P u b l i s h i n g australian s y s t e m at i c b ota n y Volume 15, 2002 © CSIRO 2002 An inter national jour nal devoted t o t h e t a x o n o my, b i o g e o g r a p hy and evolution of all plant groups All enquiries and manuscripts should be directed to: Australian Systematic Botany CSIRO Publishing PO Box 1139 (150 Oxford St) Collingwood, Vic. 3066, Australia Telephone: +61 3 9662 7613 Fax: +61 3 9662 7611 Email: publishing.asb@csiro.au Published by CSIRO Publishing for CSIRO and the Australian Academy of Science w w w. p u b l i s h . c s i ro . a u / j o u r n a l s / a s b 7 May 2002 Australian Systematic Botany 15, 181–191 The phylogenetic and taxonomic position of Lilaeopsis (Apiaceae), with notes on the applicability of ITS sequence data for phylogenetic reconstruction Gitte PetersenAC, Ole SebergA and Sidsel LarsenB A Botanical Institute, University of Copenhagen, Gothersgade 140, DK-1123 Copenhagen K, Denmark. B Present address: Zoological Museum, University of Copenhagen, Universitetsparken 15, DK-2100 Copenhagen Ø, Denmark. C Corresponding author; email: gittep@bot.ku.dk SBetPhG.0yPeal0ogte1l. 0ern7sictnnadtaoxnmoicopsit nofo G.Peterse,O.nSebgreanS.dLarsne Abstract. The relationships of the genus Lilaeopsis Greene have been difficult to determine primarily due to its simple morphology with entire, linear to spathulate leaves, simple umbels and lack of carpophore. Consequently, the genus has been referred to both Hydrocotyloideae and Apioideae. DNA sequence data from different genes (rbcL and matK) and non-coding regions (rpoC1 intron and ITS) were explored in order to determine the phylogenetic relationships of Lilaeopsis. Separate and combined analyses of rpoC1 intron, rbcL and matK data give almost congruent results with respect to a clade including Lilaeopsis. The three species of Lilaeopsis included in the analyses form a monophyletic group within the Oenanthe clade and hence belong to the Apioideae. Presently, the Mexican genus Neogoezia Hemsl. is considered its most likely sister group. Optimisation alignment of the ITS sequence data results in widely different phylogenetic hypotheses, but on the basis of congruence tests, two very similar trees can be selected. These trees are, however, incongruent with the trees obtained from analyses of the chloroplast sequences. As alignment of the ITS sequences is problematic even within the Oenanthe clade, the reliability of phylogenies on the basis of ITS data of higher taxonomic levels in the Apiaceae is questioned. However, for phylogenetic analysis within Lilaeopsis ITS sequences may prove useful and unproblematic. Introduction The taxonomic position of the genus Lilaeopsis Greene has been controversial for more than a century, the main question being whether its closest relatives are found in the Hydrocotyloideae or the Apioideae. Lilaeopsis has recently been monographed (Affolter 1985), but this work contributes little to the solution of the problems of its relationships to other genera or to elucidating interspecific relationships. The first species today recognised as belonging to Lilaeopsis was described by Linnaeus (1753) as Hydrocotyle chinensis L. It was transferred to Crantzia by Nuttall (1818) and in the subsequent years a number of new species were described. However, Crantzia is a younger homonym and Greene (1891) published the replacement name Lilaeopsis to acknowledge its vegetative resemblance to Lilaea Humb. & Bonpl. (Juncaginaceae). In the latest monograph, Affolter (1985) recognised 13 species and recently one more has been described (Petersen and Affolter 1999). The species of Lilaeopsis are either truly aquatics or occupy moist, fresh or brackish water habitats. Nine species occur in the Americas, two in Australia, two in New Zealand and one on Mauritius in the Indian Ocean. An unidentified © CSIRO 2002 species has been reported from Kerguelen and a dubious record exists from Madagascar (Raynal 1977; Affolter 1985). The predominantly Southern Hemispheric distribution pattern calls for a cladistic, biogeographic analysis (Linder and Crisp 1995; Humphries and Parenti 1999), but prior to cladistic analysis of the genus its monophyly needs to be demonstrated and its sister group relationships should be clarified. The difficulties encountered in placing the genus on the basis of morphology alone stem from its apparently strongly reduced habit. All species are tiny, perennial, rhizomatous herbs, with linear to spathulate, simple, septate leaves. The inflorescences are few-flowered, simple umbels and the schizocarp lacks a free carpophore (Affolter 1985). Early taxonomists tended to regard these characteristics as indications of relationship to genera in the Hydrocotyloideae (e.g. Nuttall 1818; Endlicher 1836–1840). Candolle (1829) also maintained Lilaeopsis (as Crantzia) in his Hydrocotylinées, but stated that despite the simple inflorescence, Lilaeopsis might be related to the apioid genus Oenanthe. Bentham and Hooker (1867) formally transferred Lilaeopsis (as Crantzia) to the tribe Seselineae, 10.1071/SB01007 1030-1887/02/02181 182 again stressing its resemblance to Oenanthe. Drude (1898), who provided the classification of the Apiaceae still largely in use, also placed Lilaeopsis (as Crantzia) close to Oenanthe in the tribe Apieae (=Ammineae), subtribe Seselinae. In recent years, a number of phylogenetic analyses of the Apiaceae and the Araliaceae have been published. On the basis of morphology, Judd et al. (1994) showed that in their traditional circumscriptions neither of the families is monophyletic, but together they form a monophyletic group. The Saniculoideae was a sister group to the Apioideae and both subfamilies plus the genus Myodocarpus Brongn. & Gris (traditionally part of the Araliaceae) form a sister group to the Hydrocotyloideae. However, the analysis was limited to only 12 taxa and 18 characters. These characters are easily scored and a re-analysis of the matrix (incl. Lilaeopsis) results in an almost unresolved tree. Only two clades, the Apioideae and the Saniculoideae, both including two taxa, remain. The unresolved relationships are probably caused by the character distribution of Lilaeopsis, which combines character states found in the Apiaceae sensu stricto and in the Araliaceae (incl. Hydrocotyloideae p.p.). A much denser taxon sampling is needed in order to substantiate the morphological analysis. Several phylogenetic analyses on the basis of molecular data and including many more taxa have been made, but none of these analyses include Lilaeopsis. An analysis of chloroplast rbcL sequence data confirmed the sister group relationship of the Apioideae and the Saniculoideae, but with the Hydrocotyloideae being polyphyletic (Plunkett et al. 1996a). The same relationships between Apioideae and Saniculoideae were revealed by sequence data from another chloroplast gene, matK (Plunkett et al. 1996b) and by a combined analysis of both data sets (Plunkett et al. 1997). The relationships within the Apioideae have been explored primarily by Downie, Katz-Downie and co-workers by the use of sequence data from introns of the chloroplast genes rpoC1 (Downie et al. 1996), rps16 (Downie and KatzDownie 1999; Downie et al. 2000b) and rpl16 (Downie et al. 2000a), the nuclear rDNA internal transcribed spacers (ITS) (Downie and Katz-Downie 1996; Katz-Downie et al. 1999) and chloroplast