ANTICANCER COMPOUNDS FROM MEDICINAL PLANTS


BIOLOGICA NYSSANA  5 (1)  September 2014: 11-15  Šarac, Z. et al.  Application of canonical discriminant analysis… 

 

 11 

 

 

 

Original Article 

Received: 28 June 2014 

Revised: 8 July 2014 

Accepted: 10 September 2014 

 

Application of canonical discriminant analysis in differentiation of 
natural populations of Pinus nigra in Serbia based on terpene 
composition 
 
 

Zorica Šarac
1*

, Srdjan Bojović
2
, Biljana Nikolić

3
, Bojan K. Zlatković

1
, and Petar D. Marin

4
 

 

 
1
University of Niš, Faculty of Sciences and Mathematics, Department of Biology and Ecology, 

Višegradska 33, 18000 Niš, Serbia 
2
University of Belgrade, Institute for Biological Research “Siniša Stanković”, Boulevard 

Despota Stefana 142, 11060 Belgrade, Serbia 
3
University of Belgrade, Institute of Forestry, Kneza Višeslava 3, 11000 Belgrade, Serbia 

4
University of Belgrade, Faculty of Biology, Institute of Botany and Botanical Garden 

“Jevremovac“, Studentski trg 16, 11000 Belgrade, Serbia 
 

* E-mail: saraczorica@gmail.com 
 

 

Abstract:  

Šarac, Z., Bojović, S., Nikolić, B., Zlatković, B., Marin, P.: Application of canonical discriminant 

analysis in differentiation of natural populations of Pinus nigra in Serbia based on terpene composition. 

Biologica Nyssana, 5 (1), September 2014: 11-15. 

The canonical discriminant analysis (CDA) was performed in order to check the hypothesis of chemical 

separation infraspecific taxa of Pinus nigra J.F. Arnold (ssp. nigra, var. gocensis, ssp. pallasiana, and var. 

banatica) in Serbia based on variability of the needle terpenes. The CDA, which maximizes variations 

between a priori groups, showed division of seven native P. nigra populations into three groups, which 

belong to three taxonomically recognized taxa (ssp. nigra, ssp. pallasiana, and var. banatica). The most 

important characters in discrimination were (E)-caryophyllene, α-humulene, terpinolene, germacrene D, α-

pinene, and myrcene. The individuals of pallasiana group were poorer in (E)-caryophyllene and terpinolene 

and richer in α-humulene contrary to the individuals of nigra group. The individuals of banatica group had 

the highest content of α-pinene and myrcene. The obtained results were compared with recently published 

data on chemodiversity of P. nigra infraspecific taxa on the territory of Serbia. 

Key words: canonical, chemotaxonomy, discriminant analysis, needle terpenes, Pinus nigra.  

 

Apstract:  

Šarac, Z., Bojović, S., Nikolić, B., Zlatković, B., Marin, P.: Primena diskriminantne kanonijske analize u 

diferencijaciji autohtonih populacija Pinus nigra u Srbiji na osnovu sastava terpena. Biologica Nyssana, 5 

(1), September 2014: 11-15. 

Diskriminatna kanonijska analiza (CDA) izvedena je u cilju provere hipoteze o fitohemijskom razdvajanju 

infraspecijskih taksona Pinus nigra J.F. Arnold (ssp. nigra, var. gocensis, ssp. pallasiana i var. banatica) u 

Srbiji na osnovu varijabilnosti terpena u četinama. CDA, koja naglašava varijabilnost između a priori 

definisanih grupa, pokazala je rezdvajanje sedam autohtonih P. nigra populacija u tri grupe (ssp. nigra, ssp. 

5 (1) • September 2014: 11-15 

 



BIOLOGICA NYSSANA  5 (1)  September 2014: 11-15  Šarac, Z. et al.  Application of canonical discriminant analysis… 

 

 12 

pallasiana i var. banatica). Najvažniji karakteri u diskriminaciji bili su (E)-kariofilen, α-humulen, terpinolen, 

germakren D, α-pinen i mircen. Individue grupe pallasiana siromašnije su u sadržaju (E)-kariofilena i 

terpinolena, a bogatije u sadržaju α-humulena, suprotno od individua grupe nigra. Individue grupe banatica 

imaju veći sadržaj α-pinena i mircena. Dobijeni rezultati upoređeni su sa nedavno objavljenim podacima o 

hemodiverzitetu P. nigra infraspecijskih taksona na teritoriji Srbije. 

