Volume 92, Issue 6 p. 1006-1016
Systematics and Phytogeography
Free Access

Patterns of molecular and morphological differentiation in Fagus (Fagaceae): phylogenetic implications1

Thomas Denk

Corresponding Author

Thomas Denk

2 Swedish Museum of Natural History, Department of Paleobotany, Box 50007, 104 05 Stockholm, Sweden;

Author for correspondence (e-mail: [email protected])Search for more papers by this author
Guido W. Grimm

Guido W. Grimm

3 Center of Plant Molecular Biology, Department of General Genetics, Eberhard-Karls-University, Auf der Morgenstelle 26, D-72076 Tübingen, Germany

Search for more papers by this author
Vera Hemleben

Vera Hemleben

3 Center of Plant Molecular Biology, Department of General Genetics, Eberhard-Karls-University, Auf der Morgenstelle 26, D-72076 Tübingen, Germany

Search for more papers by this author
First published: 01 June 2005
Citations: 48
1

The authors thank the Swedish Research Council and the German Science Foundation for financial support.

Abstract

To study phylogenetic relationships among species of Fagus, the internal transcribed spacer regions ITS1 and ITS2 of the nuclear ribosomal DNA and morphological data were analyzed. Both molecular and morphologically based phylogenies suggest that Eurasian species of Fagus subgenus Fagus are basal to the North American Fagus grandifolia. The subgenus Fagus is a paraphyletic group basal to three East Asian species forming the subgenus Engleriana. Due to a considerably large amount of DNA polymorphism, relationships among basal species of Fagus could not be entirely resolved when analyzing ITS sequences with standard methods. Morphological trees helped to resolve more clearly relationships within the subgenus Fagus. The East Asian F. hayatae is suggested to be basal to the rest of the genus. This hypothesis is further supported by distinctive patterns of nucleotide variability found for ITS regions, allowing for basic and derived types to be distinguished. The high degree of ITS polymorphism within Fagus can be explained by (1) the complex evolutionary behavior of this marker, (2) the stenoecious ecological characteristic of Fagus with respect to its continuous geographic range throughout much of the Cenozoic, and (3) the absence of major radiations into further habitats as occurred in other Fagaceae.

Fagus (beech, Fagaceae) is a small genus of 10 monoecious tree species in the northern hemisphere (Shen, 1992; Denk, 2003). It is the most abundant broadleaved forest tree in Europe and western Asia and forms an important component of mixed broadleaved evergreen–deciduous forests in North America and East Asia (Zhou and Li, 1994; Peters, 1997). The oldest fossils that can be ascribed with certainty to the genus are known from the Middle Eocene (ca. 45 million years ago [mya]) of western North America (Pigg and Wehr, 2002; Manchester and Dillhoff, 2004). By the late Early Oligocene (ca. 30 mya), Fagus covered a range from Pacific North America to Asia and Western Europe. Recent phylogenetic studies showed that Fagus constitutes an early branch within the Fagaceae (Manos and Steele, 1997; Manos et al., 2001). The dissimilarity of DNA sequences of Fagus to those of other members of the family suggests that, although basal within the family, it may represent a quite derived lineage, making it difficult to find a suitable outgroup for cladistic analysis. Shen (1992) presented the first comprehensive monograph on Fagus and established two subgenera within the genus on the basis of distinct morphological features (Table 1). Stanford (1998) and Manos and Stanford (2001) carried out the first molecular studies, which did not support Shen's subgenera. Intriguing is the fact that their studies resulted in conflicting phylogenies. Denk et al. (2002) showed that sampling a single individual for each species to infer the intrageneric phylogeny is inadequate for ITS sequences of Fagus. They detected conspicuously high intraspecific DNA polymorphism, which, linked to high morphological polymorphism, appears to be a main characteristic of the genus. Furthermore, that study strongly supports Shen's (1992) concept of two subgenera. Most recently, a morphologically based phylogenetic study (Denk, 2003) confirmed the two subgenera recognized by Shen (1992).

Because no molecular systematic study including all species of Fagus has been carried out, we used ITS sequences of all species of Fagus to infer an intrageneric phylogeny for Fagus. In addition, morphological characters were optimized on phylogenetic trees. Phylogenetic scenarios are discussed in light of molecular, morphological, and fossil evidence. Based on these considerations, the biogeographic distribution of this genus is interpreted with respect to its evolution in the Cenozoic.

