Content uploaded by Ladislav Paule
Author content
All content in this area was uploaded by Ladislav Paule on Jun 29, 2018
Content may be subject to copyright.
1 of 11Published by Po lish Botanical Socie ty
Acta Societatis Botanicorum Poloniae
ORIGINAL RESEARCH PAPER
Phylogeny of beech in western Eurasia
as inferred by approximate Bayesian
computation
Dušan Gömöry*, Ladislav Paule, Vladimír Mačejovský
Faculty of Forestry, Technical University in Zvolen, TG Masaryka 24, 960 53 Zvolen, Slovakia
* Corresponding author. Email: gomory@tuzvo.sk
Abstract
e Fagus sylvatica L. species complex in Europe and Western Asia comprises two
commonly recognized subspecies, F. sylvatica subsp. sylvatica [= F. sylvatica sensu
stricto (s. str.)] and F. sylvatica subsp. orientalis (= F. orientalis), and two putatively
hybridogenous or intermediate taxa, “F. moesiaca” and “F. ta ur ic a”. e present study
aimed to examine the demographic history of this species complex using 12 allelic
loci of nine allozymes scored in 279 beech populations in western Eurasia. ree
sets of phylogenetic scenarios were tested by approximate Bayesian computation:
one dealing with the divergence of subspecies and/or regional populations within
the whole taxonomical complex, and two others focusing on the potential hybrid
origin of “F. moesiaca” and “F. ta urica”. e best-supported scenario within the rst
set placed the time of divergence of regional populations of F. orientalis in the Early
Pleistocene (1.18–1.87 My BP). According to this scenario, the Iranian population
was the ancestral lineage, whereas F. sylvatica s. str. was the lineage that diverged
most recently. “Fagus taurica” was found to have originated from hybridization
between the Caucasian population of F. orientalis and F. sylvatica s. str. at 144 ky
BP. In contrast, there was no evidence of a hybrid origin of “F. moesiaca”. e best-
supported scenario suggested that the Balkan lineage is a part of F. sylvatica s. str.,
which diverged early from F. orientalis in Asia Minor (817 ky BP), while both the
Italian and Central-European lineages diverged from the Balkan one later, at the
beginning of the last (Weichselian) glacial period.
Keywords
Fagus sylvatica L.; Fagus orientalis Lipsky; phylogenetic scenario; allozymes
Introduction
Beech trees represent the most important component of montane forests in the temper-
ate zones of Europe and Western Asia. Beech forests with the admixture of local spruce
and r species are the most productive and stable native forest communities in the
mountains of Central Europe, as well as the Near and Middle East. As such, various taxa
of the genus Fagus L. have long been the focus of the interest of botanists. However, the
number and rank of Fagus taxa in western Eurasia have been controversial, and their
phylogeny is still not resolved. e state-of-the-art view, accepted by Flora Europaea [1]
and recent systematic revisions [2,3], considers all beech populations in western Eurasia
as belonging to a single species, Fagus sylvatica L., with two subspecies: F. s. subsp. in
most of the species’ European distribution [denoted hereaer as F. sylvatica sensu stricto
(s. str.) for simplicity] and F. s. subsp. orientalis (denoted hereaer as F. orientalis), whose
range begins in the southeastern Balkan Peninsula (Black Sea coast of Bulgaria and
northeastern Greece) and continues through Turkey and the Caucasus to the Alborz
Mountains in Iran. However, the latter taxon was described by Lipsky [4] as a separate
DOI: 10.5586/asbp.3582
Publication history
Received: 2018-03-02
Accepted: 2018-05-08
Published: 2018-06-29
Handling editor
Jan Holeksa, Faculty of Biology,
Adam Mickiewicz University in
Poznań, Poland
Authors’ contributions
LP and DG designed the
study, collected materials, and
performed analyses; DG and VM
performed ABC simulations; DG
wrote the rst draft; all authors
commented and revised the
manuscript
Funding
This study was supported by
research grant 1/0269/16 of
the Slovak Grant Agency for
Science.
Competing interests
No competing interests have
been declared.
Copyright notice
© The Author(s) 2018. This is an
Open Access article distributed
under the terms of the
Creative Commons Attribution
License, which permits
redistribution, commercial and
noncommercial, provided that
the article is properly cited.
Citation
Gömöry D, Paule L, Mačejovský
V. Phylogeny of beech in
western Eurasia as inferred
by approximate Bayesian
computation. Acta Soc Bot Pol.
2018;87(2):3582. https://doi.
org/10.5586/asbp.3582
Digital signature
This PDF has been certied using digital
signature with a trusted timestamp to
assure its origin and integrity. A verication
trust dialog appears on the PDF document
when it is opened in a compatible PDF
reader. Certicate properties provide
further details such as certication time
and a signing reason in case any alterations
made to the nal content. If the certicate
is missing or invalid it is recommended to
verify the article on the journal website.
Piotr Otręba
Digitally signed by Piotr Otręba
Date: 2018.06.29 16:42:51 +01'00'
2 of 11© The Author(s) 2018 Published by Po lish Botanical Socie ty Acta Soc Bot Pol 87(2):3582
Gömör y et al. / ABC-inferred phylo geny of Fagus sylvatica s. l .
species, F. orientalis Lipsky, and this opinion is generally shared by local botanists [5,6]
and forestry practitioners [7]. Palibin [8] distinguished beech in the Caucasus and the
Alborz Mountains as representing a separate species, F. hohenackeriana, but this view
has not been generally accepted, and Greuter and Burdet [9] included it within their
F. sylvatica subsp. orientalis. Two other taxa are mostly considered to be intermediates
between the above subspecies: “F. tau r i c a ” Popl. distributed in the mountains of the
Crimean peninsula, and “F. moesiaca” (Domin, Maly) Czeczott of the southern Balkans.
