 ） . These health benefits might be attributable to caffeic acid, chlorogenic acid, trigonelline, and other phenolic compounds Abstract: Coffee is a beverage that is consumed worldwide, and the demand for decaffeinated coffee has increased in recent years. This study aimed to investigate the effect of roasting conditions on the concentration of physiologically active compounds in coffee beans with and without supercritical CO 2 decaffeination treatment. Decaffeination treatment markedly reduced caffeine concentration and slightly reduced trigonelline concentration in the coffee beans, whereas the concentrations of chlorogenic acids (chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid) were largely unchanged. Roasting was performed using a hot-air coffee roaster machine and the coffee beans were treated at different peak temperatures (125–250℃), different hold times at the peak temperature (120–240 s), and different temperature increase times to reach the peak temperature (60–180 s). Roasting conditions such as long hold and long temperature rise times at high temperatures (≥ 225℃) significantly degraded coffee compounds except for caffeine, with similar degradation rates between non-decaffeinated and decaffeinated coffee beans. In contrast, the L * value of decaffeinated coffee decreased with less thermal history compared to that of non-decaffeinated coffee. This allowed for the complete roasting of decaffeinated coffee with a lower thermal history compared to those of non-decaffeinated counterparts, suppressing the degradation of several coffee compounds. For example, comparing the similar L * values between coffee beans with and without decaffeination treatment, it was found that the former tended to contain more chlorogenic acid. Generally, decaffeination results in the loss of physiologically active compounds along with caffeine, which is a major concern. However, this study showed that appropriate control of decaffeination and roasting conditions can limit the degradation of several valuable coffee compounds, such as trigonelline and chlorogenic acid.

2 specific to coffee 7 9 . Moreover, recently, trigonelline has received considerable attention owing to its memory improvement effects in patients with Alzheimer s disease 10,11 . Thus, the consumption of 2-4 cups of coffee per day has been recommended 12 14 . However, current evidence suggests that some population subgroups sensitive to the effects of alkaloid caffeine in coffee e.g., pregnant women and people with hypertension should avoid caffeinated drink consumption 12,15 . This alkaloid stimulates the central nervous system and affects the cardiovascular system by increasing heart output and blood pressure 16,17 . Therefore, demand for decaffeinated coffee has been increasing, and many studies have reported that decaffeinated coffee with proper decaffeination treatment can impart health benefits similar to those imparted by non-decaffeinated coffee 18 21 . Solvent extraction with organic solvents, water, or supercritical CO 2 is a commonly used decaffeination method. Ethyl acetate and dichloromethane are widely used organic solvents for decaffeination and can inexpensively produce high-quality decaffeinated beans. However, these organic solvents are highly toxic, and several physiologically active compounds, including chlorogenic acid and trigonelline, are lost during extraction 22 24 . While decaffeinating with hot water at atmospheric pressure is a very safe and efficient method, the limited selectivity with regard to other coffee compounds is a serious problem, as aroma precursor and physiologically active compounds are partially co-extracted 22 24 . Supercritical CO 2 is costly because of the required equipment and maintenance associated with the unique high-pressure technology. However, supercritical CO 2 is an ideal solvent for decaffeination because it is nontoxic, non-flammable, non-polluting, and exhibits superior performance for selective caffeine extraction 22,23,25 .
It is well known that coffee beans lose many physiologically active compounds during roasting 26 28 . For example, Trugo et al. investigated the effect of roasting degree on chlorogenic acid content in coffee beans and reported a loss of approximately 60 under medium roasting conditions 28 . Therefore, investigating the effect of roasting degree on physiologically active compounds in coffee beans is important to preserve the health benefits of coffee consumption; however, reports on the effects of decaffeinated coffee beans are limited. This study aimed to investigate the effects of roasting conditions on the degradation behavior of physiologically active compounds, namely, caffeine, trigonelline, neochlorogenic acid 3-O-caffeoylquinic acid , cryptochlorogenic acid 4-O-caffeoylquinic acid , and chlorogenic acid Fig. 1 , in supercritical CO 2 decaffeinated coffee beans and compare the behavior of these compounds with non-decaffeinated coffee beans.

Effect of Roasting on Coffee Compounds
J. Oleo Sci. 3

Materials
Arabica coffee Coffea arabica L. beans with and without decaffeination were procured from Super Critical Technology Centre Co. Ltd. Mie, Japan . Coffee bean decaffeination was performed via pilot-scale 30 L supercritical CO 2 extraction with water co-solvent , as described previously 29,30 . High-performance liquid chromatography HPLC -grade methanol was purchased from Fujifilm Wako Pure Chemical Corp. Osaka, Japan . Caffeine, caffeic acid, chlorogenic acid, nicotinic acid, theophylline, and phosphoric acid were purchased from Wako Pure Chemical Corp. Osaka, Japan . Cryptochlorogenic acid and trigonelline were procured from Toronto Research Chemicals Toronto, ON, Canada . Neochlorogenic acid and sodium 1-octanesulfonate were procured from Tokyo Chemical Industry Co. Ltd. Tokyo, Japan .

