| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
2 Department of Plant Biology, University of Illinois, Urbana, Illinois 61801 USA; and 3 Department of Plant Systematics and Geography, Warsaw University, Aleje Ujazdowskie 4, 00-478 Warszawa, Poland
Received for publication September 11, 1998. Accepted for publication April 20, 1999.
ABSTRACT
The evolutionary relationships among members of Apiaceae (Umbelliferae) tribe Scandiceae and representatives of all major lineages of Apioideae (including putatively allied Caucalideae) identified in earlier molecular studies were inferred from nucleotide sequence variation in the internal transcribed spacer regions (ITS1 and ITS2) of nuclear ribosomal DNA. In all, 134 accessions representing 18 genera commonly treated in Scandiceae were analyzed. Phylogenies estimated using maximum parsimony and distance methods were generally similar and suggest that: (1) Scandiceae form a well-supported clade, consisting of the genera Anthriscus, Athamanta (in part), Balansaea, Chaerophyllum, Conopodium, Geocaryum, Kozlovia, Krasnovia, Myrrhis, Myrrhoides, Neoconopodium, Osmorhiza, Scandix, Sphallerocarpus, and Tinguarra; (2) Athamanta is polyphyletic, with A. della-cellae allied with Daucus and A. macedonica placed close to Pimpinella; and (3) Rhabdosciadium and Grammosciadium find affinity with the Aegopodium group of umbellifers, whereas the placement of the monotypic Molopospermum cannot be inferred because of its high sequence divergence. The genus Bubon has been restored with two new combinations, B. macedonicum subsp. albanicum and B. macedonicum subsp. arachnoideum. Scandiceae arise within paraphyletic Caucalideae, the latter comprising two major lineages whose relationships to Scandiceae are not clear. Therefore, a broad treatment of Scandiceae is proposed, with subtribes Scandicinae, Daucinae, and Torilidinae (the latter two representing the Daucus and Torilis subgroups, respectively, of recent molecular systematic investigations).
Key Words: Apiaceae • Apioideae • molecular phylogeny • nuclear ribosomal DNA internal transcribed spacers • Scandiceae • Umbelliferae
Apiaceae Lindl. (Umbelliferae Juss.) is one of the best known families of flowering plants. Its members include many commonly grown vegetables (e.g., carrot, parsnip, and celery/celeriac) and condiments (e.g., coriander, anise, caraway, chervil, cumin, parsley, and dill). They owe their distinctive flavor largely to diverse volatile compounds in the fruits and leaves, which not only account for their culinary use but for their wide application in medicine. The family also encompasses widespread weeds and toxic plants, including the notorious poison hemlock used in ancient Athens to execute those sentenced to death, the most famous victim being Socrates. Fortunately, such misuses of umbellifers have been rare and the Apiaceae stand out as a family of great economic importance. Despite its long taxonomic history dating back to Morison's (1672) Plantarum umbelliferarum, the earliest systematic study of any group of plants (Constance, 1971
), the family still awaits a modern classification. The most recent treatment of umbellifers (Pimenov and Leonov, 1993
) is but an adaptation of the century-old system of Drude (1898)
, highly criticized for using subtle or poorly defined diagnostic characters (Heywood, 1982a
). Several alternative classifications have also been proposed (de Candolle, 1830
; Bentham, 1867
; Koso-Poljansky, 1916
; Cerceau-Larrival, 1962
), however, apart from the use of Cerceau-Larrival's by some French authors, none has gained wide recognition.
A molecular approach has contributed much to the understanding of evolutionary relationships of Apiaceae. Phylogenetic analyses of the family using chloroplast DNA (cpDNA) sequences (Downie, Katz-Downie, and Cho, 1996
; Plunkett, Soltis, and Soltis, 1996
; Downie et al., 1998
), cpDNA restriction sites (Plunkett and Downie, 1999
), and nuclear ribosomal DNA internal transcribed spacer (ITS) sequences (Downie and Katz-Downie, 1996
; Downie et al., 1998
; Katz-Downie et al., 1999
) have revealed that two of Drude's three subfamilies, Saniculoideae and Apioideae, are each monophyletic and sister groups, while Hydrocotyloideae are polyphyletic (containing some members allied to Araliaceae and others to Apioideae and Saniculoideae). Nevertheless, the hitherto proposed divisions of subfamily Apioideae appear to be unsound, with all tribes but one seemingly polyphyletic or paraphyletic. The only tribe that has sustained the test of molecular phylogenetics is Scandiceae (Downie et al., 1998
). However, only up to five genera in the group were included in each of these previous studies.
Apiaceae comprise 300–450 genera and 3000–3700 species (Constance, 1971
; Pimenov and Leonov, 1993
). Because of its large size, an approach to resolving the taxonomy of the family should therefore combine both a "high-level" analysis performed for a representative subset of the family as well as "low-level" revisions of particular genera and tribes. The first aims to provide an outline of the classification and may help to formulate hypotheses on the evolution of the family, while the goal of the latter is to fill this framework with more detail and to give a deeper insight into the differentiation of these plants, including their life history strategies and evolutionary pathways of particular traits. The advantage of low-level analyses is also a smaller risk of inadequate sampling.
Scandiceae are well suited for such an approach. The tribe comprises ~20 genera with some 70–90 species largely confined to southwest Eurasia. Some of the genera are monotypic, and many have been recently revised, either regionally (Schischkin, 1950a
; Tutin et al., 1968
; Davis, 1972
; Hedge et al., 1987
) or worldwide (Engstrand, 1977
; Lowry and Jones, 1984
; Spalik, 1997
). Although several different classifications of the tribe have been proposed (Table 1), there is a general agreement as to which genera constitute the core of the tribe (i.e., Anthriscus, Chaerophyllum, Grammosciadium, Myrrhis, Osmorhiza, Rhabdosciadium, and Scandix). Drude (1898)
defined Scandiceae based on the crystals of calcium oxalate in the parenchyma cells surrounding the carpophore and divided it into two subtribes according to the shape of the fruit. His subtribe Caucalidinae (Caucalineae) was united with Dauceae and treated as tribe Caucalideae (Heywood, 1971
; Hedge et al., 1987
; Pimenov and Leonov, 1993
). Molecular analyses, however, have confirmed that Scandiceae sensu Heywood (1971)
and paraphyletic Caucalideae form a well-supported clade (Downie and Katz-Downie, 1996
; Downie et al., 1998
).
|
). Particular attention has been drawn to Anthriscus sylvestris, the source of a drug used in traditional Chinese medicine (Okuyama, Sakakibara, and Shibata, 1981
). These plants contain numerous active compounds (Kozawa, Morita, and Hata, 1978
; Kurihara et al., 1978
; Kurihara and Kikuchi, 1979
; Inamori et al., 1983, 1985
), some of which inhibit the proliferation of cancer cells in vitro (Ikeda et al., 1998a, b
).
