Phylogenetic analysis of nrDNA ITS sequences reveals relationships within five groups of Iranian Apiaceae subfamily Apioideae

The varied climatic, topographic and edaphic features of Iran have resulted in a diverse flora (Zohary, 1973; Léonard, 1991–1992). Iran is a large country, with a total surface area of 1,648,000 km2. Ranges in elevation extend from 26 m below sea level, on the shores of the Caspian Sea, to 5,670 m at the summit of Mt. Damavand (Firuz, 1974). Annual precipitation varies from ca. 2,000 mm in Bandar Anzali (on the coast of the Caspian Sea in northern Iran) to less than 100 mm in the central and southwest Iranian deserts. The temperature ranges from a winter low of –35°C in the northwest to a summer high of +50°C on the Persian Gulf. Phytogeographically, Iran is divided into three regions: (1) the Irano-Turanian floristic region that covers most of the country except its most northern and southern regions; (2) the Euro-Siberian region that covers the area from the northern portion of the Alborz Mountains to the Caspian Sea; and (3) the tropical SaharoSindian region that includes a narrow strip in southern Iran (Zohary, 1963; Hedge & Wendelbo, 1978). Iran is rich in endemic species, with many of these occurring in its vast mountainous regions. Of the approximately 8,000 species of flowering plants in Iran, 1,724 (21.6%) are endemic (Rechinger, 1963–2001; Mozaffarian, 1994, 2003; Akhani, 2002, 2003, 2004). For the family Apiaceae, Iran is a major center of diversification. The country possesses one of the richest diversities of Apiaceae in the world, exceeded only by China and Turkey (Valiejo-Roman & al., 2006). A total of 363 species Phylogenetic analysis of nrDNA ITS sequences reveals relationships within five groups of Iranian Apiaceae subfamily Apioideae


INTRODUCTION
The varied climatic, topographic and edaphic features of Iran have resulted in a diverse flora (Zohary, 1973;Léonard, 1991Léonard, -1992)).Iran is a large country, with a total surface area of 1,648,000 km².Ranges in elevation extend from 26 m below sea level, on the shores of the Caspian Sea, to 5,670 m at the summit of Mt.Damavand (Firuz, 1974).Annual precipitation varies from ca. 2,000 mm in Bandar Anzali (on the coast of the Caspian Sea in northern Iran) to less than 100 mm in the central and southwest Iranian deserts.The temperature ranges from a winter low of -35°C in the northwest to a summer high of +50°C on the Persian Gulf.Phytogeographically, Iran is divided into three regions: (1) the Irano-Turanian floristic region that covers most of the country except its most northern and southern regions; (2) the Euro-Siberian region that covers the area from the northern portion of the Alborz Mountains to the Caspian Sea; and (3) the tropical Saharo-Sindian region that includes a narrow strip in southern Iran (Zohary, 1963;Hedge & Wendelbo, 1978).Iran is rich in endemic species, with many of these occurring in its vast mountainous regions.Of the approximately 8,000 species of flowering plants in Iran,1,724 (21.6%) are endemic (Rechinger, 1963(Rechinger, -2001;;Mozaffarian, 1994Mozaffarian, , 2003;;Akhani, 2002Akhani, , 2003Akhani, , 2004)).