RFLPs (Plunkett and Downie 1999). With respect to the major clades, these analyses provide largely congruent results, but discrepancies mainly within the larger clades exist. These will be addressed whenever relevant to the present study. Within the Apioideae, the phylogenies are inconsistent not only with the tribal classifications of Drude (1898), but also with more-recent classifications (Koso-Poljansky 1916; Cerceau-Larrival 1962). Hence, Plunkett et al. (1996b) coined a series of informal names for clades supported by the molecular data. In the papers by Downie and co-workers these names were adopted and will be referred to here. The present paper aims to demonstrate monophyly of Lilaeopsis and clarify phylogenetic relationship of the genus G. Petersen et al. by providing DNA sequence data from ITS, rbcL, matK and the rpoC1 intron and including these data in the already existing comprehensive matrices. Materials and methods In the lack of any infrageneric classification of Lilaeopsis three species were chosen, representing widely different parts of its geographic range, viz. L. novae-zelandiae from New Zealand (GPL9, origin unknown, cultivated), L. mauritiana from Mauritius (GPL8, H. Windeløw s.n., Le Val Nature Park, Mauritius, 5.iii.1992) and the American L. carolinensis (GPL4, J. Bogner s.n., origin unknown, cultivated). Voucher specimens are deposited at C. Table 1 lists the GenBank accession numbers and authors of the species used for phylogenetic analyses. Table 2 is a summary of morphological characters and their distribution among taxa that are considered relevant to the discussion of the relationships of Lilaeopsis. The data have been extracted from a number of published sources (see Table 2) and supplemented with personal observations. Total DNA was extracted from fresh leaves according to the procedure of Doyle and Doyle (1987). PCR and sequencing were performed under standard conditions. Products from PCR and cycle sequencing were purified by the use of the QIAquick PCR Purification Kit (QIAGEN) according to manufacturers instructions and cycle sequencing was performed by the use of the ABI PRISM Dye Terminator Cycle Sequencing Ready Reaction Kit with AmpliTaq DNA Polymerase, FS (PE Biosystems). DNA fragments were separated on an ABI 377 (PE Biosystems) automated sequencer and sequence manipulation and/or editing was done with Sequencher 3.0. For amplification of the first 1200 bp of rbcL we used the primers RH-1S and Z1204RS for PCR (from G. Zurawski, DNAX Research Institute, Palo Alto, California, but slightly reduced in length). For sequencing the primers Hv362R (TGAACCCAAATACGTTACCCA), Hv522 (TAAACCAAAATTGGGATTA-TCCGC) and Hv890 (TGCATGCAGTTATTGATAGAC) were also used. For amplification and sequencing of about 1200 bp of matK, we used the primers matK710F, matK1176F, matK2000R (Plunkett et al. 1996b) and matK1168R (Johnson and Soltis 1995). For amplification of the rpoC1 intron primers, rpoC1e1 and rpoC1e2R located in the flanking exons were used for PCR and sequencing and one additional primer, rpoC1i2, located in the intron, was used for sequencing (Downie et al. 1996). The amplified products were about 830 bp long. The ITS regions including 5.8S rDNA were amplified with the primers ITS5 L (Hsiao et al. 1994) and ITS4 (White et al. 1990) and for sequencing, primers ITS2 and ITS3 (White et al. 1990) were also included. The amplified product is about 700 bp long. Phylogenetic analyses were performed in a stepwise procedure. Initial individual analyses were performed of three relatively large matrices containing rbcL, matK and rpoC1 intron data. For reasons given below, the ITS sequences were not initially run alone or combined with the chloroplast sequences. The rbcL matrix was compiled from Plunkett et al. (1996a, 1997), Kondo et al. (1996) and Backlund and Bremer (1997), and included taxa from Apiaceae, Araliaceae and Pittosporaceae. Pittosporaceae was included as an outgroup according to its position as a sister group to Apiales (Plunkett et al. 1996a). The 48 sequences from Plunkett et al. (1996a) were supplemented with 23 Apiaceae sequences from Kondo et al. (1996) (only sequences from genera not already represented were included), plus a single sequence each from Backlund and Bremer (1997; Steganotaenia Hochst.) and Plunkett et al. (1997; Cheirodendron Nutt. ex Seem.). The final matrix, including the three Lilaeopsis sequences, contained 76 taxa and 1428 positions. With no length difference between the sequences, alignment was straightforward. Phylogenetic and taxonomic position of Lilaeopsis Table 1. 183 GenBank accession numbers for sequences from taxa included in the Oenanthe clade and used in the phylogenetic analyses Taxon matK Bifora americana (DC.) S.Watson U58551 Cicuta douglasii (DC.) J.M.Coult. & Rose U58555 Cicuta virosa L. — Cryptotaenia canadensis (L.) DC. — (syn. C. japonica Hassk.) Lilaeopsis carolinensis J.M.Coult. & Rose AF466270 Lilaeopsis mauritiana G.Petersen & Affolter AF466271 Lilaeopsis novae-zelandiae (Gand.) A.W.Hill AF466272 Neogoezia minor Hemsl. U58570 Oenanthe javanica DC. — Oenanthe pimpinelloides L. — rbcL rpoC1 ITS1/ITS2 — — D44561 D44565 — — U72447 U72445 AF466267 AF466268 AF466269 U50228 D44577 — AF466273 AF466274 AF466275 — — U72442 — — — U79613/ U79614 AF466276 AF466277 AF466278 — — U78371/ U78431 — — — U78373/ U78433 U78400/ U78460 U78370/ U78430 — — Oenanthe sarmentosa J.Presl. Oxypolis occidentalis J.M.Coult. & Rose Perideridia gairdneri (Hook. & Arn.) Mathias Perideridia kelloggii (A.Gray) Mathias U58571 — U58574 — — — — — — U72444 — U72446 Shoshonea pulvinata Evert & Constance U58579 — U72413 — — U72443 — U58580 D44587 — — — Sium latifolium L. Sium serra (Franch. & Sav.) Kitag. Sium suave Walter The matK matrix was compiled from Plunkett et al. (1996b, 1997), including taxa from Apiaceae, Araliaceae and Pittosporaceae; the latter was used as an outgroup. The matrix contained 83 taxa and 1131 positions. Alignment of the full matrix (incl. Lilaeopsis) was straightforward with gaps (all in frames of three) introduced according to Plunkett et al. (1996b, 1997). The rpoC1 intron sequences from Lilaeopsis were included in the matrix of Downie et al. (1998), kindly provided by S. Downie (UIUC, Illinois). Inclusion of the Lilaeopsis sequences in the matrix was unproblematic. The new matrix contained 99 taxa and 952 positions (an increase of four in comparison to the original). The introduction of two new gaps (both 2 bp) was unproblematic. Parsimony jackknifing was used to identify the well-supported clades found in each of the three matrices using Xac (J. S. Farris, unpubl. data). With a removal rate of P = e–1, groups that occur with a frequency higher than c. 67% are supported by at least one uncontradicted character (Farris et al. 1996). Branches with a jackknife support between 67 and 50% are expected to have some support. Xac analyses were performed with the command (1000 */5) allowing 1000 replicates, five random input orders and branch swapping. Only groups with a jackknife support >50% were retained. Subsequently, the three chloroplast data sets were combined. In order to decrease the number of missing entries in the combined matrix, genera were used as terminals and at least two sequences should be available for a genus to be included. Whenever sequence data existed for the same species these were used. If