American Journal of Botany 88(10): 17501759. 2001.

POPULATION

GENETIC STUDY IN THE

BALEARIC

DIGITALIS MINOR
(SCROPHULARIACEAE) USING RAPD MARKERS1
ENDEMIC PLANT SPECIES

ESTER SALES,2 SERGIO G. NEBAUER,3 MAURICI MUS,4

JUAN SEGURA2,5

AND

Departamento de Biologa Vegetal, Universidad de Valencia, Valencia, Spain;

Instituto Valenciano de Investigaciones Agrarias, Moncada (Valencia), Spain; and
4
Departamento de Biologa Ambiental, Universidad de las Islas Baleares, Mallorca, Spain
2

Digitalis minor (Scrophulariaceae) is a cardenolide-producing plant endemic to the eastern Balearic Islands (Mallorca, Menorca,
and Cabrera) that occurs in two morphologically distinct varieties: D. minor var. minor (pubescent) and D. minor var. palaui (glabrous).
Levels and patterns of genetic diversity in 162 individuals from 17 D. minor populations across the entire geographic range were
assessed using random amplified polymorphic DNA (RAPD) markers. Comigrating RAPD fragments tested were found to be homologous by Southern hybridization in both var. minor and var. palaui. To avoid bias in parameter estimation, analyses of population
genetic structure were restricted to those RAPD bands that fulfilled the 3/N criterion (observed frequencies were less than 1 2 [3/N]
in each population) either among or within each island. Analyses of molecular variance (AMOVAs) with distances among individuals
corrected for the dominant nature of RAPD (genotypic analysis) showed low values (1.5717.55%) of between-population variability,
indicating a relatively restricted population differentiation as expected for an outcrossing species such as D. minor. Nested AMOVAs
demonstrated, however, a not significant partitioning of genetic diversity among Mallorca, Menorca, and Cabrera islands. Estimates of
the Wright, Weir, and Cockerham and the Lynch and Milligan FST from null allele frequencies corroborated AMOVA partitioning and
provided evidence for population differentiation in D. minor. Our RAPD data did not show significant differences between pubescent
and glabrous populations of D. minor, suggesting a failure to find a correlation between the RAPD loci and this morphological trait.
Key words:

AMOVA; Digitalis minor; endemic species; population structure; RAPDs; Schrophulariaceae.

The genetic structure of plant populations reflects the interactions of different processes including long-term evolutionary
history of the species (shifts in distribution, habitat fragmentation, and population isolation), mutation, genetic drift, mating system, gene flow, and selection (Slatkin, 1987; Schaal et
al., 1998). All these factors can lead to complex genetic structuring within populations, which is often difficult to resolve.
Nevertheless, the development of a number of different DNA
markers has provided powerful tools for the investigation of
genetic variation within a species and can facilitate understanding of such complexities (Mitton, 1994).
Random amplified polymorphic DNA (RAPD) technology
via the polymerase chain reaction (PCR) has fast become a
means of investigating genetic diversity within and between
populations and has been applied to many plant species including Digitalis (see Nebauer, Del Castillo Agudo, and Segura, 1999, 2000, and references therein). In spite of this, the
various statistics used to estimate and partition genetic variation in natural populations cannot be applied easily to RAPD
data obtained from outcrossing species because complete genotypic determination is largely hampered by their dominant
nature (Isabel et al., 1999). During recent years, however, several strategies have been proposed (Lynch and Milligan, 1994;
Manuscript received 28 November 2000; revision accepted 8 March 2001.
The authors thank DGICYT, Madrid, Spain (Project PB96-0789) for financial support; the Spanish Ministerio de Educacion y Cultura for two FPI Research Fellowships to Ester Sales and Sergio G. Nebauer; the Parque Nacional
de Cabrera for facilities when sampling populations from this island; and Dr.
J. A. Rossello for valuable comments on an earlier draft of this manuscript.
5
Author for reprint requests: Departamento de Biologa Vegetal, Facultad
de Farmacia, Universidad de Valencia, Avda. Vicent A. Estelles s/n. Burjassot
46100, Valencia, Spain (phone: 134 963 864 922; FAX: 1 34 963 864 926;
e-mail: juan.segura@uv.es).
1

Apostol et al., 1996; Stewart and Excoffier, 1996) to minimize

the effects of RAPD dominance. Of all these approaches, the
pruning of fragments with low frequency of null alleles in each
population (Lynch and Milligan, 1994) resulted in FST values
concordant with those estimated from haploid sexual tissues
(Isabel et al., 1999) and codominant allozyme markers (Aagaard, Krutovskii, and Strauss, 1998; Jenczewski, Prosperi,
and Ronfort, 1999). Furthermore, simulation studies on the
effects of dominance using allozyme data suggested that
RAPDs can reasonably estimate population differentiation
(Wu, Krutovskii, and Strauss, 1999). This also holds true when
the FST value was estimated from the analysis of molecular
variance (Excoffier, Smouse, and Quattro, 1992), although
simulations were not performed due to the complexity of calculations (Isabel et al., 1999).
The Mediterranean flora has long been of interest to biologists because of their high level of endemic species richness
and complicated patterns of community organization (Cardona
and Contandriopoulos, 1979). In spite of this, only limited
information is available on the genetic structure of endemic
Mediterranean plant species (for review, see Thompson, 1999).
This is unfortunate because such information is crucial for
devising strategies to protect and preserve the genetic resources of the Mediterranean flora.
Digitalis minor L. (Synonym: Digitalis dubia J.J. Rodr.), of
the family Scrophulariaceae, is an outcrossing insect-pollinated long-lived herbaceous species with a chromosome number
of 2n 5 56 (Contandriopoulos and Cardona, 1984). Although
self-incompatibility mechanisms have not been described for
D. minor, floral dichogamy (protandry) promotes cross-fertilization of this species (Kampny, 1995). Digitalis minor shows
a high degree of morphological polymorphism, but only two

1750

October 2001]

SALES

ET AL.RAPD ANALYSIS IN

DIGITALIS MINOR

1751

widely used in some heart-failure diseases. Thus, an understanding of the extent and distribution of genetic variation
within D. minor populations is also essential for devising sampling strategies, which efficiently capture genetic diversity for
selection trials and subsequent use of material that fulfils the
dual aim of high genetic variation and reasonable performance.
In the present study the genetic diversity in 17 populations
of D. minor encompassing the natural distribution of the species in the Mallorca, Menorca, and Cabrera islands was examined using RAPD. The major objectives were to quantify
the amount and distribution of genetic variation within and
among these populations using genetic diversity measures, F
statistics, and spatial-correlation statistics. The possibilities of
using the RAPD markers to differentiate the two infraspecific
taxa of D. minor were also tested.
Fig. 1. Location of the 17 collected natural populations of D. minor included in the study (see Table 1 for population abbreviations). m, D. minor
var. minor populations; n, D. minor var. palaui populations.

