Migration between continents: geographical structure

and long-distance gene flow in Porpidia flavicunda

(lichen-forming Ascomycota)
Abstract
Historical and contemporary geographical distribution ranges with their associated gene flow patterns
interact to produce the genetic diversity observed today. Often it is not possible to separate out the
impacts of historical events, e.g. past fragmentation, and con-temporary gene flow, e.g. long-distance
dispersal. Porpidia flavicunda is a lichen-forming ascomycete occurring circumpolar in the boreal to
arctic zones for which vegetation history suggests that its distribution pattern has stayed broadly the
same over the past millennia. DNA-sequence diversity in P. flavicunda can, thus, be expected to
predominantly represent geographical population differentiation and its contemporary migration rates.
The population sample consists of 110 specimens collected in Northern Qubec, Baffin Island, Western
Greenland and Northern Scandinavia. DNA-sequence data sets of three nuclear gene fragments (LSU,
RPB2 and -tubulin) were analysed for genetic diversity within, and differentiation between, geographical
regions. Tests of population subdivision employing analyses of molecular variance and exact tests of
haplotype frequency distributions showed significant structure between the geographical regions.
However, the lack of fixed nucleotide polymorphisms and the wide sharing of identical haplotypes
between geographical regions suggest recurrent long-distance gene flow of propagules. Still, the means
by which propagules are dispersed remain to be discovered. Inference of migration rates shows that in
many cases a sufficiently high amount of migrants is exchanged between geographical regions to
prevent drastic population differentiation through genetic drift. The observed haplo-type distributions and
migration rates point to a gene flow model of isolation by distance.

Introduction
Geography is one of the major determinants of natural populations (e.g. Avise 1993; Hanski & Gilpin
1997). It influences gene flow, and thus population structure, either through a process of isolation by
distance or the existence of dispersal barriers. The geographical situation of a population also is closely
entwined with historical events, since distribution ranges can expand and shrink, and dispersal barriers
rise and fall due to climatic or geological changes. Thus, observed population structure is always a
composite of historical and contemporary geographical gene flow. One way to circumvent the problem
of separat-ing historical and contemporary factors is to investigate an organism, which geographical
situation broadly stayed the same over a long evolutionary time span and for that, historical and
contemporary gene flow patterns can thus be expected to be similar.
Lichen-forming ascomycetes depend for their survival on photoautotroph algae or cyanobacteria with
which they form intricate symbioses called lichens. This symbiotic strategy is realized not only by
major fractions of fungal and especially ascomycete diversity (Hawksworth et al.1995), but also is
found in all terrestrial ecosystems, which in the case of, for example, desert regions or arctic-alpine
habitats are dominated by lichens (Walter & Breckle 1991).
While lichen-forming ascomycetes represent important elements of terrestrial vegetation, molecular
investigations uncovering their population biology and genetics are only at the beginning. For example,
while distribution ranges in lichen-forming ascomycetes are often more extensive than those of higher
plants (Galloway 1996), information on the demography and history of the populations is generally
missing.
Lichen-forming fungi have the potential to build up strong spatial genetic structure within their
populations, due to their sedentary nature. The degree of population differentiation is determined by
the extent and range of gene flow mediated by spores and symbiotic propagules. Lichen-forming
ascomycetes do not form specialized long-distance propagules, apart from the microscopic size of their
diaspores (Bdel & Scheidegger 1996). This minute size might pose an advantage to gene flow by
propagules effectively entering the air column, or prove to be a dis-advantage since diaspores might
get lost, or damaged by UV radiation or extended desiccation in the air masses. However, it has been
shown that spores produced by lichen-forming ascomycetes can withstand extreme UV-radiation

conditions (de Vera et al. 2003).

First molecular investigations into the structure of lichen-forming ascomycete populations at the interregional and intercontinental scale provide at this point conflicting information on the extent of presentday gene flow. Observed genetic patterns have been interpreted as resulting from a process of
populations cohering due to more or less rare long-distance dispersal events (Hgberg et al. 2002), or
as the product of fragmentation subdividing a formerly continuous distribution range (Printzen & Ekman
2002; Printzen et al. 2003; Walser et al. 2005). An assessment of both explanations is difficult in the sofar analysed species, since they show presently disjunct distribution ranges that are proposed to have
been strongly influenced by past climate changes. Thus, in these species geographical structure and
historical events interact in the shaping of present-day genetic patterns. Analyses of species of lichenforming ascomycetes that were not strongly affected by population historical events will provide a way
to evaluate the question of long-range dispersal.
