Inter- and Intra-specific Variation among Five Erythroxylum Taxa Assessed by AFLP

Abstract

• Background and Aims The four cultivated Erythroxylum taxa (E. coca var. coca, E. novogranatense var. novogranatense, E. coca var. ipadu and E. novogranatense var. truxillense) are indigenous to the Andean region of South America and have been cultivated for folk-medicine and, within the last century, for illicit cocaine production. The objective of this research was to assess the structure of genetic diversity within and among the four cultivated alkaloid-bearing taxa of Erythroxylum in the living collection at Beltsville Agricultural Research Center.

• Methods Amplified fragment length polymorphism (AFLP) fingerprinting was performed in 86 Erythroxylum accessions using a capillary genotyping system. Cluster analysis, multidimensional scaling (MDS) and analysis of molecular variance (AMOVA) were used to assess the pattern and level of genetic variation among and within the taxa.

• Key Results A clear distinction was revealed between E. coca and E. novogranatense. At the intra-specific level, significant differentiation was observed between E. c. var. coca and E. c. var. ipadu, but the differentiation between E. n. var. novogranatense and E. n. var. truxillense was negligible. Erythroxylum c. var. ipadu had a significantly lower amount of diversity than the E. c. var. coca and is genetically different from the E. c. var. ipadu currently under cultivation in Colombia, South America.

• Conclusions There is a heterogeneous genetic structure among the cultivated Erythroxylum taxa where E. coca and E. novogranatense are two independent species. Erythroxylum coca var. coca is most likely the ancestral taxon of E. c. var. ipadu and a founder effect may have occurred as E. c. var. ipadu moved from the eastern Andes in Peru and Bolivia into the lowland Amazonian basin. There is an indication of artificial hybridization in coca grown in Colombia.

INTRODUCTION

The extensive living collection of Erythroxylum at the Beltsville Agricultural Research Center, Beltsville, MD, USA, which has been maintained since the early 1970s, contains the four cultivated alkaloid-bearing varieties: Erythroxylum coca var. coca Lam (E. c. var. coca); Erythroxylum coca var. ipadu Plowman (E. c. var. ipadu); Erythroxylum novogranatense var. novogranatense (Morris) Hieron (E. n. var. novogranatense); Erythroxylum novogranatense var. truxillense [Rusby] Plowman (E. n. var. truxillense). The geographical, ecological and morphological differences of these taxa were detailed as early as the 16th century (Ganders, 1979; Plowman, 1979, 1982, 1984; Rury, 1981; Schultes, 1981). However, it was not until the 1970s that cultivated coca was determined to be derived from two species of the genus Erythroxylum; E. c. var. coca Lam and E. n. var. novogranatense (Morris) Hieron (Plowman, 1979, 1982, 1984; Bohm et al., 1982). This classification, according to Plowman and Rivier (1983), has been supported multifactorially through interdisciplinary research (Johnson et al., 2003b). Furthermore, breeding evidence and eco-geographical data suggests that the most likely phylogeny for the four cultivated taxa is a linear evolutionary sequence, wherein E. c. var. coca is the ancestral taxon that gave rise to E. n. var. truxillense which gave rise, in turn, to E. n. var. novogranatense (Bohm et al., 1982).

Since the 1970s, morphology, breeding systems and chemotaxonomic data were the primary descriptors used to detail the differences between the cultivated Erythroxylum taxa (Bohm et al., 1982; Johnson et al., 1997, 1998, 2002, 2003a; Johnson and Schmidt, 1999). With the identification of molecular markers and their associated specificity, further assessment of the genetic diversity among the cultivated Erythroxylum taxa is warranted in order to accrete and refine the existing morphological and chemotaxonomic-based classification system. While a variety of molecular assays could be used to assess the genetic diversity, each method differs in principle, application, the amount of polymorphism detected, cost and time required. In a previous study, amplified fragment length polymorphism (AFLP) (Vos et al., 1995) was used to analyse 132 accessions of Erythroxylum to characterize and positively identify the four cultivated taxa, as well as a feral taxon (Johnson et al., 2003b). The first objective of the current study was to examine further the taxonomic status, and elucidate the evolutionary relationship, of the four cultivated alkaloid-bearing Erythroxylum taxa. The second objective was to detect and quantify the inter- and intra-specific genetic variation in these taxa. Eighty-five Erythroxylum accessions, which are representative of the four cultivated taxa in the living collection, were analysed using AFLP genotyping in combination with cluster and ordination analysis. The resulting information provided insights into the structure and pattern of genetic diversity of Erythroxylum in the living collection at Beltsville Agricultural Research Center and the Andes region of South America.

