Phylogeny and Biogeography of Paeonia (PAEONIACEAE)
EVOLUTION, CLASSIFICATION, AND BIOGEOGRAPHY OF PAEONIA (PAEONIACEAE)
ABSTRACT. Phylogenetic reconstructions using sequences of internal transcribed spacers (ITS) of nuclear ribosomal DNA and chloroplast DNA reveal monophyly of each of three sections of the genus Paeonia. Within section Moutan two subsections, Delavayanae and Vaginatae, are monophyletic groups in molecular phylogenies. In section Paeonia, however, two taxonomic subsections are not in agreement with phylogenetic relationships. Taxonomic difficulties within section are attributable to complex reticulate evolution. Evolution of major taxonomic characters are examined based on phylogenetic reconstructions and previous results of artificial hybridizations. Sequence divergences of ITS, matK gene, and psbA-trnH intergenic spacers of cpDNA are compared within and among sections of Paeonia. Sequences
ITS evolve slightly more rapidly than the psbA-trnH intergenic spacer, and over three times more rapidly than the matK coding region. DNA sequence divergences suggest that the earliest evolutionary split might have occurred between section Oneapia and the other two sections. Morphologically section Oneapia evolved very slowly, whereas subsection Vaginatae diverged rapidly. Among species of hybrid origin in section Paeonia, the proportion of diploids is surprisingly high, suggesting that hybrid speciation at the diploid level is quite frequent in peonies. The Eurasian and western North American disjunction between section Oneapia and the rest the genus may have resulted from interruption of a continuous distribution of peonies across the Bering land bridge during middle Miocene. Pleistocene glaciation may have been a primary factor in triggering extensive reticulate evolution within section Paeonia and may have also drastically shifted distributional ranges of species of the section.
Paeonia comprises approximately 35 species of shrubs and perennial herbs distributed widely in five disjunct areas in the northern hemisphere: eastern Asia, central Asia, the western Himalayas, the Mediterranean region, and Pacific North America (Stern, 1946; Pan, 1979; Tzanoudakis, 1983; Pei, 1993). The genus is systematically isolated, having been placed in the unigeneric family Paeoniaceae which has either been placed by itself or together with Glaucidiaceae in order Paeoniales (Takhtajan, 1969, 1987; Thorne, 1992).
Because of their great ornamental and medicinal value, peonies have been known as "king of flowers" in China and "queen of herbs" in Greece for more than one thousand years (Gambrill, 1988).
Paeonia was divided by Lynch (1890) into three subgenera, Moutan. Oneapia. and Paeon. In the latest and most widely recognized monograph of Paeonia (Stern,1946), these same subdivisions were maintained, but as sections. Section Oneapia,, endemic to Pacific North America, comprises two herbaceous species with conspicuous staminodial disks and small fleshy concave petals. Section Moutan with six species, occurring in central and western China, is divided into two subsections, Delavayanae and Vaginatae. They are shrubs with conspicuous staminodial disks and large spreading petals. Section Paeonia ("Paeon"), which includes the type species P. officinalis. is also divided into two subsections, Foliolatae and Paeonia ("Dissectifoliae"), distributed disjunctly in eastern Asia, central Asia, the western Himalayas, and the Mediterranean region. This section consists of approximately 27 herbaceous species with inconspicuous or no staminodial disks and large petals that are either spreading or cup-shaped. Sections Oneapia and Moutan contain only diploid species (2n = 10) , while one third of the species in section Paeonia are tetraploids (Stern, 1946; Tzanoudakis, 1977; Hong et al., 1988) (Table 1).
Paeonia is a phylogenetically and taxonomically complex group (Stebbins, 1938a; Hong et al., 1988). Particularly, section Paeonia may have undergone reticulate evolution which makes classification even more difficult. Regarding origins Of tetraploid species in this section. Barber (1941) and Stern (1946) considered that they were autotetraploids derived from certain extant diploid ancestors. In contrast, Stebbins (1948) argued that the majority of tetraploid species are allotetraploids based on observations of bivalents in meiosis of most tetraploid species that he studied (e.g., P. officinalis. P. peregrina. and P. wittmanniana). He further indicated that certain tetraploid species appeared to link gaps of morphological variation among some diploid species, which also suggested hybrid origins of the tetraploids. Later cytogenetic studies supported allotetraploid origin of P. officinalis and P. peregrina, and also revealed P. parnassica as an allotetraploid (Tzanoudakis, 1977; Schwarzacher-Robinson, 1986). Recent molecular phylogenetic studies using both nuclear and chloroplast DNA sequences revealed extremely complex reticulate evolution in section Paeonia (Sang et al, in press; in prep.). 'The majority of tetraploid species as well as the diploid species are derived from one or more hybridization events.