MATERIALS AND METHODS

In addition to published ITS sequences (Denk et al., 2002), samples of the following taxa are included in the analysis: Fagus crenata Blume, F. engleriana Seemen, F. grandifolia Ehrh. subsp. caroliniana (Loudon) Camp ex Shen, F. grandifolia subsp. mexicana (Martinez) Camp ex Shen, F. hayatae Palibin subsp. pashanica (C. C. Yang) Shen, F. japonica Maxim., F. longipetiolata Seemen, F. lucida Rehder and Wilson, and F. sylvatica L. Most samples were collected in the field. In all cases, several independently obtained PCR products were cloned, and up to five clones per sample (individual) were sequenced. New sequences have been submitted to NCBI GenBank (Appendix, see Supplemental Data with the online version of this article for accession numbers and voucher information). Previously published sequences of other authors are not included because they may lack crucial information due to the assembling procedure, i.e., direct sequencing of PCR products (cf. Álvarez and Wendel, 2003). For new material, the extraction of total DNA, amplification of the ITS, ligation and transformation of PCR products, isolation of plasmids, and automated sequencing followed the procedures described in Denk et al. (2002). Chromatogram files and E. coli cultures of all clones are kept at the Center for Plant Molecular Biology, Eberhard-Karls-University, Tübingen, Germany, and can be supplied upon request.

Sequences were automatically aligned with the Clustal algorithm implemented in MegAlign (DNAStar, Madison, Wisconsin, USA) and manually readjusted at two positions comprising length polymorphism within the ITS1 and ITS2. Length polymorphism is of only minor importance in the ITS of Fagus and can always be attributed to a single indel event; therefore, it does not bias the alignment process. To infer a molecular-based phylogeny, maximum likelihood (ML) via Bayesian inference (BI) was applied (MrBayes 3.0; Huelsenbeck and Ronquist, 2001), and alignment gaps were included in the analysis. Due to the structure of the ITS data set, a maximum parsimony (MP) analysis could not be used to reconstruct a molecular phylogeny (cf. Denk et al., 2002). Bayesian inference analyses were performed with the following parameters: 1 000 000 generations on five parallel Monte Carlo Markov chains, each 100th generated topology saved. A likelihood ratio test performed with Modeltest 3.06 (Posada and Crandall, 1998) proposed a general substitution model (GTR), allowing different probabilities for all kinds of substitutions. Accordingly, for the BI analyses the number of substitution types was set to six, which were assumed to be gamma-distributed and containing invariable sites (i.e., GTR + Γ + I substitution model). A consensus tree was then computed from the saved topologies, ignoring those topologies that precede the optimum plateau. Statistics for the analysis are provided in Table 2 (including permuted likelihood parameters). A total of 137 clones from 43 sampled individuals, representing all species and effectively covering the geographic range of Fagus, were included in the analysis. Two more species of Fagus that appear to be “good” species, F. chienii Cheng and F. okamotoi Shen, were not included in the ITS study. Fagus chienii has been described on the basis of a single tree, and subsequent attempts to find the type individual and further plants belonging to this species failed. Fagus okamotoi is described as a new species in Shen's (1992) dissertation but has never been formally published. Both species have been included in a phylogenetic study by Denk (2003) and have been considered for the morphological analysis of the present study. For ITS studies, isolation from herbarium material of F. okamotoi failed because DNA was already heavily decayed. ITS sequences of other Fagaceae (e.g., Trigonobalanus) turned out to be so different from Fagus sequences that any alignment with the ingroup must be considered biased for regions that are variable. In the more conserved and ± alignable regions, sequences of Fagus are basically identical and differ from all other Fagaceae; hence, they contain no information with respect to ingroup differentiation. The same is true for Nothofagus (Nothofagaceae), a possible candidate for outgroup comparisons in morphological studies (Denk, 2003).

The data matrix of Denk (2003) was used to construct phylogenetic trees based on morphological data and subject to MP analysis. Most of the characters used are discrete, but a few are continuous (chars. 4, 25, 26 in Table 6). They are considered informative, because clearly distinct categories are recognized. The choice of characters is based on extensive field sampling and herbarium studies (see Denk, 2003, 2004). Ingroup character states for Fagus were inferred using PAUP version 3.1.1 (Swofford, 1993) and McClade version 3.0 (Maddison and Maddison, 1992). Trigonobalanus or Nothofagus were used alternatively as outgroups for Fagus. Character states for the ingroup node of Fagus were compared to character states known from earliest fossils of the genus. Fossils used for comparison were cupules and nuts (Middle Eocene of western North America, Pigg and Wehr, 2002 and Manchester and Dillhoff, 2004; Early Oligocene of western North America, Meyer and Manchester, 1997), leaves (Middle Eocene of western North America, Manchester and Dillhoff, 2004; Late Eocene of Kamchatka, Budantsev, 1997; Early Oligocene of western North America, Meyer and Manchester, 1997; Early Oligocene of Japan, Tanai, 1995), and pollen (Middle Eocene of western North America, Manchester and Dillhoff, 2004; Oligocene of Europe, Schmid, 2000).