MATERIALS AND METHODS

Plant material

Young expanding leaf tissue was harvested from 86 samples, taken from the living collection at Beltsville Agricultural Research Center, of E. coca var. coca Lam, E. coca var. ipadu Plowman, E. novogranatense var. novogranatense (Morris) Hieron, E. novogranatense var. truxillense (Rusby) Plowman and E. ulei O.E. Schulz, as well as some F₁ Erythroxylum coca var. ipadu propagules (Table 1). The ethnobotany, morphological characterization, alkaloid content, breeding system, chemotaxonomic data and geographical distribution have been summarized previously (Ganders, 1979; Plowman, 1979, 1981, 1982, 1983, 1984; Rury, 1981; Schultes, 1981; Bohm et al., 1982; Plowman and Rivier, 1983; Johnson et al., 1997, 1998, 1999, 2002, 2003b; Johnson and Schmidt, 1999). Due to the confusion by investigators for the four cultivated alkaloid-bearing taxa it was considered necessary to show how the taxa differed using AFLP DNA analysis. The living collec-tion of Erythroxylum at Beltsville Agricultural Research Center was authenticated by T. Plowman in 1988 and re-authenticated by P. M. Rury in 1993. Plants derived from the living collection and authenticated by Rury were transferred to a Hawaiian field site. The Hawaiian field site was located on the Island of Kauai and was selected by the US Department of Agriculture, Agricultural Research Service and the State of Hawaii because of the similarity of soils to those found in the coca-growing regions of Bolivia and Peru. The pH of the soil at the Hawaiian field sites ranged from 4·0 to 5·7, which was ideal for coca growth. The harvested leaf tissues were separately placed in labelled Zip-Loc bags, immediately stored at 0 °C and transported to the laboratory for DNA extraction and analysis. Erythroxylum ulei in the current study was only used to verify the consistency of the AFLP analysis and was not part of the quantitative analysis.

Table 1.

Sample table of 86 Erythroxylum c. var. coca, Erythroxylum c. var. ipadu, Erythroxylum n. var. novogranatense, E. n. var. truxillense and Erythroxylum ulei samples from the living collection at the Beltsville Agricultural Research Center

Species	Accession tag	Origin	Species	Accession tag	Origin
coca	B102LS	Bolivia	novo	B292LS	Bolivia
coca	B85LS	Bolivia	novo	B253LS	Bolivia
coca	B88LS	Bolivia	novo	B205LS	Bolivia
coca	B14LS	Bolivia	novo	B300LS	Bolivia
coca	B104LS	Bolivia	novo	B201-1LS	Bolivia
coca	B110LS	Bolivia	ipadu	F-1 B508	Beltsville
coca	B56LS	Bolivia	ipadu	F-1 B508	Beltsville
coca	B96LS	Bolivia	ipadu	F-1 B501SS	Beltsville
coca	B180LS	Bolivia	ipadu	F-1 B503SS	Beltsville
coca	B105LS	Bolivia	ipadu	F-1 B508	Beltsville
coca	B63LS	Bolivia	ipadu	F-1 B501SS	Beltsville
coca	B127LS	Bolivia	ipadu	F-1 B504	Beltsville
coca	B31LS	Bolivia	ipadu	F-1 B503SS	Beltsville
coca	B60LS	Bolivia	ipadu	B501SS	Bolivia
coca	B150LS	Bolivia	ipadu	B503SS	Bolivia
coca	B94LS	Bolivia	ipadu	B507	Bolivia
coca	B98LS	Bolivia	ipadu	B504	Bolivia
coca	B9LS	Bolivia	ipadu	B505	Bolivia
coca	B80LS	Bolivia	ipadu	B508	Bolivia
coca	B147RLS	Bolivia	ipadu	F-1 B504	Beltsville
coca	B50LS	Bolivia	ipadu	F-1 B501SS	Beltsville
coca	B64LS	Bolivia	ipadu	F-1 B504	Beltsville
coca	B91LS	Bolivia	ipadu	F-1 B505	Beltsville
novo	B220LS	Bolivia	ipadu	F-1 B505	Beltsville
novo	B276LS	Bolivia	ipadu	F-1 B506	Beltsville
novo	B228-1LS	Bolivia	ipadu	F-1 B505	Beltsville
novo	B216LS	Bolivia	ipadu	BJ IP-23	Bolivia
novo	B245LS	Bolivia	ipadu	BJ IP-24	Bolivia
novo	B294LS	Bolivia	ipadu	BJ IP-25	Bolivia
novo	B225LS	Bolivia	trux	B315SS	Bolivia
novo	B201LS	Bolivia	trux	B319LS	Bolivia
novo	B242LS	Bolivia	trux	B322LS	Bolivia
novo	B223LS	Bolivia	trux	B314SS	Bolivia
novo	B286LS	Bolivia	trux	B318SS	Bolivia
novo	B251LS	Bolivia	trux	B312LS	Bolivia
novo	B206LS	Bolivia	trux	B302LSS	Bolivia
novo	B236LS	Bolivia	trux	B317SS	Bolivia
novo	B272LS	Bolivia	trux	B323LS	Bolivia
novo	B295LS	Bolivia	trux	B316	Bolivia
novo	B244LS	Bolivia	trux	B303LS	Bolivia
novo	B233LS	Bolivia	trux	B321LS	Bolivia
novo	B278LS	Bolivia	trux	B305LS	Bolivia
novo	B-205-1LS	Bolivia	ulei	#1	IIBC

