LWT - Food Science and Technology 140 (2021) 110802

Contents lists available at ScienceDirect

LWT
journal homepage: www.elsevier.com/locate/lwt

Comparison of chemical and fatty acid composition of green coffee bean

(Coffea arabica L.) from different geographical origins
Mengting Zhu a, You Long a, Yingjie Ma a, Yi Chen a, *, Qiang Yu a, Jianhua Xie a, Bin Li b,
Jinlong Tian b
a
State Key Laboratory of Food Science and Technology, Nanchang University, Nanchang, 330047, People’s Republic of China
b
College of Food Science, Shenyang Agricultural University, Shenyang, 110866, People’s Republic of China

A R T I C L E I N F O A B S T R A C T

Keywords: This study aimed to quantitatively screen and compare the chemical compounds including protein, lipid, sucrose,
Green coffee beans total phenolics (TPC), and total titratable acidity (TTA) along with the fatty acid profile of green coffee beans
Geographical origin from various geographical origins (China, Indonesia, Ethiopia, Kenya, Guatemala, Honduras, Brazil, and
Partial least squares discriminant analysis (PLS-
Colombia). Meanwhile, partial least squares discriminant analysis (PLS-DA) was used to investigate whether the
DA)
analyzed parameters’ data can be successfully applied to differentiate the geographical origins of green coffee
Fatty acid composition
Sucrose beans. The levels of protein, lipid, sucrose, TTA, and TPC in all samples were in the range of 13.06–15.98 g 100
g− 1, 12.88–16.29 g 100 g− 1, 7.30–11.45 g 100 g− 1, 231.43–361.21 mL 0.1 mol L− 1 NaOH 100 g− 1, and
37.56–50.57 mg GAE g− 1, respectively. The fatty acid composition was dominated by C18:2, C16:0, C18:1, and C18:0.
The major fatty acid group was unsaturated fatty acids (USFA), accounting for 51.66–54.28% of total fatty acids.
Statistical analysis showed significant differences in the chemical and fatty acid composition of green coffee
beans from eight geographical origins (P＜0.05). Moreover, the PLS-DA results revealed excellent discrimination
between the different geographical origins of green coffee beans, highlighting lipid, C24:0, C22:0, C18:3, C17:0, C18:0,
C20:0, C16:0, and protein as discriminatory features.

1. Introduction Occhipinti, & Maffei, 2016). Thus, Arabica coffee has major economic
significance and commercial value. Coffees are primarily planted in
Coffee is one of the most desired non-alcoholic natural beverages regions lying between the latitudes of 20◦ N and 20◦ S. The regions are
worldwide and its consumption has increased significantly over recent also known as the “coffee belt”, which encompasses about 80 countries
years. The popularity of coffee arises from its enjoyable taste sensation, of Africa, Asia, Centre-America, and South-America continents (Giraudo
its psychostimulant effects, and its appealing beneficial healthy effects et al., 2019). Unavoidably, a considerable difference is found in coffee
(Cloete et al., 2019; Dong et al., 2019). Recently, epidemiological flavor or aroma within these world’s regions. In other words, the flavor
studies have shown a significant role for coffee consumption in reducing or aroma of coffees is highly related to their geographical growing
the risk of type 2 diabetes (Carlstrom & Larsson, 2018), colorectal cancer regions.
(Li, Ma, Zhang, Zheng, & Wang, 2013), Parkinson’s, and Alzheimer’s It has been reported that the flavor or aroma of coffee produced
disease (Bae, Park, Im, & Song, 2014). during roasting was strongly concerning the chemical composition of
Coffee is obtained from the beans which grow inside the fruit of a the green coffee beans (Liu, Yang, Yang, et al., 2019; Liu, Yang, Linforth,
tropical evergreen shrub of the genus Coffea. There are about 70 species et al., 2019; Worku, de Meulenaer, Duchateau, & Boeckx, 2018). For
are reported to belong to the genus Coffea. However, only Coffea Arabica example, proteins, amino acids, and sugars are significant precursors
(Arabica coffee) and Coffea Canephora (Robusta coffee) are the primary involved in the Maillard reaction during roasting, which affects the
cultivated and marketed species. Arabica coffee is more preferred than aroma and flavor of the coffee. Similarly, the lipid is another vital
Robusta coffee because of its lower bitterness and better aroma char­ contribute participated in decomposition and auto-oxidation reactions
acteristics, which provide more than 95% of the world’s coffee (Babova, during roasting, which contributes to the formation of coffee aroma and

* Corresponding author.
E-mail address: chenyi15@ncu.edu.com (Y. Chen).

https://doi.org/10.1016/j.lwt.2020.110802
Received 22 October 2020; Received in revised form 19 December 2020; Accepted 19 December 2020
Available online 24 December 2020
0023-6438/© 2020 Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
M. Zhu et al. LWT 140 (2021) 110802