RESULTS

Genetic distances

To evaluate levels of molecular differentiation within the ITS of Fagus, pairwise genetic distances between 137 accessions, representing 43 individuals, have been calculated with MEGA 2.1 (Kumar et al., 2001) on the basis of a gamma-distributed Kimura-2-parameter substitution model (summarized in Table 3). High interspecific distances (0.074–0.119) are generally found between clones representing the subgenus Engleriana (F. engleriana, F. japonica) and the subgenus Fagus. Within the subgenus Fagus, interspecific distances are significantly lower: The lowest interspecific differences are detected between F. sylvatica and F. crenata (Japan, Ø = 0.022), F. lucida (China, Ø = 0.026), and F. longipetiolata (China, Ø = 0.027). The highest interspecific distances within subgenus Fagus are detected for F. grandifolia (N America, 0.025–0.058; see Table 3).

Highest intraspecific distances are found between clones of F. engleriana and F. japonica (subgenus Engleriana) from either the same sample or different samples and localities (Ø = 0.079–0.105). Taxa belonging to the subgenus Fagus exhibit generally lower intraspecific differences, with particularly low values occurring in F. crenata, F. grandifolia, and F. sylvatica (especially European individuals; Denk et al., 2002). Increased intraspecific differences are detected for F. hayatae subsp. pashanica.

Maximum likelihood analysis of nucleotide data

An unrooted phylogram inferred from a Bayesian analysis of ITS sequence data is shown in Fig. 1. The accessions group into four distinct lineages. High posterior probabilities can be found for the common base of the subgenus Engleriana (lineage I, 100%) and accessions of F. grandifolia (lineage II, 89%). The two species representing the subgenus Engleriana, F. engleriana (China mainland and South Korea) and F. japonica (Japan), are genetically not distinguishable and share at least three ITS subtypes. One is characterized by a prominent 13-bp indel within the ITS1. Most accessions representing the Eurasian taxa of the subgenus Fagus are not resolved as distinct clades (lineage IV), with the exception of a number of accessions from clones of F. hayatae subsp. pashanica and F. longipetiolata (lineage III, 100%). The split between lineage IV and lineages I–III is supported by a posterior probability of 72%. Within lineage IV, accessions representing F. crenata, F. hayatae subsp. pashanica, F. longipetiolata, F. lucida, and F. sylvatica intermix. The grouping of accessions, however, is never completely random, but each groups with at least one more accession of the same taxon. In lineage IV, clones of F. hayatae subsp. pashanica also consistently plot together with F. longipetiolata clones. A similar topology is produced when the strict consensus sequences for each species are computed and used for the analysis (not shown, cf. Grimm, 2003). The alignment of strict consensus sequences further demonstrates that unambiguous sites are generally missing, with the exception of accessions of F. grandifolia (cf. Table 4).

In general, a standard base-per-base cladistic analysis resolves only the systematic position of F. grandifolia and the relation of members of the subgenus Fagus to those of the subgenus Engleriana. The taxonomic position of F. hayatae subsp. pashanica and F. longipetiolata, with clones in lineages III and IV, is obscure. Moreover, relationships of F. crenata, F. lucida, and F. sylvatica still remain unresolved (cf. Denk et al., 2002).

Patterns of intraspecific nucleotide variability as phylogenetic information

To further resolve uncertain species relationships and to understand patterns of differentiation leading to the ambiguous position of taxa such as F. hayatae subsp. pashanica and F. longipetiolata, we took a closer look at the actual nucleotide composition of the ITS (summarized in Table 4). A minimal amount of length polymorphism in ITS sequences of Fagus guarantees homology of nucleotides and site variabilities at certain positions of the ITS and therefore allows detailed comparative studies. The most conspicuous feature of the nucleotide data is the large amount of subgenus Engleriana-typical site variabilities (cf. alignment sites 78f, 98ff, 108, 139, 165, 187ff, 220ff, 228, 275, 284ff, 306, 318, 475, 531ff, 552, 612, 689ff, 704, and 716, 724 in Table 4). These site variabilities generally comprise character states that correspond to the consensus nucleotide state found in many or all taxa of the subgenus Fagus (see exemplary illustration in Fig. 2). In several cases, the same types of variability found in the subgenus Engleriana are also found in certain members of the subgenus Fagus: At position 167, the same variability is found in F. crenata (very rarely in F. sylvatica) and the subgenus Engleriana and at positions 591 and 671 in F. longipetiolata and the subgenus Engleriana.