Both taxa were originally described as separate species ([10] and [11], respectively),
but again views about their taxonomical statuses are quite controversial, and they are
sometimes considered to represent lineages that originated from hybridization between
the subspecies [4,8,12,13]. e subspecies also dier in the patterns of their geographical
distributions: in Europe, the range of beech is almost continuous, except in peripheral
regions like the Cantabrian range, Apennine Peninsula, or southern Balkans, whereas
the range of oriental beech is much more fragmented into several large populations
that are relatively isolated from each other (northern Turkey, the Amanus Mountains,
the Caucasus, and the Alborz) [14,15].
Knowledge of the phylogeny of F. sylvatica is also limited. Complex studies based
on internal transcribed spacer (ITS) ribosomal DNA sequences and morphology of
this taxon plus related Asian and North American Fagus species were carried out by
Denk et al. [16,17]. orough insights into beech phylogeny based on the fossil record,
biogeography, and geological history were provided in a study by Denk and Grimm
[18]. Fagus sylvatica was represented in all three of these studies, but they did not focus
specically on this species. Generally, phylogenetic studies within the genus Fagus have
included a broad variety of species covering the whole range of the genus, but they
typically lacked the detail necessary to reveal relationships among subspecic taxa or
regional populations within F. sylvatica s. l. [19–21].
Marker-based studies focusing on beech in western Eurasia, including those based
on the same dataset as the current study [22,23] (and see below), usually inferred
the evolutionary past of beech populations from the results of clustering methods,
which were either distance-based or based on simulations (mostly Bayesian Monte
Carlo–Markov chain approaches, such as those of Pritchard et al. [24] or Corander et
al. [25]). Recently, more direct and more powerful tools based on coalescent theory
have been developed, such as those using approximate Bayesian computation (ABC),
which allow simultaneous estimation of the times of divergence of genetic lineages and
their eective population sizes [26,27]. In this study, we applied the ABC approach
to verify the conclusions about the phylogenetic relationships among beech taxa and
regional populations from earlier studies [22,23], and to make inferences about the
past demography of beech in western Eurasia.
Material and methods
is study examined 279 beech populations covering the majority of the geographic
range of the whole taxonomical complex (hereaer denoted as F. sylvatica s. l.), each
represented by a minimum of 50 adult trees. Twelve polymorphic gene loci of nine
isozyme systems were analyzed: glutamate oxaloacetate transaminase (Got-B), isocitrate
dehydrogenase (Idh-A), leucine aminopeptidase (Lap-A), malate dehydrogenase (Mdh-A,
Mdh-B, Mdh-C), menadione reductase (Mnr-A), phosphoglucomutase (Pgm-A), phos-
phoglucose isomerase (Pgi-B), peroxidase (Per-A, Per-B), and shikimate dehydrogenase
(Skdh-A). Details of enzyme extraction, electrophoresis, and staining procedures are
given in Gömöry et al. [22].
Interpretation of zymograms and genotyping followed Merzeau et al. [28] and Müller-
Starck and Starke [29]. Alleles at each locus were identied by the relative migration
rates of their respective protein fractions, expressed as a percent of the migration rate
of the most frequent allele of that gene in the Central- and Western-European beech
populations.
Speciation scenarios were tested by applying ABC analyses. However, simulations
based on the complete dataset consisting of 15,194 genotypes would last excessively
long (estimated duration in the range of several weeks per simulation), so the dataset
3 of 11© The Author(s) 2018 Published by Po lish Botanical Socie ty Acta Soc Bot Pol 87(2):3582
Gömör y et al. / ABC-inferred phylo geny of Fagus sylvatica s. l .
had to be trimmed in such a way that a certain number of randomly chosen trees per
population were retained. In the dataset of F. sylvatica s. str., trimming was the strongest,
and only two trees per population were retained except for populations that originated
from minor glacial refugia and were underrepresented in the collection, specically
the Apennine Peninsula (all samples retained) and “F. moesiaca” (southern Balkans;
14 samples per population). For F. orientalis, 24 samples per population were retained
from the Alborz and Asia Minor regions, 28 from the Caucasus, and 27 of “F. taurica”
(Crimea). In this way, the total sample size was reduced to 2,659 trees (resulting in a
reduction of computing time to few days per simulation), while all studied taxa and/
or regional populations were represented by comparable sample sizes.
e computations were performed in DIYABC v. 2.05 [30]. In all runs, 100,000
datasets were simulated for each scenario. As we have no information on the distribu-
tion of mutation rates of allozyme loci in Fagus, these were set to vary between 10−7
and 10−5, which is analogous to observed values in other organisms [31,32]. e 1%
of the simulated datasets that were the most similar to the observed data were used to
estimate relative posterior probabilities. Simulations applied the generalized stepwise
mutation (GSM) model. No prior assumptions were made of divergence times. As
the ABC algorithms use the number of generations to estimate the divergence time,
a generation turnover time of 100 years was chosen for the conversion of generations
into a time scale of years. Isolated beech trees may start owering at 15–20 years of
age, and trees in a canopy at approximately 50 years [14,33]; however, the average age
of ospring-producing trees is certainly higher in a natural forest ecosystem. At the
local scale, abundant regeneration repeatedly occurs every 100–120 years, when the
rst gaps occur in the closed canopy aer a stage of one-layer stand associated with
the culmination of the standing stock [34]. e lower limit of this interval was taken as
the estimated generation time, as the earliest survivors are likely to develop the largest
dimensions and thus contribute the most through both male and female gametes to
the gene pool of the ospring.
