Genetic differentiation and phylogeny of beech on the Balkan peninsula

& J. GRAC AN MO RY,* L. PAULE,* R. BRUS, P. ZHELEV, Z. TOMOVIC D. GO

*Faculty of Forestry, Technical University in Zvolen, T.G. Masaryka 24, SK-960 53 Zvolen, Slovakia Department of Forestry, Biotechnical Faculty, Vecna pot 83, SLO-1111 Ljubljana, Slovenia Faculty of Forestry, Higher Forestry Institute, BG-1156 Soa, Bulgaria JP Srbijasume, ure akovica 31, YU-11000 Beograd, Yugoslavia Forestry Research Institute, Cvijetno Naselje 41, HR-41420 Jastrebarsko, Croatia

Keywords:
Balkan; Fagus moesiaca (Domin Maly) Czeczott; genetic differentiation; phylogeny; taxonomy.

Summary
Genetic differentiation of 57 beech populations originating from a wide geographical space from the foothills of the Eastern Alps over the Balkan peninsula was investigated employing isozyme markers. Pooled datasets of European beech populations originating from four adjacent regions and Eastern beech populations originating from Thrace and Western Asia Minor were compared. Considerable differences of allele frequencies among regions were found in several marker loci. The highest level of genetic multiplicity and differentiation was found in the populations from the southern Balkans; however, the north-western populations showed higher genetic diversity. The pattern of genetic differentiation based on multilocus genetic distances is a clinal one. The populations belonging to the putative taxon Fagus moesiaca Czeczott seem to form an independent group. Three hypotheses of the evolutionary origin of this taxon are discussed: selection, introgressive hybridization and continuous evolution.

Introduction
Beech is one of the most represented indigenous tree species in all Balkan countries. Except for small isolated populations of Eastern beech, Fagus orientalis Lipsky, in south-eastern Bulgaria, eastern Greece and the European part of Turkey, it is generally denoted as common or European beech, Fagus sylvatica L. (Becker, 1981). However, local authors consider it mostly to be a separate species Fagus moesiaca (Domin, Maly) Czeczott. Balkan beech was rst described as a separate taxon by Josef Karel Maly in 1911. The description of this taxon was later completed by Czeczott (1933). Opinions regarding the taxonomical status of this taxon are varied. Frequently, it is described as a separate unit (Czeczott, ic (1957) considers it a phy1933; Fukarek, 1954). Mis logenetical link between F. sylvatica and F. orientalis. Sometimes it is considered a hybrid between both species morphologically closer to F. sylvatica (Becker, 1981), a
an Go univerzita, Correspondence: Dr Dus cka fakulta, Technicka mo ry, Lesn T.G. Masaryka 24, SK-960 53 Zvolen, Slovakia. Tel: +421 855 335111; fax: +421 855 22654; e-mail: gomory@vsld.tuzvo.sk

mixture of F. sylvatica and F. orientalis with the occurrence of transition forms dominated by characters of one of the nescu, two species (Stoyanoff, 1932), an ecotype (Sta 1979) or identical with the Crimean beech Fagus taurica Popl. (Didukh, 1992). The morphological description of `F. moesiaca' is rather vague. There is no agreement among different authors about the morphological traits discriminating between the Balkan and European and/or Eastern beech. For most characters, the mean values are different but the ranges of variation overlap considerably. In comparison with pure F. sylvatica, `F. moesiaca' should have larger leaves with more lateral veins, larger beechnuts and longer ic , 1957; Sta nescu, cupule peduncle (Czeczott, 1933; Mis 1979). In addition to the morphology, `F. moesiaca' differs from F. sylvatica by a high sprouting capacity and a considerably higher frequency of seed years, as well as ic , 1957). ecological requirements (Mis The description of the distribution range of Balkan beech is not unequivocal as well. The main part of the range seems to be the former Yugoslavia (Bosnia, Serbia, Montenegro, Macedonia), Albania, Bulgaria and Greece ic , 1957), but isolated occurrences (Fukarek, 1954; Mis have been reported from south-eastern Rumania, Hun-

746

J. EVOL. BIOL. 12 (1999) 746754 1999 BLACKWELL SCIENCE LTD

Genetic differentiation in Balkan beech

747

gary and even Poland and the former Czechoslovakia ti ex Fukarek, 1954; Sta nescu, 1979). Croatian and (Karpa Slovenian populations are generally considered F. sylvatica. European beech belongs to those forest tree species whose genetic variation has been very thoroughly documented within the majority of its range employing isozyme markers. However, the data for Balkan beech are scarce, especially for our main area of interest, i.e. the southern Balkans, and for Eastern beech they are practically missing. Comps et al. (1991) investigated beechwoods from the continental and Mediterranean parts of Croatia, and reported the presence of differences between these two regions. Data from Balkan countries were also included in a wide study of beechwoods in Central Europe (Comps et al., 1990), but the data from Serbia, Bulgaria and the Romanian Carpathians were pooled, so that no differentiation patterns within this large area could be identied. Recently, a study focusing on this region was published by Hazler et al. (1997). Although there is a gap in their material between Macedonia and Croatia, a north-west to south-east cline can be identied in their presentation of PCA results. In all these reports, beech in this region was denoted as Fagus sylvatica L. Recently, PCR-based cpDNA markers were utilized for the investigation of beech differentiation (Demesure et al., 1996). In Slovenian and Croatian beechwoods, no specic haplotype was found. Unfortunately, no samples from the southern Balkans were included. This study is aimed at the description of the differentiation pattern in beechwoods of the Balkan peninsula and the identication of possible evolutionary processes, contributing to the formation of this pattern.

