| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Plant Biology, University of Illinois, Urbana, Illinois 61801 USA; and 2 Royal Botanic Garden Edinburgh, 20A Inverleith Row, Edinburgh EH3 5LR, Scotland, UK
Received for publication January 5, 1999. Accepted for publication June 3, 1999.
ABSTRACT
The higher level relationships within Apiaceae (Umbelliferae) subfamily Apioideae are controversial, with no widely acceptable modern classification available. Comparative sequencing of the intron in chloroplast ribosomal protein gene rpl16 was carried out in order to examine evolutionary relationships among 119 species (99 genera) of subfamily Apioideae and 28 species from Apiaceae subfamilies Saniculoideae and Hydrocotyloideae, and putatively allied families Araliaceae and Pittosporaceae. Phylogenetic analyses of these intron sequences alone, or in conjunction with plastid rpoC1 intron sequences for a subset of the taxa, using maximum parsimony and neighbor-joining methods, reveal a pattern of relationships within Apioideae consistent with previously published chloroplast DNA and nuclear ribosomal DNA ITS based phylogenies. Based on consensus of relationship, seven major lineages within the subfamily are recognized at the tribal level. These are referred to as tribes Heteromorpheae M. F. Watson & S. R. Downie Trib. Nov., Bupleureae Spreng. (1820), Oenantheae Dumort. (1827), Pleurospermeae M. F. Watson & S. R. Downie Trib. Nov., Smyrnieae Spreng. (1820), Aciphylleae M. F. Watson & S. R. Downie Trib. Nov., and Scandiceae Spreng. (1820). Scandiceae comprises subtribes Daucinae Dumort. (1827), Scandicinae Tausch (1834), and Torilidinae Dumort. (1827). Rpl16 intron sequences provide valuable characters for inferring high-level relationships within Apiaceae but, like the rpoC1 intron, are insufficient to resolve relationships among closely related taxa.
Key Words: Apiaceae • Apioideae • Hydrocotyloideae • molecular phylogeny • rpl16 intron • rpoC1 intron • Saniculoideae • Umbelliferae
The flowering plant family Apiaceae Lindl. (Umbelliferae Juss.) comprises 300–455 genera and some 3000–3750 species (Constance, 1971
; Pimenov and Leonov, 1993
). It is cosmopolitan, being particularly abundant in the northern hemisphere. Daucus carota subsp. sativus (Hoffm.) Arcang., the common cultivated carrot, is by far its most economically important member. Other familiar vegetables, flavorings, or garnishes include angelica, anise (aniseed), caraway, celeriac, celery, chervil, coriander (cilantro), cumin, dill, fennel, lovage, parsley, and parsnip. Deadly poisonous plants include water hemlock, poison hemlock, hemlock water-dropwort, and fool's parsley. The obvious distinctive characters of many of these plants, such as herbs with hollow or pith-filled stems, pinnately divided leaves with sheathing bases, small unspecialized flowers in compound umbel inflorescences, and specialized fruits, make them easily identifiable to family (likely making them one of the first families of flowering plants to be generally recognized). However, despite their large size, widespread distribution, and economic importance, no widely acceptable modern classification is available.
The most recent treatment of the family (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
). Radically different classifications exist (such as those of Koso-Poljansky, 1916
, and Cerceau-Larrival, 1962
), but have proved unworkable in practice and are rarely used. Drude recognized three subfamilies of Apiaceae (Apioideae, Saniculoideae, and Hydrocotyloideae), dividing each into a series of tribes and subtribes. Molecular systematic investigations have confirmed the monophyly of Apioideae and demonstrated its sister-group status to subfamily Saniculoideae, but have also shown that all of Drude's tribes (and other reclassifications of the family) are largely unsound (Downie and Katz-Downie, 1996
; Downie, Katz-Downie, and Cho, 1996
; Kondo et al., 1996
; Plunkett, Soltis, and Soltis, 1996a, b, 1997
; Downie et al., 1998
; Valiejo-Roman et al., 1998
; Katz-Downie et al., 1999
; Plunkett and Downie, 1999
). Umbellifers display a remarkable array of morphological and anatomical modifications of their fruits, many of which are adaptations for various modes of seed dispersal. Not surprisingly, these characters are prone to convergence, and their almost exclusive use to delimit suprageneric groups has confounded estimates of relationship.
Our goal over the past few years, and that of our collaborators, has been to resolve the "higher level" relationships within subfamily Apioideae. This is necessary in order to provide the framework for "lower level" revisions of particular tribes and complexes of genera, so important in such a group of plants where suprageneric relationships have been largely speculative and ever changing. Eventually, this will lead to the production of a modern classification (i.e., a "new Drude"). To achieve this goal, a variety of molecular characters have been used, such as chloroplast gene (rbcL, matK) and intron (rpoC1, rps16), and nuclear ribosomal DNA internal transcribed spacer (ITS) sequences. A recent study examined restriction site variation of chloroplast DNA (cpDNA; Plunkett and Downie, 1999
); further examination of chloroplast genomic structure is in progress (G. Plunkett and S. Downie, unpublished data). While these characters have been important in providing insight into evolutionary relationships, not all have been useful at the same hierarchical level. Moreover, because many existing data sets are not parallel in construction, opportunities to combine data have been few.
Noncoding regions of cpDNA, such as introns and intergenic spacers, tend to evolve more rapidly than coding loci, both in nucleotide substitutions and in the accumulation of insertion and deletion events (indels), presumably because they are less functionally constrained (Curtis and Clegg, 1984
; Palmer, 1991
; Clegg et al., 1994
). Because these noncoding regions can potentially supply more informative characters than coding regions of comparable size, they have become popular for phylogenetic studies among taxa that are recently diverged. The chloroplast gene rpl16, encoding the ribosomal protein L16 (Posno, Van Vliet, and Groot, 1986
), is interrupted by an intron in many, but not all, land plants (Campagna and Downie, 1998
). In most flowering plants, this intron is ~1 kilobase in length (Campagna and Downie, 1998
). Pairwise comparisons of the 17 chloroplast introns shared between tobacco and rice indicate that the rpl16 intron is most divergent, with 64.5% sequence similarity (Downie, Katz-Downie, and Cho, 1996
). Wolfe, Li, and Sharp (1987)
reported that this intron has an exceptionally high rate of sequence change when Spirodela is compared with tobacco, and Small et al. (1998)
concur that this intron is rapidly evolving, at least in the context of the seven noncoding cpDNA loci examined in a group of recently radiated tetraploid cottons. Given its large size relative to other plastid introns and potential for much variation, we have chosen to examine the historical relationships of subfamily Apioideae and allied taxa using the rpl16 intron. Previous studies have already demonstrated the utility of this region for phylogenetic inferences in Lemnaceae (Jordan, Courtney, and Neigel, 1996
), Poaceae (Kelchner and Clark, 1997
), and Cactaceae (Dickie, 1996
; R. Wallace, unpublished data).