Model checking was done by ranking the summary
statistics of the observed data (mean number of alleles
and mean heterozygosity across loci, FST) against the
posterior predictive distributions, as well as by principal
component analysis (PCA), which placed the observed
data onto the cloud of datasets simulated with the prior
distributions of parameters and datasets from the posterior
predictive distribution.
The scenarios tested are shown in Fig. 1, and they
are displayed geographically in Fig. S1. e rst set of
scenarios focused on the phylogeny within F. sylvatica s. l.,
namely on the divergence of regional populations within
F. orientalis (Asia Minor, Caucasus, and Alborz) and the
potential paraphyly of F. sylvatica s. str. In the second
round of simulations, we tested whether Crimean beech
(“F. taurica”) is a lineage within F. orientalis or a taxon that
originated from hybridization between F. sylvatica s. str.
and F. orientalis. Finally, the last set of scenarios addressed
the potential hybrid origin of Balkan beech (“F. moesiaca”)
and the phylogenetic position of beech populations in
the Apennine Peninsula. Only scenarios with realistic
biogeographic settings were considered.
Results
Allelic variation within F. sylvatica s. l.
Allele frequency distributions in both subspecies of F.
sylvatica are shown in Tab. 1. Even though dierences be-
tween subspecies were sometimes substantial, most alleles
Fig. 1 Evolutionary scenarios tested in the ABC simulations.
Abbreviations of regional populations: Alb – Alborz; Ami – Asia
Minor; Ape – Apennine Peninsula; Balk – southern Balkans
(“F. moesiaca”); Cau – Caucasus; CEur – Central Europe; Cri –
Crimea (“F. tau r i c a ”); Eur – Europe (F. sylvatica s. str.). Lines:
shades of red – F. sylvatica s. str.; shades of blue – F. orientalis;
shades of violet – potential hybrid taxa.
4 of 11© The Author(s) 2018 Published by Po lish Botanical Socie ty Acta Soc Bot Pol 87(2):3582
Gömör y et al. / ABC-inferred phylo geny of Fagus sylvatica s. l .
were shared between them. In the cases when an allele was
absent in one subspecies, it was found at very low frequency
in the other one (always <0.2%). In multiallelic loci, relative
migration rates of the respective protein fractions mostly
changed by regular steps. Moreover, the more the migra-
tion rate diered from that of the most common protein
fraction, the lower the respective allele frequency was, with
a few exceptions (Mdh-A, Mdh-B, and partly Skdh-A).
Phylogeny simulations
ABC simulations were done under three sets of scenarios. e
performance of simulations was quite good, as the observed
data were placed fairly well in the center of the point clouds
produced from simulated datasets in the case of all three
selected scenarios on PCA plots (Fig. S2), and (with two
exceptions) summary statistics were always within the interval
of 5% to 95% of simulated values (Appendix S1).
e parameter estimates for the most probable simulated
scenarios are presented in Tab. 2. In most simulations, the
posterior estimates of the mutation rate were around 6 ×
10−6.
e rst set of simulations comprised four scenarios,
successively testing F. sylvatica s. str. and each of the major
regional populations of F. orientalis as the ancestral lineage
from which the others diverged. Scenario 3, in which the
Alborz population of beech was treated as the ancestral one,
received the best support (posterior probability of 0.63).
However, the probability of Scenario 1, supporting the current
taxonomical classication of the two subspecies, was also
relatively high (0.22), which means that the odds for Scenario
3 compared to this one were approximately 3:1. Divergence
times for all major regional populations under Scenario 3
ranged between 11,800 and 18,700 generations, placing the
divergence events in the Early or Middle Pleistocene. Eective
population size estimates for the Caucasus and Asia Minor
populations were approximately threefold of those from the
Alborz and for F. sylvatica s. str.
Simulations in the second round of scenarios rendered
Crimean beech a product of an ancient hybridization between
F. sylvatica s. str. and a regional population of F. orientalis.
e geographically more plausible Scenario 4, wherein the
second parental population was that in the Caucasus, re-
ceived higher support (posterior probability of 0.69), but
again the probability of Scenario 3 (hybridization with the
beech population in Asia Minor) was also relatively high
(0.25). e timing of hybridization in Scenario 4 was quite
late, 1,440 generations, meaning that the event would have
occurred at the beginning of the Holocene, but since the
upper limit of the 95% condence interval was high, 6,400
generations, there was much uncertainty in this estimate. e
genetic contribution of F. sylvatica s. str. to the gene pools of
Crimean beech in Scenario 4 was 30%.
In the last set of simulations, focusing on the origin of
Balkan beech, the odds in favor of the best scenario of suc-
cessive splits were from 16:1 to 74:1 compared to scenarios
involving hybridization, so only this scenario was taken into
consideration. A hybrid origin of Balkan beech was thus
ultimately not supported, and it seemed to rather represent
Tab. 1 Allelic frequencies in subspecies of Fagus sylvatica s.
l. across all regional populations. Alleles are labeled by their
relative migration rates.
Locus Allele F. sylvatica s. str. F. orientalis
Per-A 105 0.35724 0.32377
100 0.64276 0.67623
Per-B 52 0.00072 0.00103
39 0.17966 0.49362
26 0.74256 0.43821
13 0.07707 0.06598
10 - 0.00115
Lap-A 106 0.01899 0.01363
100 0.51255 0.63448
97 0.30979 0.21413
94 0.15866 0.13756
91 - 0.00019
Got-B 54 0.00009 0.00121
36 0.33814 0.05692
18 0.66166 0.93887
6 0.00009 0.00300
Mnr-A 152 - 0.00153
139 - 0.00035
126 0.02101 0.23546
100 0.91569 0.73794
74 0.00469 0.01704
63 0.05857 0.00766
52 0.00004 -
Idh-B 132 - 0.00018
116 0.28021 0.26672
100 0.71458 0.73258
84 0.00517 0.00053
Mdh-A 135 0.00005 -
130 0.00449 0.02238
125 0.99496 0.81711
113 0.00014 0.00693
101 0.00037 0.15358
Mdh-B 118 0.05825 0.03575
109 0.00250 0.00105
100 0.93677 0.92932
78 0.00195 0.03013
56 0.00054 0.00375
Mdh-C 22 0.22599 0.01888
18 0.77388 0.98112
14 0.00014 -
Pgi-B 126 - 0.00034
113 0.00792 0.07257
100 0.97899 0.89924
87 0.01300 0.02136
74 0.00008 0.00649
5 of 11© The Author(s) 2018 Published by Po lish Botanical Socie ty Acta Soc Bot Pol 87(2):3582
Gömör y et al. / ABC-inferred phylo geny of Fagus sylvatica s. l .