For the family Apiaceae, Iran is a major center of diversification.The country possesses one of the richest diversities of Apiaceae in the world, exceeded only by China and Turkey (Valiejo-Roman & al., 2006).A total of 363 species and 114 genera of Apiaceae are known from Iran, of which 114 species and 12 genera are endemic (Mozaffarian, 1996;Pimenov & Leonov, 2004;Valiejo-Roman & al., 2006).In the Flora Iranica area (which includes Iran, Afghanistan, and the adjacent mountain regions of Iraq, Turkmenistan, and western Pakistan), about 140 genera of Apiaceae are represented.In this region, the family is poorly known; therefore, many of the accounts provided in Flora Iranica are provisional (Hedge & Lamond, 1987).The region contains members of several large and taxonomically complex genera, such as Ferula L., Prangos Lindl.and Heracleum L., and many small endemic genera whose phylogenetic affinities are obscure, such as Alococarpum Riedl & Kuber, Haussknechtia Boiss., Azilia Hedge & Lamond, and Johrenia DC.It is not clear, however, whether this high number of small, endemic genera accurately reflects phylogeny or is a result of taxonomic splitting.In general, the taxonomic confusion surrounding many genera of Apiaceae with similar morphologies has resulted in the creation of many monotypic genera when distinctive morphological characteristics are observed (Spalik & al., 2001).Within Iran, some of these endemic species have very restricted habitat requirements or narrow geographic distributions.As examples, Haussknechtia elymaitica Boiss.and Azilia eryngioides (Pau) Hedge & Lamond occur only in a few specialized niches in western Iran and Johrenia golestanica Rech.f. is restricted to Golestan National Park in northeastern Iran.The major objective of this paper is to ascertain the taxonomic and phylogenetic relationships within five groups of genera native to the Flora Iranica area.Historically, these taxa were treated in Apiaceae tribes Apieae, Peucedaneae, and Smyrnieae (sensu Drude, 1898), but the results of molecular phylogenetic studies have revealed that each of these tribes is highly polyphyletic (Katz- Downie & al., 1999;Downie & al., 2000c).Thus, the phylogenetic placements of many of these Iranian taxa, several of which are narrow endemics or unusual morphologically, are also unknown or controversial.The precise composition of each of these groups and the monophyly of their constituent genera will be assessed through the results of phylogenetic analyses of nuclear ribosomal DNA (nrDNA) internal transcribed spacer (ITS) sequences and, for some of the included taxa, morphological comparisons.For one of these groups, ancillary data will come from petiole anatomical and palynological studies.Additionally, the phylogenetic placements of these genus groups in Apiaceae subfamily Apioideae are inferred.These five groups represent some of the most outstanding problems of umbellifer taxonomy in Iran.We refer to these five groups of genera as the Cachrys group, Cymbocarpum group, Ferula group, Johrenia group, and Opopanax group.In this section, we provide a provisional circumscription of each of these groups based on histori-cal treatments of these taxa and/or the results of previous molecular systematic investigations and discuss additional specific objectives.
Cachrys group.-In Flora Iranica, Hedge & Lamond (1987) recognized two morphologically similar species in Diplotaenia Boiss.(D. cachrydifolia Boiss., D. damavandica Mozaffarian, Hedge & Lamond).They believed that Diplotaenia allied closely with Ferula and Prangos by the common possession of lateral (secondary) umbels with male and hermaphrodite flowers and central (primary) umbels with hermaphrodite flowers.Ferula and Prangos (and Ferulago W.D.J. Koch) have a common inflorescence structure, with Prangos distinguished from the other two by fruit shape.Azilia is a morphologically unique, monotypic genus endemic to Iran and its simple to once-pinnate leaves with thick, broad segments and spiny margins are reminiscent of some Eryngium L. species.Previously, A. eryngioides was treated in Prangos (as P. eryngioides Pau), but according to Hedge & Lamond (1987) this species "has, of course, no similarity to or connection with any species of Prangos."Instead, its strongly dorsally compressed fruit with thickened or winged margins and flat albumen at the commissural surface suggest its placement near such Iranian genera as Diplotaenia, Dorema D. Don, Ferula, Ferulago, and Peucedanum L. (Hedge & Lamond, 1987).Rechinger (1990), however, in revising Pau's types, concluded that Azilia is distantly related to Ferula and Peucedanum.On the basis of previous molecular evidence, Prangos pabularia Lindl.and Ferulago galbanifera (Mill.)W.D.J. Koch comprised a clade sister group to Azilia eryngioides (Downie & al., 2000c).This group was poorly supported (56% bootstrap value) and, again, distantly related to Ferula and Peucedanum.Immunological studies support a close relationship between Prangos and Ferulago (and also Bilacunaria Pimenov & V.N.Tikhom.), but not between these genera and Ferula (Shneyer & al., 1995).Therefore, we provisionally recognize the Cachrys group as comprising the genera Azilia (monotypic), Diplotaenia (2 spp.), Prangos (38 spp.), Ferulago (45 spp.), and possibly Bilacunaria (4 spp.).(Numbers in parentheses here and afterwards in this section refer to numbers of species in each genus recognized by Pimenov & Leonov, 1993.)In this study, we ascertain the phylogenetic relationships among these genera.In particular, we assess the phylogenetic position of Azilia relative to the other members of this group.We refer to this assemblage as the Cachrys group because the results of our study place Cachrys libanotis L. within this clade.Cachrys L. is the oldest (Linnaean) generic name among those taxa included in the clade and C. libanotis is its nomenclatural type.As additional species are added in subsequent study, the name Cachrys will be retained whereas other potential names may be subject to change and eventual synonymization.Iran. In Flora Iranica, Alava (1987) transferred C. marginatum into the monotypic genus Kalakia Alava (as K. marginata (Boiss.)Alava) based on its thickened mericarp margins, and Rechinger (1987) recognized only two species within Cymbocarpum (C.anethoides, C. erythraeum).Previously, C. marginatum was treated in Ducrosia Boiss.(as D. stenocarpa Bornm.& Gauba), suggesting an affinity to that genus as well.Therefore, we provisionally circumscribe the Cymbocarpum group to include the genera Cymbocarpum (3 spp.), Ducrosia (3 spp.), and Kalakia (monotypic).In this study, we confirm the monophyly of Cymbocarpum and ascertain its relationship to Ducrosia and Kalakia.
