Molecular Ecology (2004) 13, 349– 370 doi: 10.1046/j.1365-294X.2003.02064.

Extensive chloroplast haplotype variation indicates

0Blackwell Publishing, Ltd.

Pleistocene hybridization and radiation of North American

Arabis drummondii, A. × divaricarpa, and A. holboellii
(Brassicaceae)
C . H . D O B E S,* T . M I T C H E L L - O L D S † and M . A . K O C H *‡§
*Institute of Botany, University of Natural Resources and Applied Life Sciences Vienna, Gregor-Mendel-Strasse 33, A-1180 Vienna,
Austria, †Max-Planck Institute of Chemical Ecology, Winzerlaer Strasse 10, D-07745 Jena, Germany, ‡Institute of Plant Sciences,
University of Heidelberg, Im Neuenheimer Feld 345, 69120 Heidelberg, Germany

Abstract
Arabis drummondii, A. holboellii and their hybrid A. × divaricarpa are widespread peren-
nials of open habitats in North America. A phylogenetic analysis based on noncoding chloro-
plast DNA sequences (trnL intron and trnL/F intergenic spacer) resolved A. drummondii as
a monophyletic taxon, but found A. holboellii to bear chloroplast haplotypes from highly
diverged evolutionary lineages. This raised the question of a possible polyphyletic origin
of A. holboellii. Arabis × divaricarpa was found to be of recent and polytopic origin, a result
consistent with its presumed hybrid origin. One hundred and three chloroplast haplotypes
were detected within 719 Arabis accessions investigated. The majority of chloroplast-types
were estimated to have arisen prior to the Wisconsin glaciation. Phylogeographical analysis
using nested clade analysis, suggested for A. holboellii (i) past fragmentation events,
partitioning genetic variation in several instances between the Sierra Nevada, the Southern
Rocky Mountains and the Colorado Plateau on the one hand and the Central to Northern
Rockies of the United States and adjacent Cascades on the other; and for both parental spe-
cies (ii) recolonization of major areas formerly covered by the Wisconsin glaciation by three
haplotypes; and (iii) restricted gene flow indicating isolation by distance in areas south of
the last glacial maximum. Arabis × divaricarpa is closely codistributed with its parental
species and resampled their haplotypes. The highest genetic diversity was found in the
Rocky Mountains from Idaho and Montana south to Utah and Colorado. This area was
further hypothesized to have played a major role in the origin of both parental species
and probably represented an important glacial refugium. However, evidence for glacial
refugia was also found in arctic and boreal regions of Alaska and near the Great Lakes.
In comparison to nuclear ribosomal internal transcribed spacer data, chloroplast DNA
divergence was very high and evidently predated the origin of A. drummondii and possibly
A. holboellii. Divergence of major chloroplast lineages dates back to the middle of the Pleis-
tocene at least. Extensive hybridization is the most likely evolutionary factor working on
A. holboellii to explain the revealed discrepancy in nuclear DNA and chloroplast DNA
diversification.
Keywords: Arabis, Boechera, glacial refugia, nested clade analysis, North America, phylogeography

Received 31 July 2003; revision received 10 September 2003; accepted 15 October 2003

Correspondence: Marcus A. Koch. § University of Heidelberg, Institute of Plant Sciences, Department Biodiversity and Systematics,
Im Neuenheimer Feld 345, 69120 Heidelberg, Germany. Fax: + 49-(0)-6221-54-5508; E-mail: marcus.koch@urz.uni-heidelberg.de

© 2004 Blackwell Publishing Ltd

350 C . H . D O B E S, T . M I T C H E L L - O L D S and M . A . K O C H
Introduction
Native North American Arabis (sometimes grouped in the
genus Boechera) stands in the centre of interest of several
research activities focusing on the genetic basis of apomixis
(Roy 1995; Naumova et al. 2001; Sharbel & Mitchell-Olds
2001), on ecology and evolution (Mitchell-Olds 2001), hybrid
speciation (Koch et al. 2003b) and on the taxonomy of the
genus and its allies (Koch et al. 1999, 2001). Depending
on the underlying taxonomic concept, ‘Boechera’ comprises
50–80 species centred in the western United States (Rollins
1993). Arabis holboellii, A. drummondii and their putative hybrid
A. × divaricarpa (Rollins 1983; Koch et al. 2003b) were chosen
as a model system to understand which genetic and biological
processes drive the radiation of the genus.
These species are widespread biennials or short-lived
hemicryptophytes of open habitats in western and north-
eastern North America (see http://sis.agr.gc.ca/pls/spec/
brassget). Plants reach heights of up to 50 cm and put out
dry, wingless or narrowly winged seeds, 1–2 mm in dia-
meter and borne in terminal fructifications.
Based on morphological features, several intraspecific
taxa were accepted by Rollins (1993) for A. holboellii, a poly-
morphic species which occurs from lowland to alpine habitats.
Using nuclear ribosomal DNA internal transcribed spacer
(ITS) sequences, multiple intraindividual ITS copies were
detected in several A. holboellii accessions and it was demon-
strated that intraspecific hybridization is a common pheno-
menon in this genetically polymorphic species (Koch et al.
2003b). Diploid and triploid A. holboellii are often apomictic
and pseudogamous (Böcher 1951, 1969; Roy & Rieseberg
1989; Roy 1995; Sharbel & Mitchell-Olds 2001; Sharbel et al.
2004), but diploid cytotypes can also reproduce via sex
(Böcher 1951, 1969). However, outbreeding seems to be
infrequent in A. holboellii. In contrast, A. drummondii, mostly
diploid and sexual (Rollins 1993; Roy 1995; Koch et al. 2003b),
is a morphologically well-defined and moderately variable
species of montane to alpine habitats and, despite consider-
able intraspecific genetic variability, intraindividual ITS
polymorphism was found to be extremely rare (Koch et al.
2003b). The latter finding might be explained by the pre-
dominant reproduction via self-fertilization of this species
(Roy 1995). Species-specific ITS polymorphism supported
genetic differentiation between A. drummondii and A. holboellii
(Koch et al. 2003b). However, a few A. holboellii accessions
investigated in this study showed evidence of introgres-
sion by A. drummondii. The widespread A. × divaricarpa is
largely codistributed with its parents and is considered to
have arisen several times (Rollins 1983). This taxon is almost
exclusively polyploid (triploid, tetraploid) and was sug-
gested to reproduce preferably via apomixis (Rollins 1983,
1993). The above findings are supported by ITS data (the
hybrid resampled the parental sequence variation in different
geographical regions) and by pollen size measurements
and estimation of pollen quality (Koch et al. 2003b; Sharbel
et al. 2004).
Hybridization and apomixis in conjunction with poly-
ploidy were found to play a major role in the origin and main-
tenance of intraspecific and interspecific polymorphism also
in other Boechera species (Böcher 1969; Rollins 1983). However,
classification within the genus has been mainly based on
morphological characters, and diagnostic features discrim-
inating between taxa are often limited. Therefore, species
delimitation seems to be highly artificial, especially with
respect to apomictic complexes (cf. Böcher 1969). Little is yet
known about the phylogeny or phylogeography of ‘Boechera’,
although a number of molecular markers and biogeograph-
ical studies are available for species in the Brassicaceae (see
Koch 2003; Koch et al. 2003a for overview).
Since its development by Avise et al. (1987) phylogeo-
graphy has become an increasingly important field of research
within biogeography. Originally the aim was to describe
the distribution of genetic variation in space and time. More
recently, understanding of historical and population pro-
cesses has emerged as a central focus. Besides its strength
in testing for explicit evolutionary relationships between
geographical areas occupied by phylogenetically related
taxa, a phylogeographical approach traces the history of
genealogic lineages. This advantage may be of special import-
ance at the junction of the intra- and interspecific level, where
partially divergent lineages may occur sympatrically. Indeed,
knowledge on the geographical distribution and history of
genealogical lineages can further improve our understand-
ing of the dynamics of speciation via hybridization in a
spatial and temporal context (e.g. Wittzell 1999; Avise 2000;
Comes & Abbott 2001). Such an approach appears highly
relevant in the case of the genus Boechera, given the suggested
evolutionary importance of hybridization (Böcher 1969;
Rollins 1983; Roy & Rollins 1989; Sharbel & Mitchell-Olds
2001; Koch et al. 2003b) and the availability of highly variable
DNA markers (Sharbel & Mitchell-Olds 2001).
Templeton and colleagues (Templeton et al. 1987, 1992,
1995; Templeton & Sing 1993; Crandall & Templeton 1993;
Templeton 1998) proposed nested clade analysis (NCA),
a method aiming to assess the historic causes of geo-
graphical variation. However, as outlined by Knowles &
Maddison (2002) the method does not estimate the con-
fidence limits for inferred historical processes. Further-
more, Petit & Grivet (2002) demonstrated that inferences
drawn by this approach depend on the method used to
make the permutations when there are several individuals
per population. Several other methods are available and
commonly used in phylogeographical data analysis, e.g.
analysis of molecular variance (amova); Mantel tests; assess-
ment of allelic richness and genetic diversity or of dispersal
distances; various methods of phylogenetic inference;
or ordination methods. However, although the algorithms
provide statistical probabilities for the resolved patterns,
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349– 370
 P L E I S T O C E N E H Y B R I D I Z A T I O N O F A R A B I S S P E C I E S 351
none of them make explicit statements on the underlying
historic processes. Therefore, the accuracy in interpreting
phylogeographical distribution patterns still relies on a
thorough consideration and synthesis of genetic as well as
of geological, historic geographical, environmental, or
biological data. As no single technique is available today
that is capable of resolving reliably the various historic events
leading to contemporary geographical distribution patterns,
usually a combination of methods is applied to address
these questions (e.g. Stehlik 2002; Tribsch et al. 2002).
Phylogeographical studies in plants using molecular
markers have so far focused primarily on Europe (Hewitt
2001; Petit et al. 2001). Considerable knowledge is now avail-
able for this continent, supporting detailed hypotheses on
Pleistocene migration routes and possible ice age refugia
(Comes & Kadereit 1998). Yet, there is limited information
available to address these questions for North American
plant species (e.g. Soltis et al. 1997; Hamann et al. 1998;
Tremblay & Schoen 1999; Ritland et al. 2001). Recently,
Brunsfeld et al. (2001) outlined three historical scenarios
for contemporary biogeographical patterns of plants and
animals in the Pacific Northwest. Glacial refugia were
proposed for the Pacific coast as well as for central parts
of the Rocky Mountains. Furthermore, models invoking
vicariance are compared against hypotheses of postglacial
range expansions.
The general aim of this study is to develop a phylogeo-
graphical framework for the model species A. drummondii,
A. holboellii and A. × divaricarpa to provide detailed informa-
tion on the centres of genetic biodiversity, on past refugia
of the genus, and on possible secondary contact and to
better understand the dynamics of hybridization. For this
purpose, chloroplast DNA (cpDNA) markers will be used,
which are usually considered to provide, as a result of a
lack of recombination and maternal inheritance (Birky et al.
1989; Ennos 1994; Martinez et al. 1997), a more conservative
historical view than nuclear markers. Here we will focus
on the following questions. (i) Are A. drummondii and A.
holboellii genetically differentiated at cpDNA sequences,
as suggested by nuclear ITS sequences? (ii) Does cpDNA
genetic variation allow us to identify separate geograph-
ical regions within North America? (iii) Can centres of
diversity be localized? (iv) Do maternally inherited markers
support the polyphyletic/polytopic origin of A. × divaricarpa?
This work is part of an ongoing, comprehensive investigation
into the systematics, taxonomy, phylogenetic relationships,
evolutionary history and breeding system of these species.
Materials and methods
Plant material and sampling design
Plant material was obtained from Gray Herbarium, Harvard
University, Cambridge (acronym GH), from Missouri
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349–370
Botanical Garden, St Louis (MO), from the California
Academy of Sciences, San Francisco (CAS and DH), from
Agriculture and Agri-Food Canada, Ottawa (DAO), and
from the National Herbarium of Canada, Ottawa (CAN).
In total, 717 accessions from these herbaria (212 of Arabis ×
divaricarpa, 173 of A. drummondii, 332 of A. holboellii) were
investigated, covering the entire geographical range of
the three taxa in North America. Vouchers were chosen
randomly and the sampling should therefore approximately
reflect the relative abundances of the three species across
North America (for full information on their taxonomic
determination and geographical origin see Supplementary
material: Table 1, Table 2 and Fig. 1). In addition, two acces-
sions (J49, J52) from the collection of the Max-Planck
Institute of Chemical Ecology, Jena, Germany, were included
in this study. Each accession represents one population.
Vouchers were labelled with a material notation to allow
for consistent cross-references between previous (Koch
et al. 2003b) and future studies. For computation purposes
and presentation of data, origins noted on the voucher
labels were translated into geographical coordinates. Most
of the material was either revised or collected by the mono-
grapher of the genus, Reed C. Rollins.
DNA isolation, polymerase chain reaction and DNA
sequencing
Total DNA was obtained from 50–75 mg dried leaf tissue
from single individuals. At date of isolation the ages of the
herbarium material ranged from 5 to 108 years, but the
majority of accessions were then either 10–20 or 40 –
70 years old. Extraction followed the procedure of Doyle &
Doyle (1987) (the CTAB method), but some modifications
were applied, involving grinding of only 50–75 mg of dry
leaf tissue in 2-mL tubes using a Retsch swing mill (MM
200), addition of 2 units of ribonuclease per extraction to
the isolation buffer, and washing of the DNA pellet twice with
70% ethanol. DNA was finally dissolved in 50 µL TE-buffer
for long-term storage. For each DNA sample 20–40 µL has
been deposited at the herbaria GH and MO. Before use,
DNA was diluted 1 : 3 in TE-buffer.
Fifty-microlitre polymerase chain reactions (PCR) were
performed in a master mix containing 1 × PCR buffer
(10 mm Tris–HCl/50 mm KCl buffer, pH 8.0), 3 mm MgCl2,
0.4 µm of each primer, 0.2 mm of each dNTP, 1 unit Taq
DNA polymerase (Schott-Eppendorf), and approximately
1 ng of template DNA using an ABI 9700 (Applied Biosystems)
thermal cycler. Thermal cycling started with a denatura-
tion step at 95 °C lasting 5 min; followed by 35 cycles each
comprising 60-s denaturation at 95 °C, 45-s annealing
at 38 °C (trnL intron)/45 °C trnL/F IGS/48 °C rpoC1 intron,
and 60-s elongation at 72 °C. Amplification ended with
an elongation phase at 72 °C lasting 10 min, and a final
hold at 4 °C.
352 C . H . D O B E S, T . M I T C H E L L - O L D S and M . A . K O C H