a lineage that diverged from F. orientalis in Asia Minor ap-
proximately 8,200 generations ago.
Discussion
Suitability of the stepwise mutation model for allozymes
e choice of an appropriate mutation model is an essential
issue in model-based phylogeny reconstruction. For protein
markers such as allozymes, the innite allele model of Kimura
and Crow [35] is mostly considered suitable. Under this
model, each mutation is expected to produce a new allele,
while all alleles are considered equally dierent from each
other. However, we used a dierent model from this, which assumed stepwise mutations,
i.e., alleles are ranked by some property of themselves or their products and a mutation
can only change a particular allelic state into either of the two possible adjacent states.
Stepwise models of mutations are generally preferred in studies employing repetitive
sequences, such as minisatellites or microsatellites, among which mutations mostly
consist in the addition or deletion of a short sequence or motif and alleles thus dier
in their total sequence length (i.e., number of repetitions).
However, the stepwise mutation model was originally developed for electrophoreti-
cally detectable variants diering in electric charge [36,37]. Brown et al. [38] found
good support for the occurrence of stepwise manner of mutations in the available
experimental allozyme data. In our case, the relative migration rates of protein frac-
tions produced by dierent isozyme alleles diered at quite regular intervals. Even
though the migration rate is not necessarily a linear function of charge and depends
also on the shape and size of a protein molecule, such regularity is a strong indication
of stepwise changes in the net charge [39]. Moreover, the frequency distributions of
protein fractions are unimodal for most allozyme loci, which also suggests that muta-
tions mostly change protein charge by one unit and backward mutations may occur.
Taking this into account, we consider the mutation model used in our simulations to
be untraditional but adequate.
Phylogeny of beech in western Eurasia
Formulating a priori hypotheses about the phylogeny of F. sylvatica s. l. is not easy:
molecular studies have mostly covered only a part of the range of F. orientalis [40,41],
and those based on more representative sampling did not yield a clear picture of the
relationships among genetic lineages within the species (e.g., the ITS study of Denk
et al. [17]).
In earlier studies [22,23] based on the material examined in this one, a much
more detailed view of beech phylogeny in terms of the number (24) of small regional
populations sampled was attained. However, such a detailed subdivision would result
in an enormous number of potential evolutionary scenarios with equal or comparable
posterior probabilities. erefore, the small regions were merged into four geographic
areas in the present study: Europe (i.e., F. sylvatica s. str.), Asia Minor (including the
Ponthic range, Amanus Mountains, and the southeastern tip of the Balkan Peninsula),
the Caucasus (including the Greater and the Lesser Caucasus and the Colchis), and
the Alborz.
e results of ABC simulations in this study did not dier substantially from the
conclusions of earlier studies derived from distance-based (neighbor-joining tree,
reticulogram) and model-based (Bayesian) analyses; however, they were not identical
to these either. e Structure analysis [24] yielded two clusters within F. sylvatica s.
l.: one corresponding to F. sylvatica s. str., and the other (a more homogeneous one)
to F. orientalis, with the gene pools of “F. moesiaca” and “F. t a u r i c a ” each containing
a mixture of samples from both of these clusters [22]. In terms of the evolutionary
scenarios tested herein by ABC, this corresponds to an early split of the F. sylvatica
Tab . 1 Continued
Locus Allele F. sylvatica s.str. F. orientalis
Pgm-A 112 0.00181 0.02895
100 0.99508 0.96952
97 0.00312 0.00154
Skdh-A 114 0.00410 0.00671
100 0.96727 0.73705
86 0.01128 0.02426
72 0.01729 0.23163
58 0.00005 0.00034
6 of 11© The Author(s) 2018 Published by Po lish Botanical Socie ty Acta Soc Bot Pol 87(2):3582
Gömör y et al. / ABC-inferred phylo geny of Fagus sylvatica s. l .
Tab. 2 Posterior estimates of the parameters of the demographic inference based on the approximate Bayesian computation for the
best-supported scenarios in dierent constellations of taxa and regional populations of Fagus sylvatica s. l.