The position of several genera in Scandiceae is dubious as they have also been placed in other tribes. Drude (1898)
underlined similarities between Geocaryum and Butinia (
Conopodium subgenus Butinia) from Apieae-Apiinae; however, he could have been referring to Butinia capnoides, which has also been placed in Chaerophyllum (Bentham, 1867
; Hedge and Lamond, 1980
) and eventually in a separate genus, Neoconopodium, within the Scandiceae (Pimenov and Kljuykov, 1987
). Calestani (1905)
placed Geocaryum in tribe Bunieae, together with Conopodium and Bunium; such a treatment was also adopted by Engstrand (1973, 1977)
in his revision of Geocaryum, contrary to Spalik (1997)
who allied it with core Scandiceae. Calestani (1905)
also included several members of Apieae-Apiinae in Scandiceae, while removing some genera, like Anthriscus and Myrrhoides. He was followed by Koso-Poljansky (1916)
, who added genera recognized presently in tribe Echinophoreae. Molopospermum peloponnesiacum is the only member of Scandiceae with distinctly winged fruits and has been placed either in Smyrnieae (Bentham, 1867
; Pimenov and Leonov, 1993
), Apieae (Calestani, 1905
), or in a separate tribe close to Smyrnieae (Cerceau-Larrival, 1962
). Kozlovia is recognized either in Scandiceae (Heywood, 1971
; Pimenov and Leonov, 1993
) or in Caucalideae (Heywood, 1982b
; Hedge et al., 1987
). Bentham (1867)
, admitting the similarity between Tinguarra and Athamanta, placed the former in Scandiceae and the latter in Seseleae; this treatment was also adopted by Drude (1898)
. Athamanta was transferred to Scandiceae by Cerceau-Larrival (1962)
based on pollen and cotyledon morphology. Monotypic Balansaea was recognized in Scandiceae as a separate genus (Heywood, 1971
) or included in Geocaryum (Drude, 1898
; Pimenov and Leonov, 1993
). Balansaea was also synonymized with Conopodium (Engstrand, 1973
), which is usually placed in Apieae. Although the affinities of Grammosciadium and Rhabdosciadium to Scandiceae are rarely questioned, Hedge and Lamond (1987)
indicated that both genera differ from other members of the tribe in not having a deeply sulcate endosperm. Tamamschian and Vinogradova (1969)
suggested that Grammosciadium occupies an intermediate position between Scandiceae and Caucalideae.
The major objectives of this study were: (1) to ascertain the monophyly of Scandiceae and its relationship to the other currently recognized lineages of Apiaceae inferred from molecular studies; (2) to verify the monophyly of the largest genera, i.e., Anthriscus, Chaerophyllum, and Osmorhiza; and (3) to ascertain the evolutionary affinities of the monotypic members, i.e., Balansaea, Myrrhis, Myrrhoides, Kozlovia, Krasnovia, and Sphallerocarpus. The reassessment of qualitative morphological characters and the interpretation of their evolution will be presented in a subsequent study.
MATERIALS AND METHODS
Plant accessions
We have chosen to examine variation in nuclear ribosomal DNA (rDNA) internal transcribed spacers (ITS) sequences, regions that have been useful in estimating infrafamilial relationships in Apiaceae (Downie and Katz-Downie, 1996
; Downie et al., 1998
; Katz-Downie et al., 1999
) as well as in other groups of angiosperms (reviewed in Baldwin et al., 1995
). A total of 134 accessions representing 64 genera and 119 species was considered. Complete ITS1 and ITS2 sequences for 85 accessions representing 29 genera and 73 species are reported here for the first time (Table 2); ITS data for the remaining 49 accessions were published previously (Downie et al., 1998
). The ingroup has been chosen based on the analysis of different classification systems of Scandiceae (Table 1). The most important accounts of the tribe are those of de Candolle (1830)
, Bentham (1867)
, Boissier (1872)
, Drude (1898)
, Calestani (1905)
, Koso-Poljansky (1916)
, Cerceau-Larrival (1962)
, Heywood (1971)
, Hedge et al. (1987)
, and Pimenov and Leonov (1993).
We have omitted those sources in which Scandiceae sensu Drude (1898)
were not divided into Scandicinae and Caucalidinae, as in Schischkin (1950a)
. Not all of the cited accounts give a complete treatment of the tribe. Moreover, older accounts (i.e., de Candolle, 1830
) lack some later described genera. Boissier (1872)
, Calestani (1905)
, and Hedge et al. (1987)
revised regional floras, while Heywood (1971)
considered solely Old World umbellifers. The revision of Cerceau-Larrival (1962)
contained merely a sketch of the system, listing some six genera in Scandiceae. The most influential classification was that of Drude (1898)
, and the accounts of Heywood (1971)
, Hedge et al. (1987)
, and Pimenov and Leonov (1993)
generally follow it.
|
, for instance, is in fact synonymous with Chaerophyllum. The real puzzle, due to somewhat obscure synonymy and controversial taxonomic decisions, is the classification of Koso-Poljansky (1916)
. He included Osmorhiza (as Uraspermum) in Scandix while retaining Glycosma, which is now recognized as a subgenus of Osmorhiza. He also divided Grammosciadium into two genera and included Falcaria, Ptychotis, and Hladnikia into one of them. Consequently, the comparison of classifications based solely on the lists of genera may be highly misleading. The summary presented in Table 1 is based therefore on the lists of species included in particular genera.