In this paper, we (1) characterize the molecular evolution of the rpl16 intron in Apiaceae and related taxa and assess its utility in estimating phylogeny, (2) present results based on phylogenetic analyses of these rpl16 intron sequences, and (3) for a subset of the taxa, compare the phylogenetic results obtained to those inferred using rpoC1 intron sequences. These intron data are then combined and the resultant estimate of relationship compared to phylogenies for the group inferred using other characters, such as nuclear ribosomal DNA ITS (Downie et al., 1998
; Katz-Downie et al., 1999
) and chloroplast matK (Plunkett, Soltis, and Soltis, 1996b
) sequences, and chloroplast restriction sites (Plunkett and Downie, 1999
). Based on consensus of relationship, we take the first steps towards a "new Drude" by formally recognizing seven groups of apioids at the tribal level and, in so doing, provide the requisite framework for "lower level" systematic study.
MATERIALS AND METHODS
Plant accessions
One hundred and nineteen species from 99 genera of Apiaceae subfamily Apioideae, five species (five genera) each from Apiaceae subfamilies Hydrocotyloideae and Saniculoideae, 11 species (ten genera) of Araliaceae, and seven species (five genera) of Pittosporaceae were examined for rpl16 intron sequence variation (Table 1). In total, 147 species representing 124 genera were considered, with 84 of these species included in a previous phylogenetic analysis of rpoC1 introns (Downie et al., 1998
). RpoC1 intron sequences for Billardiera scandens and Bursaria spinosa (Pittosporaceae) were procured as part of this study, for a total of 86 matching rpl16 and rpoC1 intron sequences (Table 1). With the exception of Anethum graveolens and Crithmum maritimum, where different accessions of the same species were examined, both rpl16 and rpoC1 intron data for these 86 species were obtained from precisely the same specimens.
|
). Details of the DNA extraction procedures have been presented in Downie and Katz-Downie (1996)
. The complete rpl16 intron from all 147 species, including portions of its flanking exons and the intergenic spacer between genes rps3 and rpl16, was amplified using the polymerase chain reaction (PCR) method and primers "rps3" and "L16 exon2" in an equimolar ratio (Fig. 1). These primers were designed by comparing published rpl16 exon2 or rps3 sequences from tobacco, spinach, Epifagus, Vigna, rice, maize, and Marchantia, and choosing regions highly conserved among them (Ohyama et al., 1986
; Shinozaki et al., 1986
; McLaughlin and Larrinua, 1987
; Hiratsuka et al., 1989
; L. Arief, B. Entsch, and R. Wicks, unpublished data). In tobacco cpDNA, the rpl16 intron is 1020 base pairs (bp) in size, the 3' end of primer "rps3" is 377 bp upstream from the exon1-intron junction, and the 3' end of primer "L16 exon2" is 18 bp downstream from the intron-exon2 junction (Shinozaki et al., 1986
). Five internal primers were constructed to facilitate manual sequencing; these are labeled "L16 exon1" and "intron 1–4" in Fig. 1. All seven primers were synthesized by Integrated DNA Technologies, Inc. (Coralville, Iowa, USA).
|
). Each set of PCR amplifications was monitored by the inclusion of positive (tobacco cpDNA) and negative (no template) controls. Successful PCR amplifications resulted in a single-band product of ~1400 bp. Of the 147 accessions sequenced, 37 were done so with the seven primers identified in Fig. 1 using manual sequencing methods. Here the sequence data were obtained through direct sequencing of double-stranded templates derived from the PCR procedure. The remaining species, including ten of those sequenced manually, were sequenced using an Applied Biosystem's, Inc. (Foster City, California, USA) 373A Automated DNA sequencer with Stretch upgrade. Cycle sequencing reactions were carried out in a PTC-100 thermocycler (M. J. Research, Inc., Cambridge, Massachusetts, USA) using the purified PCR products, AmpliTaq DNA polymerase, and fluorescent dye-labeled terminators (Perkin-Elmer Corp., Norwalk, Connecticut, USA). The reaction conditions were as specified by the manufacturer, with the addition of 5% dimethylsulfoxide (DMSO). The sequencing products, after purification with Centri-Sep spin columns (Princeton Separations, Adelphia, New Jersey), were resolved by electrophoresis in 4% acrylamide gels. Sequencing primers "L16 exon1," "L16 exon2," and "intron3" (Fig. 1) were each used in the sequencing of each DNA template. All automated output was checked visually and edited for correct automated base-calling.
Sequence alignment and intron secondary structure
The DNA sequences were aligned initially using CLUSTAL W version 1.7 (Thompson, Higgins, and Gibson, 1994
), copied into the data editor of PAUP version 3.1.1 (Swofford, 1993
), and realigned manually. Gaps were positioned to minimize nucleotide mismatches. Consideration was also given to the probable mechanism of DNA evolution giving rise to the mutation, as described by Kelchner and Clark (1997)
. For example, many insertions were inferred to be the result of a single inserted direct repeat, a highly probably mutational event in noncoding DNA. Only sequence data from the rpl16 intron were included in the analysis, because data from the rps3-rpl16 intergenic spacer were not available for many taxa. Predictions of the Anethum graveolens (dill) rpl16 intron secondary structure were made using the free-energy minimization method of MULFOLD version 2.0 (Jaeger, Turner, and Zuker, 1989
; Zuker, 1989
).
