FINAL CONSOLIDATED REPORT

Task1 Construction of a geographic map of genetic types

Subtask 1.1. Standardization of molecular techniques for analysis of chloroplast DNA polymorphism in oaks.

1.1.1. Molecular methods

            A technical workshop was organized at P1 during one week (March 11th to March 15 th  1996) at the beginning of the project. Ten participants from all the institutions involved in the cpDNA mapping project took part in the workshop. The molecular assay (PCR-RFLP), that had earlier been implemented by P1 (Demesure et al., 1995; Dumolin-Lapègue et al., 1997), was described and tested on a few material coming from the different labs. The protocols of the methods are described in the first annual report (March 96 to March 97), and are available on the website of the project (http://www.pierroton.inra.fr/Fairoak/). During the second coordination and technical meeting, that was held from March 13 th to March 21st 1997, the technical protocols were rediscussed according to the experience gained in the first year. The routine methods follow those of Dumolin-Lapègue et al. (1997), with some usually minor modifications across the 10 laboratories carrying out the molecular analyses. For DNA isolation, the methods either followed that of Dumolin et al. (1995) based on CTAB, or used the QIAGEN plant DNA kit. Then, four (mostly) non-coding cpDNA fragments (DT, AS, CD, TF) were studied, each with one restriction enzyme: DT-TaqI, AS-HinfI, CD-TaqI, and TF (either HinfI or AluI). For establishing the phylogenetic trees, the haplotypes identified during the survey were further characterised for two additional point mutations involving two more primers-enzyme combinations: DT-AluI and TF-CfoI. This latter polymorphism had been used in other cpDNA surveys in Europe (Ferris et al. 1993, 1998). The restriction fragments were separated by electrophoresis on 8% polyacrylamide gels as described in Dumolin et al. (1995), or in more resolutive system as described in Csaikl et al. (2000).

1.1.2. Genetic inventory at the continental scale

              A full inventory of cpDNA haplotypes was realised at the European scale. Whereas at the beginning of the project, it was planned to monitor cpDNA polymorphism in 1187 populations (see technical annex), these numbers increased during the course of the project to 2673, representing a total number of 12714 trees (Table 1.1). The increase in efforts consisted both in an increase of the density of populations per country, and in an increase of the number of countries. Whereas at the beginning 12 countries were concerned by the project, the final inventory is based on a systematic sampling in 22 countries (Figure 1.2). Many partners associated the financial support provided by the current project to other resources coming mainly from their own country. An additional partner joined the project during the last year. Finally other countries not yet covered by the inventory have started with their own means the monitoring of cpDNA polymorphism.

              The inventory represents an unprecedented achievement in population genetics. No other plant, nor tree species has received such attention for the description of its genetic diversity.

Table 1.1.  Sampled sizes used for the construction of the synthetic maps of cpDNA haplotypes

1.1.3. Haplotype identification

The haplotypes have been detected on the basis of the information provided by four PCR fragments, each digested by one restriction enzyme: DT-TaqI, AS-HinfI, CD-TaqI, and TF (either HinfI or TaqI). 25 of these haplotypes have been described before (22 + haplotype 23 = Q. cerris in Dumolin-Lapègue et al. 1997, 3 more in Dumolin-Lapègue et al. 1998). Seventeen haplotypes are described for the first time, along with the patterns corresponding to three other European oak species non included in the survey (Q. cerris [=haplotype 23 in Dumolin-Lapègue 1997, due to a problem of identification], Q. suber, Q. ilex). Four restriction-diagrams obtained in small acrylamide gels are provided for the 42 types (DT-TaqI, AS-HinfI, CD-TaqI, and TF-HinfI). Other haplotypes that have been discovered during the course of the project are not described here for one or more of the following reasons: (i) they occurred in only one locality, (ii) their detection was based on more resolutive method than those used by the majority of the participants, and they could not be reliably determined in small acrylamide gels, (iii) they had not been checked in laboratory of P1 in time to be incorporated in the present synthesis. Some participants did not analysed all fragments systematically, but only those that had been shown to discriminate among the haplotypes discovered in preliminary screening within a given region. Moreover, the resolution was not always as good as in the gels used to prepare the restriction-diagrams presented here. Finally, two alternative restriction enzymes had been used in the analysis of fragment TF (HinfI and TaqI). As a consequence, some haplotypes shown in the diagrams are likely to be mistaken one for the other by some participants; in such a case, they were given subscripts (e.g., haplotypes 5a, 5b, and 5c) solely to indicate this. All haplotypes are presented in the phylogenetic tree (Figure 1.1), but distribution maps will be presented only for the ‘merged’ haplotypes (i.e., a map will be presented for haplotype 5 (=haplotypes 5a + 5b +5 c), rather than three separate maps; hence, the color code is the same for these haplotypes (Annexe 1.7 to 1.22).  In some cases, within regional papers, the distribution of these haplotypes may be given. The remaining haplotypes would have been detected by all participants if  they had happened to be present among their  samples: at least one fragment (or combination of fragments) of these haplotypes allows their clear identification. Nevertheless, a bias in favor of the most frequent haplotypes is possible. Additional fragments not considered systematically or corresponding to levels difficult to differentiate during a systematic and large-scale study sometimes provide additional characters useful to identify their relationships with the other haplotypes. For instance, fragment AS6 (which is highly polymorphic and may harbor cpSSRs) was not scored by most groups but is represented here to provide a complete description of the haplotypes and to improve the characterization of the haplotypes for establishing the phylogenetic tree.

Following these protocols,  in total 42 haplotypes were found during the survey that comprised 12714 trees.

1.1.4. Phylogenetic relationships between the haplotypes

Among the 42 haplotypes, 67 length variants were detected in 24 polymorphic restriction fragments (i.e., a mean of 2.8 length variants per restriction fragment). The inclusion of three other species used as outgroups (Q. cerris, Q. ilex and Q. suber) raised the number of length variants to 75. In total, they were 57 phylogenetically informative characters (shared by at least two haplotypes) and 10 autapomorphies (variant found in a single haplotype) within the white oak group. There were two more autapomorphies present in Q. ilex and one for Q. suber. Five characters were unique to both Q. suber and Q. ilex. The results obtained using the Fitch and the Kitsch phylogenetic analyses were largely consistent and clearly showed that the 42 haplotypes could be clustered in six lineages (Figure 1.1).  Lineage C (haplotypes 1, 2 and 3, located mainly from Italy to Scandinavia, Figure 1.3, Annexe 1.7 and 1.8) ) includes two divergent group of haplotypes and is therefore poorly resolved. It was slightly better individualised in the Neighbour-joining analysis obtained by Dumolin-Lapègue et al. (1997), where other fragments had been studied. On the other hand, lineage B (haplotypes located mostly in western Europe: 10,11,12, 24, 25 and 32, , Figure 1.3 and Annexe 1.13, 1.14, 1.15, 1.21) and lineage A (haplotypes 4-7, 26, and 30-31, Figure 1.3 and Annexe 1.9, 1.10, 1.11, 1.12, 1.22) are clearly identified in both analyses, although not as well as in the previous study in the case of the A lineage. The two algorithms give a somewhat different picture for the remaining haplotypes. Among them, a first group of haplotypes can be distinguished, particularly when using the Kitsch procedure. It includes haplotypes found around the western part of the Mediterranean region: from eastern Spain (27-29, 33, Annexe 1.20 ), to south-western France (21, Annexe 1.20), and Algeria (22, Annexe 1.20). We will call it thereafter lineage D (Figure 1.3). Among the remaining haplotypes, number 8 and 9 form a separate group which was well supported by bootstrap values in Dumolin-Lapègue et al. (1997); we will call it lineage F. These two haplotypes are found near the Black sea. The remaining haplotypes (13-20, Annexe 1.16, 1.17, 1.18, 1.19) have a relatively basal location in the Fitch tree, between lineages B and D. However, they are grouped by the Kitsch method, though haplotype 14 is somewhat more divergent and haplotypes 17d and 20 are in an intermediate position with lineage D. This lineage (called E, Figure 1.3) is therefore likely to be provisory. However, all haplotypes belonging to this group have a more eastern distribution (from Italy to Rumania, up to Turkey) (Figure 1.3 ).

Subtask 1.2. and 1.3.  CpDNA analysis of populations established in provenance tests and  sampled on a grid system

1.2.1. Geographic distribution of cpDNA diversity

There were 2,264 populations where more than three individuals had been sampled, which have been used for estimating the diversity statistics. These populations can be split in 6 species and 7 geographic regions. We used these subdivisions to calculate the diversity measures. The frequencies of the haplotypes as well as the distances between them (number of different restriction fragments) were used to compute diversity and differentiation measures, following Pons and Petit (1996), using softwares available at http://www.pierroton.inra.fr/genetics/labo/Software/. All measures of diversity as well as their standard errors were computed by taking into account (v) or not (h) the distances between haplotypes. The resulting coefficients of differentiation are called N_ST (for v) and G_ST (for h), and can be directly compared using the U_NG  test. The overall total diversity was 0.848, and the coefficient G_ST was 0.828, significantly lower than the N_ST (0.876) (U_NG = 13.1, P<0.001; see Table 4).