Materials and methods

Fifty-seven indigenous beech populations were examined, originating from different altitudes and different locations. The geographical situation is shown in Fig. 1. In each beech stand, twigs with dormant buds were sampled from approximately 50 nonadjacent trees chosen at random. For the evaluation, they were pooled into seven geographical regions: Pre-Alpine region (northern Slovenia, abbreviation used further in the text PrAl), northern part of the Dinarian range (southern Slovenia, south-western Croatia, western Bosnia NDin), hills on the southern edge of the Pannonian lowland (Pann), Central Dinarian mountains (central Bosnia, western Serbia CDin), Southern Dinarian mountains (central and eastern Serbia, Macedonia, western Bulgaria SDin), Stara Planina mountains (northern Bulgaria StPl), and the Rodopi mountains (southern Bulgaria Rodo). The ka gora population (marked with an arrow in Fig. 1) Frus exhibits morphological features of F. moesiaca, and therefore it was classied to the Central Dinarian region, although, from a purely geographical point of view, it should belong to the Pannonian hills. Pooled Eastern

beech populations originating from Thrace (eastern Bulgaria and European Turkey) and western Asia Minor (abbreviation TAMi), as well as pooled common beech populations from four regions: Calabria (Cala), Central and Southern Alps, Jura and Schwarzwald (Alps), Hercynian region and Western Carpathians (HWCa), and Eastern and Southern Carpathians (ESCa), were used for comparison (Fig. 1). Enzymes were extracted from buds and cortical tissues of each individual, and were separated by means of starch electrophoresis. Protein separation and staining procedures were described by Merzeau et al. (1989) and Mu ller-Starck & Starke (1993). Nine isozyme systems coded by 12 loci were scored: isocitrate dehydrogenase (Idh-A), malate dehydrogenase (Mdh-A, Mdh-B, Mdh-C), menadione reductase (Mnr-A), phosphoglucomutase (Pgm-A), phosphoglucose isomerase (Pgi-B), peroxidase (Px-A, Px-B), glutamate-oxaloacetate transaminase (GotB), leucine aminopeptidase (Lap-A) and shikimate dehydrogenase (Skdh-A). Where possible, the designation of alleles followed that of Merzeau et al. (1989). Allelic frequencies at each locus were calculated on the basis of diploid genotypes. Differences of allelic frequencies among populations and regions were tested using the probability test (Raymond & Rousset, 1995a). Subsequently, a global test across loci was calculated using Fisher's method (Rousset & Raymond, 1995). Expected HardyWeinberg heterozygosity, effective number of alleles and total number of alleles were used to characterize the allelic diversity. Since the last measure is very sensitive to sample size, minimum frequency of an allele for being identied with a 95% probability was calculated following Gregorius (1980). Genetic distances between populations were calculated following Nei (1978). Principal coordinate analysis (Gower, 1966) was utilized for interpretation of the matrix of genetic distances. To quantify the degree of differentiation among populations, the FST statistic (Weir & Cockerham, 1984) was used. The possible isolation by distance was tested according to Rousset (1997). As recommended for a two-dimensional case, regression of estimates of FST/(1 FST) against logarithm of distance was evaluated. Since we have no information about migration paths, simple air distances between populations were used. For the calculations, B I O S Y S - 1 (Swofford & Selander, 1981) and G E N E P O P v.3.1b (Raymond & Rousset, 1995b) were employed.

Results
Mean allelic frequencies in the investigated regions are presented in Table 1. The abbreviations of regions of primary interest are in bold type. There were very few alleles identied which were specic for one or few adjacent regions. Most alleles were found in F. orientalis, `F. moesiaca' or Calabrian populations (Px-B/52, Mdh-A/ 130, 101, 88, Mdh-B/78, 56, Pgi-B/126, Skdh-A/58). No

J. EVOL. BIOL. 12 (1999) 746754 1999 BLACKWELL SCIENCE LTD

748

MO RY ET AL. D. GO

Fig. 1 Location of the investigated populations and the populations in the comparison regions.