The lichen-forming ascomycete Porpidia flavicunda (Ach.) Gowan (Porpidiaceae, Lecanorales) is
associated with unicellular green algae of the genus Trebouxia with which it forms brightly orangecoloured crustose thalli on siliceous rock surfaces in the boreal to arctic zones (Gowan 1989). It has a
continuous circumpolar distribution range on the Northern Hemisphere (Hertel 1977; Inoue 1983;
Gowan 1989; Gowan & Ahti 1993; Fryday 2005). Within its distribution range, it is generally common
and widespread.
Previously, the species was described as a so-called species pair, consisting of two taxa that differ
primarily in their reproductive modes: Porpidia flavocoerulescens (Hornem.) Hertel & A.J. Schwab [now
P. flavicunda (Ach.) Gowan (Fryday 2005)], the exclusively sexual member of the pair, and Porpidia
melinodes (Krb.) Gowan & Ahti, the predominately vegetatively reproducing member of the species
pair. Several hypotheses were proposed to explain the evolutionary processes within species pairs. The
pairs were considered to represent two independently evolving lineages with contrasting reproductive
modes, a single outcrossing species with a polymorphic reproductive mode, a sexual mother lineage
giving rise repeatedly to short-lived vegetative spin-offs or a complex of cryptic species. Based on
phylogenetic analyses at the interspecific (Buschbom & Mueller 2004; Buschbom & Barker 2006) and
intraspecific (Buschbom & Mueller 2006) levels, it could be shown that the here-investigated species
pair is actually a single species with a variable reproductive mode. The genetic structure within P.
flavicunda is characterized by five divergent genetic lineages. Buschbom & Mueller (2006) proposed
that the observed patterns of genetic differenti-ation and of conflict within and between analysed loci
and lineages can be best explained by a model in which repro-ductive and nutritional requirements
need to be balanced, resulting in switches between sexual and vegetative reproduction depending on
symbiont interactions. Such a system is characterized by recurrent selective sweeps producing a
pattern of coexisting genetic lineages that are more or less deeply divided depending on recombination
rate and selection coefficients.
The five genetic lineages found in the phylogenetic analyses of the evolutionary relationships
between the specimens (Buschbom & Mueller 2006) show no strict correlation with geography. Lineage
I is the only lineage that was found exclusively in a single geographical region (Baffin Island, Table 1).
However, this geographical restric-tion might be deceptive, since the lineage is represented by only two
specimens in the sample. All other genetic lineages are present in at least three of the four analysed
geographical regions. Often, a single haplotype per locus is found in several of the geographical
regions. Conversely, several of the genetic lineages are commonly present in a single sampling site. For
example, in one sampling locality on Baffin Island (Tarr Creek), all five genetic lineages were collected
within a few hundred metres of each other. Similarly, thalli belonging to lineages II and III were found
bordering each other on a single rockface in the valley Krkevagge near Abisko (Sweden).
The data set analysed phylogenetically by Buschbom & Mueller (2006) is re-analysed in the present
study using population-genetic approaches. The data set was re-analysed, since tree-based
phylogenetic analyses at the intraspecific level assume that the reconstructed single topology reflects
important population processes. Phylo-genetic approaches, thus, provide fundamental insights into the
evolution at the population level. For a more detailed investigation of population processes, including

parameter estimates and model testing, coalescence-based population-genetic methods are necessary
that consider the actually realized genealogy within a population as unknowable and take into account
all possible relation-ships. Phylogenetic and population-genetic approaches, thus, complement each
other.
The goal of the present investigation is to provide insight into the extent of geographical
differentiation and gene flow in the lichen-forming ascomycete species P. flavicunda. In a first step,
mismatch analyses were conducted to investi-gate the importance of historical population size changes
within the species. In a subsequent step, the amount of genetic diversity observed was described and
different hypotheses of population subdivision were tested on a regional and intercontinental scale.