infraspecific taxa are currently recognized according to its differences in pubescence (Hinz, 1987): D. minor var. minor (pubescent) and D. minor var. palaui (glabrous).
Digitalis minor is an endemic species to the Balearic archipelago (Spain). This species is believed to be an schizoendemic vicariant of D. purpurea ssp. purpurea (Contandriopoulos
and Cardona, 1984) and occurs on three islands of eastern
Balearics or Gymnesian Islands (Mallorca, Menorca, and Cabrera). According to paleohistory and present-day geography
of the western Mediterranean basin, the differentiation of D.
minor seems to be of pre-Messinian origin (Hinz, 1990). The
isolation of the Gymnesian Islands, which constituted on several occasions a single landmass, was probably a more recent
event (Cardona and Contandriopoulos, 1979), since it would
date from the last glaciation of the Pleistocene Era (the
Wurm).
To our knowledge, there is no report on the application of
molecular markers to study the population structure of D. minor. As other Digitalis species, it contains cardiac glycosides,
TABLE 1.
Origin

MATERIALS AND METHODS

Plant materialThe study was undertaken on 17 populations of Digitalis
minor L. (Scrophulariaceae). Sampling sites were chosen to give a good representation of the distribution areas of the species in the Mallorca, Menorca,
and Cabrera islands (Fig. 1) of the Balearic archipelago (Spain). The number
of individuals sampled and census estimates for each population are shown
in Table 1. Individuals were chosen at random and sampled directly using
silica gel to dry and preserve leaf material until DNA extraction.
DNA extraction and amplificationGenomic DNA extractions and PCR
amplifications were carried out as described in Del Castillo-Agudo et al.
(1995) and Nebauer, Del Castillo Agudo, and Segura (2000), respectively.
RAPD products were separated by size on 1.5% agarose (SeaKem LE, FMC
Bioproducts, Rockland, Maine, USA) gel run in TBE buffer (89 mmol/L Tris
base, 89 mmol/L boric acid, and 2 mmol/L EDTA, pH 8), stained with ethidium bromide, and photographed in UV light (Nebauer, Del Castillo Agudo,
and Segura, 2000). To aid interpretation of band identity between gels, each
contained Gene Ruler DNA ladder mix (Fermentas AB, Vilnius, Lithuania).
Primers were initially screened to identify well-amplified, polymorphic
bands among populations. Individual Digitalis DNA samples were appropriately diluted and bulked by populations to screen 60 decamer primers (Series
OPA, OPB, and OPC from Operon Technologies, Alameda, California, USA).
Six primers from the initial screening process (OPA2, OPA20, OPC9, OPC14,

Sampled populations of Digitalis minor with abbreviations, geographical location, and distances between populations.
Population

Infrasp. taxon

Longitude
(E)
Latitude (N)

Population
size

Mallorca
Deia` (D)
Gorg Blau (G)
Ses Ortigues (O)
Torrent de Mortitx (T)
Puig Galatzo (P)
Son Fortuny (Y)
Formentor (F)
Coma de nArbona (C)

minor
minor
palaui
minor
palaui
palaui
minor
minor

28399
28499
28259
28549
28299
28279
38109
28479

398449
398489
398379
398529
398389
398389
398569
398479

3040
2535
3040
100120
5070
20
100
3040

Sa Torre Nova (MS)

Llinaritx Nou (ML)
Sa Mesquida (MM)
Binialcala` (MB)
Fava`ritx (MX)
Ermita Ferreries (ME)
El Toro (MT)

minor
palaui
minor
palaui
minor
palaui
minor

48069
48049
48179
48049
48159
48019
48069

398529
398599
398549
408019
398599
398599
398599

30
20
2030
3040
5070
8
70100

Punta Coll Roig (CP)

Cala Galiota (CG)

minor
minor

28559
28559

398089
398099

3040
1520

Menorca

Cabrera

Sample
size

86
14
10
12
12
12
11
6
9
57
9
6
9
10
6
8
8
19
8
11

Distances between populations (km)

16.0
23.5
29.0
19.0
19.0
50.5
12.0
MS

39.5
13.0
34.5
35.0
34.0
4.0
ML

52.5
6.0
5.0
74.0
35.5
MM

47.0
48.0
21.0
17.0
MB

3.0
68.0
30.5
MX

69.0
31.0
ME

38.0

20.7
2.6
16.7
4.3
3.6

22.0
10.0
24.6
17.2

16.8
5.7
5.0

21.3
13.3

8.0

14.0
16.3
16.3
19.3
15.0
12.7
CP
1.5

1752

AMERICAN JOURNAL

and OPC15 from Operon, and OPB7 from Amersham Iberica, Madrid, Spain)
that exhibited a high polymorphism and showed the best readability were
chosen for further study of the 162 individual genotypes.
Reproducibility and repeatability of amplification profiles were tested for
each primer. Control samples containing all reaction material except DNA
were used to test that no self-amplification or DNA contamination occurred.
Only those bands consistently reproduced in different analyses were considered. Poor amplifications occurred systematically with individuals from different populations; these were excluded from the analysis and they mainly
account for the different sample sizes of this study. At least two replicates
per sample were amplified and DNA from all the individuals was extracted
twice.
Southern blot of PCR productsTo check that the same-size products
were homologous among individuals from different populations, hybridizations on Southern transfer of some RAPD gels were performed as described
by Nebauer, Del Castillo Agudo, and Segura (2000).
Statistical analysisSince RAPD markers are dominant, we assumed that
each band represented the phenotype at a single biallelic locus (Williams et
al., 1990). Amplified fragments, named by the primer used and the molecular
mass in base pairs (bp), were scored as presence (1) or absence (0) of homologous bands, and a matrix of the different RAPD phenotypes was assembled. Three different types of data were used for the analyses performed in
this study for RAPDs: (1) total scorable bands (hereafter referred to as data
sets Baleares141, Mallorca141, Menorca141, and Cabrera141); (2) bands that
fulfilled the Lynch and Milligan (1994) criterion (those whose observed frequency in each population is less than 1 2 [3/N], where N is the number of
sampled plants in the population) among the three islands (hereafter referred
to as data sets Baleares39, Mallorca39, Menorca39, and Cabrera39); and (3)
bands that fulfilled the Lynch and Milligan criterion within each island (hereafter referred to as Mallorca52, Menorca35, and Cabrera22 data sets; see Table
3). When appropriate, and assuming that all populations are in Hardy-Weinberg equilibrium, we estimated qj(i), the frequency of the null allele a at locus
i in population j, as: qj(i) 5 [xi(i)]1/2, where xi(i) is the frequency of null
recessive homozygotes in population j at locus i. The frequencies were computed in RAPDBIOS from RAPDPLOT (Black, 1998) using Lynch and Milligans (1994) correction factor for small sample sizes.
Pairwise distance matrices were computed based on both the Euclidean
metric (Excoffier, Smouse, and Quattro, 1992) and Apostol (Apostol et al.,
1993) coefficients using RAPDistance (Armstrong et al., 1996) software. A
Mantel test (1967) showed a complete correlation (r 5 1) when both matrices
were compared in NTSYS-pc (Rohlf, 1997). The former distance was chosen
due to its adequacy for the analysis of molecular variance (see below) and
the latter one because it uses both presence and absence of matches, providing
more information regarding the phenotypic similarity among pairs of individuals (Apostol et al., 1993).
The relationships among RAPD phenotypes were assessed as follows. First,
the Apostol distance matrices were used to produce dendrograms using the
neighbor-joining (NJ) cluster analysis as implemented in NEIGHBOR from
the PHYLIP 3.57c package (Felsenstein, 1993). To give a measure of the
variability in the data, bootstrap analysis was conducted and 100 similarity
matrices were produced using RAPDPLOT. The NEIGHBOR and CONSENSE programs in PHYLIP were used to generate the 100 trees that were
then used to produce a consensus tree. Principal coordinate analyses (PCO)
were also performed, using the Apostol distance matrices (DCENTER and
EIGEN in NTSYS).
The distance matrices between RAPD patterns were also used to calculate
pairwise genetic distances between populations. These distance matrices were
used to construct dendrograms using the NJ method. The relationships between matrices of genetic and linear geographic distances were examined with
a Mantel (1967) test in NTSYS. Resulting r values were interpreted as correlation coefficients. Additionally, and for comparative purposes, Neis unbiased genetic distances (Nei, 1978) were calculated among populations for
each of our data sets with POPGENE software (Yeh et al., 1997). All den-