Parameter
Best-supported scenario Alternative scenario
Mode
95% condence
interval Mode
95% condence
interval
F. sylvatica s. l. Scenario 3 Scenario 1
Probability of scenario 0.6329 0.6004–0.6654 0.2179 0.1879–0.2478
Ne (Alb) 26,500 12,600–63,300 30,200 13,200–84,200
Ne (Cau) 82,700 41,700–98,600 71,300 27,300–97,400
Ne (AMi) 88,000 42,400–98,300 74,100 30,900–97,900
Ne (Eur) 30,100 12,500–90,000 25,300 10,800–57,600
t1 (divergence Cau from Alb) 16,200 (1.62 My) 6,870–26,300
t2 (divergence Eur from AMi) 11,800 (1.18 My) 2,590–25,300
t3 (divergence AMi from Alb) 18,700 (1.87 My) 8,270–29,400
t1* (divergence Cau from AMi) 15,200 (1.67 My) 3,580–26,400
t2* (divergence Alb from AMi) 18,700 (2.06 My) 6,190–27,700
t3* (divergence AMi from Eur) 25,800 (2.84 My) 8,060–29,300
µ 6.46 × 10−6 2.48 × 10−7 – 9.92 × 10−6 6.35 × 10−6 2.88 × 10−6 – 9.91 × 10−6
F. sylvatica s. str. – F. orientalis –
“F. taurica” Scenario 4 Scenario 3
Probability of scenario 0.6925 0.6431–0.7420 0.2456 0.1989–0.2923
Ne (Alb) 31,700 19,800–69,300 27,700 15,200–62,500
Ne (Cau) 127,000 60,600–148,000 118,000 54,400–147,000
Ne (AMi) 95,200 38,700–146,000 91,900 39,500–145,000
Ne (Cri) 44,000 13,700–95,200 54,100 17,100–94,600
Ne (Eur) 38,600 18,400–89,000 36,600 16,200–87,900
t1 (hybridization Cau × Eur → Cri) 1,440 (144 ky) 532–6,400
Hybridization rate (Eur) 0.307 0.0757–0.855
t1* (hybridization AMi × Eur →
Cri)
3,070 (307 ky) 820–10,500
Hybridization rate* (Eur) 0.484 0.304–0.953
t2 (divergence Eur from AMi) 6,740 (674 ky) 2,300–9,760 9,330 (933 ky) 3,640–9,860
t3 (divergence AMi from Alb) 18,200 (1.82 My) 9,720–27,400 15,600 (1.56 My) 6,990–25,400
t4 (divergence Cau from Alb) 22,700 (2.27 My) 8,630–29,500 27,300 (2.73 My) 10,500–29,400
µ 5.80 × 10−6 1.14 × 10−6 – 9.45 × 10−6 7.05 × 10−6 1.91 × 10−6 – 9.89 × 10−6
F. sylvatica s. str. – F. orientalis –
“F. moesiaca” Scenario 1
Probability of scenario 0.9297 0.9153–0.9442
Ne (AMi) 108,000 51,900–146,000
Ne (Balk) 46000 14,600–138,000
Ne (Apen) 131,000 39,900–476,000
Ne (CEur) 74,900 22,700–145,000
t1 (divergence CEur from Balk) 963 (96 ky) 257–7,190
t2 (divergence Ape from Balk) 702 (70 ky) 193–5,700
t3 (divergence Balk from AMi) 8,170 (817 ky) 3,320–28,300
µ 1.5 × 10−6 2.66 × 10−7 – 4.11 × 10−6
* Parameters of the alternative scenario. t – timing of the event (number of generations before present); µ – mutation rate.
7 of 11© The Author(s) 2018 Published by Po lish Botanical Socie ty Acta Soc Bot Pol 87(2):3582
Gömör y et al. / ABC-inferred phylo geny of Fagus sylvatica s. l .
s. str. clade from some of the regional populations of F. orientalis, and is also in ac-
cordance with morphology-based relationships previously determined among taxa
and regional populations [17]. ABC support for such a scenario in this study was not
negligible (posterior probability of 0.22), but was still much lower than the support
for the scenario in which the easternmost beech population in Iran was the ancestral
one. e best-supported scenario placed the divergence of regional populations in the
Early Pleistocene, with F. sylvatica s. str. being the latest-diverging lineage. is timing
corresponds with the history of beech in western Eurasia as reected in the fossil record
[18]. is would mean that barriers to gene ow (probably due to range fragmenta-
tion associated with climate uctuations) appeared earlier within the Asian range of
beech than those between Asia Minor and Europe arose. Fossils reveal a permanent
presence of the genus Fagus in Europe and Western Asia since at least the Oligocene
[42], and a wide distribution during the Late Tertiary [16,18,43,44]. ese fossils
have been described using many dierent names, but those that were suggested to be
linked with F. sylvatica s. l. can probably all be subsumed under the name F. haidingeri
Kováts, which was documented in the Miocene and Pliocene of Europe and Georgia
[18]. Although the distribution of F. sylvatica-like fossils is not completely identical
with the current range of the extant species, their number and geographic coverage
suggest that the range of beech in the Pliocene was quite continuous. Fragmentation,
which began with the onset of Pleistocene climatic uctuations, may have disrupted
gene ow among regional “islands” and promoted genetic dierentiation, while later re-
established contacts during Pleistocene warm phases [16] were too short to homogenize
the gene pools again. Such a course of events is compatible with both scenarios tested
for the F. sylvatica s. l. complex in this study. Alternatively, beech may have eectively
disappeared from Europe at the end of the Pliocene, with a subsequent recolonization
of Europe from Asia Minor, while remnant populations surviving in Europe until the
Pleistocene (if any) were overlaid by this recolonization event.
In the case of Crimean beech, ABC simulations conrmed a hybrid origin of “F.
taurica”, with F. sylvatica s. str. as one of the parental lineages and the Caucasus/Colchis
regional population of F. orientalis as the second one. Alternatively, beech in Asia Minor
may have participated in the hybridization, as the posterior probability of this scenario
(0.25) is too large for it to be ignored. e timing of hybridization (mode of 144 ky BP)
was placed at the end of the Saalian glacial period, but the condence interval is broad
enough to include the whole Eemian interglacial (130 to 114 ky BP), and is thus not
unrealistic. Beech was certainly present in Crimea during the Pleistocene [45,46], and
it was also present (although not dominant) in the forests on the other side of the cur-
rent Kerch Strait [47]. A contact between the Crimean and the Caucasian population is
thus easy to imagine. e more problematic aspect is a contact with Balkan populations
of F. sylvatica s. str. Fagus pollen was discovered in Late Pleistocene sediments along
the northern Black Sea coast [48], but a continuous presence of beech in this area is
not suciently documented. A certain anity of Moldovan beech populations to the
Crimean ones [23,49] may support the existence of such contact, but this would require
a permanent presence of beech in southern Romania throughout the Weichselian glacial
period, which cannot be excluded but to date has not been corroborated by the fossil
record (see [50]). No nal conclusions can thus be made concerning the plausibility
of a hybrid origin of Crimean beech.