These hypotheses state that there is no population subdivision (i) between localities within each of the
four investigated geographical regions, and also not (ii) between the four regions, which are situated on
different continents. The extent and direction of gene flow between geographical regions was inferred
using a Markov chain Monte Carlo (MCMC)-based approach in a likelihood framework. The results are
discussed with regard to gene flow following a model of isolation by distance.
Materials and methods
Population and character sampling
A detailed list of collected specimens can be found in Buschbom & Mueller (2006). The only difference
between the data set analysed by Buschbom & Mueller (2006) and the one here is that specimen 113 is
excluded from the RPB2 alignment due to its shorter nucleotide sequence. Specimens of Porpidia
flavicunda were collected in four geographical regions (Fig. 1), including sites around Schefferville,
Northern Qubec (Canada), Iqaluit on Baffin Island, Nunavut (Canada) and Qeqertarsuaq on Disko
Island (Western Greenland). In Northern Scandinavia, localities near Abisko (Norrbotten, Sweden), Kevo
(Lappi, Finland) and along the coast of Finnmark (Norway) were sampled. Within each region (called
Qubec, Baffin Island, Greenland and Europe), collected thalli were grouped into three to five collection
localities that followed landscape features and represent geographical distance between specimens.
The data set consists of a total of 110 specimens, with 35 thalli originating from Qubec, 37 from Baffin
Island, 13 from Greenland and 25 from Europe.
Information on population genetic parameters can be gained through analyses of the neutral
nucleotide vari-ation in the genome. However, one has to take into account that population
demographic factors and gene-specific selective forces produce similar patterns of nucleotide
distributions (Nordborg 2001; Depaulis et al. 2003). In the present study, a multilocus approach was
employed to differentiate between these two evolutionary factors, since population demography will
influence the whole genome in the same way, that is, all analysed gene regions will provide the same
signal, while selection is gene-specific and thus will show different effects between genes.
Population-genetic analyses are based on DNA-sequences from three loci: a 0.6-kb fragment at the 5
end of the nuclear large subunit ribosomal RNA gene (LSU), and two nuclear protein-coding genes a
1-kb fragment at the 3 end of the gene coding for -tubulin, and a 1.2-kb fragment of the gene coding
for the second largest subunit of DNA-dependent RNA polymerase II (RPB2) spanning regions 5 7 as
described by Liu et al. (1999). GenBank accession numbers of the sequences are AY532945 AY532949,
AY532954 AY532958, and DQ314895 DQ314994 for the LSU, AY536804 AY536808, AY536813
AY536817, DQ315069DQ315153, and DQ315155 for -tubulin, and DQ314996DQ314997 and
DQ314999 DQ315068 for RPB2. Discussions of locus selection and details of laboratory procedures can
be found in Buschbom & Mueller (2004, 2006).
Genetic data analysis
The number of polymorphic sites and haplotypes, the haplotype diversity [Nei 1987; equation 8.4 (gene
diversity) in which 2n is replaced by (n)], as well as the nucleotide diversity (Nei 1987; equations 10.5
and 10.6) and the popu-lation mutation parameter per site (Nei 1987; equation 10.3) are used to
describe the genetic diversity within the analysed four geographical regions. The extent of population
divergence between geographical regions in pairwise comparisons is described by the number of fixed
and shared differences between localities and the number of mutations that are fixed within one
locality, but poly-morphic in the other.
Population size changes were investigated using mis-match distributions, that is, the distribution of

pairwise nucleotide site differences. Graphs were produced that show the observed mismatch
distribution and the expected distribution for a population of constant size (Rogers & Harpending 1992).
The R2 test statistic (Ramos-Onsins & Rozas 2002) was used to evaluate if for the investigated data set,
a model of constant population size can be rejected. To test the significance of the observed R2-value,
coalescent simulations were performed conditional on the number of polymorphic sites and assuming
no recombination (resulting in a conservative test, see Ramos-Onsins & Rozas 2002). Simulations are
based on 10 000 replicates.