OF

BOTANY

[Vol. 88

drograms were displayed and printed using TREEVIEW software (Page,

1996).
Two different analyses of molecular variance (AMOVA) were performed
to study the genetic structure of D. minor populations. First, the selected bands
were analyzed directly as phenotypes using the Euclidean distance matrix of
Excoffier, Smouse, and Quattro (1992). Second, we used the extension of the
AMOVA developed by Stewart and Excoffier (1996), which allows the estimation of population genetic parameters at the genotypic level with RAPD
profile data. In this case, the previously estimated null allele frequencies were
used to generate a genotypic distance matrix assuming random mating (S 5
0). The nested AMOVAs were used to estimate the partition of total genetic
diversity in variance components among individuals within populations,
among populations, and among regions (islands). AMOVAs were also performed for the individuals within each of the three islands: Mallorca, Menorca, and Cabrera. The AMOVA procedure was finally implemented to study
the relationships between the two infraspecific taxa of D. minor. The whole
set of bands as well as the data sets that fulfilled the Lynch and Milligan
criterion were employed for these last analyses. The resulting variance components were used as estimates of the genetic divergence among the taxa.
A nonparametric test for homogeneity of molecular variance (HOMOVA),
based on the Barlett statistic (Barlett, 1937), was also applied to test whether
populations have different amounts of RAPD variation (see Stewart and Excoffier, 1996). Both HOMOVA and AMOVA analyses were performed using
the WINAMOVA 1.5 program (available from L. Excoffier, Genetics and
Biometry Laboratory, University of Geneva, Switzerland).
The WINAMOVA program extracts analogs of F statistics (so-called F
statistics). Then, and for comparative purposes, we used F statistics as a second approach to study population genetics in D. minor. FST was estimated
from those RAPD bands that fulfilled Lynch and Milligans assumptions for
the analysis of dominant markers using RAPDFST from RAPDPLOT (the
corresponding statistical methods and equations are given in Apostol et al.,
1996). This program estimates FST according to Lynch and Milligan (1994),
Wright (1951), and as u (Weir and Cockerham, 1984). When appropriate, a
subsequent chi-square value was calculated to determine whether these estimates of FST varied from zero (significant population differentiation). Pairwise
FST between populations were also calculated using RAPDDIST from RAPDPLOT, applying the Lynch and Milligan (1994) correction when estimating
allele frequencies. The resulting distance matrices were compared to test their
correlation with the FST distance matrices from AMOVAs, as described
above.

RESULTS
Homology of RAPD markersBands used as digoxigeninlabelled probes were either monomorphic for the 162 sampled
individuals (OPC9-1180, isolated from individual O2) or almost exclusive of Menorca populations (OPC9-700, isolated
from individual MM5). The RAPD gels of 12 individuals from
var. palaui (Y2, O2, P3, MB1, ML3, and ME1) and var. minor
(F3, C2, T4, MS7, MM5, and MX3) populations were selected
for Southern blots (Fig. 2A). Probe OPC9-1180 hybridized
with similar-size fragments in 12 individuals (Fig. 2B), while
probe OPC9-700 recognized only co-emigrating fragments in
the individuals from Menorca (Fig. 2C). Thus, it appears that
the same-size RAPD markers can be treated as homologous
between the two infraspecific taxa of D. minor.
Fingerprinting of D. minor populationsThe six selected
primers generated a total of 141 consistently well-amplified
bands, ranging in size from 420 to 3200 bp. Most of these
bands (96%) were polymorphic among the 17 populations (Table 2), and none of the 162 individuals was characterized by
the same RAPD phenotype. Figure 2A illustrates a typical example of the band pattern generated with the OPC9 primer.

October 2001]

SALES

ET AL.RAPD ANALYSIS IN

DIGITALIS MINOR

1753

Fig. 2. (A) RAPD profiles in D. minor var. minor (lanes 5, 6, 7, 11, 12, 13) and in D. minor var. palaui (lanes 2, 3, 4, 8, 9, 10) generated by primer OPC9.
In the two lines marked with arrows, fragments corresponding to individuals O2 (lane 3) and MM5 (lane 12) were used as probes. (B) and (C) Southern blots
of the RAPDs profiles showing the hybridization of the probes O2-1180 (B) and MM5-700 (C) with all fragments of similar size. Lane M, Gene Ruler DNA
ladder mix.

TABLE 2. Primers employed and RAPD markers obtained for the 162
individuals studied.
Number of bands
Primers

OPA2
OPA20
OPB7
OPC9
OPC14
OPC15

Sequence (59 39)

TGCCGAGCTG
GTTGCGATCC
GGTGACGCAG
CTCACCGTCC
TGCGTGCTTG
GACGGATCAG

Size (bp) min-max

4802900
6302300
5401900
5503200
4202300
5003000

Polymor- Monophic morphic

22
24
20
19
29
22

2
0
1
2
0
0

Total

24
24
21
21
29
22

The distribution of RAPD bands by geographical location is

shown in Table 3.
Population-specific bands were found for Son Fortuny
(OPB7-1050, OPC14-1600) and Formentor (OPC14-2100,
OPC15-1700) in Mallorca, for Sa Torre Nova (OPC15-1180
and -1300), Llinaritx (OPA2-1290), Sa Mesquida (OPC93000), Ermita Ferreries (OPC14-590), and Favaritx (OPC93200) in Menorca; and for Punta Coll Roig (OPA2-620) and
Cala Galiota (OPA20-1020) in Cabrera. Genotype-specific
bands were only detected in Llinaritx (ML1: OPA2-840), Deia`
(D6: OPC14-610), and Favaritx (MX4: OPC9-550, -580, and
-610). Differential band patterns between var. minor and var.
palaui individuals were not found among the RAPD profiles.