flavor. Besides, the bitterness and sourness of coffee are generated from 2.3. Moisture content determination
chlorogenic acids and corresponding organic acid precursors during
roasting. Although it is known that the composition of green coffee 2 g of green coffee powder was placed in an electric constant tem­
beans strongly affects coffee flavor or aroma, which are highly related to perature drying oven at 105 ◦ C until the difference between successive
their geographical growing regions, little quantitative information is weight was less than or equal to 2 mg. The results were present as a
telling the effect of geographical origins on the chemical and fatty acid percentage of green coffee beans.
composition of the green coffee beans. Besides, identification of green
coffee bean’s geographical origins is a stimulating issue, owing to the 2.4. Sugar analysis
diverse commercial value of green coffee beans based on their
geographical origins. Sugar analysis was performed according to the high-performance
It’s known that the partial least squares (PLS-DA) method was liquid chromatography (HPLC) method described by National Stan­
powerful in studying differences among coffee beans from different dards of the People’s Republic of China (China Food and Drug Admin­
classes (Garrett et al., 2013; Malta et al., 2020; Monteiro et al., 2019). istration, 2016). In a 50 mL centrifuge tube, 3 g of green coffee powder
It’s developed from partial least squares (PLS) algorithms but applied for was dissolved in 15 mL petroleum ether, then the mixture was shaken by
sorting. The variable importance in projection (VIP) value of the PLS-DA an IKA Vortex Genius 3 shaker (IKA, Staufer, Germany) for 2 min to
model was a parameter of screening biomarkers in metabolomics. remove oil. After centrifuging the mixture at 2 000 rpm for 10 min, the
Generally, the variable with a VIP value greater than 1 can be considered supernatant was removed and the residue was dried at 40 ◦ C. Then the
as a differential variable between groups. Thus, previous studies have dried residue was transferred carefully into a 50 mL volumetric flask and
been carried out with the PLS-DA method for the classification of coffee extracted with deionized water. Before adjusting the solution volume to
beans and the identification of important variables (potential bio­ 50 mL, the sample solution was treated with 2 mL each of zinc acetate
markers) involved in the coffee beans classification. For instance, the (21.9 g 100 mL− 1) and potassium ferrocyanide (10.6 g 100 mL− 1) so­
PLS-DA model was applied for the discrimination of the genealogical lution to remove polymeric components. The extraction process was
groups of Coffea arabica green coffee beans by Malta et al. (2020), and carried out under an ultrasonic wave for 30 min at room temperature.
the most important variables related to the discrimination of coffee Finally, the sugar extracts were filtered through a 0.22 μm polytetra­
beans genealogical groups were also identified. Similarly, Garrett et al. fluoroethylene (PTFE) filter.
(2013) using the PLS-DA model successfully differentiate the Sarchimor The sugar contents were determined using high-performance liquid
and Catuaí cultivars of Arabica coffee, and revealed the correlations chromatography (HPLC, Agilent 1260) coupled with a refractive index
between the coffee metabolites with the Arabica cultivars. detector (RI). A CNW Athena NH2-RP (250 × 4.6 mm, 5 μm) column was
Therefore, the objectives of this study were to chemical character­ used to separate via isocratic elution with acetonitrile/deionized water
ization of commercially available green coffee beans from some of the (7.5:2.5, v/v) at a flow rate of 1.0 mL/min. The injection volume was 10
world’s top coffee production regions to identify specific chemical pat­ μL, column, and detector temperatures were both 35 ◦ C. Sugar contents
terns and to differentiate green coffee beans for their geographical ori­ were quantified based on an external calibration curve of reference
gins. Specifically, quantitatively screen and compare the chemical standards expressed in g⋅100− 1 green coffee beans (dry weight).
compound contents including protein, lipid, sucrose, total phenolics
(TPC), and total titratable acidity (TTA) as well as the fatty acid profile 2.5. Protein and lipid content determination
of green Arabica coffee beans from China, Indonesia, Ethiopia, Kenya,
Guatemala, Honduras, Colombia, and Brazil. PLS-DA models were The gross protein content was determined by the Kjeldahl nitrogen
established to help explain whether the analyzed chemical compound’s method. After digestion of the organic substance of green coffee powder
data can be successfully applied to differentiate the geographical origins (approximately 0.5 g) with sulfuric acid, the nitrogen content was
of green coffee beans. Further, potential discriminatory features for measure using Hanon K9860 Automatic Kjeldahl Analyzer (Jinan Hanon
differencing the geographical origin of green coffee beans were also Instrument Co., Ltd., China), then the protein content was calculated by
identified. 6.25 times nitrogen content.
The lipid extraction was adapted from (Dong, Hu, Chu, Zhao, & Tan,
2. Materials and methods 2017). Briefly, 3 g green coffee powder was mixed with 60 mL of pe­
troleum ether in a round bottom flask. The mixture was refluxed at 55 ◦ C
2.1. Samples until the petroleum ether was colorless using a water bath. The results of
protein and lipid content were both expressed in g 100 g− 1 green coffee
Fifty green Arabica coffee beans from the crop year 2018–2019 were beans (dry weight).
collected from several origins around the world that included Brazil (n
= 7), Colombia (n = 7), Ethiopia (n = 5), Guatemala (n = 5), Honduras 2.6. Fatty acid composition analysis
(n = 7), Indonesia (n = 5), Kenya (n = 6), China (n = 8). All samples
were crushed by a grinder and sieved through a No. 40 mesh (380 μm). The fatty acid composition in each green coffee bean lipid extract
was studied by Agilent 6890N gas chromatography (GC) coupled with a
flame ionization detector (FID) with and a split/splitless injector. The
2.2. Chemicals and reagents separation of fatty acids was performed on a CP-Sil 88 capillary column
(100 m × 0.25 mm × 0.20 μm, Varian Inc., USA), according to the
Fatty acid methyl ester standards (GLC-463), triheneicosanoin (in­ method reported in the published work (Chen et al., 2014).
ternal standard, C21:0, TAG), and methyl heneicosanoate (C21:0, FAME)
were obtained from Nu-Chek Prep, Inc. (Elysian, MN, USA). HPLC grade 2.7. Total phenolics and titratable acidity analysis
n-heptane and chloroform were bought from Fisher Scientific (Fair
Lawn, NJ, USA). HPLC-grade acetonitrile and methanol were obtained 1 g of green coffee powder was weighed and extracted with 100 mL
from Merck (Darmstadt, Germany). Guaranteed grade potassium hot deionized water (95 ◦ C) for 20 min, then filtered through a mem­
hydroxidewere, Folin-Ciocalteu’s phenol reagent, sucrose, gallic acid, brane (Whatman No. 4) under a vacuum. The filtered liquid (extracts)
and 0.1 mol L− 1 NaOH standard solution purchased from Aladdin was applied for total phenolics and titratable acidity analyses.
(Shanghai, China). Ultrapure water was used throughout the experi­ The total phenolic content (TPC) of 2X diluted extracts was deter­
ments (Milli Q-System, Millipore, Milford, MA, USA). mined using the Folin-Ciocalteu method. The procedure is a