Coffee bean roasting
Roasting of coffee beans with or without decaffeination treatment was performed using a hot-air coffee roaster machine The Roast; Panasonic Corporation, Osaka, Japan . The effects of peak temperature 125-250 , hold time at peak temperature 120-240 s , and temperature rise time to reach the peak temperature 60-180 s on the thermal degradation of coffee compounds were investigated it should be noted that the above temperatures were measured inside the roasting chamber . The roasting conditions are summarized in Table 1, and the temperature profiles inside the roasting chamber are shown in Figs. S1-S3. The chamber was preheated to 100 before roasting, and air was blown into the chamber for rapid quenching after roasting was completed. The degree of coffee bean roasting was evaluated using an L* value with a color meter ZE 6000; Nippon Denshoku Industries Co., Ltd., Tokyo, Japan . Before L* value analysis, the coffee beans were ground using a coffee grinder KCG17; Kalita Co., Ltd., Kanagawa, Japan in the finest ground mode particle size: approximately 0.5 mm . The cross-sections of coffee beans were imaged using scanning electron microscopy SEM, Hitachi S-4800; Hitachi Ltd., Tokyo, Japan . After roasting, the coffee beans were stored at 4 until just before the extraction operation described in Section 2.3.

Extraction of physiologically active compounds from
coffee beans Coffee compounds were extracted from coffee beans using 15 methanol following a previously described procedure 31 . Briefly, the coffee beans were crushed using a food processor IFM-C20G; Iwatani Corporation, Osaka, Japan for 30 s before extraction. Approximately 50 mg of each sample was then weighed into a 50-mL screw-capped glass bottle, and 30 mL of 15 methanol was added. The coffee compounds were extracted via ultrasonic treatment CPX1800H-J; Yamato Scientific Co., Ltd., Tokyo, Japan at 80 W and 38 kHz for 30 min. The resulting solution was filtered through a 0.22-μm PTFE membrane filter Osaka Chemical Co., Ltd., Osaka, Japan and analyzed by reversed-phase HPLC, as described in Section 2.4.

HPLC analysis
Physiologically active compounds in the coffee bean extract were analyzed by reversed-phase HPLC equipped with a photodiode array detector SPD-M20A; Shimadzu Corp., Kyoto, Japan according to a previously reported method 31 . Briefly, an InertSustain C18 column 150 mm 4.6 mm, 3 μm, GL Sciences Inc., Tokyo, Japan was used as the stationary phase. A mixture of water/methanol 85:15, v/v containing 0.1 phosphoric acid and 4 mM sodium 1-octanesulfonate was used as the mobile phase. The mobile phase flow rate and column temperature were adjusted to 1.0 mL/min and 35 , respectively. The quantification of the compounds in the coffee beans was performed via peak area integration at 220 nm. The coffee compounds were identified by comparing HPLC retention times and spectral data spectral shape and absorption maxima with those of the corresponding standards. Herein, eight characteristic coffee compounds, namely, caffeine, caffeic acid, chlorogenic acid, cryptochlorogenic acid, neochlorogenic acid, nicotinic acid, theophylline, and trigonelline 2, 8, 26 28, 31 , were analyzed. The peak temperature was measured inside the roasting chamber. 4