Table 1 Results of the nested analysis of geographical distance for the Arabis chloroplast haplotypes

Chain of Interior
Clade inference clades Tip clades Inferred historic event

Lineage I
1–24 1-2-3-4-NO U V restricted gene flow with isolation by distance
2–8 1-2-3-5-15-NO 1–20 D, F past fragmentation
3–6 1-2-3-4-9-NO 2–9 1–23 past fragmentation
4–2 1-2-3-4-9-NO 3–5 2–10 past fragmentation
Lineage II
1–1 1-2-3-4-NO AS AK, BC, BD restricted gene flow with isolation by distance
2–1 1-2-3-4-NO 1–1 1–6, 1–7 restricted gene flow with isolation by distance
2–3 1-2-3-4-NO 1–8 AJ restricted gene flow with isolation by distance
3–1 1-2-3-4-NO 2–1 2–2 restricted gene flow with isolation by distance
3–2 1-2-3-4-NO 2–3 1–9 restricted gene flow with isolation by distance
3–3 1-2-3-4-9-NO 2 – 4! 2–5 past fragmentation
4–1 1-2-11-12-NO 3–2 3–1 contiguous range expansion
total network I 1-2-11-12-NO 3–3 4–1, 4–2 contiguous range expansion
Lineage III
1–13 1-2-11-12-NO BU BY Contiguous range expansion
1–17 1-2-3-4-NO CI CJ, CL restricted gene flow with isolation by distance
2–7 1-2-3-4-NO 1 – 17 CM, CQ, CR, 1–19 restricted gene flow with isolation by distance total
network II 1-2-11-12-NO 1 – 12! 2–6, 2–7 contiguous range expansion

Only clades for which a statistically significant and conclusive geographical pattern was found are tabled. Only tip clades showing
statistically significant Dc or Dn values are listed. Exclamation marks mark most ancient interior clades for which only few accessions
were found. The analyses on lineage II (except for clade 4-1) was performed twice, with and without A. holboellii accessions. However,
both analyses inferred the same processes.

Table 2 Test of correspondence between phylogenies of haplotypes of lineage I, II, III, the entire network, and the three species under study,
respectively, and their geographical distribution according to Pons & Petit (1996)

NST (variance) GST (variance)

mean for 300 5% of permutated

Haplotypes of: observed permutations values > observed

lineage I 0.200 (0.0965) 0.192 0.297 0.208 (0.0561)

lineage II 0.248 (0.0553) 0.307 0.381 0.318 (0.0682)
lineage III 0.276 (0.0757) 0.534 0.296 0.229 (0.0659)
entire network 0.154 (0.0315) 0.174 0.236 0.179 (0.0444)
A. drummondii 0.228 (0.0617) 0.320 0.409 0.339 (0.0906)
A. × divaricarpa 0.286 (0.0858) 0.574 0.326 0.259 (0.0632)
A. holboellii 0.281 (0.0510) 0.205 0.255** 0.204 (0.0488)

**P > 0.01.

Except for Arabis holboellii, none of the N ST values observed was statistically significantly greater than GSTs as at least 5% of the permutated
NST values were greater than the observed NSTs.

Fig. 1 Maximum parsimony analysis of 103 chloroplast haplotypes (A to CY): strict consensus tree of 10 000 most parsimonious trees.
Rectangles mark the homoplastic occurrence of indels as listed in the legend. Synapomorphies supporting particular clades are indicated
by , homoplasies by ↓; bootstrap values (greater than 50%) are given in bold above and at the bottom of branches, respectively. Dashed
lines indicate A. drummondii-like haplotypes (clades), solid lines represent haplotypes characteristic for A. holboellii. Vertical lines to the right
of the tree specify in detail, which haplotypes are present in A. drummondii (white), A. × divaricarpa (grey), and A. holboellii (black),
respectively (see Fig. 6 for haplotype frequencies).