2
M. Zhu et al. LWT 140 (2021) 110802

modification of previously reported work (Kim, Kim, Kim, Kim, & Baik, between 8.00 and 10.45% (except an Indonesian sample, 7.64%). Out of
2018). Quantification of TPC was achieved via an external standard the eight geographical origins of green coffee beans, it can be observed
curve of gallic acid and expressed as milligrams of gallic acid equivalents that the highest moisture content was observed for Guatemalan coffee
per gram of green coffee bean (mg GAE g− 1) beans, which were significantly higher than those from the other
The total titratable acidity (TTA) of 10X diluted extracts was deter­ countries (P＜0.05). Nevertheless, no significant differences were ob­
mined by titration with 0.1 mol L− 1 NaOH solution. The phenolphtha­ tained in the moisture content of green coffee beans from the other re­
lein solution (1%) was used as an indicator, and the results are expressed gions (P > 0.05).
in mL 0.1 M NaOH 100 g− 1 of green coffee beans (Barbosa, Scholz, As shown in Table 1, lipid content values were in the range of
Kitzberger, & Benassi, 2019). 12.88–16.29 g 100 g− 1. Guatemalan samples were characterized by the
highest lipid content (15.23–16.29 g 100 g− 1), which were significantly
greater than that of Honduran, Colombian, Brazilian and Indonesian
2.8. Data analysis
samples. While green coffee beans from Guatemala, China, Kenya, and
Ethiopia had statistically similar lipid concentrations. Similarly, Hon­
Statistical analysis was evaluated by the one-way analysis of variance
duran, Colombian and Brazilian samples have statistically similar lipid
(ANOVA), followed by Duncan’s multiple range test. All statistical tests
concentrations. On the other hand, Indonesian samples had the lowest
were two-tailed and differences at P < 0.05 were considered to be sig­
lipid content with values ranging from 12.88 to 13.62 g 100 g− 1, which
nificant. ANOVA analyses were performed using the statistical package
were significantly lower than that of the other countries. The results of
SPSS 23.0 (SPSS Inc., Chicago, IL, USA). Partial least squares discrimi­
this study were similar to those reported earlier. Scholz et al. (2016)
nant analysis (PLS-DA) was built to differentiate the geographical ori­
reported that the green Arabica coffee beans from Ethiopia possess a
gins of Arabica green coffee beans with chemical parameters and fatty
lipid content ranged between 13.58 and 16.58 g 100 g− 1. Besides, an
acid composition. PLS-DA was performed using SIMCA-P+ version 13
extensive range of lipid contents (11.5–16.24 g 100 g− 1) in Brazilian
(Umetrics, Umea, Sweden). To investigate the potential variables for
green coffee beans was reported by Barbosa et al. (2019) and Wage­
discriminating the geographical origin of green coffee beans from the
maker, Carvalho, Maia, Baggio, and Guerreiro (2011).
eight countries, variable influence on projection (VIP) values were
In the current study, the levels of total protein were between 13.06
suggested. The potential variables were selected based on VIP values >
and 15.98 g 100 g− 1. As can be observed from Table 1, the statistical
1.
analysis clearly showed that protein content was significantly influenced
by geographical origins (P < 0.05). Green coffee beans from Brazil have
3. Results and discussion a statistically greater amount of protein content (14.34–15.98 g 100 g− 1)
than those from the other regions. The second highest protein values
3.1. Chemical composition of green coffee beans from different were found in green coffee beans from China (13.95–15.75 g 100 g− 1)
geographical origins and Guatemala (14.33–14.75 g 100 g− 1), which were statistically
greater than the concentration in green coffee beans from the other five
Table 1 presents the minimum and maximum, the mean concentra­ countries. On the contrary, the protein values of Colombian, Ethiopian,
tion, and the standard deviations of the chemical parameters of the Indonesian, and Honduran green coffee beans showed no obvious dif­
green coffee beans from each country. The results of the significance ference. The lowest protein values were recorded in Kenyan green coffee
analysis were also present in Table 1. Meanwhile, Fig. 1 showed the beans (13.06–13.67 g 100 g− 1). Previous studies have reported the
values of the chemical parameters in each green coffee bean and the protein contents of Brazilian green coffee beans ranged from 12.08 to
distributions of the chemical parameters in each country. 16.36 g 100 g− 1 (Barbosa et al., 2019; Scholz, Kitzberger, Prudencio, &
The quantification of moisture content of green coffee beans is Silva, 2018), and Ethiopian green coffee beans from 12.33 to 16.78 g
necessary because the composition of foods is often reported based on 100 g− 1 (Scholz et al., 2016). As can be seen, the data published by other
dry matter (Ozbekova & Kulmyrzaev, 2019) and it is the foremost authors coincided with our results.
parameter to ensure safe transport and storage (Caporaso, Whitworth, Sucrose accounts for more than 90% of the total low molecular
Grebby, & Fisk, 2018). Generally, the moisture content of green coffee carbohydrates in green coffee beans (Cheng, Furtado, Smyth, & Henry,
beans between 8.0 and 12.5% could maintain the quality of green coffee 2016). However, there are only trace amounts of glucose and fructose.
beans (Adnan, Horsten, Pawelzik, & Morlein, 2017). As shown in The sucrose content of all analyzed green coffee beans varied between
Table 1 and Fig. 1, the moisture content of all analyzed samples ranged

Table 1
The ranges and average values of chemical compounds determined in green Arabica coffee beans from different geographical origins. Values in the same rows with
different letters are significantly different (P < 0.05).
China (n = 8) Indonesia (n = Kenya (n = 6) Ethiopia (n = 5) Guatemala (n = Honduras (n = Colombia (n = Brazil (n = 7)
5) 5) 7) 7)