1.2.1.1 Variation across species.

Measures of diversity: For both within- and total-population diversity, highly significant differences were found between h-type and v-type measures, the latter being in most cases lower (Table 1.2). This witnesses the strong phylogeographic component of the genetic structure, in these oaks: although they share many haplotypes when in sympatry, they have different distribution ranges, and therefore do not have a balanced representation of all existing chloroplast lineages. Indeed, these lineages are not distributed at random in Europe (Figure 1.2.). Not unexpectedly, the three most abundant and widespread oak species in Europe (Q. robur, Q. petraea, and Q. pubescens) have the highest levels of total chloroplast diversity, as measured by h_T. The only significant difference is between these three species and Q. pyrenaica, which is characterised by a lower diversity (h_T = 0.67, i.e. about 80% of that of the other species).

Table 1.2.  Levels of diversity and differentiation by species

Coefficients of differentiation: G_ST varies from a low 0.784 in Q. robur to a high 0.962 in Q. pyrenaica, whereas N_ST values are always higher than the corresponding G_ST values, as a consequence of this phylogeographic structure. The N_ST values are therefore more similar among species, from 0.846 in Q. robur to 0.971 in Q. pyrenaica. In three cases the difference between the two coefficients of differentiation is significant: for Q. robur, Q. petraea, and Q. pubescens, the largest difference being observed in Q. robur. The ranking across species is the same for G_ST and N_ST: Q. robur < Q. frainetto < Q. petraea < Q. pubescens < Q. faginea < Q. pyrenaica.

1.2.1.2. Variation across regions.

Haplotype  richness: Among the eight regions compared, the observed number of haplotypes varies from five in Great Britain to 11 in the Balkans or 12 in Spain (Table 1.3).  However, some of these haplotypes are very rare or of dubious origin. For instance, the number of autochthonous haplotypes in Great-Britain is more likely three than five. Moreover, sample sizes vary across regions (from 855 trees in Scandinavia to 3,587 in France). To obtain a more representative idea of the haplotype richness, it was standardised to a common sample size of 100 trees. The estimates now vary from 4.2 in Great Britain, 5.1 in Switzerland/Austria, to 9.7 in Spain and 9.9 in the Balkans.

Genetic diversity: As in the comparison across species, the v-type measures are generally (often highly significantly) lower than their h-type equivalents. The only exceptions are the Alpine region, and central Europe, which have a mixture of divergent haplotypes belonging to lineages A and C in the first case, A, B and C in the second one (Table 1.3). The lowest levels of total diversity (h_T) are again observed in Great Britain and in the Alpine region (Switzerland and Austria); however, contrary to what was found in the analysis based on allelic richness, the lowest value is observed for the Alpine region, where two haplotypes only (1 and 7) predominate, whereas there are three abundant haplotypes in Great Britain. The most diverse regions based on this parameter are again the Iberian peninsula and the Northern Balkans, this time followed by Italy.

Table 1.3. Levels of diversity and differentiation by region.

Coefficients of genetic differentiation: The differentiation G_ST among the eight regions is 0.18; it is however significantly higher (U_NG = 2.0, P<0.05) when taking into account phylogenetic relationships: N_ST = 0.29. This means that the differentiation among regions is a significant component of the overall differentiation among populations. This overall phylogeographic effect, although quite reduced, is still discernible at the within region level. Indeed, in seven of the eight regions, N_ST is larger than G_ST, and the differences are significant in four cases. It is in France that the estimates differ most (by 0.08); relatively large differences are also present in the Balkans, Scandinavia and Spain, whereas the Alpine region and Great Britain present no indication of a phylogeographic component of the genetic structure; this was expected given their genetic composition, as there are two dominant haplotypes in the Alpine region and three equally divergent haplotypes in Great Britain (and hence no way to detect a phylogeographic structure). In the other regions, the potentialities to detect a difference between both measures are higher; in France, the three closely related haplotypes (10, 11 and 12) of the B lineage have a similar distribution and are comparatively often found in the same populations. Hence, mixtures of related haplotypes will be much more frequent than mixtures of more divergent haplotypes, which should account for the observed difference between both differentiation  coefficients. The Iberian and the Italian peninsula, as well as Scandinavia, partition a higher proportion of their diversity among populations (as judged by both G_ST and N_ST).

1.2.2. Geographic distribution of cpDNA haplotypes

Maps of cpDNA haplotypes  were produced using the software MapInfo Professional version 3.5 (Figure 1.1, 1.2 and Annexes 1.1 to 1.22). Symbols of different sizes were used to indicate the level of reliance in the authochthony of each population. The largest symbols indicate populations fixed for a given haplotype; smaller symbols were used for populations represented by a single haplotype; still smaller symbols were used for indicating populations that comprised other haplotypes than the one considered; finally, populations considered a priori of dubious autochthony were represented with the smallest symbols.

1.2.2.1. Haplotypes originating from the Iberian peninsula

The four chloroplast lineages recognised so far in Europe  are all represented in the Iberian peninsula (Figure 1.2). However, two of them are represented by a single haplotype (lineage C with haplotype 1, and lineage A with haplotype 7) in the north of the country. On the other hand, lineages B and D are largely represented within the Iberian peninsula (Goicoechea et al. 2000; Petit et al. 2000). These two lineages have very distinct, non-overlapping distributions. Lineage B occurs to the west of a line running roughly from Cádiz to Navarra, whereas lineage D is found exclusively east of this line. The first one appears to have considerably expanded throughout western Europe, reaching Scandinavia, northern Poland and northern Scotland, whereas the other lineage remained trapped in eastern Spain and did not reached the Pyrenees. Compared to previous investigations (notably Dumolin-Lapègue et al. 1997), the recognition of lineage D, which had not been detected before in the Iberian peninsula, represents a major biogeographic finding. It points to the existence of at least two major refugia in Spain. Moreover, the existence of oak species endemic to the Iberian peninsula and to North Africa (Q. faginea, Q. canariensis) or shared only with western France (Q. pyrenaica) point clearly to a wide and diversified refugial zone in this part of Europe.

Lineage B includes three abundant haplotypes (10-11-12, Annexe 1.14, 1.15 and 1.16) and a few less frequent ones. Among the latter, the newly recognised haplotype 25 (Annexe 1.21) is common to southern Andalusia and to the Rif region in Morocco. The southernmost location of haplotype 10 is in southern Portugal, but this very frequent haplotype then extends over most of Portugal, a large part of western Spain and France, Great-Britain, Benelux, north and west Germany, and part of Scandinavia. The distribution of this haplotype is strikingly similar to that of haplotypes 11 and 12, although these three haplotypes are usually exclusive of each other in smaller regions. The fact that all three haplotypes readily colonised  Great Britain and Ireland and were not lost during the spread northward throughout France, Germany and up to Scandinavia indicates that at least for those haplotypes that are frequent enough, the mechanism of recolonisation does not induce a rapid loss of diversity.

            The absence of trees characterised by haplotype 12 below the river Douro (northern Portugal) could indicate that the refugium where this  haplotype was restricted during the last ice-age was located at much higher latitudes compared to the inferred distribution of oaks during this period (Brewer et al., 2000, Huntley and Birks 1983). However, it is possible that populations characterised by this haplotype were present further south and have disappeared or have been overlooked. The distribution of haplotype 11 in Spain (Annexe 1.14) illustrates this possibility: if the four populations located in western Andalusia (Sierra Morena) had been overlooked, the southern limit for this haplotype would now lie in northern Spain. Compared with that of haplotype 10, the distribution of haplotype 12 is slightly more westerly. This haplotype is indeed particularly abundant in Galicia, in the Cordillera Cantabrica and in western France (especially in Brittany and Normandy). In Great Britain, it is also slightly more abundant westward (in Cornwall, Wales and along the west coast of Scotland), although it is by no way restricted to these regions. Haplotype 11 is present in Norway, but in a single population along the south-eastern coast, so it would be worth to check whether it is autochthonous to this area by further sampling in the same forest. On the other hand, there is good evidence for the migration of haplotypes 10 into Scandinavia, again along the south-eastern coast of Norway; there is also some scattered evidence of the presence of this haplotype in Sweden. The abundance of haplotype 10 in northern Denmark may explain this pattern.