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349– 370

 P L E I S T O C E N E H Y B R I D I Z A T I O N O F A R A B I S S P E C I E S 353

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349–370

354 C . H . D O B E S, T . M I T C H E L L - O L D S and M . A . K O C H
The trnL intron was amplified using the forward primer
5′-CGA AAT CGG TAG ACG CTA CG-3′ and the reverse
primer 5′-GGG GAT AGA GGG ACT TGA AC-3′ (primer
c and d according to Taberlet et al. 1991), which anneals in
the first and second exon of the trnL gene, respectively.
Sequences comprised the complete intron and the second
exon of the trnL gene. For amplification of the trnL/F IGS
primers 5′-GGT TCA AGT CCC TCT ATC CC-3′ (primer e
according to Taberlet et al. 1991) and 5′-GAT T TT CAG
TCC TCT GCT CTA C-3′ (designed in this study) annealing
in the second exon of the trnL gene and the trnF gene,
respectively, were used. Amplified sequences included the
complete IGS and the first 18 bases of the trnF gene. Firty-
four accessions comprising the whole variation of the trnL
intron-trnL/F IGS were deliberately chosen for amplifica-
tion of the rpoC1 intron. As a result of the length of the
rpoC1 intron [> 700 base pairs (bp)] the region was divided
into two overlapping fragments, which were amplified
using primer pairs 5′-CAT GAA AGT TTC CTC TCA TCC-
3′/5′-ATG TCA TGG T TG CAG AAG TC-3′, and 5′-GAA
TGC CGG AAT ACT CT T TAC-3′/5′-ATA CTG CGA TGT
GTG ATT TG-3′, respectively.
PCR products were checked for length and concentrations
on 1.5% agarose gels. No purification of PCR products was
necessary for subsequent sequence reactions. Cycle sequen-
cing was performed using the TaqDyeDeoxy Terminator
Cycle Sequencing Kit (ABI Applied Biosystems, Inc.) and
the original amplification primers. However, the reverse
trnL/F IGS primer was modified adding an additional
cytosine to its 3′-end. Products were analysed on an ABI
377XL automated sequencer. Cycle sequencing was
performed on both strands; in the majority of cases each
reaction spanned the complete sequence.
Data analysis: phylogenetic inference and nested clade
analysis
A phylogenetic analysis was performed on trnL intron-
trnL/F IGS haplotypes. In order to verify the topology of
the obtained tree a phylogenetic analysis including also the
rpoC1 intron was performed on 44 accessions. TrnL intron-
trnL/F IGS sequences from Halimolobos parryii (AF307539)
and Polyctenium fremontii (AF183043, AF183044) served in
both analyses as outgroups. TrnL/F IGS (AY257692–
AY257794) and rpoC1 (AY257795–AY257838) sequences were
submitted to GenBank with the complete corresponding
alignment information. The maximum parsimony optimality
criterion was applied for reconstruction of phylogenies
using paup 4.0 version 4.0b10 (Swofford 2002). The heuristic
search algorithm was chosen, using the random addition
of taxa and the tree bisection–reconnection (TBR) option
for branch swapping, which was restricted to 10 000 retained
trees in the case of analysis of trnL intron-trnL/F IGS sequence
types. The parsimonious uninformative start and end regions
of the trnL intron-trnL/F IGS alignment (positions 1–105;
1104–1125) as well as a duplicated region (positions 890–981)
were excluded from the analyses. Two poly(T) stretches
found in the rpoC1 intron, which showed homoplastic
length variation, were likewise excluded. gapmode was
set to missing, but selected indels within the trnL/F IGS,
according to the criteria outlined in the Results section, and
all indels present in both the trnL and the rpoC1 intron were
coded as additional single binary 0/1 characters in the
nexus input file. A strict consensus tree was constructed
from 10 000 shortest trees retained, respectively, from all
shortest trees found by the analysis of the combined trnL/
F-rpoC1 data. Bootstrapping was carried out on 1000 replicates
using the heuristic search option.
Haplotype networks were constructed for the trnL intron-
trnL/F IGS region. For this purpose the same sequence align-
ment assembled for the maximum parsimony analysis was
used, except that all indels were coded as additional single
characters. Haplotype networks were constructed using tcs
version 1.13 (Clement et al. 2000) according to acceptance
criteria outlined in Templeton et al. (1992). The network-
building algorithm joins haplotypes according to the number
of mutational differences between sequences. Therefore,
the loss of genetic distance, as a result of the later deletion
of informative DNA-regions (which means that they car-
ried mutations), will not be considered by the algorithm,
which results in low topological distances between more
diverged haplotypes. To overcome this problem, node
sequences were reconstructed for the strict consensus
tree (see above) using the reconstruct command imple-
mented in paup. Gaps were closed with the corresponding
sequence of the inferred node adjacent to the haplotype
bearing the deletion. Evolutionary ambiguities as expressed
by reticulate connections within the observed haplotype
network were treated according to the criteria outlined in
Crandall & Templeton (1993). Haplotypes within the inferred
networks were nested into hierarchically interlocking groups
(‘clades’ sensu Templeton et al. 1987) following the rules
provided in Templeton et al. (1987, 1992).
A quantitative analysis of geographical data, as described
in Templeton (1998) was performed using geodis 2.0.
Historic events underlying the contemporary geographical
distribution of A. drummondii were inferred in excluding
all A. holboellii accessions from lineage II (cf. Figs 1 and 6).
Hybrid and parental accessions were jointly analysed as
their haplotypes were highly codistributed. Statistically
significant large or small Dc values (a measure for the geo-
graphical range of a particular clade), Dn (the geographical
range of a particular clade relative to its closest sister
clades) (I–T)D (the Dc of interior minus that of tip clades),
c
or (I–T)D (Dn interior–Dn tip) were interpreted using the
n
inference key available from http://inbio.byu.edu/
Faculty/kac/crandall_lab/geodis.htm (cf. also Templeton
1998).
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349– 370
 P L E I S T O C E N E H Y B R I D I Z A T I O N O F A R A B I S S P E C I E S 355
Genetic differentiation among geographical regions:
NST statistics and pairwise population comparisons
Fourteen geographical regions corresponding to major
land features of North America were defined. To obtain
geographical units comprising approximately equal numbers
of individuals on which the population-genetic calculations
were performed, several of the major regions were further
divided, resulting in 23 subregions (then comprising between
20 and 45 accessions each). Genetic differentiation among
these subregions was tested applying NST statistics (Pons &
Petit 1996) and using the program permut (available from
http://www.pierroton.inra.fr/genetics/labo/Software).
This test compares NST against values of GST. NST estimates
consider not only differences in the frequencies of haplo-
types between populations, as with GST, but also genetic
distances between haplotypes. In case of correspondence
between haplotype phylogenies and their geographical
distribution, estimates for NST will be greater than GST
values (Pons & Petit 1996). Average numbers of pairwise
distances (Nei & Li 1979) between chloroplast haplotypes
distributed in the above subregions using uncorrected
P-distances were calculated using arlequin vers. 2.000. The
analyses was performed separately on A. drummondii-like
(haplotypes AD to BL; Fig. 1) and A. holboellii-like haplo-
types (remaining ones), respectively. Based on these distances
neighbour-joining trees were constructed using paup 4.0.
Estimation of genetic diversity
Geographic distribution of genetic diversity was estimated
based on the 23 subregions described above. Effective divers-
ity according to Gregorius (1978) was calculated for each
subregion:
 n 
νa = 
 ∑p i 
2
 Absolute frequencies of particular haplotypes varied from
 i =1 
where pi is the frequency of haplotype i.
Time estimates and molecular clock assumptions
The approximate age of haplotypes within the ingroup
was inferred by calculating the mutation rate for the trnL
intron using sequences of A. thaliana (X74573), A. drummondii
(haplotype AS), A. lignifera (AF182999), Halimolobos perplexa
(AF312971), Rorippa palustris (AF100854), Barbarea vulgaris
(AF198119) and Lepidium campestre (AY015845) as outgroups.
Times of divergence among these taxa were taken from
Koch et al. (2000, 2001). A test of rate heterogeneity was
performed using the RR-Test program (Robinson et al. 1998).
A phylogeny based on the trnL intron was reconstructed
using the maximum parsimony optimality criterion and
paup 4.0.
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349–370
Results
Sequence analysis and inference of chloroplast haplotypes
The length of the trnL intron was almost invariably 505 bp
within the in-group (except haplotype CM: Arab77, Arab479,
500 bp; and haplotype AR: Arab597, 506 bp), whereas,
length of the trnL/F IGS varied between 309 and 493 bp.
Ninety-one single nucleotide polymorphisms (SNPs; includ-
ing 40 singleton mutations) in the combined data set, trnL
intron + trnL-trnF IGS, were identified. Eighty-seven of these
polymorphic sites (96%) were biallelic. The length of the
alignment was 1125 bp (Fig. 2).
Considerable differences regarding the trnL intron
and trnL/F IGS region were recognized with respect to
insertion–deletion polymorphisms (Fig. 2). In the trnL
intron one single base insertion (position 241; haplotype AR)
and one deletion (5 bp at alignment position 222–226;
haplotype CM) were found, whereas in the trnL/F IGS, indels
were detected at alignment positions 629–716, 797–834**,
839–1007, 890–981*, 898–980, 982–1078*, 990–995, 990–1086,
1004–1100**, 1019–1100**, 1037–1041 and 1075–1082 (Fig. 2).
The indels marked by * and ** were characterized by com-
mon start (AGTA*; ATTA or CGTA**) and end motifs (CATT*;
CTTC**) as well as by high sequence homology among
themselves. Tandem repeats comprising up to four copies
were observed, numbered I to IV according to their 5′ to 3′
order in the (trnL intron)–trnL/F IGS alignment (see above).
As a result of their lengths (up to 115 bp), they were respons-
ible for the largest amount of length polymorphism observed:
the shortest IGS sequence found was 309 bp long and pos-
sessed only copy I and IV (e.g. cp-type M or AI); whereas,
the length of this sequence was 493 bp in type AZ, CA,
or CV, in which the repeat was made up of all four copies
(Fig. 2). The observed mutations define 103 chloroplast
haplotypes (subsequently named haplotype A to CY).
singletons, found only once, to 147 in case of cp-type AS
( Table 2 — Supplementary material).
Taxonomic and phylogenetic relationship of chloroplast
haplotypes
In Arabis drummondii 29 different cp-types were found, in
A. holboellii 64, and in A. × divaricarpa 33 (Table 2 — Supple-
mentary Material). Of these, 19, 50 and 16 cp-types were
species-specific for A. drummondii, A. holboellii, and A. ×
divaricarpa, respectively. Arabis drummondii shared six
cp-types with A. holboellii; and A. × divaricarpa shared eight
cp-types with A. drummondii and 13 with A. holboellii.
Frequencies of particular haplotypes varied considerably
between species (Fig. 6, Table 2 — Supplementary Material)
as most haplotypes found in A. drummondii were absent or
very rare in A. holboellii.
356 C . H . D O B E S, T . M I T C H E L L - O L D S and M . A . K O C H

Fig. 2 Characterization of 103 chloroplast haplotypes by single nucleotide and indel polymorphism observed in the trnL intron and the
trnL-trnF IGS. Sequences are compared to haplotype A. Order of haplotypes follows their position in the strict consensus tree resolved by
the maximum parsimony analysis (Fig. 1); singleton mutations (observed only once) are given in italic; for convenience SNPs and indels
are presented in two separate blocks; numbers mark base positions in the alignment.