Representatives of some 40 currently recognized genera have been placed in Scandiceae; eight of these, however, have already been excluded from the tribe based on molecular studies (Downie and Katz-Downie, 1996
; Plunkett, Soltis, and Soltis, 1996
; Downie et al., 1998
; Lee and Downie, 1999
) and are therefore omitted from Table 1. These are Cryptotaenia (= Deringa; Koso-Poljansky, 1916
), Cuminum (Calestani, 1905
), Falcaria (Calestani, 1905
; Koso-Poljansky, 1916
, in Prionitis), Petagnaea (= Heterosciadium; Koso-Poljansky, 1916
), Physospermum (Koso-Poljansky, 1916
, in Chaerophyllum), Rhodosciadium (= Velaea; de Candolle, 1830
), Tauschia (Koso-Poljansky, 1916
), and Yabea (Koso-Poljansky, 1916
). From among those remaining, we generally considered all genera that have been classified in Scandiceae by more than one author and included in one of the more recent treatments. Several genera have been recognized in the tribe only once, mostly by Calestani (1905)
or Koso-Poljansky (1916)
. These are Anisosciadium (Koso-Poljansky, 1916
), Bunium (Koso-Poljansky, 1916
, = Elwendia), Chaerophyllopsis (Heywood, 1971
), Echinophora (Koso-Poljansky, 1916
), Kundmannia (Calestani, 1905
), Microsciadium (Calestani, 1905
), Ottoa (Bentham, 1867
), Portenschlagiella (Calestani, 1905
), Pycnocycla (Koso-Poljansky, 1916
), Rhopalosciadium (Pimenov and Leonov, 1993
), and Scaligeria (Bentham, 1867
, in Conopodium). While most of these genera were omitted from our analysis, we did include members of Bunium and Scaligeria (sensu lato, i.e., including Elaeosticta) to test for the monophyly of Bunieae sensu Calestani (1905)
, and four accessions of tribe Echinophoreae (Echinophora, Pycnocycla, and Dicyclophora), whose members have not been previously analyzed. We omitted Chaerophyllopsis, a monotypic endemic of China recognized in Scandiceae by Heywood (1971)
, and Rhopalosciadium, transferred to Scandiceae from Caucalideae by Pimenov and Leonov (1993)
, as both species were not available for molecular study (the latter is known exclusively from the type collection).
In summary, the 134 accessions examined reflect 64 genera and 119 species, including 18 genera (with 67 species and 82 accessions) classified more than once in Scandiceae. Fifteen genera, including 11 of Scandiceae, were represented by more than one accession. Due to limited availability of material, different proportions of taxa from particular genera were included in the analysis. The most extensively sampled genera were Osmorhiza, Anthriscus, Tinguarra, and Neoconopodium, which included 90–100% of all currently recognized species, subspecies, and varieties, while Geocaryum, Rhabdosciadium, and Conopodium were not so well represented. Although only four of a maximum 20 species of Scandix were analyzed, these included those species retained in major taxonomic treatments (e.g., Cannon, 1968
; Hedge and Lamond, 1972, 1987
).
Close outgroups include members of tribes Caucalideae (i.e., Caucalis, Chaetosciadium, Cuminum, Daucus, Orlaya, Pseudorlaya, Torilis) and Laserpitieae (Laserpitium, Monizia). Also, representatives of all major lineages of Apioideae identified by earlier molecular studies have been included (Plunkett, Soltis, and Soltis, 1996
; Downie et al., 1998
). We have added the sequences of Nirarathamnos and Naufraga, monotypic endemics whose affinities are unclear, and of Deverra triradiata, a representative of a genus occurring in arid regions of Africa and the Arabian Peninsula. The sequence of Heteromorpha arborescens, a basal apioid identified by earlier studies (Rodríguez, 1971
; Plunkett, Soltis, and Soltis, 1996
; Downie et al., 1998
), has been included to root the trees.
DNA extraction, PCR amplification, and sequencing
Total genomic DNAs were extracted from dried leaf or flower material using the modified hexadecyltrimethylammonium bromide (CTAB) protocol of Doyle and Doyle (1987)
or the Plant DNeasy Extraction Kit (Qiagen, Inc., Valencia, California). Some samples required further purification by centrifugation to equilibrium in cesium chloride-ethidium bromide gradients or by applying Qiagen genomic-tip purification columns. These methods allowed us to sample fairly old accessions: many specimens were over 30 yr old, while three taxa, Anthriscus sylvestris subsp. fumarioides, Chaerophyllum magellense, and Neoconopodium capnoides, were represented by century-old collections. Details of the PCR amplification reactions and sequencing strategies are provided elsewhere (Downie and Katz-Downie, 1996
; Downie et al., 1998
). The whole ITS region was amplified using primers ITS4 and ITS5 (White et al., 1990
). For one accession (Chaerophyllum nivale), each spacer region was amplified separately using internal primers ITS2 and ITS3. For some samples, particularly those isolated from old herbarium material, the annealing temperature was lowered from 53°C to 46°C or even to 37°C. Each PCR product was then electrophoresed in a 1% agarose gel and stained with ethidium bromide. The DNA band was subsequently excised and eluted using the Elu-Quick DNA Purification Kit (Schleicher & Schuell, Keene, New Hampshire) or the QIAEX II Gel Extraction Kit (Qiagen). Both manual and automated sequencing strategies were used. Cycle sequencing reactions were performed using the purified PCR product, AmpliTaq DNA polymerase, and fluorescent dye-labeled terminators (Perkin-Elmer Corp., Norwalk, Connecticut). The products were resolved by electrophoresis using Applied Biosystem's, Inc. (Foster City, California) 310 and 373A DNA sequencing systems.
Sequence analysis
Alignment of the sequences was done manually starting from a subset of the data matrix from previously published studies (Downie and Katz-Downie, 1996
; Downie et al., 1998
) and consecutively adding new accessions. The initial data matrix was aligned using the program CLUSTAL V (Higgins, Bleasby, and Fuchs, 1992
) and then adjusted manually when necessary. Boundaries of the coding and spacers regions were determined by comparison of the sequences to the respective boundaries in Daucus carota (Yokota et al., 1989
). Positions with ambiguous alignment were excluded from the analysis. Pairwise nucleotide differences were determined using the distance matrix option in PAUP version 3.1.1 (Swofford, 1993
). All gaps were treated as missing data, while the sequence divergence was calculated as a proportion of different sites without taking into account multiple substitutions. Transition/transversion ratios were calculated using MacClade version 3.01 (Maddison and Maddison, 1992
) over a subset of the maximally parsimonious trees. The sequences reported in this study have been deposited with GenBank (Table 2), and their alignment may be obtained directly from the authors.
Phylogenetic analysis
The resulting data matrix was analyzed using PAUP version 3.1.1 (Swofford, 1993
) or PAUP* version 4.0.0d64 (D. Swofford, Smithsonian Institution, Washington, DC) run on Power Macintosh computers. Unordered characters states were assumed. To locate possible islands of most parsimonious trees, 500 heuristic searches were initiated using random addition starting trees, with TBR (tree bisection-reconnection) branch swapping, mulpars, and steepest descent selected. Only two shortest trees were saved from each search; these trees were subsequently used as starting trees for TBR branch swapping. This search was stopped when the number of trees reached the memory limit of 12 000. The strict consensus of these trees served as a topological constraint in a further heuristic search using the inverse constraint approach of Catalán, Kellogg, and Olmstead (1997). This time, 5000 searches were initiated, saving no more than two trees per replicate. However, in this analysis, only those trees that did not fit the constraint tree were saved. Since no additional trees, shorter or equal to those previously obtained, were found, this then suggests that the strict consensus tree satisfactorily summarizes the available evidence, even though the exact number of trees at that length is not known.