            This lineage (B) includes another haplotype of intermediate abundance, restricted to the Iberian peninsula and to France: haplotype 24 (Annexe 1.21). It is also the most basal haplotype within that lineage. The migration route that can be traced from the distribution of this haplotype is similar to that of haplotypes 10-11-12. There are 3 more haplotypes of lineage B that have been detected in the Iberian peninsula (10b, 12b, and 34), each related to the three most abundant haplotypes (10a, 12a and 11, respectively).

1.2.2.2. Haplotypes originating from the Italian peninsula.

The cpDNA haplotypes detected in this country belong to three of the four major lineages (i.e., all except lineage B). As can be seen in Figure 1.2 and 1.3, lineage A (in which we have distinguished four haplotypes) could have been restricted to the Italian refugia during the last ice-age, whereas the situation is more complex for haplotypes belonging to lineages C and D which include haplotypes shared by other refugial areas, especially in the Balkans.

            Among the four haplotypes belonging to lineage A, all except one (32) are present in Italy. The southernmost location of haplotype 1 (Annexe 1.7) is in Sicily, where it is abundant, especially in the east; this haplotype is also found along the Apennines and in Corsica and Sicily; it is completely absent from the Balkan peninsula but was detected on the south-eastern slopes of the Pyrenees, down to the Barcelona region. Haplotype 2 (Annexe 1.8) was also found in many populations in Sicily (to which the related haplotype 3 (Annexe 1.8) is restricted). There is then an interesting disjunction since haplotype 2 is no longer found in southern Italy but appears again slightly north of Roma (i.e., over 600 km further north). As discussed above, intermediate populations may have disappeared or may have been overlooked (although the sampling is quite intensive in this area). It would be interesting to check that we are dealing with exactly the same haplotype rather than with two closely related haplotypes (one in Sicily and the other in north Italy), by looking at additional cpDNA fragments. This haplotype is then found more or less continuously from Croatia to Austria, Hungary and Slovakia, with more isolated occurrences in Poland and Lithuania.   

There are two other frequent haplotypes that have been found in Italy, one belonging to the C lineage (haplotype 5, Annexe 1.10) and the other belonging to the D lineage (haplotype 17, Annexe 1.19). Actually, we know that these haplotypes include several molecular types that could not be systematically resolved during this survey, making the interpretations more difficult (haplotypes 5a,b,c and haplotypes 17a,b,c). In any case, their wide distribution from Italy throughout the east of the Balkan peninsula points to some important exchanges between these regions, although some of these exchanges may have been anterior to the last postglacial.

1.2.2.3. Haplotypes originating from the Balkan peninsula

 Haplotype 7 (Annexe 1.12), although present in northern Italy, is absent from the south of the Italian peninsula (except in two isolated populations where it is mixed with other haplotypes, suggesting recent introduction). It is also abundant in the north of Croatia, but is then absent further south. Contrary to the previous haplotypes, however, it does not extend very far to the east (the two points in Rumania could represent contamination through plantation). Haplotype 7 has a large distribution in western Europe. This may be explained by an early colonisation due to a relatively northern refugium for this haplotype (for instance in Slovenia or in northern Croatia). A related haplotype (n°26) has been described earlier (Dumolin-Lapègue et al. 1998), in the French Alps. It could represent a recent, post-colonisation mutation event (see previous paper).

            Haplotype 4 (Annexe 1.9) is represented in the southern part of the investigated range in two separate groups of populations, one in the south-east of the Carpathian mountains in Rumania, and the other in the north of Hungary. If we opt for a refugia in the eastern Balkans for this haplotype (for instance somewhere in Bulgaria), then the most likely route would be around the northern side of the Carpathian mountains, from east to west, reaching the eastern side of Poland, from where further migration would have occurred towards the north, up to the Baltic countries, and towards the west, to reach eastern Germany.  The distribution of haplotype 6 (annexe 1.11) seems to point to a central Balkanic refugium (for instance in southern Serbia or further south in Macedonia), leading to the invasion of western Rumania and Hungary,  and then eastern Austria, but many other scenario are possible due to the lack of data in the southern Balkans.

             Haplotype 13 (annexe 1.16) is found only in Rumania, then in Crimea (Ukrainia) and further east in Russia, along the coast of the Black sea. The absence of this haplotype west of Rumania (in Hungary or Croatia) supports the view of deciduous oak refugia in south-eastern Balkans, at a place which remains to be identified, for instance along the western coast of the Black sea. By analogy to the colonisation routes inferred for haplotype 6, we suggest that the distribution of haplotypes 14, 15 and 16 (Annexe 1.16, 1.17 and 1.18) in populations located from Rumania to Russia also witnesses eastward movements of populations initially located in an eastern Balkan refugium (rather than a movement from east to west). All of them are located south of the Carpathian mountains in Rumania.

            Two related haplotypes (8 and 9) have been found east and north (Crimea) of the Black sea, pointing to other oak refugia in these areas (Dumolin-Lapègue et al. 1997). The sole population from Turkey (south of the Black sea) that was included presents also a distinct haplotype (n° 18).

Subtask 1.4 Analysis of pollen deposits

1.4.1. Pollen Analysis

Pollen analysis has been well-established as a means for reconstructing vegetational history, and the period between the end of the last glacial period and the present day is well documented (Berglund et al., 1996). The pollen grains and spores produced by plants preserve well in anoxic environments, e.g. lake sediments, peat bogs. They may be extracted from these deposits by sampling exposed sections, or more commonly, by taking sediment cores. The resulting sequences are subsampled along their length, to provide a chronologically ordered set of samples.

1.4.2. Chronologies

The principle form of independent dating used for pollen sequences is radiocarbon measurements on the sediments or, preferably, plant remains derived from the same fossil record. A number of measurements are made throughout the core, and the chronology is obtained by interpolation between these dates, giving a date for each sample. There are, however, a number of problems in using radiocarbon dates, mainly contamination (see Bowman, 1990 or Lowe and Walker, 1997 ch. 5 for more details), which must be taken into account when constructing a chronology. Also, it should be stressed that ages expressed in ¹⁴C years, which are a measure of the ratio of ¹⁴C/¹²C, are not equal to calendar ages due to variations in the atmospheric production of ¹⁴C (Stuiver et al. 1998). However, sequences with some form of objective dates are preferable when studying time-dependent phenomena, such as the migration of a tree species. It is worth noting that with a few exceptions in areas where no dated information was available, all sequences used in this project have chronologies based on independent dates. 

1.4.3. Database of Pollen Sequences

The European Pollen Database (E.P.D.) was established in 1991 to archive, in a relational database, the original data from pollen analyses performed across Europe. Of the 875 sequences which are currently held in the EPD, 483 have chronologies based on radiocarbon dates, as described above. These formed the basis of the data set used to map the oak pollen percentages. To this data set, information was added from fifty sequences held in the Alpine Palynological Database (ALPADABA), covering Switzerland and Austria, plus several sequences supplied for use in this project. Finally, in areas where no original data was available for use in this project, the percentages of Quercus were digitised directly from the published pollen diagrams. The digitising method used to recover the data was tested on sites where the original counts are available and the accuracy is greater than 99% (Brewer unpubl.). In all, approximately 600 sequences were used to build the time slice maps.

1.4.4. Data extraction and mapping

Oak pollen percentages were calculated for every dated sample based on a pollen sum of trees, shrubs and herbaceous plants. The set of sample percentages obtained have been plotted as a series of maps showing the distribution of the oak pollen percentages for a series of time-slices between 15 ka BP and 6 ka BP (Figure 1.4).

In order to summarise the data, a single map (figure 1.5) was produced showing isochrones for the arrival of Quercus across the European continent (see Birks, 1989 for a review of the method). Arrival times were estimated using the rational limit, the age corresponding to the rise to high percentages (Smith and Pilcher, 1973). In some areas, these were supplemented using diagrams available in the literature. The resulting set of times was interpolated onto a 15 arc-minute grid using ordinary kriging (Deutsch and Journel, 1998), to show the progressive migration of the deciduous oak across the European continent, and contoured to produce the isochrone map (Figure 1.5).

1.4.5. Location of Glacial Refugia

The refugia for deciduous Quercus are most easily identified from the single map dated to the end of the glacial period at 15 ka BP (fig 1.4). The isochrone map (fig 1.5) further indicates that the source populations for the modern day forest were restricted to these sites where local mild condition prevailed. These maps identify three possible regions of Europe that acted as main refugia for the oak during the last glacial period:

1.      southern Iberian peninsula (type site Padul (Pons and Reille, 1988))

2.      central Italy (type site Laghi di Monticchio (Watts et al., 1996))

3.      the southern Balkan peninsula (type sites Ioannina  and the Black Sea (Shopov et al., 1992))

The Balkan refugium contains two refugial areas, Greece and the western coast of the Black Sea. This second refugia is not shown on the map at 15 ka BP, but is represented on the isochrone map. At the site in south-west Turkey which shows a presence of approximately 20% oak pollen in the sediments at this time, the pollen has been identified as that of Quercus cerris (Bottema and Van Zeist 1991). It is not therefore considered as a refugium for the species studied in this project (Q. robur and  Q. petraea).