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349– 370

 P L E I S T O C E N E H Y B R I D I Z A T I O N O F A R A B I S S P E C I E S 357

Fig. 2 Continued

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349–370

358 C . H . D O B E S, T . M I T C H E L L - O L D S and M . A . K O C H
Fig. 2 Continued
The maximum parsimony analysis assigned the cp-types
to three approximately equally sized clades (subsequently
named lineage I, II, III) in the strict consensus tree of 10 000
shortest trees (Fig. 1; tree length = 169, CI = 0.89, RI = 0.96).
Rooting the tree, using Polyctenium fremontii and Halimolobus
parryii as outgroup, made lineage I sister to a combined
clade made up of lineage II + III.
TrnF tandem repeat copy II, copy III, and the indel at
alignment position 1004 –1100 (Fig. 2) were each assigned
to several well-separated clades. As inferred from Fig. 1
these repeats have been inserted or deleted at least four
times or eight times, respectively, and twice in parallel. For
this reason all indels marked with * or ** (see above) were
neglected in the reconstruction of the phylogeny. The com-
mon start and end motifs of these indels probably indicate
a common genetic mechanism responsible for their parallel
occurrence (Koch et al. 2003c).
Arabis × divaricarpa was scattered throughout the tree,
appearing in all major and most minor clades with high fre-
quencies, while all A. drummondii accessions grouped (except
Arab728, haplotype CW) in a monophyletic branch of lineage
II. Arabis holboellii showed in this respect a somewhat inter-
mediate behaviour. It dominated in lineage I and III (aside
from A. × divaricarpa and the previously mentioned A. drum-
mondii accession), but A. holboellii also occurred (19 accessions)
within the ‘A. drummondii-branch’ as well as in paraphyletic
position at the base of this A. drummondii lineage. However,
the two parental species were almost confined to different
clades. Therefore, a phylogeographical analysis of the major
lineages (II vs. I and III) is representative for the species (as
taxonomically perceived today). The various morpho-
logically defined varieties of A. holboellii (for details
refer to Table 1 – Supplementary Material) did not form
monophyletic groups, but rather took several polyphyletic
positions within all lineages.
Chloroplast phylogeny including information on
the rpoC1 intron
In order to verify the phylogenetic relationships based on
the trnL intron-trnL/F IGS, in addition the rpoC1 intron
was analysed for 44 selected accessions. The length of the
rpoC1 alignment was 747 bp. Within the 44 ingroup taxa 27
SNPs (including five singletons), four indels, and two poly
T stretches of variable length (alignment positions 560-570-
573; 600-606-608) were observed. The poly T stretches were
excluded from subsequent phylogenetic analysis because
of homoplastic occurrence of length variants. The phylogeny
of chloroplasts including the rpoC1 intron was completely
consistent with the relationships revealed using solely the
trnL intron-trnL/F IGS region (see above). Furthermore,
the enlarged data set resulted in a higher bootstrap support
for most clades (Fig. 3).
Geographic distribution of haplotypes
The geographical distribution of cp-types is shown in
Fig. 4(a)–(d). The combined distribution ranges of all
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349– 370
 P L E I S T O C E N E H Y B R I D I Z A T I O N O F A R A B I S S P E C I E S 359
cp-types grouped in lineage II of the strict consensus tree
covered the whole geographical range of the three species
investigated, whereas, those forming lineage III were distri-
buted in western North America only (aside from a few
geographical outliers in the east). Lineage I finally was con-
fined to the western United States (except haplotype G,
Whitehorse, southern Yukon). In this latter area therefore
the three lineages were codistributed.
Among haplotypes (with a frequency of at least N = 3),
most exhibited obvious restrictions in their individual geo-
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349–370
Fig. 3 Maximum parsimony analysis of 44
Arabis accessions based on trnL intron,
trnL/F IGS and rpoC1 intron sequences:
strict consensus tree of 54 most parsimoni-
ous trees. Bootstrap values (greater than
50%) are given above branches. Chloroplast
haplotypes are provided to the left of the
accession numbers.
graphical distributions. Their ranges varied in size, and
most of them overlapped with at least one other haplotype,
especially in southwestern North America. The regions
covered by the different cp-types correlated in several
cases with major land features (e.g. with stretches of
the Rocky Mountains or the Sierra Nevada), but no easily
perceivable general pattern emerged from examination
of their geographical ranges. All these features were
encountered independently for each of the three major
clades (Fig. 4a–d).
360 C . H . D O B E S, T . M I T C H E L L - O L D S and M . A . K O C H

Fig. 4 (a) to (d) Schematic geographical distribution ranges of particular haplotypes and historic events suggested by NCA to explain for
their mutual geographical associations. Letters indicate haplotypes; ‘+ d’ means ‘inclusively descendants’. Ancient haplotypes are given in
bold and underlined. Dashed, one-headed arrows represent range expansions, solid, double-headed arrows represent fragmentation
events. For all other haplotypes restricted gene flow with isolation by distance was suggested. The black, bold line indicates the extent of
the Wisconsin glaciation at the last glacial maximum. Some simplifications have been applied. (a) Historic events inferred for A. drummondii;
the six areas not named in the figure represent from south to north haplotypes BD, BA + BB, BF + BG, AK, AO + AP, and BC; the question
mark stands for the high age and low frequencies of haplotypes AA, AB, and AC, which makes it uncertain that range expansion of
haplotype AH started from this region. (b) to (d) A. holboellii.

Several geographical outliers were found for both A. ×
divaricarpa and A. holboellii at two locations: around the Great
Lakes (haplotype AH: Arab367, Arab368; BU: Arab753) and
near Anchorage, Alaska, close to the Pacific (CI: Arab593,
Arab596, Arab604).
As mentioned, Arabis × divaricarpa accessions shared almost
all chloroplast haplotypes (except singletons) with their
parental species, with a few exceptions occurring with low
frequency in the data set. Further haplotypes found in
A. × divaricarpa did not show distinct distribution ranges,
but were always mixed with parental ones. In A. holboellii
effective diversity (Gregorius 1978) of cp-types was high-
est in the Rocky Mountains from Idaho and Montana south
to Utah and Colorado as well as in the Sierra Nevada and
adjacent Cascades (Fig. 5a). In A. drummondii the highest
diversity scores were found in the Rocky Mountains of Colo-
rado. Lowest diversity was observed for the regions formerly
covered by the continuous ice sheet in all three species. A
similar picture emerged from the number of singleton haplo-
types (Fig. 5b). Out of 55 singletons only six were found
north of or at the southern margin of the last glacial maxi-
mum (LGM). Therefore, the majority of singleton haplotypes
was restricted to the area south of the LGM, 31 of these (15
haplotypes from the ‘A. drummondii-clade’; 16 characteristic
for A. holboellii) were restricted to the Rocky Mountains.
However, in contrast only three A. holboellii–like singletons

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349– 370

 P L E I S T O C E N E H Y B R I D I Z A T I O N O F A R A B I S S P E C I E S 361

Fig. 5 (a) Geographic distribution of effective genetic diversity of haplotypes according to Gregorius (1978) for 23 geographical subregions
(dotted lines) in North America. Bars give the individual genetic diversities (compare to scale in the bottom left) for Arabis drummondii
(white), A. × divaricarpa (grey) and A. holboellii (black). (b) Location of singleton haplotypes observed in A. drummondii + A. × divaricarpa of
lineage II (except basal clade 3-3; dots) and in A. holboellii + A. × divaricarpa of lineages I, III and of clade 3-3 (circles).

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349–370

362 C . H . D O B E S, T . M I T C H E L L - O L D S and M . A . K O C H

Fig. 6 Minimum-spanning network of 103 chloroplast haplotypes. Connections between haplotypes having a probability of at least 95% of
being parsimonious are indicated by solid, black lines. Grey, dotted lines are nonparsimonious connections between network I, network II
and haplotype Z, respectively. Dot-dashed lines represent ambiguous connections, which were rejected using the criteria suggested by
Crandall & Templeton (1993). The network was rooted at the reconstructed haplotype marked by a black star (in clade 3-5) applying the
root of the strict consensus tree resolved by the maximum parsimony analysis (Fig. 1). Reconstructed haplotypes (= intermediate haplotypes
not present in the data set) are depicted as black, small dots. Sampled haplotypes are indicated by circles, the surface area of which is
proportional to the number of individuals bearing it (N ranges from 1 to 147), while sectors indicate the partition of this haplotype between
species (white sectors Arabis drummondii, grey sectors A. × divaricarpa, black sectors A. holboellii). Nesting of haplotypes into hierarchically
interlocking clades, as outlined in Templeton et al. (1987, 1992), is represented by labelled boxes. Four-step clades are shown in a separate
diagram placed in the left lower corner; clades containing only one sampled subclade are given in dotted rectangles and were not numbered,
as they are uninformative in the subsequent test for association of genetic variation with geography. Clades, for which subclades showed
a statistically significant geographical pattern (see Table 1) are marked by an asterisk.

were observed in the Sierra Nevada. No singletons at all
were found for A. drummondii in the Sierra Nevada.
Haplotype network estimation and nested clade analysis
According to the 95% confidence limit of parsimonious
connections (P = 0.953) as revealed by the estimation
procedure (Templeton et al. 1992) and implemented in tcs
version 1.13 (Clement et al. 2000), the network-building algo-
rithm joined haplotypes up to four mutational steps apart,
resulting in two major networks as well as in one isolated
haplotype (Z) (Fig. 6). The single haplotype was excluded
from further analysis. The two networks and the isolated
haplotype could be manually joined via two five-step muta-
tional connections to provide information on their evolu-
tionary relationship. However, because this connection
exceeded the 95% confidence limit of parsimony (P = 0.929),
the networks were treated subsequently as independent
entities. The topology of the haplotype network was con-
gruent [except for the connection of clade 1-23 ( = haplotype
S + T) with M instead with K] with that of the strict con-
sensus tree resolved by the maximum parsimony analysis
(Fig. 1). As a result of the inclusion of information on indels
in the repeats located in the trnL/F IGS, additional resolution
for several terminal clades was evident in this analysis.
Two obvious features were observed for the polarized
networks (Fig. 6): (i) The frequencies of particular haplo-
types increased from the root toward the middle portion of
the networks, and again declined for most terminal haplo-
types. (ii) Inferred (nonobserved) haplotypes were much
more frequent at the base of the networks. As shown in
Fig. 7 there is a significant increase in Dcs (a measure for