For bootstrap analysis, 100 resampled data sets were generated and, for each, 500 heuristic searches were initiated, saving no more than two shortest trees per search, and no more than 100 trees per bootstrap replicate. The frequencies of particular groups of taxa were then calculated over the set of all trees saved. Decay values were estimated using the converse constraint method of Baum, Sytsma, and Hoch (1994)
. Each clade was defined as a constraint tree and a heuristic search with 100 replicates was performed with the converse of the constraint tree enforced; TBR branch swapping was selected and mulpars turned off. The decay value for the clade was calculated as the difference between the length of the shortest trees for the converse constraint of that clade and the length of the most parsimonious trees previously obtained.
Distance trees were obtained from neighbor-joining analyses, estimated with the variety of distance measures available in PAUP* version 4.0.0d64 (such as the two- and three-parameter methods of Kimura, the Jukes-Cantor method, and maximum likelihood distance), using an IBM ThinkPad 365X computer. A bootstrap analysis was performed using 1000 resampled data sets.
RESULTS
Sequence analysis
Of the 134 umbellifer accessions examined (including 18 genera with 67 species and 82 accessions classified more than once in Scandiceae), the following four pairs of accessions each had identical ITS sequences and were each represented by one terminal taxon in the phylogenetic analysis: Rhabdosciadium aucheri (two accessions); Osmorhiza depauperata (two accessions); Chaerophyllum macrospermum and C. hakkiaricum; Athamanta cretensis and A. turbith. Due to difficulty in aligning the sequence of Molopospermum peloponnesiacum it too was eliminated. Therefore, only 129 sequences were retained, and their alignment resulted in a matrix of 499 characters (Table 3). As the alignment of ten positions in ITS1 and 14 positions in ITS2 were ambiguous, these positions were excluded from the analysis. No evidence of obvious ITS length polymorphism within each accession examined was found.
|
|
|
|
|
two-parameter distance, is presented in Fig. 4. Similar trees (not shown) were obtained using other distance measures, such as Jukes-Cantor, Kimura's three-parameter, and maximum likelihood. Selecting gamma distribution as an approximation of substitution rate among sites and trying different values of the parameter a did little to change the topology of the trees, other than collapsing or rearranging a few weakly supported clades.
|
, Downie et al. (1998)
, and Plunkett and Downie (1999)
. These include the Daucus clade (with three subgroups, identified herein as the Scandix, Daucus, and Torilis subclades), the Aciphylla clade, the Angelica clade, the Apium clade, the Aegopodium clade, and the Oenanthe clade. Additionally, three other minor clades occur, and are identified as the Pimpinella, Physospermum, and Conioselinum clades. Scandiceae sensu stricto are equivalent to the Scandix subclade, while the Daucus and Torilis subclades constitute, to a large extent, tribe Caucalideae. The results of the maximum parsimony analysis (Fig. 2) suggest that Scandiceae sensu stricto (the Scandix subclade) arise within paraphyletic Caucalideae, with the Daucus subclade being the closest relative and the Torilis subclade representing the next most basal branch. The distance-based analyses (e.g., Fig. 4) also support paraphyletic Caucalideae, but the relationship of the Daucus and Torilis subclades to each other is not clear, as the putative sister-group relationship between the Scandix and Daucus clades is very weakly supported. Although Scandiceae form a well-supported monophyletic group, several taxa previously included in the tribe are placed elsewhere. Two genera, Grammosciadium and Rhabdosciadium, fell within the Aegopodium clade, with the former sister to Carum. Athamanta is polyphyletic. Although the type of the genus, A. cretensis, is retained in Scandiceae, another European species, A. macedonica, clustered with Arafoe aromatica and Pimpinella peregrina in the Pimpinella clade. The North African Athamanta della-cellae was placed alongside Daucus and Pseudorlaya in the Daucus subclade. The positioning of the various Athamanta species was not only supported by high bootstrap and decay values (Fig. 2) but also by synapomorphic indels (Fig. 3). Athamanta cretensis and A. turbith, indistinguishable so far as ITS sequences are concerned, formed a strongly supported group with Tinguarra, Conopodium, and Balansaea; however, the phylogenetic relationships within this clade are ambiguous. In the neighbor-joining tree (Fig. 4), Tinguarra and Athamanta form a clade, denoted as the Athamanta group, and arise within a paraphyletic Conopodium. There is little support for the monophyly of Tinguarra, as T. sicula is sister to Athamanta cretensis/A. turbith and not to the two remaining congeners. Conopodium bourgaei is placed as a basal branch in this clade; however, since all internal nodes in this clade are weakly supported, the rejection of monophyly of Conopodium may be unsound. Therefore, an additional heuristic search was initiated with Conopodium and Balansaea constrained to form a clade sister to Athamanta and Tinguarra; the shortest trees inferred in this search were only two steps longer than those obtained without the constraint invoked.
Of the 11 genera of Scandiceae represented by more than one accession, eight are monophyletic (Table 4). Athamanta is polyphyletic, while the monophyly of each of Conopodium and Tinguarra is dubious. With the exception of Scandix, sequence divergence estimates within each of the monophyletic genera were relatively low, with mean divergence values ranging from identity to 6.5% of nucleotides (Table 4).
In all analyses, the Athamanta-Conopodium clade (e.g., Fig. 4) is sister to all other members of Scandiceae sensu stricto. The position of the monotypic Sphallerocarpus, however, is equivocal. It is either placed, by itself, as the next branch up the tree (Fig. 4) or sister to Chaerophyllum and Myrrhoides (Fig. 2). The placement of Myrrhoides is also ambiguous, as it is either included in Chaerophyllum (Fig. 4) or placed as its sister taxon (Fig. 2). Nevertheless, both Chaerophyllum and Myrrhoides comprise a well-supported monophyletic group.