All refugia type sites are located in or near mountainous areas: the Sierra Nevada in southern Spain, the central Italian mountains and the Pindos mountains in Greece. This supports the idea proposed by Beug (1975) that the refugia for the deciduous trees would have been located at mid-altitude sites, where the precipitation would have been higher than on the plains during the arid glacial period.

Previous work on the location of glacial refugia of the deciduous oak has indicated eight possible refugia (Huntley and Birks, 1983; Bennett et al., 1991). The inclusion of new information and the extension of the mapping into the glacial period has enabled us to confirm half of these locations, and more clearly define the position of these refugia. No evidence was found for either the refugia proposed in the northern Spain, southern France, or in Alps and northern Balkans (Huntley and Birks, 1983). The zone of identified refugia is constrained to the extreme south of the continent.

1.4.6. Migration

This first step was a spread from the glacial refugia (see above) northwards across southern Europe. During the period between 13 and 11 ka BP, the NAPF (North Atlantic Polar Front) retreated northwards as far as Iceland (Ruddiman and MacIntyre, 1981), with increased temperatures and moisture availability across the European continent. At the same time, there is an increase in the percentages recorded of oak pollen in sites south of the main European mountain ranges, concurrent with new occurrences in these regions. Notably, there is a strong increase throughout Italy from 12.5 ka BP, and on the western Iberian coast from 11.5 ka BP. However, any corresponding expansion on the eastern Adriatic coast is obscured due to a lack of sites.

In Italy, the oak appears to have migrated both north and south from the central refugia. In the north, the oak spread rapidly along the Apennine chain, reaching the north-west between 12.5 and 12 ka BP (Lowe et al, 1996). From here, the oak spread west beyond the Alps along the Mediterranean coast by 11 ka BP (Nicol-Pichard, 1987). The dispersal into the north-east was slower, reaching the southern Alps by the end of the lateglacial interstadial (Schneider and Tobolski, 1985). As there is no information available from the extreme south of the peninsula, the possibility exists of a refuge in this area. However, the dynamic of the migration suggests that the spread did not start there. In the Italian peninsula, the high percentages of deciduous oak pollen suggest that it formed a dominant part of the vegetation assemblage during the late-glacial period.

In the Iberian peninsula, this spread was equally rapid, reaching the region of Galicia in the north-west  by 12 ka BP. By 11.5 ka BP, there is a significant presence along the west coast, and south of the Pyrenees, and by the end of the lateglacial interstadial, oak pollen is recorded in sites in both the west and east Pyrenees. It is, however, difficult to discern the pattern of migration, due to the lack of sites in the centre and south-west of the Iberian peninsula. The dispersal may have been fan-shaped, spreading out across the peninsula from the south-east, or may have followed a south-north path along the coasts. This second hypothesis gives rise to the possibility of a second glacial refuge in the south-west. The site of El Asperillo in the south-west of Spain, (Stevenson, 1984), indicates a possible regional presence of Quercus at the start of the lateglacial interstadial period (ca 13 ka BP). However, the date attributed to this pollen deposit is too imprecise to confirm or reject this hypothesis. However, the oak did not spread across the entire peninsula at this time. At a site on the east coast of Spain, oak pollen is not recorded before 6 ka BP, some 5000 years after it was recorded further north in the Pyrenees (Carrion and Dupré, 1996). This may indicate that the migration along the Mediterranean coast was blocked, possibly by the presence of a pre-established Pinus forest (Carrion and Dupré, 1996).

In contrast to the expansion of oak seen at sites on the Iberian and Italian peninsulas, there is little observed increase in the distribution range at the Greek sites. Bottema (1979) suggests that the expansion in this area was limited by the lack of available precipitation. As stated above, the interpretation of the vegetation history of the Balkans is made more difficult by the lack of sites in the central region, which cover the lateglacial interstadial. Two sites on the Adriatic coast, the island of Mljet (Beug, 1960) and Vid (Brande, 1973) both show a strong presence of oak from the early Holocene onwards, but no inference can be made from these about a potential occurrence in the lateglacial. However, a new dated diagram from Taul Zanogotti (Farcas et al, 1999) in the west of in the Carpathian chain indicates a low but regular occurrence of deciduous Quercus pollen in the lateglacial interstadial. There are also several sites that give evidence from the lateglacial in the north of the Balkan peninsula. The extensively studied Ljublana moor in Slovenia indicates an arrival of Quercus toward the end of the lateglacial interstadial (Culiberg, 1991). In the absence of any direct evidence from the central Balkans, our interpretation is based on this evidence and a latitudinal comparison across Europe: between 13 and 11ka BP, the oak spread northwards from its southern glacial refugia, to reach the Carpathian chain, and the south-eastern Alps.

Little change in distribution is seen between 11 and 10 ka BP: there is no significant outward spread, and the values in southern Europe decline. This period corresponds to a dry, cold period, termed the Younger Dryas (YD), and an advance of the NAPF to the level of northwest Spain (approx 43°N, Ruddiman and MacIntyre, 1981). In pollen records across Europe, the impact of this period can be seen as a decline in the mesophilous taxa, and an increase in steppic elements (e.g. Artemsia). The reaction of Quercus to this climatic change appears to vary between sites. At some sites, Quercus pollen is no longer deposited during this period, whilst at other sites there is continuous but reduced deposition of pollen. This reduction is more pronounced in the northern parts of the range. The sites with a contiuned presence of Quercus pollen appear to act as temporary, secondary refugia, enabling Quercus to remain further north during this relatively short cold period, than in the full glacial period. However, it is possible that the impact of the YD is not fully represented on the map at 10.5 ka BP. The 500-year time interval may therefore represent the limit, imposed by the problems described above, of the temporal resolution possible for mapping pollen data at this scale.

After approx. 10ka BP, there was a shift in the climate from the cold, dry conditions to warmer conditions (Huntley and Prentice, 1993). This warming coincided with a shift northwards of the North Atlantic Polar Front to a position north of Iceland (approx 65° N, Ruddiman and MacIntyre 1981), and a resulting increase in moisture availability across Europe.

Coupled with the increased summer insolation, the period 10-9ka BP was an optimal period for the expansion of mesophilious tree species. From the beginning of the Holocene, Quercus spread rapidly north and west into France, reaching the south of Ireland and England between 9.5 and 9ka BP. However, the Alps played a major role in slowing the spread of oak northward into central Europe. The Vosges mountains may have also prevented the effects of the Gulf Stream from reaching southern Germany. These geological barriers explain the much later appearance of oak to the north of the Alps. The maps indicate the presence of oak south of the Alps as early as 11.5 ka BP while the presence on the northern slopes is dated around 9 ka BP. Once the oak has reached the southern part of the central Europe, around 9 ka BP, its spread was much faster, reaching southern Scandinavia in less than 1000 years. To the east of the Alps, the oak had reached northern Belorussia by 8 ka BP. However, to the east the dispersal is patchy, in contrast to the relatively smooth spread in the west, indicating that the expansion was limited by a secondary factor, possibly competition. By approximately 6 ka BP, however, the oak had filled the majority of its modern-day range.

The rapidity of the dispersal into northern Europe makes it hard to define the dynamics of migration in the Holocene. Three possible migratory pathways may be identified from the isochrone map: western, central and eastern. The west path follows the French atlantic coast from the Pyrenees to the British Isles. In the east, a path can be discerned spreading north from the Carpathian mountains into the region east of the Baltic sea. In central Europe, the situation is unclear: the dispersal into Germany, and ultimately Scandinavia may have come from the Pyrenees, the central  or south-eastern Alps, or from a combination of all three.

The complete late and postglacial migration of deciduous Quercus can be seen more clearly on the summary map (figure 1.5), which shows the spread from the southern refugia to the northern mountains, and then rapidly into northern Europe with the start of the current warm period. It is important to note that the area bounded by the contours on this map (figure 1.5) do not indicate the presence of oak forest. They should be interpreted as the areas indicated as suitable by the interpolation to have contained small populations of oak.