Site variabilities characteristic of both F. hayatae and F. longipetiolata suggest a basal position for either F. hayatae and/or F. longipetiolata within the genus. For instance, at position 121 F. hayatae and F. longipetiolata clones have a guanine (G), adenosine (A), or cytosine (C). Clones of F. crenata and F. sylvatica may have a G or an A, whereas F. lucida has either G or C (Table 5). The same pattern is found at positions 152ff (Fig. 3), 521ff, 562ff, and 653, and is reflected in the grouping of accessions representing these taxa in lineages III and IV. Several other site variabilities restricted to F. hayatae and F. longipetiolata (at positions 163, 212, 547; Table 5) may be derived or basal.

At positions 180ff and 505f (Fig. 4) are complex patterns of site variability that appear to represent the differentiation from simple to more complex patterns from F. hayatae to Fagus subgenus Engleriana or to F. grandifolia. Besides, a few variabilities shared between F. hayatae/longipetiolata and F. engleriana may have evolved in parallel, perhaps originating from an old polymorphism or from incomplete concerted evolution in combination with ancient hybridization events.

Morphology: “hayatae basal” vs. “lucida basal” hypothesis

Monophyly of the subgenus Engleriana and the position of F. grandifolia basal to it have previously been suggested by a morphologically based phylogenetic study (Denk, 2003). This study suggested that either F. hayatae or F. lucida is basal within the genus, depending on two different sets of outgroup taxa. In addition, sister group relationships were recognized for F. sylvatica + F. crenata, F. lucida + F. chienii, and F. hayatae + F. longipetiolata. In the present study, we used the data matrix of Denk (2003; characters that are informative at the ingroup level are listed in Table 6). First, we tested two competing hypotheses, F. hayatae most basal vs. F. lucida most basal within Fagus. We used Nothofagus as the outgroup for Fagus, resulting in a topology with F. hayatae as most basal within Fagus (Fig. 5), or Trigonobalanus resulting in a topology with F. lucida as the most basal species within Fagus (not shown). Then we reconstructed character states for the ingroup nodes of the trees obtained from these analyses and compare them to character states known from the oldest fossils of different organs belonging to Fagus.

Ingroup node character states differ depending on which outgroup has been chosen (Table 7). In principal, the reconstruction for the cladogram with Nothofagus as outgroup and F. hayatae as basalmost species within Fagus matches well with character states known from oldest fossils of Fagus. Oldest fossil cupules and nuts have short peduncles, medium to small cupules with spine like appendages, and conspicuously winged nuts that are as long as the cupule valves (Pigg and Wehr, 2002; Manchester and Dillhoff, 2004). Most of these features are reconstructed for the ingroup node of the hayatae-basal tree but not for the lucida-basal tree. For pollen, the reconstruction suggests large pollen for the ingroup node, regardless of the outgroup used. This is at odds with the fossil record and with the relatively smaller pollen of other Fagaceae. For vegetative characters, there is a fairly good consensus between reconstructed and fossil character states for discrete morphological characters with the exception of stomata size (Table 7).

Character optimization

Morphological characters are mapped on the single most parsimonious cladogram obtained from the analysis with Nothofagus as outgroup (Fig. 5). Vegetative as well as reproductive characters display a high degree of polymorphism, which is reflected in few non-homoplastic apomorphies (full circles) and many homoplastic characters (open circles). Clade support through non-homoplastic apomorphies is restricted to the clades F. crenata and F. sylvatica (small ratio of length to width of leaf), F. grandifolia + subgenus Engleriana (anomocytic stomata), and subgenus Engleriana (multistemmed, low-branching trees; large lenticels; stipitate buds). All other apomorphies are autapomorphies.

A number of characters appear to have evolved more than once. For example, wax glaucosity and papillate lower leaf surfaces, characteristic of species of the subgenus Engleriana, can also be found in F. longipetiolata; pubescent lower leaf surfaces occur in some populations of F. longipetiolata and F. grandifolia. These characters are not known from the fossil record and may have evolved very late in the history of Fagus. In contrast, small stomata occur in F. hayatae, F. grandifolia, and the subgenus Engleriana, and have been found in fossils from the Oligocene. Nevertheless, both character optimizations (outgroup Nothofagus, outgroup Trigonobalanus) indicate large stomata to be the character state at the ingroup node, because outgroup taxa have significantly larger stomata than ingroup taxa. Within the cupule–nut complex, closely related species display transitions between different character states. For instance, F. japonica has unwinged nuts, which are longer than the cupule valves, whereas F. okamotoi has obviously winged nuts, which are as long as the cupule valves. The same can be observed in F. lucida + F. chienii. In F. crenata and F. longipetiolata nuts may be winged or unwinged within the same species, whereas in F. grandifolia and F. hayatae nuts may be as long as the cupule valves or considerably longer. Eocene fossils clearly suggest conspicuously winged nuts to be the ingroup character state for Fagus.