General pro le of coffee compounds
Typical HPLC profiles of the standard mixture and coffee bean extract are shown in Fig. 2. In this HPLC system, eight physiologically active coffee compounds caffeine, caffeic acid, chlorogenic acid, cryptochlorogenic acid, neochlorogenic acid, nicotinic acid, theophylline, and trigonelline were clearly separated, as reported by Arai et al. 31 . The coffee beans used herein contained high amounts of caffeine, trigonelline, chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid Table 2 . In contrast, caffeic acid, nicotinic acid, and theophylline were not detected at any time point, even after decaffeination and roasting treatments. Hence, the effects of roasting conditions on the concentrations of caffeine, trigonelline, chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid, which were detected consistently in the coffee beans used herein, were evaluated.
Supercritical CO 2 decaffeination treatment with water was found to efficiently remove caffeine, as its concentration decreased by 91.5 after decaffeination. Although trigonelline concentrations decreased with decaffeination, they did so to a less drastic extent at 13.8 . In contrast, chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid increased slightly after decaffeination. This can be explained by the coffee bean mass reduction during decaffeination. Overall, the coffee bean mass decreased by approximately 3 in terms of dry weight. Furthermore, Farah et al. demonstrated that cryptochlorogenic acid and neochlorogenic acid in coffee increased after thermal treatment, likely due to the hydrolysis and/or isomerization of chlorogenic acid analogs 26 . Such reactions that underwent during decaffeination may have contributed to the increase in cryptochlorogenic acid and neochlorogenic acid in the decaffeinated coffee beans. The fact that caffeine is selectively removed from coffee beans via the extraction using supercritical CO 2 with water is important information. Machmudah et al. reported a method for the selective removal of caffeine from coffee beans using supercritical CO 2 with water using lab-scale apparatus 60-250 mL 32,33 . Herein, we demonstrated that this method could be applied to pilot-scale apparatus 30 L . Furthermore, although trigonelline was well separated from caffeine using supercritical CO 2 with water, the authors considered only the separation of caffeine and chlorogenic acid 32, 33 .
3.2 Effect of roasting degree on major coffee compounds 3.2.1 Effect of peak temperature The effects of peak roasting temperature 125-250 on the concentrations of caffeine, trigonelline, chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid in coffee beans with or without decaffeination treatment were evaluated. The increase/decrease trends of the coffee compounds with different peak temperatures did not differ sig-nificantly between the two types of beans Fig. 3 . Caffeine concentration did not decrease at temperatures up to 250 because of the high thermal stability of caffeine, in agreement with the results of previous studies 27,34,35 . Although trigonelline concentration in both bean types showed no reduction up to 200 , it started to decrease Fig. 2 Reversed-phase HPLC chromatograms of A standard mixture and extracts of coffee beans B without and C with decaffeination before roasting and D without and E with decaffeination after roasting at a peak temperature of 225 with a hold time of 180 s and a temperature rise time of 1 2 0 s . P e a k 1 t r i g o n e l l i n e ; p e a k 2 neochlorogenic acid; peak 3 nicotinic acid; peak 4 theophylline; peak 5 caffeine; peak 6 cryptochlorogenic acid; peak 7 chlorogenic acid; peak 8 caffeic acid.

Effect of Roasting on Coffee Compounds
J. Oleo Sci. 5 slightly at 225 decrease of approximately 10 in both bean types and drastically at 250 decrease of approximately 40 in both beans . Similar trends were observed in several previous studies 27,34 . The three chlorogenic acids chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid showed different behaviors. The concentration of chlorogenic acid decreased in a temperature-dependent manner, whereas cryptochlorogenic acid and neochlorogenic acid increased up to approximately 200 and subsequently decreased at ≥ 200 . Similar trends were reported by Farah et al., where the increase in cryptochlorogenic acid and neochlorogenic acid at peak temperatures in the range of 125-200 was attributed to the conversion of chlorogenic acid analogs via hydrolysis and/ or isomerization reactions 26 . The decrease in their concentrations at ≥ 200 is likely due to the markedly enhanced thermal degradation reaction at that temperature. These data are extremely important to guide consumers in the prevention of coffee compound degradation during roasting. For example, roasting under mild 225 conditions can be recommended to enrich trigonelline in coffee beans with or without decaffeination treatment.

Effect of hold time
The effect of hold time 125-240 s at the peak temperature 225 or 250 on the coffee compound concentrations was investigated for regular and decaffeinated beans. The concentration of caffeine remained largely unchanged; however, those of trigonelline, chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid decreased with increasing hold time at both peak temperatures Fig. 4 . Moreover, the rates of decrease between non-decaffeinated and decaffeinated coffee beans were approximately equal. For trigonelline, the effect of hold times in the range of 120-240 s on thermal degradation was moderate at 225 . In contrast, at a peak temperature of 250 , although the decrease in trigonelline was relatively suppressed at a hold time of 120 s decrease of approximately 25 in both bean types , it was significant after 180 s decrease of 55 in both bean types . In addition, chlorogenic acids chlorogenic acid, cryptochlorogenic acid, and neochlorogenic  acid were almost completely degraded by extended roasting at 250 . Therefore, longer roasting at high temperatures e.g., 250 should be avoided to increase the residual concentrations of trigonelline and chlorogenic acids in coffee beans.