sequences from different species existed from each of the two or three data sets, the genus was a priori assumed to be monophyletic and the sequences combined irrespective of any indications of non-monophyly by the individual analyses. If two or more sequences from different species existed for the same data set and none of the species were included in any of the other data sets, a consensus sequence was constructed. The consensus sequences were constructed by the use of the IUPAC ambiguity code for sites with nucleotide variation and question marks were used when length differences were encountered. The combined matrix contained 72 terminals (genera) and 3452 positions. The number of positions is lower than the sum of positions in the original matrices as some gaps were removed due to exclusion of taxa. A phylogenetic analysis was performed with PAUP*4.0b2a (Swofford 1999) by the use of 100 random addition orders, TBR (tree bisection and regrafting) swapping, steepest decent and by keeping five trees per step. Parsimony jackknifing was performed as for analyses of the individual data sets by the use of Xac. Whereas sequences from the three chloroplast regions were readily alignable, the ITS1 and ITS2 sequences from the Apiaceae (Downie et al. 1998) showed considerable length differences. Downie et al. (1998) used CLUSTAL V (Higgins et al. 1992) to make a preliminary alignment (subsequently adjusted manually) of 95 Apiaceae ITS sequences. The final alignment had a length of 485 positions of which 48 were considered ambiguous and excluded from the phylogenetic analyses (Downie et al. 1998). The aligned matrix was kindly put at our disposal. However, we failed to add the Lilaeopsis sequences to this matrix in an ‘unambiguous manner’ and inspection of the matrix did not convince us that the original alignment was unambiguous either [see Downie and Katz-Downie (1996) for a published alignment of 40 taxa]. Hence, no attempts were made to run the chloroplast sequences and the ITS sequences in a combined analysis. CLUSTAL V alignments are based on a guide tree which is based on distance measures, whereas a methodology on the basis of the principle of optimisation alignment is preferable in a phylogenetic context (Wheeler 1996). Optimisation alignment is implemented in the programme POY version 2.0 (Gladstein and Wheeler 2000). POY analysis is, however, extremely computer intensive and we were only able to perform it on a subset of the sequences. To select a relevant subset we used the results from our previous analysis of the combined chloroplast data. Sequences from all taxa included in the same clade as 184 G. Petersen et al. Table 2. Morphological characters for taxa included in the Oenanthe clade by one or more cladistic analyses Data from Coulter and Rose (1900), Drude (1898), Hegi (1926), Constance (1987), Evert and Constance (1982), Hasskarl (1858), Mathias and Constance (1944–1945), Affolter (1985) HabitatA Roots Leaves Umbels Simple, linear, septate Pinnate Pinnate Pinnate Simple Simple Compound Compound Pinnate/ternate/ simple, linear, septate Pinnate Ternate Digitate/pinnate/ simple, linear, septate Pinnate Lilaeopsis Greene Neogoezia Hemsl. Bifora Hoffm. Sium L. s.str. Wet Wet Dry Wet Berula Besser & W.D.J.Koch Wet Thin, fibrousC Tuberous fascicled Slender taproots Fusiform-thickened/ fibrous fascicled Fibrous fascicled Oxypolis Raf. Wet Tuberous fascicled Helosciadium W.D.J.Koch Cryptotaenia DC. Cynosciadium DC. Wet Dry Wet Fibrous Slender fascicled Fibrous fascicled Cicuta L. Wet Fibrous/tuberous Oenanthe L. fascicled Wet Fibrous/tuberous Pinnate/ simple, linear, septate Perideridia Rchb. Dry Shoshonea Evert & Constance Dry Fusiform/tuberous fascicled Woody taproot Pinnate/ternate/ simple, linear Pinnate Pinnate AerenchymeB Carpophore Absent Free, bipartite or bifid Free, bipartite Bipartite, adnate to the mericarps Compound Bipartite, adnate to the mericarps Compound Free, bipartite Present Absent Absent Present Compound Free Compound Free, bipartite-bifid Compound Free, bifid Present No data Present Compound Free, bipartite (but splits late, also from mericarps) Compound Bipartite, adnate to the mericarps (mericarps split late) Compound Free, bipartite Present Compound Bipartite, adnate to the mericarps Present Absent Present Absent Absent A Dry refers to a variety of habitats: woodlands, meadows. Data collected from text and drawings. Any indication of ‘corky’ or ‘spongy’ tissue has been interpreted as aerenchyme, although it may not consist of similar cell types. C Minor root swellings observed in one species. B Lilaeopsis were selected (see Results and Table 1) and the available ITS sequences (eight) from species belonging to the genera included in this clade were used in the POY analyses. Only Lilaeopsis was represented by more than one species. On the basis of its position in the combined phylogenetic analysis Shoshonea was used as an outgroup. ITS1 and ITS2 sequences were treated as independent blocks. No external evidence is available for selecting the necessary parameters used in POY, e.g. gap versus substitution costs or transversion versus transition costs, for ITS. Hence, to select a ‘best’ set of costs, a sensitivity analysis (Wheeler 1995) exploring congruence measured by the ILD (incongruence-length difference) (Mickevich and Farris 1981) between the results from separate and combined analyses of ITS1 and ITS2 was performed. A total of 30 POY analyses with different cost matrices was performed. Transversion costs of 0.5, 1, 2, 4, 8 and 16 times transition costs were applied for each of the gap costs of 1, 2, 4, 8 and 16. Each analysis was performed with the commands: -random 20, -multibuild 20, -checkslop 100, -slop 50, -norandomizeoutgroup. The topological incongruence-length difference, TILD (Wheeler 1999), was calculated between all trees obtained from analyses of ITS1 + ITS2 and the topology of the clade including Lilaeopsis was derived through analysis of the chloroplast data (with the modification that all species of Lilaeopsis were included). Results The rbcL sequences from the three Lilaeopsis species contain only eight variable sites. After jackknifing of the rbcL matrix, including a total of 200 parsimony informative sites, Lilaeopsis was placed in a largely unresolved clade with Sium, Cicuta (C. virosa), Cryptotaenia, Oenanthe and Neogoezia (corresponding to the Oenanthe clade of Plunkett et al. 1996b) (Fig. 1). The only supported branches in the clade support monophyly of Lilaeopsis (jackknife proportions of 99), with L. novae-zelandiae sister group of Neogoezia Sium Cicuta Cryptotaenia 73 Oenanthe L. novae-zelandiae L. mauritiana 99 63 L. carolinensis Fig. 1. Topology of the Oenanthe clade following parsimony jackknife analysis of rbcL sequence data. Numbers below branches are jackknife support values. L. = Lilaeopsis. Phylogenetic and taxonomic position of Lilaeopsis the other two species. Only the Oenanthe clade is shown in Fig. 1 and in subsequent figures. The jackknife trees produced here are largely similar but less-resolved than the trees already shown in the publications presenting the original sequence data. The complete jackknife trees are available at www.bot.ku.dk/groups/monocot/. The matK sequences from the three species of Lilaeopsis contain a total of 10 variable sites. Analysis of the matK matrix, including a total of 