1.4.7.  Migration pathways inferred from cpDNA and fossil pollen analysis

The distribution of haplotypes (Figure 1.2 and 1.3) and the palynological information available (Figure 1.4 and 1.5)  were used to infer colonisation routes out of the ice-age refugia (Figure 1.6). In western Europe, in particular, clear routes out of the Iberian and the Italian peninsula can be traced. Haplotype lineage B clearly witnesses northwards migration from the Iberian peninsula (yellow and orange arrows on figure 1.6); similarly lineage C strongly suggests movements from Italy (red arrows). Some movements resulted in the exchange of haplotypes among separate refugia (especially between the Balkans and the Italian peninsula, lineage A, blue and green arrows on figure 1.6) during the present interglacial and probably also during earlier cycles. This has lead to parallel colonisation routes for phylogenetically divergent haplotypes, and thereby some obscuring of the phylogeographic structure. Cases of disjunction in the present-day distribution of haplotypes are also apparent and could have been induced by the existence of a cold phase (the Younger Dryas, from 11,000 to 10,000 BP), which resulted in range restriction following an early warm period which lead to the first expansion of the oak from its refugia. This was followed by a new period of expansion in the postglacial, in some cases involving ‘secondary’ refugia. All this would have led to a reassortment of the distribution of the haplotypes. Early association between haplotypes and oak species is also suggested by the data, although extensive introgression among species has partly obscured the pattern. This means that colonisation routes may have been initially constrained by the ecological preferences of the species hosting each chloroplast variant. We suggest for instance that two oak species distributed in the north of the Iberian peninsula (Quercus petraea and Q. robur) are recent postglacial immigrants. Altogether, the combination of molecular data and palynological information allows much more precise reconstructions than those based on only one method.

Task 2 Provenance evaluation in experimental plantations

Subtask 2.1 Database of oak provenances in Europe.

            During the first cordination meeting, decisions were taken on the type of information that should be included in the data base. A set of seven files were identified: PROVENANCE, CYTOPROV, CYTOGRID, INSTITUTE, TEST, PROVECO, TESTECO. These files contain passport information (geographic location) and ecological data  of provenances and test sites. In total 376 provenances, growing on 62 test sites are recorded in the data base.

Subtask 2.2. Joint evaluation of the S. Madsen range wide provenance test

The basis for the work of task 2.2 was the international 1989 seed collection in Quercus petraea (Matt.) Liebl.stands, that comprised in total 19 different provenances  across the natural area of distribution. From the project it has become clear that this basic material was used for the establishment of 27 different provenance field experiments across Europe and Asia Minor, mainly during the years 1992-1993. The main part of the task 2.2 work has been concentrated on the joint evaluation of 13 of these experiments, here of three Danish, seven German, two British, and one French experiment.

The aim of the research work has been to i) present the total series of provenance field experiments established; ii) describe the variation in different tree character values found by a joint analysis within the group of 13 experiments being included in the EU-research project; iii) form a basis for future analyses of the series; iv) contribute to the general knowledge on variation of leaf nutrient concentrations between sites (i.e. field experiments); v) contribute to the general knowledge on variation of nutrient concentrations between provenances; vi) from literature, to get an idea whether a deficiency of nutrients in some of the field experiments might be assumed; and, vii) analyze the effect of nutrient concentration on present height and survival in the field experiments.

Table 2. 1. Field experiments holding material from the international 1989 Quercus petraea (Matt.) Liebl. seed collections.

2.2.1. Field evaluation for phenotypic traits

According to the guidelines worked out during the coordination meetings held in 1996 and 1997  a joint data collection program was carried out during 1998. The data sets consist partly of plot character values of plant loss, stand height, height increment, flushing, growth cessation, tree form etc., further of leaf nutrient concentrations found per plot for three specific provenances, and finally of geographic, historical, and environmental information per field experiment. These data are supplementary to the general information already given in the data sets TEST and PROVENANCE, and they have been organized in the different data files. The field evaluation for phenotypic traits is based on the international 1989-series of sessile oak provenance experiments, representing in total 27 experiments across Europe and Asia Minor (Table 2.1). Further, in the report, various sections of thirteen of these experiments has been analyzed for variation of early plant loss 1993-1995, frequency of live trees of oaks and other tree species in the autumn 1997, heights for various groups of experiments in the autumn 1997, time of flushing in the spring of 1997, and the rate of height growth before and after August 1, 1997. Generally, for all traits, significant effects were found for experiment, blocks within experiment, and provenance sources of variation. However, significant interaction effects between experiments and provenances were also found for many traits. For each trait has been calculated an lsmean value for each experiment, assuming the existence of all provenances on all sites, and for each provenance has been calculated a best linear unbiased predicted (blup) value, reflecting the result to be expected for the provenance on the same site under the same climatic conditions, but using other random designs. Analyses of the general effect of provenances in eleven west-European experiments showed the highest plant losses for the south-eastern provenances (Hungary, Turkey). The fastest height growth was found for two British and two German provenances (Table 2.2), and the difference between the earliest and the latest provenance in time of flushing was not more than 4 days. The lsmean values of the experimental sites did show no clinal structures for survival or height, whereas a north-south trend was very clear for the date of 50 % flushing, which in 1997 occurred in France on April 11, but 18-30 days later in various British and German sites, and 40 days later in the Danish experiment. Analyses of pairs of experiments in Germany, France and Denmark did reveal two different zones of provenance performance, i) an Atlantic zone including France, western Germany and Denmark, and ii) a Continental zone including central and eastern Germany. Within each of these zones, for all pairs of field experiments, was found nearly perfect correlations of height growth between west European provenances, whereas the correlations for pairs of field experiments in different zones were weaker. This means that provenance height ranks found in a field experiment in a specific zone have shown valid also for other sites within the zone concerned.

2.2.2. Field evaluation for foliar nutrient concentrations

Foliar nutrient concentrations were analysed for the thirteen elements C, N, P, K, Mg, Ca, Na, S, Fe, Mn, Zn, Cu, Al, and for the three ratio’s N/P, N/K, and K/P as well. The analyses were based on leaf collections from three common provenances troughout 13 field experiments (Table 2.1) Pooled analyses of the foliar concentrations showed statistically significant effects of field experiments for all elements and ratios, but the level of concentrations found in each experiment only seldom exceeded intermediate ranges according to general experience. Pooled analyses of foliar concentrations further showed statistically significant effects of provenances for the elements Ca, Mg, and P, where the level of concentrations varied six to ten percent between provenances. The different levels of element concentration between the various provenances may indicate a possible different need for specific nutrients between provenances. The variation in survival of plants between field experiments in the period 1-3 years after the stand establishment was probably heavily influenced by the extremely severe draught situation during the months May, June, and July, 1992. Analyses showed that survival of plants were significantly dependent on foliar concentrations of the macro nutrients K and P. The concentration of K had a positive effect on survival, whereas P, when seen in connection with N and K only, had a negative effect. Significant positive effect on survival was also found for the micro nutrients Mg and Cu. No significant effects of N, P, K or any other supplementary element was found on height growth.

Table 2.2.  Lsmean values for experiments and predicted deviations for provenances. All calculations based on non-transformed data.

Subtask 2.3. Analysis of national provenance tests and comparison of results with the CpDNA map.

The comparison of the genetic data (cpDNA) with the fossil pollen data confirmed earlier hypothesis that the observed divergence between chloroplast lineages results from isolation of the oaks into separate refugia during the glacial periods. That the four most frequent chloroplast lineages (A,B,C,E, see Figure 1.3) originated from 3 major refugial zones is a strong indication of genetic differentiation related to a long period of genetic isolation between refugia. In this subtask, we tested whether the nuclear differentiation is still present in the today's populations, or if it has disappeared as a result of pollen flow among population and/or local selection pressures. We compared if these trends of variation of nuclear differentiation are parallel to those observed for cpDNA variation, that witness the maternal origin of the provenances. The tests are based on two different statistical methods. First of all ANOVA will be used to evaluate differences of mean provenances values for various phenotypic traits among the different maternal lineages. Second pairwise genetic distances between populations based on maternal lineages will be correlated to differences in phenotypic traits and tested by using Multiple Mantel tests (Smouse et al., 1986).

2.3.1. Testing for a correlation between chloroplast divergence and nuclear divergence

            Two series of data concerning traits controlled by nuclear genes were available. First of all, phenotypic traits were assessed in  16 provenance tests established in Western Europe since the fifties (Table 2.3) (Great Britain, Germany, France). Traits that were assessed in the tests corresponded to four different categories: survival, growth, stem form, and phenology (Table2.3). All over the data compiled here represent 62 traits, over all tests in the two species (Q. petraea and Q. robur). Differences between provenances for phenotypic traits were tested separately within each provenance plantation by using standard analysis of variance. Provenance mean values were then compiled in separate files by each partner of the project for each test for further analysis

            The second series of data corresponds to genetic diversity surveys of Q. petraea populations across Europe based on isozymes and DNA markers (Table 2.4.). Results of the survey were published in earlier reports (Zanetto and Kremer., 1995; Le Corre et al., 1998; Le Corre et al., 1997). Frequencies of 56 alleles belonging to 8 isozyme loci were available for the 89 populations.