1754
TABLE 3.

AMERICAN JOURNAL

OF

BOTANY

[Vol. 88

Distribution of RAPD markers by geographical location. L and M, bands that fulfilled the Lynch and Milligan criterion.
Number of bands (%)

Island

Population

Mallorca
Deia` (D)
Gorg Blau (G)
Ses Ortigues (O)a
Torrent de Mortitx (T)
Puig Galatzo (P)a
Son Fortuny (Y)a
Formentor (F)
Coma de nArbona (C)
Menorca
Sa Torre Nova (MS)
Llinaritx Nou (ML)a
Sa Mesquida (MM)
Binialcala` (MB)a
Fava`ritx (MX)
Ermita Ferreries (ME)a
El Toro (MT)
Cabrera
Punta Coll Roig (CP)
Cala Galiota (CG)
Gymnesian
a

Total

122
81
85
76
77
72
79
75
77
121
73
83
77
68
78
78
79
88
73
76
141

Monomorphic

11
30
37
35
33
27
28
39
28
14
33
57
34
40
42
49
37
39
51
50
5

(9)
(37)
(44)
(45)
(41)
(37)
(36)
(52)
(37)
(11)
(43)
(69)
(46)
(58)
(53)
(63)
(47)
(43)
(69)
(64)
(4)

Polymorphic

111
51
48
41
44
45
51
36
49
107
40
26
43
30
26
29
42
49
22
26
136

(91)
(63)
(56)
(55)
(59)
(63)
(64)
(48)
(63)
(89)
(57)
(31)
(54)
(42)
(47)
(37)
(53)
(57)
(31)
(36)
(96)

L and M

52
39
37
29
34
33
36
22
34
35
25
15
28
18
22
22
27
22
16
18
39

(43)
(48)
(45)
(38)
(46)
(46)
(44)
(30)
(44)
(39)
(36)
(18)
(34)
(25)
(28)
(26)
(35)
(26)
(22)
(24)
(28)

D. minor var. palaui populations.

Relationship among RAPD phenotypesFirst, the 141

RAPD phenotypes (the Baleares141 data set) were analyzed
using the Apostol coefficient. The unrooted NJ dendrogram
based on these data clustered the 162 individuals within distinct groups according to their geographical origin. Three major clusters were formed, each grouping individuals from Mallorca, Menorca, and Cabrera, respectively, and all populations
formed subclusters of their own. The bootstrapped (100 replications) dendrogram confirmed the reliability of the differences among islands and the grouping of individuals within
their own population (data not shown). This also holds true
for the NJ dendrograms performed separately with the 86 individuals from Mallorca, the 57 from Menorca, and the 19
from Cabrera (the Mallorca141, Menorca141, and Cabrera141
data sets), although the individual D9 clustered apart of the
other 13 samples from Deia` population in Mallorca. It is worth
noting that the tree from the Mallorca141 data set revealed a
clear differentiation between individuals from var. minor and
var. palaui populations. In contrast, the tree from the Menorca141 data set showed that individuals could not be separated
as expected from their morphological characters (data not
shown).
Individual plants were also separated by their area location
using PCO of the Baleares141 data set. The first three coordinate axes accounted for 25.0, 10.5, and 8.4% of the total
variance, respectively, and identify three main groups corresponding roughly to Mallorca, Menorca, and Cabrera populations (data not shown). The results showed similar clustering
to the NJ analysis and confirmed the grouping of individuals
from var. palaui populations in Mallorca, probably due to the
more isolated location of these populations (Fig. 1).
When the NJ analysis was restricted to those bands whose
observed frequency was ,1 2 3/(N) among the three islands
(Baleares39 data set), there was no grouping of the individuals
into three different geographic regions (islands). Although
some individuals from the same population appeared in the
same subcluster, the general trend was for individuals to be

distributed throughout several subclusters, along with those

from other populations (data not shown). This lack of concordance between individuals and their geographical location was
also observed when NJ cluster analyses were performed using
the Mallorca52, Menorca35, and Cabrera22 data sets. In the
three cases, RAPD phenotypes were intermingled in different
parts of the dendrograms (data not shown). Trees from Baleares39, Mallorca52, and Menorca35 data sets also failed to
differentiate individuals from var. palaui and var. minor populations.
Divergence at the population levelFirst, interpopulation
distances were calculated using the Apostol coefficient based
on Baleares141 data set (data not shown). The mean distance
among the 17 populations was 0.1799 6 0.0327 (0.1095
0.2553). The NJ dendrogram obtained from this matrix (Fig.
3A) revealed a good grouping of the D. minor populations
into three main clusters, each one corresponding to their respective island location (Mallorca, Menorca, and Cabrera; see
Fig. 1). Note that within the main clusters, var. palaui populations formed a distinctive subcluster only in Mallorca, corroborating results from PCO analysis. Interpopulation genetic
distances were lower when the Baleares39 data set was employed (mean distance 0.0655 6 0.0467) and the corresponding NJ dendrogram did not show any relation between populations and their geographic region (data not shown).
The matrices of interpopulation distances, obtained for each
island separately, using the Apostol coefficient (data sets Mallorca141 and 52, Menorca141 and 35, and Cabrera141 and 22)
are shown in Table 4. The Mallorca and Menorca NJ dendrograms, constructed from the Mallorca141 and Menorca141
data sets, showed a tendency for the populations to cluster
according to their geographical location, especially in Mallorca
(data not shown). This, however, did not hold true when the
NJ dendrograms were constructed from the Mallorca52 (Fig.
3B) and Menorca35 (Fig. 3C) data sets.
To investigate a possible association between genetic relat-

October 2001]

SALES

ET AL.RAPD ANALYSIS IN

DIGITALIS MINOR

1755

0.0388 or 0.9904 6 0.0053 when either Baleares141 or Baleares39 data sets were employed. Similar values were obtained
when the Nei identity was averaged within the 11 var. minor
populations or the 6 var. palaui populations.
Population genetic structureThe analyses were exclusively based on those RAPD bands (39 for the 17 sampled
populations or 52, 35, and 22 for the 8, 7, and 2 sampled
populations from Mallorca, Menorca, and Cabrera islands, respectively) that fulfilled the Lynch and Milligan (1994) condition for obtaining unbiased estimates of population-genetic
parameters.