Molecular makers that have been used for phylogenetic studies include internal transcribed spacers (ITS) of nuclear ribosomal DNA, matK gene coding region of cpDNA, and cpDNA intergenic spacers psbA-trnH and trnL-trnF (Sang et al., in press, in prep.; Sang, in prep.) Results described in the previous chapters include the ITS phylogeny of section Paeonia, matK phylogeny of the genus, psbA-trnH spacer phylogeny of the genus, and comparison of the- average rates of sequence divergence in the genus among matK, psbA-trnH intergenic spacer, and trnL-trnF intergenic spacer (Sang et al., in press, in prep; Sang, in prep. In the present chapter, the ITS phylogeny of the entire genus is presented and compared with cpDNA phylogenies. Sequence divergences of ITS, matK. psbA-trnH intergenic spacer are calculated within and among sections. Based on molecular phylogenies and DNA sequence divergences, patterns and rates of morphological evolution through divergent and reticulate speciation in Paeonia are examined. Taxonomic problems of the genus, especially involving reticulate evolution in section Paeonia, are discussed. Modes of speciation, rates of DNA sequence divergence, and concerted evolution of ITS sequences are compared at diploid and tetraploid levels.
Paeonia, with widely disjunct distributions and rich endemism, provides a favorable system for studying historical biogeography of the northern hemisphere. Historical biogeography of this region has been strongly impacted by Pleistocene glaciation, which drastically altered distributions of organisms and caused significant extinction
92 (Noonan, 1988; Potts and Behrensmeyer, 1992). 'Pleistocene' glaciation has been suggested as a primary factor triggering extensive hybridization in Paeonia *(Sang et al., in press). Rapid speciation via hybridization and drastic changes of distributional ranges of peony species since Pleistocene glaciation are examined more carefully for biogeographic implications. Intercontinental disjunct distribution between section Oneapia (western North America) and the other two sections (Eurasia) is also discussed based on phylogenetic reconstructions and the molecular clock hypothesis. >
MATERIALS AND METHODS
Accessions of Paeonia species sequenced for ITS, matK, psbA-trnH intergenic spacer, and trnL-trnH intergenic spacer are given in Table 1. For most species , fresh leaves used as sources of DNA were collected from natural populations in Bulgaria, China ,Greece, and Spain. The remaining species were collected, from The Royal Botanic Garden, Kew (Table 1) . Total DNA was isolated from leaf tissues using the CTAB method (Doyle and Doyle, 1987), and purified in CsCl/ethidium bromide gradients. Double-stranded DNAs were amplified by 30 cycles of symmetric PCR. Details regarding PCR and primers are described in the previous chapters. The amplification products were purified by electrophoresis through 1.0% agarose gel followed by use of Bioclean (U. S. Biochemical) . Purified double-stranded DNAs were used for sequencing reactions employing Sequenase Version 2.0 (U. S. Biochemical), deoxyadenosine 5^-ta-[35S]thio]triphosphate, and the same primers used in PCR (Sang et at., 1995). The sequencing reaction products were separated electrophoretically in 6% acrylamide gel with wedge spacers for 3 hr at 1500 V. After fixation, gels were dried and exposed to Kodak XAR x-ray film for 18-48 hr. DNA sequences were aligned manually.
Sequence divergence between species for each DNA region was calculated using DNADIST program of PHYLIP version 3.5c (Felsenstein, 1994) . The Jukes-Cantor model was used as the method of correcting possible multiple hits of nucleotide substitutions (Jukes and Cantor, 1969).
Mutations including nucleotide substitutions and indels were analyzed by unweighted Wagner parsimony using PAUP version 3.1.1 (Swofford, 1993). Indels are coded as binary characters in the analysis. The shortest trees were searched with TBR Branch Swapping of the heuristic method, and character changes were interpreted with the ACCTRAN optimization. Bootstrap analyses were carried out with 1000 replications using TBR Branch Swapping of the Heuristic search (Felsenstein, 1985). Section Oneapia of Paeonia was chosen as the outgroup for the cladistic analysis of the genus (Sang et al., in press; also see discussion).
Morphological characters were examined from literature (Stebbins, 1938a; Stern, 1946; Fang, 1958; Pan, 1979), field collections, and herbarium specimens during visits to or on loan from ATHr.GRA, GH, K, KUN, NY, PE, SO, SOM, SZ, UQ, UPA, US. and WUK.
Phylogenetic relationships of Paeonia species not showing ITS sequence additivity, and thus ostensibly not derived through hybridization, were reconstructed using parsimony analysis of 47 variable sites of the ITS sequences. The strict consensus tree (CI = 0.911, RI == 0.959) was generated from two equally most parsimonious trees (CI == 0.927, RI = 0.967) (Fig. 1.). Species of section Paeonia that show ITS sequence additivity, and thus presumably were derived through reticulate evolution, were not included in this phylogenetic tree. A more comprehensive reconstruction of reticulate evolution in section Paeonia was obtained by comparing ITS and matK phylogenies (Fig. 6).