private allele was found in any of the regions of occurrence of pure F. sylvatica. Although both European and Eastern beeches are generally recognized as different species, their distributions of alleles (at least at the investigated set of nuclear genes) are very similar. The most frequent alleles in both species were always identical. Allelic proles of the putative taxon `F. moesiaca' are more similar to those of European than to those of Eastern beech. Despite the fact that the easternmost populations of Balkan beech in Stara Planina and the Rodopi mountains are situated geographically very closely to the range of the Eastern beech (few tens of kilometres), the change of frequencies of several alleles is quite abrupt (e.g. Got-B/36, Mdh-A/125, Pgi-B/113). At Mdh-C and Skdh-A, allelic proles of Balkan beech exhibit transitional tendencies. Among the populations, which are generally classied as pure F. sylvatica, the Balkan populations are most similar to Calabrian provenances: they exhibit the same deviations from allelic proles common to the remaining part of the distribution range (absence and/or decreased frequency of Px-B/13, decreased frequencies of Got-B/36, Lap-A/94, Mdh-C/22, presence of Mdh-B/78, 56). The only locus displaying a different allelic distribution is Skdh-A, where the allele 72, quite frequent in Balkan populations, is absent in Calabria.

Multilocus patterns of the genetic differentiation were expressed by genetic distances. Owing to relatively small sample sizes (50100 individuals per population), the pattern of differentiation among individual populations is slightly distorted. In fact, using larger samples, a smaller dispersion of populations around the gravity centres of regions could be expected. Despite this fact, a clear northwest to south-east cline can be identied (Fig. 2). Although the space between the pure F. sylvatica and the putative taxon F. moesiaca is not wide, the points displaying populations belonging to these two taxa are not intermixed. Eastern beech is clearly separated from the rest of the presented material, but Balkan beech populations, located near its range (eastern Stara Planina mountains, Rodopi mountains), exhibit the greatest similarity to this species. Among European beech, Calabrian beech is most similar to Balkan populations, whereas the remaining comparison regions appear similar to Pre-Alpine and North-Dinarian populations. However, it is necessary to emphasize that a considerable portion of variation (approximately 43%) is not displayed. The regions are genetically heterogeneous (at least at some loci), as indicated by tests of the genic differentiation as well as FST values (Table 2). Pooling the populations belonging to one region does not permit recognition of this intraregional variation. On the other hand, the substantially increased summary sample size reduces the random

J. EVOL. BIOL. 12 (1999) 746754 1999 BLACKWELL SCIENCE LTD

Genetic differentiation in Balkan beech

749

Table 1 Mean allelic frequencies in the investigated regions of Balkan beech and comparison regions of European and Eastern beech.
Allele Rf 105 100 52 39 26 13 106 100 97 94 36 18 126 100 74 63 52 116 100 84 135 130 125 101 88 118 109 100 78 56 22 18 126 113 100 87 112 100/6 94 114 100 86 72 58 F. sylvatica Cala 0.339 0.661 0.165 0.835 0.019 0.696 0.247 0.038 0.174 0.826 0.013 0.987 0.234 0.763 0.003 0.022 0.962 0.016 0.072 0.003 0.897 0.006 0.022 1.000 0.003 0.991 0.006 0.006 0.994 0.003 0.984 0.013 Alps 0.375 0.625 0.155 0.703 0.142 0.032 0.429 0.347 0.192 0.358 0.642 0.021 0.897 0.001 0.081 0.328 0.663 0.009 1.000 0.050 0.041 0.909 0.305 0.695 0.999 0.001 1.000 0.003 0.984 0.004 0.009 HWCa 0.318 0.682 0.192 0.694 0.114 0.021 0.498 0.303 0.178 0.375 0.625 0.015 0.913 0.004 0.068 0.312 0.680 0.008 0.001 0.999 0.060 0.003 0.937 0.286 0.714 0.003 0.983 0.014 0.002 0.996 0.002 0.005 0.980 0.009 0.006 ESCa 0.338 0.662 0.155 0.741 0.104 0.013 0.531 0.317 0.139 0.385 0.615 0.030 0.891 0.009 0.070 0.288 0.709 0.003 0.004 0.996 0.049 0.949 0.002 0.229 0.771 0.003 0.988 0.009 0.004 0.993 0.003 0.004 0.972 0.023 0.001 PrAl 0.408 0.592 0.209 0.729 0.062 0.015 0.455 0.327 0.203 0.364 0.636 0.001 0.924 0.011 0.064 0.264 0.730 0.006 1.000 0.046 0.004 0.950 0.290 0.710 0.003 0.979 0.018 0.001 0.994 0.005 0.008 0.987 0.001 0.004 NDin 0.385 0.615 0.246 0.689 0.065 0.027 0.427 0.354 0.192 0.338 0.662 0.018 0.922 0.001 0.059 0.333 0.661 0.006 0.002 0.998 0.027 0.002 0.971 0.245 0.755 0.002 0.984 0.014 0.999 0.001 0.004 0.987 0.002 0.007 Pann 0.371 0.629 0.236 0.736 0.028 0.008 0.465 0.345 0.182 0.379 0.621 0.032 0.935 0.002 0.031 0.193 0.807 1.000 0.041 0.957 0.002 0.217 0.783 0.002 0.984 0.014 0.993 0.007 0.013 0.974 0.013 `F. moesiaca' CDin 0.401 0.599 0.168 0.805 0.027 0.024 0.503 0.310 0.163 0.294 0.706 0.012 0.965 0.022 0.001 0.302 0.692 0.006 0.001 0.998 0.001 0.073 0.926 0.001 0.160 0.840 0.003 0.981 0.016 0.992 0.008 0.971 0.003 0.026 SDin 0.393 0.607 0.002 0.171 0.779 0.048 0.031 0.500 0.306 0.163 0.217 0.783 0.016 0.933 0.051 0.215 0.781 0.004 0.001 0.004 0.995 0.100 0.001 0.897 0.002 0.098 0.902 0.002 0.973 0.025 0.992 0.008 0.001 0.967 0.001 0.031 StPl 0.379 0.621 0.006 0.197 0.754 0.043 0.022 0.609 0.259 0.110 0.267 0.733 0.025 0.909 0.005 0.061 0.182 0.804 0.014 0.031 0.967 0.001 0.001 0.059 0.935 0.005 0.001 0.115 0.885 0.034 0.955 0.011 0.002 0.998 0.006 0.912 0.016 0.066 Rodo 0.353 0.647 0.211 0.705 0.084 0.003 0.646 0.245 0.106 0.267 0.733 0.043 0.914 0.003 0.040 0.190 0.810 1.000 0.084 0.910 0.006 0.073 0.927 0.003 0.095 0.881 0.021 0.006 0.994 0.003 0.860 0.003 0.134 F. orientalis TAMi 0.283 0.717 0.214 0.654 0.132 0.014 0.854 0.100 0.032 0.058 0.942 0.056 0.906 0.026 0.012 0.343 0.657 0.121 0.875 0.004 0.012 0.002 0.952 0.021 0.013 0.002 0.998 0.180 0.820 0.049 0.951 0.591 0.010 0.398 0.001