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349– 370

 P L E I S T O C E N E H Y B R I D I Z A T I O N O F A R A B I S S P E C I E S 363
Fig. 7 Correlation between age of haplotypes and their individual
geographical distribution expressed by Dc computed using
geodis 2.0. The most ancient cp-types of lineages I, II, and III
haplotypes B, AB, and BR show a drop in size of their geographical
ranges. Whereas haplotypes AS, BY, CG (AH, BU) occupy areas
larger than average.
the geographical range of a particular clade) with the age
of haplotypes (Pearson’s r = 0.63; singletons and outliers
haplotype AS, BY, CG were excluded from the calculation).
Dc s of haplotypes ranged from 3.9 to 1455.0 kilometres.
However, the diagram shows a drop in the size of geo-
graphical ranges for the oldest haplotypes B, AB and BR.
This decrease may be causally related to the low numbers
of these haplotypes within the investigated material (Fig. 6),
if one assumes gradual (and/or accidental) extinction of
haplotypes by genetic drift over time (Avise 2000). An old
haplotype will then be under-represented in the sampled
material and underestimation of the size of its distribution
range will be likely. The test of association of haplotype
variation with geography using geodis 2.0, revealed statis-
tically significant values for Dc, Dn (the geographical range
of a particular clade relative to its closest sister clades)
(I–T)D (Dcs of interior minus that of tip clades), or (I–T)Dn
c
(Dn interior–Dn tip) within several clades (see Table 1 for
details). The analysis was rerun for lineage II/A. drummondii-
like-haplotypes after excluding all A. holboellii accessions,
but results were identical with respect to inferred historic
events (the test could not be performed on clade 4-1 as its
interior clade was made up by A. holboellii).
Three different types of historic events were inferred by
NCA: restricted gene flow with isolation by distance; past
fragmentation; and contiguous range expansion (Table 1,
Fig. 4a–d). These historic processes were each associated
with general features with respect to geographical distri-
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349–370
bution of the haplotypes involved. All haplotypes for
which restricted gene flow with isolation by distance was
suggested were nested within the geographical ranges of
their ancestors and were found south of the last glacial maxi-
mum. Past fragmentations were deduced for haplotypes
geographically separated between the Sierra Nevada, the
southern Cascades, the Southern Rockies, or the Colorado
Plateau in the south and the Central and Northern Rocky
Mountains of the US and the adjacent Cascades in the north.
Contiguous range expansions in contrast were suggested
for all haplotypes (except for haplotype CG and singletons),
found within regions formerly covered by the contiguous
ice shield.
All three processes were inferred for A. holboellii, but only
restricted gene flow and range expansion were suggested
for A. drummondii. However, according to these results the
two evolutionary lineages made up by A. holboellii behaved
conspicuously differently. Lineage III was dominated by
accessions bearing haplotypes suggested to have under-
gone range expansions (Fig. 4a), whereas the NCA method
applied to lineage I revealed past fragmentation events
(Fig. 4b). Another interesting observation was that seven
out of eight events of restricted gene flow by isolation by
distance suggested for A. drummondii, respectively, lineage
II were located in (the majority restricted to) the Rockies,
with the only remaining one (haplotype AJ) confined to the
northern Sierra Nevada (Fig. 4a). However, haplotype AJ
was exceptional as 10 out of 11 accessions belonged to
A. × divaricarpa but only one to A. drummondii.
Genetic differentiation between major land features
according to NST statistics and pairwise population
comparisons
NST values did not exceed GST values in the test of genetic
differentiation between subregions, neither for the total
sample nor for major lineages or species (except A. holboellii;
Table 2). Therefore, according to this test, there is no or
only weak overall correspondence between these phylogenies
and geographical distribution. This result is presumably
the result of the wide geographical codistribution of the
most numerous haplotypes. However, the neighbour-joining
trees based on average numbers of pairwise distances
between chloroplast haplotypes often joined adjacent sub-
regions. Arabis drummondii-like haplotypes were assigned
to three clusters made up by regions corresponding to the
Cordilleran ice sheet, to the Laurentide ice sheet, and to
all regions unglaciated during the LGM (Western US),
respectively (Fig. 8a). In contrast, A. holboellii-like haplotypes
showed a differentiation between the Western Cordillera
of the United States and a combined Eastern area (‘Rocky
Mountains plus Great Basin’), while regions formerly covered
by the contiguous ice shield of the LGM formed a third
cluster (Fig. 8b).
364 C . H . D O B E S, T . M I T C H E L L - O L D S and M . A . K O C H
Fig. 8 Neighbour joining trees of 23 geographical subregions based on average numbers of pairwise distances (Nei & Li 1979) between A.
drummondii-like and A. holboellii-like chloroplast haplotypes using uncorrected P-distances.
Discussion
Age estimation of chloroplast haplotypes
Based on plastid matK-sequences, nuclear chalcone synthase
and alcohol dehydrogenase data Koch et al. (2000, 2001)
estimated that divergence between Arabidopsis and North
American Arabis (= Boechera) occurred between 6 and 15
million years ago. In comparison, sequence divergence of
the trnL intron was 4 –5% (uncorrected P-distance) between
Arabidopsis thaliana and Arabis. That means, considering
the length of the intron of approximately 500 bp, an overall
difference of 20–25 mutations separating the two lineages.
As the null hypothesis of rate heterogeneity was rejected
for the trnL intron using rr-test (P values of pairwise lineage
comparisons ranged from 0.81 to 0.17), it was estimated
that there would be approximately one mutational change
per 250 000–800 000 years. Within the investigated Arabis
material, haplotypes differed from zero to eight muta-
tions in the trnL intron. Therefore, this estimate suggests a
divergence date of roughly 1–3 million years ago for the
most distantly related cpDNA lineages. The timescale covered
by the phylogeny would therefore span at least the second
half of the Pleistocene (Richmond & Fullerton 1986). From
these data we infer that plastome type differentiation
predates the speciation of A. drummondii and possibly
that of A. holboellii. However, nuclear data indicate a more
recent Pleistocene origin of the two species and their
hybrid as nuclear DNA ITS pairwise sequence differences
(Koch et al. 2003b) were lower (0.3–1.3%) between these
parental species and four closely related Arabis species
than sequence variation of the trnL intron and the trnL-F
spacer region (up to 2.1% sequence differences), although
these latter sequences evolve with much lower substitution
rates than the ITS region (Hewitt 2001).
Comparative phylogeography of species and of major
evolutionary lineages resolved by the gene tree
Very high genetic polymorphism was detected within the
sequenced cpDNA region. This finding is unusual, as most
intrageneric phylogeographical studies in plants have
found little sequence divergence (Schaal et al. 1998; Hewitt
2001). The observed degree of divergence is all the more
remarkable considering that the corresponding ribosomal
ITS displayed comparably low variation (Koch et al. 2003b).
The ratio between divergence of the trnL intron and the
ribosomal ITS was 1 : 1, while, in contrast, ratios in other
Brassicaceae genera varied from 1 : 1.5 in Draba (Koch &
Al-Shebaz 2002) to 1 : 18 in Yinshania (Koch & Al-Shebaz
2000). A number of inferred ancestral haplotypes dating
back to the divergence of the three major evolutionary
lineages were not detected in the investigated material
(Fig. 6). Either these missing haplotypes still exist outside
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349– 370
 P L E I S T O C E N E H Y B R I D I Z A T I O N O F A R A B I S S P E C I E S 365
our sample or, more likely, have been lost through extinc-
tion. Nevertheless, the three major lineages observed are
clearly distinct (e.g. extant haplotypes of lineages I and III
are separated by six mutational steps).
Templeton’s NCA seeks to distinguish among a diverse
array of historical processes (Templeton et al. 1995). Although
this approach has performed well in the case of range
expansions (Templeton 1998), NCA does not provide
confidence limits on the reconstructed history (Knowles &
Maddison 2002). However, the criticism of Petit & Grivet
(2002), that genetic fixation within populations can bias the
outcome of a NCA, should not apply to our data, as we
sampled single individuals from isolated localities.
Haplotypes observed in A. × divaricarpa did not show
distinct distribution ranges but were always found mixed
with parental ones. This observation is in concordance with
the polyphyletic origin of the hybrid, as has been demon-
strated by Koch et al. (2003b) and suggested by Rollins (1983).
For these reasons data on the hybrid were pooled with those
on its parents. In contrast, parental species are genetically
well differentiated as revealed by this study and suggested
by Koch et al. (2003b). Hence, A. drummondii-like and
A. holboellii-like haplotypes were analysed separately.
One principle of NCA is that the geographical distribu-
tion of derived haplotypes is compared against that of their
ancestors. Two major shortcomings related to the distribu-
tion of interior clades may be encountered in this analysis.
First, the nesting procedure often does not only include
the ancestral haplotype but also includes closely related
descendants into the interior clade (Templeton et al. 1987,
1992). Such a nesting design can be considered contra-
dictory to the basic idea of NCA to compare the distribution
ranges of derived haplotypes against that of their ances-
tors. Second, unless populations grow, haplotypes will
theoretically become extinct over time (Avise 2000, p. 45).
Therefore, there will be a limit to the detection of the actual
geographical distribution range of old haplotypes at least
by the number of samples affordable. However, the accur-
ate estimation of the distribution ranges of haplotypes is
crucial to the outcome of NCA.
Disregarding the hybrid species A. × divaricarpa, recoloniza-
tion of the area covered by the Cordilleran and Laurentide
ice shields was restricted to three taxa, A. drummondii, A.
holboellii var. collinsii and var. retrofracta (cf. Rollins 1993),
and involved only four haplotypes (AH, AS, BY, CG; plus
a few of their descendants; Fig. 4a,c). The colonization was
explained by NCA (except for haplotype CG) to have taken
place via expansion of the ranges of these haplotypes. It is