Analyses of genetic diversity, population divergence and population size changes, as well as
coalescence simu-lations were conducted using DNASP 3.53 and 4.10 (Rozas et al. 2003). The genomic
phase was set to diploid, since the organism possesses a haploiddiploid life cycle, i.e. meiosis takes
place each generation and provides the opportunity of recombination, despite the fact that the thalli
are haploid. Sites with missing data were automatically excluded from the analyses.
Geographical structure within P. flavicunda was tested using analyses of molecular variance (AMOVA)
and exact tests of haplotype frequency distributions as implemented in ARLEQUIN 2.001 (Schneider et
al. 2000). AMOVA partitions the observed genetic variation into within- and between-group covariance
components (Excoffier et al. 1992). The significance of the proposed structure is then tested using a
nonparametric permutation procedure. While the AMOVA takes into account information on the genetic
distance between haplotypes, the employed exact tests of haplotype frequency distributions only
consider haplotypes as identical or different. Exact tests of haplotype frequency distributions test if the
observed distribution of haplotypes is as likely as a random distribution of haplotypes (Raymond &
Rousset 1995a; Goudet et al. 1996). First, it was inves-tigated if population structure among collection
localities within each of the four geographical regions can be identified using both approaches
[hypothesis (i)]. In a second step, population subdivision between geographical regions on an
intercontinental scale was investigated again using both approaches [hypothesis (ii)].
ARLEQUIN was also used to calculate pairwise FST values between geographical regions. The
significance of popu-lation differentiation in pairwise comparisons was tested based on FST values and
employing exact tests of haplotype distributions.
Migration rates between geographical regions were inferred in a likelihood framework employing an
MCMC-based approach to parameter estimation. Hereby, the like-lihood surface is explored through the
iterative updating of a model parameter and the rejection or acceptance of the new parameter value
based on the Metropolis-Hastings criterion. The procedure was conducted as implemented in the
program MIGRATE 2.03 (Beerli & Felsenstein 1999; Beerli & Felsenstein 2001). An unrestricted
migration model with variable values for geographical regions and potentially asymmetric migration
rates between all populations was applied. Initial and population migration rate M (migra-tion rate m
per generation/mutation rate per generation) estimates were obtained from FST calculations. A single
run consisted of 10 initial short chains, each of a length of 1 million generations from which 5 000
genealogies were recorded. These short chains were followed by two long
chains of a length of 10 million generations from which 50 000 genealogies were recorded. In both
short and long chains, 100 000 initial genealogies were discarded as burn-in before the sampling of
genealogies started. Within each run, three replicates were averaged and a heating scheme with four
temperatures (1.00, 1.20, 1.50, 3.00) employed. The numbers of immigrants per generation (4Nem)
were calculated by multiplying of the receiving population with the population migration rate M.
Finally, a model of isolation by distance was evaluated using a Mantel test (Mantel 1967). Here,
correlations between geographical distances and genetic distances are tested among the four regions.
Following Rousset (1997), genetic distances are expressed as FST/(1 FST). Geographical distances were
transformed using their natural logarithm as appropriate for sampling points that are located in a twodimensional system. The input matrix of FST values was produced by pairwise comparisons between
geographical regions as implemented in Arlequin. The web version of the program Genepop 3.4
(Raymond & Rousset 1995b) was used to perform the Mantel tests, the statistics package R 2.0.1 (R
Development Core Team 2004) was used to compute correlation coefficients.
Results
Genetic diversity
While the LSU fragment could be amplified and sequenced for all 110 specimens, sequences of only 96
and 72 specimens could be obtained for -tubulin and RPB2, respectively. Still, the five lineages
observed in phylo-genetic analyses of Porpidia flavicunda are represented in the data sets of all gene

regions. Alignments of all specimens contain neither gaps nor ambiguously aligned regions. Excluding a
few positions with missing data, the alignment is 580 nucleotides long for the LSU, 940 nucleotides long
for -tubulin and spans 1166 sites for RPB2. Across all analysed specimens, -tubulin has the highest
number of polymorphic sites with 69 variable positions and the highest nucleotide diversity with a value
of 0.01749 (Table 2). RPB2 shows an intermediate level of genetic diversity with 47 polymorphic sites
and a nucleotide diversity of 0.01243. The LSU is the least diverse with 24 polymorphic sites and a
nucleotide diversity of 0.00756. Despite the varying number of polymorphic sites and nucleotide
diversities per locus, the number of haplotypes found is rather similar in the loci (LSU: 15, -tubulin: 18,
RPB2: 13).