Moisturea Range 8.27–9.19 7.64–10.23 8.23–9.56 8.61–10.45 9.15–10.43 8.80–9.67 8.74–9.77 8.55–9.17
Average 8.79b ± 0.30 9.01b ± 1.15 8.74b ± 0.46 9.29b ± 0.72 9.99a±0.51 9.17b ± 0.37 9.34b ± 0.39 8.87b ± 0.21
a
Lipid Range 14.75–15.67 12.88–13.62 14.14–16.15 14.33–15.19 15.23–16.29 14.05–15.53 13.37–14.59 12.88–14.86
Average 15.22ab ± 0.34 13.26d ± 0.26 15.13ab ± 0.77 14.98ab ± 0.37 15.66a±0.41 14.63bc±0.55 14.18c±0.48 14.00c±0.80
Proteina Range 13.95–15.75 13.52–14.11 13.06–13.67 14.01–14.13 14.33–14.75 13.20–14.07 13.63–14.03 14.34–15.98
Average 14.58b ± 0.63 13.81c±0.23 13.34d ± 0.26 14.06c±0.05 14.53b ± 0.17 13.60cd ± 0.30 13.85c±0.15 15.27a±0.54
Sucrosea Range 7.30–9.50 8.68–10.01 9.56–10.81 8.74–9.59 8.58–9.60 7.66–11.46 8.44–9.58 8.06–8.99
Average 8.57cd ± 0.79 9.43ab ± 0.57 10.09a±0.45 9.19bcd ± 0.41 9.02bcd ± 0.40 9.35abc±1.14 9.01bcd ± 0.39 8.51d ± 0.39
TPCb Range 41.88–49.11 44.15–50.57 39.57–49.11 43.19–48.14 42.11–47.65 42.22–49.87 39.74–43.88 37.56–43.85
Average 44.62bc±2.43 48.05a±2.69 43.76bcd ± 45.44abc±1.88 45.01abc±2.17 46.73ab ± 2.78 42.45cd ± 1.55 41.17d ± 2.32
3.54
TTAc Range 270.3–361.2 231.4–309.3 275.8–306.8 273.7–310.7 260.6–300.8 300.3–307.0 273.6–308.9 263.7–291.2
Average 306.5a±32.9 286.9ab ± 31.4 292.1ab ± 12.8 299.0ab ± 14.6 279.3b ± 14.5 303.5ab ± 3.1 290.1ab ± 13.1 279.3b ± 8.7
a
g 100 g− 1 green coffee bean.
b
Total phenolic content - mg GAE g− 1 green coffee bean.
c
Total titratable acidity - mL 0.1 mol L− 1 NaOH 100 g− 1 green coffee bean.

3
M. Zhu et al. LWT 140 (2021) 110802

Fig. 1. Boxplot of main components in different origins of green coffee beans presented as (A) moisture, (B) lipid, (C) protein, (D) sucrose, (E) total phenolic content
(TPC), and (F) total titratable acidity (TTA). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

7.30 and 11.45 g 100 g− 1. Kenyan and Indonesian green coffee beans coffee beans were distinguished by the lowest TPC values (37.56–43.85
showed the highest sucrose values (9.56–10.81 g 100 g− 1, 8.68–10.01 g mg GAE g− 1), while the highest TPC values ranging from 44.15 to 50.57
100 g− 1), while green coffee beans collected from Brazil showed the mg GAE g− 1 were determined in Indonesian samples. Statistical analysis
lowest sucrose values (8.06–8.99 g 100 g− 1). For those from other revealed that green coffee beans from Indonesia have a statistically
countries, the value of sucrose of the green coffee beans was not greater value of TPC than those from China, Kenya, Colombia, and Brazil
significantly different (P > 0.05). Our results were similar to those (P＜0.05). However, the TPC of green coffee beans from Indonesia,
described for Ethiopian Arabica green coffee beans by Scholz et al. Honduras, Ethiopia, and Guatemala have no statistical difference (P >
(2016) and Stredansky, Redivo, Magdolen, Stredansky, and Navarini 0.05). Geographical origin did not result in indistinguishable total
(2018), who found sucrose values ranging between 6.36 and 10.42 g phenolic content between Brazil and Colombia originated coffee sam­
100 g− 1. However, the sucrose contents were greater than those re­ ples, which is in accord with the previous report (Bilge, 2020). Cheong
ported by Barbosa et al. (2019) for green coffee beans from Brazil et al. (2013) reported that the TPC of green coffee beans from Indonesia
(5.13–7.76 g 100 g− 1) and by Stredansky et al. (2018) for green coffee ranged from 48.51 to 49.88 mg GAE g− 1, which fall within the range
beans from Indonesia (7.27 g 100 g− 1) and Colombia (7.0 g 100 g− 1). determined in the present study. Still, in that study, the amount of total
The total phenolic content (TPC) of the green coffee beans from each phenolics of Chinese green Arabica coffee bean was 53.76 mg GAE g− 1,
origin was present in Table 1 and Fig. 1. TPC values of all green coffee which was greater than those reported in our study (41.88–49.11 mg
beans varied from 37.56 to 50.57 mg GAE g− 1. The Brazilian green GAE g− 1).