Ključne reči: kanonijska, hemotaksonomija, diskriminantna analiza, terpeni u ;etinarima, Pinus nigra.  

 

 

Introduction 
 

Pinus nigra J.F. Arnold (Pinaceae) is a 

species of the tertiary origin belonging to the group 

of taxa typical for Mediterranean and 

submediterranean coniferous forests (B o g u n i c  et 

al., 2007). It is one of the widespread, 

morphologically as well as taxonomically very 

variable pine species in Europe (G a u s s e n  et al. 

1993), with highly fragmented distribution range 

that extends from North Africa through the northern 

Mediterranean and eastwards to the Black Sea. In 

the flora of Serbia, J o v a n o v i ć  (1992) recognized 

two subspecies of black pine that can be 

geographically differentiated, each further divided 

into several varieties. Generally, in the western, 

southwestern and central parts of Serbia, P. nigra 

ssp. nigra is widespread, within which a distinct 

variety is recognized (var. gocensis Đ o r đ e v i ć ), 

while in eastern Serbia P. nigra ssp. pallasiana (D. 

Don) Holmboe has an extremely narrow distribution 

range. Namely, Crimean black pine (P. nigra ssp. 

pallasiana) was only recorded in southeastern 

Serbia (R a j e v s k i , 1950) on the mountain Crnook 

(Jarešnik). Apart from this, in eastern and 

northeastern Serbia three additional populations of 

black pine are known, which  J o v a n o v i ć  (1992) 

describes as var. banatica Georg. et Ion. within ssp. 

pallasiana. 

The chemodiversity aspect of black pine on 

the territory of Serbia has been addressed 

concerning n-alkane (B o j o v i ć  et al., 2012) and 

terpene (Š a r a c  et al., 2013) variability. In the 

second study, 58 essential-oil components were 

identified in the needles of 195 trees from seven 

populations. Unlike principal component analyses 

(PCA), which showed an overlap of all populations, 

cluster analyses (CA) separated analysed samples 

into three basic groups: the first group consisted of 

populations from western (Banjska stena, Omar, 

Zmajevački potok), southwestern (Priboj) and 

central Serbia (Goč) (considered as ssp. nigra 

group), the second of population from southeastern 

Serbia (Jarešnik) (ssp. pallasiana group), and the 

third of population from northeastern Serbia 

(Lazareva reka), which had the most distinct oil 

composition (ssp. banatica group) (Š a r a c  e t  a l ., 

2013). 

Bearing this in mind, the aim of the present 

study was to check the hypothesis of chemical 

separation of infraspecific taxa of P. nigra from 

Serbia by performing a canonical discriminant 

analysis (CDA). This method maximizes variations 

between a priori defined groups and thus 

characterize the degree of divergence among 

analyzed populations (J a m e s  &  M c C u l l o c h , 

1990). Also, CDA was performed in order to 

determine the relative importance of terpenes as 

discriminators between a priori groups. 
 

Material and methods 

Plant material 

The samples from seven typical populations 

of P. nigra infraspecific taxa - I - Banjska stena, II - 

Omar, III - Zmajevački potok (ssp. nigra), IV - 

Priboj-Crni vrh, V - Goč-Gvozdac (var. gocensis), 

VI - Jarešnik (ssp. pallasiana), and VII - Lazareva 

reka (var. banatica), growing wild in Serbia were 

analyzed. Details about sampling, locations and 

ecological conditions of the selected populations, 

numbering of populations, the extraction of the 

essential oils, the GC-FID and GC/MS analyses, and 

the identification of terpenes have been reported 

previously (Š a r a c  e t  a l ., 2013). 

Statistical Analysis 

The CDA was carried out in order to check 

the hypothesis that the analyzed sample was 

composed of discrete groups, which are chemically 

differentiated one from other. From the total data set 

of 75 original compounds (for details s e e  Š a r a c  

e t  a l . ,  2013), 14 compounds were selected for 

CDA, viz., α-thujene, α-pinene, camphene, β-pinene, 

myrcene, limonene, (E)-β-ocimene, terpinolene, (E)-

caryophyllene, α-humulene, γ-muurolene, 

germacrene D, δ-cadinene, and germacrene D-4-ol. 

The 61 compounds which were found in traces 

(content <0.5%, according to R u d l o f f  e t  a l . ,  

1975; L i e u t i e r  e t  a l ., 1991) were excluded 

from further analysis. The CDA was computed on 

the selected data set (195 samples x 14 variables) 

with a priori defined groups. Statistical analyses 

were performed using the software Statistica 5.1 

(S t a t S o f t , 1997). 