336 parsimony informative sites, also placed Lilaeopsis in the Oenanthe clade (Fig. 2). Cryptotaenia, which was placed in the Oenanthe clade by the rbcL data, was not included in the matK analysis. The genus Cicuta, here represented by C. douglasii, is not a member of the Oenanthe clade, but is placed in the Angelica clade sensu Plunkett et al. (1996b). Bifora, Shoshonea and Perideridia are new members of the clade, with respect to the rbcL analysis in which none of the genera were represented. The clade is well-resolved and most branches are very well supported. The monophyly of Lilaeopsis is confirmed (jackknife proportions of 100) and again L. novae-zelandiae is placed as the sister group of the other species. The rpoC1 intron sequences from the three species of Lilaeopsis contain 11 variable sites (10 substitutions and one 1-bp indel). Analysis of the rpoC1 matrix, including a total of 272 parsimony informative sites, again placed Lilaeopsis in the Oenanthe clade (Fig. 3), together with Perideridia, Oenanthe, Cicuta (C. virosa), Sium, Cryptotaenia and Oxypolis. The latter genus was not represented in any of the previous analyses. Shoshonea, which was placed in the Oenanthe clade by the matK data, is not included in this clade according to the analysis of the rpoC1 intron data. Bifora and Neogoezia included in the rbcL or matK matrices were not included in this analysis. The Oenanthe clade is only moderately resolved, but Lilaeopsis remains monophyletic 185 Perideridia Oenanthe Cicuta 81 Sium 52 Oxypolis 75 Cryptotaenia L. novae-zelandiae L. mauritiana 100 98 L. carolinensis Fig. 3. Topology of the Oenanthe clade following parsimony jackknife analysis of rpoC1 intron sequence data. Numbers below branches are jackknife support values. L. = Lilaeopsis. (jackknife proportions of 100) again with L. novae-zelandiae being the sister group to the other species. Phylogenetic analysis of the combined matrix resulted in more than 57 900 equally parsimonious trees (length = 1635, ci = 0.50, ri = 0.83). Although the trees were distributed on three islands (Maddison 1991), the topological differences between the consensus trees constructed from each of the islands reside in a large clade, including Araliaceae and part of the Hydrocotyloideae only. Lilaeopsis was placed in a clade together with Cryptotaenia, Sium, Neogoezia, Oenanthe, Perideridia and Shoshonea (Fig. 4). Shoshonea was placed as the sister group to the remaining taxa in the clade, but this position did not have jackknife support above 50%. Cicuta (C. virosa and C. douglasii combined) was Perideridia 97 75 Oenanthe Shoshonea Shoshonea Perideridia 67 78 Sium Oenanthe Bifora Cryptotaenia 79 98 Neogoezia 86 89 Neogoezia L. novae-zelandiae 94 L. mauritiana 100 88 L. carolinensis Fig. 2. Topology of the Oenanthe clade following parsimony jackknife analysis of matK sequence data. Numbers below branches are jackknife support values. L. = Lilaeopsis. Sium Lilaeopsis Fig. 4. Topology of the Oenanthe clade in the strict consensus tree (length = 1635, ci = 0.50, ri = 0.83) based on more than 57 900 equally parsimonious trees reconstructed from a combined phylogenetic analysis of data from three chloroplast sequences. Numbers below branches are jackknife support values. 186 G. Petersen et al. placed as a sister group to a larger clade including 19 other Apioideae genera. Oxypolis and Bifora were not included in the combined analyses as data from only one gene were available for each. The complete consensus tree is available at www.bot.ku.dk/groups.monocot/. One ITS1 and ITS2 sequence from each of the genera Cryptotaenia, Sium, Oenanthe, Perideridia and Shoshonea plus the three Lilaeopsis sequences were included in POY analyses. Following the previously described search strategy, most analyses recovered the shortest tree(s) in all of the 20 replicates and never in fewer than five of them. The lowest ILD value (0.00459) was obtained at a transversion : transition cost of 2 : 1 and a gap cost of 1. Optimisation alignment of ITS1 + ITS2 under these costs resulted in only one tree (Fig. 5a). A Shoshonea Perideridia Cryptotaenia Sium Oenanthe L. carolinensis L. mauritiana L. novae-zelandiae B Shoshonea Perideridia Cryptotaenia Sium Oenanthe L. novae-zelandiae L. mauritiana L. carolinensis Fig. 5. Topologies of the Oenanthe clade according to optimisation alignment of ITS1 and ITS2 sequences. (A) Topology obtained with the cost matrix TV: TS : gap 2 : 1 : 1 minimising ILD between ITS1 and ITS2. (B) Topology obtained with cost matrices TV : TS : gap 1 : 2 : 8, 2 : 1 : 2, 4 : 1 : 2, 8 : 1 : 4, 16 : 1 : 8 minimising TILD between topologies obtained from ITS and the combined chloroplast data. L. = Lilaeopsis. Under the 30 different cost functions, only nine different trees were found from analyses of ITS1 + ITS2. None of these were congruent with the topology of the tree derived through combined analysis of the three chloroplast genes and the only clade in common between all nine trees was a monophyletic Lilaeopsis. The lowest TILD value (0.0769; TILDN = 0.0588) was for a tree (Fig. 5b) obtained by the use of several cost matrices (TV : TS : 3gap 1 : 2 : 8, 2 : 1 : 2, 4 : 1 : 2, 8 : 1 : 4, 16 : 1 : 8). This tree differs from the tree obtained by the use of the 2 : 1 : 1 cost matrix by the resolution within Lilaeopsis only. Discussion Applicability of ITS for phylogenetic reconstruction Under the assumption that the ITS sequences included in the present analyses are homologous, i.e. the copies have been completely homogenised, it should be possible to make homology statements for every site included in the sequences. However, a very large number of different hypotheses of positional homology (alignments) can be made and our possibilities for objectively choosing any one as ‘the best’ are very limited. Computer programmes performing multiple alignment or optimisation alignment remove part of the subjectivity that burdens manual alignments and manually adjusted alignments. However, the choice of cost matrices is still subjective. Hence, alternatively, cost matrices specifying a broader spectrum of different evolutionary patterns observed in other sequences or other organisms may form the basis of a sensitivity analysis by testing the robustness of the resulting tree topology when the underlying cost matrix is perturbed (Wheeler 1995; Phillips et al. 2000). If alternative evidence exists for the phylogeny of a given group, congruence between this phylogeny and the trees constructed with different cost matrices may be used as a means of selecting the best estimate of the underlying evolutionary pattern. This approach is, however, not free from shortcomings. Any real conflict between data sets, e.g. conflict between nuclear and chloroplast data caused by hybridisation, will be hidden. In the present case, we observe that the tree topologies obtained by the ITS data are sensitive to changes in the cost matrix and that monophyly of Lilaeopsis is the only resolution shared between the trees. No cost matrix results in a tree congruent with the tree derived the combined chloroplast data. Because of assumed differences in the evolutionary constraints on ITS and the chloroplast sequences, ILD tests have not been performed. In the protein-coding genes rbcL and matK, only indels maintaining the reading frame (i.e. in multiples of three) occur (in rbcL, no