Pairwise nucleotide differences of unambiguously aligned positions were determined using the distance matrix option in PAUP. Alignment gaps in any one sequence were treated as missing data for all taxa. These divergence values were calculated simply as the proportion of divergent sites in each direct pairwise comparison with no provision made to account for multiple hits. Transition/transversion (Ts/Tv) ratios over a subset of the maximally parsimonious trees were calculated using MacClade version 3.01 (Maddison and Maddison, 1992
). Polytomies were arbitrarily resolved. To assess variation in levels of base substitution among sites across a subset of the maximally parsimonious trees obtained, the number of steps per four consecutive bases was estimated using MacClade. The nucleotide sequence data reported in this study have been deposited with the GenBank Data Library; accession numbers are provided in Table 1.
Phylogenetic analysis
Phylogenetic analyses were carried out on the complete 147-species rpl16 intron data matrix and, for 86 of these species, separately and in combination with available rpoC1 intron sequences (Table 1). The data were analyzed using Macintosh Quadra 700 or Power Macintosh computers. All trees computed were rooted with the Pittosporaceae accessions. Phylogenetic analyses of molecular data (Xiang et al., 1993
; Plunkett, Soltis, and Soltis, 1996a
) corroborate traditional taxonomic evidence (Jay, 1969
; Dahlgren, 1980
; Thorne, 1992
; Judd, Sanders, and Donoghue, 1994
) in suggesting that Pittosporaceae are likely sister to Apiaceae + Araliaceae.
Maximally parsimonious (MP) trees were sought using PAUP and the heuristic search strategies described in Downie et al. (1998)
, based on those presented in Catalán, Kellogg, and Olmstead (1997)
. The length of the shortest trees was obtained by initiating at least 500 searches, each using random addition starting trees, with tree bisection-reconnection (TBR) branch swapping and MULPARS selected, but saving no more than five of the shortest trees from each search. The equally MP trees were then used as starting trees for TBR branch swapping. In all analyses, the maximum number of trees to be saved was set at 5000. The strict consensus of these 5000 trees was subsequently used as a topological constraint. Once more, 500 random-order-entry replicate searches were initiated as above, saving no more than five trees from each search. However, only those trees that did not fit the constraint tree were saved. As no additional trees were found at the length of the initial 5000 trees, this suggested strongly that the strict consensus tree does adequately summarize the available evidence, even though the exact number of trees at that length is not known. Bootstrap values (Felsenstein, 1985
) were calculated from 100 replicate analyses using a heuristic search strategy, simple addition sequence of the taxa, and TBR branch swapping. Owing to the large size of the data matrix, a maxtree limit of 200 trees per replicate was set. Gaps were incorporated into the analysis by scoring each insertion or deletion as a separate presence/absence (i.e., binary) character (Swofford, 1993
). The resultant topology was then compared to the one inferred when gaps were omitted as additional characters. Previous investigations of cpDNA rpoC1 intron (Downie et al., 1998
) and other noncoding sequences (e.g., van Ham et al., 1994
) have revealed that indels contain much phylogenetic information and, indeed, may provide particularly clear indications of relationship.
Distance trees were constructed using the neighbor-joining (NJ) method (Saitou and Nei, 1987
), implemented using the NEIGHBOR program in PHYLIP version 3.5 (Felsenstein, 1993
). Distance matrices were calculated using the DNADIST program of PHYLIP, and the numbers of nucleotide substitutions were estimated using Kimura's (1980)
two-parameter method. Length mutations were not incorporated into the analysis. Two Ts/Tv rate ratios were used (1.0 and 2.0), with the former approximating the expected ratio of Ts to Tv as inferred by the MP analysis. A bootstrap analysis was done using 100 resampled data sets generated with the SEQBOOT program prior to calculating the distance matrices and NJ trees. PHYLIP's CONSENSE program was then used to construct a consensus tree.
The maximum likelihood (ML) method was also applied to these substitution data using the program fastDNAml version 1.0.6 (Olsen et al., 1994
). ML trees were inferred using a Ts/Tv rate ratio of 1.0, randomizing the input order of sequences (jumble), and by invoking the global branch swapping algorithm. Empirical base frequencies were derived from the sequence data and used in the ML calculations. The ML analyses, however, could not be carried out to completion, given the large size of the data matrix and the time required to complete the global branch swapping. Despite four weeks of computer run time, none of the 12 Macintosh computers running simultaneously completed their searches or converged on the same highest (least negative) log likelihood value. After completing one round of branch swapping, the best ML tree had a log likelihood value of -8123.195.
RESULTS
Rpl16 intron sequence characterization
The sequenced rpl16 introns varied in length from 892 bp (in Cicuta virosa; Apiaceae subfamily Apioideae) to 1021 bp (in Bursaria spinosa; Pittosporaceae). Within Apioideae, their length ranged from 892 to 982 bp (Heracleum lanatum). Alignment of all 147 sequences resulted in a matrix of 1444 positions. However, because of frequent length mutations of varying sizes confounding alignment interpretation, it was necessary to exclude 26 regions from the matrix in the distance calculations and phylogenetic analyses. These ambiguous regions ranged in size from one to 90 positions (averaging ~14 positions each), with several characterized by tracts of poly-A's, G's, and T's of variable length. We have taken a conservative approach to sequence alignment in excluding regions where alternative alignments are possible and that may result in conflicting phylogenetic signal. Alternating gap penalty or substitution costs were not considered. A region representing an unambiguous 92-bp deletion in all Apiaceae and Araliaceae relative to Pittosporaceae was also excluded. As a result, 452 alignment positions (or about one-third of the entire matrix) were excluded from subsequent analyses. Characteristics of the remaining 992 unambiguously aligned positions, including the numbers of constant, parsimony informative, and autapomorphic positions, are provided in Table 2. The ratio of terminal taxa (147) to parsimony-informative nucleotide substitutions (378) is 1 : 2.6. Measures of pairwise sequence divergence ranged from identity to 11.3% across 119 accessions (99 genera) of subfamily Apioideae, and from identity to 18.1% across all 147 accessions. A total of 90 unambiguous gaps was required for proper alignment of these sequences. These gaps ranged from 1 to 92 bp, with the average size being ~7 bp; the number of gaps with respect to their size is presented in Fig. 2. Thirty-seven gaps were parsimony informative (Table 2), with three of these (including the large 92-bp deletion) distinguishing all Apiaceae and Araliaceae from Pittosporaceae. Percentage G + C content across all 147 intron sequences ranged from 28.3 to 33.2%, and averaged 30.8%.