 



BIOLOGICA NYSSANA  5 (1)  September 2014: 11-15  Šarac, Z. et al.  Application of canonical discriminant analysis… 

 

 13 

Results and Discussion 
 

The CDA based on individuals from seven 

native P. nigra populations showed that the first two 

discriminant functions participated in 79.6% of the 

total discrimination, of which the first function with 

64.7% (Tab. 1). The first discriminant function is 

approximately equal determined by the content of 

terpinolene, α-thujene, (E)-caryophyllene, α-pinene, 

and germacrene D. The second function is mostly 

defined by the content of germacrene D, α-pinene, 

and β-pinene (Tab. 1). 

 

Table 1. Standardized coefficients for the first two 

canonical axes (CA) of variation in terpene variables 

from the discriminant functional analysis of 7 a 

priori groups. 

Variables CA1 CA2 

α-Thujene -1.087 0.404 

α-Pinene 1.017 -1.823 
Camphene -0.408 -0.482 

β-Pinene 0.829 -1.618 
Myrcene 0.683 0.643 

Limonene -0.086 -0.408 

(E)-β-Ocimene 0.498 -0.494 

Terpinolene 1.260 -0.758 

(E)-Caryophyllene 1.053 -0.140 

α-Humulene -0.880 -0.840 

γ-Muurolene 0.016 -0.051 

Germacrene D 1.010 -2.305 
δ-Cadinene 0.729 -0.230 

Germacrene D-4-ol 0.432 -0.959 

Eigenvalue 1.493 0.344 

% explained variation 64.7 79.6 

 

The first discriminant function mainly 

separated population VI while the second function 

separated population VII from the other populations 

(Tab. 2, Fig. 1). The CDA with 7 populations was 

correctly classified only 53% of individuals on 

average (classification matrix values, data from the 

program). However, its results are sufficient to 

assume the existence of three groups of individuals: 

(1) populations I, II, III, IV, and V which can be 

designated as ssp. nigra, (2) population VI as ssp. 

pallasiana, and (3) population VII as ssp. banatica, 

in accordance with previous results of terpene 

analysis (Š a r a c  e t  a l . ,  2013). However, 

departing from the first group of individuals (ssp. 

nigra), population V was partially overlapped with 

the second group (ssp. pallasina) by axis 1. 

Differentiation suggestion obtained from seven 

populations was the inducement for performing the 

CDA with three a priori defined groups. 

 

 

Table 2. Means of Canonical Variables. 

Population CA1 CA2 

I 1.414 -0.241 

II 0.481 -0.247 

III 0.468 -0.149 

IV 0.881 -0.106 

V -0.654 0.041 

VI -2.366 -0.273 

VII -0.053 1.988 

 

 
Fig. 1. Canonical discriminat analysis (CDA) based 

on the content of 14 terpenes isolated from 195 

Pinus nigra samples of seven populations (I-VII). 

 

The CDA based on individuals from three 

groups (taxa) showed that the first discriminant 

function explained 73.3% and the second 26.7% of 

the total discrimination (Tab. 3). The first function 

was mostly determined by the content of (E)-

caryophyllene, α-humulene and terpinolene (Tab 3). 

The second function was mainly defined by the 

content of germacrene D, α-pinene, and myrcene. 

The first discriminant function separated nigra and 

pallasiana group, while the second function 

banatica group from the other groups (Fig 2). The 

individuals of pallasiana group had lower content of 

(E)-caryophyllene and terpinolene and higher 

content of α-humulene (considering that the 

majority of individuals was on the right side of the 

figure). The individuals of nigra group had higher 

content of (E)-caryophyllene and terpinolene, and 

lower content of α-humulene (the majority of 

individuals was on the left side of the figure). The 

individuals of banactica group had higher content of 

α-pinene and myrcene. The CDA with 3 a priori 

defined gropus was correctly classified 88% of 

individuals on average (classification matrix values, 

data from the program). 

 

 

 



BIOLOGICA NYSSANA  5 (1)  September 2014: 11-15  Šarac, Z. et al.  Application of canonical discriminant analysis… 

 

 14 

Table 3. Standardized coefficients for the first two 

canonical axes (CA) of variation in terpene variables 

from the discriminant functional analysis of 3 a 

priori groups. 