In contrast, there is no support for a hybrid origin of Balkan beech. e simple
divergence scenario without reticulations received the best support (posterior prob-
ability of 0.92), and the odds against the next-best scenario were 16:1. is scenario
again suggests an early divergence of F. sylvatica s. str. from the Turkish lineage of F.
orientalis occurring during the early Pleistocene (the mode of 817 ky BP falls into the
Cromerian interglacial). On the other hand, the divergence of both the Italian and
Central-European lineages from the Balkan one was suggested to occur at the begin-
ning of the last (Weichselian) glacial. In light of the Quaternary history of vegetation
in Europe, such a course of events is plausible: the range of beech expanded and con-
tracted along with glacial/interglacial cycles, with a maximum extension during the
Holsteinian interglacial [51], and such uctuations may have contributed to recurrent
contacts among local populations and thus have maintained the integrity of gene pools.
e expansion during the Eemian was limited to southern Europe, and the onset of
the Weichselian glacial may have led to range fragmentation, interruption of gene ow
8 of 11© The Author(s) 2018 Published by Po lish Botanical Socie ty Acta Soc Bot Pol 87(2):3582
Gömör y et al. / ABC-inferred phylo geny of Fagus sylvatica s. l .
among refugial areas, and, consequently, formation of the present-day genetic lineages.
is view was also supported by the discovery of pollen, the chloroplast haplotype of
which was identical to that of the Balkan lineage (“F. moesiaca”), in the Venice lagoon
from the last pleniglacial [52]; apparently, the formation of the present-day lineages
within F. sylvatica s. str. may have begun quite late.
Conclusions
e number of scenarios tested in this study was limited, but not primarily for technical
reasons: we chose only those that make sense in terms of biogeography, and which were
supported by fossil evidence. For instance, the splitting of the European lineage from
the Caucasian one was not considered, as it would suggest the migration of beech into
Europe along the northern Black Sea coast, and there is not enough fossil evidence
to support such a scenario. Taking this limitation into account, this study partly cor-
roborated earlier views on the phylogeny of beech in the western Eurasian area. ABC
simulations conrmed that F. sylvatica subsp. orientalis is a paraphyletic subspecies, as
it excludes F. s. subsp. sylvatica, which diverged from the Asia-Minor lineage. A hybrid
origin was conrmed for one transitional taxon (Crimean “F. t au r ica ”), but not for the
other (Balkan “F. moesiaca”).
e applicability of the stepwise mutation model to allozyme data is a crucial point in
judging whether the above conclusions are realistic. Posterior model checking indicated
that both the model and the choice of prior parameters were appropriate for these data
in the present study. It seems that the mutation rates of allozymes t well within the
time frame of the processes under study herein. is does not imply that simulations
using this model on longer temporal scales (e.g., phylogeny of the family or the genus
as a whole) or shorter scales (e.g., postglacial migration) would be equally successful.
Nevertheless, since a broad-scale application of allozyme analysis in the 1970s, plenty
of data have been gathered in many plant taxa. Our study suggests that a reanalysis of
these data, employing new methodological approaches such as ABC and revision of
earlier views of phylogeny, may be meaningful.
Acknowledgments
Experimental material for this study was collected in collaboration with M. Akhalkadze, A.
Andonoski, D. Ballian, N. Bilir, R. Brus, M. Dida, J. Gračan, V. Hynek, R. Longauer, F. Popescu,
P. Salehi-Shanjani, I. M. Shvadchak, M. Sulkowska, Z. Tomović, J. Vyšný, and P. Zhelev.
Supplementary material
e following supplementary material for this article is available at http://pbsociety.org.pl/
journals/index.php/asbp/rt/suppFiles/asbp.3582/0:
Fig. S1 Geographical display of the scenarios tested in ABC analyses.
Fig. S2 Model checking for ABC simulations of the chosen scenarios based on principal com-
ponent analysis (PCA) of summary statistics for: F. sylvatica s. l., F. sylvatica s. str. – F. orientalis
– “F. ta ur i ca”, F. sylvatica s. str. – F. orientalis – “F. moesiaca”.
Appendix S1 Outcomes of the ABC simulations.
References
1. Tutin TG, Heywood VH, Burges NA, Valentine, DH, Moore DM, editors. Flora
Europaea. Volume 1, Psilotaceae to Platanaceae. Cambridge: Cambridge University Press;
1993.
2. Denk T. e taxonomy of Fagus in western Eurasia, 1: Fagus sylvatica
subsp. orientalis (= F. orientalis). Feddes Repert. 1999;110:177–200.
https://doi.org/10.1002/fedr.19991100305
9 of 11© The Author(s) 2018 Published by Po lish Botanical Socie ty Acta Soc Bot Pol 87(2):3582
Gömör y et al. / ABC-inferred phylo geny of Fagus sylvatica s. l .
3. Denk T. e taxonomy of Fagus in western Eurasia, 2: Fagus sylvatica subsp. sylvatica.
Feddes Repert. 1999;110:381–412. https://doi.org/10.1002/fedr.19991100510
4. Wul EV. e beech in Crimea, its systematic position and origin. In: Rübel E, editor. Die
Buchenwälder Europas. Bern: Hans Huber; 1932. p. 223–261.