|
|
, and the results of the MULFOLD analysis. It should be noted, however, that minor differences in free energy exist between this model and other conformations that can be drastically different. These differences are particularly evident within intron domains III and IV. Therefore, this model should be interpreted as a provisional estimate. Like other group II introns, a conserved core structure is evident, consisting of six major domains (I–VI) radiating from a central wheel. Domain I is divided into several subdomains and other regions, of which we have identified subdomains IC and ID, and exon binding sites 1 and 2 (EBS 1 and EBS 2).
|
|
|
|
Distance trees obtained from the NJ analysis, estimated from the two-parameter method of Kimura (1980)
with Ts/Tv rate ratios of 1.0 or 2.0, were topologically congruent. The tree constructed with a rate ratio of 1.0 is presented in Fig. 6. While the ML analysis could not be completed with global branch swapping invoked, the best results obtained (not shown) were consistent with those inferred using MP and NJ methods with respect to the major clades distinguished. Within Araliaceae and many clades of Apioideae, branch lengths are quite short, whereas among the hydrocotyloids (e.g., Bolax, Centella, Didiscus, and Hydrocotyle), the branches are long. While branch lengths can vary substantially among closely related taxa possessing different life-history strategies (compare predominantly woody Araliaceae vs. herbaceous Hydrocotyloideae, for example), the variation exhibited within subfamily Apioideae is not so readily explained (Downie et al., 1998
).
|
Within Apiaceae subfamily Apioideae, similar groupings of taxa occur in all trees. These clades, identified by numbered brackets, coincide with those groups recognized on the basis of parsimony analysis of rpoC1 intron sequences (Downie et al., 1998
). The most basal elements in the subfamily belong to a well-supported clade comprising the genera Anginon, Glia, and Heteromorpha (group 12, the "Heteromorpha" clade). Progressing upwards in the trees, groups 10 (Aulacospermum, Eleutherospermum, Physospermum, and Pleurospermum; the "Physospermum" clade) and 11 (Bupleurum; the "Bupleurum" clade) each arise separately (Fig. 6), or unite as sister taxa (Fig. 4). In the ML tree (not shown), groups 10 and 11 form two branches of a trichotomy, the third branch representing all other members of Apioideae except group 12. Pairwise sequence divergence values in group 10 range from 1.6 to 3.9%. Next is group 9, comprising Komarovia and Parasilaus (the "Komarovia" clade), followed by the phylogenetically isolated Erigenia. Next, the genera Cicuta, Cryptotaenia, Oenanthe, Oxypolis, Perideridia, and Sium comprise a well-supported clade (group 6, the "Oenanthe" clade of Plunkett, Soltis, and Soltis, 1996b
, and Downie et al., 1998
); the relationships within this clade, however, are not consistent. While Oenanthe and Cicuta unite in all analyses, as do Cryptotaenia, Sium, and Oxypolis, their relationships to each other and to Perideridia are not clear. Divergence values in this clade range from 0.9 to 3.9%. Group 7, comprising Aciphylla, Anisotome, Smyrnium, and Lecokia (the "Aciphylla" clade of Plunkett, Soltis, and Soltis, 1996b
, and Downie et al., 1998
), arises next in the NJ (Fig. 6) and ML (not shown) trees, but is sister (albeit with poor bootstrap support) to group 5B in the MP tree (Fig. 4). In group 7, sequence divergence values vary between 0.4 and 4.3%.
Group 5, the "Daucus" clade of Plunkett, Soltis, and Soltis (1996b)
and Downie et al. (1998)
, consists of representatives of Drude's tribes Dauceae, Laserpitieae, and Scandiceae (the latter including subtribes Scandicinae and Caucalidinae). Within this clade, two major subgroups (5A and 5B) are recognized (Figs. 4 and 6). Subgroup 5A comprises representatives of Drude's Dauceae, Laserpitieae, and Scandiceae subtribe Caucalidinae. Subgroup 5B reflects Drude's Scandiceae subtribe Scandicinae. Subgroup 5A can be further subdivided in the NJ (Fig. 6) and ML (not shown) trees. The first group consists of Daucus (two species), Pseudorlaya, Agrocharis, Laserpitium, Orlaya (two species), Laser, and Polylophium. The genus Daucus is not monophyletic, with the North American D. pusillus allied with African Agrocharis, and D. carota allied with Pseudorlaya. Drude's tribe Laserpitieae, exemplified by Laserpitium, Laser, and Polylophium, is also not monophyletic. The second group consists of Turgenia, Lisaea, Caucalis, Szovitsia, Astrodaucus, Torilis, Chaetosciadium, Glochidotheca (= Turgeniopsis), and Yabea. In subgroup 5A, pairwise sequence divergence values among congeners range between 0.5 and 4.8%. Subgroup 5B comprises the genera Anthriscus, Myrrhis, Chaerophyllum (two species), Osmorhiza (two species), and Scandix (two species), and parallels Heywood's (1971)
tribe Scandiceae. The last three genera are each monophyletic, but their relationships differ depending upon method of tree construction used. Among congeners, divergence values in this subgroup range between 1.2 and 5.6%. Variously associated with subgroups 5A and 5B are Ferula kokanica, Ligusticum scoticum, and Conioselinum chinense. In all analyses, F. kokanica, L. scoticum, and Caucalideae comprise a well-supported clade. Conioselinum chinense is sister to this clade in the MP and NJ trees.
All remaining species belong to groups 1–4, the "Angelica," "Crithmum," "Apium," and "Aegopodium" clades, respectively, of Downie et al. (1998)
. Resolution here is poor, with none of these four clades distinguishable. However, six smaller clades can be inferred with varying degrees of bootstrap support and include: (1) Aethusa, Exoacantha, and Peucedanum caucasicum; (2) Heracleum (two species), Pastinaca (two species), and Malabaila; (3) Apium, Anethum, Foeniculum, and Ridolfia; (4) Cnidiocarpa, Cnidium, Ligusticum ferulaceum, and L. physospermifolium; (5) Heracleum rigens, Zosima, and Tordylium; and (6) Prangos, Smyrniopsis, and Opopanax. With the exception of only a few branches, such as those leading to Peucedanum caucasicum, Apium, Oedibasis, and Pimpinella, the branch lengths within apioid groups 1–4 are relatively short. In this group, pairwise sequence divergence estimates reach a maximum value of 6% (between Peucedanum caucasicum and Pimpinella).
