Improvement of Robusta Coffee Aroma with L-leucine Powder

L-leucine powder (LP) were added to improve the aroma of Robusta coffee beans. Treatment was a short soaking (M1) or spraying procedure (M2), then LP was added at varying levels up to 3% (w/w). All samples were roasted (240 °C/15 min) and extracted using an espresso machine. Volatile compounds were analysed by solid-phase microextraction−gas chromatography−mass selective detection. Thirty volatile compounds (6 pyrroles, 8 pyrazines, 3 phenols, 9 furans, 2 ketones, 2 aldehydes) were analysed. In 15 coffee samples, the levels of total volatile compounds (based on peak area ratios) ranged from 8.9 (M1-1) to 15.5 (non-treated Robusta: NTR). Robusta coffee has lower levels of bitter aroma compounds when pre-treated with LP. The sum of bitter volatiles (phenols, pyrroles, pyrazines) was lowest in M1-5 (3% LP), M2-1 (1% LP; both dried at 50 °C/15 min) and M2-7 (3% LP, dried at 70 °C/15 min) compared with NTR (p < 0.05). bitter aroma (pyrazines, pyrroles and phenols) of Robusta beans by adding L-leucine powder (LP). The hypothesis of this study was that the addition of large amounts of specic amino acid such as LP, which is less-reactive in Maillard reaction, could induce competitive inhibition with presence amino acids in coffee beans. Treatment was a short soaking (M1) or spraying procedure (M2), then LP was added at varying levels up to 3% to affect the chemical reaction of volatiles during the roasting process. varieties NTA). volatile compounds, including 6 pyrroles, 8 pyrazines, 3 phenols, 9 furans, 2 ketones and 2 aldehydes, were identied by quantitative and qualitative analysis.


Analysis of volatile compounds in coffee by GC-MS
Volatile compounds in coffee were analysed by headspace−solid-phase microextraction−gas chromatography−mass spectrometry (HS-SPME-GC-MS). Extracted coffee solution (10 ml) and sodium chloride (1 g) were added to 20-ml headspace vial, and then 10 µl of quinoxaline as the internal standard (1,000 µg/ml) and 20 µl of n-alkane standard (10 µg/ml) were spiked. The mixture was equilibrated at 70 °C for 10 min using a hot plate. For adsorption of the volatile compounds, the SPME bre (DVB/CAR/PDMS) was exposed to the vial headspace at 70 °C for 40 min. The bre was then inserted into the GC injection port for desorption of the volatile compounds at 230 °C for 10 min. Gas chromatography−mass spectrometry (GC-MS) was performed using an Agilent 7820A GC with 5977E MS detector. Chromatographic separation of the volatile compounds was achieved using a DB-WAX column (60 m × 250 µm × 0.25 µm). The GC oven was set at 44 °C for 5 min, increased to 170 °C at 3 °C/min and held for 10 min and, nally, increased to 240 °C at 8 °C/min and held for 5 min.
GC-MSD was performed for qualitative and quantitative analysis of the volatile compounds. For qualitative analysis, each peak in the total ionisation chromatogram (TIC) obtained by GC-MSD was identi ed by co-injection, retention index (RI) on the DB-WAX column and mass spectrum in the Wiley Mass Spectral database. The quantitative analysis was calculated to peak area ratio (peak area of each peak/peak area of internal standard) of each compound with quinoxaline as an internal standard.