5. Гулисашвили [Gulisashvili] ВЗ [VZ], Махатадзе [Mahatadze] ЛБ [LB], Прилипко
[Prilipko] ЛИ [LI]. Растительность Кавказа [Rastitel’nost’ Kavkaza]. Москва [Moskva]:
Наука [Nauka]; 1975.
6. Mobayen S, Tregubov V. La carte de la végétation naturelle de l’Iran. Tehran: Université
de Tehran; 1970.
7. Mayer H, Aksoy H. Wälder der Türkei. Stuttgart: Gustav Fischer; 1986.
8. Czeczott H. Distribution of Fagus orientalis Lipsky. In: Rübel E, editor. Die Buchenwälder
Europas. Bern: Hans Huber; 1932. p. 362–387.
9. Greuter W, Raus T. Med-checklist notulae, 4. Willdenowia. 1981;11:271–280.
10. Poplavskaja GI. Contribution to experimental studies of the systematics of Crimean
beech. Trudy Leningradskogo Obshchestva Estestvoispytatelej. 1936;3:353–371.
11. Czeczott H. Studium nad zmiennością liści buków: F. orientalis Lipsky, F. sylvatica L.
i form pośrednich. Cz. I. Rocznik Dendrologiczny. 1933;5:45–121.
12. Mišić V. Varijabilitet i ekologija bukve u Jugoslaviji. Beograd: Biološki Institut N. R.
Srbije; 1957.
13. Borza A. Le genre Fagus dans la République Populaire Roumaine. Biológia. 1965;20:367–
373.
14. von Wühlisch G. EUFORGEN technical guidelines for genetic conservation and use for
European beech (Fagus sylvatica). Rome: Bioversity International; 2008.
15. Kandemir G, Kaya Z. EUFORGEN technical guidelines for genetic conservation and use
of oriental beech (Fagus orientalis). Rome: Biodiversity International; 2009.
16. Denk T, Grimm GW, Hemleben V. Patterns of molecular and morphological
dierentiation in Fagus (Fagaceae): phylogenetic implications. Am J Bot. 2005;92:1006–
1016. https://doi.org/10.3732/ajb.92.6.1006
17. Denk T, Grimm G, Stögerer K, Langer M, Hembelen V. e evolutionary history of Fagus
in western Eurasia: the evidence from genes, morphology and the fossil record. Plant Syst
Evol. 2002;232:213–236. https://doi.org/10.1007/s006060200044
18. Denk T, Grimm GW. e biogeographic history of beech trees. Rev Palaeobot Palynol.
2009;158:83–100. https://doi.org/10.1016/j.revpalbo.2009.08.007
19. Göker M, Grimm GW. General functions to transform associate data to host data, and
their use in phylogenetic inference from sequences with intra-individual variability BMC
Evol Biol. 2008;8:86. https://doi.org/10.1186/1471-2148-8-86
20. Grimm GW, Denk T, Hemleben V. Coding of intraspecic nucleotide polymorphisms:
a tool to resolve reticulate evolutionary relationships in the ITS of beech trees (Fagus L.,
Fagaceae). Syst Biodivers. 2007;5:291–309. https://doi.org/10.1017/S1477200007002459
21. Renner SS, Grimm GW, Kapli P, Denk T. Species relationships and divergence
times in beeches: new insights from the inclusion of 53 young and old fossils in
a birth–death clock model. Philos Trans R Soc Lond B Biol Sci. 2016;371:20150135.
https://doi.org/10.1098/rstb.2015.0135
22. Gömöry D, Paule L, Vyšný J. Patterns of allozyme variation in western-Eurasian beeches.
Bot J Linn Soc. 2007;154:165–174. https://doi.org/10.1111/j.1095-8339.2007.00666.x
23. Gömöry D, Paule L, Reticulate evolution patterns in western Eurasian beeches. Botanica
Helvetica. 2010;120(1):63–74. https://doi.org/10.1007/s00035-010-0068-y
24. Pritchard JK, Stephens M, Donnelly P. Inference of population structure from multilocus
genotype data. Genetics. 2000;155:945–959.
25. Corander J, Waldmann P, Sillanpää MJ. Bayesian analysis of genetic dierentiation
between populations. Genetics. 2003;163:367–374.
26. Beaumont MA, Zhang WY, Balding DJ. Approximate Bayesian computation in
population genetics. Genetics. 2002;162:2025–2035.
27. Csilléry K, Blum MGB, Gaggiotti OE, Francois O. Approximate Bayesian
computation (ABC) in practice. Trends Ecol Evol. 2010;25:410–418.
https://doi.org/10.1016/j.tree.2010.04.001
28. Merzeau D, Di Giusto F, Comps B, iébaut B, Letouzey J, Cuguen J. e allozyme
10 of 11© The Author(s) 2018 Published by Polis h Botanical Societ y Acta Soc Bot Pol 87(2):3582
Gömör y et al. / ABC-inferred phylo geny of Fagus sylvatica s. l .
variants of beech (Fagus sylvatica L.): inheritance and application to a study of the mating
system. Silvae Genet. 1989;38:195–201.
29. Müller-Starck G, Starke R. Inheritance of isoenzymes in European beech (Fagus sylvatica
L.). J Hered. 1993;84:291–296.
30. Cornuet JM, Pudlo P, Veyssier J, Dehne-Garcia A, Gautier M, Leblois R, et al. DIYABC
v2.0: a soware to make approximate Bayesian computation inferences about population
history using single nucleotide polymorphism, DNA sequence and microsatellite data.
Bioinformatics. 2014;30:1187–1189. https://doi.org/10.1093/bioinformatics/btt763
31. Voelker RA, Schaer HE, Mukai T. Spontaneous allozyme mutations in Drosophila
melanogaster: rate of occurrence and nature of the mutants. Genetics. 1980;94:961–968.