Sequence divergences of ITS, matK, and psbA-trnH intergenic spacer were calculated. Average percent sequence divergences of these three DNA regions were compared within and among sections (Table 5). Sequence divergence of trnL-trnF intergenic spacer that evolves very slowly is not presented here. Comparisons of sequence divergences indicate that ITS sequences evolve slightly more rapidly than psbA-trnH intergenic spacer, and over three times more rapidly than the matK coding region.
Classification. Molecular phylogenies support recognition of three sections within Paeonia and two subsections in section Moutan. Separation of section Oneapia and the other two sections is strongly supported by both ITS (99% bootstrap value) and cpDNA (100% bootstrap value) phylogenies (Figs. 9, 11) . Monophyly of section Moutan is also strongly supported by high bootstrap values on ITS and cpDNA phylogenies. Within this section, monophyly of each of two subsections is relatively strongly supported on the ITS and matK phylogenies (Figs. 9, 11), whereas no mutations were found in either cpDNA intergenic spacer to support subsection Delavayanae (Sang, in prep.). This is apparently due to insufficient time for the accumulation of mutations in the two short intergenic spacers in the common ancestor of this subsection (Sang, in prep.). Recognition of the two subsections, therefore, is still supported by overall phylogenetic information.
Support for monophyly of section Paeonia is very strong on the cpDNA phylogenies (100% bootstrap value. Fig. 9), but rather weak on the ITS phylogeny (only one nucleotide substitution, and 65% bootstrap value. Fig. 11). Hybridization between an early evolutionary lineage in the larger ITS clade (as the maternal parent) and the ancestor of the smaller ITS clade (as the paternal) has been suggested to account for this difference (Sang et al., in prep.). Subsequent to this hybridization, hybrids fixed the ITS sequences of the smaller clade type received cpDNA from the larger clade. Therefore, two basal clade are still seen on the ITS phylogeny, but there is only one basal clade on the cpDNA phylogeny. The two ITS clades appear to have diverged soon after the ancestor of section Paeonia separated from that of section Moutan. because there is only one substitution supporting this section on the ITS phylogeny (Fig. 11). Hybridization between these two early divergent groups within section Paeonia may have mixed morphological distinctions between them and thus led to easier taxonomical recognition of this section.
Within section Paeonia. complex reticulate evolution has occurred, which makes natural classification difficult. In this regard, two questions should be addressed: (1) to what extent does the previous classification, based on morphology, cytology, and distribution, reflect the phylogeny of the section; and (2) how should species within section Paeonia, which has undergone extensive reticulate evolution, be treated taxonomically at the subsectional level. Stern (1946) recognized two subsections, Paeonia and Foliolatae, based on number and degree of dissection of leaflets. According to Stern (1946), leaves of subsection Paeonia are very dissected and cut into usually 25 or more lobed or/and toothed leaflets, while leaves of subsection Foliolatae are divided into usually 9-16 (sometimes up to 23) entire leaflets. This classification of section Paeonia was criticized by Stebbins (1948) and Tzanoudakis (1977). Stebbins (1948) argued that this classification ignores that many of Mediterranean tetraploid species, such as P. arietina and P. officinalis, may be allotetraploids with similar origins, and thus separation of them into different subsections is artificial. Tzanoudakis (1977), in his studies of Greek peony species, indicated that P. clusii of subsection Paeonia is the closest relative to P. rhodia of subsection Foliolatae, and he further reduced the former to a subspecies of the latter based on more detailed morphological cytological data.
Interestingly, in the divergent portion of the ITS phylogeny, each of the two major clades of section Paeonia includes species of each of the currently recognized two subsections (Fig. 11) . Although extant species on the smaller ITS clade are derived through hybridization, the ancestor of this smaller clade may have had less dissected leaves. Further, different degrees of dissection of leaves may have occurred in certain species through later hybridization between lineages of these two ITS clades. Most species of subsection Foliolatae are hybrid species derived through hybridization involving directly or indirectly the lineages of the smaller ITS clade (Figs. 6, 11). Hybrid species in these lineages involved directly include: P. obovata. P. japonica. P. banatica. P. cambessedesii. P. russi P. clusii. P. rhodia. P. broteri. P. coriacea, P. mlokosewitschi, P. mascula ssp. hellenica. P. mascula ssp. mascula, and P. sterniana. Paeonia wittmanniana is derived indirectly. Of these species, P. sterniana and P. clusii are placed in section Paeonia. We may conclude, therefore, that section Foliolatae accommodates species possessing the genotype of the smaller ITS clade which are characterized by less dissected leaves. Segregation after hybridization, however, created wide variation in this character, such that P.clusii has the most dissected leaves and is now placed in subsection Paeonia. Paeonia arietina is the only species that apparently does not have the genotype of the smaller ITS clade, but it is placed in subsection Foliolatae. This species and P. officinalis and P. parnassica, having meiotic behavior typical of allotetraploids (Stebbins, 1938b), may also be of hybrid origin even though molecular evidence is still lacking. Therefore, the possibility of involvement of the smaller ITS clade in the origin of P.arietina cannot be ruled out.