Accession identifications correspond to the labels in the MDS (Fig. 2).

Isolation of DNA from leaf tissue

Genomic DNA (i.e. total DNA) was extracted from leaf tissue of the five taxa using a modification of the Qiagen DNA Stool Mini Kit™ protocol (Qiagen Inc., Valencia, CA, USA). One hundred milligrams f. wt (i.e. 20 mg d. wt.) of leaf tissue were weighed and placed into a 2-mL lysing matrix cylinder containing two 0·53-cm ceramic spheres (QBiogene, Inc., Carlsbad, CA, USA) with 35 mg of polyvinylpolypyrrolidone (Sigma Chem., Co., St Louis, MO, USA). Then 1·4 mL of the buffer ASL was added and the tissue homogenized with a FastPrep 120 tissue homogenizer (Savant Instruments, Holbrook, NY, USA) at a speed setting of 6·5 for 45 s. The sample was then incubated for 5 min in a 70 °C water bath (Lab Line, Barnstead International, Dubuque, IA, USA). During incubation, the sample was mixed several times by inverting the cylinder and then centrifuged at 16 100 g (23 °C) for 10 min (Eppendorf 5415D/R Centrifuge, Hamburg, Germany). The supernatant was transferred to a 1·5-mL microfuge tube containing one InhibitEX™ tablet, vortexed for 1 min and then incubated at 23 °C for 1 min. The sample was centrifuged at 16 100 g (23 °C) for 3 min and the supernatant transferred to a 1·5-mL microfuge tube and the pellet discarded. The sample was then centrifuged for 10 min at 16 100 g (23 °C) and 200 μL of the supernatant transferred to a 1·5-mL microfuge tube containing 15 μL of proteinase K. To this tube, 200 μL of AL lysis buffer were added, the microfuge tube was vortexed, and then incubated at 70 °C for 10 min. Following incubation, 200 μL of 100 % ethanol were added, the sample vortexed and centrifuged as above for 1 min. The lysate was transferred to a QIAamp spin column in a 2-mL microfuge tube and centrifuged (16 100 g (23 °C) for 1 min. The filtrate was discarded and 500 μL of AW1 wash buffer added. It was then centrifuged for 1 min (16 100 g (23 °C) and the QIAamp spin column transferred to a 2-mL microfuge tube. This wash step was repeated twice using 500 µl of AW2 wash buffer. The spin column was transferred to a 1·5-mL microfuge tube and 200 μL of AE elution buffer added. The sample was allowed to stand for 1 min and then centrifuged at 16 100 g (23 °C) for 1 min to elute the DNA. All samples were stored at −80 °C.