length variation is observed at all) and in the rpoC1 intron, most length differences seem to be caused by smaller, 1–7-bp duplications. No such simple patterns are discernable in the ITS sequences. In the present Phylogenetic and taxonomic position of Lilaeopsis implementation of POY, different data sets have to be analysed by the use of the same cost matrix; consequently, simultaneous analyses of ITS and the chloroplast sequences have not been attempted. As an alternative, two different strategies were followed. First, an ILD test was performed with ITS1 and ITS2 treated as independent data sets. The ITS1 and ITS2 sequences are likely to evolve according to the same pattern (Baldwin et al. 1995) and ILD values are indeed very low for all sets of cost (data not shown). The cost matrix resulting in the lowest ILD value has a moderate TV : TS cost ratio (2 : 1) and a low gap cost (1). Second, congruence between the topologies on the basis of ITS under different costs and the topology obtained from the combined analyses of the three chloroplast data sets (Fig. 4) was explored by the use of the TILD, which avoids the problems of weight functions and differences in taxon sampling. This approach showed that a single tree (Fig. 5b), obtained by the use of a number of different cost matrices, was most congruent with the chloroplast-based tree. The tree differs from the one obtained by the 2 : 1 : 1 cost matrix only by topology of the Lilaeopsis clade. The latter approach assumes a priori that the topology constructed from the chloroplast data is preferable, but this assumption may of course be wrong. If, instead, any of the nine topologies derived from the ITS data was assumed to represent the ‘true phylogeny’, it would be equally relevant to explore topological congruence between this and the topologies derived under different weighting schemes applied to the chloroplast data. However, this approach has not been followed as most of the weights used in analyses of the ITS data lie far beyond empirically derived weighting schemes for chloroplast data. For example, for rbcL and rpoA, the estimated transition : transversion bias resulted in no more than a 1.2 upweighting of transversions (Albert et al. 1993; Petersen and Seberg 1997). The problems handling the ITS data even within the Oenanthe clade have far-reaching implications for the reliability of the phylogenetic implications drawn from the analyses of higher taxonomic levels in the Apiaceae on the basis of ITS (e.g. Valiejo-Roman et al. 1998; Katz-Downie et al. 1999; Downie et al. 2000b). Therefore, any reference below to relationships suggested by ITS analyses should be treated with caution. The distribution of morphological characters in the Oenanthe clade Despite the differences in phylogenetic relationships suggested by the ITS data and the chloroplast data, the lack of full resolution in some of the phylogenetic trees presented here and the uncertainty about the inclusion of some taxa in the Oenanthe clade, it is still possible to discuss some of the more or less unusual morphological character traits possessed by Lilaeopsis and other members of the clade. The characters relevant to the following discussion are listed in Table 2. 187 Downie et al. (2000a) characterised the Oenanthe clade as consisting of glabrous plants with clusters of tubers or tuberous roots, often growing in wet or moist habitats. Truly, all the presently recognised members of the clade are glabrous or almost so, as are most taxa in the Apiaceae. The more- or less-thickened, often fascicled roots also occur in many taxa outside the Oenanthe clade and thin, fibrous roots are not uncommon within the clade, even if the questionable members (see below) are excluded. The preference for wet or even truly aquatic habitats is more pronounced, especially as the inclusion of Bifora and Shoshonea in the Oenanthe clade may be questioned (see below). A number of morphological traits of limited occurrence in the Apioideae characterise some members of the Oenanthe clade. The linear, septate leaves of Lilaeopsis and other taxa in the clade are usually interpreted as homologous to the rachis of a normal compound leaf and the position of septae corresponds to the position of pinnae in compound leaves (Kaplan 1970; Charlton 1992). According to Affolter (1985), similar leaves can be found in the genera Cynosciadium, Eryngium L., Limnosciadium Mathias & Constance, Oenanthe, Ottoa Humb., Bonpl., & Kunth, Oxypolis, Perissocoeleum Mathias & Constance and Ptilimnium Raf. Affolter (1985) suggested that this trait had evolved repeatedly in the Apiaceae. Obviously, the simple leaves found in Eryngium, which by all available evidence belongs to the Saniculoideae (e.g. Plunkett et al. 1997; Downie et al. 2000a), are not homologous to the simple leaves found in members of the Oenanthe clade. The following question remains then: how many of these genera belong to the Oenanthe clade. Both according to the present (Fig. 3) and previous phylogenetic analyses of rpoC1 intron data (Downie et al. 1998) and according to analyses of other chloroplast regions (Downie et al. 2000a), Oxypolis is a member of the Oenanthe clade. On the basis of unpublished data, Downie et al. (2000a) provided phylogenetic evidence also for the inclusion of Cynosciadium in the Oenanthe clade. This genus comprises wetland species with aerenchymatic tissue in the mericarp, but with free carpophores (see below). Also Ptilimnium and Limnosciadium are strong candidates for inclusion in the Oenanthe clade and Downie et al. (2000a, 2001) include both genera in the Oenanthe clade. However, they do not provide any data supporting the inclusion. Both genera consist of glabrous, annual, wetland species with pinnate or simple leaves, compound umbels, free bifid carpophores and aerenchymatic tissue in the mericarps and they are usually considered closely related to Cynosciadium (Limnosciadium was originally included in Cynosciadium). However, Perissocoeleum and Ottoa are less-likely candidates for inclusion in the Oenanthe clade. The South American genus Perissocoeleum consists of glabrous or pubescent perennials, with either linear, septate (one species) or pinnately decompound (three species) leaves, compound umbels and a free, bipartite carpophore (Mathias and 188 Constance 1952, 1967). The mericarps do not posses aerenchymatic tissue and the fruit is described as ‘peucedanoid’ by Mathias and Constance (1952). Perissocoeleum is more likely a member of the Angelica clade. The monotypic Central American Ottoa is a glabrous perennial with fusiform roots, simple, septate leaves and a free bipartite carpophore (Mathias and Constance 1944–1945). Information about the texture of the fruits is unavailable. Although originally placed next to Oenanthe (Humboldt et al. 1815–1825), Coulter and Rose (1895) considered Ottoa related to Arracacia Bancr., which, although being non-monophyletic, is a member of the Angelica clade (e.g. Downie et al. 2000a). Hence, Affolter (1985) was probably correct in postulation that the linear, septate leaves had evolved repeatedly. Even within the Oenanthe clade, linear, septate leaves cannot be considered a synapomorphy for any subset of taxa; on the contrary, the reduction from pinnate to simple leaves must have occurred several times. It is likely that the linear leaves are an adaptation to an