The complex reticulate evolution in section Paeonia poses a challenging problem for developing an appropriate system to classify these species satisfactorily. If Sterns (1946) classification is followed, it is phylogenetically misleading to place P. clusii and P. sterniana in subsection Paeonia and to place P. arietina and P. officinalis in two separate subsections (Fig. 6) . Also, P. parnassica described by Tzanoudakis (1983 has leaf morphology of subsection Foliolatae, but its ITS and cpDNA sequences are identical to P. humilis and P. officinalis of subsection Paeonia. If P. clusii, P. sterniana, P. humilis. and P. officinalis are transferred from subsection Paeonia to subsection Foliolatae to maintain consistency between classification and phylogeny, degree of leaf dissection loses its taxonomic importance
Previous cytogenetic data (Stebbins, 1938b; Tzanoudakis, 1977, 1983) and new molecular phylogenetic data (Sang et al., in press, in prep.; Sang, in prep.) suggest modifications in the classification of section Paeonia. Three subsections may be designed to accommodate morphological and molecular phylogenetic information. One subsection would include Asian species with dissected leaves that were derived through strict divergent evolution and the hybrid species derived among them (i.e., P. anomala, P. veitchii, P. lactiflora. P. emodi, P. sterniana, and P. xinjiangensis) . The second subsection would contain eastern Asian species with fewer and broader leaflets belonging to the smaller ITS clade (i.e., P.mairei. P. obovata. and P. japonica). The third subsection would include all the Mediterranean species that may have been derived through hybridization between the first two subsections. This subsection would accommodate a wide range of morphological variation (see later discussion), and thus may be subject to further subdivision for a maximally predictive classification. Paeonia tenuifolia with extremely dissected leaves, does not appears to fall clearly in this subsection. Further investigations are needed to clarify the maternal parent of this species, which was assumed to be an extinct lineage (Fig. 6) . Likewise, more detailed molecular phylogenetic studies should be done to verify possible hybrid origins for P. arietina P. humilis, P. officinalis, and P. parnassica. Final decisions on major changes in the current classification of section Paeonia, therefore, might be held until these studies are completed and correlated with ongoing taxonomic' revisionary studies by D. Hong (Institute of Botany, Beijing; pers. comm.) and N. Rowland (University of Reading, London; pers. comm.).
Morphological evolution. Based on molecular phylogenies, evolution of some major taxonomic characters in Paeonia can be examined. The shrubby habit delimiting section Moutan was previously considered to be primitive (Stebbins, 1938a). Phylogenetic reconstruction using DNA sequence data, however, reveals that the herbaceous habit is more primitive and the shrubby habit is a synapomorphy of section Moutan (Figs. 6, 11).
Length of staminodial disk is also regarded as a taxonomically important character (Stebbins, 1938a). Sections Oneapia and Moutan have very conspicuous staminodial disks that are over 3 mm long. The staminodial disk of section Oneapia is fleshy and lobed and 1/3 to 1/2 as long as carpels. In section Moutan. subsection Delavayanae has a lobed disk that is usually less than 1/3 length of the carpels, whereas subsection Vaginatae has a more specialized disk that envelopes from 1/2 to the entire length of the carpels (Pei, 1993). The staminodial disk is much shorter in section Paeonia. The disk is relatively conspicuous (1-2 mm long) in P. lactiflora and P. veitchii. and their hybrid species P. emodi. It is much less conspicuous, however, in P. xinjiangensis an older hybrid between P. lactiflora and P. veitchii. The staminodial disk is very inconspicuous or absent in P. obovata and P. mairei. Variation of this character among the remaining hybrid species seems to be unpredictable. Disks are relatively conspicuous (ca. 1 mm) in P. coriacea. P. mlokosewitschi. and P. peregrina. Particularly, P. peregrina has a more conspicuous disk than either of its putative parents. Sepal morphology was considered to be a very important phylogenetic character by Stebbins (1938a). He suggested that innermost sepals with conspicuous terminal appendages were primitive. Species with this characteristic include P.anomala, P. lactiflora. P. emodi, and P. veitchii which were thus considered to be more primitive species (Stebbins, 1938a). Paeonia xinjiangensis, described by Pan (1979), also has long terminal appendages. Therefore, this characteristic may represent the genotype of divergent species on the larger ITS clade and their direct hybrids. On the other hand, sepals of P. obovata do not have terminal appendages, and only one or two sepals of P. mairei have very small appendages. The remaining species of hybrid origin have variable intermediate conditions, of which sepals of P.peregrina and P. sterniana have relatively conspicuous appendages. Saunders and Stebbins (1938) observed that sepals of artificial F, hybrids were more nearly like their more primitive parents, and suggested that genes for the more primitive type of sepals are partially dominant. Conspicuous sepal appendages found in P. peregrina and P. sterniana appear 'to support this hypothesis. However, more complex variation found in other hybrid species implies that segregation of these genes may have occurred during hybrid speciation.