Seven genera of Scandiceae sensu stricto belong to a highly supported clade that is sister group to the genus Scandix. However, the relationships among these seven genera are unclear. Monotypic Kozlovia and Krasnovia, and both species of Neoconopodium, form a relatively well supported lineage, but the affinities of Anthriscus, Myrrhis, Geocaryum, and Osmorhiza are unclear. For instance, maximum parsimony (Fig. 2) favors Myrrhis as sister to Geocaryum while the distance trees (e.g., Fig. 4) ally it with Osmorhiza.
Infrageneric relationships within the three most extensively sampled genera—Chaerophyllum, Osmorhiza, and Anthriscus—are also ambiguous. Most species of Chaerophyllum form a highly supported but poorly resolved branch (herein called the C. aureum group; Fig. 4). The remaining species comprise two other major branches, herein called the C. hirsutum group and the C. temulentum group (Fig. 4). The latter group includes the North American representatives of the genus, C. procumbens and C. tainturieri. Although the monophyly of Osmorhiza is well supported, most taxa form a polytomous branch in the strict consensus tree (Fig. 2). It is noteworthy, however, that the basal species, O. aristata, is the only Old World member of the genus, while the next branch contains O. longistylis and O. claytonii, both of eastern North America. The remaining species of Osmorhiza constitute a poorly resolved group, which also includes O. occidentalis, usually placed in the monotypic subgenus Glycosma. Our study fails to confirm unequivocally close affinities between the two subspecies of O. mexicana. Anthriscus is likely monophyletic, although with bootstrap values of 51% or less (Figs. 2, 4) its monophyly is not strongly supported. Moreover, the relationships among its species are ambiguous. Two annual species, A. caucalis and A. cerefolium, appear to be unrelated, while A. sylvestris is paraphyletic with A. lamprocarpa, A. schmalhausenii, and A. nitida arising within. A basal branch in the A. sylvestris clade (A. sylvestris subsp. sylvestris or A. keniensis H. Wolff) is from Tanzania, while the European accessions of A. sylvestris subsp. sylvestris unite with European A. sylvestris subsp. alpina and Levantine A. lamprocarpa.
A group of tuberous plants constituting Calestani's (1905)
tribe Bunieae (i.e., Geocaryum, Bunium, and Conopodium, to which Engstrand (1973
) included Balansaea) is not supported as monophyletic. The genera arise in at least three distant lineages. Bunium, together with Scaligeria and its segregate Elaeosticta, is placed close to Trachyspermum ammi in the Aegopodium clade, and Conopodium and Geocaryum belong to two separate branches within Scandiceae. Balansaea is indeed closely related to Conopodium, although monophyly of the latter has not been confirmed.
The four examined members of tribe Echinophoreae (i.e., Echinophora, Pycnocycla, and Dicyclophora), are sister group to the clade comprising Nirarathamnos and a yet to be described taxon from Socotra, tentatively considered a species of Peucedanum (M. Watson, Royal Botanic Garden Edinburgh, UK, unpublished data). This clade is allied with Heracleum, Pastinaca, Aethusa, and Peucedanum morisonii, all members of the Angelica clade. Although the Angelica clade is relatively well supported by high bootstrap and decay values, it is noteworthy that its members do not share any particular length mutation (Fig. 3). Deletion J, which would otherwise be synapomorphic for the whole branch including the Apium and Pimpinella clades, is lost in Echinophoreae, Nirarathamnos, and a Peucedanum species. Monizia edulis, an endemic of Madeira, is placed in the Daucus subclade. Naufraga balearica is sister to Apium graveolens, and Deverra triradiata is sister to Petroselinum. These last four genera all fall within the Apium clade.
DISCUSSION
Comparison to previous treatments
Existing classifications of Scandiceae are generally quite similar. Drude's (1898)
system, containing elements from earlier authors, such as de Candolle (1830)
and Bentham (1867)
, is most commonly used despite its imperfections. The later accounts of Heywood (1971)
, Hedge et al. (1987)
, and Pimenov and Leonov (1993)
did not introduce many changes, sometimes reverting to earlier treatments (e.g., Boissier, 1872
; Bentham, 1867
) with respect to the exclusion of Molopospermum. Notably different is the classification proposed by Koso-Poljansky (1916)
, based on a revision of European umbellifers by Calestani (1905)
. Both authors excluded several core members of Scandiceae, such as Anthriscus, and added some miscellaneous taxa, such as Falcaria, Ptychotis, and Hladnikia. The system of Cerceau-Larrival (1962)
differs from all others in the inclusion of Athamanta and Conopodium.
Phylogenetic analyses of ITS sequences reveal that the majority of Scandiceae representatives examined constitute a monophyletic taxon (designated herein as the Scandix subclade or Scandiceae sensu stricto). This monophyly is supported by relatively high bootstrap and decay values and a low average sequence divergence. Of the 18 putative ingroup genera included in this study, very few have been eliminated from the tribe. These eliminated taxa include Grammosciadium, Rhabdosciadium, and two species of Athamanta. The position of Molopospermum peloponnesiacum could not be determined with certainty; phylogenetic analyses of those ITS sequence characters that could be readily aligned with the matrix suggest an affinity with Physospermum. It does appear, however, that the Molopospermum ITS sequence is quite divergent from all other Scandiceae representatives. Those taxa that should be maintained within the tribe, as a result of this investigation, include Anthriscus, Athamanta (in part), Chaerophyllum, Conopodium (including Balansaea, discussed below), Geocaryum, Kozlovia, Krasnovia, Myrrhis, Myrrhoides, Neoconopodium, Osmorhiza, Scandix, Sphallerocarpus, and Tinguarra. The classification of Koso-Poljansky (1916)
has found little support from these molecular data. Surprisingly, the treatment of Drude (1898)
appears to be quite good, with only a few genera misplaced. The only major deficiency is his exclusion of Athamanta and Conopodium from the tribe. In this study, these taxa fall basal within Scandiceae.
Major clades identified in Scandiceae sensu stricto
Nine distinct clades, six of which are equivalent to generally recognized genera, are distinguished within the tribe. These clades are identified in Fig. 4.