The amount of specimens sampled per geographical region varies threefold, so that estimates of
genetic diversity are not necessarily comparable between regions. However, there is no close
correlation between regional sample size and the analysed genetic diversity measures in the present
study, suggesting that the estimates represent essential characteristics of the regions that thus can be
keeping sample size in mind compared. Greenland, represented by the smallest number of
specimens, has low numbers of polymorphic sites and haplotypes; however, its haplotype and
nucleotide diversities and its estimates of are inter-mediate. On the other hand, Baffin Island, which is
rep-resented by the largest sample size, is also the genetically most diverse geographical region among
the four regions analysed across all loci. It possesses the highest number of polymorphic sites and
haplotypes, the highest haplotype and nucleotide diversities and the largest values for . Taking the
comparably high sample size into account, Qubec seems to be the least diverse region. Here, the
lowest nucle-otide diversities are found, values for S and are generally low, and comparably low
numbers of haplotypes are present. The sample from Europe resulted in intermediate diversity
measures across loci.
Mismatch analyses conducted for each of the three genes across the whole sample of specimens are
ragged and generally multipeaked, the typical distribution for a population of constant size (Fig. 2). A
model of constant population size was not rejected using the R 2 statistic (LSU: R2 = 0.090, P = 0.53;
RPB2: R2 = 0.153, P = 0.97; -tubulin: R2 = 0.118, P = 0.84).
Pairwise population differentiation
The number of fixed polymorphisms between populations is a measure for the extent of isolation
between popu-lations. No fixed differences are observed in any pairwise comparison between the
geographical regions for all three loci (Table 3). Instead, a large amount of nucleotide polymorphisms is
shared between geographical regions (average fraction of shared substitutions to total variable
substitutions LSU: 37%, RPB2: 53%, and -tubulin: 45%). Similarly high numbers of sites that are
polymorphic within one of the regions, but not the other, suggest that a non-negligible amount of the
genetic variation is situated within the geographical regions and not between regions. A com-parison of
the average numbers of nucleotide substitutions per site D xy and the net numbers of nucleotide
substitutions per site Da supports this conclusion (data not shown).
Pairwise analyses of population differentiation between the four geographical regions based on FST
values (Table 4) show that the populations in Greenland and in Europe are not differentiated, that is, FST
values are not different from zero. Tests based both on the significance of FST values and on exact tests
of haplotype distributions return nonsignifi-cant results for these two populations. The highest amount
of population differentiation is observed between the populations in Qubec and Greenland, and in
Qubec and Europe, for all three loci. Intermediate levels of differenti-ation are found in analyses
involving Baffin Island. For these comparisons, pairwise tests of population differentiation provided
mixed results, depending on the test statistic used and the locus analysed.
Overall geographical structure
The distribution of haplotypes within each of the four geographical regions [hypothesis (i)] is not
significantly different from random according to exact tests of haplo-type frequency distributions (Table
5). Furthermore, no population structure was generally found based on the results of the AMOVAs, the
only exceptions are the results for the LSU and RPB2 in Qubec. However, an AMOVA conducted with
the -tubulin data set and the results of the exact tests for all three loci do not support population
structure within that geographical region. The amount of the total variation already explained by the
within-population component is (with the exception of the LSU and RPB2 results for Qubec) above 87%
for all loci and geographical regions. Values of above 100% are artefacts of the calcula-tions and not
significantly different from 100%. Similarly, the corresponding negative FST values have no biological
explanation and are not different from zero. FST values per locus within the four geographical regions
are in the range between 0 and 0.12325, again with the exception of the two loci in Qubec, which had

FST values of 0.46525 (LSU) and 0.34431 (RPB2).

An AMOVA and exact test of haplotype frequency dis-tribution returned significant P values for
geographical population structure between the four analysed regions for each of the three loci
[hypothesis (ii)]. Between 61% and 85% of the total genetic variation can be explained by variation
within the geographical regions. FST values range from 0.14864 to 0.38558 for the loci.