Each provenance was assigned to a given maternal lineage according to its most frequent cpDNA haplotype. Because cpDNA is highly differentiated among populations, the assignment to a given lineage was usually straightforward, except in a few cases of mixed populations. Provenances that comprised haplotypes from different lineages were discarded from the analysis and considered as mixed populations. Differences between provenances were tested using either classical  ANOVA or the Mantel test.

(1)   BB: Bud burst; D: breast height diameter; DA: damages on the tree; F: form of the trunk; H: height; (2) LR: leaf retention in winter; S: survival V: volume

(3)   Numbers after each character indicate the age of the tree when the phenotypic trait was assessed.

(4)   N : total number of provenances used for the analysis

Nb, Nc, Na, Ne : Number of provenances belonging respectively to the B, C, A and E lineage (see Figure 1.3). Nm : Number of provenances comprising different lineages

Table 2.4.  Description of the gene diversity survey

N: total number of provenances

Nb, Nc, Na, Ne : Number of provenances belonging respectively to the B, C, A and E lineage (see Figure 1.3). Nm : Number of provenances comprising different lineages

A, Ae: number of alleles, effective number of alleles.  Ho, He: observed and expected heterozygosity

F: fixation index

The correlation between cytoplasmic and nuclear divergence was calculated as follows. A chloroplastic genetic distance based on maternal lineages (CGD) was calculated for all pairs of provenances within a provenance test. CGD was defined as the number of restriction fragment polymorphisms separating the two populations. Similarly for phenotypic traits, a  differentiation index (DI) was computed between all pairs of populations as the absolute value of the difference between mean values of phenotypic traits of two provenances. For each provenance test, a square matrix of chloroplastic genetic distances (CGD) is constructed and compared with a square matrix of phenotypic distances (DI). The comparison is made after computing the product moment correlation between CGD and DI. Significance of the correlation coefficient was tested with the help of the Mantel test (Mantel, 1967). Geographic distances (GeoD) were used for correcting the calculation of the correlation between CGD and DI. This is the reason why the computation of the partial correlation between DI  and CGD at constant GeoD  (r(DI, CGD),  GeoD) is proposed as an alternative to test whether the maternal origin of a provenance has still an impact on nuclear controlled traits (Smouse et al., 1986). Interestingly this partial correlation coefficient was also  compared to r (DI, GeoD), CGD) (correlation of DI phenotypic distance and geographic distance at constant chloroplastic distance).

2.3.2. Lack of association between chloroplastic and nuclear divergence.

Among the 62 phenotypic traits,  only 7 exhibited  significant  associations with maternal lineages  using ANOVA (mostly growth traits, Table 2.5), and 6 using the Mantel test. This number even decreased to 2 once correction for geographic distance was introduced in the calculation of  the Mantel test. There were stronger associations with gene markers. Among the 18 allozyme frequencies that were available, 8 showed significant differences among maternal lineages corresponding to 7 loci (among the 8 that were used for the allozyme  diversity study (Table 2.6.).  These differences followed either an increasing or decreasing order in the following sequence of maternal lineage: B, C, A and E. Furthermore, there were significant differences among lineages for gene diversity statistics. In general, the observed and heterozygosity values differed markedly among the 4 lineages (Table 2.6); heterozygosities values were higher for populations belonging  the B lineage , than in the A or C. The lowest values were observed in the E lineage.

Table 2.5. Differences among maternal lineages for phenotypic traits using ANOVA.

Table 2.6.  Differences among maternal lineages for nuclear diversisty statistics.

The computation of product moment correlation between the differentiation index (DI) for phenotypic traits and genetic chloroplast distance (CGD) revealed only 6 cases of significant relationships based on the Mantel test among the 62 tests that were made. In contrast to phenotypic traits, nuclear genetic distances (GD) computed  for isozymes  were significantly correlated to CGD, and the correlation remained significant for the RAPDs when it was computed as partial correlation. Furthermore DI values of diversity parameters (Allelic richness, observed and expected heterozygosities) were strongly correlated to CGD, but an important part of the relationship was due to the correlation between CGD and geographic distance. When the impact of geographic distance is removed by the calculation of partial correlation, only the correlation for DI of  He remained significant. Again correlations between DI and GeoD were higher and remained significant when calculated as partial correlation coefficients.

            The two methods (ANOVAs and Mantel tests) that were used to detect associations between cpDNA and nuclear controlled traits lead to congruent conclusions. The extensive computation of data from various provenance tests installed in different countries has indicated that the imprint of the maternal origin on extant populations has in general no significant impact on the phenotypic traits which are of interest in forestry. Neither the ANOVA nor the correlation between chloroplastic and nuclear divergence has revealed any important trend of variation in growth, flushing, form that could be attributed to the origin of the provenances. The lack of  association between chloroplastic divergence and phenotypic traits may be related to the differential impact of the evolutionary history on these two traits.  To sum up our conclusions for phenotypic traits, the scenario that lead to the today's structure of chloroplastic diversity and variation for phenotypic traits  can be subdivided in four major phases. (1) The glacial period ended with the subdivision of the deciduous oaks in three major refugia that were genetically differentiated for chloroplastic and nuclear genes. At the end of the glacial period, there was most likely an association between chloroplastic and nuclear divergence. (2) As the recolonization begun, oaks migrated northwards and during migration installed their spatial distribution for chloroplastic genes. Because later seed migration between installed oak stands were limited, the chloroplastic spatial structure that was established during recolonization has never been changed (Petit et al., 1998). (3) As oaks occupy progressively the mid and northern part of Europe, pollen flow established communication between stands originating from eastern and western refugia. Pollen migration increased as stands from different origins merged, and reduced the initial nuclear differentiation between  the gene pools of the different refugia. At the same time the chloroplastic divergence installed during colonization remained untouched due to limited effective seed flow among established stands. (4) Finally local selection pressures acting on the installed populations contributed to their genetic differentiation, which constantly increased with time. New patterns of differentiation were installed that are totally different from the one left over after colonization. At the same time, chloroplastic divergence remained still untouched. As a result, there is no association any more between chloroplastic divergence and phenotypic traits.

Task 3 Assessments and dynamics of levels of genetic diversity

Subtask 3.1. Sampling of populations.

3.1.1. The Intensive Studied Plots (ISP)

            During the first coordination meeting, criteria for sampling populations where gene diversity would be assessed were discussed and decided. The selection of 9 ISP (Intensive studied plots, Table 3.1) was based on the following rules:

- Mixed stand comprising Q. petraea and Q. robur approximately in equal proportions. Mixing would be pure zones of each species adjacent to each other. This intermixing may be spatially subdivided in three zones: 2 pure zones and one zone were the two species are confronted to each other.

- Stand of natural origin

- Adult trees (more than 120 years old)

- Population size for each species should be close to 200. This is the real size and not a sample size.

Table 3.1 List of the ISP and their composition

3.1.2. Identification of species

            During the course of the project it appeared that the distinction between the two species was not based on the same criteria in the different countries. We therefore established a standard protocol of 14 leaf morphological assessments as a consensus mean for species identification in the different ISP. These characters were measured on 16055 leaves (3025 trees) representing the total number of trees in all ISPs. These data were analysed with three different multivariate methods, in order to ensure a relative independance of our conclusions from the data analysis process: (i) canonical discriminant analysis (CDA)  using the tree as the classification variable; (ii) principal component analysis (PCA) of the 14 x 14 correlation matrix computed from the 14 variables x 3025 trees table; (iii) multiple correspondence analysis (MCA) of the 14 variables x 3025 trees table, each variable being divided into 15 classes of nearly equal weight.

Table 3.2- Correlation between variables and the first axis.

Whatever the method used, the first axis explained two to three times more variance than the second one and the distribution of the variables corresponding to the first axis followed a clear two bell-shaped distribution (Figure 3.1).  Variables of highest correlations with this first axis are the same for the three methods : PL (Petiole length)  and PR (Petiole ratio) , NV (Number of intercalary veins) and PV (percentage of venation). SW (Sinus width), LDR (Lobe depth ratio) and HR (Hairiness) are also correlated to the first axis. Thus, the first axis can be interpreted as a gradient between Q. robur and Q. petraea shapes. The bimodal distribution was used to separate the trees corresponding to the two species. The interesting point is that there is a clear gap between the two morphotypes (Q. robur and Q. petraea) whatever the method used (Figure 3.1): the distribution is a clear mixture of two bell-shaped distributions.