For those 16 genera of subfamily Apioideae where more than one species was examined, 11 are not monophyletic. These genera include Aciphylla, Arracacia, Conioselinum, Daucus, Ferula, Heracleum, Ligusticum, Pastinaca, Peucedanum, Pleurospermum, and Seseli. Of the eight tribes of Apioideae recognized by Drude, we have sampled extensively from three (Apieae, Peucedaneae, and Smyrnieae). Not surprisingly, none of these tribes are monophyletic, with multiple independent derivations inferred in all cladograms. The only suprageneric taxon in Apioideae that is maintained as monophyletic is Drude's Scandiceae subtribe Scandicinae (= tribe Scandiceae sensu Heywood, 1971
).
Rpl16 and rpoC1 intron sequence characterization
For those 86 species where both rpl16 and rpoC1 intron data are available (Table 1), the data sets were analyzed separately and together using PAUP. Sequence characteristics of these separate and combined matrices are presented in Table 4. These comparisons show that sequence data from the rpl16 intron are more variable than those of the rpoC1 intron, as evidenced by the greater number of positions excluded due to alignment ambiguity (30.8 vs. 16.8%), the fewer invariant positions (50.7 vs. 57.9%), the greater number of positions informative for parsimony analysis (31.7 vs. 24.6%), and the greater number of unambiguous alignment gaps (72 vs. 47). Across all 86 species, pairwise sequence divergence was higher for rpl16 than for rpoC1 (17.8 vs. 12.3%, respectively). Within subfamily Apioideae, however, divergence values were approximately the same. In the combined matrix, pairwise sequence divergence estimates ranged between 0.2 and 15.2%.
|
65%) are treated as unresolved (i.e., they are collapsed to yield polytomies), the trees are consistent, with the only remaining area of discord being the relative positions of Bupleurum and Physospermum. In general, there is greater resolution at the base of the rpl16 intron tree than that of the rpoC1 tree (i.e., the "Oenanthe" clade is resolved in the former, its placement relative to other major clades is clearer, and greater resolution is achieved in Araliaceae). In contrast, among those apioids belonging to groups 1–4 (i.e., the clade extending from Arracacia brandegei to Aegopodium; Fig. 7), greater resolution is seen in the rpoC1 intron tree.
|
|
), and include: group 1—the "Angelica" clade; group 2—the "Crithmum" clade; group 3—the "Apium" clade; group 4—the "Aegopodium" clade; group 5—the "Daucus" clade, comprising subgroups 5A and 5B; group 6—the "Oenanthe" clade; group 7—the "Aciphylla" clade; groups 8, 9, 10, and 11, with Conioselinum chinense, Komarovia anisosperma, Physospermum cornubiense, and Bupleurum as their sole representatives, respectively; and group 12—the "Heteromorpha" clade. Resolution of relationships among many of these clades is poor. DISCUSSION
Molecular evolution and phylogenetic utility of the rpl16 intron
The chloroplast gene rpl16 is interrupted by an intron in many, but not all, land plants. Sequencing, PCR surveys, and blot-hybridization assays have revealed that this intron is absent from several Geraniaceae, Plumbaginaceae, and Goodeniaceae cpDNAs (Campagna and Downie, 1998
). Among those species possessing an intron, it varies considerably in size, from 536 bp in Marchantia polymorpha (Ohyama et al., 1986
) to 1411 bp in Spirodela oligorhiza (Posno, Van Vliet, and Groot, 1986
). Evidently, this locus is able to withstand much variation in length so long as it does not fall below the minimum size (~500 bp) required for intron splicing (Doyle, Doyle, and Palmer, 1995
). In most angiosperms, the rpl16 intron is ~1 kb in size (Campagna and Downie, 1998
), and its size in Apiaceae, Araliaceae, and Pittosporaceae cpDNAs (892–1021 bp) is consistent with these results.
Group II introns are excised from mRNA transcripts via a series of self-catalyzed reactions (Michel, Umesono, and Ozeki, 1989
) and show a strong relationship between the functional importance of its structural features and probability of evolutionary change (Clegg et al., 1994
). Of these introns' six major structural domains, domains V and VI are required for processing of the transcript and, therefore, evolve most slowly (Learn et al., 1992
). Portions of domain I, such as the region housing exon binding site 1, are also conserved evolutionarily. In contrast, domains II and III, apparently dispensable in self-splicing introns (Michel, Umesono, and Ozeki, 1989
), have the highest rates of sequence change (Learn et al., 1992
; Downie et al., 1998
). With regard to the rpl16 intron, domains III and IV were inferred to be the most variable and domains V and VI the least variable. Domain IV is characterized by numerous indels, and, for those positions that can be aligned unambiguously, divergence values approached 31%.
Across a comparable array of taxa, the rpl16 intron is more variable than that of the rpoC1 intron. This variation extends from the former possessing proportionally more informative nucleotide substitutions and length mutations to having a greater number of positions excluded from the analysis because of alignment ambiguity (Table 4). However, the amount of homoplasy in each data set is similar. Within Apioideae, sequence divergence estimates for both introns were approximately the same, whereas across basal Apiaceae, Araliaceae, and Pittosporaceae, values are much higher for rpl16. Despite the greater variability of the rpl16 intron, these regions are useful at different levels. While phylogenetic analysis of rpl16 intron sequences fail to resolve relationships among apioid groups 1–4, some resolution is achieved using rpoC1. In contrast, rpl16 intron data generally provide greater resolution among basal apioids. Considered separately, rpoC1 and rpl16 intron sequences have little power to resolve relationships among closely related taxa. By combining these data, greater resolution and higher bootstrap support are achieved.
The rpl16 intron has several other properties, making it attractive for comparative sequencing studies. A pair of universal primers, anchored in the exons, are sufficient to amplify the entire intron. The region is easily amplified once the PCR protocol has been optimized, and among closely related taxa the alignment of sequences is generally straightforward. Deep-level comparisons, however, result in frequent length mutations and regions of high variability. With the exception of conserved domains V and VI, variation is generally equally distributed over the length of the entire intron. Within Apiaceae, estimates of sequence divergence are comparable to those of other chloroplast intron (rpoC1) and gene (matK) sequences, but lower than that of the nuclear rDNA ITS region (Downie et al., 1998
).