Fig. 3. Neighbor-joining trees of mean interpopulation distances inferred

from matrices calculated using the Apostol coefficient with the Baleares141
(A), Mallorca52 (B), and Menorca35 (C) data sets. * D. minor var. palaui
populations.

edness and the geographical origin of sampling sites, we compared, separately for Mallorca and Menorca, the matrices of
Apostol genetic distances (Table 4) with the corresponding
matrices of geographical distances (Table 1). Mantels test
showed that these matrices were correlated in Mallorca, but
only when the 141 RAPD phenotypes (the Mallorca141 data
set) were analyzed (r 5 0.47, P 5 0.01). The remaining pairwise genetic distances, constructed from the Menorca141 data
set or RAPD phenotypes that fulfilled Lynch and Milligan criterion in each island, were not correlated with the actual geographical distances between populations (r 5 0.09, P 5 0.41
for the Menorca141 data set; r 5 0.02, P 5 0.50 for the
Mallorca52 data set; and r 5 0.39, P 5 0.07 for the Menorca35 data set).
The investigated var. palaui and var. minor populations (Table 1) were genetically very similar. The average Neis unbiased genetic identity among the 17 populations was 0.8524 6

AMOVA analysisAMOVA analysis from the phenotypic

distance matrix (DP) for the 162 individuals (the Baleares39
data set) permitted a partitioning of the overall variation into
three levels (Table 5). Although most of this variation was
found within populations (71.4%), there was also evidence for
a significant phenotypic structure of the populations (FST 5
0.286, P , 0.001). The remaining phenotypic diversity was
distributed between islands (2.11%) and between populations
(26.47%). Similar results were found for the Mallorca and Menorca analysis, but the Cabrera analysis showed a different
partitioning (42.94% among populations and 57.06% among
individuals within populations). The bias of these results may
be related to the low number of polymorphic bands (six) for
the individuals of Cabrera in the Baleares39 data set.
To avoid this bias, a second AMOVA was performed using
data sets that fulfilled the Lynch and Milligan criterion within
each island (Mallorca52, Menorca35, and Cabrera22) (Table
6). The phenotypic diversity partitioning corroborated results
previously obtained for Mallorca and Menorca, and the partitioning of variation in Cabrera was more in accordance with
the expected structure (32.05% among populations and
67.95% among individuals within populations).
These significant results allow us to compute RAPD-site
frequencies separately for each population and to generate genotypic distance matrices (DG) from the Baleares39, Mallorca52, Menorca35, and Cabrera22 data sets. The AMOVA analysis of the Baleares39 matrix corroborated the outbreeding behavior of D. minor. As shown in Table 5, the proportion of
variation attributable to within-populations differences was
very high (92.87%). Nevertheless, this analysis also indicated
a significant population differentiation (FST 5 0.071, P ,
0.001). For the within-region analyses, most of the genetic
variation resided between individuals within populations, but
again a significant population structure was evident in each
island, especially in Cabrera (FST 5 0.176 vs. FST 5 0.062
or 0.025 in Mallorca and Menorca, respectively). This variance
partitioning was corroborated when AMOVAs were performed
with DG matrices derived from the Mallorca52, Menorca35,
and Cabrera22 data sets (Table 6). Note however, that the three
within-region analyses showed a lower among-population differentiation (FST 5 0.046, 0.023, and 0.072 for Mallorca, Menorca, and Cabrera, respectively).
Irrespective of the data set used (Baleares39, Mallorca52,
or Cabrera22), all pairwise FST values between Mallorca and
Cabrera populations derived from genotypic AMOVAs were
significant. Pairwise FST values for Menorca populations were
also significant, except comparisons among ML-MS, ML-MB,
ML-MX, ML-MT, and MB-MT in data set Baleares39, and
ML-MB and MB-MT in Menorca35.
It is worth noting that nested AMOVAs (Table 5) showed

1756

AMERICAN JOURNAL

OF

BOTANY

[Vol. 88

TABLE 4. Interpopulation distances using the Apostal coefficient based on 141 (above diagonal) and 52, 35, or 22 (below diagonal) RAPD
phenotypes produced in D. minor individuals from Mallorca, Menorca, and Cabrera, respectively.
Population

Mallorca
D
G
Oa
T
Pa
Ya
F
C
Menorca
MS
MLa
MM
MBa
MX
MEa
MT
Cabrera
CP
CG
a

G
0.1534

0.1241
0.0957
0.1136
0.0769
0.1278
0.1282
0.1459
MS
0.0760
0.1132
0.0593
0.0956
0.0993
0.0931
CP

0.1176
0.1185
0.0887
0.1317
0.1217
0.1350
MLa
0.2153
0.1213
0.0546
0.0931
0.1090
0.0642
CG
0.1221

Oa
0.1446
0.1516
0.0952
0.0670
0.1132
0.1186
0.1429
MM
0.2015
0.1932
0.0501
0.1197
0.1229
0.0973

T
0.1237
0.1538
0.1408
0.0697
0.1139
0.1156
0.1184
MBa
0.2012
0.1856
0.1481
0.0739
0.0758
0.0433

Pa
0.1235
0.1637
0.1154
0.1296

Ya
0.1485
0.1651
0.1095
0.1622
0.1136

0.0798
0.0871
0.1110
MX
0.1851
0.1581
0.1876
0.1774

0.1206
0.1322
MEa
0.2018
0.1573
0.1865
0.1927
0.1604

0.1056
0.0757

F
0.1602
0.1543
0.1715
0.1775
0.1529
0.1611

C
0.1565
0.1346
0.1542
0.1693
0.1498
0.1494
0.1617

0.1353
MT
0.2026
0.1620
0.1791
0.1658
0.1506
0.1688

0.0896

0.2752

D. minor var. palaui populations.

a nonsignificant partitioning of genetic differentiation among

islands (FCT 5 0.211, P 5 0.104; and FCT 5 0.016, P 5
0.082 for both phenotypic and genotypic analyses, respectively).
Except for the two Cabrera populations and the phenotypic
HOMOVA from the Menorca35 data set, the computation of
the Barlett test for the homogeneity of variance indicated significant levels of RAPD variation (both phenotypic and genotypic) among populations as a whole (Tables 5, 6). Of the
589 pairwise Barlett tests of homogeneity of population variation, 554 tests indicated significant differences between populations (data not shown).