4
M. Zhu et al. LWT 140 (2021) 110802

As shown in Table 1, the TTA of the green coffee extract was in the amounts of arachidic acid were determined in Ethiopian samples
range of 231.43–361.21 mL 0.1 M NaOH 100 g− 1, among which, the (2.13–2.30%). For linolenic acid, green coffee beans from Ethiopia were
Chinese samples had the highest TTA values varied from 270.27 to distinguished by the highest linolenic acid contents (1.33–1.45%), while
361.21 mL 0.1 M NaOH 100 g− 1, whereas the lowest TTA values were the lowest linolenic acid values ranging from 1.06 to 1.25% were
found in Guatemalan samples and its level was between 260.63 and determined in Chinese samples. Other fatty acids including myristic acid
300.81 mL 0.1 mol L− 1 NaOH 100 g− 1. Statistical analysis revealed that (C14:0), margaric acid (C17:0), gondoic acid (C20:1), behenic acid (C22:0),
there was no significant effect of the geographical origins on TTA values, lignoceric acid (C24:0) were minor fatty acids, which were detected at
except the differences of TTA between Chinese and Guatemalan sam­ percentages of less than 1%. Joet et al. (2010) reported fatty acid
ples. Other studies in the literature showed that TTA values of green compositional data consistent with the results of the present study. The
Arabica coffee beans varied from 182.45 to 409.85 mL 0.1 mol L− 1 major fatty acid in green coffee beans was linoleic acid (43.58–44.02%),
NaOH 100 g− 1 (Barbosa et al., 2019; M.; Scholz et al., 2018). followed by palmitic (35.42–35.64%), oleic (7.21–7.44%), stearic
(7.02–7.09%), arachidic (2.36–2.37%) and linolenic acid (1.88–1.90%).
The rest fatty acids were also detected at percentages of less than 1%,
3.2. Determination of fatty acid composition however, myristic and margaric acid wasn’t found in that study.
When considering the fatty acid groups it was found that the per­
A typical fatty acid profile chromatogram of the green coffee bean centage of saturated fatty acids (SFA) was less than that of unsaturated
was shown in Fig. 2, eleven fatty acids including seven saturated fatty fatty acids (USFA). Meanwhile, the content of polyunsaturated fatty
acids (myristic acid (C14:0), palmitic acid (C16:0), margaric acid (C17:0), acids (PUFA) was greater than that of monounsaturated fatty acids
stearic acid (C18:0), arachidic acid (C20:0), behenic acid (C22:0), ligno­ (MUFA). The findings were consistent with the reports by Mehari et al.
ceric acid (C24:0) and four unsaturated fatty acids (oleic acid (C18:1), (2019) and Joet et al. (2010). Statistical differences were found in the
linoleic acid (C18:2), linolenic acid (C18:3), gondoic acid (C20:1) were fatty acid group contents among green coffee beans from different ori­
detected in all samples. Fig. 3 and Table 2 illustrate the fatty acids profile gins (Table 2). The SFA content ranged from 45.72% to 48.34%, with
of green coffee beans from each country and statistical analysis. It can be maximum values in Chinese samples and minimum values in Guate­
observed that significant differences were present in the fatty acids malan samples. Conversely, green coffee beans from Guatemala con­
contents of green coffee beans from different geographical origins (P＜ tained the maximum values of unsaturated fatty acids (USFA) and from
0.05). China characterized by the minimum values. The unsaturated fatty acids
Firstly, the most principal fatty acid found in all green coffee beans (USFA = MUFA + PUFA) accounted for 51.66–54.28% of total fatty
was linoleic acid (C18:2), representing 41.31–45.37% of the total fatty acids. Mehari et al. (2019) found that the content of SFA ranging from
acids. Guatemalan samples were distinguished by the highest amount of 35.08 to 40.04%, which was smaller than that determined in this study.
linoleic acid (44.36–45.34%), while the lowest amount of linoleic acid Yet, these authors obtained higher USFA values (60.54–64.55%) than
was found in Chinese samples (41.31–42.58%). Secondly, Palmitic acid ours. Similar fatty acid group results were obtained by Joet et al. (2010),
(C16:0) was the second most prevalent fatty acid, accounting for with the PUFA and SFA ranging from 53.39 to 54.09% and
34.11–35.90% of the total fatty acids. The palmitic acid found in Bra­ 45.49–45.83%, respectively.
zilian samples (33.53–35.93%) was significantly lower than those from
the other regions, however, the content of palmitic acid found in green
coffee beans from the rest of the countries showed no significant dif­ 3.3. PLS-DA analysis
ferences. The other major fatty acids were oleic acid (C18:1), stearic acid
(C18:0), arachidic acid (C20:0), and linolenic acid (C18:3), accounting for The chemical (lipid, protein, sucrose, TPC, and TTA content) and
6.86–9.57%, 6.34–8.57%, 2.13–3.04%, and 1.06–1.55%, respectively, fatty acid composition of green coffee beans were submitted directly to
of the total fatty acids. Honduran samples showed the lowest oleic and PLS-DA. Before PLS-DA, auto-scaling (UV) was done for preprocessing.
stearic acid value (6.86–7.72%, 6.34–6.68%), while the highest oleic Geographical origin discrimination is presented in Fig. 4. The PLS-DA
and stearic acid was determined in Brazilian samples (8.60–9.57%, model fits two principal components (PCs). The first principal compo­
7.77–8.30%). Also, Brazilian samples were distinguished by the highest nent covered 36.8% of variables and the second component covered
amount of arachidic acid (2.80–3.04%), which was found statistically 17.5% of variables. The goodness of fit of the model was R2X = 0.542
greater than those from the other countries. On the contrary, the lowest and R2Y = 0.223. As shown in Fig. 4A, the PLS-DA score plot presents a

Fig. 2. Typical chromatograms of the fatty acids quantified in the analyzed green coffee beans. *myristic acid (C14:0), palmitic acid (C16:0), margaric acid (C17:0),
stearic acid (C18:0), oleic acid (C18:1), linoleic acid (C18:2), linolenic acid (C18:3), arachidic acid (C20:0), gondoic acid (C20:1), behenic acid (C22:0), lignoceric
acid (C24:0).

5
M. Zhu et al. LWT 140 (2021) 110802

Fig. 3. Fatty acids profiles of green coffee beans from different geographical origins. (For interpretation of the references to color in this figure legend, the reader is
referred to the Web version of this article.)

Table 2
The ranges and average values of individual fatty acids determined in green Arabica coffee beans from different geographical origins.
Fatty acids China (n = 8) Indonesia (n = Kenya (n = 6) Ethiopia (n = 5) Guatemala (n = Honduras (n = Colombia (n = 7) Brazil (n = 7)
5) 5) 7)

Major fatty acidsa (% of total FA)