Effect of temperature rise time
The effect of temperature rise time 60-180 s at a peak temperature 225 or 250 on the concentration of coffee compounds in coffee beans with and without decaffeination treatment was evaluated. Similar to the results described above in sections 3.2.1 and 3.2.2 , the concentration of caffeine negligibly changed, whereas those of the other compounds decreased in response to the thermal history, that is, the temperature rise time Fig. 5 . Regarding trigonelline, the concentration decreased in accordance with the temperature rise time at 250 but not at 225 . Thus,  Effect of Roasting on Coffee Compounds 7 trigonelline was considered relatively stable at 225 . The decrease rates of trigonelline, chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid in non-decaffeinated and decaffeinated coffee beans were approximately equal. From the above results, the effects of roasting conditions peak temperature, hold time, and temperature rise time on the concentration of coffee compounds did not differ significantly between the beans with and without decaffeination treatment. Therefore, decaffeination does not affect the behavior of physiologically active compounds in subsequent roasting. This suggests that majority of the reported information regarding the effects of roasting conditions on compounds of non-decaffeinated coffee beans 26 28 can be applied to decaffeinated coffee beans.
3.3 Correlation between the concentration of coffee compounds and L* value of coffee beans The L* value is a common index used to evaluate the degree of coffee bean roasting. In general, coffee beans can be categorized into light 23.5 L* 25.0 , medium 21.0 L* 23.5 , and dark 19.0 L* 21.0 roasted degrees, according to the color changes after roasting. The effects of roasting conditions on the L* values of coffee beans with and without decaffeination treatment are summarized in Table 3. The coffee beans required roasting treatment at peak temperatures of ≥ 225 to reach a light or more The peak temperature was measured inside the roasting chamber. Results are expressed as mean±standard deviation (n = 3). 8 roasted degree. Interestingly, we found that the L* value of decaffeinated coffee beans decreased with lower thermal history than that of non-decaffeinated beans. This might be attributable to the changes in some compounds of the coffee beans and/or structural changes in the beans caused by decaffeination treatment. In fact, we observed the cross-sections of coffee beans by SEM and found that there were more voids in the decaffeinated coffee beans than in the non-decaffeinated beans Fig. 6 . In other words, decaffeination treatment increased the number of voids in the beans and thus improved their heat transfer efficiency during roasting, resulting in a reduced L* value with less thermal history. This indicates that decaffeinated coffee beans can reach a given roasting degree under milder roasting conditions than non-decaffeinated beans, which may suppress the thermal degradation of the physiologically active compounds during roasting. The correlation between the concentrations of caffeine, trigonelline, chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid and the L* value of coffee beans under different roasting conditions is shown in Fig. 7. The concentrations of the coffee compounds, except for caffeine, decreased with decreasing L* value, that is, with increasing roasting degree. Comparing the same roasting conditions peak temperature, hold time, and temperature rise time , the concentrations of these compounds trigonelline, chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid tended to be higher in the non-decaffeinated coffee beans than in the decaffeinated beans Figs. 3-5 . In contrast, comparing similar L* values, it was found that the concentrations of trigonelline, cryptochlorogenic acid, and neochlorogenic acid were almost identical between the non-decaffeinated and decaffeinated coffee beans, whereas the chlorogenic acid concentrations were higher in the decaffeinated beans Fig. 7 . These results support the above-mentioned suppression of coffee compound loss due to the shorter roasting time of decaffeinated coffee beans. Namely, several coffee compounds are lost during decaffeination, whereas their loss during roasting can be suppressed, resulting in

Effect of Roasting on Coffee Compounds
J. Oleo Sci. 9 their concentrations in the roasted coffee beans with decaffeination treatment being equivalent or more compared to those of roasted regular coffee beans. Although decaffeination treatment typically decreases the concentrations of physiologically active compounds along with that of caffeine 22 24 , which is a major concern, it was revealed that appropriately controlling the conditions of supercritical CO 2 decaffeination and roasting can limit this effect for several valuable coffee compounds, such as trigonelline and chlorogenic acid.

Conclusion
This study aimed to examine the effects of roasting conditions on the concentrations of physiologically active compounds caffeine, trigonelline, chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid in coffee beans prepared with and without supercritical CO 2 decaffeination treatment. High-temperature roasting ≥ 250 markedly degraded trigonelline, chlorogenic acid, cryptochlorogenic acid, and neochlorogenic acid but not caffeine. The degradation rates of these compounds in non-decaffeinated and decaffeinated coffee beans were largely the same. However, the L* value of decaffeinated coffee beans decreased with less thermal history compared with that of non-decaffeinated beans. Therefore, because the roasting of decaffeinated coffee beans can be completed more rapidly, the loss of coffee compounds was suppressed. This represents an important finding that certain coffee compounds in decaffeinated coffee beans can be maintained at the same or higher levels compared to those in non-decaffeinated coffee beans by adjusting roasting conditions.