The AMOVA procedure was also implemented to study the

relationships between the two infraspecific D. minor taxa (var.
minor and var. palaui). The AMOVAs performed with data
sets Baleares141 and Baleares39 showed no significant differences between groups (FCT 5 0.002, P 5 0.3247; FCT 5
20.001, P 5 0.3076, respectively), when the 17 populations
were subdivided according to their corresponding infraspecific
taxon (see Table 1). These results were corroborated when the
Menorca141 (FCT 5 20.021, P 5 0.7103), Menorca35 (FCT
5 20.01, P 5 0.7612), and Mallorca52 (FCT 5 0.0, P 5
0.3077) data sets were analyzed. Nevertheless, the AMOVA
performed with the Mallorca141 data set showed a significant

TABLE 5. Summary of the AMOVA and HOMOVA analyses performed with the Baleares39 data set. Population statistics were estimated according
to different assumptions of the data. DP and DG: analysis based on the matrices of phenotypic and genotypic distances, respectively, between
individuals (see text for their definitions). Statistics include: degrees of freedom (df), sum of squares (SSD), variance-component estimates
(CV), percentages of the total variance (%total) contributed by each component, and the Barlett test (B).
Analysis

DP
Three Islands
Mallorca
Menorca
Cabrera
DG
Three Islands
Mallorca
Menorca
Cabrera

Source of variation

df

SSD

CV

%Total

Among groups
Among populations within groups
Within populations
Among populations
Within populations
Among populations
Within populations
Among populations
Within populations

2
14
145
7
78
6
50
1
17

15.7
83.0
190.4
44.4
104.3
33.7
75.6
4.9
10.5

0.04
0.49
1.31
0.47
1.34
0.51
1.51
0.47
0.62

2.11 NS
26.47***
71.41***
25.95***
74.05
25.04***
74.96
42.94***
57.06

3.91***
14.26***

Among groups
Among populations within groups
Within populations
Among populations
Within populations
Among populations
Within populations
Among populations
Within populations

2
14
145
7
78
6
50
1
17

18.9
87.7
579.6
47.8
312.2
33.6
231.2
6.3
36.2

0.07
0.24
3.99
0.26
4.00
0.12
4.62
0.45
2.13

1.57 NS
5.56***
92.87***
6.20***
93.80
2.53***
97.47
17.55***
82.45

* P , 0.05; ** P , 0.01; *** P , 0.001; NS 5 not significant (significance test after 5000 permutations).

6.46***
3.47***
0.16 NS

2.84*
16.58***
7.35***
5.83***
0.19 NS

October 2001]

SALES

ET AL.RAPD ANALYSIS IN

DIGITALIS MINOR

1757

TABLE 6. Summary of the AMOVA and HOMOVA analyses performed separately for each island with the appropriate set of bands (52 in
Mallorca, 35 in Menorca, and 22 in Cabrera). Population statistics were estimated according to different assumptions of the data. DP and DG:
analysis based on the matrices of phenotypic and genotypic distances, respectively, between individuals (see text for their definitions). Statistics
include: degrees of freedom (df), sum of squares (SSD), variance-component estimates (CV), percentages of the total variance (%total) contributed by each component, and the Barlett test (B).
Analysis

DP
Mallorca
Menorca
Cabrera
DG
Mallorca
Menorca
Cabrera

Source of variation

df

SSD

CV

%Total

Among populations
Within populations
Among populations
Within populations
Among populations
Within populations

7
78
6
50
1
17

101.7
256.1
37.7
85.7
15.9
50.4

14.5
3.3
6.28
1.71
15.9
2.9

24.28***
75.72
24.69***
75.31
32.1***
67.9

3.09***

Among populations
Within populations
Among populations
Within populations
Among populations
Within populations

7
78
6
50
1
17

105.2
775.6
37.1
259.9
17.7
175.1

15.0
9.9
6.2
5.2
17.7
10.3

4.57***
95.43
2.28***
97.72
7.16***
92.84

4.56***

1.70 NS
0.03 NS

3.48***
0.05 NS

* P , 0.05; ** P , 0.01; *** P , 0.001; NS 5 not significant (significance test after 5000 permutations).

variance among var. minor and var. palaui populations (FCT

5 0.079, P 5 0.005), which was in agreement with results
from the NJ dendrogram of this data set.
Estimation of FSTEstimates of FST and u averaged over
the 39 loci for the 162 individuals and those derived from
separate analyses performed for each island with the appropriate set of bands are shown in Table 7. The correlations
across loci between all the values were high (.91%, P 5
0.0000). Irrespective of the method employed, the estimated
FST and u were significantly different from zero, indicating
among-population genetic differentiation. Interpopulation distances obtained using FST from AMOVA and Lynch and Milligans FST or u showed a significant positive correlation (r .
0.94, P 5 0.0002). This correlation was slightly lower with
Wrights FST pairwise distances (r 5 0.80, P 5 0.02). All these
results corroborated those above described in the extension of
the AMOVA analysis (Stewart and Excoffier, 1996).
DISCUSSION
Results herein represent the first use of molecular (DNA)
markers to characterize genetic diversity in D. minor, a Balearic endemic outcrossing cardenolide-producing plant species. Our study on the spatial structure of D. minor was exclusively based on those RAPD bands that fulfilled Lynch and
Milligans assumptions for the analysis of dominant markers.
Since the variance for population genetic parameters did not
TABLE 7. Estimates for D. minor populations of FST, u, and FST according to Wright (1951), Weir and Cockerham (1984), Lynch and
Milligan (1994), and Stewart and Excoffier (1996). Results obtained with the Baleares39, Mallorca52, Menorca35, and Cabrera22
data sets, respectively.
Item

Baleares39
Mallorca52
Menorca35
Cabrera22
*** P , 0.001.

FST (Wright)

u (Weir and Cockerham)

FST (Lynch
and Milligan)

FST (Stewart and

Excoffier)

0.184***
0.150***
0.141***
0.109***

0.167***
0.141***
0.146***
0.181***

0.200***
0.165***
0.165***
0.167***

0.071***
0.046***
0.023***
0.072***

decrease substantially beyond 30 bands (Aagaard, Krutovskii,

and Strauss, 1998), the number of RAPD markers used in our
study seems to be enough to produce reliable estimates of
genetic structure. Furthermore, the high level of conformity
among the different statistical approaches used for RAPD data
analysis demonstrates the suitability of these molecular markers to accurately resolve the population structure of D. minor.
Both phenotypic and genotypic nested AMOVA showed a
nonsignificant partitioning of genetic diversity among groups
(islands). Since the current distribution of D. minor is relatively recent (Cardona and Contandriopoulos, 1979), the absence of among-island differentiation may only be a reflection
of historical gene flow between D. minor populations when
the Gymnesian Islands constituted a single landmass. Within
species, genetic exchange rather than historical relationships
(e.g., persistent ancestral polymorphisms) has traditionally
been emphasized as the determinant of genetic structure. Nevertheless, shared common ancestry and similar selective regimes could also account for the observed genetic cohesion of
plants (Schaal et al., 1998). Thus, in many groups, genetic
exchange across the species range is severely restricted, either
by wide geographical distribution of populations or by limited
pollen and seed dispersal. In these cases, historical events such
as range expansion, range fragmentation, and population bottlenecks will be strong determinants of population genetic
structure (Schaal et al., 1998). Thus, the observed genetic similarity between D. minor populations belonging to different
islands probably owes more to recent common ancestry than
to any ongoing process of genetic exchange. Nevertheless, further studies based on cpDNA markers can help discriminate
between limited contemporary gene flow and patterns caused
by ancestral polymorphism (Avise, 1994; Schaal et al., 1998).
Our RAPD-based AMOVA studies show that most genetic
variation in D. minor is distributed within populations rather
than between them, indicating a relatively restricted population
differentiation as expected in outcrossing species. Such a pattern of population genetic structure has been previously reported for the mainland congener species D. obscura (Nebauer,
Del Castillo Agudo, and Segura, 1999) and many other outcrossing species (Hamrick and Godt, 1996). Furthermore, the
percentages of within-population variability found in our study