obvious that these haplotypes have indeed experienced fast
range expansions following the deglaciation of this vast
region. Analogous scenarios have been described by a series
of studies (e.g. Graham & Mead 1987; Hewitt 1993; Lewis
& Crawford 1995; Templeton 1998; Marshall et al. 2002).
The failure to detect range expansion also for haplotype
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349–370
CG was the result of inflation of the geographical range
of the interior clade by inclusion of derived haplotype
BY. Thus range expansion was inferred for haplotype CG
when BY was excluded from the test (results not shown).
That range expansion was indeed the likely process
leading to the contemporary geographical distribution of
haplotype CG is also suggested by Fig. 7, as this haplotype
covers an area of similar size to haplotypes AS and BY.
However, the data do not allow pinpointing of the geo-
graphical origin of the expansions, and possible refugia in
boreal and arctic regions might have contributed to the
recolonization beside the mountain ranges south of the
continuous ice sheet. The migration patterns of our sampled
taxa and haplotypes might be explained by selective advant-
ages of ecotypes preadapted to boreal and arctic conditions
(e.g. A. drummondii is a montane to alpine plant; and the
variety ‘collinsii’ is known primarily from boreal habitats;
Rollins 1993). Interestingly the involved haplotypes belonged
to genealogical lineages II and III only. On the other hand,
it is known from simulation studies and observation that
early colonizers established by chance may suppress later
invaders, hence precluding subsequent increases in diversity
(Hewitt 1993; Ibrahim et al. 1996; Corre et al. 1997; Hewitt &
Ibrahim 2001). The current study cannot distinguish between
these two alternatives, as all lineages occupy similar areas
and habitats south of the former contiguous ice sheet, and
include all varieties of A. holboellii investigated.
The only range expansion event outside the continu-
ous ice sheet of the Wisconsin glaciation was suggested
for haplotype CI (respectively clade 2-7; lineage III; Fig.
4d). However, this inference needs further evidence to be
accepted, as the interior clade, on which this test was based,
comprised only three haplotypes (six accessions), the ancestor
being a singleton (haplotype AB). Several fragmentation
events were inferred for A. holboellii by NCA in lineage I.
This inference was based on the geographical separation
of derived southern haplotypes (haplotypes D, F, S, U +
descendants; Fig. 4b) from their ancestors (haplotypes B
and M in the Central to Northern Rocky Mountains and
adjacent Cascades). This pattern coincided with analogous
disjunct distributions of several haplotypes (A, K, M, BT),
an observation which is in favour for the past fragmenta-
tion hypothesis. However, two arguments favour more
recent colonization (range expansion, long-distance dis-
persal) as an alternative explanation for the contemporary
distribution of haplotypes: (i) that the southern haplotypes
are descendants of the northern ones and (ii) the extremely
low numbers of singleton haplotypes for A. holboellii found
in the Sierra Nevada (Fig. 5b). However, southerly areas
are now dominated by more derived haplotypes. The
major uplift of the Sierra Nevada did not occur until 2 to 5
million years ago (King 1959; Brunsfeld et al. 2001). Given
the above estimates of haplotype ages, the uplift of these
mountains could have influenced the isolation of ancient
366 C . H . D O B E S, T . M I T C H E L L - O L D S and M . A . K O C H
populations within the range. The ‘modern’ haplotypes might
have replaced ancient lineages because of climatically
unfavourable conditions (Wahrhaftig & Birman 1965). The
extensive past glaciations of this mountain chain (Wahrhaftig
& Birman 1965) would fit such a scenario.
The past fragmentation event suggested for A. holboellii
involving haplotypes located at the basis of lineage II (clade
3-3) constituted another case of rare haplotypes building
up the interior clade (ancestral haplotype AB being a
singleton). The partition of populations between southern
Nevada (interior clade 2-4) and the Rocky Mountains of
Montana and Wyoming (tip clade 2-5) should therefore not
be accepted unless confirmed by additional evidence.
The historic event most frequently inferred by NCA was
restricted gene flow by isolation by distance (Table 1). This
pattern was observed for both parental species alike and
was explained by limited movement of individuals bear-
ing newly arisen haplotypes (Neigel et al. 1991; Neigel &
Avise 1993; Templeton 1998). This process was inferred from
our data when young, derived haplotypes were embedded
within the geographical range of their ancestors. Further-
more, this pattern was only observed in areas which
remained largely ice-free during past glaciations. Young
haplotypes showed restricted, often nonoverlapping, geo-
graphical distribution ranges within formerly unglaciated
areas, whereas only haplotypes of the middle portions of
the phylogenetic network were geographically widespread
(e.g. haplotypes U, AH, AS, BU, CI; Fig. 4a– d). Therefore,
it seems likely that limited spread of seeds constrains the
geographical distribution of cpDNA haplotypes. Accept-
ing the validity of the applied molecular clock chloroplast
haplotypes (Fig. 7) colonized in about 250 000 –1 000 000
years the whole western USA. Therefore, time rather than
physical or ecological barriers was the main factor limiting
gene flow via seeds within the period covered by our data.
However, distribution rates of seeds may become extremely
high in the case of recolonizations as formerly ice-covered
locations at least 3000 km distant from possible refugia
have been reached within the Holocene (Hewitt 1993; Hewitt
& Ibrahim 2001; Petit et al. 2001; Palmé et al. 2003).
For the above reasons all frequent haplotypes were to
some extent codistributed across North America. This
pattern is reflected in NST values not exceeding the GST values
(Pons & Petit 1996) indicating that there is no overall
correspondence between the phylogenies of haplotypes and
their geographical distribution (Table 2). This result demon-
strates that the contemporary geographical distribution
of genetic variation is the outcome of several incongruent
historical processes (as revealed by NCA), which were
blurred by pooling the data. However, the neighbour-
joining tree based on average numbers of pairwise dis-
tances between subregions resolved areas formerly covered
by the Cordilleran and Laurentide ice sheet as well as the
Sierra Nevada/Southern Cascades as distinct entities (Fig. 8).
This finding is consistent with the fragmentation and range
expansion events inferred by NCA for the Sierra Nevada
and Cascades (Fig. 4b) in the case of A. holboellii, as well as
with the inferred major range expansion of A. drummondii–
like haplotypes into western vs. eastern deglaciated regions
in the north (Fig. 4a,c).
The whole phylogeographical scenario fits the ‘Category
II phylogeographical pattern: Deep gene tree, major
lineages broadly sympatric’ suggested by Avise (2000).
This pattern applies to A. holboellii and may be either
explained by (i) the maintenance of large effective popu-
lation sizes and of high gene flow retaining ancient, co-
distributed lineages over time, or by (ii) secondary admixture
between allopatrically evolved populations. Hybridization
is a likely phenomenon associated with the latter explana-
tion (Barton & Hewitt 1985). One might speculate that the
frequent intraspecific hybridization events observed for
A. holboellii (Koch et al. 2003b) were promoted by secondary
admixture of the three major evolutionary lineages observed
in this study. However, to demonstrate the validity of such
a scenario would demand the pinpointing of different
geographical centres of origin for these lineages. But such
evidence was not provided by the data. The phylogeography
of A. drummondii in contrast is only part of this ‘Category II-
pattern’ and fits the ‘Category V-pattern’ suggested by Avise
(2000), which involves common, ancestral lineages that
are widespread plus closely related lineages that are
geographically restricted.
Hybridization
Arabis × divaricarpa accessions shared almost all chloroplast
haplotypes with their parental species, with a few excep-
tions occurring at low frequency in the dataset. This means
also, that no obvious preferences with respect to phylogenetic
age of haplotypes are evident. However, interestingly
A. drummondii served more frequently as mother of A. ×
divaricarpa than A. holboellii (149 vs. 56 times). This finding
was unexpected as pollen quality is often decreased in
A. holboellii (Koch et al. 2003b) and self-fertilization was
suggested to predominate in A. drummondii (Roy 1995).
Further, haplotypes found in A. × divaricarpa did not form
distinct distribution ranges, but were always found mixed
with parental ones. These findings suggest that the evolu-
tion of A. × divaricarpa is closely related to that of both
parental species, implying repeated hybridization, as
suggested by Rollins (1983) and Koch et al. (2003b).
Arabis drummondii accessions constituted a monophyletic
group within lineage II, while ancestors of this lineage
consisted of haplotypes which have only been observed in
A. holboellii (Figs 1 and 6). This phylogenetic pattern might
be interpreted as an ancestor–descendant relationship.
However, differential sorting of haplotypes predating the
divergence of the two species, as well as introgression of
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349– 370
 P L E I S T O C E N E H Y B R I D I Z A T I O N O F A R A B I S S P E C I E S 367
one species into the other resulting in chloroplast capture,
should also be considered. Two related observations are
important: (i) A. drummondii and A. holboellii accessions
bearing shared haplotypes were codistributed (e.g. geo-
graphically restricted haplotypes AO and BJ), and (ii) 16
out of 18 A holboellii accessions (except singleton haplotype
AG and Arab671: AH) were confined to the Central Rocky
Mountains of Colorado, Utah, Wyoming and southwestern-
most Montana. Therefore, lineage sorting seems unlikely,
as one would not expect congruence in distribution of
genetic variability in evolutionarily unrelated species (Schaal
et al. 1998). Based on a comprehensive investigation of
concerted evolution of ribosomal ITS, Koch et al. (2003b)
suggested for five A. holboellii accessions introgression by
A. drummondii. These accessions carry haplotypes AS, AT and
BJ. On the other hand, AH, which is basal, was represented
by six accessions in which only A. holboellii-ITS-types were
found. Therefore, both common ancestry and reticulation
may explain the co-occurrence of haplotypes in A. drummondii
and A. holboellii. However, the phylogenetic origin of A.
drummondii must be addressed in a broader taxonomic con-