pH and colour measurements
Green/roasted coffee bean powder (25 g) was extracted with 200 ml of ltered water using an espresso coffee machine (BCC-480ES, Bean Cruise). The pH of the coffee was measured using a pH meter (SevenEasy, Mettler Toledo Co., Ltd., USA) at room temperature in triplicate.
Colour measurement of coffee powder was conducted in re ection mode using a colour meter from Nipon Denshoku Industries Co., Ltd. (Tokyo, Japan).  The volatile compounds of 15 coffee samples were analysed by SPME-GC-MSD. The pro le contained 30 volatile compounds, including 6 pyrroles, 8 pyrazines, 9 furans, 3 phenols, 2 ketones and 2 aldehydes (Table 2). For qualitative analysis, the peaks were identi ed by Kováts RI, co-injection and the Wiley Mass Spectrum Library database. A quantitative analysis was performed by calculating the peak area ratio of each volatile compound and the internal standard. All values are represented as the peak area ratio (peak area of each peak/peak area of internal standard).   3) The N.D. means that R.I was not found in the reference. Table 3 shows the peak area ratio of the volatile compounds for all 15 coffee samples: 13 samples of green coffee beans treated with LP, including M1 samples (M1-1 to M1-5) and M2 samples (M2-1 to M2-8), and the non-treated coffee beans of both varieties (NTR and NTA). Thirty volatile compounds, including 6 pyrroles, 8 pyrazines, 3 phenols, 9 furans, 2 ketones and 2 aldehydes, were identi ed by quantitative and qualitative analysis.   were detected at higher concentrations than non-heterocyclic compounds, including ketones and aldehydes. In coffee, pyrroles, pyrazines and phenols impart bitter, woody and smoky aromas. NTR showed higher peak area ratios of pyrroles, pyrazines and phenolic compounds compared to NTA (p < 0.05).
Ketones and aldehydes contributed the lowest peak area ratios among the 6 groups of volatile compounds. Comparing NTR and NTA, the peak area ratio of NTR was detected higher in total ketones and total aldehydes. Comparing NTR and M1 samples, NTR had a higher peak area ratio for total ketones and aldehydes, except for M1-1 (p < 0.05). In M1-1, the levels of total ketones were detected at 0.089 ± 0.005, more than twice that in NTR. Previously, Robusta beans were found to have more ketones than Arabica beans 21,22 .
The aroma of non-treated and treated coffee is described in Fig. 1, in which the peak area ratio of 30 volatile compounds in non-treated and treated coffee is expressed relative to their concentrations in NTA (100%). Figure 1(a−e) shows the aroma pro les of NTR, NTA and M1 samples. Volatile compounds in M1-1 were signi cantly lower than those in NTR, excluding acetoxyacetone (Fig. 1a). The peak area ratio of acetoxyacetone, related to buttery odour in coffee, was 66.85% and 71.61% higher than those in NTR and NTA, respectively. Figure 1(b) reveals signi cant differences in 28 volatile compounds, excluding 2,3dimethylpyrazine and acetoxyacetone, between M1-2 and NTA (p < 0.05). Signi cant differences were also observed between M1-3 and NTA, except for 2acetylpyrrole and acetoxyacetone (Fig. 1c), and between M1-4 and NTA, except for 2-formyl-1-methylpyrrole, furfuryl methyl ether and acetoxyacetone ( Fig. 1d; p < 0.05). Figure 1(e) highlights signi cant differences in 26 compounds, except for ketones (buttery odour), 2-acetyl-1-methylpyrrole and 2-acetylpyrrole (bitter odour) (p < 0.05). M1-5 had a 1.6-fold higher peak area ratio of total bitter aroma compounds (pyrazines, pyrroles, phenols) than that of NTA but a 24.61% lower peak area ratio of total bitter aroma compounds compared with NTR (p < 0.05). Figure 1(A−H) shows the aroma pro le of NTR, NTA and M2 samples. In Fig. 1(A), 24 volatile compounds in M2-1 showed signi cant differences compared with NTA (p < 0.05). The peak area ratio of volatile compounds in M2-1 was lower than that of NTA, except for 2-methoxy-4-vinylphenol and benzaldehyde. In addition, the peak area ratio of all volatiles in M2-1 was lower than that of NTR, especially the peak area ratio of pyrroles, pyrazines and phenols, which were 49.05%, 49.46% and 28.60% lower, respectively (p < 0.05). Figure 1(B) indicates the volatile compounds of M2-2, NTR and NTA. There were no signi cant differences between M2-2 and NTA in 2-formyl-1-methylpyrrole, furfuryl methyl ether and benzaldehyde. Although the peak area ratio of 1-furfuryl-2formylpyrrole, 4-hydroxy-3-methylacetophenone and 2-methoxy-4-vinylphenol in M2-2 was higher than those of NTR and NTA, the remaining volatile compounds in M2-2 indicated a signi cant decrease compared with NTR (p < 0.05). The peak area ratio of phenols, one of the bitter avours, was 9.21% and 57.10% higher, respectively, in M2-4 than in NTR and NTA (p < 0.05). Figure 1(E) shows signi cant differences in 27 compounds, except for 2-formyl-1-methylpyrrole, 2-acetyl-5-methylfuran and 2phenyl-2-butenal between M2-5 and NTA (p < 0.05). The peak area ratio of phenols in M2-5 was 10.88% and 57.89% higher than those in NTR and NTA, respectively, (p < 0.05). Figure 1(F) shows signi cant differences in 28 compounds, except for 2-formyl-1-methylpyrrole and benzaldehyde between M2-6 and NTA. The peak area ratio of total volatiles in M2-6 was the highest among the treated samples, and the sum of volatile compounds of bitter aroma (pyrazines, pyrroles and phenols) in M2-6 was the highest among all samples (p < 0.05). Figure 1(G) indicates signi cant differences in 25 volatiles between M2-7 and NTA. The peak area ratio of all volatile compounds in M2-7 was lower than that in NTR, especially the peak area ratio of pyrazines, one of the bitter avours, which was lower by 42.92% (p < 0.05). Figure 1(H) shows the pro les of volatile compounds in M2-8, NTR and NTA. The peak area ratio of volatile compounds in M2-8 indicated signi cant differences in 28 volatile compounds compared with NTA. Furthermore, M2-8 displayed a 9.54% and 57.26% higher peak area ratio of phenols compared with NTR and NTA, respectively (p < 0.05). Table 4 shows the pH and colour values in non-treated and treated coffee. The pH and colour of coffee are important sensory properties determining the quality of the beverage 15 . The pH values showed signi cant differences between NTR and treated coffee (p < 0.05). The pH of NTR was signi cantly higher than that of NTA (p < 0.05). Furthermore, the pH of treated coffee was signi cantly reduced compared with NTR (p < 0.05   Colour measurement of coffee powder was conducted in re ection mode using a colour meter. There were signi cant differences in ΔE* values between NTR and treated coffee (p < 0.05). The minimum colour difference that the naked eye can detect is ΔE* = 3.0 13 . The colour of NTR and NTA can be distinguished, while NTR and M1 samples are di cult to distinguish by the naked eye. The colour differences between NTR and M2-3 and M2-6 were both ΔE* 3.11 (p < 0.05). The colour of Arabica beans and Robusta beans can be distinguished, with Arabica beans displaying a higher L* value than Robusta beans 23 .