The close relationship among Athamanta, Tinguarra, and Conopodium is a rather unexpected conclusion of this study as these genera appear to be morphologically distinct. Athamanta and Tinguarra are hemicryptophytes while Conopodium is a geophyte and, as a result, usually allied with Bunium. These taxa, however, share a West/Central Mediterranean distribution, unlike the remaining Scandiceae, which have an East Mediterranean/Central Asian distribution. The relationships within this clade are unresolved; better sampling of Conopodium may improve the resolution. It is rather unlikely that Conopodium is paraphyletic with regard to Athamanta and Tinguarra, as its characteristic morphology strongly suggests monophyly. The genus Balansaea should probably be included in Conopodium, as proposed by Engstrand (1973)
. Similarly, the monophyly of Tinguarra is dubious. Tinguarra sicula should be recognized in Athamanta, as in the treatment for the Flora Europaea (Tutin, 1968
) but contrary to later authors (e.g., Knees, 1996
). The relationship between T. montana and T. cervariifolia is also poorly supported. Based on further ITS sequence comparisons, another endemic of the Canaries, Todaroa aurea, is sister to Tinguarra cervariifolia (M. Watson, K. Spalik, and S. Downie, unpublished data), although the latter was the only species of Tinguarra included in that investigation. Tinguarra montana was originally described in Todaroa. Tinguarra, Todaroa, and Athamanta are undoubtedly closely related and, perhaps, should be recognized as a single genus. However, as revealed in this study, the delimitation of Athamanta is open to question, and that decision would require a better sampling of Athamanta combined with a closer look at the morphology and anatomy of these genera. Such a study is under way, and it should confirm whether these two genera deserve to be kept separate or included in Athamanta.
The monotypic Sphallerocarpus appears to occupy an isolated position in Scandiceae. This isolation parallels its geographic distribution, for it occurs in Siberia, Northeast China, and Korea, far removed from the center of diversity of Scandiceae in the Mediterranean region. The position of Sphallerocarpus varies depending upon the method of analysis used. While the results of the maximum parsimony analysis place it, albeit with weak bootstrap support, as sister to Chaerophyllum and Myrrhoides (Fig. 2), the neighbor-joining analysis places it away from this group (Fig. 4).
Chaerophyllum is the largest genus in Scandiceae. However, contrary to other large genera in the tribe, it has not been recently revised. The only modern account, which includes infrageneric divisions, is that of Schischkin (1950b)
who recognizes three subgenera: Nomochaerophyllum (= Chaerophyllum), Golenkinianthe, and Buniomorpha. As the genus includes more than 30 species, a natural infrageneric classification would be beneficial. However, the classification of Schischkin (1950b)
is not supported by the molecular data. Subgenus Nomochaerophyllum comprises representatives of all distinct lineages within Chaerophyllum identified in this study, and many of its distinguishing morphological features have been found to be homoplastic (K. Spalik and S. Downie, unpublished data). The ITS data suggest that Chaerophyllum comprises three distinct clades. Two clades, exemplified by C. temulentum and C. hirsutum, comprise only a few species each, while the majority belong to the C. aureum group (Fig. 4). The C. temulentum group, in addition to C. temulentum, includes the two American representatives of this genus: C. tainturieri and C. procumbens. The taxa constituting the C. hirsutum group are generally regarded as being closely related because they are so morphologically similar. The C. aureum group is much diversified with respect to its habit, life history, and ecology. As a consequence, it includes representation of all three subgenera of Chaerophyllum. Myrrhoides is unequivocally allied with Chaerophyllum, although its position here is somewhat ambiguous. The phylogeny inferred using parsimony confirms its position as a separate lineage sister to Chaerophyllum, while distance methods place it inside the latter, close to the C. temulentum group. This group includes both annual and biennial taxa somewhat similar in habit to annual Myrrhoides nodosa.
Scandix is clearly defined by its long-beaked fruits and an annual habit. Its monophyly is unambiguous. Scandix is sister to a large clade encompassing Osmorhiza, Myrrhis, Geocaryum, Kozlovia, Krasnovia, Neoconopodium, and Anthriscus. Monophyly of this group is supported by morphology (K. Spalik and S. Downie, unpublished data), and its members have already been considered closely allied (Pimenov and Kljuykov, 1987
; Spalik, 1997
).
Central Asiatic Kozlovia, Krasnovia, and Neoconopodium unite, albeit with weak support. They also share a geophytic habit. Although close relatives of Geocaryum have not been previously identified, this genus is also characterized by underground tubers. Geocaryum is taxonomically complex, with these difficulties partly accounting for its poor representation in this study. Ball (1968)
recognized only three species, while Engstrand (1977)
has shown that the most widespread G. cynapioides is diversified both morphologically and in chromosome number. By carrying out a series of crosses he has shown that there is a genetic barrier among many populations once regarded as conspecific. As a result, he raised the number of species to 13. This revision, however, was based mostly on variation observed in cultivated material; he did not provide a list of representative herbarium specimens seen, material that would have made the identification of species less difficult. Both of our accessions of Geocaryum were originally determined as G. cynapioides, which according to Engstrand (1977)
does not occur in the area in which they were collected. Based on their distribution and morphology, we have identified these accessions as G. macrocarpum, however, the high sequence divergence between them suggests that they may actually represent different species. The position of Myrrhis is not clearly resolved; in the strict consensus tree (Fig. 2) it allies, with weak support, with Geocaryum, whereas in the neighbor-joining tree (Fig. 4) it shows affinity with Osmorhiza. Morphologically, Myrrhis is more similar to Osmorhiza than it is to Geocaryum.
Although the relationships within Osmorhiza are mostly unresolved, some hypotheses on their evolution can be presented based on the results obtained. Lowry and Jones (1984)
suggested that North America is both the center of diversity and place of origin for the genus. Although they did not provide a phylogenetic tree, their classification suggests that the most basal member within the genus might be O. occidentalis, the only representative of subgenus Glycosma. The remaining taxa constitute three lineages, i.e., sections, the typical section Osmorhiza containing Euroasiatic O. aristata and two morphologically similar eastern North American species, O. claytonii and O. longistylis. This hypothesis has not been confirmed. In contrast, we provide evidence for the Asiatic origin of the genus, as O. aristata is sister to all other examined species. The two other representatives of sect. Osmorhiza, O. claytonii and O. longistylis, constitute the next branch. Their similarity is therefore plesiomorphic. The trees inferred from these ITS data are therefore consistent with the hypothesis that the radiation of Osmorhiza occurred during the migration of its representatives from Asia and, in America, from north to south.
The genus Anthriscus was divided into three sections based on life history and habit (Spalik, 1996, 1997
), with section Anthriscus including three annuals, A. caucalis, A. cerefolium, and A. tenerrima. Based on our analyses of ITS data, this section is likely polyphyletic, as A. caucalis and A. cerefolium do not appear to be closely related. Similarly, the relationships within the A. sylvestris group (Fig. 4), although mostly unresolved, are also somewhat different from those inferred from morphology (Spalik, 1996
). The basal position of the African representatives and the close similarity of the European taxa suggest that the distribution of A. sylvestris in African montane "islands" are postglacial relics and that the differentiation of the taxa occurred while they migrated northwards from these refugia.