Patterns of gene flow
High numbers of immigrants into populations originate from adjacent geographical regions (Fig. 1). The
highest rates were observed for migration from Qubec [28.07 immigrants per generation (Imm)] and
Greenland (22.45 Imm) to Baffin Island, respectively, and from Greenland to Europe (8.30 Imm).
However, the extent of migration is not necessarily symmetrical between geographical regions. In
contrast to the high number of immigrants from Qubec and Greenland to Baffin Island, respectively,
the backward rates of migrants originating from Baffin Island are below one migrant per generation. In
approximately half of all cases the number of immigrants per generation is above one.
A positive correlation between geographical and genetic distance is observed for the three loci (Fig.
3). This suggests that gene flow between geographical regions follows an isolation-by-distance
relationship. However, while cor-relation coefficients for the three gene regions are generally high (R for
LSU: 0.43, RPB2: 0.33, -tubulin: 0.40), the null hypothesis of no correlation between both distance
meas-ures cannot be rejected (P value for LSU: 0.46, RPB2: 0.38, -tubulin: 0.46).
Discussion
Since the Miocene of the Tertiary, well-defined climatic belts exist on the Northern Hemisphere that
move north to south depending on overall climatic conditions (Hohl 1981). Resulting from the ongoing
cooling of the Earth, the circumarctic belt developed and was in place 3 million years ago (Matthews
1979). While the arctic biome with the exception of proposed high-latitude refugia (Abbott &
Brochmann 2003), moved north to south during the increasingly severe glaciation cycles and required
arctic organisms to migrate at proposed high speeds (Hewitt 2001), it presented a more or less
continuous, circumpolar habitat. For example, in Europe large parts of the continent were covered by
tundra vegetation during past ice ages, representing rather a distribution range expansion than a
restriction into southern European refugia as proposed for many temperate species. Accordingly, a
reduction in genetic diversity in Europe due to past bottlenecks as proposed for the tree-inhabiting
temperate to boreal species Letharia vulpina (Hgberg et al. 2002) was not observed for Porpidia
flavicunda.
A sign of the latitudinal movement of the distribution range in the genetic diversity of P. flavicunda
was not observed. One might expect a correlation between time since retraction of the ice at the end of
the last ice age [deglaciation dates were set to be 6000 BP for Qubec, 8000 BP for southern Baffin
Island, 9500 BP for Disko Island in Western Greenland and 7500 BP for northernmost Scandinavia
(Barnekow 1999; Dyke 2004; Dyke et al. 2004)] and genetic diversity, due to limited migration rates
and founder effects. However, such a relationship was not observed. Only the low genetic diversity
found in Qubec corresponded with the long ice cover of that area. Assum-ing comparable conditions
for all regions, e.g. that source populations recolonizing the regions are similar or the same, it seems
that the migration rate of P. flavicunda is large enough to allow the species to follow the north-to-south
movement of its distribution range without difficulties and attain locality-specific genetic diversities
since ice retraction. While some arctic organisms show reductions in genetic diversity in previously
glaciated areas due to founder effects (Hewitt 2004), the picture emerging for P. flavicunda is similar to
those for the arctic saxifrage,
Saxifraga oppositifolia (Gabrielsen et al. 1997), Siberian lemming, Lemmus sibiricus (Fedorov et al.
1999), collard lemming, Dicrostonyx groenlandicus (Ehrich et al. 2000), arctic fox, Alopex lagopus
(Daln et al. 2005) and bog bilberry,
Vaccinium uliginosum (Alsos et al. 2005). In Saxifraga, Alopex and Vaccinium, the lack of a reduction in
genetic diversity in nonglaciated vs. glaciated areas was brought into con-nection with the efficient
long-distance dispersal capabilities of these organisms and a broad-fronted recolonization from large
and diverse periglacial populations.