32. Baer CF, Miyamoto MM, Denver DR. Mutation rate variation in multicellular
eukaryotes: causes and consequences. Nat Rev Genet. 2007;8:619–631.
https://doi.org/10.1038/nrg2158
33. Houston Durrant T, de Rigo D, Caudullo G. Fagus sylvatica and other beeches in Europe:
distribution, habitat, usage and threats. In: San-Miguel-Ayanz J, de Rigo D, Caudullo G,
Houston Durrant T, Mauri A, editors. European atlas of forest tree species. Luxembourg:
Publications Oce of the European Union; 2016. p. e012b90+.
34. Korpeľ Š. Die Urwälder der Westkarpaten. Jena: Gustav Fischer Verlag; 1995.
35. Kimura M, Crow J. e number of alleles that can be maintained in a nite population.
Genetics. 1964;49:725–738.
36. Ohta T, Kimura M. A model of mutation appropriate to estimate the number of
electrophoretically detectable alleles in a nite population. Genet Res. 1973;22:201–204.
https://doi.org/10.1017/S0016672300012994
37. Barbadilla A, King LM, Lewontin RC. What does electrophoretic variation
tell us about protein variation? Mol Biol Evol. 1996;13(2):427–432.
https://doi.org/10.1093/oxfordjournals.molbev.a025602
38. Brown AHD, Marshall DR, Weir BS. Current status of the charge state model for protein
polymorphism. In: Gibson JB, Oakeshot JG, editors. Genetic studies of Drosophila
populations. Canberra: Australian National University Press; 1981. p. 15–43.
39. Richardson BJ, Baverstock PR, Adams M. Allozyme electrophoresis: a handbook for
animal systematics and population studies. Sydney: Academic Press; 1986.
40. Demesure B, Comps B, Petit R. Phylogeography of the common beech
(Fagus sylvatica L.) in Europe inferred by restriction studies of PCR
amplied chloroplast DNA fragments. Evolution. 1996;50:2515–2520.
https://doi.org/10.1111/j.1558-5646.1996.tb03638.x
41. Gailing O, von Wühlisch G. Nuclear markers (AFLPs) and chloroplast microsatellites
dier between Fagus sylvatica and F. orientalis. Silvae Genet. 2004;53:105–110.
https://doi.org/10.1515/sg-2004-0019
42. Kvaček Z, Walther H. Revision der mitteleuropäischen tertiären Fagaceen nach
blattepidermalen Charakteristiken. IV. Teil Fagus Linné. Feddes Repert. 1991;102:471–
534. https://doi.org/10.1002/fedr.19911020702
43. Tralau H. Die spättertiären Fagus-Arten Europas. Botaniska Notiser. 1962;115:147–177.
44. Denk T. Phylogeny of Fagus L. (Fagaceae) based on morphological data. Plant Syst Evol.
2003;240:55–81. https://doi.org/10.1007/s00606-003-0018-x
45. Gerasimenko N. Environmental changes in the Crimean mountains during the Last
Interglacial–middle pleniglacial as recorded by pollen and lithopedology. Quat Int.
2007;164–165:207–220. https://doi.org/10.1016/j.quaint.2006.12.018
46. Cordova CE, Gerasimenko NP, Lehman PH, Kliukin AA. Late Pleistocene and Holocene
paleoenvironments of Crimea: pollen, soils, geomorphology, and geoarchaeology.
In: Buynevich IV, Yanko-Hombach V, Gilbert AS, Martin RE, editors. Geology and
geoarchaeology of the Black Sea Region: beyond the ood hypothesis. Boulder, CO:
Geological Society of America; 2011. p. 133–164. (Special Papers, Geological Society of
America; vol 473). https://doi.org/10.1130/2011.2473(09)
47. Zubakov VA. Climatostratigraphic scheme of the Black-Sea Pleistocene and its
correlation with the oxygen isotope scale and glacial events. Quat Res. 1988;29:1–24.
https://doi.org/10.1016/0033-5894(88)90067-1
48. Dodonov AE, Zhou LP, Markova AK, Tchepalyga AL, Trubikhin VM,
Aleksandrovski AL, et al. Middle-Upper Pleistocene bio-climatic and magnetic
records of the northern Black Sea coastal area. Quat Int. 2006;149:44–54.
11 o f 11© The Author(s) 2018 Published by Po lish Botanical Socie ty Acta Soc Bot Pol 87(2):3582
Gömör y et al. / ABC-inferred phylo geny of Fagus sylvatica s. l .
https://doi.org/10.1016/j.quaint.2005.11.017
49. Gömöry D, Paule L, Shvadchak IM, Popescu F, Sulkowska M, Hynek V, et al. Spatial
patterns of the genetic dierentiation in European beech (Fagus sylvatica L.) at allozyme
loci in the Carpathians and the adjacent regions. Silvae Genet. 2003;52:78–83.
50. Magri D, Vendramin GG, Comps B, Dupanloup I, Geburek T, Gömöry
D, et al. Palaeobotanical and genetic data outline the Quaternary
history of European beech populations. New Phytol. 2006;171:199–222.
https://doi.org/10.1111/j.1469-8137.2006.01740.x
51. Peters R. Beech forests. Dordrecht: Springer Science & Business Media; 1997.
(Geobotany; vol 24). https://doi.org/10.1007/978-94-015-8794-5
52. Paetti D, Vettori C, Caramelli D, Vernesi C, Lari M, Paganelli A, et al. Unexpected
presence of Fagus orientalis complex in Italy as inferred from 45,000-year-old
DNA pollen samples from Venice lagoon. BMC Evol Biol. 2007;7(2 suppl):S6.
https://doi.org/10.1186/1471-2148-7-S2-S6