Apioideae phylogenetic resolutions and classification
Phylogenetic analysis of combined rpoC1 and rpl16 intron data supports 12 major clades within subfamily Apioideae (Fig. 8) and demonstrates patterns of relationship consistent with previously published cpDNA and ITS based phylogenies (Kondo et al., 1996
; Plunkett, Soltis, and Soltis, 1996b
; Downie et al., 1998
; Valiejo-Roman et al., 1998
; Katz-Downie et al., 1999
; Plunkett and Downie, 1999
). These groups are discussed below, with formal tribal recognition given to those seven clades that are consistently recognized and usually well supported in all analyses. To facilitate communication, several additional informal groups are described. However, not all present analyses support these groups as distinct, and their recognition is highly provisional.
Groups 1–4, the "Angelica," "Crithmum," "Apium," and "Aegopodium" clades
Analyses of rpoC1 intron sequences alone (Fig. 7) or in combination with rpl16 intron data (Fig. 8) reveal the presence of four distinct yet closely allied groups of taxa, previously called the "Angelica" (group 1), "Crithmum" (group 2), "Apium" (group 3), and "Aegopodium" (group 4) clades (Downie et al., 1998
). Separate analyses of rpl16 intron sequences, however, fail to resolve these groups (Figs. 4 and 6). Upon consideration of all available molecular evidence, the "Angelica" and "Apium" clades cannot be circumscribed unambiguously, and in those studies where the latter occurs as monophyletic it is supported only weakly (Fig. 8; Plunkett, Soltis, and Soltis, 1996b
; Downie et al., 1998
). In the ITS studies, taxa attributed to the "Apium" clade comprise at least four lineages basal to the "Angelica" clade (Downie et al., 1998
; Katz-Downie et al., 1999
). The "Crithmum" clade is variously positioned, sister group to either the "Angelica" clade (Figs. 7–8) or, when ITS data are considered, to the "Aegopodium" clade (Downie et al., 1998
; Katz-Downie et al., 1999
). Plunkett, Soltis, and Soltis (1996b)
included members of the "Aegopodium" and "Crithmum" clades in an expanded "Apium" clade; a subsequent study considered the "Aegopodium" clade as distinct, but included within it the genera Crithmum and Trachyspermum, and the saniculoid genus Lagoecia (Plunkett and Downie, 1999
). These four groups of Apioideae have been collectively termed the "apioid superclade" (Plunkett and Downie, 1999
), for while they unite as a strongly supported monophyletic group in all phylogenetic analyses to date, the relationships among them are equivocal.
Within the "Angelica" clade four major groups of taxa are distinguishable. Not all molecular studies, however, support these groups as distinct and their recognition here is highly provisional. The first includes a group of palaeopolyploid, meso-American genera (Arracacia, Coulterophytum, Dahliaphyllum, Donnellsmithia, Enantiophylla, Prionosciadium, and Rhodosciadium, and possibly Coaxana, Mathiasella, and Myrrhidendron). These genera are endemic to the highland regions of Mexico and neighboring Central America, one of the two centers of diversity of Apioideae in the western Northern Hemisphere (Mathias, 1965
). We recognize this group as the "Arracacia" clade. The second group comprises the genera Heracleum, Malabaila, Pastinaca, Tordylium, and Zosima. While plastid DNA data do not advocate monophyly of this group, the ITS data do and with high bootstrap support (Downie et al., 1998
; Katz-Downie et al., 1999
). These plants are characterized generally by fruits with thickened wing margins and a rich diversity of furanocoumarins. We recognize this group as the "Heracleum" clade. A third group includes Aletes, Cymopterus, Lomatium, Musineon, Neoparrya, Podistera, Shoshonea, Taenidia, Thaspium, and Zizia. The majority of these species occur in the dry, sandy, or alkaline regions of western North America, and often at high elevations. These western members are primarily herbaceous perennials and are frequently caespitose. Current data from the plastid genome cannot unequivocally support monophyly of this group. ITS data, however, provide weak support for this clade (S. Downie and R. Hartman, unpublished data). We have recognized this group as the "Rocky Mountain" umbellifers, realizing that many species extend beyond this range. A fourth group consists of Dicyclophora, Echinophora, and Pycnocycla, and their monophyly has been confirmed on the basis of ITS data (Downie, Katz-Downie, and Spalik, 2000). These taxa possess unique inflorescence and infructescence morphology and have been treated by Drude in tribe Echinophoreae. Before any of these four groups can be recognized formally, further investigations are necessary to confirm their monophyly; in the case of Echinophoreae, data from the plastid genome are required to confirm its position within the "Angelica" clade.
The "Crithmum" clade is often represented by only two taxa: Crithmum maritimum and Trachyspermum ammi. This group is recognized in studies of ITS and rpoC1 intron sequences; separate analysis of rpl16 intron data fail to resolve it. Pyramidoptera and Oedibasis can be added to this group, but this association is weakly supported (Katz-Downie et al., 1999
). Additional evidence suggests an affinity with Scaligeria moreana Engstrand, Elaeosticta allioides (Regel & Schmalh.) Kljuykov, Pimenov & V. N. Tichom., and Bunium elegans (Fenzl) Freyn (Downie, Katz-Downie, and Spalik, 2000). We continue to recognize this group as the "Crithmum" clade. If further investigation supports this grouping, and if it is to be recognized at the tribal level, the earliest name Pyramidoptereae Boiss. (1871) should be applied. Within the "Apium" clade, one lineage is consistently recognized in all analyses. This group includes Ridolfia segetum, Deverra triradiata Hochst. ex Boiss., and Naufraga balearica Constance & Cannon, and the cultivated species of Ammi, Anethum, Apium, Foeniculum, and Petroselinum.
Of the "Angelica," "Crithmum," "Apium," and "Aegopodium" clades, the only one that is unambiguously circumscribed in all (but the rpl16 intron) analyses and contains more than a few members is the "Aegopodium" clade (Aegopodium, Carum, and Falcaria, in this study). ITS studies add Aegokeras (syn. Olymposciadium), Fuernrohria, Rhabdosciadium, and Grammosciadium (Downie et al., 1998
; Katz-Downie et al., 1999
; Downie, Katz-Downie, and Spalik, 2000
), and the matK study of Plunkett, Soltis, and Soltis (1996b)
adds Cyclospermum. In the rpl16 intron trees (Figs. 4 and 6), Fuernrohria and Carum are strongly supported sister taxa, and the work of Vinogradova (1995)
supports the close relationship between Fuernrohria and Grammosciadium. The matK study also suggests a possible affinity to the saniculoid genus Lagoecia. Here Lagoecia is allied to Crithmum, and both taxa are considered within the "Aegopodium" clade. If future studies indicate that Lagoecia is to be included within a formally described "Aegopodium" clade, its name will have priority (Lagoecieae Lange, 1880). If on the other hand it is shown that Lagoecia should be excluded from the group (and placed within a separate "Crithmum" clade), priority would extend to Carum (Careae Adanson ex Kuntze, 1904).