aquatic or semi-aquatic environment, although in some species (Perideridia) only the upper leaves are reduced. Presence of simple umbels characterises the Hydrocotyloideae and the Saniculoideae, but is extremely rare in the Apioideae. Apart from Lilaeopsis, the only apioid genera with simple umbels are Neogoezia, Oreomyrrhis Endl. and possibly Naufraga Constance & Cannon (Constance and Cannon 1967; Constance 1987; Downie et al. 2000b). According to the present phylogenetic analyses, the Mexican genus Neogoezia may be the sister group to Lilaeopsis (Fig. 4). This relationship is strongly supported by the shared presence of simple umbels. The monotypic, Balearic endemic Naufraga was originally included in the Hydrocotyloideae because of the simple umbel and complete lack of carpophore (Constance and Cannon 1967) and a recent phylogenetic analysis on the basis of morphological characters places Naufraga within Hydrocotyloideae (Henwood and Hart 2001). However, analyses of rps16 and ITS data place Naufraga as the sister group to Apium graveolens L. in the Apium clade (Downie et al. 2000b). The inflorescence of Apium L. s.str. is a sessile umbel, lacking involucre, but subtended by stem leaves. Given their position as sister groups, the Naufraga inflorescence may be interpreted in the same way. The typical Gwondanan distribution of the genus Oreomyrrhis largely resembles that of Lilaeopsis and Drude (1898) placed Oreomyrrhis next to Neogoezia. However, in their monograph of the genus, Mathias and Constance (1955, p. 349) did not consider Oreomyrrhis closely related to any other genus with simple umbels, but ‘having a vague relationship to Chaerophyllum’, which according to the molecular analyses is a member of a monophyletic group within the Daucus clade, corresponding quite closely to Drude’s Scandiceae, subtribe Scandicinae (Downie et al. G. Petersen et al. 2000a). Ongoing molecular studies seem to confirm a close relationship to Chaerophyllum (Kuo-fang Chung, WU, Missouri, pers. comm.). The species of Oreomyrrhis are alpine, occasionally wetland perennials with taproot, glabrous to tomentose pinnate or simple leaves, a free bipartite or bifid carpophore and mericarps ‘sometimes slightly corky’ (Mathias and Constance 1955, p. 355). The simple leaves possessed by two species from New Guinea are not septate, but they may have a number of incurved, often opposite teeth at the apex, which may suggest that they are reduced forms of pinnate leaves consisting primarily of the petiole. Whether the corky tissue in the mericarps is similar to the aerenchyma found in Lilaeopsis remains unknown (see below). But given the more likely relationship to Chaerophyllum, there is little evidence for the inclusion of Oreomyrrhis in the Oenanthe clade. The lack of a free carpophore is not common in the Apiaceae. The Hydrocotyloideae is characterised by the lack of a free carpophore, but in the remaining subfamilies the character state is rare. However, in the Oenanthe clade several genera have an either poorly developed (Oenanthe) or entirely lacking carpophore (Lilaeopsis, Shoshonea) or a carpophore that is bipartite with the halves more or less strongly adnate to the mericarps (Sium, Berula and Shoshonea) (Drude 1898; Hegi 1926; Mathias and Constance 1944–45; Evert and Constance 1982). According to Hegi (1926), Cicuta has a free, bipartite carpophore, which, however, splits only reluctantly from the mericarps. Several members of the Oenanthe clade have free, bipartite or bifid carpophores as do most apioid taxa; hence, the absence of a free carpophore is neither a synapomorphy for the clade nor for any subset of taxa. Lacking a free carpophore may be found also outside the Oenanthe-clade taxa, e.g. Naufraga (Apium clade) and Thaspium Nutt. (Angelica clade). Thus, reduction of the carpophore or adnation to the mericarps seem to have occurred several times during the evolution of the Apioideae. The presence of aerenchymatic tissue (‘spongy cells’) in the mericarps was discussed by Affolter (1985), who stated that this character might be related to habitat preference. The aerenchymatic tissue enhances the buoyancy of the mericarps and is primarily, but not exclusively, found in species occupying wet habitats. The aerenchymatic tissue of Lilaeopsis consists of storage tracheids (Affolter 1985), but detailed anatomical studies of the fruits from most other genera are lacking. Hence, the scoring of aerenchymatic tissue in Table 2 is based on drawings of fruit transections or descriptions of the fruits as ‘corky’, ‘schwammig’, or with ‘schwimmgewebe’ and may not reflect homologous structures. Similar descriptions can be found for a number of genera outside the Oenanthe clade, but it is possible that the type of aerenchymatic tissue found in most members of the Oenanthe clade is different from that found in taxa outside the clade. Phylogenetic and taxonomic position of Lilaeopsis In conclusion, none of the characters discussed above can be considered a synapomorphy for the Oenanthe clade. In general, there is only limited agreement between the traditional taxonomy and the molecular hypotheses for the entire Apioideae as discussed by Downie and co-workers (e.g. Downie et al. 2001; Spalik and Downie 2001). A largescale phylogenetic analysis of morphological characters is highly needed. Mapping the characters on the phylogenetic hypothesis for the Oenanthe clade presented here seems premature. A better resolved phylogeny is highly desirable, more taxa need to be added and there is uncertainty about the inclusion of some taxa. Taxa with questionable relationship to the Oenanthe clade The monotypic genus Shoshonea has been included in the analyses of matK (Plunkett et al. 1996b), the rpoC1 and rpl16 introns and ITS (Downie et al. 1998, 2000a). Whereas matK places Shoshonea in the Oenanthe clade, all other analyses place Shoshonea within the Angelica clade (Downie et al. 1998, 2000a), related to a group of primarily western North American taxa, among which Aletes J.M.Coult. & Rose, Neoparrya Mathias and Musineon Raf. have been suggested as close relatives by Evert and Constance (1982). Extensive incongruence between phylogenies on the basis of data from the plastids is rare, whereas phylogenies on the basis of data from different genomes, e.g. plastid versus nuclear sequences, frequently differ (e.g. Petersen and Seberg 1997). Plunkett and Downie (1999) considered the matK results spurious and Shoshonea misplaced in the Oenanthe clade. This is supported by Shoshonea not being a wetland plant and by a lack of aerenchymatic tissue in the mericarps. However, Shoshonea frequently lacks a free carpophore (Evert and Constance 1982). The relationships of Shoshonea are in need of further investigation. The genus Bifora has been included in three of the molecular studies. The Mediterranean B. radicans is included in the ITS data set. Another Mediterranean species, B. testiculata (L.) Spreng. ex Schult., is included in the rps16 data set and the North American B. americana is included in the matK data set. The ITS and rps16 analyses show B. radicans as a member of the Angelica clade, whereas B. testiculata according to the rps16 analysis appears to be a member of the Apium clade (Downie et al. 1998, 2000b). In contrast, the matK analysis places B. americana in the Oenanthe clade as the sister group to Neogoezia, Lilaeopsis or both (Fig. 2; Plunkett et al. 1996b). If the conflict is real, Bifora is not monophyletic. Bifora americana is the only American member of this otherwise Eurasian genus and was described by Candolle (1829) as Atrema americana DC., mainly deviating from the genus Bifora by its prominent calyx teeth. The presence of a free carpophore, the lack of aerenchymatic tissue in the mericarps and an overall very different fruit structure do not indicate relationships to the Oenanthe clade, but further studies are needed. 