Taxa excluded from Scandiceae and their placement
According to our results, Grammosciadium and Rhabdosciadium belong to the Aegopodium group of umbellifers. Although these two genera have consistently been placed in Scandiceae, some have underlined the differences between them and the remaining members of the tribe. Their exclusion, therefore, is not totally surprising. Drude (1898)
recognized Grammosciadium in Scandiceae but placed Caropodium, a genus typified by C. meoides (= Grammosciadium platycarpum), in Apieae-Apiinae (his Ammineae-Carinae). Tamamschian and Vinogradova (1969)
regarded Grammosciadium as occupying an intermediate position between Scandicinae and Caucalidinae, while Hedge and Lamond (1987)
commented on its nonsulcate endosperm, so atypical in Scandiceae. More recently, Vinogradova (1995)
transferred the genus to Apieae and suggested that Fuernrohria setifolia is its closest relative. Based on phylogenetic analysis of ITS data, Fuernrohria also belongs to the Aegopodium clade (Katz-Downie et al., 1999
). Rhabdosciadium is a poorly known genus of which only limited material is available; this material, however, usually lacks leaves and mature fruits. The few fruits available for sectioning show a flat commissural face, as opposed to a deeply sulcate commissural face characteristic of other Scandiceae (Hedge and Lamond, 1987
).
Since its validation in Species plantarum (Linnaeus, 1753
), Athamanta appears to be a rather artificial assemblage of species. From among the ten names introduced in Athamanta by Linnaeus, only A. cretensis, the type of the genus, has been retained. Athamanta sicula is sometimes recognized in Tinguarra (Bentham, 1867
; Knees, 1996
), while two other Linnaean species now placed in Athamanta were originally described in Bubon and Seseli (Jarvis and Knees, 1988
). This study restores A. sicula in the genus, but removes A. macedonica and A. della-cellae.
Athamanta macedonica falls within the Pimpinella group of umbellifers, which also includes Aphanopleura and Psammogeton, recently removed from Caucalideae by Katz-Downie et al. (1999)
. Morphological comparisons place Registaniella in this clade also (Rechinger, 1987b
). These genera share several morphologic features, such as hispid, ovoid fruits. Athamanta macedonica is generally similar in habit to some species of Psammogeton and Pimpinella, the similarity between the latter two genera having already been noted by Boissier (1872). Pimpinella, the largest genus in the group, includes some 150 species distributed throughout the Old World (Pimenov and Leonov, 1993
). It encompasses a diverse array of species, contrary to the more narrowly defined Aphanopleura, Registaniella, and Psammogeton. Recently, three species of Psammogeton were transferred from Pimpinella (Rechinger, 1987a
). Pimpinella may therefore constitute a paraphyletic aggregate of species requiring division into smaller, more natural genera. Some close relatives of Athamanta macedonica may eventually be identified, however, until a better understanding of the phylogeny of the Pimpinella clade is achieved, it seems reasonable to restore the Linnaean genus Bubon, typified by Bubon macedonicum L. As three subspecies are generally recognized in B. macedonicum (Tutin, 1968
), two new combinations are necessary: Bubon macedonicum subsp. albanicum (Alston & Sandwith) Spalik & S. R. Downie comb. nov. (basionym: Athamanta albanica Alston & Sandwith, Journ. Bot. 78: 193. 1940), and Bubon macedonicum subsp. arachnoideum (Boiss. & Orph.) Spalik & S. R. Downie comb. nov. (basionym: Athamanta arachnoidea Boiss. & Orph. in Boiss., Fl. or., suppl. 262. 1888).
Another species removed from Athamanta is A. della-cellae, which in the present study is placed close to Daucus. The taxonomy of Daucus, albeit after some intense investigations (i.e., Small, 1978
; Okeke, 1982
; Heywood, 1983
), is still not clear; molecular data suggest that the genus may be paraphyletic, with A. della-cellae and several other genera of Caucalideae nested within (Lee and Downie, 1999). A detailed morphological study is necessary in order to decide whether these taxa should be included in Daucus or whether Daucus should be divided into smaller units. We therefore refrain from making a new combination that may later prove provisional.
Molopospermum peloponnesiacum was placed in Scandiceae by de Candolle (1830)
, and its retention there throughout the systems of Drude (1898)
and followers reflects an inability to find a better place rather than a well-justified taxonomic decision. Creating a monotypic tribe, as did Cerceau-Larrival (1962)
, simply reflects that no relatives have been identified. Molopospermum is the only member of Scandiceae with distinctly winged fruits. Bentham (1867)
placed the genus in Smyrnieae, a decision confirmed by Krähenbühl and Küpfer (1992) and followed by Pimenov and Leonov (1993)
. However, tribe Smyrnieae apparently represents an artificial group, as molecular analyses scatter its representatives among most major clades of Apioideae (Downie et al., 1998
). This analysis has not resolved the question of taxonomic affinity of Molopospermum, as its sequence was too divergent to allow a reasonable alignment. Based on the analysis of partially aligned ITS data, however, an affinity to Physospermum may be apparent, as also suggested by Shneyer et al. (1992)
based on serological data. Additional data are necessary to confirm its placement here, especially from the more conservatively evolving chloroplast genome.
Main divisions in Apiaceae subfamily Apioideae
This study included representation of the Daucus, Aciphylla, Angelica, Apium, Aegopodium, Oenanthe, Physospermum, and Conioselinum clades, i.e., major lineages of Apioideae delimited on the basis of earlier molecular studies (Plunkett, Soltis, and Soltis, 1996
; Downie et al., 1998
). Based on the inclusion of a subset of taxa from each of these groups, the same clades were retained in this study, although their relationships differed from the earlier, more comprehensive analyses. In addition to these previously published ITS sequences, we included representation from Echinophoreae, a small tribe of Irano-Turanian distribution whose members have not been previously analyzed in a molecular systematic investigation. We also included Bunium and Scaligeria (sensu lato, i.e., including Elaeosticta) to test the monophyly of tribe Bunieae sensu Calestani (1905)
. Two new accessions of Laserpitieae, Laserpitium petrophilum and Monizia edulis, were surveyed, as well as material of Nirarathamnos, Naufraga, and Deverra, genera whose affinities are largely unknown.