Group 5, the "Daucus" clade
Recent molecular systematic studies confirm three distinct and well-supported groups within the "Daucus" clade (Plunkett and Downie, 1999
; Downie, Katz-Downie, and Spalik, 2000
; Lee and Downie, 1999, and unpublished data). These three groups, coinciding herein with subgroup 5B and the two clades comprising subgroup 5A as a result of the NJ analysis (Fig. 6), have been designated as subtribes Scandicinae Tausch (1834), Daucinae Dumort. (1827), and Torilidinae Dumort. (1827), respectively, of tribe Scandiceae Spreng. (Downie, Katz-Downie, and Spalik, 2000). Included in subtribe Daucinae (the "Daucus" subclade) are the genera Agrocharis, Ammodaucus, Cuminum, Daucus, Orlaya, Pseudorlaya, and Pachyctenium, and representatives of Drude's tribe Laserpitieae (Laser, Laserpitium, Melanoselinum, Monizia, Polylophium, and Thapsia). Laserpiteae is not monophyletic, with four separate lineages arising within Daucinae. Subtribe Torilidinae (the "Torilis" subclade) includes Astrodaucus, Caucalis, Chaetosciadium, Glochidotheca, Lisaea, Szovitsia, Torilis, Turgenia, and Yabea. With the exception of Laserpitieae, Scandiceae subtribes Daucinae and Torilidinae (collectively forming group 5A) coincide closely with S. Jury and V. Heywood's (in Heywood, 1982b
) circumscription of tribe Caucalideae Spreng. Subtribe Scandicinae (group 5B, the "Scandix" subclade), representing Drude's Scandiceae subtribe Scandicinae [= Heywood's (1971)
tribe Scandiceae], comprises Anthriscus, Athamanta, Balansaea, Chaerophyllum, Conopodium, Geocaryum, Kozlovia, Krasnovia, Myrrhis, Myrrhoides, Neoconopodium, Osmorhiza, Scandix, Sphallerocarpus, and Tinguarra. Because the relationships among the three groups comprising the "Daucus" clade are equivocal given all available evidence, three distinct yet closely related groups have been recognized at the subtribal level (Downie, Katz-Downie, and Spalik, 2000).
Associated closely with tribe Scandiceae (the "Daucus" clade) in all but a few studies are Ligusticum scoticum and Ferula kokanica. Kondo et al. (1996)
, using rbcL sequences, placed L. scoticum next to Daucus and Torilis. The ITS trees, on the other hand, show L. scoticum (with weak bootstrap support) falling alongside Lecokia and Smyrnium in the "Aciphylla" clade (group 7); this clade is sister to group 5, the "Daucus" clade (Downie et al., 1998
; Katz-Downie et al., 1999
). The anomalous placement of Ferula kokanica near Scandiceae is just as intriguing. Ferula is a large, morphologically variable genus of some 170 species (Pimenov and Leonov, 1993
) and, based on our results, its monophyly, although supported most recently by Shneyer, Borschtschenko, and Pimenov (1995)
, is brought into question. Ferula is thought to be allied with Peucedanum (Drude, 1898
; Bernardi, 1979
) and, in this study, F. assa-foetida falls within the "Angelica" clade, well away from F. kokanica. Valiejo-Roman et al. (1998)
, using ITS sequences, showed F. tenuisecta Korovin and F. violacea Korovin (each representing a different section within Ferula) arising within or close to the "Daucus" clade. Furthermore, ITS data for F. kingdon-wardii H. Wolff show that this species, together with F. kokanica, F. tenuisecta, and F. violacea, constitute a clade sister to tribe Scandiceae (S. Downie and M. Watson, unpublished data). Additional studies of Ferula and Ligusticum are required to investigate their suspected nonmonophyly and unique phylogenetic positions.
Group 6, the "Oenanthe" clade
In this study, the genera Cicuta, Cryptotaenia, Oenanthe, Oxypolis, Perideridia, and Sium are treated in the "Oenanthe" clade; other studies include Berula (Downie et al., 1998
), Neogoezia (Plunkett, Soltis, and Soltis, 1996b
), and Cynosciadium, Lilaeopsis, and several species of Apium attributable to Helosciadium (S. Downie and M. Watson, unpublished data). In all analyses, this clade is very well supported, with bootstrap support values >90% and often 100%. These plants possess glabrous stems and leaves, clusters of tubers or tuberous roots, and commonly grow in moist to wet habitats. We recognize this group as tribe Oenantheae Dumort. (1827).
The taxonomic history of this group of genera is extraordinarily complex, confounded by the use of many longstanding names that are now considered synonyms, and although they share some characters in common, they are rather heterogeneous. Dumortier (1827) described tribe Oenantheae for the genera Aethusa, Coriandrum, and Oenanthe, defined by the presence of radiately ribbed fruits. This artificial assemblage was not followed by later authors, nor is it supported by molecular studies. We have used Dumortier's name, but stress that our circumscription of the tribe is radically different from his (and many others, such as Koso-Poljansky, 1916
, and Cerceau-Larrival, 1962
).
Group 7, the "Aciphylla" clade
The "Aciphylla" clade is recognized as comprising the genera Aciphylla, Anisotome, Lecokia, and Smyrnium (Plunkett, Soltis, and Soltis, 1996b
; Downie et al., 1998
; Katz-Downie et al., 1999
; Plunkett and Downie, 1999
). While phylogenetic analysis of rpl16 intron sequences supports this clade strongly (with bootstrap values of 95 or 98%, Figs. 4 and 6), separate analyses of rpoC1 intron (Fig. 7; Downie et al., 1998
), ITS (Downie et al., 1998
; Katz-Downie et al., 1999
), or cpDNA restriction site data (Plunkett and Downie, 1999
) do not. Furthermore, the ITS studies include Ligusticum scoticum alongside Lecokia and Smyrnium, but with weak bootstrap support (Downie et al., 1998
; Katz-Downie et al., 1999
). In all analyses, two distinct and well-supported subclades are apparent (each supported by bootstrap values of 91–100%): one comprising the Australasian endemics Aciphylla and Anisotome, and the other comprising largely Eurasian genera Lecokia and Smyrnium. Given their geographic isolation and morphological differences, we treat these subclades as separate tribes. Their union cannot readily be supported on the basis of morphological characters.
