Chemical diversity in coffee species of genebank of Instituto Agronômico do estado

The objective of this work was to group different coffee species present in the genebank of the Instituto Agronômico by using some chemical variables. A total of thirty-nine plants belonging to seven species were analyzed for chemical seed components (soluble solids, lipids, trigonelline, chlorogenic acids and caffeine). The results evidenced that species could be grouped by lipids, chlorogenic acid, trigonelline and caffeine. The existence of great variation among and within species ranging from 24.12 to 30.65% for soluble solids; 8.68 to 17.49% for lipids; 0.32 to 2.15% for trigonelline; 2.91 to 6.38% for chlorogenic acid and 0.80 to 2.50% for caffeine was also observed, indicating the possibility to select plants of interest for the improvement of cultivated coffee species. 1 Centro de Café “Alcides Carvalho”, Instituto Agronômico (IAC/APTA), C. P. 28, 13.001-970, Campinas, SP, Brasil. *E-mail: aguiar@iac.sp.gov.br 2 Departamento de Produção Vegetal, ESALQ/USP, C. P. l 9, 13.418-900, Piracicaba, SP, Brasil Adriano Tosoni da Eira Aguiar1*, Luiz Carlos Fazuoli1, Terezinha de Jesus Garcia Salva1, and José Laércio Favarin2 Received 8 April 2005 Accepted 20 September 2005


INTRODUCTION
Coffee plants belong to the family Rubiaceae and genus Coffea and comprise approximately 100 identified taxa, which include all species of agronomical importance.The two commercially most important species are C. arabica and C. canephora, usually known as Arabica and Robusta coffee, respectively.
Coffee species are geographically distributed across the tropical area of Africa.The establishment of germplasm banks of the known species at a global level stationed in many countries is very important to include the wild forms, which can have extremely advantageous agronomic traits such as resistance and tolerance to diseases, nematodes, insects, drought, frost, and other biotic and abiotic factors, as well as differing plant traits (root system, stem, leaves, flowers, fruits, and seeds).
Coffee genebanks exist in a few countries although these collections have few accesses of species aside from C. arabica and C. canephora.At the Instituto Agronômico in Campinas (IAC), state of São Paulo, it was possible to collect 16 of the main species of the genus so far, besides different varieties of C. arabica, C. canephora and C. liberica species.
This plant material presents great genetic variability in several traits such as: stem traits and size of plant, leaves, fruits and seeds, resistance to biotic and abiotic factors, root system and cup quality.Furthermore, it forms a valuable genepool for different breeding purposes which does not only present differences among morphological Chemical diversity in coffee species of genebank of Instituto Agronômico do estado de São Paulo and agronomical traits but also at biochemical and molecular levels (Ky et al. 2001).
Numerous criteria have been used to determine genetic diversity of coffee genus such as some morphological, agronomical, cytological, molecular, and geographical traits.
The objective of this research was to group different species present in the IAC Coffee genebank and to evaluate the existing variability by using chemical variables.

MATERIAL AND METHODS
Seeds of 39 plants from seven species were used.The plants were provided by the Coffee genebank of the Instituto Agronômico -IAC/APTA, at the Experimental Center of Campinas (Santa Elisa farm) in Campinas, SP.
Table 1 shows the relationship among species, their respective varieties and analyzed plant samples.The seven following species used in this investigation were: C. canephora, C. liberica, C. congensis, C. eugenioides, C. stenophylla, C. racemosa, and C. kapakata.Mature fruits of each plant were collected individually at the berry stadium in 2002, according to the species' maturation time.The dried seeds were finely ground to perform the analyses.
The soluble solids content was determined using 10 grams of ground coffee according to methodology number 15.034 of AOAC (1997).
To estimate the total oil content, 10 grams of ground coffee were extracted with 100 mL petroleum ether in a Butt apparatus for 16 hours.The solid material was weighed after drying (30 minutes at 105 ºC).Lipid concentrations were calculated by the difference between the initial and degreased mass weight (Mazzafera et al. 1998).
Trigonelline, chlorogenic acid and caffeine were extracted in 70% methanol at 60 ºC for an hour (Mazzafera 1999) and quantified by High Performance Liquid Chromatography (HPLC) (Casal et al. 2000).The elution was performed in isocratic mode using a mobile phase 30%; 0.3%; 69.7%, v/v/v, methanol/ acetic acid/bidistilled water (pH 3.0) at a flow rate of 1mL min -1 .Trigonelline, chlorogenic acid and caffeine were quantified by comparing peak heights with standard values and the results presented as dry base.
The experiment comprised randomized complete block designs with two replications.Variance analyses were applied for the soluble solids, lipids, trigonelline, chlorogenic acid, and caffeine content.The Tukey test (P = 0.05) was applied to species means to identify statistically different species in relation to the evaluated traits.Multivariate analysis was used to determine the principal components (PCA), discriminant factors (DFA), the classification matrix, and Mahalanobis distance.All statistical analyses were performed on software Statistica (StatSoft Inc. 1999).