C16:0 Range 35.47–37.64 35.5–36.67 36.03–38.03 36.42–36.87 36.11–36.99 36.22–37.57 35.83–36.71 33.53–35.93
Average 36.67a±0.79 36.04a±0.50 36.67a±0.69 36.70a±0.17 36.40a±0.35 36.67a±0.47 36.45a±0.31 34.59b ± 0.81
C18:0 Range 6.66–8.57 6.91–7.46 6.96–7.36 6.71–7.04 6.47–6.63 6.34–6.68 6.77–7.13 7.77–8.30
Average 7.62b ± 0.62 7.16c±0.20 7.21c±0.17 6.81de ± 0.13 6.53e±0.07 6.48e±0.12 7.01cd ± 0.12 8.11a±0.18
C18:1 Range 8.54–9.08 7.06–8.96 7.43–9.26 7.60–7.99 7.07–8.01 6.86–7.72 8.40–9.46 8.60–9.57
Average 8.85a±0.17 7.90bc±0.87 8.26b ± 0.71 7.76bc±0.18 7.58c±0.36 7.41c±0.35 9.01a±0.33 9.07a±0.36
C18:2 Range 41.31–42.58 42.14–44.90 42.23–44.36 43.13–44.15 44.36–45.34 43.17–45.77 41.73–43.23 42.24–43.02
Average 41.86f±0.49 43.64cd ± 1.32 43.04de ± 0.79 43.89bc±0.43 44.77a±0.36 44.69ab ± 0.94 42.42ef±0.45 42.61ef±0.30
C20:0 Range 2.41–2.78 2.54–2.79 2.23–2.50 2.13–2.30 2.25–2.38 2.18–2.51 2.45–2.72 2.80–3.04
Average 2.62b ± 0.14 2.65b ± 0.09 2.34c±0.12 2.20d ± 0.06 2.30cd ± 0.06 2.33c±0.11 2.58b ± 0.09 2.89a±0.09
C18:3 Range 1.06–1.25 1.17–1.55 1.22–1.37 1.33–1.45 1.17–1.23 1.19–1.32 1.21–1.32 1.22–1.43
Average 1.17e±0.06 1.35ab ± 0.14 1.28bcd ± 0.07 1.40a±0.05 1.21de ± 0.02 1.25de ± 0.05 1.26cd ± 0.04 1.34abc±0.09
Minor fatty acidsb (% of total FA)
C14:0 Range 0.05–0.07 0.05–0.06 0.06–0.07 0.06–0.06 0.05–0.06 0.05–0.06 0.05–0.06 0.06–0.07
Average 0.06abc±0.01 0.06cd ± 0.00 0.06ab ± 0.00 0.06abc±0.00 0.06bcd ± 0.00 0.05d ± 0.00 0.06abcd±0.00 0.06a±0.00
C17:0 Range 0.08–0.12 0.08–0.08 0.07–0.09 0.09–0.10 0.10–0.10 0.08–0.08 0.08–0.09 0.09–0.09
Average 0.10a±0.01 0.08c±0.00 0.08c±0.01 0.09ab ± 0.00 0.10a±0.00 0.08c±0.00 0.09bc±0.00 0.09b ± 0.00
C20:1 Range 0.25–0.28 0.27–0.29 0.25–0.27 0.24–0.25 0.27–0.27 0.27–0.30 0.28–0.29 0.25–0.26
Average 0.26bc±0.01 0.28a±0.01 0.26bc±0.01 0.25d ± 0.00 0.27b ± 0.00 0.28a±0.01 0.29a±0.00 0.25cd ± 0.01
C22:0 Range 0.51–0.62 0.57–0.63 0.51–0.65 0.55–0.66 0.51–0.59 0.48–0.58 0.54–0.65 0.70–0.79
Average 0.58b ± 0.04 0.58b ± 0.03 0.57b ± 0.05 0.59b ± 0.05 0.55bc±0.03 0.52c±0.03 0.59b ± 0.04 0.73a±0.03
C24:0 Range 0.20–0.23 0.24–0.26 0.21–0.24 0.24–0.26 0.22–0.24 0.22–0.25 0.23–0.25 0.24–0.27
Average 0.21d ± 0.02 0.25a±0.01 0.23c±0.01 0.25a±0.01 0.23bc±0.01 0.23bc±0.01 0.24ab ± 0.01 0.24a±0.01
Fatty acid groups (% of total FA)
SFAc Range 47.33–48.34 46.21–47.63 46.73–48.22 46.37–47.29 45.79–46.72 45.75–47.6 46.66–47.28 45.72–47.61
Average 47.85a±0.34 46.82bc±0.58 47.15b ± 0.56 46.71bcd ± 0.35 46.17d ± 0.34 46.37cd ± 0.68 47.02b ± 0.23 46.72bcd ± 0.61
c
USFA Range 51.66–52.67 52.37–53.79 51.78–53.27 52.71–53.63 53.28–54.21 52.4–54.25 52.72–53.34 52.39–54.28
Average 52.15d ± 0.34 53.18bc±0.58 52.85c±0.56 53.29abc±0.35 53.83a±0.34 53.63ab ± 0.68 52.98c±0.23 53.28abc±0.61
MUFAc Range 8.83–9.36 7.33–9.25 7.69–9.53 7.85–8.25 7.34–8.28 7.14–8.01 8.68–9.75 8.86–9.84
Average 9.12a±0.16 8.18bc±0.87 8.52b ± 0.72 8.01bc±0.18 7.85c±0.36 7.69c±0.35 9.29a±0.33 9.33a±0.37
PUFAc Range 42.49–43.7 43.45–46.39 43.59–45.58 44.46–45.56 45.54–46.54 44.45–46.96 43.01–44.49 43.54–44.45
Average 43.03e±0.45 45.00bc±1.40 44.32cd ± 0.73 45.28ab ± 0.46 45.98a±0.37 45.94a±0.92 43.69de ± 0.45 43.95d ± 0.37

Values in the same lines with different letters are significantly different (P < 0.05).
a
>1% of total FA.
b
≤1% of total FA; e: not detected.
c
SFA-saturated fatty acids; USFA-unsaturated fatty acids; MUFA-monounsaturated fatty acids; PUFA-polyunsaturated fatty acids.

clear differentiation among green coffee beans collected from China, Kenya). The model also fits a total of two principal components and the
Indonesia, Colombia Brazil, and other counties (Kenya, Ethiopia, fit of the model was R2X = 0.568, R2Y = 0.27; the variance of PC1 and
Guatemala, and Honduras). Samples collected from Guatemala and PC2 was 38.2% and 18.6%, respectively, representing 56.8% of the total
Honduras grouped with minor overlapping, and the samples of Kenya X variables, allowing a good differentiation of the green coffee beans
and Ethiopia were grouped with the maximum amount of overlapping. collected from seven geographical origins. The classification results
Then, a new PLS-DA model was built using the chemical and fatty were obtained in Fig. 4B. In this case, the classification was more
acid composition of green coffee beans from seven countries (without distinct. The PLS-DA score plot for all groups shows strong clustering

6
M. Zhu et al. LWT 140 (2021) 110802

Fig. 4. The score plot of the first two dimensions of PLS-DA analysis.

according to geographical origin, without any overlap. chemical marker to discriminate green coffee beans from different
geographical origins. To verify this result, another PLS-DA model was
built with the fatty acid composition of green coffee beans from seven
3.4. Potential biomarkers for discriminating green coffee beans countries (without Kenya) as input variables. The PLS-DA score plot was
geographical origin shown in Fig. 4C, the first two principal components explained 63.5% of
the total X variables. It can be observed that green coffee beans from
The variable importance in the projection (VIP) value of the PLS-DA Honduras slightly overlapped with those from Guatemala and Colombia.
model is a parameter of screening biomarkers in metabolomics (Hu However, green coffee beans collected from other countries showed
et al., 2020; Medina, Perestrelo, Santos, Pereira, & Camara, 2019). In the clear classification. The result further determined by the VIP score plot
present study, the variables with VIP values higher than 1 would be (Fig. 5B) that the fatty acids mentioned above (except C17:0) had VIP
identified as a differential variable between groups. As shown in Fig. 5A, values greater than 1. Besides, a previous study also informed that oleic,
eleven (class) compounds including lipid, C24:0, C22:0, C18:3, C17:0, C18:0, linoleic, palmitic, stearic, and arachidic acids were identified as the most
C20:0, C16:0, protein, C18:1, and C18:2 had VIP values greater than 1, discriminating compounds for the differentiation of green coffee beans
indicating that these compounds contributed to discriminating the green from the major producing regions Ethiopia (Mehari et al., 2019), which
coffee beans from the studied seven geographical origins. It’s demon­ agrees with our findings. Based on the results, it can be said that fatty
strated that the fatty acid composition can be suggested as a potential