Locus Px-A Px-B

Lap-A

Got-B Mnr-A

Idh-A

Mdh-A

Mdh-B

Mdh-C Pgi-B

Pgm-A

Skdh-A

variation due to sampling error. The regions of occurrence of pure F. sylvatica (except Calabria) are clearly less differentiated than Balkan beech, although they cover an incomparably larger area with very heterogeneous

climatic and soil conditions (see the dispersion of regions in Fig. 3 or FSTtax values in Table 2). The cline from Central Europe through the whole Balkan peninsula, which may even be prolonged to Bulgarian and Turkish Eastern

J. EVOL. BIOL. 12 (1999) 746754 1999 BLACKWELL SCIENCE LTD

750

MO RY ET AL. D. GO

Fig. 2 Principal coordinate analysis of the matrix of genetic distances between investigated populations projection into the 1st and 2nd principal axes.

Table 2 Characteristics of allelic multiplicity, diversity and differentiation of the investigated regions.
ne Species F. sylvatica Region Cala Alps HWCa ESCa PrAl NDin Pann CDin SDin StPl Rodo TAMi Np 3 6 47 46 11 7 4 7 12 13 3 10 Ni 158 372 2225 2299 716 419 198 546 630 637 161 497 N 32 31 35 35 34 34 32 34 36 38 33 35 a (95%) 0.039 0.019 0.003 0.003 0.011 0.017 0.032 0.016 0.012 0.012 0.038 0.015 mean* pooled 1.222 1.382 1.364 1.341 1.354 1.352 1.329 1.308 1.302 1.324 1.340 1.349 He mean* pooled (SE) 0.182 0.277 0.267 0.259 0.262 0.261 0.244 0.238 0.232 0.245 0.255 0.051 0.068 0.066 0.062 0.066 0.068 0.063 0.062 0.057 0.051 0.050 FSTreg* FSTtax* Ndiff*

1.342 1.346 1.319 1.306 1.295 1.314 1.330

0.259 0.260 0.244 0.236 0.230 0.241 0.250

0.0127 0.0067 0.0143 0.0085 0.0171 0.0153 0.0227

0.0125

7 3 3 6 9 8 5

F. moesiaca

0.0193

F. orientalis

0.259 0.050

*Comparison regions excluded from the calculation. Comparison regions included in the calculation. Np number of populations included; Ni total number of analysed individuals; N total number of alleles; a(95%) minimum frequency of an allele to be found with a 95% probability; ne effective number of alleles; He expected heterozygosity (SE standard error); FSTreg differentiation among populations within regions; FSTtax differentiation among populations within taxa; Ndiff number of loci exhibiting signicant (a < 0.05) differences of allelic frequencies between populations within region.

beech, is even more pronounced than in the previous case (Fig. 3). The pairwise evaluation of differentiation between regions (Table 3) conrms these trends. Global tests of differentiation across the investigated loci proved that there are highly signicant differences between all pairs of regions. Eastern beech is clearly differentiated from the remaining regions (FST ranges from 0.0622 to 0.1442, 10 12 loci exhibiting signicant differences of allelic frequencies). Beech in Calabria exhibits more similarity to Balkan beech (FST 0.01530.0211) than to typical European beech (FST 0.04280.0664). Differentiation increases with growing geographical distance between regions. Figure 3 as well as Table 3 indicate that the Central Dinarian region occupies an intermediary position between both taxa.