The predominantly alpine, Australasian genera Aciphylla, Anisotome, Gingidia, Lignocarpa, and Scandia constitute a closely related group (Dawson and Webb, 1982
; Webb and Druce, 1984
; Webb, 1986
). Indeed, phylogenetic analysis of ITS sequences supports their monophyly (Mitchell, Webb, and Wagstaff, 1998
; L. Radford and M. Watson, unpublished data). All are glabrous, long-lived perennials endemic to New Zealand and Australia. The genera are mostly sexually dimorphic, with Aciphylla and Anisotome dioecious and the other three usually gynodioecious (Webb, 1979
). We treat this group as tribe Aciphylleae M. F. Watson & S. R. Downie.
Aciphylleae M. F. Watson & S. R. Downie, Trib. Nov. Tribus generum dioicorum vel gynodioicorum, praesertim alpinorum, distributionis australasicae. Type genus Aciphylla J. R. Forst. & G. Forst., Char. Gen. Pl.: 68 (1775). Other included genera: Anisotome Hook. f., Gingidia J. W. Dawson, Lignocarpa J. W. Dawson, Scandia J. W. Dawson.
The tribal name Smyrnieae (or subtribe Smyrniinae) has been used by virtually all authors of Apiaceae suprageneric classifications. First described by Sprengel (1820) and later modified by Koch (1824), de Candolle (1830)
, Bentham (1867)
, and Drude (1898)
, the size and composition of the tribe have varied considerably. Members of Drude's Smyrnieae were united on the basis of a deep groove on the commissural side of the seeds (campylospermy) and to a lesser extent on their nonelongate fruits. The artificiality of the tribe, however, has been demonstrated repeatedly (Shneyer et al., 1992
; Downie and Katz-Downie, 1996
; Plunkett, Soltis, and Soltis, 1996b
; Downie et al., 1998
; Valiejo-Roman et al., 1998
). Of the 12 genera that Sprengel (1820) cited in his original publication of the tribe, only the type Smyrnium remains in our treatment. Lecokia was first included in Smyrnieae by de Candolle (1830)
, and it has remained there ever since. This narrow circumscription of Smyrnieae parallels, in part, the treatment of Hedge et al. (1987)
where only Smyrnium and (the unrelated) Smyrniopsis are included in the tribe, and the immunochemical study of Shneyer et al. (1992)
where Smyrnium occupies an isolated position away from all other Smyrnieae and other Apiaceae investigated (Lecokia was not considered). Shneyer et al. suggested that Smyrnium may be best treated within a monotypic tribe. Of the 29 genera in Drude's Smyrnieae, all but eight have been considered in molecular systematic investigations to date; in all of these analyses Smyrnium and Lecokia comprise a strongly supported clade. We treat this group as tribe Smyrnieae Spreng. (1820).
Group 8, the "Conioselinum chinense" clade
Phylogenetic analysis of ITS data indicates a close relationship among C. chinense (the only member of the group considered in this study), C. scopulorum (A. Gray) J. M. Coult. & Rose, Ligusticum porteri J. M. Coult. & Rose, and L. canadense (L.) Britton (Katz-Downie et al., 1999
). Conioselinum and Ligusticum are each cosmopolitan and, based on increasing evidence, are clearly not monophyletic. Interestingly, the type species for each of these genera (C. tataricum and L. scoticum) do not occur in this clade. Conioselinum and Ligusticum are both in need of revision, and our molecular results indicate that these four species must be transferred to a new genus or genera. A tribal name cannot be assigned until a new genus is described, but it would be premature to alter the nomenclature at this stage. Pending a full revision of this group, we continue to refer to it as the "Conioselinum chinense" clade.
Group 9, the "Komarovia" clade
Included in the "Komarovia" clade are the genera Komarovia and Parasilaus (and Hansenia in Katz-Downie et al., 1999
). These monotypic genera comprise a group of distinctive umbellifers but, at present, there are insufficient data to confidently delimit a tribe. ITS studies reveal a weak association among these taxa, monotypic Erigenia, and members of group 10, the "Physospermum" clade. Plastid DNA data (e.g., Figs. 4 and 6) fail to support this relationship.
Group 10, the "Physospermum" clade
United in the "Physospermum" clade are the genera Aulacospermum, Eleutherospermum, Physospermum, and Pleurospermum. On the basis of ITS (Downie, Katz-Downie, and Spalik, 2000) and serological (Shneyer et al., 1992
) studies, Molopospermum joins the group. ITS data (Katz-Downie et al., 1999
) support this group strongly (with a 100% bootstrap value), whereas rpl16 intron data (Figs. 4 and 6) support it only moderately (71–84% bootstrap values). We recognize this clade as tribe Pleurospermeae M. F. Watson & S. R. Downie.
Pleurospermeae M. F. Watson & S. R. Downie, Trib. Nov. Tribus generum bracteis latis, saepe albomarginatis, sed haud omnino sic. Type genus: Pleurospermum Hoffm. in Gen. Pl. Umbell., ed. 1: VIII (1814). Other included genera: Aulacospermum Ledeb., Eleutherospermum K. Koch, Molopospermum W. D. J. Koch, Physospermum Cusson ex Juss.
The majority of taxonomists have regarded Physospermum and Pleurospermum sensu lato (including Aulacospermum, Eleutherospermum, and several other genera) as related. Sprengel (1820) treated both in his tribe Smyrnieae, and most authors have followed this example with some adjustment of rank. Molopospermum has a complicated taxonomic history; Bentham (1867) was the first to treat it alongside Physospermum, and Cerceau-Larrival (1962) placed the monotypic (and invalidly published) Molopospermeae next to a reduced Smyrnieae (Physospermum, Pleurospermum, and Smyrnium). Kr&aum