text, and nuclear markers not prone to the effects of concerted
evolution, like microsatellites, will be used to answer this.
A single origin of the morphologically defined varieties
of A. holboellii (Rollins 1993) is not supported by this study,
as the phylogeny of chloroplast haplotypes was incon-
gruent with the proposed taxonomic concept. Lineage
sorting might explain this contradiction, but seems unlikely,
as few disjunct distribution patterns of haplotypes were
observed (Schaal et al. 1998). Therefore, pollen-mediated
gene transfer seems more likely to explain the observed
discrepancy. Reticulation would also explain the strik-
ing discrepancy between nuclear DNA (ITS) and cpDNA
divergence in A. holboellii. Monophyly of the taxonomically
well-defined taxon A. drummondii, on the other hand, is con-
cordant with the marked geographical concentration of chloro-
plast haplotype diversity and a possible monotopic origin of
the species in the Central and Southern Rocky Mountains.
Centres of diversity and possible glacial refugia
High diversity of chloroplast haplotypes was found for the
Rocky Mountains south of the contiguous ice sheet (Fig. 5a).
Likewise, singleton haplotypes showed highest frequencies
within this mountain area and adjacent regions of the Great
Basin (Fig. 5b). These observations suggest centres of genetic
diversity in these areas (Hewitt 1996; Comes & Kadereit
1998; Widmer & Lexer 2002; Slatkin 1985). Interestingly,
almost all ancient haplotypes (B, K, M, U, AH, AS, BU, BR,
CI; except AB) occurred or were restricted to this mountain
area and the majority of derived haplotypes were concen-
trated here (Figs 4 and 5). Especially A. drummondii showed
by far the highest chloroplast diversity within this area
(e.g. no singleton haplotypes were found for this species in
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349–370
the Sierra Nevada). Therefore, past conditions during the
Pleistocene allowed for the accumulation and preservation
of genetic variation in this part of the Rocky Mountains. In
contrast, in the Sierra Nevada ancient haplotypes (except
widespread haplotypes AH, AS and CI) were either missing
or comparably rare. As outlined above, Pleistocene fragmenta-
tion or colonization events, as well as subsequent radi-
ations, are likely to have played an important role in shaping
the contemporary distribution of genetic variation within
this mountain range. However, both mountain ranges are
known to have been important glacial refugia (Brunsfeld
et al. 2001) and evidently played a similar role in the evolu-
tion of these Arabis species.
Several haplotypes were detected in boreal and arctic
regions (especially the Great Lakes: AH, BU, Fig. 4a,c; and
near the Pacific coast: G, Fig. 5b; CI, Fig. 4d), which were
geographically isolated from their ancestors. Only a few
derivatives (AL, AU, AZ, CF, CU) of those haplotypes
(AH, AS, BY, CG), which have spread into areas formerly
covered by the Laurentide and Cordilleran ice sheet, were
found within these newly recolonized regions (Fig. 5b).
This latter observation, together with a rate of one mutation
per 250 000–800 000 years for the trnL intron as suggested
above, implies that almost all observed haplotypes origin-
ated prior to the Wisconsin glaciation. Reliable evidence
for glacial refugia in the far north is known for Beringia,
for areas southeast of the contiguous ice sheet from New
Jersey to Quebec, and for the eastern Arctic (e.g. Hultén
1937; Tremblay & Schoen 1999; McCusker et al. 2000;
Fleming & Cook 2002; Abbott & Brochmann 2003; Abbott
et al. 2003). The present distribution of haplotypes G
(southwesternmost Yukon); AH and BU (Great Lakes),
AL and AU (at the St Lawrence in Quebec and in Maine)
could be explained by survival within these glacial refugia.
Also, the Canadian Archipelago was suggested as a poten-
tial glacial refugium (Tremblay & Schoen 1999; Abbott
et al. 2003), but this hypothesis is still based on little evid-
ence. The possible role of the southern Alaskan Pacific
coast as glacial refugium was discussed by Demboski et al.
(1999), Smith et al. (2001) and Fleming & Cook (2002). The
observation of three isolated accessions of haplotype CI
near Anchorage, Alaska, was of special interest, as this area
was heavily glaciated during the LGM (Péwé et al. 1965 cf.
http://instaar.colourado.edu/QGISL/ak_palaeoglacier_
atlas/gallery/index.html) and is separated from inland-
Beringia by the Alaska Range. The opening of continental
shelves along the Pacific coast presumably provided some
ice-free areas (Frenzel 1968; Frenzel et al. 1992; Mann &
Peteet 1994), e.g. in the Caribou Hills east of lower Cook
Inlet (Reger & Pinney 1996), which might have served as a
glacial refugium for these haplotypes. However, no indis-
putable evidence exists for the persistence of glacial refugia
along the Pacific margin of the Cordilleran ice sheet (Mann
& Hamilton 1995).
368 C . H . D O B E S, T . M I T C H E L L - O L D S and M . A . K O C H
Acknowledgements
This work was supported by a grant to M. K. from the Austrian
Science Foundation (FWF; P14655-GEN). T. M. O. was supported
by NSF grant DEB-9527725 and by the Max-Planck Gesellschaft.
We are grateful to D. Schnabelrauch (Jena) for technical support
in the laboratory and to the curators of herbaria MO and GH for
providing leaf material for destructive sampling. Special thanks
goes to Timothy Sharbel for critically reading the manuscript and
to two anonymous reviewers and the subject editor Rémy Petit for
numerous suggestions.
Supplementary material
The following material is available from
http://www.blackwellpublishing.com/products/journals/
suppmat/MEC/MEC2064/MEC2064sm.htm
Table S1. List of investigated accessions of Arabis drummondii, A.
holboellii and their hybrid A. × divaricarpa Accessions are ordered
within taxa by material number (in bold characters and at the
end of the record) assigned by the authors; additional information
provided by the authors as well as herbarium abbreviations are put
into brackets; the material number is followed by the chloroplast
haplotype
Table S2. List of investigated accessions of Arabis drummondii, A.
holboellii and their hybrid A. × divaricarpa accessions are ordered by
chloroplast haplotype, taxon and material number. Geographic
coordinates are provided in order to allow mapping of exact dis-
tribution of each haplotype
Figure S1. Collection sites of accessions (populations) listed in
Supplementary Table 1.
References
Abbott RJ, Brochmann Ch (2003) History and evolution of the arc-
tic flora: in the footsteps of Eric Húlten. Molecular Ecology, 12,
299–313.
Abbott RJ, Smith LC, Milne RI et al. (2003) Molecular analysis of
plant migration and refugia in the arctic. Science, 289, 1343–1346.
Avise JC (2000) Phylogeography. The History and Formation of Species.
Harvard University Press, Cambridge, London.
Avise JC, Arnold J, Ball RM (1987) Intraspecific phylogeography:
the mitochondrial DNA bridge between population genetics
and systematics. Annual Review of Ecology and Systematics, 18,
489–522.
Barton NH, Hewitt GM (1985) Analysis of hybrid zones. Annual
Review of Ecology and Systematics, 16, 113 –148.
Birky CM, Fuerst P, Maruyama T (1989) Organelle gene diversity
under migration, mutation, and drift: equilibrium expectations,
approach to equilibrium, effects of heteroplasmic cells, and
comparison to nuclear genes. Genetics, 121, 613 – 627.
Böcher TW (1951) Cytological and embryological studies in
the amphi-apomictic Arabis holboellii complex. Det Kongelige
Danske Videnskabernes Selskab Biologiske Skrifter (København),
6, 1–59.
Böcher TW (1969) Further studies in Arabis holboellii and allied
species. Saertryk Af Botanik Tidsskrift, 64, 141–161.
Brunsfeld SJ, Sullivan J, Soltis DE, Soltis PS (2001) Comparative
phylogeography of northwestern North America: a synthesis.
In: Integrating Ecology and Evolution in a Spatial Context (eds
Silvertown J, Antonovics J), pp. 319–339. Blackwell Science,
Oxford.
Clement M, Posada D, Crandall KA (2000) TCS: a computer program
to estimate gene genealogies. Molecular Ecology, 9, 1657–1659.
Comes HP, Kadereit JW (1998) The effect of Quaternary climatic
changes on plant distribution and evolution. Trends in Plant
Science, 3, 432–438.
Comes HP, Abbott RJ (2001) Molecular phylogeography, reticula-
tion, and lineage sorting in Mediterranean Senecio sect. Senecio
(Asteraceae). Evolution, 55, 1943–1962.
Corre V, Machon N, Petit RJ, Kremer A (1997) Colonization with
long-distance dispersal and genetic structure of maternally
inherited genes in forest trees: A simulation study. Genetical
Research, 69, 117–125.
Crandall K, Templeton AR (1993) Empirical tests of some pre-
dictions from coalescent theory with applications to intraspecific
phylogeny reconstruction. Genetics, 134, 959–969.
Demboski JR, Stone KD, Cook JA (1999) Further perspectives on
the Haida Gwaii glacial refugium. Evolution, 53, 2008–2012.
Doyle JJ, Doyle JL (1987) A rapid DNA isolation procedure for
small amounts of fresh leaf tissue. Phytochemical Bulletin, 19, 11–
15.
Ennos RA (1994) Estimating the relative rates of pollen and seed
migration among plant populations. Heredity, 72, 250–259.
Fleming MA, Cook JA (2002) Phylogeography of endemic ermine
(Mustela erminea) in southeast Alaska. Molecular Ecology, 11,
795–807.
Frenzel B (1968) The Pleistocene vegetation of northern Eurasia.
Science, 161, 637–649.
Frenzel B, Pécsi M, Velichko AA (1992) Atlas of Palaeoclimates and
Palaeoenvironments of the Northern Hemisphere. Gustav Fischer,
Stutlgart.
Graham RW, Mead JI (1987) Environmental fluctuations and evo-
lution of mammalian faunas during the last deglaciation in North
America. In: North America and Adjacent Oceans During the Last
Deglaciation, the Geology of North America, V K-3 (eds Ruddiman WF,
Wright HE Jr), pp. 371–402. Geological Society of North America,
Boulder, CO.
Gregorius H-R (1978) The concept of genetic diversity and its
formal relationship to heterozygosity and genetic distance.
Mathematical Biosciences, 41, 253–271.
Hamann A, El K, Koshy MP, Namkoong G (1998) Multivariate
analysis of allozymic and quantitative trait variation in Alnus
rubra: geographic patterns and evolutionary implications. Cana-
dian Journal of Forest Research, 28, 1557–1565.
Hewitt GM (1993) Postglacial distribution and species substructure:
lessons from pollen, insects and hybrid zones. In: Evolutionary
Patterns and Processes (eds Lees DR, Edwards D), pp. 98–123.
Academic Press, London.
Hewitt GM (1996) Some genetic consequences of ice ages, and
their role in divergence and speciation. Biology Journal of the
Linnean Society, 58, 247–276.
Hewitt GM (2001) Speciation, hybrid zones and phylogeography
— or seeing genes in space and time. Molecular Ecology, 10, 537–
549.
Hewitt GM, Ibrahim KM (2001) Inferring glacial refugia and his-
torical migrations with molecular phylogenies. In: Integrating
Ecology and Evolution in a Spatial Context (eds Silvertown J,
Antonovics J), pp. 271–294. Blackwell Science, Oxford.