DNA quantification

DNA was quantified by fluorimetric analysis using a Fluoroskan Ascent microplate reader using 485/538 nm excitation/emission filter settings (LabSystems, Helsinki, Finland). All samples were diluted 1:20 with Tris-buffered EDTA (TE) and analysed using the PicoGreen dsDNA quantitation kit (Molecular Probes, Eugene, OR, USA) in a 96-well platform plate (Greiner Bio-one, Longwood, FL, USA). Fifty microlitres of PicoGreen solution (1:200 dilution), 2·5 μL of diluted DNA (1:20) and 47·5 μL of de-ionized water were added to each well. A standard curve was generated from DNA standards (PicoGreen kit) ranging from 10 to 500 ng mL⁻¹ on the same plate. Final sample dilution was 1:800 and all measurements were repeated three times.

AFLP analysis

DNA fragments were amplified using a modification of the procedure by Vos et al. (1995). The modification was as follows: template DNA (500 ng) was digested with EcoRI and MseI (New England BioLabs, Beverly, MA, USA) and ligated in a single step to commercial EcoRI and MseI oligonucleotide adapters (Applied Biosystems, Foster City, CA, USA) by incubation overnight at room temperature. Solutions were prepared as previously described, (Johnson et al., 2003b), except as noted below.

The first, preselective amplification of the restricted and ligated fragments utilized commercial EcoRI and MseI AFLP preselective primers and AFLP core mix (both from Applied Biosystems). The thermocycling programme for this amplification was: 94 °C for 3 min, followed by 20 cycles of the following profile: 94 °C for 20 s, 56 °C for 30 s and 72 °C for 2 min with a final hold of 60 °C for 45 min (PTC-200 Peltier Thermal Cycler, MJ Research, Waltham, MA, USA). For the second, selective amplification, primers with a FAM, HEX or NED active ester dye attached to the 5′ end of each EcoRI primer, and non-tagged MseI primers (Applied Biosystems) were used. The products from the preselective amplification were diluted as described previously (Johnson et al., 2003b) and used as templates for the selective amplification. The labelled EcoRI primer and unlabelled MseI primer were both used at a concentration of 0·10 µM. The thermocycling profile was: 94 °C for 2 min, followed by 10 cycles of 94 °C for 20 s, 1 degree per cycle step-down of annealing temperature from 66 °C held for 30 s and 72 °C for 2 min. This was followed by 25 cycles of 94 °C for 20 s, 56 °C for 30 s, 72 °C for 2 min and a final hold at 60 °C for 45 min.

Capillary electrophoresis and genotyping

Capillary electrophoresis of all samples was performed on a MegaBACE™ 500, 48-capillary system (Amersham Biosciences, Piscataway, NJ, USA) equipped with MegaBACE™ Instrument Control software Ver. 2·5 (Amersham Biosciences). Samples were run under the following genotyping parameters: injection voltage, 3 kV; injection time, 45 s; run voltage, 10 kV; run time, 105 min; dyes, GT dye set 2 [ET-Rox (900), FAM, NED and HEX]. Some samples were prepared for analysis by diluting the final amplified product 1:30 (v/v) in 0·1 % Tween-20. All samples included 1 % (v/v) Amersham Bioscience ET-Rox 900 bp DNA size standard. Electropherograms were analysed with Fragment Analyzer™ 1.2 Software (Amersham Biosciences) and all dendrograms with bootstrap values (1000 repetitions) were produced using Free Tree (Pre-released Version 0.9.1.50; © Adam Pavlieek, Tomal Pavlieek and Jaroslav Flegr, 1998–1999) and Tree View (Version 1.6.6; © Roderic D. M. Page, 2001).

Twenty primer pairs were evaluated for the current research. Sixteen primer pairs were successful; however, only the four primer pairs with optimal resolution were selected for this study. A previous study with Erythroxylum revealed that three primer pairs were sufficient to determine relationships in a population containing up to 132 samples (Johnson et al., 2003b).

The fragment data (100–550 nt range) were analysed using Amersham's Fragment Analyzer™ 1.2 Software (Amersham Biosciences). For AFLP analysis, the maximum bin width was 1·00 nt, and no further Y threshold was applied. Each sample was scored for each bin as ‘1’ if a fragment of that size was present, and ‘0’ if not. Fully populated and unpopulated bins were excluded (i.e. peaks present in all or no accessions). In rare cases where two fragments were present in one bin, the bin was scored as ‘1’ for that sample.