Subtask 3.2. Standardization of molecular techniques for the analysis of nuclear polymorphisms

            A workshop was held from March 17th to 21st 1997 to share the techniques for microsatellites analysis developped by P1 among the other participants. Protocols for analyzing 6 microsatellite loci were described and implemented by all participants involved in Task 3 (2 originating from Dow and Ashley (1996) and four from Steinkellner et al. (1997)). These protocols are available on the website  (http://www.pierroton.inra.fr/Fairoak/) and used mainly the silver staining method. At the end of the workshop, primer pairs corresponding to the 6 microsatellites were distributed together with reference DNA samples corresponding to standards to be used for comparative purposes across laboratories.

            A test cross was further implemented in order to compare the scoring procedure between  laboratories. Each partner sent 10 DNA extracts to P1 who performed the comparison between his own scoring and the scoring procedure of the partner. The test cross was further needed because the partners used finally different methods (silver staining, and different sequencing machines). The results of the test cross indicated five different scoring discrepancies among partners resulting

-          (E1) Systematic shift of allele size. For example the allele that was scored 202 by P1 was actually scored 205 by P8. The difference of allele size (here 3bp) was constant across the range size of alleles.

-          (E2) Systematic shift of allele size. However the difference of allele size was not constant across the range size of alleles. The difference amounted to a certain value when the allele size was lower than a given threshold, and then changed above the threshold.

-          (E3) Random variation of allele sizes. There were occasionally discrepancies between  allele sizes. One allele would be scored 203 by one partner, and 204 by partner P1.

-          (E4) Differences in genotype identification, especially confusions between heterozygotes and homozygotes. In some cases a tree bearing alleles 203 and 205  for example was scored as heterozygote by partner P1 and homozygote (205-205)  by another partner. This occurred mainly when the two alleles exhibited a small difference in size.

-          (E5) Miscoring of unusual large size alleles. For a few loci there were alleles of unusual extreme size corresponding to either deletion or insertion in flanking regions. They were usually small differences in allele size for these alleles.

Overall some loci were more affected than others  (particularly MSQ4 and AG104 Table 3.3).

Table 3.3. Discrepancies of scoring procedures between laboratories

For the definition of discrepancies (E), see text.

0: no discrepancy, +: 10% discrepancies: ++:20% discrepancies

Subtask 3.3. Estimation of levels of diversity

3.3.1. Measuring levels of diversity in oak species

Standard genetic parameters were estimated within each species and within each ISP (Nei, 1987): allelic richness (A), effective number of alleles (Ae), observed heterozygosity (Ho), expected heterozygosity (He), and fixation index (Fis). The estimates of the parameters were made as the mean value over the different loci. Resampling methods were used to test for significant differences between two populations a and b. For example, in the case of He, the procedure consisted in resampling individuals with replacement within each population a and b, calculate the mean value for He in each population and the difference of He between the two population. Resampling was then repeated 1000 times. The distribution of (He_a-He_b) was then compared to the null hypothesis that He_a-He_b=0.

Because there were discrepancies of microsatellite scoring across the different laboratories the estimation of genetic diversity was done in two different ways:

-          The comparison of diversity between Q. petraea and Q. robur was done separately within each ISP, by using the scoring procedure developed by the partner in charge of the given ISP. Since the two species within an ISP were scored the same way, the comparison of species was not affected by the discrepancies of scoring.

-          The comparison of diversity of Q. petraea populations across sites was done after transforming the original data by taking into account the discrepancies. The transformation of data could be done when systematic discrepancies were identified ( E1, E2 and E5). In this case corrections were made according to the results obtained by the test cross and allele sizes were shifted adequately. Furthermore alleles differing by one base pair and present in low frequencies were merged in allelic classes in order to correct for E3. Furthermore, the comparison of diversity was restricted to the expected heterozygosity (He) that is less affected by small changes of allele frequencies.

3.3.2. Difference of gene diversity between Q. petraea and Q. robur.

            In each ISP, there were clear indication that there is a higher genetic variation within populations for Q. petraea than for Q. robur (Table 3.4). There was at least one measure of diversity  (among A, Ae, He and Ho) per ISP that showed higher values in Q. petraea than in Q. robur , whereas there was never any measure showing the opposite. The differences in levels of diversity were usually small, but significant. These results confirm earlier reports based on other markers and characters:

- Gene diversity surveys based on isozymes (Müller Starck et al., 1993; Kremer et al; 1991;  Zanetto et al., 1994) indicate that heterozygosity values were higher in Q. petraea,than in Q. robur.

- Data from DNA analyses conducted in both species (Moreau et al., 1994; Bodénès et al, 1997) reached also similar conclusions.

            The species differences in levels of diversity may be related to their social status. Stands of Q. petraea are usually pure and of larger size than those of Q. robur that can be more generally mixed with other species. Furthermore the co-evolution of the two species, the so called “regeneration” of Q. petraea from successive unidirectional hybridization with Q. robur  (Petit et al., 1998) can also be considered as a mechanism contributing to the enrichment of genetic diversity in Q. petraea . The enrichment results from the addition of the diversity coming from hybridization to the diversity already existing in Q. petraea per se. Lastly results from mating system (Bacilieri et al., 1997) have shown that the outcrossing rate is larger in Q.  petraea than in Q. robur. The difference in outcrossing rates is likely to contribute also to the difference in the fixation index (Fis). As shown on table (3.4), in three ISPs there is a higher fixation index in Q. robur than in Q. petraea, indicating a higher excess of homozygotes in Q. robur than in Q. petraea.

Table 3.4 Diversity statistics in different oak populations

p values obtained by resampling methods

3.3.3. Geographic variation in levels of diversity

            For each species, the within population diversity (expected heterozygosity) was compared among the different ISPs. The calculations were done after transforming the original data to account for the differences in scoring procedures (Table 3.5). The estimated values were compared by using the boostrapping resampling method. Despite the low number of populations, some trends of variation can be identified.

            In Q. petraea and Q. robur, populations from the central and north central area exhibit higher values than others. This is a confirmation of earlier reports (Zanetto and Kremer, 1995) that showed that allozyme diversity was higher in the central region where the species occupies large areas and where the most “famous” stand exist. Various hypotheses can be advocate to interpret these results :

-          the central zone of the natural distribution can be considered as a hot spot of diversity because it constitutes a “merging zone” of the different recolonisation routes from the different refugial zones (Figure 1.2 and 1.3).

-          The central zone may also be considered as an area where the human interference was the most pronounced either by local transfer of material or by using even aged silvicultural regimes. The former impact contributes to the “hot spot” hypothesis whereas the latter results in maintaining larger populations sizes.

Table 3.5. Comparisons of levels of diversity (He, expected heterozygosity) across sites

* values not significantly different are separated by the same letter.

Subtask 3.4. Estimation of genetic variation for phenotypic traits

            Due to uneven and heterogeneous fruiting this Subtask could not be addressed. Collections of open pollinated progenies were done, when fruiting was sufficient by P1, P5, P8 and P11. The material is being raised and will be evaluated during the next coming years.

References

Bacilieri R., Ducousso A., Petit R.J., Kremer A., 1996. Mating system and directional gene flow in a mixed oak stand. Evolution 50: 900-908

Bennett, K.D., Tzedakis, P.C. and Willis, K.J. 1991. Quaternary refugia of north European trees. Journal of Biogeography, 18, 103-115.

Berglund, B.E., Birks, H.J.B, Ralska-Jasiewiczowa, M. and Wright, H.E. 1996. Palaeoecological Events during the last 15 000 years. John Wiley & Sons, Chichester, 764 pp.

Beug, H.J. 1960. Beitrage zur postglazialen floren und vegetationsgeschichte in Suddalmatien: der see "Malo Jesero" auf Mljet. Flora, 150(4), 600-631.

Beug, H.J. 1975. Changes of climate and vegetation belts in the mountains of mediterranean Europe during the Holocene. Biuletyn Geologiczny, 19, 101-110.

Birks, H.J.B. 1989. Holocene isochrone maps and patterns of tree-spreading in the British isles. Journal of Biogeography, 16, 503-540.

Bodénès C., Labbé T., Pradère S., and Kremer A., 1997. General vs local differentiation between two closely related white oak species. Molecular Ecology 6: 713-724

Bottema, S. 1979. Pollen analytical investigations in Thessaly (Greece). Palaeohistoria, XXI, 20-40.

Bottema, S. and Van Zeist, W. 1991. Late Quaternary Vegetation of the Near East. Beihefte Zum Tübinger Atlas Des Vorderen Orients, Reihe A (Naturwissenschaften) Nr 18. Dr Ludwig Reichert Verlag, Wiesbaden, 156 pp.

Bowman, S. 1990. Radiocarbon dating. Interpreting the past. British Museum Press, 64 pp.

Brande, A. 1973. Unterschungen zur postglazialen Vegetationgeschite im Gebiet der Neretva-Niederungen (Dalmatioan, Herzegowina). Flora, 162, 1-44.