7
M. Zhu et al. LWT 140 (2021) 110802

Fig. 5. Variable importance in the projection (VIP) score.

acids contain adequate information for use as descriptors of the geographical origin of green coffee. Therefore, present findings can
geographical region of green coffee beans. potentially have practical importance for both producers and suppliers
for preventing the deliberate mislabeling of the geographical origin of
4. Conclusion green coffee beans.

To sum up, significant differences in the chemical and fatty acid CRediT authorship contribution statement
composition of green coffee beans collected from eight geographical
origins were observed. Based on the studied parameters, green coffee Mengting Zhu: Conceptualization, Methodology, Validation,
beans according to their geographical origins were effectively identified Formal analysis, Investigation, Writing - original draft, Project admin­
with the help of the PLS-DA model. Moreover, potential chemical istration. You Long: Methodology, Validation, Formal analysis, Inves­
markers including lipid, C24:0, C22:0, C18:3, C17:0, C18:0, C20:0, C16:0, tigation, Writing - review & editing, Visualization. Yingjie Ma:
protein, C18:1, and C18:2 for discriminating the geographical origin of Methodology, Validation, Formal analysis, Investigation, Writing - re­
green coffee beans were highlighted. It’s demonstrated that fatty acid view & editing, Visualization. Yi Chen: Conceptualization, Methodol­
composition could serve as authenticity indicators to verify the ogy, Validation, Writing - review & editing, Visualization, Supervision,