The total number of alleles is very sensitive to sample size. Therefore, it is not surprising that in the comparison regions of the Carpathians and Hercynian range, represented by the largest pooled samples (over 2000 trees), a relatively high number of alleles was found. Here, a frequency of approximately 0.003 is sufcient for an allele to be found with a probability of 95%. In the Rodopi mountains, where only three populations were sampled, the necessary frequency is about 13 times higher. Considering this fact, F. moesiaca seems to be a taxon with an increased total number of alleles (a maximum of 38 alleles found in the Stara Planina region). The trend of the total number of alleles seems to be opposite to that of expected heterozygosity and

J. EVOL. BIOL. 12 (1999) 746754 1999 BLACKWELL SCIENCE LTD

Genetic differentiation in Balkan beech

751

Fig. 3 Principal coordinate analysis of the matrix of genetic distances between pooled regions projection into the 1st and 2nd principal axes.

Table 3 Differentiation among pairs of regions: the number of loci exhibiting signicant (a < 0.05) differences of allelic frequencies based on probability test (above diagonal) and pairwise FST (below diagonal).
Region Cala Alps HWCa ESCa PrAl NDin Pann CDin SDin StPl Rodo TAMi Cala 0.0664 0.0539 0.0428 0.0576 0.0563 0.0511 0.0270 0.0171 0.0153 0.0211 0.0875 Alps 8* 0.0024 0.0057 0.0033 0.0032 0.0114 0.0144 0.0287 0.0316 0.0417 0.1434 HWCa 8* 4* 0.0025 0.0042 0.0041 0.0084 0.0128 0.0258 0.0245 0.0336 0.1312 ESCa 8* 8* 10* 0.0054 0.0063 0.0057 0.0086 0.0194 0.0165 0.0261 0.1272 PrAl 9* 5* 6* 9* 0.0025 0.0032 0.0098 0.0218 0.0227 0.0352 0.1442 NDin 9* 4* 6* 8* 3* 0.0067 0.0086 0.0215 0.0246 0.0355 0.1356 Pann 8* 8* 7* 5* 4* 3* 0.0093 0.0160 0.0143 0.0253 0.1398 CDin 10* 9* 7* 9* 7* 6* 7* 0.0061 0.0107 0.0212 0.1175 SDin 9* 10* 11* 11* 9* 7* 7* 8* 0.0057 0.0135 0.1083 StPl 10* 9* 10* 10* 11* 9* 10* 9* 9* 0.0033 0.0857 Rodo 9* 9* 9* 9* 9* 8* 8* 8* 6* 5* 0.0622 TAMi 10* 12* 10* 12* 12* 10* 11* 11* 11* 11* 10*

* Global test of differentiation across loci signicant at a < 0.001.

effective number of alleles. The lowest values of both these measures were found in south-eastern regions (Central and Southern Dinarians, Stara Planina mountains). Beechwoods in these regions contain most alleles, but apparently they are less regularly represented than in pure European beech, except in the Calabrian populations. Rousset's (1997) model of isolation by distance provided a regression FST a1 FST 0X0262 0X0085 ln d which is highly signicant (P > 0.9999).

Discussion
The authors who recognized F. moesiaca as a separate taxon formulated three principal hypotheses about its taxonomical status and evolutionary origin:

nescu (1979) considered Balkan beech as an eco1 Sta type of F. sylvatica, i.e. a product of selection, mainly due to the specic climatic conditions. 2 Balkan beech is a hybrid between F. sylvatica and F. orientalis (Stoyanoff, 1932; Becker, 1981). This hypothesis assumes that European beech is considered a phylogenetically more ancient form than Balkan beech. After the formation of European beech (which must have occurred more to the north or to the west), the species spread back to the Balkan peninsula, where it came into contact with Eastern beech. Both species hybridized, having formed Balkan beech as a transitional form. 3 Balkan beech constitutes an intermediary stage of the evolution from F. orientalis to F. sylvatica. This hypothesis assumes that European beech evolved from Balkan beech through selection and genetic drift during glaciation (i.e. Balkan beech is a more ancient taxon).