Brewer S, Cheddadi R, de Beaulieu J.-L. and Data contributors 2000.  The migration of deciduous Quercus throughout Europe since the last glacial period. For. Manag. Ecol.

Carrion, J.S. and Dupré, M. 1996. Late Quaternary vegetational history at Navarrés, eastern Spain. A two core approach. New Phytologist, 134, 177-191.

Culiberg, M. 1991. Late glacial vegetation in Slovenia. Academia Scientarium et Artium Slovenica, 29. Ljubljana, 52 pp.

Csaikl U., Chloroplast DNA  variation of white oaks in the alpine region. Forest Ecology and Management (Special Issue)

Demesure B.1995, Sodzi N., Petit R.J., 1995. A set of universal primers for amplification of polymorphic non-coding regions of mitochondrial and chloroplast DNA in plants. Molecular Ecology 4:129-131

Demesure, B., B. Comps, and R. J. Petit, 1996 Chloroplast DNA phylogeography of the common beech (Fagus sylvatica L.) in Europe. Evolution 50: 2515-2520.

Deutsch, C.V. and Journel, A.G. 1998. GSLIB Geostatistical software library and user's guide. Applied geostatistics series. Oxford University Press, 2^nd edition, 369 pp.

Dow, B.D., and Ashley, M.V., 1996. Microsatellite analysis of seed dispersal and parentage of saplings in bur oak., Quercus  macrocarpa. Molecular Ecology, 5, 615-627

Dumolin-Lapègue S, Demesure B, Le Corre V, Fineschi S, Petit RJ (1997) Phylogeographic structure of white oaks throughout the European continent. Genetics, 146, 1475-1487.

Dumolin S., Demesure B, Petit R.J., 1995. Inheritance of chloroplast and mitochondrial genomes in pedunculate oak investigated with an efficient PCR method. Theoretical and Applied Genetics 91:1253-1256

Dumolin-Lapègue S, Pemonge M-H, Petit R J (1998) Association between chloroplast and mitochondrial lineages in oaks. Molecular Biology and Evolution, 15, 1321-1331.

Farcas S., Beaulieu J.-L. de, Reille, M., Coldea G., Diaconeasa B., Goeury C., Goslar T., Jull T. (1999) Premières datations 14C de séquences polliniques tardiglaciaires et holocènes des carpates roumaines. C.R. Acad. Sci. Paris. Sciences de la Vie, 322, 799-807.

Ferris, C., R. P. Oliver, A. J. Davy, and G. M. Hewitt, 1993 Native oak chloroplasts reveal an ancient divide across Europe. Mol. Ecol. 2: 337-344.

Ferris, C., King R. A., Vainola R, and G. M. Hewitt, 1998 Chloroplast DANN recognises three refugial sources of European oaks and shows independent eastern and western immigrations to Finland. Heredity 80: 584-593.

Goiecoechea  P et al., 2000 White oak phylogeography in the iberian Peninsula, Forest Ecology and Management (Special Issue)

Huntley, B. and Birks, H.J.B. 1983. An atlas of past and present pollen maps for Europe: 0-13000 years ago. Cambridge University Press, 667 pp.

Huntley, B. and Prentice, I.C. 1993. Holocene vegetation and climates of Europe. In: Wright Jr., H.E., Kutzbach, J.E., Webb III, T., Ruddiman, W.F., Street-Perrott, F.A. and Bartlein, P.J. (Eds.), Global climates since the Last Glacial Maximum. University of Minnesota Press, Minnesota and London, pp. 136-168.

Kremer A., Petit R.J., Zanetto A., Fougère V., Ducousso A., Wagner D., and Chauvin C., 1991. Nuclear and organelle gene diversity in Quercus  robur and Q. petraea. In: Genetic variation in European Population of Forest Trees (Müller-Starck G., and Ziehe M. eds), Sauerländer’s Verlag. Frankfurt am Main.

Le Corre,V., Dumolin-Lapègue, S., Kremer, A., 1997. Genetic variation at allozyme and RAPD loci in sessile oak Quercus  petraea (Matt.) Liebl.: the role of history and geography. Molecular Ecology 6, 519-529

Le Corre,V., Roussel, G., Zanetto, A., Kremer, A., 1998. Geographic structure of gene diversity in Quercus  petraea (Matt.) Liebl. III Patterns of variation identiifed by geostatistical analysis. Heredity 80, 464-473

Lowe, J.J, Accorsi, C.A, Asioli, A., Van Der Kaars, S and Trincardi, F. 1996. Pollen-stratigraphical records of the last glacial-interglacial transition (ca. 14-9 ¹⁴C ka BP) from Italy: a contribution to the PALICLAS project. Il Quaternario, 9(2), 627-642.

Lowe, J.J. and Walker, M.J.C. 1997. Reconstructing Quaternary environments. Longman, 2^nd edition, 446 pp.

Mantel, N.A., 1967. The detection of disease clustering and a generalized regression approach.  Cancer Research 27, 91-99

Moreau F., Kleinschmit J., and Kremer A., 1994. Molecular differentiation between Q. petraea and Q. robur assessed by random amplified DNA fragments. Forest Genetics 1(1) :51-64

Müller-Starck G, Herzog S. and Hattemer H.H., 1993. Intra- and interpopulation genetic variation in juvenile populations of Quercus  robur L., and Quercus  petraea Liebl. Ann. For Sci  50: 233s-244s

Nei M., 1987. Molecular evolutionary genetics. Columbia University Press, New York.

Nicol-Pichard, S. 1987. Analyse pollinique d’une séquence tardi et postglaciare à Tourves (Var, France). Ecologia Mediterranea, 13, 29-42.

Petit, R. J., E. Pineau, B. Demesure, R. Bacilieri, A. Ducousso, and A. Kremer 1998 Chloroplast DNA footprints of postglacial recolonization by oaks. PNAS

Petit R.J. et al., 2000 Postglacial colonisation routes of European white oaks inferred from the variation of chloroplast DNA: synthesis based on more than 2600 populations. Forest Ecology and Management.

Pons, O., Petit, R.J., 1995. Estimation, variance and optimal sampling of gene diversity. I. Haploid locus. Theoretical and Applied Genetics 90: 462-470.

Pons, A. and Reille, M. 1988. The Holocene and upper Pleistocene pollen record from Padul (Granada, Spain): a new study. Palaeogeography, Palaeoclimatology, Palaeoecology, 66, 243-263.

Ruddiman, W.F. and McIntyre, A. 1981. The mode and mechanism of the last deglaciaction: oceanic evidence. Quaternary Research, 16, 125-134.

Schneider, R. and Tobolski, K. 1985. Lago Di Ganna - Late-glacial and Holocene environments of a lake in the southern Alps. Dissertationes Botanicae, 87, 229-271.

Shopov, V.L., Bosilova, E.D. and Atanasova, J.R. 1992. Biostratigraphy and radiocarbon data of Upper Quaternary sediments from western part of Black Sea. Geologica Balcanica, 22(2), 59-69.

Smith, A.G. and Pilcher, J.R. 1973. Radiocarbon dates and vegetational history of the British Isles. New Phytologist, 72,903-914.

Smouse P.E., Long J.C., Sokal R., 1986. Multiple regression and correlation extensions of the Mantel test of matrix correspondence. Systematic Zoology 35, 627-632

Steinkellner H., Fluch S., Turetschek E., Lexer C., Streiff R., Kremer A., Burg K., and J. Glössl, 1997. Identification and characterization of (GA/CT)n microsatellite loci from Quercus  petraea. Plant Molecular Biology 33: 1093-1096

Stevenson, A.C. 1984. Studies in the vegetational history of S.W. Spain. III. Palynological investigations at El Asperillo, Huelva. Journal of Biogeography, 11, 527-551.

Stuiver, M., Reimer, P.J., Bard, E., Beck, J.W., Burr, G.S., Hughen, K.A., Kromer, B., McCormac, G., Van der Plicht, J. and Spurk, S. 1998. Intcal98 radiocarbon age calibration, 24,000-0 cal BP. Radiocarbon, 40(3), 1041-1083.

Watts, W.A., Allen, J.A., Huntley, B. and Fritz, S.C. 1996. Vegetation history and climate of the last 15,000 years at Laghi di Monticchio, southern Italy. Quaternary Science Reviews, 15, 113-132.

Zanetto, A., and A. Kremer, 1995 Geographic structure of gene diversity in Quercus petraea (Matt.) Liebl. 1. Monolocus patterns of variation. Heredity 75: 506-517.

Zanetto A., Roussel G., and A. Kremer 1994. Geographic variation of inter-specific differentiation between Quercus  robur and Quercus  petraea Matt.) Liebl. Forest Genetics 1:111-123

Annexes