The appendages of the cupule valves are a highly complex character. Spine-like appendages are suggested to be the ingroup character state for the F. hayatae-basal phylogeny and are found in the earliest Fagus cupules from the Middle Eocene. The scale-like and subulate appendages in F. lucida and F. chienii are suggested to be derived types. A number of more derived species display leaf-like basal appendages in addition to spine-like appendages. When the occurrence of leaf-like appendages is optimized on the cladogram in Fig. 5, leaf-like basal appendages appear to have evolved in a lineage of hypothetical ancestors to the clade (F. crenata + F. sylvatica)– (F. grandifolia–(subgenus Engleriana)) and subsequently been lost in F. japonica and F. okamotoi.

DISCUSSION

Reliability and coherence of phylogenetic hypotheses

Intrageneric relationships within Fagus could not be entirely resolved using standard cladistic analyses of ITS sequence data because of the very complex patterns of intraindividual and intraspecific nucleotide variability. However, thorough investigation of the types of variability within the ITS, coupled with the reconstruction of morphological character evolution, allows detailed insights into the intrageneric differentiation of Fagus. Within the subgenus Engleriana, a number of typical variability patterns can be detected (Tables 3, 4), but they are not restricted to one of the two morphologically distinct species, F. japonica or F. engleriana. This suggests that the morphological differentiation of these two species outran the fixation of a specific ITS type, probably as a result of incomplete concerted evolution and/or extended unhindered gene flow between populations. The same applies to most of the species of the subgenus Fagus, in which overlapping patterns of intraspecific ITS variability (Tables 4, 5), along with comparatively low levels of interspecific genetic divergence (Fig. 1; Table 3), substantially hamper standard phylogenetic analyses. It is, however, important that the ML phylogeny and patterns of molecular differentiation obtained from ITS data are fundamentally similar to morphologically based phylogenies. Some of the relationships that are resolved by morphological analyses, for instance, the basal position of Fagus hayatae–(F. longipetiolata), are also indicated by patterns of intraspecific genetic variability. A polymorphic basal F. hayatae–(F. longipetiolata) may also explain the occurrence of F. hayatae and F. longipetiolata sequences in two lineages of the ML phylogram. Also, the two subgenera Fagus and Engleriana and the isolated position of F. grandifolia within the subgenus Fagus are strongly supported by both data sets.

Two previous studies on Fagus (Stanford, 1998; Manos and Stanford, 2001) used the same nuclear rDNA regions and an additional chloroplast DNA marker to construct MP-based phylogenies for the genus. Very few samples per taxon were included in these studies, and single-gene phylogenies were more or less unresolved (Stanford, 1998), although combined analysis resulted in appreciable cladistic structure. Rydin and Källersjö (2002) found that phylogenetic studies on seed plants may be affected considerably by the numbers of terminal taxa used. Unlike Rosenberg and Kumer (2001), they found that changing the set of terminal taxa is likely to produce contradictory phylogenies. Similarly, the previously mentioned studies on Fagus, which used only one or two samples per species and direct sequencing of PCR products, are likely to have covered only a very random set of the genetic variability present in the group under study (cf. Álvarez and Wendel, 2003). This may explain the conflicting results of these studies. In contrast, our much larger data set produces a less resolved phylogeny with standard methods, which in return is not affected by adding or omitting terminal taxa.

Most recently, Álvarez and Wendel (2003) expressed several objections against the utilization of ITS data, which hinged largely on the intraindividual variability encountered in ITS sequences. Because the ITS is part of a multicopy gene region, paralogous sequence relationships may confound phylogenetic reconstructions. At the same time, variable marker regions such as the ITS allow insights into organismal reticulation and ancient hybridization, provided that ITS sequences are not generated from direct sequencing of single PCR products, and clones are sampled to assess sequence diversity (Álvarez and Wendel, 2003). Ancient hybridization events must be assumed for Fagus on the basis of the assembled morphological, fossil, and molecular data, and with respect to the biogeographic history and ecology of the genus. Bailey et al. (2003) noticed that a broader assembling of intraindividual, interindividual, and interspecific ITS variability may considerably enhance the resolution of phylogenetic reconstructions. Additionally, such a broad database allows the evaluation of whether or not paralogous and/or pseudogenous sequences affect phylogenetic reconstruction. For our data, we can assume that the large congruence between morphological phylogenies, the occurrence of certain character states in the early fossil record of Fagus, the ML-based phylogeny, and the differentiation patterns of nucleotide variability (see Table 4, Fig. 4) is not completely accidental and thus reflects fairly well the actual evolution of Fagus.