Echinophoreae are a small tribe comprising six genera (Pimenov and Leonov, 1993
). In this study, Echinophora, Pycnocycla, and Dicyclophora form a monophyletic group. Sister to this group is a clade comprising the two Socotran endemics, Nirarathamnos asarifolius and a yet to be described species of Peucedanum. A detailed revisionary study of these Socotran umbellifers and their continental allies is currently being carried out by M. Watson and colleagues (Royal Botanic Garden Edinburgh, UK). All of these taxa, with the addition of Pastinaca, Heracleum, Aethusa, and Peucedanum, are nested within the Angelica group of umbellifers.
The Aegopodium group has also been enlarged based on the results of this study. It now encompasses Grammosciadium and Rhabdosciadium (both transferred from Scandiceae) and also Bunium, Scaligeria, and Elaeosticta, the last three forming a separate lineage together with Trachyspermum ammi. This entire clade is supported by two synapomorphic indels (Fig. 3). Engstrand (1977)
, following Calestani (1905)
, regarded Bunium as closely related to Conopodium and Geocaryum, as these genera share a geophytic habit. They appear, however, to be distantly related, with the similarities among them homoplastic. Scaligeria and Elaeosticta are sister taxa. Based on ITS1 sequence comparisons, Scaligeria setacea also belongs to this clade (K. Spalik and S. Downie, unpublished data). Therefore, the segregation of Elaeosticta from Scaligeria, advocated by Kljuykov, Pimenov, and Tikhomirov (1976)
, is optional rather than necessary; Elaeosticta may be satisfactorily reduced to a lower taxonomic rank, as in Rechinger (1987c)
. Pimenov and Kljuykov (1995)
regarded Scaligeria and Elaeosticta as closely related to Physospermum; this affinity is not supported by ITS data.
The two new accessions of Laserpitieae, Laserpitium petrophilum and Monizia edulis, fall within the Daucus subgroup. This confirms earlier suggestions that Laserpitieae do not constitute a monophyletic entity but rather independent lineages differing from the remaining Daucus relatives in the homoplastic absence of spines (Downie et al., 1998
). The monotypic genus Naufraga from Baleares, once also present in Corsica (Gamisans et al., 1996
), appears to be related to Apium graveolens. Another addition to the Apium clade is Deverra triradiata.
The three subtribes of Scandiceae
According to the phylogenies presented herein, Scandiceae sensu stricto (the Scandix subclade) arise within paraphyletic Caucalideae (sensu Heywood, 1971, 1982b
). The strict consensus tree (Fig. 2) shows the Daucus subgroup being their closest relative and the Torilis subgroup the next basal branch. The neighbor-joining tree (Fig. 4), however, does not support this relationship as strongly. Here both the Daucus and Torilis subgroups are likely contenders for sister taxon to Scandiceae sensu stricto. All of these taxa have been considered previously as belonging to the Daucus group of umbellifers (Plunkett, Soltis, and Soltis, 1996
; Downie et al., 1998
). In this analysis, as in the studies of Plunkett and Downie (1999)
and Lee and Downie (1999)
, the Daucus group encompasses three distinct lineages, albeit the relationships among them are equivocal. Therefore, the most natural treatment seems to be the inclusion of the entire group into the tribe Scandiceae Spreng. in Roem. & Schult., Syst. Veg. 6: xlii. August–December 1820, with the division of the tribe into three subtribes: (1) Scandicinae Tausch, Flora 17: 342. 14 June 1834; (2) Torilidinae Dumort., Fl. Belg.: 81. 1827; and (3) Daucinae Dumort., Fl. Belg.: 81. 1827. These subtribes exemplify the Scandix, Torilis, and Daucus subgroups, respectively, of previous studies. The recognition of these three distinct yet closely related groups, the result of collapsing one of the basal nodes of the previously delimited Daucus clade, may achieve the stability of classification much desired by students of this important group of flowering plants.
Molecular evolution of ITS sequences
The ITS1 and ITS2 regions are part of the transcriptional unit of nuclear rDNA and appear to play a significant role in the maturation of nuclear rRNA. They are therefore subject to evolutionary constraints. The length of each spacer is relatively stable in angiosperms; the entire region including the intervening 5.8S rDNA is usually less than 700 bp in size (Baldwin et al., 1995
). Several relatively conserved sequences have been identified in both spacers (Liu and Shardl, 1994
; Hershkovitz and Lewis, 1996
; Coleman and Mai, 1997
), some of these apparently constituting cleavage sites (Allmang et al., 1996a, b
). The ITS2 region generally possesses more constant positions than ITS1 (Hershkovitz and Zimmer, 1996
; Coleman and Mai, 1997
; Mai and Coleman, 1997
), although no unambiguous conserved motifs shared between algae, fungi, and plants have been identified (Hershkovitz and Lewis, 1996
). It is the secondary structure of ITS2 that is conserved despite wide intra- and interfamilial primary sequence divergence (Mai and Coleman, 1997
). Therefore, the determination of secondary structure for both ITS regions may improve alignment of these sequences at deep levels (Coleman and Mai, 1997
).
Different rates of DNA evolution relative to generation time are usually evoked to explain differences in branch length (Wu and Li, 1985
; Wilson, Gaut, and Clegg, 1990
; Gaut et al., 1996
), with long-lived woody species likely having slower rates than annuals. Branch lengths leading to the annual and biennial members of the Scandix subgroup, such as Scandix, Anthriscus caucalis, A. cerefolium, and species within the Chaerophyllum temulentum clade are generally longer, but not much longer than the others. One possible reason for this is that generation time is not simply a function of life cycle, usually understood as either annual vs. perennial habit. For example, taxa from the Anthriscus sylvestris group are usually iteroparous (polycarpic) perennials, but they also may be semelparous (monocarpic) biennials (or even annuals). Its sister group, A. caucalis, is annual (or winter annual, i.e., biennial). However, the seeds of A. sylvestris remain viable for only a single season, while those of A. caucalis may persist in soil for 3–5 yr (Roberts, 1979, 1986
).
Conclusions
Subtribe Scandicinae, forming tribe Scandiceae along with subtribes Daucinae and Torilidinae, emerges as the only natural suprageneric division in subfamily Apioideae, as defined by morphology and confirmed by cladistic analysis of molecular data. However, careful analysis of different accounts, including that of Drude (1898)
, reveals that there is no single morphological or anatomical character identifying this clade. Therefore, in the past, the included taxa were grouped intuitively, i.e., based on general
) or ITS2 (
) sequences, where only a single DNA strand was sequenced, all accessions were sequenced from both strands. These ITS data were deposited with GenBank as separate ITS1 and ITS2 sequences