Data analysis

For cluster analysis, matrix of pairwise distances (Nei and Li, 1979) between all pairs of individuals, were calculated using Free Tree (Pre-released Version 0.9.1.50; © Adam Pavlieek, Tomal Pavlieek and Jaroslav Flegr, 1998–1999). A neighbour-joining tree was constructed and the bootstrap value for internal branches of the tree was computed with 1000 repetitions. The software Tree View (Version 1.6.6; © Roderic D. M. Page, 2001) was used to draw the tree. Erythroxylum ulei was included as a check in the cluster analysis and is a taxon of Erythroxylum which does not contain the cocaine alkaloid.

For the assessment of within taxon genetic variation, the mean distance (Nei and Li, 1979) of each taxon was calculated using the SAS program of Dubreuil et al. (2002). The mean distance was defined as the average of all pair-wise distances between individuals within each taxon (Nienhuis et al., 1994). The mean distances then were compared using the T-TEST procedure of SAS (SAS, 1999).

For ordination analysis, a matrix of Euclidian distances was calculated using a program written in Visual Basic Macro language embedded in MS Excel. The distances between the 85 accessions were then presented in a two-dimension scaling plot using the Multi-Dimensional Scaling (MDS) procedure of SAS (SAS, 1999). The analysis of molecular variance (AMOVA) procedure, based on the squared Euclidian distance was used to estimate variance components for AFLP genotypes (Excoffier et al., 1992). Variation was partitioned between species, within species and among varieties, within variety and among individuals. The variance components of interest were extracted and tested using nonparametric permutation procedures. Variation between taxa was then partitioned into pair-wise distances between taxa to examine their relative contribution to the total molecular diversity (Excoffier and Smouse, 1994).

RESULTS

AFLP polymorphisms and cluster analysis

The four primer pair combinations used for AFLP analysis generated 1667 fragments for the 85 accessions of which 1585 were polymorphic (94·4 %; Table 2). Between the two species studied, the polymorphic fragments accounted for 98·5 % in E. c. var. coca and 95·3 % in E. n. var. novogranatense (Table 2). Within each individual taxon, the number of polymorphic fragments was significantly smaller than at the inter-specific level, ranging from 386 for E. n. var. truxillense to 403 for E. c. var. ipadu and E. n. var. novogranatense (Table 2). The large number of polymorphic fragments at the inter- and intra-specific level demonstrated that there is a high level of genetic variation across the four cultivated Erythroxylum taxa. This variation is sufficient to permit the assessment of inter- and intra-specific diversity in Erythroxylum using AFLP analysis.

Table 2.

AFLP markers generated among 85 Erythroxylum accessions representing four taxa using four EcoRI + MseI primer pairs

Erythroxylum species	Total no. of markers	No. polymorphic markers	% polymorphic markers
E. c. var. ipadu	427	403	94·4
E. c. var. coca	399	393	98·5
E. n. var. truxillense	418	386	92·3
E. n. var. novogranatense	423	403	95·3
Total	1667	1585	94·4

Clustering of the 85 accessions based on Nei's genetic distance (Nei and Li, 1979) resulted in two distinct clusters at the coefficient level around 0·36 (Fig. 1). The first cluster includes most accessions in E. coca, while the second cluster includes most accessions in E. novogranatense. Within the cluster of E. coca, almost all accessions from the two E. coca varieties grouped within their own taxon, which supports their current systematic status as two different varieties. However, in the cluster of E. novogranatense, a high frequency of overlapping occurred between E. n. var. novogranatense and E. n. var. truxillense (Fig. 1).

Fig. 1.

Dendrogram of 86 Erythroxylum c. var. coca, Erythroxylum c. var. ipadu, Erythroxylum n. var. novogranatense, E. n. var. truxillense and Erythroxylum ulei samples from the living collection at the Beltsville Agricultural Research Center. Accession labels correspond to the sample list in Table 1.

Genetic diversity within and among the four Erythroxylum taxa

Mean genetic distance, as another measure of internal genetic diversity within each taxon, varied significantly across the four varieties (Table 3). While there were no significant differences between E. c. var. coca, E. n. var. novogranatense, and E. n. var. truxillense, the accessions of E. c. var. ipadu had a significantly smaller mean distance than the rest of the three taxa.

Table 3.

Comparison of mean genetic distances of four Erythroxylum taxa

Taxa	No. of individuals	Mean distance among individuals within taxa
Erythroxylum coca var. coca Lam	23	0·347a
Erythroxylum novogranatense var. novogranatense (Morris) Hieron	25	0·358a
Erythroxylum novogranatense var. truxillense (Rusby) Plowman	13	0·331a
Erythroxylum coca var. ipadu Plowman	24	0·302b

^a,bValues in the columns followed by the same letter are not significantly different at the 0·05 level from the other values in the same column.