Basal species within Fagus

The types of nucleotide variability point to a Fagus hayatae basal hypothesis. Fagus hayatae and F. longipetiolata share a similar intraspecific ITS variability, including molecular variability found in other species (Tables 4, 5). In particular, their variability includes types that are restricted to the genetically (BI analysis) and morphologically distinct species F. grandifolia and to members of the derived subgenus Engleriana. One explanation would be that ancient genetic polymorphisms have been retained in modern F. hayatae and F. longipetiolata (genetic living fossils) with partial loss of genetic diversity in subsequent lineages. Other explanations would require the fixation of numerous, convergently developed, identical mutation events at the same positions within the ITS, or unhindered gene flow via frequent introgression–hybridization events until recent times. The conspicuously lower degree of intraspecific variability in F. crenata, F. lucida, F. sylvatica, and F. grandifolia, however, invalidates the assumption that randomly occurring mutational events are fixed at a constant rate within the ITS. Frequent hybridization and introgression until recent times is not probable in view of the biogeographic history and modern distribution of the genus.

A phylogenetic study based on morphology by Denk (2003) could not unambiguously resolve relationships within basal species of Fagus. Best candidates for basalmost lineages within the genus were F. lucida and F. hayatae. When fossil species were included in the analysis, they consistently grouped with F. hayatae (and F. longipetiolata). For the present study, we chose an outgroup basal to Fagus and to Fagaceae (Nothofagus), which resulted in a topology with F. hayatae basalmost within Fagus. This, and the fact that character optimization on the single most parsimonious cladogram reconstructed character states that match well the states known from earliest fossils of the genus, is another indication that F. hayatae represents a basal lineage within Fagus.

Reconstructing ingroup node character states, outgroups, and the fossil record

Reconstructing ingroup node character states for the morphological tree recovered from the data set with Nothofagus as outgroup (Fig. 5) showed that reconstructed character states agree well with the evidence from the earliest fossils attributable to Fagus. However, stomata are reconstructed to be large at the ingroup node, which is at odds with the trend from smaller stomata to slightly larger stomata in the fossil record for the Cenozoic of Europe (Kvaček and Walther, 1991). In this case, character state reconstruction may be affected by the large size of stomata found in the outgroup taxa (see Denk, 2003), but does not necessarily reflect the actual state for the ingroup node. Among living species, small stomata occur in F. hayatae, F. grandifolia, and the subgenus Engleriana. In light of the fossil record, these species are more likely to have retained the primitive character state, while the remaining species of the subgenus Fagus have evolved larger stomata. The leaf margin in oldest fossils of Fagus (Manchester and Dillhoff, 2004) varies from entire to conspicuously dentate. These leaves belong to a species that cannot be assigned to one of the two extant subgenera (Denk, 2004), which may explain why the present phylogeny based on modern species suggests that the dentate leaf margin is the character state for the ingroup node.

For the size of pollen, large pollen at the ingroup node requires two steps to optimize this character on the cladogram. Small pollen for the ingroup node would require one step more, and therefore be less parsimonious. Nevertheless, it appears more likely that small pollen is plesiomorphic within Fagus. This is based on the fossil record (Walther and Zetter, 1993; Schmid, 2000; Manchester and Dillhoff, 2004) and the much smaller pollen in other taxa of Fagaceae and Nothofagaceae. Oldest known Fagus pollen from the Middle Eocene is conspicuously smaller than in other fossil and modern species of Fagus (cf. Denk, 2003; Manchester and Dillhoff, 2004) and displays long and narrow colpi, whereas Middle Oligocene pollen (Schmid, 2000) displays shorter and broader colpi. It seems plausible that from a small pollen type with long and narrow colpi, a slightly larger one with long colpi (subgenus Engleriana) and with short colpi (F. hayatae) evolved. Large pollen with short colpi or with either short or long colpi appears to be derived.

Biogeography and ecology

Tiffney and Manchester (2001) pointed out the importance of fossils for biogeographic consideration. In the case of Fagus, the earliest fossils have been described from the Middle Eocene of western North America (Pigg and Wehr, 2002; Manchester and Dillhoff, 2004). From there, the genus spread to the northwest and reached Eurasia (Kamchatka) via the Bering Strait by the Late Eocene (Fotjanova, 1982). This would point to a Northern Pacific origin of Fagus rather than to a Chinese origin, as suggested by phylogenies based exclusively on modern species (Manos and Stanford, 2001). The genus reached Western Europe by the late Early Oligocene (Schmid, 2000) and was distributed continuously in (western) North America and Eurasia in the Late Oligocene (Tanai, 1974; Iljinskaja, 1982; Kvaček and Walther, 1991). The absence of major radiations during this period may explain shared morphological characteristics and patterns of molecular variability within and between the two modern subgenera. The fossil record of Fagus is scarce for China (Liu et al., 1996), and ancestors of Chinese species likely grew in adjacent areas of the Pacific such as Taiwan, Japan, Korea, and Sakhalin (e.g., Tanai, 1974, 1995). In the Late Cenozoic, Fagus in western Eurasia combines features of different modern eastern Asian species (Zetter, 1984; Kvaček and Walther, 1991) and is similar to coeval eastern Asian species (Leng, 2000). Modern species belonging to the subgenus Fagus appear to have taken shape only in the Latest Cenozoic, which would explain the poor (morphological and molecular) resolution among these modern species. In this context, we should point out that even though the continuous Eurasian distribution area had become disrupted by the end of the Middle Miocene, mosaic types with features of both modern western Eurasian and Eastern Asian species persisted in Europe until the Pliocene (Denk, 2004).