J. EVOL. BIOL. 12 (1999) 746754 1999 BLACKWELL SCIENCE LTD

752

MO RY ET AL. D. GO

There does not seem to be enough experimental evidence to validate the rst hypothesis. Under the assumption that selection is mainly responsible for the observed patterns, beech in the other parts of its distribution range should respond to similar environmental gradients by similar changes of morphology as well as genetic structures. There are further regions where beech exhibits similar morphological features, e.g. Crimea, Podolia or Moldova. That is why Didukh (1992) and further Ukrainian authors designated Crimean and Podolian beech as F. moesiaca. However, it is necessary to emphasize that these beechwoods, although being taxonomically intermediate as well, are genetically differentiated from Balkan beech (Go mo ry et al., 1998). In addition, they have developed in quite different environmental (at least climatic) conditions. The patterns of spatial distribution of allozyme frequencies, expected under the assumption of selection, depend on whether isozymes are selectively neutral markers. We do not intend to re-initiate the never ending discussion on this topic. If the observed trends of morphological change are produced by selection, but isozyme loci are considered neutral, then the pattern of spatial distribution of isozyme allele frequencies is expected to be random, and no trends or clines should be observed. If allozymes possess an adaptive value, then their frequencies should follow climatic gradients. In our material, neither the one nor the other was observed. At some loci, allelic frequencies of Balkan beech populations were different from those of typical European beech. However, these changes do not seem to exhibit any clear trends, comparable with those in the other parts of the Mediterranean region. For the locus Px-B, the allele 13 is less represented in the Balkan populations, in contrast to the Alps and Carpathians. A similarly decreased frequency was reported by Comps et al. (1987) for north-western Spain. On the other hand, even in the Pyrenees, its frequency increased again, whereas a less frequent occurrence was reported for regions further to the north and east (Comps et al., 1987; Go mo ry et al., 1992). Similarly, the representation of the Mdh-C/22 allele was generally reduced in the Balkans. Its frequency also decreased over the Apennine Peninsula from north to south (Leonardi & Menozzi, 1995); it is even absent in Calabria. However, for Corsica with a similarly hot Mediterranean climate, higher values (over 0.2), comparable with Central Europe, were reported by Comps et al. (1990). ic (1957) identied a gradual change of traits of Mis leaves, owers and fruits from the Caucasus over the Balkan peninsula up to Germany. We observed a similar tendency of genetic differentiation based on nuclear protein markers. This fact would support the selection hypothesis, since clines are mostly produced by selection. However, gene exchange between neighbouring populations can also produce similar patterns, and it is

sometimes difcult to distinguish the effects of selection from those of isolation by distance. In addition, there was no allele in our sample exhibiting an unequivocal clinal pattern of frequency change. Only summarizing the genetic differences over the whole set of investigated loci produced a more or less smooth cline. In any case, only discovering the background physiological mechanisms causing the changes of allelic frequencies (e.g. different activity of gene products enzymes under different environmental conditions) can prove or disprove the action of selection. Assuming Balkan beech is a hybrid between F. sylvatica and F. orientalis, it must be phylogenetically younger than both `parental' species. Eastern beech is generally recognized as the most ancient taxon within this complex, with undoubtedly Tertiary origin. As stated by Czeczott (1933), fossil leaves with the morphological features of F. orientalis have been found in Tertiary sediments throughout Europe and the Caucasus, although they were frequently classied as F. pliocenica Sap. This assumption is in full concordance with the information on genetic structure of this species. Eastern beech is characterized by an incomparably higher level of differentiation. The differences of allelic frequencies such as, for instance, between east and west Caucasian beechwoods (Paule et al., unpublished) were not found within European beech poulations (including Calabrian beech). F. sylvatica is considered a more recent species; however, there is no unanimity about the time when it split from the common ancestor (more similar to F. orientalis). Most authors assume that it is of Quaternary origin (Vernet, 1981) and that it was formed through selection owing to severe climate changes during glacial periods. There is a possibility for hybridization between F. sylvatica and F. orientalis. The ranges of both species are very close. Eastern beech occurs undoubtedly in East Bulgaria (Strandz a mountains, hills near Varna), whereas the limits of European beech distribution run through the eastern Stara Planina and Rodopi mountains. The distance separating the ranges of both species in eastern Bulgaria is not more than several kilometres. In fact, both species do not immediately touch each other, but pollen can surely be transported over such small distances. The existence of a continuous north-west to south-east clinal variation of morphological characters accompanied by similar pattern of genetic differentiation supports the hypothesis of hybridization: the European beech populations located near to the range of Eastern beech should be more affected by gene ow through pollen transport. In addition, a similar westeast cline over the Stara Planina mountain range was also observed. However, the Eastern beech populations from Thrace do not seem to be affected by gene ow in the opposite direction at all. They are even more differentiated from common beech than the populations from Asia Minor (Go mo ry et al., 1995). Theoretically, this pattern could be explained by asymmetric