pH and colour
For evaluating the correlation among pH, L* and volatile compounds, the results for all samples are shown in Table 5 (*p < 0.05 and **p < 0.01). The pH values increased signi cantly as the levels of phenolic compounds and pyrazines increased (p < 0.01). That is, M1-5 (3% LP, dried at 50 °C for 15 min), M2-1 (1% LP, dried at 50 °C for 15 min) and M2-7 (3% LP, dried at 70 °C for 15 min) with low levels of phenols and pyrazines also had low pH values (p < 0.05). The levels of pyrroles, phenolic compounds and pyrazines increased signi cantly as the levels of total volatile compounds increased (p < 0.05). The correlation between L* values [white (L* = 100) and black (L* = 0)] and the levels of furans were also signi cant (p < 0.05).

Conclusion
In this study, LP was blended with green Robusta beans in two different ways to reduce the bitter aroma (pyrazines, pyrroles, phenols). The sum of bitter volatiles (pyrazines, pyrroles, phenols) in M1-5 (3% LP, dried at 50 ℃ for 15 min) was lower than that in NTR by 24.61%, and the sum was the lowest among the M1 samples (p < 0.05). The sum of bitter volatiles in M2-1 (1% LP, dried at 50 ℃ for 15 min) was the lowest among all treated samples. In particular, the peak area ratio of pyrazines in M2-1 was lower than that of NTR by 49.46% (p < 0.05). The sum of bitter volatiles in M2-7 (3% LP, dried at 70 ℃ for 15 min) was 18.38% lower than that of NTR by (p < 0.05).
This study shows that pre-treatment of Robusta beans with LP affects the chemical reactions responsible for the generation of volatiles during the roasting process. Pre-treatment with LP reduced the bitter aroma of Robusta beans. Unlike previous studies that soaked the beans in acetic acid or sugar solutions to improve the aroma, this study pre-treated the surface of green Robusta beans with LP. The results of this study can suggest a new manufacturing method for coffee.