Number of flowers per stem is another significant taxonomic character which has also been studied by artificial hybridization (Saunders and Stebbins, 1938). Species in section Oneapia and subsection Delavayanae of section Moutan have more than one flower per stem. In section Paeonia, P. lactiflora, P. veitchii and P. emodi have more than one flower per stem, while P.veitchii has sometimes one terminal flower and weakly developed flower buds on side branches. All the other species have only one flower per stem. Saunders and Stebbins (1938) suggested that the state of more than one flower per stem is primitive and appeared to be partially dominant in F, hybrids. Distribution of this character state on the phylogeny, i.e., occurring in section Oneapia. subsection Delavayanae of section Moutan, and divergent species of section Paeonia. supports the primitive nature of having of more than one flower per stem (Figs. 9, 11) . Unlike results of artificial hybridization, however, genes for one flower per stem are more likely to be dominant because a hybrid species with parents having the two different character states always has one flower per stem (Fig. 6).
Degree of dissection of leaves, as discussed earlier, is a distinguishing feature between two basal ITS clades of section Paeonia. A high degree of dissection of leaves appears primitive because it is also found in section Oneapia and subsection Delavayanae of section Moutan. Hybridization studies revealed that this character tended to be intermediate in F, hybrids (Saunders and Stebbins, 1938). Such intermediate conditions can be observed clearly in P. xinjiangensis. P. peregrina, and P. sterniana. Paeonia cambessedesii and P. russi have entire and broad leaflets which resemble those of their parent, P. mairei. Paeonia clusii has highly lobed leaves, distinguishing it from other species of the same hybrid origin. Within the same population of P. broteri found in Granada, Spain, individuals were found with highly dissected leaves resembling those of P. clusii. and individuals with broad leaflets similar to those of P. coriacea. Therefore, there is no clear tendency for dominance of either character state in species of hybrid origin, and segregation is likely to have occurred making this character even more taxonomically confusing.
Speciation at different ploidy levels. Allopolyploidy has been considered a primary mode for formation of fertile and stable hybrid species (Grant, 1981) Frequency of natural hybrid speciation at the diploid level, however, has been controversial (Rieseberg et al., 1990; Reiseberg, 1991; Wendel et al., 1991; Wolfe and Elisens, 1994). According to the phylogenetic reconstruction (Fig. 6), species that are not derived from hybridization include P. anomala. P. veitchii, P. lactiflora, P. arietina. P. humilis. P. officinalis and P. parnassica, of which only the first three species are diploids. As discussed earlier, the remaining four species that have identical ITS and cpDNA sequences may have been derived through hybridization. The species of hybrid origin identified by ITS and cpDNA sequences include seven diploid and six tetraploid species, and three species with both diploid and tetraploid populations (Fig. 6). The proportion of diploids among the hybrids species is surprisingly high, suggesting that. hybrid speciation at the diploid level has been quite successful in peonies. An even more striking phenomenon is the co-existence of diploids and tetraploids in the same species or a group of species with the same origin (Fig. 6) . It is noteworthy that at least some diploid species seem to be as fertile as their sister tetraploid species based on DNA sequence data. For example, Paeonia broteri (diploid) and P. coriacea (tetraploid) may have the same origin because they share a substitution in matK phylogeny (Fig. 9). Paeonia broteri has three autapomorphic substitutions in matK sequences while P. coriacea has none, suggesting that the diploid hybrid species may have a shorter generation time than its sister tetraploid species. Studies of reproductive biology and application of more sensitive molecular markers at the populational level are necessary for understanding relative fertility of the diploid, and tetraploid species of hybrid origin in Paeonia. Extensive vegetative reproduction by rhizomes in peonies may have facilitated survival of initial diploid populations of hybrids until they became fertile or polyploidized. Regardless of mechanisms responsible for maintenance of both diploid and tetraploid populations, existence of different ploidy levels in the same or very closely related species may eventually result in reproductive isolation and further speciation. Molecular evolution in P. cambessedesii and P.russi, two endemic species in the western Mediterranean islands, is of interest. ITS sequences of P. russi show nucleotide additivity at all sites that are variable between P. lactiflora and P. mairei strongly suggesting that P. russi is derived via hybridization between these two species (Fig. 12). The origin of P. cambessedesii. whose ITS sequences show partial additivity (only at site 1, 2, and 3; Fig. 12), is not so clear because of two possible alternatives: (l) p. lactiflora and P. mairei are the parents, and additivity at sites 6-12 has been homogenized by gene conversion; or (2) instead of P. mairei. P. obovata is one of the parents, and the additivity at sites 4 and 5 has also been homogenized by gene conversion. The matK phylogeny supports the sister group relationship of P. russi and P. cambessedesii. and