 
Variables CA1 CA2 

α-Thujene 0,741 -0,436 

α-Pinene -0.548 1,682 
Camphene 0,496 0,411 

β-Pinene -0,464 1,523 

Myrcene -0,739 -0,621 
Limonene 0,329 0,281 

(E)-β-Ocimene -0,348 0,464 

Terpinolene -0,781 0,785 

(E)-Caryophyllene -1,069 0,294 

α-Humulene 0,942 0,670 

γ-Muurolene 0,177 -0,102 

Germacrene D -0,477 2,169 
δ-Cadinene -0,719 0,345 

Germacrene D-4-ol -0,128 0,832 

Eigenvalue 0,922 0,336 

% explained variation 0,733 1,00 

 

 
Fig. 2. Canonical discriminat analysis (CDA) based 

on the content of 14 terpenes isolated from 195 

Pinus nigra samples of three groups (ssp. nigra, ssp. 

pallasiana and var. banatica). 

 

In accordance with previously published 

results of CA (Š a r a c  e t  a l ., 2013), CDA 

confirmed that based on the terpene variability in 

native populations of black pine in Serbia, three 

discrete groups can be distinguished. The first group 

(nigra) consisted of populations described as var. 

nigra (populations I-III) and var. gocensis 

(populations IV and V) within ssp. nigra according 

to existing morho-anatomical and phytocoenological 

studies (J o v a n o v i ć , 1992). Also, our very recent 

(Š a r a c  et al. (2013) and present study, showed that 

populations determined as var. gocensis exhibited a 

moderate tendency of separation from the first group 

(nigra) (population IV in CA and population V in 

CDA). Based on n-alkanes variability, the 

populations assigned as var. gocensis (IV and V) 

showed even grater tendency of splitting (Bojović et 

al., 2013). Population V has been classified into the 

first group (nigra), characterized as chemotype 1 by 

B o j o v i ć  et al. (2012), while population IV 

belonged to the second group (pallasiana), assigned 

as chemotype 2. It was assumed that populations 

considered as var. gocensis could be transitional 

form between subspecies nigra in the west and 

subspecies pallasiana in the southest of Serbia. 

Taxonomic position of var. gocensis within P. 

nigra complex is still not completely resolved. This 

variety was primarily described by Đ o r đ e v i ć  

(1931) from the slopes of the mountain Goč in 

central Serbia. Later, V i d a k o v i ć  (1955), 

determined the wider distribution of this variety, and 

based on the leaf anatomy raised it to the subspecies 

level (ssp. gocensis Vid.). Comparing to the typical 

var. nigra it is different in some tree traits (bark of 

mature trees had prominent transverse and 

longitudinal furrows, resembling the bark of Pinus 

heldreichii H. Christ), as well as in needles 

structure. However, according to results of our 

analyses there is no such clear phytochemical 

distinction between var. gocensis and typical var. 

nigra. 

The second group is composed mainly of 

trees of population VI from southeastern Serbia 

belonging to ssp. pallasiana. The presence of 

Crimean pine in Serbia was originally reported by 

A d a m o v i ć  (1909), but without precise data on its 

distribution. R a j e v s k i  (1950) was the first who 

established the position of this population within the 

territory of Serbia. The single enclava of Crimean 

pine is situated at the locality Jarešnik in the 

surroundings of Bosilegrad growing on the 

crystalline slates. The population in Serbia 

represents the northernmost disjunction of the 

Crimean black pine area spreading from the 

southern parts of Balkan Peninsula. Based on , n-

alkane and terpene composition, ssp. pallasiana is 

clearly distinguished as separate group (taxon and 

chemotype) from ssp. nigra according to this and 

previous studies (Bojović et al., 2012; Šarac et al., 

2013). 

Finally, the third group consisted of trees 

from population Lazareva reka, determined as var. 

banatica within ssp. pallasiana (J o v a n o v i ć , 

1992). This taxon is previously described in 

Romania (Domogled - Valea Cernei) and according 

to the literature data the black pine populations from 

northwestern Romania have a very controversial 

taxonomic position (B o ş c a i u  &  B o ş c a i u , 

1999). In the second edition of Flora Europea, 

G a u s s e n  et al. (1993) treated these populations as 

an independent species, i.e., P. banatica. Analysis 



BIOLOGICA NYSSANA  5 (1)  September 2014: 11-15  Šarac, Z. et al.  Application of canonical discriminant analysis… 

 

 15 

of the chemodiversity of the black pine populations 

from Serbia, based on both n-alkane and terpene 

composition, supports the view that var. banatica 

should be considered as distinct subspecies or even 

species (B o j o v i ć  et al., 2012; Š a r a c  et al., 

2013). 