Fossil leaves clearly pointing to the modern subgenus Engleriana (genetically and morphologically derived) occur for the first time in the Miocene of Sakhalin (F. evenensis; Chelebaeva, 1980), while the highly derived reproductive structures typical of the subgenus Engleriana are not known from the fossil record (Denk and Meller, 2001).

The genus Fagus consists of rather few closely related species that are ecologically very similar to each other (cf. Cao, 1995). The main habitats of Fagus are humid cool to warm temperate forests of lowlands (North America, western Eurasia) and forests of the montane vegetation belt in temperate and subtropical areas (North America, western Eurasia, East Asia; Peters, 1997). Growing mostly under conditions close to the ecological optimum, beeches are highly stenoecious trees. Based on the fossil record, Fagus appears to have occupied very similar habitats throughout the Cenozoic (Kvaček and Walther, 1991; Meyer and Manchester, 1997; Knobloch, 1998; among many others). This may be a reason for the absence of major radiations during the geological history of Fagus and could explain the difficulties that arise when phylogenetic studies are undertaken, both at the morphological and the molecular marker levels. Other genera in the Fagaceae, such as Quercus, are much more species rich (see, e.g., Mabberley, 1997) and underwent major radiations into diverse habitats.

Conclusions

Both morphological and genetic evidence clearly points toward a derivation of the subgenus Engleriana from the subgenus Fagus. ITS sequences of the subgenus Engleriana are characterized by the co-occurrence of putative ancestral nucleotide states, which are the dominant elements in most individuals of the subgenus Fagus, and derived nucleotide states, which are specific for accessions of the subgenus Engleriana. The presence of ancestral or derived nucleotide states and variability patterns within the ITS allows us to recognize several ITS subtypes (cf. Fig. 2). In the case of the subgenus Engleriana, various subtypes show different derived nucleotide states that co-occur with the ancestral state. The resulting increased intraspecific genetic variability shared by both subgenera can be explained by increased mutation rates and evolutionary speed on the one hand, and incomplete concerted evolution at the genomic level on the other (cf. Álvarez and Wendel, 2003; Volkov et al., 2004). The inferred increased evolutionary rate for members of the subgenus Engleriana correlates well with the accumulation of derived morphological characters. Because F. engleriana and F. japonica are morphologically clearly distinct but cannot be separated based on ITS sequence data, it must be assumed that speciation processes within this group are still in progress. Clearly, more detailed studies of members of the subgenus Engleriana, including F. okamotoi Shen, are needed.

Fagus grandifolia, native to eastern North America, is the most distinct taxon within the subgenus Fagus. Morphological as well as genetic data point toward a rather early differentiation of F. grandifolia from the rest of the subgenus Fagus. In addition, patterns of molecular differentiation and the assumption of a “basal” F. hayatae allow us to interpret the sister taxon relationship between F. grandifolia and the subgenus Engleriana, as inferred from the BI analysis and morphology, within a global paleobiogeographic framework. Complete evidence points to a Northern Pacific origin of the genus, instead of an East Asian origin as previously suggested (Manos and Stanford, 2001). Eocene and Lower Oligocene taxa (F. langevinii Manchester and Dillhoff, F. napanensis Fotjanova, F. uemurae Tanai, F. kitamiensis Tanai) that cannot be assigned to either of the two modern subgenera may have differentiated into a fraction that comprised those populations that spread westward to Central Asia and Europe, and a Northern Pacific fraction that gave rise to a lineage ancestral to the modern F. grandifolia and to the subgenus Engleriana. The remaining species of the subgenus Fagus must be considered descendants of a genetically and morphologically rather weakly differentiated, widely distributed Eurasian Paleogene taxon (Fagus castaneifolia Unger; Denk, 2004) that migrated to Europe as soon as the Turgai strait closed and eastern and western Eurasia were connected. In modern Eurasian species of the subgenus Fagus, speciation processes have not been fully completed, and because of repeated phases of area expansion and shrinkage, originally occurring genetic variability eventually has been lost and is preserved to a certain degree in the morphologically and genetically basalmost taxa (F. hayatae + F. longipetiolata) and in geographic refugia (Georgia, Transcaucasia). Differentiation into morphologically distinct species happened as a consequence of the fragmentation (Miocene or Pliocene) of the originally continuous distribution area and, hence, the interruption of horizontal gene flow.