In arctic organisms, the long-lasting glacial periods are proposed to have resulted in population
expansions, while the short-term interglacials might be periods of population bottlenecks. A population
bottleneck with subsequent expansion dating to the last interglacial was found for the arctic fox (Daln
et al. 2005). Population reductions might produce isolated populations and subsequent phylogeographical structure. The lack of such phylogeographical structure in the arctic fox was explained by
current extensive gene flow. Mismatch analyses in P. flavicunda did not reject a hypothesis of constant
population size. However, the result of the mismatch analyses in P. flavicunda are strongly influenced
by the genetic structure, i.e. the genetic lineages, found within the species that are proposed to be the
result of symbiont interactions in the lichen symbiosis. It is pos-sible that P. flavicunda experienced
population bottlenecks resulting in isolated lineages in previous interglacials. If so, the last ice age
seems to have provided the opportunity for population homogenization through migration and
recombination, resulting in e.g. the wide distribution of haplotypes throughout the investigated area.
Overall, a picture emerges of a species that is character-ized by a continuous circumarctic distribution
that reaches back at least to the beginning of the last ice age. This popu-lation history distinguishes P.
flavicunda from previously investigated Mediterranean and temperate species of lichen-forming
ascomycetes, whose distributions were strongly affected by past climate change. Furthermore, there is
no indication that the species has difficulties to follow the latitudinal displacement of its distribution
range. Thus, the observed population structure within the species can be expected to be less
influenced by historical events but primarily reflect contemporary geographical structure and
population biological factors.
The focus of the present study is on the relationships between geographical regions on an
intercontinental scale. The presence of local population structure within regions was investigated to
exclude the possibility of local struc-ture interfering with the results at the larger scale. The lack of
population subdivision within geographical regions [hypothesis (i) cannot be rejected, see Table 5] has
to be taken with caution, since sample sizes for individual localities generally were small. A detailed
study will need to be conducted to fully answer the question of the extent of population differentiation
on a local to intracontinental scale in P. flavicunda.
Between continents and geographical regions, significant geographical population structure exists in P.
flavicunda as was shown for all gene partitions [rejection of hypothesis (ii), see Table 5]. However, no
population differentiation due to geography was found if the observed genetic lineages were
considered individually (data not shown). AMOVAs and exact tests of haplotype frequency distributions
performed for lineages III and IV, which showed sufficient variation for the analysis, returned (with the
exception of the AMOVA based on the LSU for lineage IV) nonsignificant results.
Despite the significant result for the whole data set, an interpretation of the extent of population
differentiation that is reflected in the calculated FST values, however, will be context-specific. While
Avise (1993) maintains that an FST value of 0.2 is marginally sufficient in theory to prevent dramatic
genetic differentiation (p. 208), Hartl (1988) considers the same value to indicate great genetic
differ-entiation (p. 90). The range of FST values observed here coincides with the average FST estimate
of 0.2 for fungi reported by Morjan & Rieseberg (2004) in their review of gene flow studies. They
interpret these estimates to represent moderate to low levels of gene flow.
In P. flavicunda, patterns of substitutions and haplotype distributions suggest that a non-negligible
amount of gene flow occurs between geographical regions. In the species, no fixed differences are
observed between geographical regions, while a large fraction of nucleotide polymorph-isms are shared
between continents and none of the major haplotype lineages represented by more than a couple of
specimens was found exclusively within any of the sampled regions. In addition, identical haplotypes
were found in up to all four regions. This is either the picture of very recently diverged populations or of
continuing gene flow between the populations (Hey 1991; Wakeley & Hey 1997). Since it is argued in
this study that populations of P. flavicunda on the investigated continents possess com-parably old and
continuous histories, this leaves population cohesion due to recurrent gene flow.
The occurrence of long-distance gene flow is also sup-ported by the reconstruction of migration rates
between the investigated geographical regions. The employed maximum-likelihood reconstructions that
are based on DNA-sequence substitution patterns have the advantage that they use the most
information present in the data. They, thus, allow inference in parameter-rich models (e.g.
asymmetrical migration rates, population-specific ). However, doubts arise with regard to the
convergence of the MCMC procedure within commonly used lengths of the chains and, in addition,
parameter estimates are often associated with large confidence intervals. Simulation studies

investigating the performance of the program MIGRATE employed here showed that is estimated
reasonably well, while migration rates generally are underestimated with occasional overestimates and
associated with large confid-ence intervals (Abdo et al. 2004; Beerli 2006). Confidence in the overall
pattern of the numbers of migrants recon-structed here between populations results from a second run
inferring the model parameters with MIGRATE conducted with the same settings (data not shown). The
two runs pro-duced comparable results with regard to relative migration rates between populations and
magnitudes of rates.