J. EVOL. BIOL. 12 (1999) 746754 1999 BLACKWELL SCIENCE LTD

Genetic differentiation in Balkan beech

753

hybridization such as between sessile and pedunculate oaks (Bacilieri et al., 1996), but there is no experimental evidence for such a phenomenon in beeches. The patterns of genetic differentiation are mostly compared with and explained by the putative refugial areas during the last (Wu rm/Weichsel/New Drift) glaciation and the migratory routes in the Holocene. For European beech, the principal refugium is assumed to be located in the southern part of the Dinarian range (i.e. within the present range of the putative taxon F. moesiaca). Further refugial areas were situated in Italy, southern France and the southern Carpathians (Huntley & Birks, 1983; Horvat-Marolt, 1992). However, small sheltered beechwoods seem to be dispersed over ne valley, the whole of southern Europe Pyrenees, Rho Southern Alps at least during the late Wu rm (Vernet, ercelj, 1996). The question is, to what extent were 1981; S individual refugial populations differentiated? The assumption that all refugia from which beech expanded in the Holocene were independent or that they even had continuously been inhabited by beech since the Tertiary is hardly admissible. During the Pleistocene, beech very probably experienced several expansions during warm periods, similar to those in the Holocene, followed by retreats during cold periods. For example, in Slovenia it was present in the Tertiary, but no traces of beech pollen from Mindel and Riss glacials were identied. It emerges again in mid-Wu rm the beech charcoal found in ercelj, Palaeolithic settlements was dated to this era (S 1996). It is probable that only some of the Wu rm refugia preserved their continuity, whereas the other ones were occupied by migratory streams originating from the same source in previous interglacials. The major part of Europe was probably settled by beech originating either from one source or from several nondifferentiated refugia, since except for Italy and Balkans, the genetic differentiation monitored by isozyme loci (Comps et al., 1990, 1991; Go mo ry et al., 1992, 1995; and others) or by cpDNA markers (Demesure et al., 1996) is quite weak. A frequently presented hypothesis of the Holocene history of beech in Europe supposes that beech started to spread from the refugium located in the southern Dinarians to the north-west and, after reaching the southern slopes of the Alps, it expanded into the whole of Europe, mixing with the migratory streams from other secondary refugia (Comps et al., 1991; Hazler et al., 1997; and others). This hypothesis is based on fossil pollen maps published by Huntley & Birks (1983). However, the maps can be interpreted in a different way as well. Beech surely started to expand from the southern Dinarians and, at 9000 B P , this migration stream covered the whole former Yugoslavia. Some 500 years later, pollen values decreased again in the southern part of the Dinarian range and high values occurred on southern slopes of the Alps. There is no evidence of continuity between these two phenomena; the expansion at 8500 B P might have occurred also from local sources. This could explain the

differentiation between Balkan beech and the beechwoods in the major part of the range. Striking genetic similarity between Calabrian and Balkan beech, despite present geographical isolation, indicates the possibility of a common evolutionary origin. During the Pliocene, the present-day Apennine and Balkan Peninsulas were connected so that there might have been a possibility for migration and/or gene exchange between these populations. It should be stated that the Calabrian beechwoods, although displaying transitional gene frequencies between F. orientalis and F. sylvatica, could hardly be a product of introgression. If we admit that they constitute an evolutionary link between the Pliocene or early Pleistocene form of F. orientalis and the recent F. sylvatica, there is no reason to refuse this hypothesis for the case of the southern Balkan populations. The clinal pattern of genetic variation observed in our material might have originated also from subsequent gene exchange between Slovenian or Croatian common beech (more or less identical with beech in the major part of the range) and the transitional Balkan beech, whereby the most distant populations (Rodopi mountains, Stara Planina mountains) preserved their identity. This hypothesis is supported by highly signicant results of isolation-by-distance testing: the geographically extreme populations are less affected by gene ow than the populations in central regions (Central Dinarians, Pannonia). The nal question to be solved is the taxonomical status of Balkan beech. Unfortunately, the criteria for distinguishing species in the plant kingdom are very vague. The populations in this region can be distinguished from the remaining common beech by morphology, and they are genetically differentiated, so that they can be considered a separate taxon. Nevertheless, the rank of a separate species seems to be too high. There are other beechwoods, e.g. in Calabria, which are even more differentiated, but they are denoted as F. sylvatica. Therefore, the rank of a subspecies appears to be more appropriate for Balkan beech.

Acknowledgments
, We are grateful to Dr Yane Acevski, Dr Predrag Aleksic Prof. Aleksandar Andonoski, Prof. Vladimir Beus, Prof. Velichko Gagov, Prof. Sonja Horvat-Marolt, Dr Nikola , Dr Sas a Orlovic , and Ing. Sead Vojnikovic for Janjic their assistance with the procurement of material and organizing the stays in Zvolen, where the analyses were performed. We also wish to express our gratitude to Dr Yousry El-Kassaby, who checked the grammar of the paper. The study was supported by research grant No. GL-1032 from the Slovak Grant Agency for Science.

References
Bacilieri, R., Ducousso, A., Petit, R.J. & Kremer, A. 1996. Mating system and asymmetric hybridization in a mixed stand of European oaks. Evolution 50: 900908.