1758

AMERICAN JOURNAL

(92.87% for the among-islands analysis and 82.4597.72% for

the within-island analyses) were either similar or somewhat
higher than those reported in other outcrossing plant species
using RAPD-based AMOVA analysis (see Tables 5 and 6 from
Busell, 1999, and Bartish, Jeppsson and Nybom, 1999, respectively).
Our estimates of the fixation indices (FST or FST) also demonstrate that some structure can be discerned among D. minor
populations. Genotypic AMOVA analyses resulted in FST estimates lower than those given by RADPLOT. Note, however,
that all FST or FST pairwise distance matrices showed a significant positive correlation. Our estimates of FST and FST are
close to those given in the literature for analysis of population
structure in mixed and outcrossing species: 0.10.24 (Loveless
and Hamrick, 1984); 0.0990.216 (Hamrick and Godt, 1990);
and 0.030.31 (Heywood, 1991). The relatively low FST values
corroborate that most of the genetic diversity resides within
D. minor populations and that these populations have experienced gene flow in recent history or diverged a relatively short
time ago. Except for the interpopulation genetic distances derived from the Mallorca141 data set, there was no significant
correlation between the genetic and the actual geographical
distances among populations. This is an indication that isolation by distance is not the process accounting for the distribution of genetic variation among populations within the Mallorca or Menorca islands.
The NJ and PCO analyses revealed a similar, overall geographical structuring of the RAPD variation in D. minor when
the 141 markers were used. These analyses, however, seem to
overestimate the relatedness among individuals since all analyses based on RAPD markers that fulfilled the Lynch and Milligan (1994) criterion failed to follow a consistently geographical pattern either among islands or within each of the islands.
The lack of separation in the dendrograms reflects a weak
genetic differentiation among populations, corroborating results from AMOVA and traditional F statistics.
The most accepted systematic positioning of D. minor takes
into account the plasticity of the species and considers two
infraspecific taxa, varieties minor and palaui, in virtue of their
differences in pubescence (Hinz, 1987). According to this
character, six out the 17 sampled populations (see Table 1)
would belong to D. minor var. palaui. Our RAPD data do not
support, however, this taxonomic differentiation of D. minor.
First of all, our cluster analysis from the data set that fulfilled
the Lynch and Milligan criterion does not show a closer relationships among var. palaui populations than among any other of the sampled populations; thus, the different RAPD phenotypes of var. palaui and var. minor plants were intermingled
in different parts of the NJ and PCO analyses. Also, genetic
differences among populations are small. Crawford (1989)
stated that the mean identity for populations belonging to the
same taxon is often above 0.90. We calculated population
mean identities of 0.8201 6 0.0327 and 0.9345 6 0.0467 (for
the Baleares141 and Baleares39 data sets, respectively) using
the Apostol coefficient (Apostol et al., 1993). Note, however,
that the population mean identities derived from Neis unbiased distance (Nei, 1978) were higher (0.8524 6 0.0388 and
0.9904 6 0.0053 for the Baleares141 and Baleares39 data sets,
respectively). The Apostol coefficient measures the similarity
of pairs of individuals by examining both the shared presence
and the shared absence of bands and then provides a clearer
separation than the Nei distance when very related individuals
are compared (Apostol et al., 1993). Finally, the AMOVA pro-

OF

BOTANY

[Vol. 88

cedure, implemented as an alternative approach to study the

relationships between var. minor and var. palaui populations,
demonstrates that there were no significant differences between groups when the populations were subdivided according
to their corresponding infraspecific taxon. The reliability of
this result is endorsed by the accuracy of RAPD-based AMOVA to detect dissimilarities among several plant species, including Digitalis (see Nebauer, Del Castillo Agudo, and Segura, 2000, and references therein). Corroborating all these
findings, analyses of the sequences of the nuclear rDNA internal transcribed spacers (ITS region) did not discriminate var.
minor and var. palaui (J. A. Rossello, personal communication).
In conclusion, our results indicate that RAPDs are sufficiently informative and powerful to assess genetic variability
in D. minor. Estimates of genetic variation reported herein
provide a basis for the in situ conservation and exploitation
of genetic resources in this species. With knowledge of the
available genetic structure, an appropriate strategy for sampling and propagation of D. minor may be easily formulated
when ex situ conservation is required. Also, information on
the spatial structure of natural populations of D. minor provides important insights into the colonization history, isolation, and diversification of this species.
LITERATURE CITED
AAGAARD, J. E., K. V. KRUTOVSKII, AND S. H. STRAUSS. 1998. RAPDs and
allozymes exhibit similar levels of diversity and differentiation among
populations and races of Douglas-fir. Heredity 81: 6978.
APOSTOL, B., W. C. BLACK IV, B. R. MILLER, P. REITER, AND B. J. BEATY.
1993. Estimation of family numbers at an oviposition site using RAPDPCR markers: applications to the mosquito Aedes aegypti. Theoretical
and Applied Genetics 86: 9911000.
APOSTOL, B., W. C. BLACK IV, P. REITER, AND B. R. MILLER. 1996. Population genetics with RAPD-PCR markers: the breeding structure of Aedes
aegypti in Puerto Rico. Heredity 76: 325334.
ARMSTRONG, J., A. GIBBS, R. PEAKALL, AND G. WEILER. 1996. RAPDistance
programs: version 1.04 for the analysis of patterns of RAPD fragments.
Australian National University, Canberrra, Australia.
AVISE, J. C. 1994. Molecular markers, natural history and evolution. Chapman and Hall, London, UK.
BARLETT, M. S. 1937. Some examples of statistical methods of research on
agriculture and applied biology. Journal of the Royal Statistic Society,
Supplement 4: 137170.
BARTISH, I. V., N. JEPPSSON, AND H. NYBOM. 1999. Population genetic structure in the dioecious pioneer plant species Hipphophae rhamnoides investigated by random amplified polymorphic DNA (RAPD) markers.
Molecular Ecology 8: 791802.
BLACK, IV W. C. 1998. Fortran programs for the analysis of RAPD-PCR
markers in populations. Colorado State University, Fort Collins, Colorado, USA.
BUSSELL, J. D. 1999. The distribution of random amplified polymorphic DNA
(RAPD) diversity amongst populations of Isotoma petraea (Lobeliaceae).
Molecular Ecology 8: 775789.
CARDONA, M. A., AND J. CONTANDRIOPOULOS. 1979. Endemism and evolution in the islands of the western Mediterranean. In D. Bramwell [ed.],
Plant and islands, 133169. Academic Press, London, UK.
CONTANDRIOPOULOS, J., AND M. A. CARDONA. 1984. Caracte`re original de
la flore endemique des Baleares. Botanica Helvetica 94: 101131.
CRAWFORD, D. J. 1989. Enzyme electrophoresis and plant systematics. In D.
E. Soltis and P. S. Soltis [eds.], Isozymes in plant biology, 146164.
Dioscorides Press, Portland, Oregon, USA.
DEL CASTILLO-AGUDO, L., I. GAVIDIA, P. PEREZ-BERMUDEZ, AND J. SEGURA.
1995. PEG precipitation, a required step for PCR amplification of DNA
from wild plants of Digitalis obscura L. Biotechniques 18: 766768.
EXCOFFIER, L., P. E. SMOUSE, AND J. M. QUATTRO. 1992. Analysis of molecular variance inferred from metric distances among DNA haplotypes:

October 2001]

SALES

ET AL.RAPD ANALYSIS IN

applications to human mitochondrial DNA restriction data. Genetics 131:

479491.
FELSENSTEIN, J. 1993. PHYLIP: Phylogeny inference package, version 3.57c.
Department of Genetics, University of Washington, Seattle, Washington,
USA.
HAMRICK, J. L., AND M. J. W. GODT. 1990. Allozyme diversity in plant
species. In A. H. D. Brown, M. T. Clegg, A. L. Kahler, and B. S. Weir
[eds.], Plant population genetics, breeding and genetic resources, 4363.
Sinauer, Sunderland, Massachusetts, USA.
HAMRICK, J. L., AND M. J. W. GODT. 1996. Effects of life history traits on
genetic diversity in plant species. Philosophical Transactions of the Royal Society of London, Series B, 351: 12911298.
HEYWOOD, J. S. 1991. Spatial analysis of genetic variation in plant populations. Annual Review of Ecology and Systematics 22: 335355.
HINZ, P. A. 1987. Etude biosystematique de lagregat Digitalis purpurea L.
(Scrophulariaceae) en Mediterranee occidentale. VIII. Digitalis minor L.
endemique des Baleares. Candollea 42: 693716.
HINZ, P. A. 1990. Etude biosystematique de lagregat Digitalis purpurea L.
(Scrophulariaceae) en Mediterranee occidentale. XII. Synthe`se. Candollea 42: 693716.
ISABEL, N., J. BEAULIEU, P. THERIAULT, AND J. BOUSQUET. 1999. Direct
evidence for biased gene diversity estimates from dominant random amplified polymorphic DNA (RAPD) fingerprints. Molecular Ecology 8:
477483.
JENCZEWSKI, E., J. M. PROSPERI, AND J. RONFORT. 1999. Differentiation
between natural and cultivated populations of Medicago sativa (Leguminosae) from Spain: analysis with random amplified polymorphic DNA
(RAPD) markers and comparison to allozymes. Molecular Ecology 8:
13171330.
KAMPNY, C. M. 1995. Pollination and flower diversity in Scrophulariaceae.
Botanical Review 61: 350366.
LOVELESS, M. D., AND J. L. HAMRICK. 1984. Ecological determinants of
genetic structure in plant populations. Annual Review of Ecology and
Systematics 15: 6595.
LYNCH, M., AND B. G. MILLIGAN. 1994. Analysis of population genetic structure with RAPD markers. Molecular Ecology 3: 9199.
MANTEL, N. A. 1967. The detection of disease clustering and a generalized
regression approach. Cancer Research 27: 209220.
MITTON, J. B. 1994. Molecular approaches to population biology. Annual
Review of Ecology and Systematic 25: 4569.
NEBAUER, S. G., L. DEL CASTILLO AGUDO, AND J. SEGURA. 1999. RAPD

DIGITALIS MINOR

1759

variation within and among natural populations of outcrossing willowleaved foxglove (Digitalis obscura L.). Theoretical and Applied Genetics
98: 985994.
NEBAUER, S. G., L. DEL CASTILLO AGUDO, AND J. SEGURA. 2000. An assessment of genetic relationships within the genus Digitalis based on
PCR-generated RAPD markers. Theoretical and Applied Genetics 100:
12091216.
NEI, M. 1978. Estimation of average heterozygosities and genetic distances
from a small number of individuals. Genetics 89: 583590.
PAGE, R. D. M. 1996. TREEVIEW: an application to display phylogenetic
trees on personal computers. Computer Applications in the Biosciences
12: 357358.
ROHLF, F. J. 1997. NTSYS: numerical taxonomy and multivariate analysis
system, version 2.0. Exeter Software, Setauket, New York, USA.
SCHAAL, B. A., D. A. HAYWORTH, K. M. OLSEN, J. T. RAUSCHER, AND W.
A. SMITH. 1998. Phylogeographic studies in plants: problems and prospects. Molecular Ecology 7: 465474.
SLATKIN, M. 1987. Gene flow and the geographic structure of populations.
Science 236: 787792.
STEWART, C. N., AND L. EXCOFFIER. 1996. Assessing population genetic
structure and variability with RAPD data: application to Vaccinium macrocarpon (American cranberry). Journal of Evolutionary Biology 9:
153171.
THOMPSON, J. D. 1999. Population differentiation in Mediterranean plants:
insights into colonisation history and the evolution and conservation of
endemic species. Heredity 82: 229236.
WEIR, B. S., AND C. C. COCKERHAM. 1984. Estimating F-statistics for the
analysis of population structure. Evolution 38: 13581370.
WILLIAMS, J. G. K., A. R. KUBELIK, K. J. LIVAK, J. A. RAFALSKI, AND S. V.
TINGEY. 1990. DNA polymorphisms amplified by arbitrary primers are
useful as genetic markers. Nucleic Acids Research 18: 65316535.
WRIGHT, S. 1951. The genetical structure of populations. Annals of Eugenics
15: 323354.
WU, J., K. V. KRUTOVSKII, AND S. H. STRAUSS. 1999. Nuclear DNA diversity, population differentiation, and phylogenetic relationships in the California closed-cone pines based on RAPD ands allozyme markers. Genome 42: 893908.
YEH, F. C., R. C. YANG, T. BOYLE, Z. H. YE, AND J. X. MAO. 1997. POPGENE, the user friendly shareware for population genetic analysis. Molecular Biology and Biotechnology Centre, University of Alberta, Canada.