Table 1. Classification of Fagus based on Shen (1992) and Denk (2003). Species that are not italicized are not recognized as distinct species here and in a previous study by Denk (2003). No subspecies are indicated
image
Table 2. Statistics for Bayesian inference analysis based on 9173 samples of 10 001 samples recorded
image
Table 3. Mean genetic distances between species and individuals of Fagus originating from different geographic regions. Pairwise distances calculated with MEGA 2.1, using a gamma-distributed Kimura 2-parameter model (gaps pairwise deleted). Bold font indicates mean interindividual distance for each group (species, subspecies, geographic origin). Light to dark gray shading indicates increasing genetic distances.
image
Table 4. Patterns of intrageneric differentiation in the ITS of Fagus. a–e indicate detected nucleotide compositions at a certain position (gray background) or intraindividual or intraspecific variability (white background). Details are given in the right column. Sites in bold indicate key positions for the relationship between subgenus Engleriana and Fagus.
image
Table 5. Variability patterns of F. hayatae and F. longipetiolata indicating archaic polymorphism in Fagus.
image
Table 6. Morphological character states that are informative at the ingroup level
image
Table 7. Ingroup node character states for two maximum parsimony analyses using different outgroups and character states known from oldest fossils of Fagus
image
Details are in the caption following the image

Unrooted maximum likelihood phylogram from a Bayesian inference analysis of ITS sequence data, with posterior probabilities indicated for each lineage. Branches supported with less than 50% are collapsed

Details are in the caption following the image

Types of ITS variability: incomplete concerted evolution (site 180ff, Table 4). A basic motif (CACAAA + d!; gray background) can be found in all individuals of Fagus. Conspicuous intraspecific variability in this region (comprising a total length of 27 bp) is restricted to the subgenus Engleriana, F. grandifolia, and F. crenata. In F. grandifolia, the basic motif is gradually replaced by a grandifolia-specific motif (bottom left; subsp. mexicana: one of 4 clones; subsp. caroliniana: 3 of 5; subsp. grandifolia: 4 of 6). The pathway of molecular differentiation in individuals of the subgenus Engleriana (dotted circle) is exemplary for the complete ITS: the basic motif is still retained (9 of 28 accessions, 1st ITS subtype); 10 accessions have a 2- bp shorter motif (CAAA + d!, 2nd subtype) that is also found in a single F. sylvatica clone (ho-1904, Georgia). In the remaining accessions, this motif occurs in combination with a subgenus Engleriana-typical 13-nt elongation at position 187ff (CAAA + i!, 3rd subtype; Table 4)

Details are in the caption following the image

Types of ITS variability: ancient polymorphisms (sites 152–159, ITS1). All ITS accessions of Fagus have one of the three shown nucleotide motives at alignment positions 152–159. Only a single fixed mutational event is necessary to derive a “Pacific” 6-nt motif (C-3G-AA, top left), which is typical of accessions of F. grandifolia and the subgenus Engleriana, and a “Chinese” 8-nt motif (C-5G-AG, bottom right, common in F. lucida) from the assumed basic motif C-5G-AA (center), which is found in all Eurasian representatives of the subgenus Fagus. The co-occurrence of the Pacific and/ or Chinese motif in individuals of F. sylvatica (especially in Georgian individuals) and F. hayatae/F. longipetiolata (shaded) is best explained by ancient polymorphism

Details are in the caption following the image

Types of ITS variability (sites 505–526, 5′ end of ITS2). The 5′ end of the ITS2 illustrates the mode of intrageneric differentiation within the ITS of Fagus. Nucleotide motifs can be derived from each other by single mutational events; both putative ancestral (motif B) and derived motifs are found among ITS copies. Motif A is characteristic for the subgenus Fagus. The systematic position of F. grandifolia as sister taxon to the remaining species is reflected by the occurrence of a further derived motif in subsp. caroliniana, which becomes the dominating motif in subsp. grandifolia. The basic motif B, found in all Eurasian Fagus subgenus Fagus, is completely lacking in F. grandifolia. The subgenus Engleriana (hatched) exclusively exhibits YY…CCCTC-type motifs, which can be derived from motif C, also found in the polymorphic “genetic living fossils” F. hayatae and F. longipetiolata (shaded)

Details are in the caption following the image

Morphological characters optimized on the single maximum parsimony cladogram with Nothofagus as outgroup. Full circles indicate apomorphies with a consistency index (CI) of 1.0, and open circles are characters with homoplasy (consistency index below 1.0). Bootstrap values from 1000 replicates are indicated as bold numbers on the left of the branches. Tree length 65, consistency index 0.846, homoplasy index 0.538, retention index 0.767, rescaled consistency index 0.649. For characters and character states see Table 6