J. EVOL. BIOL. 12 (1999) 746754 1999 BLACKWELL SCIENCE LTD

754

MO RY ET AL. D. GO

res botaniques In: Le He tre Becker, M. 1981. Taxonomie et caracte (E. Teissier du Cros, ed.), pp. 3546. INRA, Paris. re, G., Merzeau, D. & Letouzey, J. 1987. La Comps, B., Barrie alloenzymatique des he traies dans les sous-dovariabilite maines medio- et eu-atlantiques d'Europe. Can. J. For. Res. 17: 10431049. baut, B., Paule, L., Merzeau, D. & Letouzey, J. Comps, B., Thie 1990. Allozymic variability in beechwoods (Fagus sylvatica L.) over central Europe: spatial differentiation among and within stands. Heredity 65: 407417. baut, B., Sugar, I., Trinajstic , I. & Plazibat, M. Comps, B., Thie 1991. Genetic variation of the Croatian beech stands (Fagus sylvatica L.): spatial differentiation in connection with the environment. Ann. Sci. For. 48: 1528. cia lis ci buko w: Fagus Czeczott, H. 1933. Studium nad zmiennos rednich. Cz. 1. Rocznik orientalis Lipsky, F. sylvatica L. i form pos Dendrologiczny 5: 45121. Demesure, B., Comps, B. & Petit, R.J. 1996. Chloroplast DNA phylogeograpgy of the common beech (Fagus sylvatica L.) in Europe. Evolution 50: 25152520. Diduch, Ya. P. 1992. Rastitel'nyj Pokrov Gornogo Kryma (Struktura, Dinamika, Evolyutsia i Okhrana). Naukova Dumka, Kiev. Fukarek, P. 1954. Neki osnovni podaci u vezi sa pitanjem bukve umar 78: 120. u Bosni i Hercegovini. Narodni S ny Go , J. 1998. mo ry, D., Shvadchak, I.M., Paule, L. & Vys Genetic diversity and differentiation of beech populations in Crimea. Russ. J. Genet. 34: 6370. ny baut, B. 1992. Go , J., Comps, B. & Thie mo ry, D., Vys Geographical patterns of genetic differentiation and diversity in European beech (Fagus sylvatica L.) populations in France. gia (Bratislava) 47: 571579. Biolo ny Go , J. & Paule, L. 1995. Genetic differentiation mo ry, D., Vys of populations in the transition zone between Fagus sylvatica L. & Fagus orientalis Lipsky. Forskningserien 11: 238244. Gower, J.C. 1966. Some distance properties of latent root and vector methods used in multivariate analysis. Biometrika 53: 325338. Gregorius, H.-R. 1980. The probability of losing an allele when diploid genotypes are sampled. Biometrics 36: 643652. ugar, I., Melovski, L., Tashev, A. & Grac an, Hazler, K., Comps, B., S J. 1997. Genetic structure of Fagus sylvatica L. populations in Southeastern Europe. Silvae Genet. 46: 229236. Huntley, B. & Birks, H.J.B. 1983. An Atlas of Past and Present Pollen Maps for Europe: 013000 Years Ago. Cambridge University Press, London.

Leonardi, S. & Menozzi, P. 1995. Genetic variability of Fagus sylvatica L. in Italy: the role of postglacial recolonization. Heredity 75: 3544. baut, B., Letouzey, J. Merzeau, D., Di Giusto, F., Comps, B., Thie & Cuguen, J. 1989. The allozyme variants of beech (Fagus sylvatica L.): inheritance and application to a study of the mating system. Silvae Genet. 38: 195201. ic , V. 1957. Varijabilitet i ekologija bukve u Jugoslaviji. Biol. Mis Inst. N. R. Srbije, Beograd. Mu ller-Starck, G. & Starke, R. 1993. Inheritance of isoenzymes in European beech (Fagus sylvatica L.). J. Hered. 84: 291296. Nei, M. 1978. Estimation of average heterozygosity and genetic distance from a small number of individuals. Genetics 89: 583 590. Raymond, M. & Rousset, F. 1995a. An exact test of population differentiation. Evolution 49: 12801283. Raymond, M. & Rousset, F. 1995b. GENEPOP (Version 1.2): population genetics software for exact tests and ecumenicism. J. Hered. 86: 248249. Rousset, F. 1997. Genetic differentiation and estimation of gene ow from F-statistics under isolation by distance. Genetics 145: 12191228. Rousset, F. & Raymond, M. 1995. Testing heterozygote excess and deciency. Genetics 140: 14131419. ercelj, A. 1996. The origins and development of forests in S Slovenia. Opera Acad. Sci. Art. Slovenicae, Classis IV: Historia Naturalis 35: 81142. nescu, V. 1979. Dendrologie. Universitatea din Brasov, Brasov. Sta Stoyanoff, N. 1932. The beech woods of the Balkan Peninsula. In: Die Buchenwa lder Europas (E. Ru bel, ed.), pp. 182219. Vlg. Hans Huber, Berne. Swofford, D.L. & Selander, R.B. 1981. BIOSYS-1: a Forstran program for the comprehensive analysis of electrophoretic data in population genetics and systematics. J. Hered. 72: 281 283. tre. In: Le He tre (E. Teissier du Vernet, J.-L. 1981. Histoire du he Cros, ed.), pp. 4958. INRA, Paris. Weir, B.S. & Cockerham, C.C. 1984. Estimating F-statistics for the analysis of population structure. Evolution 38: 1358 1370. Received 30 September 1998; revised 25 November 1998; accepted 13 January 1998

J. EVOL. BIOL. 12 (1999) 746754 1999 BLACKWELL SCIENCE LTD