thus suggests the same origin for both species. This result further confirms that gradients of gene conversion is the mechanism responsible for partial homogenization of ITS sequence additivity in hybrid species. More interestingly, gene conversion has operated more rapidly in diploid P. cambessedesii than in tetraploid P. russi. This is reasonable because in diploids, loci of nrDNA are more easily brought together during meiosis, which allows more effective interaction among these loci and thus more rapid gene conversion (Arnheim, 1983). However, this hypothesis does not apply to a group of diploid and tetraploid species, P. clusii. P. rhodia. P. broteri, P. coriacea P. mlokosewitschi. and two P. mascula subspecies, with the same origin and almost identical ITS sequences. Particularly, diploid P. broteri and tetraploid P. coriacea. whose common origin is supported by one substitution in matK (Fig. 9), have the same pattern of partial additivity of ITS sequences.
DNA sequence and morphological divergence. A comparison of sequence divergence between any one section and the other two indicates that section Oneapia is the roost divergent group within the genus, because it has the highest percent sequence divergence values of ITS (4.38), psbA-trnH intergenic spacer (3.74), and matK (1.33) (Table 5), suggesting that the earliest evolutionary split in Paeonia might have occurred between section Oneapia and the rest of the genus. Morphologically, section Oneapia is very distinct from the other sections by its small flowers (2-3 cm in diameter versus larger than 5 cm in sections Moutan and Paeonia) with fleshy and strongly concave petals.
Sequence divergence within sections is also highest in Oneapia, 2.94 for ITS, 2.30 for psbA-trnH intergenic spacer, and 0.74 for matK, supporting an oldest age of this section (Table 5). However, there are only very minor morphological differences between the two species of this section, P. brownii and P. californica. Since its recognition, P. californica had been treated as a synonym or variety of P. brownii until detailed morphological, ecological, and cytological studies (Stebbins, 1938c; Stebbins and Ellerton,1939). They suggested that the two species were likely to have differentiated genetically rather than to have undergone morphological modification due to different environmental stimuli. The two species are distributed allopatrically, i.e., P. californica is endemic to southern California from San Diego to Monterrey, whereas P. brownii is found from Santa Clara of California to British Columbia. Paeonia californica flowers from February to April, but P. brownii flowers during June and July. Paeonia californica is adapted to warmer and wetter climates, whereas P.brownii is semi-xerophytic able to grow up through banks of snow and complete the latter part of growth during the dry season (Stebbins, 1938c). The exact morphological differences between the two species, however, are not easily defined because variation between them is overlapping (Stebbins, 1938c). DNA sequence data support Stebbins' hypothesis that the two species have undergone considerable genetic divergence. Morphological evolution in section Oneapia. therefore, has been remarkably slow compared with such high level of sequence divergence (see Sytsma and Smith, 1992 for other examples).
Another major evolutionary split within Paeonia is separation of sections Moutan and Paeonia. Sequence divergence values between these two sections are not much higher than within section Oneapia. but morphological differences between them are pronounced. Within section Moutan, two distinct subsections are recognized in spite of quite low sequence divergence values. In subsection Delavayanae. P. lutea was also treated as a variety of P. delavayi because they differ only by flower color and slightly different sizes of leaflets (Finet and Gagnepain, 1904). This taxonomic treatment was supported by Stebbins (1938a) because these two taxa could be artificially hybridized easily to produce intermediate forms. He also suggested that some herbarium specimens have intermediate flower color and may be natural hybrids between the two taxa. For each of the three DNA regions sequenced here, the two taxa differ by one nucleotide substitution. Species-level problems of this group obviously require more careful field and population-level investigations.
Within subsection Vaginatae. three species studied by DNA sequences are morphologically distinct and allopatrically distributed. However, there is no nucleotide substitution among them in ITS sequences, P. spontanea and P. szechuanica have identical sequences of psbA-trnH intergenic spacer, and P. rockii and P. szechuanica have identical matK sequences. Morphological divergence in this case obviously exceeds DNA sequence divergence. Within section Paeonia, ITS sequence divergence is much higher than cpDNA divergence because one type of cpDNA has been lost after hybridization between two basal ITS clades (Sang et al., in prep.). Species in this section derived through divergent evolution are quite distinct morphologically from each other. Species of hybrid origin apparently have not accumulated novel nucleotide substitutions subsequent to hybridization. Only P. mascula ssp. hellenica has one autopomorphic substitution in - ITS relative to its parental species (Sang et al., in press). The hybrid species, therefore, are not distinguished primarily by divergent evolution, but by different combinations or differentiated segregation of parental features after hybridizations.