189 The genus Cicuta is represented by two different species in the phylogenetic analyses. Plunkett et al. (1996b) included the North American C. douglasii in their analysis of the matK data, whereas the Eurasian C. virosa was included in the analyses of rbcL (Kondo et al. 1996), the rpoC1 and rpl16 introns and ITS (Downie et al. 1998, 2000a). All analyses including C. virosa place the species in the Oenanthe clade (Kondo et al. 1996; Downie et al. 1998, 2000a) and C. douglasii as a member of the Angelica clade (Plunkett et al. 1996b). Although Mulligan (1980) proposed a hybrid origin of C. douglasii, there are no morphological indications that the genus is non-monophyletic. The genus Cicuta comprises wetland species with aerenchymatic tissue in the mericarps (Hegi 1926). Although usually supposed to have a free carpophore, the mericarps of C. virosa split only with difficulty from each other and from the carpophore (Hegi 1926). Hence, the genus seems to fit well into the Oenanthe clade, but the issue requires further studies. In addition to the taxa included in the Oenanthe clade in this study, Downie et al. (1998, 2000b) proposed Berula (B. erecta (Huds.) Coville and B. thunbergii (DC.) H.Wolff) and Helosciadium (H. inundatum W.D.J.Koch and H. nodiflorum W.D.J.Koch) as members of the Oenanthe clade on the basis of analyses of ITS and the rps16 intron (Downie et al. 1998, 2000b). These studies also included Oxypolis, Perideridia, Cryptotaenia, Cicuta, Sium and Oenanthe in the Oenanthe clade and both Berula and Helosciadium appear as members of a subclade including Oxypolis, Cryptotaenia and Sium (Downie et al. 1998, 2000b). Morphologically and ecologically, the two genera fit well in the Oenanthe clade. Berula is often included in Sium, with which it shares most of its characteristics. Helosciadium is often included in Apium s.lat., but according to the analyses of Downie et al. (2000b), Apium in this sense is non-monophyletic. Helosciadium comprises wetland species with aerenchymatic tissue in the mericarps and free carpophores. Aerenchymatic tissue is not present in Apium s.str. Phylogenetic reconstruction within Lilaeopsis The interspecific relationships in Lilaeopsis are largely unknown, but all the present phylogenetic analyses confirm monophyly of the genus. A phylogenetic analysis of the entire genus is highly desirable. Due to the simple morphology and the pronounced morphological similarity between the species, it will probably be possible to use only a few morphological characters in a phylogenetic analysis. The genes sequenced in the present study are most likely sufficiently variable for a phylogenetic analysis of the genus. The level of variation is low and only marginally different among the three chloroplast sequences: there are eight variable sites in rbcL and 11 in rpoC1 and 10 in matK and all sequences should be included to provide a sufficient number of characters. As a complement to the chloroplast sequences, 190 nuclear ITS sequences are most likely useful for a phylogenetic analysis within Lilaeopsis, although their value for higher-level analyses may be limited. According to the alignment implied by the cost matrix TV : TS : GAP 2 : 1 : 1, the three Lilaeopsis species differ at 54 positions. Also traditional multiple alignment of ITS sequences from Lilaeopsis seems possible and unambiguous. Given the uncertain phylogenetic relationships of members of Bifora, an appropriate outgroup in studies of the phylogeny of Lilaeopsis could include Neogoezia and other members of the Oenanthe clade confirmed by multiple studies. Acknowledgments We thank Stephen Downie (UIUC, Illinois) for kindly putting his molecular matrices at our disposal and for valuable comments and suggestions relating to the manuscript. Charlotte Hansen and Lisbeth Knudsen provided skilful technical assistance. The Faculty of Natural Sciences supported (M. Curie Grant) Gitte Petersen. References Albert VA, Chase MW, Mischler BD (1993) Character-state weighting for cladistic analysis of protein coding DNA sequences. Annals of the Missouri Botanical Garden 80, 752–766. Affolter JM (1985) A monograph of the genus Lilaeopsis (Umbelliferae). Systematic Botany Monographs 6, 1–140. Backlund A, Bremer B (1997) Phylogeny of the Asteridae s.str. based on rbcL sequences, with particular reference to the Dipsacales. Plant Systematics and Evolution 207, 225–254. Baldwin BG, Sanderson MJ, Porter JM, Wojciechowski MF, Campbell CS, Donoghue MJ (1995) The ITS region of nuclear ribosomal DNA: a valuable source of evidence on angiosperm phylogeny. Annals of the Missouri Botanical Garden 82, 247–277. Bentham G, Hooker JD (1867) ‘Genera plantarum. Vol. 1.’ (Reeve and Co.: London) Candolle AP de (1829) ‘Collection de mémoires pour servir á l’histoire du règne végétal. 5. Mémoire sur la famille des ombellifères.’ (Treuttel and Würtz: Paris) Cerceau-Larrival M-Th (1962) Plantules et pollens d’ombelliferes. Leur intérêt systématique et phylogénique. Mémoires du Muséum national d’Histoire naturelle, série B, Botanique 14, 1–166. Charlton WA (1992) The rachis-leaves of Lilaeopsis brasiliensis (Glaziou) J.M.Affolter (Apiaceae). Botanical Journal of the Linnean Society 109, 259–280. Constance L (1987) Neogoezia (Apiaceae), a very distinct and elegant genus of Mexican Umbelliferae. Opera Botanica 92, 59–71. Constance L, Cannon JFM (1967) Naufraga—a new genus of Umbelliferae from Mallorca. Feddes Repertorium 74, 1–4. Coulter JM, Rose JN (1895) Report on Mexican Umbelliferae, mostly from the state of Oaxaca, recently collected by C. G. Pringle and E. W. Nelson. Contributions from the US National Herbarium 3, 289–309. Coulter JM, Rose JN (1900) Monograph of the North American Umbelliferae. Contributions from the U. S. National Herbarium 7, 1–256. Downie SR, Katz-Downie DS (1996) A molecular phylogeny of Apiaceae subfamily Apioideae: evidence from nuclear ribosomal DNA internal transcribed spacer sequences. American Journal of Botany 83, 234–251. G. Petersen et al. Downie SR, Katz-Downie DS (1999) Phylogenetic analysis of chloroplast rps16 intron sequences reveals relationships within the woody southern African Apiaceae subfamily Apioideae. Canadian Journal of Botany 77, 1120–1135. Downie SR, Katz-Downie DS, Cho K-J (1996) Phylogenetic analysis of Apiaceae subfamily Apioideae using nucleotide sequences from the chloroplast rpoC1 intron. Molecular Phylogenetics and Evolution 6, 1–18. Downie SR, Ramanath S, Katz-Downie DS, Llanaas E (1998) Molecular systematics of Apiaceae subfamily Apioideae: phylogenetic analyses of nuclear ribosomal DNA internal transcribed spacer and plastid rpoC1 intron sequences. American Journal of Botany 85, 563–591. Downie SR, Katz-Downie DS, Watson MF (2000a) A phylogeny of the flowering plant family Apiaceae based on chloroplast DNA rpl16 and rpoC1 intron sequences: towards a suprageneric classification of subfamily Apioideae. American Journal of Botany 87, 273–292. Downie SR, Watson MF, Spalik K, Katz-Downie DS (2000b) Molecular systematics of Old World Apioideae (Apiaceae): relationships among some members of tribe Peucedaneae sensu lato, the placement of several island-endemic species, and resolution within the apioid superclade. Canadian Journal of Botany 78, 506–528. Downie SR, Plunkett GM, Watson MF, Spalik K, Katz-Downie DS, Valiejo-Roman CM, Terentieva EI, Troitsky AV, Lee B-Y, Lahham J, El-Oqlah A (2001) Tribes and clades within Apiaceae subfamily Apioideae: the contribution of molecular data. Edinburgh Journal of Botany 58, 301–330. Doyle JJ, Doyle JF (1987) A rapid DNA isolation procedure for small quatities of fresh leaf tissue. Phytochemical Bulletin 19, 11–15. Drude O (1898) Umbelliferae. In ‘Die natürlichen Pflanzenfamilien, vol. 3(8)’. (Eds A Engler, K Prantl) pp. 63–250. (Wilhelm Engelmann Verlag: Leipzig) Endlicher SL (1836–1840) ‘Genera plantarum secundum ordines naturales disposita.’ (Fr Beck: Vienna) Evert EF, Constance L (1982) Shoshonea pulvinata, a new genus and species of Umbelliferae from Wyoming. Systematic Botany 7, 471–475. Farris JS, Albert VA, Källersjö M, Lipscomb D, Kluge AG (1996) Parsimony jackknifing outperforms neighbor-joining. Cladistics 12, 99–124. Gladstein D, Wheeler WC (2000) ‘POY. Phylogeny reconstruction via direct optimization of DNA data. Version 2. 0.’ (American Museum of Natural History: New York) Greene EL (1891) Against the using of revertible generic names. Pittonia 2, 185–195. Hasskarl JK (1858) ‘Hortus bogoriensis descriptus.’ (F. Günst: Amsterdam) Hegi G (1926) ‘Illustrierte Flora von Mittel-Europa, vol. 5(2).’ (J. F. Lehmannns Verlag: München) Henwood MJ, Hart JM (2001) Towards an understanding of the relationships of Australian Hydrocotyloideae (Apiaceae). Edinburgh Journal of Botany 58, 269–289. Higgins DG, Bleasby AJ, Fuchs R (1992) CLUSTAL V: improved software for multiple sequence alignment. Computer Applications in the Biosciences 83, 189–191. Hsiao C, Chatterton NJ, Asay KH, Jensen KB (1994) Phylogenetic relationships of 10 grass species: an assessment of phylogenetic utility of the internal transcribed spacer region in nuclear ribosomal DNA in monocots. Genome 37, 112–120. Humboldt A von, Bonpland A, Kunth CS (1815–1825) ‘Nova genera et species plantarum.’ (Librariae graeco-latini-germanicae: Paris) Humphries CJ, Parenti LR (1999) ‘Cladistic biogeography. Interpreting patterns of plant and animal distributions. 2nd Edn.’ (Oxford University Press: Oxford) Phylogenetic and taxonomic position of Lilaeopsis 191 Johnson LA, Soltis DE (1995) Phylogenetic inference in Saxifragaceae sensu stricto and Gilia (Polemoniaceae) using matK sequences. Annals of the Missouri Botanical Garden 82, 149–175. Judd WS, Sanders RW, Donoghue MJ (1994) Angiosperm family pairs: preliminary phylogenetic analyses. Harvard Papers in Botany 5, 1–51. Kaplan DR (1970) Comparative development and morphological interpretation of ‘rachis-leaves’ in Umbelliferae. Botanical Journal of the Linnean Society 63 (suppl. 1), 101–125. Katz-Downie DS, Valiejo-Roman CM, Terentieva EI, Troitsky AV, Pimenov MG, Lee B, Downie SR (1999) Towards a molecular phylogeny of Apiaceae subfamily Apioideae: additional information from nuclear ribosomal DNA ITS sequences. Plant Systematics and Evolution 216, 167–195. Kondo K, Terabayashi S, Okada M, Yuan C, He S (1996) Phylogenetic relationships of medically important Cnidium officinale and Japanese Apiaceae based on rbcL sequences. Journal of Plant Research 109, 21–27. Koso-Poljansky BM (1916) Sciadophytorum systematis lineamenta. Bulletin de la Société impériale des Naturalistes (Moscou) 29, 93–222. Linder HP, Crisp MD (1995) Nothofagus and Pacific biogeography. Cladistics 11, 5–32. Linnaeus C (1753) ‘Species plantarum.’ (L. Salvii: Stockholm) Maddison D (1991) The discovery and importance of multiple islands of most parsimonious trees. Systematic Zoology 40, 315–328. Mathias ME, Constance L (1944–1945) Umbelliferae. In ‘North American flora. Vol. 28B’. pp. 43–295. (New York Botanical Garden: New York) Mathias ME, Constance L (1952) New South American Umbelliferae. Bulletin of the Torrey Botanical Club 79, 359–370. Mathias ME, Constance L (1955) The genus Oreomyrrhis (Umbelliferae). A problem in South Pacific distribution. University of California Publications in Botany 27, 347–416. Mathias ME, Constance L (1967) Some Umbelliferae of the Andean paramos of South America. Brittonia 19, 212–226. Mickevich MF, Farris JS (1981) The implications of congruence in Menidia. Systematic Zoology 30, 351–370. Mulligan GA (1980) The genus Cicuta in North America. Canadian Journal of Botany 58, 1755–1767. Nuttall T (1818) ‘The genera of North American plants. 2 vols.’ (D. Heart: Philadelphia) Petersen G, Affolter J (1999) A new species of Lilaeopsis (Apiaceae) from Mauritius. Novon 9, 92–94. Petersen G, Seberg O (1997) Phylogenetic analysis of the Triticeae (Poaceae) based on rpoA sequence data. Molecular Phylogenetics and Evolution 7, 217–230. Phillips A, Janies D, Wheeler W (2000) Multiple sequence alignment in phylogenetic analysis. Molecular Phylogenetics and Evolution 16, 317–330. Plunkett GM, Downie SR (1999) Major lineages within Apiaceae subfamily Apioideae: a comparison of chloroplast restriction site and DNA sequence data. American Journal of Botany 86, 1014–1026. Plunkett GM, Soltis DE, Soltis PS (1996a) Higher level relationships of Apiales (Apiaceae and Araliaceae) based on phylogenetic analysis of rbcL sequences. American Journal of Botany 83, 499–515. Plunkett GM, Soltis DE, Soltis PS (1996b) Evolutionary patterns in Apiaceae: inferences based on matK sequence data. Systemtic Botany 21, 477–495. Plunkett GM, Soltis DE, Soltis PS (1997) Clarification of the relationship between Apiaceae and Araliaceae based on matK and rbcL sequence data. American Journal of Botany 84, 565–580. Raynal J (1977) Le genre Lilaeopsis (Ombelliferes) à Madagascar. Adansonia 17, 151–154. Spalik K, Downie SR (2001) The utility of morphological characters for inferring phylogeny in Scandiceae subtribe Scandicinae (Apiaceae). Annals of the Missouri Botanical Garden 88, 270–301. Swofford DL (1999) ‘PAUP* 4. 0b2a: phylogenetic analysis using parsimony (*and other methods).’ (Sinauer Associates: Sunderland, MA) Valiejo-Roman KM, Pimenov MG, Terentieva EI, Downie SR, KatzDownie DS, Troitsky AV (1998) Molecular stystematics of the Umbelliferae: using nuclear ribosomal DNA internal transcribed spacer sequences to resolve issues of evolutionary relationships. Botanicheskii Zhurnal 83, 1–22. Wheeler WC (1995) Sequence alignment, parameter sensitivity, and the phylogenetic analysis of molecular data. Systematic Biology 44, 321–331. Wheeler WC (1996) Optimization alignment: the end of multiple sequence alignment in phylogenetics? Cladistics 12, 1–9. Wheeler W (1999) Measuring topological congruence by extending character techniques. Cladistics 15, 131–135. White TJ, Bruns T, Lee S, Taylor J (1990) Amplification and direct sequencing of fungal ribosomal RNA genes for phylogenetics. In ‘PCR protocols: a guide to methods and applications’. (Eds MA Innis, DH Gelfand, JJ Sninsky, TJ White) pp. 315–322. (Academic Press: San Diego) Manuscript received 9 April 2001, accepted 8 November 2001 http://www.publish.csiro.au/journals/asb