RESULTS AND DISCUSSION
The results obtained for chemical (soluble solids, lipids, trigonelline, chlorogenic acid, and caffeine) variables for the seven species used in the present investigation are shown in Table 1.
Experimental coefficients of variation for soluble solids and trigonelline indicated a good precision ranging from 1.60 to 4.59%, respectively.Significant variation was observed for the contents of all analyzed chemical variables (Table 1).Variation was also observed among plants within species.These variations can be attributed to genetic differences among them.
The extreme values observed in the different species ranged from 24.12 to 30.65% for soluble solids; 8.68 to 17.49% for lipids; 0.32 to 2.15% for trigonelline; 2.91 to 6.38% for chlorogenic acid and 0.80 to 2.50% for caffeine.The soluble solids contents ranged from 24.12 to 30.35%.C. canephora (25.52% c) is a cultivated species with more soluble solids than C. arabica, another cultivated species.The trait high soluble solids is desirable for the soluble coffee industry.C. liberica var.dewevrei (29.02%) and C. congensis (27.94%), both crossable with C. arabica (24.1%) have the potential to raise soluble solids contents to about 14% higher than in C. canephora var.robusta.There is additional phenotypic varibility available among C. liberica accesses, ranging from 26.22 to 30.46%.There are Arabica phenotypic selections with good agronomic traits and with C. liberica genes that can be explored to improve this trait in C. arabica selections such as the germplasm of Catuaí Sh 3 and Mundo Novo Sh 3 .
Lipid contents are very important because of the high value of coffee oil on the market and the tocopherol content in the oil.Lipid contents in coffee species range from 10.69 to 16.75%.One of the ancestors of C. Arabica, The bioflavonoid chlorogenic acid is very important as antioxidant (Amorin and Silva 1968) and ranged from 3.29 (C.liberica) to 6.22% (C.stenophylla).If breeding aims at reduced contents, the best source is C. liberica (3.02 to 3.98%), while C. arabica cultivars presented 3.2% (Martín et al. 1998).Four species that are easy to cross with C. arabica but can not be used to obtain a lower chlorogenic acid content in Arabica cultivars are C. eugenioides (4.76%), C. congensis (4.86%), C. racemosa (4.43%) and C. canephora (5.48).
Based on the results of principal components analysis (PCA) species can be positioned and the chemical variables associated in a biplot (Figure 1).The first two principal components explained 79.3% of the total variation.In the biplot, the X axis characterized by the chlorogenic acid, caffeine and trigonelline accounted for 43.4% and Y axis, which was characterized essentially by the soluble solids and lipids, accounted for 35.9%.The variables chlorogenic acid and caffeine are highly correlated to the first principal component, demonstrating effectiveness at differentiating coffee species.Theses results show that the variables lipids, chlorogenic acid, trigonelline and caffeine contents can separate coffee species relative effectively, evidencing the existence of three species groups, as indicated in Figure 1.
Figure 1 shows that species on lower right (C. eugenioides and C. kapakata) formed one group and species on the upper left (C.congensis, C. canephora, C. stenophylla and C. racemosa) formed another.The species C. liberica on the upper right formed one more group, though a little distant from the other species.Our results agree well with earlier studies in literature, where the three identified groups were identified using chemical traits (Carvalho and Monaco 1967, Clifford et al. 1989, Anthony et al. 1993, Bridson 1994and Mazzafera and Guerreiro Filho 1998).It should be emphasized that in an investigation realized by Mazzafera and Guerreiro Filho (1998), the authors worked with coffee pulp (a maternal tissue) to detect chemical components through HPLC, in contrast to our study.
In general, the results evidenced that most of the species from central and west Africa have superior caffeine contents (C.canephora, C. congensis, and C. stenophylla), with exception of C. kapakata and C. liberica, in contrast to eastern Africa species (C.eugenioides and C. racemosa).The low caffeine content in C. kapakata and C. liberica species could be explained by the following reason: in agreement with Anthony et al. (1993), based on observations of taxonomic criteria, C. kapakata has many affinities with species from east Africa.With regard to C. liberica, these differences between the present study and literature data are due to genetic variability present among coffee plants of this species.Is known that C. liberica is widely distributed across the African continent, extending from Guinea (west) until Zaire (central) (Dussert et al. 1999).
Table 2 presents the classification matrix and plant numbers of the different analyzed materials established through the discriminating factorial analysis (DFA).In the mean, about 97.4% of total plants concerning the different species were classified in the respective species.In general, the existence of a low variability was observed in the classification within species among species (92.0 to 100.0%).
The plants of the C. canephora, C. liberica, C. eugenioides, C. stenophylla, C. racemosa and C. kapakata species were showed to be similar, indicating that all were classified in their proper species (Table 2).In an inferior position, although very well classified, is C. congensis.This species presen ted 92.0% of classification.It was observed that from the total of the C. congensis plants, only one was not classified as of the proper species, but as belonging to C. racemosa.These results confirm the fidelity of the groupings obtained by PCA.
The Mahalanobis' distances of the different analyzed coffee species were obtained ( 2. The variable soluble solids was not effective at discriminating coffee species.
3. Results evidenced the existence of great variation among species and within species for all analyzed variables, indicating the possibility of selecting plants of interest to achieve enhanced cultivated species with higher trigonelline, lower caffeine and lower chlorogenic acid.

Table 1 .
Soluble solids, lipids, trigonelline, chlorogenic acid and caffeine content in coffee species Chemical diversity in coffee species of genebank of Instituto Agronômico do estado de São Paulo

Table 3 )
, whose values represent the divergence among the materials.The largest distance was verified between C. kapakata and C. liberica species (196.2), while C. eugenioides and C. kapakata presented the greatest similarity (12.4) among the analyzed materials.C. canephora and C. congensis are very similar, as well C congensis and C. racemosa, and C. eugenioides and C. kapakata.On the other hand, C. kapakata is very different from C. canephora, C. liberica and C. congensis, as well as C. stenophylla from C. liberica, C. eugenioides from C. canephora and C. liberica, C. congensis from C. eugenioides, and C. canephora from C. kapakata.These results suggest that the distribution of biochemical diversity does not coincide with the species geographic distribution.

Table 2 .
Classification matrix and plants number among coffee species used in the present study

Table 3 .
Estimates of the Mahalanobis distances among coffee species used in the present study