In over half of all cases, more than one migrant per generation is inferred between populations (Fig.
1). According to Wright (1931), this level of exchange between populations is sufficient to prevent
population differenti-ation due to genetic drift. In three cases, the number of immigrants is larger than
four, suggesting homogenization of neutral alleles between the involved populations (Hartl & Clark
1997). However, in these cases reconstructed migration rates were drastically asymmetrical. Such
asym-metric migration rates seem to produce intermediate levels of genetic differentiation between
populations. In these cases, the results of pairwise tests of population differ-entiation depended on the
test statistic used and the locus analysed. Only in the case of migration between Greenland and
Europe, for which forward and backward migration rates were found to be larger than one, no
population differentiation was observed consistently.
The extent of the migration rates on a yearly basis is dif-ficult to assess in the present species. For a
transformation of the number of immigrants per generation to those exchanged per year, the
generation time of P. flavicunda needs to be known. The species can be expected to form a very slow
growing, long-lived symbiotic system that has the potential to possess long generation cycles. The
length of the generation time will depend on the growth rate of the organism as a measure of the onset
of reproductive maturity and especially the rate of survival of propagules into the next generation.
Patterns of haplotype distributions and reconstructed migration rates suggest that gene flow follows a
model of isolation by distance. While migration rates are highest between adjacent geographical
regions, the distribution of the five genetic lineages produces an east-to-west gradient. For example,
considering the LSU, specimens in Qubec frequently possess haplotypes belonging to lineages IV and
V, rarely show haplotypes belonging to lineage III, and do not have haplotypes from lineages II and I. On
Baffin Island, lineages II to V are present with intermediate frequencies. However, specimens from
Greenland are only found in lineages II and III, and collections from Europe are commonly found in
lineages II and III and rare in line-ages IV and V. Haplotype distributions, migration rate patterns and FSTbased Mantel tests provide independent evidence for a process of isolation by distance. Taking into
account the results of the first two data systems, the Mantel procedure seems to provide a conservative
test. More extensive sampling of localities at varying distances and representing the whole distribution
range of the spe-cies will show if a model of isolation by distance will be supported in the future to
describe the process of differentiation between the geographical regions.
The data for P. flavicunda suggest that long-distance dispersal instead of past fragmentation underlies
the differentiation between the analysed regions. Not only does vegetation history suggest that the
species possessed a more or less continuous distribution range over an evolutionary long time span,
but the species also lacks fixed polymorphisms within its geographical regions and shares identical
haplotypes between continents. Long-distance dispersal was already proposed by Hertel (1987) for
another, closely related, epilithic crustose genus of lichen-forming ascomycetes. Several species within
Lecidea show bipolar distribution ranges, and Hertel (1987) sug-gested that these distribution patterns
can be explained by long-distance dispersal of the microscopically small diaspores (ascospores, or
vegetative propagules that consist of small packets of algal cells wrapped together by fungal hyphae
called soredia) via global movement of high-altitude air masses or migratory birds.
Porpidia flavicunda is not a typical candidate for dispersal by migratory birds, since it is not a member
of bird perch communities. Dispersal by jet streams seems to be a more probable option. This requires
the propagules to be effec-tively transferred into the air column. This is especially imaginable for the
ascospores, which are actively ejected from the asci, while the vegetative propagules (soredia) would
require air turbulence for the transition out of the boundary layer of the rock surface into the air
column. All four geographical regions sampled lie in the west-wind drift that moves air masses
circumpolar. However, no correlation between reconstructed migration rates and prevailing wind
direction is observed. An ubiquitous presence of the spores in the air would explain such a pattern. This
would require the long-term survival of the propagules in the air masses. At first sight, no specific

adaptations to prolonged desiccation and intense UV-radiation (as for example dark colouration or thick
cell walls) are obvious in P. flavicunda. However, the answers to those requirements might be found in
the inherent desiccation tolerance of the lichen symbionts and the abundance and diversity of UVabsorbing biochemical secondary com-pounds present in P. flavicunda.