 

Conclusion 
 

The canonical discriminant analysis (CDA) of 

recently reported terpene composition of P. nigra in 

Serbia (Š a r a c  et al., 2013), also showed division 

of populations into three discrete groups. The CDA 

confirmed the attitude that terpenes are good 

taxonomic markers at infraspecific level in the 

Pinales. Nevertheless, all these assumptions should 

be checked by further extending studies to the entire 

Balkan peninsula and the Mediterranean region, and 

especially including the classical locality (locus 

classicus) of P. nigra var. banatica from Romania 

(Domogled - Valea Cernei). Also, detail molecular 

analysis is needed to clarify relationships witin this 

extremely variable and complex taxon. 

 

References 
 

Adamović, L. 1909: Die Vegetationsverhältnisse der 

Balkanländer, Leipzig, 258-262. 

Bogunic, F., Muratovic, E., Ballian, D., Siljak-

Yakovlev, S., Brown, S.C. 2007: Genome size 

stability of five subspecies of Pinus nigra Arnold 

s.l. Environmental and Experimental Botany, 59: 

354-360. 

Bojović, S., Šarac, Z., Nikolić, B., Tešević, V., 

Todosijević, M., Veljić, M., Marin, P. D. 2012: 

Composition of n-alkanes in natural populations 

of Pinus nigra from Serbia – chemotaxonomic 

implications. Chemistry & Biodiversity, 9: 2761-

2774. 

Boşcaiu, N., Boşcaiu, M. 1999: On the presence of 

Pinus nigra subsp. pallasiana in Romania. 

Wissenschaftliche Mitteilungen aus dem 

Niederösterreichischen Landesmuseum, 12: 21-

24. 

Đorđević, P. 1931: Pinus nigra Arn. var. gočensis, 

n. var. Izdanje Ministarstva Šuma i Rudnika, 

Beograd. 

Gaussen, H., Heywood, V.H., Charter, A.O. 1993. 

Pinus L.. In: Tutin, T.G., Heywood, V.H., 

Burges, N.A., Valentine, D.H., Walters, S.M., 

Webb, B.A. (eds.), Flora Europea 1: 40-44, 

Cambridge University Press, Cambridge. 

James, F.C., McCulloch, C.E. 1990: Multivariate 

analysis in ecology and systematics: panacea or 

Pandoras box?. Annuual Reviews of Ecology and 

Systematics, 21: 129-166. 

Jovanović, B. 1992. P. nigra Arn. In: Sarić, M. 

(ed.), Flora Srbije 1: 200-202, Srpska Akademija 

Nauka i Umetnosti, Beograd. 

Lieutier, F., Berryman, A.A., Millstein, J.A. 1991: 

Preliminary study of the monoterpene response 

of three pines to Ophiostoma clavigerum 

(Ascomycetes: Ophiostomatales) and two 

chemical elicitors. Annals of Forest Science, 48: 

377-388. 

Rajevski, L. 1950: Nalazište Pinus nigra Arn. var. 

pallasiana (Lamb.) u okolini Bosilegrada. 

Zbornik radova Instituta za ekologiju i 

biogeografiju, Srpska Akademija Nauka i 

Umetnosti, Beograd. 

StatSoft 1997. Statistica for Windows, version 5.1. 

StatSoft Inc., Tulsa. 

Šarac, Z., Bojović, S., Nikolić, B., Tešević, V., 

Đorđević, I., Marin, P. D. 2013: 

Chemotaxonomic significance of the terpene 

composition in natural populations of Pinus 

nigra J. F. Arnold from Serbia. Chemistry & 

Biodiversity, 10: 1507-1520. 

Vidaković, M. 1955: Značenje anatomske građe 

iglica kod svojta crnog bora u Jugoslaviji. 

Šumarski list, Zagreb, 79: 244-253. 

von Rudloff, E. 1975: Volatile leaf oil analysis in 

chemosystematic studies of North America 

conifers. Biochemical Systematics and Ecology, 

2: 131-167. 
 

  

 

 

 



BIOLOGICA NYSSANA  5 (1)  September 2014: 11-15  Šarac, Z. et al.  Application of canonical discriminant analysis… 

 

 16