8
M. Zhu et al. LWT 140 (2021) 110802

Project administration, Funding acquisition. Qiang Yu: Resources, Dong, W., Hu, R., Long, Y., Li, H., Zhang, Y., Zhu, K., et al. (2019). Comparative
evaluation of the volatile profiles and taste properties of roasted coffee beans as
Writing - review & editing, Visualization, Supervision. Jianhua Xie:
affected by drying method and detected by electronic nose, electronic tongue, and
Resources, Writing - review & editing, Visualization, Supervision. Bin HS-SPME-GC-MS. Food Chemistry, 272, 723–731. https://doi.org/10.1016/j.
Li: Resources, Writing - review & editing, Visualization, Supervision. foodchem.2018.08.068
Jinlong Tian: Resources, Writing - review & editing, Visualization, Garrett, R., Schmidt, E. M., Pereira, L. F. P., Kitzberger, C. S. G., Scholz, M. B. S.,
Eberlin, M. N., et al. (2013). Discrimination of arabica coffee cultivars by
Supervision. electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry
and chemometrics. Lwt-Food Science and Technology, 50(2), 496–502.
Giraudo, A., Grassi, S., Savorani, F., Gavoci, G., Casiraghi, E., & Geobaldo, F. (2019).
Declaration of competing interest Determination of the geographical origin of green coffee beans using NIR
spectroscopy and multivariate data analysis. Food Control, 99, 137–145. https://doi.
org/10.1016/j.foodcont.2018.12.033
The authors declare no conflict of interest in this work. Hu, G., Peng, X., Gao, Y., Huang, Y., Li, X., Su, H., et al. (2020). Effect of roasting degree
of coffee beans on sensory evaluation: Research from the perspective of major
chemical ingredients. Food Chemistry, 331, 127329. https://doi.org/10.1016/j.
Acknowledgments
foodchem.2020.127329
Joet, T., Laffargue, A., Descroix, F., Doulbeau, S., Bertrand, B., de Kochko, A., et al.
The financial supports from the National Key Research and Devel­ (2010). Influence of environmental factors, wet processing and their interactions on
opment Program of China (2019YFE0106000, 2017YFC1600405) and the biochemical composition of green Arabica coffee beans. Food Chemistry, 118(3),
693–701. https://doi.org/10.1016/j.foodchem.2009.05.048
the National Natural Science Foundation of China (No: 31471647, Kim, W., Kim, S. Y., Kim, D. O., Kim, B. Y., & Baik, M. Y. (2018). Puffing, a novel coffee
21866021) are gratefully acknowledged. bean processing technique for the enhancement of extract yield and antioxidant
capacity. Food Chemistry, 240, 594–600. https://doi.org/10.1016/j.
foodchem.2017.07.161
References Li, G., Ma, D., Zhang, Y., Zheng, W., & Wang, P. (2013). Coffee consumption and risk of
colorectal cancer: A meta-analysis of observational studies. Public Health Nutrition,
Adnan, A., Horsten, D. V., Pawelzik, E., & Morlein, A. D. (2017). Rapid prediction of 16(2), 346–357. https://doi.org/10.1017/S1368980012002601
moisture content in intact green coffee beans using near infrared spectroscopy. Liu, C., Yang, Q., Linforth, R., Fisk, I. D., & Yang, N. (2019). Modifying Robusta coffee
Foods, 6(5). https://doi.org/10.3390/foods6050038 aroma by green bean chemical pre-treatment. Food Chemistry, 272, 251–257.
Babova, O., Occhipinti, A., & Maffei, M. E. (2016). Chemical partitioning and antioxidant https://doi.org/10.1016/j.foodchem.2018.07.226
capacity of green coffee (Coffea arabica and Coffea canephora) of different Liu, C., Yang, N., Yang, Q., Ayed, C., Linforth, R., & Fisk, I. D. (2019). Enhancing Robusta
geographical origin. Phytochemistry, 123, 33–39. https://doi.org/10.1016/j. coffee aroma by modifying flavour precursors in the green coffee bean. Food
phytochem.2016.01.016 Chemistry, 281, 8–17. https://doi.org/10.1016/j.foodchem.2018.12.080
Bae, J. H., Park, J. H., Im, S. S., & Song, D. K. (2014). Coffee and health. Integrative Malta, M. R., Fassio, L. D., Liska, G. R., Carvalho, G. R., Pereira, A. A., Botelho, C. E., …
Medicine Research, 3(4), 189–191. https://doi.org/10.1016/j.imr.2014.08.002 Pereira, R. G. F. A. (2020). Discrimination of genotypes coffee by chemical
Barbosa, M. S. G., Scholz, M., Kitzberger, C. S. G., & Benassi, M. T. (2019). Correlation composition of the beans: Potential markers in natural coffees. Food Research
between the composition of green Arabica coffee beans and the sensory quality of International, 134.
coffee brews. Food Chemistry, 292, 275–280. https://doi.org/10.1016/j. Medina, S., Perestrelo, R., Santos, R., Pereira, R., & Camara, J. S. (2019). Differential
foodchem.2019.04.072 volatile organic compounds signatures of apple juices from Madeira Island according
Bilge, G. (2020). Investigating the effects of geographical origin, roasting degree, particle to variety and geographical origin. Microchemical Journal, 150, 104094. https://doi.
size and brewing method on the physicochemical and spectral properties of Arabica org/10.1016/j.microc.2019.104094
coffee by PCA analysis. Journal of Food Science & Technology, 57(9), 3345–3354. Mehari, B., Redi-Abshiro, M., Chandravanshi, B. S., Combrinck, S., McCrindle, R., &
https://doi.org/10.1007/s13197-020-04367-9 Atlabachew, M. (2019). GC-MS profiling of fatty acids in green coffee (Coffea arabica
Caporaso, N., Whitworth, M. B., Grebby, S., & Fisk, I. D. (2018). Rapid prediction of L.) beans and chemometric modeling for tracing geographical origins from Ethiopia.
single green coffee bean moisture and lipid content by hyperspectral imaging. Journal of the Science of Food and Agriculture, 99(8), 3811–3823. https://doi.org/
Journal of Food Engineering, 227, 18–29. https://doi.org/10.1016/j. 10.1002/jsfa.9603
jfoodeng.2018.01.009 Monteiro, P. I., Santos, J. S., Rodionova, O. Y., Pomerantsev, A., Chaves, E. S.,
Carlstrom, M., & Larsson, S. C. (2018). Coffee consumption and reduced risk of Rosso, N. D., et al. (2019). Chemometric authentication of Brazilian coffees based on
developing type 2 diabetes: A systematic review with meta-analysis. Nutrition chemical profiling. Journal of Food Science, 84(11), 3099–3108.
Reviews, 76(6), 395–417. https://doi.org/10.1093/nutrit/nuy014 Ozbekova, Z., & Kulmyrzaev, A. (2019). Study of moisture content and water activity of
Cheng, B., Furtado, A., Smyth, H. E., & Henry, R. J. (2016). Influence of genotype and rice using fluorescence spectroscopy and multivariate analysis. Spectrochimica Acta
environment on coffee quality. Trends in Food Science & Technology, 57, 20–30. Part A: Molecular and Biomolecular Spectroscopy, 223, 117357. https://doi.org/
https://doi.org/10.1016/j.tifs.2016.09.003 10.1016/j.saa.2019.117357
Chen, Y., Yang, Y., Nie, S. P., Yang, X., Wang, Y. T., Yang, M. Y., … Xie, M. Y. (2014). The Scholz, M. B. D., Kitzberger, C. S. G., Pagiatto, N. F., Pereira, L. F. P., Davrieux, F.,
analysis of trans fatty acid profiles in deep frying palm oil and chicken fillets with an Pot, D., et al. (2016). Chemical composition in wild ethiopian Arabica coffee
improved gas chromatography method. Food Control, 44, 191–197. https://doi.org/ accessions. Euphytica, 209(2), 429–438. https://doi.org/10.1007/s10681-016-1653-
10.1016/j.foodcont.2014.04.010 y
Cheong, M. W., Tong, K. H., Ong, J. J. M., Liu, S. Q., Curran, P., & Yu, B. (2013). Volatile Scholz, M., Kitzberger, C. S. G., Prudencio, S. H., & Silva, R. (2018). The typicity of
composition and antioxidant capacity of Arabica coffee. Food Research International, coffees from different terroirs determined by groups of physico-chemical and sensory
51(1), 388–396. https://doi.org/10.1016/j.foodres.2012.12.058 variables and multiple factor analysis. Food Research International, 114, 72–80.
China Food and Drug Administration. (2016). GB 5009.8-2016 determination of fructose, https://doi.org/10.1016/j.foodres.2018.07.058
glucose, sucrose, maltose, lactose in national food safety standard. Beijing, China: The Stredansky, M., Redivo, L., Magdolen, P., Stredansky, A., & Navarini, L. (2018). Rapid
standard Press of the People’s Republic of China. sucrose monitoring in green coffee samples using multienzymatic biosensor. Food
Cloete, K. J., Smit, Z., Minnis-Ndimba, R., Vavpetic, P., du Plessis, A., le Roux, S. G., et al. Chemistry, 254, 8–12. https://doi.org/10.1016/j.foodchem.2018.01.171
(2019). Physico-elemental analysis of roasted organic coffee beans from Ethiopia, Wagemaker, T. A. L., Carvalho, C. R. L., Maia, N. B., Baggio, S. R., & Guerreiro, O. (2011).
Colombia, Honduras, and Mexico using X-ray micro-computed tomography and Sun protection factor, content and composition of lipid fraction of green coffee
external beam particle induced X-ray emission. Food Chemistry X, 2, 100032. https:// beans. Industrial Crops and Products, 33(2), 469–473. https://doi.org/10.1016/j.
doi.org/10.1016/j.fochx.2019.100032 indcrop.2010.10.026
Dong, W., Hu, R., Chu, Z., Zhao, J., & Tan, L. (2017). Effect of different drying techniques Worku, M., de Meulenaer, B., Duchateau, L., & Boeckx, P. (2018). Effect of altitude on
on bioactive components, fatty acid composition, and volatile profile of robusta biochemical composition and quality of green arabica coffee beans can be affected
coffee beans. Food Chemistry, 234, 121–130. https://doi.org/10.1016/j. by shade and postharvest processing method. Food Research International, 105,
foodchem.2017.04.156 278–285. https://doi.org/10.1016/j.foodres.2017.11.016