The MDS plot, based on Euclidian distance, divided the 85 accessions (Erythroxylum ulei was excluded in the ordination analysis) into two main groups, showing a clear distinction between E. c. var. coca and E. n. var. novogranatense (Fig. 2). Within the species of E. coca, the pattern of differentiation between E. c. var. coca and E. c. var. ipadu is clearly shown, although one outlier accession of E. c. var. ipadu, BJIP-23, was closely grouped with E. c. var. coca (Table 1 and Figs 1 and 2). In addition, three accessions of E. c. var. ipadu, BJ IP-24, BJ IP-25 and B507, were located intermediately between E. c. var. ipadu and E. c. var. coca. In contrast to E. coca, the two varieties in E. novogranatense are not distinguishable (Figs 1 and 2). One outlier of E. coca var. coca, accession B147RLS, was close to E. novogranatense and will be discussed later (Table 1 and Figs 1, 2 and 3B).

Fig. 2.

Multidimensional scaling plot of 85 Erythroxylum accessions based on Euclidian distance calculated from AFLP data (MDS badness of fit = 0·229). All accession identifications correspond to sample list in Table 1.

Fig. 3.

Three Erythroxylum accessions from the multidimensional scaling plot: (A) E. n. var. truxillense (B305LS); (B) E. c. var. coca (B147RLS) and (C) E. c. var. coca (B14LS). Note the phenotypical similarity between (A) and (B) and the dissimilarity to (C). B147RLS was shown to be more distanced from B14LS, but taxonomically identified as E. c. var. coca in the living collection at the Beltsville Agricultural Research Center.

The results of AMOVA revealed significant genetic variation hierarchically at the level of inter-species, inter-variety and inter-individual (Table 4). Variation from the three sources accounts for 35·7 %, 13·0 % and 51·3 % of the total molecular variance, respectively. Partitioning the inter-variety variability into pair-wise distances showed the amount that each variety contributes to the total molecular diversity (Table 5). Pair-wise inter-variety distances differed greatly. Within the species of E. novogranatense, the distance between the two varieties, E. n. var. novogranatense and E. n. var. truxillense, was negligible (distance = 0·060), whereas the distance between the two varieties in E. coca was substantial (distance = 0·299; Table 5). This demonstrates that the differentiation between E. c. var. coca and E. c. var. ipadu contributed most to the inter-variety variability in this analysis. The inter-individual variation also differed among the four taxa. E. n. var. novogranatense had the largest mean squared deviation (28·7), whereas E. c. var. ipadu had the smallest one (23·7). E. c. var. coca and E. n. var. truxillense had intermediate mean squared deviation levels of 26·1 and 26·4, respectively (Table 4).

Table 4.

Analysis of molecular variance (AMOVA) for AFLP variation among and within four Erythroxylum taxa

Source of variation	d.f.	SSD*	MSD†	Variance component	% Total‡	P value§
Between species	1	1659·4	1659·4	32·2	35·7	<0·001
Among variety within species	2	569·8	284·9	11·8	13·0	<0·001
Individual within variety	81	3743·3	46·2	27·5	51·3	<0·001
E. c. var. coca	22	959·1	26·1
E. c. var. ipadu	23	928·3	23·7
E. n. var. novogranatense	24	1215·2	28·7
E. n. var. truxillense	12	640·6	26·4
Total	84	5972·6

^*Sum of squared deviations.

^†Mean squared deviations.

^‡Percentage of total molecular variance.

^§Probability of obtaining a larger component estimate. Number of permutations = 1000.

Table 5.

Genetic distances between the cultivated Erythroxylum taxa

Name of taxa	Genetic distance
E. c. var. coca vs. E. c. var. ipadu	0·299
E. c. var. coca vs. E. n. var. novogranatense	0·451
E. c. var. coca vs. E. n. var. truxillense	0·459
E. c. var. ipadu vs. E. n. var. novogranatense	0·510
E. c. var. ipadu vs. E. n. var. truxillense	0·519
E. n. var. novo. vs. E. n. var. truxillense	0·060