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349– 370
 P L E I S T O C E N E H Y B R I D I Z A T I O N O F A R A B I S S P E C I E S 369
Hultén E (1937) Outline of the History of Arctic and Boreal Biota
During the Quaternary Period. Lehre J. Cramer, New York.
Ibrahim KM, Nichols RA, Hewitt GM (1996) Spatial patterns
of genetic variation generated by different forms of dispersal
during range expansion. Heredity, 77, 282 –291.
King PB (1959) The Evolution of North America. Princeton Univer-
sity Press, Princeton, NJ.
Knowles LL, Maddison WP (2002) Statistical phylogeography.
Molecular Ecology, 11, 2623 –2635.
Koch M (2003) Molecular phylogenetics, evolution and popula-
tion biology in the Brassicaceae. In: Plant Genome: Biodiversity
and Evolution, Vol. 1: Phanaerogams (eds Sharma AK, Sharma A).
Science Publishers, Enfield, NH.
Koch M, Al-Shebaz IA (2000) Molecular systematics of the Chinese
Yinshania (Brassicaceae): evidence from plastid and nuclear ITS
DNA sequence data. Annals of the Missouri Botanical Garden, 87,
246–272.
Koch M, Al-Shebaz IA (2002) Molecular data indicate complex
intra- and intercontinental differentiation of American Draba
(Brassicaceae). Annals of the Missouri Botanical Garden, 89, 88–
109.
Koch M, Bishop JG, Mitchell-Olds Th (1999) Molecular systemat-
ics and evolution. Of Arabidopsis and Arabis Plant Biology, 1, 529–
537.
Koch M, Haubold B, Mitchell-Olds Th (2000) Comparative
evolutionary analysis of chalcone synthase and alcohol dehydro-
genase loci in Arabidopsis, Arabis, and related genera (Brassicaceae).
Molecular Biology and Evolution, 17, 1483 –1498.
Koch M, Haubold B, Mitchell-Olds Th (2001) Molecular system-
atics of the Brassicaceae: evidence from coding plastidic matK
and nuclear Chs sequences. American Journal of Botany, 88, 534–
544.
Koch M, Al-Shebaz IA, Mummenhoff K (2003a) Molecular system-
atics, evolution, and population biology in the mustard family
(Brassicaceae). Annals of the Missouri Botanical Garden, 90, 151–
171.
Koch M, DobeS Ch, Mitchell-Olds Th (2003b) Multiple hybrid
formation in natural populations: concerted evolution of the
internal transcribed spacer of nuclear ribosomal DNA (ITS)
in North American Arabis divaricarpa (Brassicaceae). Molecular
Biology and Evolution, 20, 338 – 350.
Koch M, DobeS Ch, Matschinger M (2003c) The trnF (GAA) gene
in cruciferous plants: extensive duplication, variation in copy
number and parallel evolution. Biodiversity and Evolutionary
Biology, 17th Senckenberg Conference. Palmarum Hortus
Francofurtensis 7. Frankfurt am Main.
Lewis PO, Crawford DJ (1995) Pleistocene refugium endemics
exhibit greater allozymic diversity than widespread congeners
in the genus Polygonella (Polygonaceae). American Journal of Botany,
82, 141–149.
Mann DH, Peteet DM (1994) Extent and time of the last glacial
maximum in Southwest. Alaska Quaternary Research, 42, 136 –148.
Mann DH, Hamilton TD (1995) Late Pleistocene and Holocene
palaeoenvironments of the north Pacific coast. Quaternary Sci-
ence Reviewss, 14, 449 – 471.
Marshall HD, Newton C, Ritland K (2002) Chloroplast phylo-
geography and evolution of highly polymorphic microsatellites
in lodgepole pine (Pinus contorta). Theoretical and Applied Genetics,
104, 367–378.
Martinez P, Lopez C, Roldan M, Sabater B, Martin M (1997) Plastid
DNA of five ecotypes of Arabidopsis thaliana: sequence of ndhG
gene and maternal inheritance. Plant Science, 123, 113 –122.
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349–370
McCusker MR, Parkinson E, Taylor EB (2000) Mitochondrial
DNA variation in rainbow trout (Oncorhynchus mykiss) across its
native range: testing biogeographical hypotheses and their
relevance to conservation. Molecular Ecology, 9, 2089–2108.
Mitchell-Olds Th (2001) Arabidopsis thaliana and its wild relatives:
a model system for ecology and evolution. Trends in Ecology and
Evolution, 16, 693–700.
Naumova TN, van der Laak J, Osadtchiy J et al. (2001) Reproduc-
tive development in apomictic populations of Arabis holboellii
(Brassicaceae). Sexual Plant Reproduction, 14, 195–200.
Nei M, Li W-H (1979) Mathematical model for studying genetical
variation in terms of restriction endonucleases. Proceedings of the
National Academy of Sciences of the United States of America, 76,
5269–5273.
Neigel JE, Avise JC (1993) Application of random-walk model to
geographic distributions of animal mitochondrial DNA vari-
ation. Genetics, 135, 1209–1220.
Neigel JE, Ball RM, Avise JC (1991) Estimation of single generation
migration distances from geographic variation in animal mito-
chondrial DNA. Evolution, 45, 423–432.
Palmé AE, Su Q, Rautenberg A, Manni F, Lascoux M (2003)
Postglacial recolonization and cpDNA variation of silver birch
Betula pendula. Molecular Ecology, 12, 201–212.
Petit RJ, Grivet D (2002) Optimal randomization strategies when
testing, the existence of a phylogeographic structure. Genetics,
161, 469–471.
Petit RJ, Bialozyt R, Brewer S, Cheddadi R, Comps B (2001) From
spatial patterns of genetic diversity to postglacial migration
processes in forest trees. In: Integrating Ecology and Evolution in
a Spatial Context (eds Silvertown J, Antonovics J), pp. 295 –318.
Blackwell Science, Oxford.
Péwé TL, Hopkins DM, Giddings JL (1965) The Quaternary geo-
logy and archeology of Alaska. In: The Quaternary of the United
States (eds Wright HE Jr, Frey DG), pp. 355 – 374. Princeton
University Press, Princeton, NJ.
Pons O, Petit RJ (1996) Measuring and testing genetic differentiation
with ordered versus unordered alleles. Genetics, 144, 1237–1245.
Reger RD, Pinney DS (1996) Late Wisconsin glaciation of the Cook
Inlet region with emphasis on Kenai lowland and implications
for early peopling. In: The Anthropology of Cook Inlet: Proceedings
from a Symposium (eds Davis NY, Davis WE), pp. 5–23. Cook
Inlet Historical Society, Anchorage.
Richmond GM, Fullerton DS (1986) Summation of Quaternary
glaciations in the United States of America. Quaternary Science
Reviewss, 5, 183–196.
Ritland C, Pape T, Ritland K (2001) Genetic structure of yellow
cedar (Chamaecyparis nootkatensis). Canadian Jorunal of Botany, 79,
822–828.
Robinson M, Gouy M, Gautier C, Mouchiroud D (1998) Sensitivity
of the relative-rate test to taxonomic sampling. Molecular Biology
and Evolution, 15, 1091–1098.
Rollins RC (1983) Interspecific hybridization and taxon uniformity
in Arabis (Cruciferae). American Journal of Botany, 70, 625– 634.
Rollins RC (1993) The Cruciferae of Continental North America. Stan-
ford University Press, Stanford.
Roy BA (1995) The breeding system of six species of Arabis (Brassi-
caceae). American Journal of Botany, 82, 869–877.
Roy BA, Rieseberg L (1989) Evidence for apomixis in Arabis.
Journal of Heredity, 80, 506–508.
Schaal BA, Hayworth DA, Olsen KM, Rauscher JT, Smith WA
(1998) Phylogeographic studies in plants: problems and pros-
pects. Molecular Ecology, 7, 465–474.
370 C . H . D O B E S, T . M I T C H E L L - O L D S and M . A . K O C H
Sharbel TF, Mitchell-Olds Th (2001) Recurrent polyploid origins
and chloroplast phylogeography in the Arabis holboellii complex
(Brassicaceae). Heredity, 87, 59–68.
Sharbel TF, Voigt ML, Mitchell-Olds Th, Dobeß Ch, Jong H (2004) Are
the aneuploid chromosomes in the apomictic Boechera holboellii
complex B chromosomes? Cytogenetic Genome Research, in press.
Slatkin M (1985) Rare alleles as indicators of gene flow. Evolution,
39, 53–65.
Smith ChT, Nelson RJ, Wood ChC, Koop BF (2001) Glacial bio-
geography of North American coho salmon (Oncorhynchus
kisutch). Molecular Ecology, 10, 2775 –2785.
Soltis DE, Gitzendanner MA, Strenge DD, Soltis PS (1997) Chloro-
plast DNA intraspecific phylogeography of plants from the
Pacific Northwest of North America. Plant Systematics and Evo-
lution, 206, 353–373.
Stehlik I (2002) Glacial history of the alpine herb Rumex nivalis
(Polygonaceae): a comparison of common phylogeographic
methods with nested clade analysis. American Journal of Botany,
89, 2007–2016.
Swofford DL (2002) Phylogenetic Anaylsis Using Parsimony (and
Other Methods), Version 4. Sinauer Associates, Sunderland MA.
Taberlet P, Gielly L, Pautou G, Bouvet J (1991) Universal primers
for amplification of three non-coding regions of chloroplast
DNA. Plant Molecular Biology, 17, 1105 –1109.
Templeton AR (1998) Nested clade analyses of phylogeographic
data: testing hypotheses about gene flow and population his-
tory. Molecular Ecology, 7, 381–397.
Templeton AR, Sing ChF (1993) A cladistic analysis of phenotypic
associations with haplotypes inferred from restriction endo-
nuclease mapping. IV. Nested analyses with cladogram
uncertainty and recombination. Genetics, 134, 659 – 669.
Templeton AR, Boerwinkle E, Sing ChF (1987) A cladistic ana-
lysis of phenotypic associations with haplotypes inferred from
restriction endonuclease mapping. I. Basic theory and an
anaylsis of alcohol dehydrogenase activity in Drosophila. Genetics,
117, 343–351.
Templeton AR, Crandall KA, Sing ChF (1992) A cladistic analysis
of phenotypic associations with haplotypes inferred from
restriction endonuclease mapping and DNA sequence data.
III. Cladogram estimation. Genetics, 132, 619–633.
Templeton AR, Routman E, Philips CA (1995) Separating popula-
tion structure from population history: a cladistic analysis of the
geographical distribution of mitochondrial DNA haplotypes
in the tiger salamander Ambystoma tigrinum. Genetics, 140, 767–
782.
Tremblay NO, Schoen DJ (1999) Molecular phylogeography of
Dryas integrifolia: glacial refugia and postglacial recolonization.
Molecular Ecology, 8, 1187–1198.
Tribsch A, Schönswetter P, Stuessy TF (2002) Saponaria pumila
(Caryophyllaceae) and the ice age in the European alps. American
Journal of Botany, 89, 2024–2033.
Wahrhaftig C, Birman JH (1965) The Quaternary of the Pacific
mountain system in California. In: The Quaternary of the United
States (eds Wright HE Jr, Frey DG), pp. 299–340. Princeton Uni-
versity Press, Princeton, NJ.
Widmer A, Lexer Ch (2002) Glacial refugia: sanctuaries for allelic
richness, but not for gene diversity. Trends in Ecology and Evolu-
tion, 16, 267–269.
Wittzell H (1999) Chloroplast DNA variation and reticulate evolu-
tion in sexual and apomictic sections of dandelions. Molecular
Ecology, 8, 2023–2035.
The present investigation is part of a comprehensive research
project on the evolution of the North American Boechera group.
The central aim is to introduce this species complex as a model
group for comparative evolutionary studies making extensive
usage of molecular methods primarily developed for Arabidopsis
thaliana. Phylogeographical studies are performed by Christoph
Dobeß to obtain a PhD under the supervision of Marcus Koch.
Marcus Koch is mainly investigating evolutionary processes
and biological diversification among cruciferous plants from the
populational to the family level. Thomas Mitchell-Olds is
focusing on the interface between molecular biology, evolution
and ecology.
© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 349– 370