Biogeography. Paeonia occurs widely in five disjunct areas in the northern hemisphere. The endemic section Oneapia in western North America and the other two sections found in Eurasia form an intercontinental disjunction. Phylogenetically, separation of these two disjunct groups possibly represents the earliest evolutionary split within the genus. The biogeographical implication of this phylogenetic information is that the first major geographic isolation between ancestral populations of Paeonia occurred between Eurasia and western North America. The isolation may have resulted either from a vicariance event disrupting continuous distribution of the ancestral populations between Eurasia and western North America, or a long distance dispersal from one region to the other. The vicariance explanation is favored here because Paeonia, with follicle fruits and seeds having smooth surfaces and diameters of 7-13 mm, does not appear to have great dispersal ability.
Continuous distribution of ancestral populations of Paeonia between Eurasia and western North America is likely to have occurred through the Bering land bridge which allowed exchange of temperate plants between eastern Asia and western North America until late Tertiary or Quaternary (Wolfe, 1975, 1980; Tiffney, 1985). Disruption of this continuous distribution may have been due to climatic cooling at high latitudes and/or submergence of the Bering land bridge (Tiffney, 1985).The time of such a vicariance event can be estimated using a molecular clock. Time of divergence may be calculated as the value of DNA sequence divergence divided by twice the sequence divergence rate. For peonies, sequence divergence rates of the sequenced DNA regions are unknown, and cannot be estimated by either fossil records or biogeographic events. We can only use rates estimated in other plant groups. Divergence rates of ITS sequences have been calculated in several plant groups, but vary considerably among them (Suh et al.', 1993; Sang et al., 1995) . Further, ITS sequences are very short and may be subject to large statistical errors when used to calculate divergence times. ITS sequences, therefore, are probably not a good choice for use as a molecular clock. Divergence rates of matK sequences, which have not yet been estimated directly, have been suggested as being about twice as fast as that of rbcL sequences (Steele and Vilgalys, 1994). If a divergence rate of 1 X 10"9 per site per year is used for rbcL sequences (Zurawski et al., 1984; Zurawski and Clegg, 1987; Parks and Wendel, 1990) , a rate of 5 X 10'10 per site per year can be used for matK sequences. The divergence time between section Oneapia and the rest of the genus, therefore, is estimated to be 13.3 million year ago (Ma). This estimate, however, is subject to several sources of errors. First, the divergence rate of rbcL may not apply to peonies, because it can vary in different groups with different generation times (Li et al., 1987; Clegg, 1990; Gaut et al., .1992). Vegetative reproduction by rhizomes is very common in peonies, which may significantly prolong generation time, and consequently yield slower rates of DNA divergence. In this context, the divergence time may be estimated as more recent than it really is. Second, the estimation that matK evolves twice as fast as rbcL is very rough and may actually be different in peonies. Nevertheless, although estimation of divergence rates or times using the molecular clock hypothesis has largely been based on uncertain assumptions and roughly approximate values, it continues to be useful in helping understand tempos of evolution and historical biogeography (Parks and Wendel, 1990; Crawford et al., 1992; Wendel and Albert, 1992).
The estimated time for formation of the intercontinental disjunction in peonies, 13.3 Ma, is middle Miocene. Tiffney (1985) suggested that during the Miocene, temperatures at higher latitude allowed exchange of deciduous temperate plants via the Bering land bridge. Further, many herbaceous angiosperm groups evolved during the Miocene and exhibited an eastern Asian and eastern North American disjunct distribution (Tiffney, 1985). Formation of intercontinental disjunction in peonies, therefore, may well be a result of disruption of continuous distribution through the Bering land bridge during Miocene time. For greater accuracy, additional calculations from other DNA regions are needed. Addressing the question of whether this time indicates occurrence of a common vicariance event at the Bering land bridge awaits similar studies of additional taxa with eastern Asia-western America disjunctions.