DISCUSSION

The neotropical taxa of Erythroxylum have been cultivated in South America for at least 5000 years. While intensive ethnobotanical studies have been conducted on these taxa (Plowman, 1984), the taxonomic status, evolutionary relationship, and organization of genetic diversity have been poorly understood. Various taxonomic treatments have been proposed for the cultivated cocaine-bearing Erythroxylum, ranging from one to three separate species (Plowman, 1979, 1981, 1982, 1984; Bohm et al., 1982). The recent treatment recognizes two species (Erythroxylum coca var. coca Lam and Erythroxylum novogranatense var. novogranatense (Morris) Hieron) each with a variety (Erythroxylum coca var. ipadu Plowman and Erythroxylum novogranatense var. truxillense (Rusby) Plowman (Plowman, 1981; Bohm et al., 1982; Johnson, 1997, 1998, 2002, 2003b). All four of these taxa have the same chromosome number, n = 12 and are known only in cultivation, or as semi-wild escapees from cultivation (Plowman, 1979). Based on the results of artificial hybridization and the comparison of flavonoids in the four taxa, Bohm et al. (1982) proposed a hypothesis of a linear evolution series suggesting that (a) E. n. var. novogranatense and E. n. var. truxillense are two varieties of a species distinct from E. coca; (b) E. c. var. ipadu was independently derived from E. c. var. coca; (c) E. c. var. coca is the ancestral taxon from which E. n. var. truxillense was derived; and (d) E. n. var. novogranatense was derived from E. n. var. truxillense.

In the present study, 86 accessions were genotyped from the living collection of Erythroxylum at the Beltsville Agricultural Research Center using AFLP, and this result provided insight into the taxonomic relationships and evolution. Overall, the pattern of genetic variation in these taxa agrees in principle with known geographic and phenotypic data. There is a clear separation between E. coca and E. novogranatense (Table 5 and Figs 1 and 2), thus supporting the current taxonomic treatment recognizing E. coca and E. novogranatense as two different species. However, the level of genetic diversity present in E. novogranatense is as high as that of E. coca, as quantified by the mean genetic distance and the mean squared deviation from AMOVA (Tables 3 and 4). This high level of diversity in E. novogranatense, together with its significant geographic and genetic separation from E. coca (Fig. 2) suggests that E. novogranatense may not be derived from E. coca. The geographical distribution of these two taxa is completely allopatric (Plowman, 1979). The ecological habitat of E. novogranatense is predominately in northern Peru, Ecuador, Colombia, and Venezuela, ranging from desert to moist forest, whereas E. coca inhabits the pre-montané wet forest of Ecuador, Peru and Bolivia, as well as the lowland rain forest of the Amazonian Basin (Ganders, 1979; Schultes, 1981; Plowman, 1982, 1984). This geographical barrier between the two taxa suggests that E. novogranatense may have an independent origin from E. coca, as suggested by Bohm et al. (1982), but later refuted in favour of the E. coca lineage.

At the inter-variety level, a significant differentiation was detected between E. c. var. ipadu and E. c. var. coca (Table 5 and Fig. 2). Erythroxylum c. var. ipadu is considered to be a cultigen of E. c. var. coca based on their morphological similarities and identical flavonoid profiles (Bohm et al., 1982). The significantly lower level of genetic diversity in E. c. var. ipadu, relative to that in E. c. var. coca, supports the current view that E. c. var. ipadu is a derived cultigen of E. c. var. coca. Its lower diversity, in conjunction with its clear separation from E. c. var. coca, is most likely to be the result of a founder effect as E. c. var. ipadu moved from the eastern Andes in Peru and Bolivia into the lowland Amazonian basin. Since E. c. var. ipadu can only be propagated vegetatively, it is unable to reproduce in nature (Plowman, 1979, 1982, 1984). This cultigen was traditionally cultivated under conditions of shifting ‘slash and burn’ agriculture by Amazonian tribes, where cultivation plots were shifted annually or biennially (Plowman, 1982). This continued artificial breeding and migration is likely to result in future genetic drift and population isolation.

Within E. novogranatense, the rational of recognizing E. n. var. truxillense as a different variety is questionable. As shown in the MDS plot and the result of AMOVA (Table 5 and Fig. 2), the genetic distance between E. n. var. novogranatense and E. n. var. truxillense is negligible. The two varieties are known to have similar morphological characters and crosses between the two varieties are capable of producing morphologically normal and fully fertile F₁ hybrids (Bohm et al., 1982). Therefore, main genetic differences that may justify the separation of E. n. var. truxillense from E. n. var. novogranatense is that the former has better adaptability to drier ecological habitats in northern Peru and southern Ecuador, as well as the distinct leaf flavonoids (Johnson et al., 1997, 1998, 2002, 2003a).