Possible existence of ancestral populations of Paeonia in high latitudes around the Bering land bridge in Miocene is concordant with warm climate during this period (Potts and Behrensmeyer, 1992). But did Paeonia originate in this area in the Miocene? Because of lack of fossil record, one way to determine the age of peonies is by estimating divergence time between Paeonia and its sister group using the molecular clock. Unfortunately, the systematic position of Paeonia is problematical. A variety of families including Ranunculaceae, Berberidaceae, Magnoliaceae, Dilleniaceae, Glaucidiaceae have been suggested as close relatives Paeoniaceae (Hallier, 1905; Kumzawa, 1935; Corner, 1946; Sawada, 1971; Keefe and Moseley, 1978; Dahlgren, 1983; Melville, 1983). Recent phylogenetic analysis of rbcL sequences, however, places Paeonia as a sister group 01 Crassulaceae (Lehase et al., 1993). Uncertain systematic relationships of Paeonia may imply that Paeonia did not share a long evolutionary history with any existing related family, suggesting further that the time of its origin could be quite ancient and much earlier than Miocene. If this were the case, then ancestral populations of Paeonia might have beer rather homogeneous until the first evolutionary split in Miocene, or alternatively the ancestral populations might represent the only lineage that survived through evolution of this plant group. In this instance, ancestral populations of Paeonia may have originated somewhere else and migrated to the area near the Bering land bridge during the warm period in Miocene. In any event, even though the actual time and place of origin of the genus are unknown, we assume that populations ancestral to extant species of Paeonia were in the Bering land bridge area in Miocene.
After intercontinental separation of ancestral populations of Paeonia, those taxa in eastern Asia underwent another major divergent event, i.e., isolation between ancestral populations of sections Moutan and Paeonia. This event is likely to have taken place in eastern Asia where both sections currently occur. The shrubby section Moutan. endemic to high mountains in southwestern and central China, probably was never dispersed into Europe after its origin in eastern Asia. The largest section Paeonia. however, has been dispersed most widely and thus has undergone the highest level of speciation (Fig. 4) A marked difficulty in understanding historical biogeography of the northern hemisphere is that shifts of distributional ranges and extinction of taxa have been caused by Pleistocene glaciation (Noonan, 1988). Fossil records seem to be the only source of evidence for detection of such historical processes. Lack of a fossil record, therefore, may seriously hinder development of biogeographic hypotheses. Recent molecular phylogenetic and biogeographic studies in peonies have suggested a possible new avenue for solving this problem (Sang et al., in press). Documentation of reticulate evolution in section Paeonia reveals that most hybrid species are found in the Mediterranean region, whereas their parental species are restricted to Asia. Detection of parental type of DNA <sequences in species of hybrid origin, therefore, provides gene records for suggesting historical Mediterranean distributions of the Asian peony species. It has also been suggested that the extensive reticulate evolution may have been triggered by Pleistocene glaciation when parental species were forced into refugia in the Mediterranean region where they became sympatric and hybridized (Stebbins, 1948; Sang et al., in press).
Based on phylogenetic reconstructions (Figs. 6), a group of Mediterranean species, P. clusii. P. rhodia, P. broteri. P. coriacea. P. mlokosewitschi, and P. mascula ssp. hellenica and ssp. mascula. apparently has the same hybrid origin. However, they are endemic to widely different areas in the Caucasus and .throughout the southern Mediterranean (Fig. 4). If this group of species had a single hybrid origin, then it seems unlikely that the hybrid populations would have been dispersed so widely in the southern Mediterranean region. It seems more likely that the ancestral hybrid populations migrated northward and extended their distributional ranges both eastward and westward during a glacial interval. During a subsequent glaciation, these populations might have been forced into isolated areas in the southern Mediterranean region. Geographic isolation consequently led to speciation among these populations and created the complex diversity now seen for this group of species.
Another group of Mediterranean species, P. arietina. P. humilis, P. officinalis,, and P. parnassica, which have the same ITS and cpDNA sequences, are also distributed in widely disjunct areas (Fig. 4). As discussed earlier, they are also possibly of hybrid origin based on meiotic behavior of their chromosomes (Stebbins, 1938b; Tzanoudakis, 1977), although molecular data so far provide no support. If they had the same hybrid origin, their distributional patterns may have been reached in the same way as suggested above. If this group of species is also considered to be of hybrid origin, then all the Mediterranean species within the genus are of hybrid origin. Their direct or indirect parental species are all found currently in eastern Asia or central Asia (only P. anomala).
It is striking that after hybridization, European populations of the present Asian species were completely replaced by their hybrids. Extensive hybridization of peony species must have produced a wide spectrum of different genome combinations upon which natural selection could act. Hybrids that adapted to drastic climatic changes during Pleistocene in Europe are currently distributed in the Mediterranean region. Asian species did not survive such changes in Europe, and their distributions became more restricted. Hybrid species, P. xinjiangensis. P. emodi. and P. sterniana. may represent footprints of the eastern Asian species, P. lactiflora. P.veitchii. and P. mairei, when their distributional ranges became reduced to only eastern Asia. Eastern Asia was much less seriously affected by Pleistocene glaciation and may have provided refugia for these peony species (Potts and Behrensmeyer, 1992; Tao, 1992).