While there are distinguishable AFLP profiles for the cocaine-bearing Erythroxylum, both at species and variety level, it is noteworthy that some outlier accessions do not fit into the typical variety or species proximity. This happened mostly in the two varieties of E. coca, which may have practical implications for monitoring the coca production in the Andes. The E. c. var. coca accession B147RLS is grouped toward E. novogranatense (Figs 1, 2 and 3B). This is of interest because, in their breeding experiment, Bohm et al. (1982) showed that crosses between E. c. var. coca and E. n. var. novogranatense were incompatible, while those between E. c. var. coca and E. n. var. truxillense were compatible. The accession B147RLS, therefore, is likely a hybrid between E. c. var. coca and E. n. var. truxillense (Table 1 and Figs 1–3).

A similar outlier pattern, as described above, is shown for E. c. var. ipadu accessions B507, BJIP-23, BJIP-24 and BJIP-25 where they appear as intermediates between E. c. var. coca and E. c. var. ipadu (Table 1 and Figs 1 and 2). According to Bohm et al. (1982), where there are barriers to out-crossing, hybrid weakness and reduced hybrid fertility have occurred and three of the four cultivated coca taxa have become genetically isolated from each other (E. c. var. coca, E. n. var. novogranatense and E. n. var. truxillense). Bohm (1982) and Johnson et al. (1997, 1998, 2002, 2003a; Johnson and Schmidt, 1999) characterized the leaf flavonoids of the four cultivated Erythroxylum taxa. Johnson et al. (1997, 1998, 2002, 2003a) and Johnson and Schmidt (1999) showed that each taxon possessed distinct leaf flavonoids, as well as unique morphological features. Johnson et al. (2002) further compared the leaf flavonoids in E. c. var. ipadu (collected about 1970) from the living greenhouse collection at the Beltsville Agricultural Research Center with those of E. c. var. ipadu currently under cultivation in Colombia and found that they differed greatly. The flavonoids of the greenhouse-grown E. c. var. ipadu were similar to those present in E. c. var. coca (Johnson et al., 1998), indicating that the former should be derived from the latter, whereas the E. c. var. ipadu currently under cultivation in Colombia contained a mixture of the flavonoids found in both E. c. var. coca and E. n. var. truxillense, as well as a new found flavonoid not present in E. c. var. coca or E. c. var. ipadu in the living collection at the Beltsville Agricultural Research Center. This suggested that the material presently in cultivation has resulted in a hybrid from a cross between the E. c. var. coca and E. n. var. truxillense (Johnson et al., 2002, 2003a). The leaf flavonoid data coincides with the genetic distance data in the current study (Table 5 and Fig. 2), wherein B147RLS is representative of E. c. var. ipadu currently under cultivation in Colombia (Table 1 and Figs 1 and 2). The morphology of B147RLS is more similar to that of E. n. var. truxillense than to that of E. c. var. coca (Fig. 3) which would account for its genetic intermediacy (Fig. 2). The separation from other accessions of the taxon may not be explained by population differentiation and it is likely that this accession represents a hybrid between E. c. var. coca and E. n. var. truxillense, as was also suggested by the authors' previous flavonoid data (Johnson et al., 2002).

It was stated above that BJIP-23, BJIP-24 and BJIP-25 are intermediates between E. c. var. coca and E. c. var. ipadu (Table 1 and Figs 1 and 2). However, BJ IP-23 is separated from BJIP-24 and BJIP-25 and grouped within E. c. var. coca (Fig. 2). Hitherto, there are no reported crosses between E. c. var. coca and E. c. var. ipadu, and Bohm et al. (1982) concluded that preliminary crosses between E. c. var. ipadu were self-incompatible. At the Hawaiian field site, it was observed that E. c. var. ipadu F₁ progeny clustered (AFLP) well with E. n. var. novogranatense (Johnson et al., 2003b). This may indicate that BJ IP-23 is a cross between E. c. var. ipadu and E. n. var. novogranatense. Therefore it is distanced between E. c. var. coca and E. n. var. novogranatense (Fig. 2). In future research it is intended to examine the relationship between the parental, F₁ and F₂ progeny in crosses between the four cultivated Erythroxylum taxa.