Ferula group.-Plants of Dorema are monocarpic, large, possess thickened storage roots, and have large simple umbels with male flowers on the lower branches and hermaphrodite flowers on the upper branches.Twelve to sixteen species are recognized, and these are distributed primarily in the Caucasus, southern parts of central Asia, Iran, Afghanistan, and Baluchistan (Schisch kin, 1951;Pimenov & Leonov, 1993) (Rechinger, 1987;Mozaffarian, 2003).Dorema kopetdaghense Pimenov was treated as a synonym of D. hyrcanum in Flora Iranica (Rechinger, 1987).Schisch kin (1951) and Pimenov (1988) reported that Dorema closely resembles Ferula and immunological comparisons revealed that Dorema is serologically similar to Ferula and Leutea Pimenov (Shneyer & al., 1995).In contrast, and rather surprisingly, a recent study of ITS sequences placed Dorema (represented by D. aucheri ) sister group to Seseli mucronatum (Schrenk) Pimenov & Sdobnina in tribe Selineae, well away from Ferula (Valiejo-Roman & al., 2006).Haussknechtia is a monotypic genus endemic to southwest Iran.These plants are highly unusual in the family because of their compressed, globose, compound umbels.Koso-Poljansky (1925, as cited in Pimenov, 1988) suggested an affinity between Haussknechtia and Dorema ; Rechinger (1987) did as well, but emphasized their differences in fruit shape and inflorescence type.Recently, however, analyses of ITS sequence data showed that Haussknechtia is a sister group to the Iranian endemic monotypic genus Demavendia Pimenov in tribe Pimpinelleae (Valiejo-Roman & al., 2006).We provisionally recognize the Ferula group as comprising the genera Dorema (12 spp.), Ferula (170 spp.), and Leutea (6 spp.).In this study, we confirm the mono-phyly of Dorema and assess its relationship to Ferula and Leutea.

MATERIALS AND METHODS
Morphology. -Observations and measurements were taken from herbarium specimens and plants in the field.An examination of living plants was necessary because some species are large and poorly represented by herbarium specimens.Dorema, for example, may be up to 3 m in height and have a stem diameter of 10 cm or more.The morphological characters examined include those traditionally important in circumscribing genera and species, such as habit, stems (height, color, indumentums or surface sculpturing and furrowing), branching patterns, leaf segments (size, shape, margin, and surface features such as indumentum), bracts, bracteoles, rays, inflorescences, pedicels, petals (length and color), stylopodia, and mericarps.All relevant material from herbaria TARI, IRAN, TUH and the private herbarium collection of Dr. H. Akhani (Department of Biology, Faculty of Science, University of Tehran, Iran) was examined.
Petiole anatomy.-Petiole anatomy was examined in Johrenia paucijuga, J. golestanica, and J. aromatica.Material collected from the field was fixed in FAA.Freehand sections of petioles were stained with carmine (to identify the phloem) and diluted methyl green (xylem), mounted permanently on microscope slides using Canada balsam (Chamberlain, 1930), and preserved as vouchers in the private herbarium of Dr. H. Akhani.Photographs were taken using an Olympus BX51 compound microscope and an Olympus SZX12 stereomicroscope using an Olympus DP12 digital camera.Petiole characters included crosssectional outline shape, number of peripheral vascular bundles, and features of sclerenchyma around the vascular bundles (Pimenov & al., 1986).Herbarium voucher specimens are deposited in the aforementioned herbaria.Petiole data for Johreniopsis seseloides (C.A. Mey.) Pimenov and Holandrea caucasica (M.Bieb.)Spalik, Reduron & S.R. Downie were obtained from literature (Salimian, 2002).
Pollen morphology.-Pollen was examined from anthers collected directly in the field or from herbarium specimens from three species of Johrenia.Pollen grains were acetolized (Erdtman, 1969) and viewed using both an Olympus BX51 compound microscope and a Zeiss SEM-960A scanning electron microscope.Characters observed included pollen shape, length and width, exine thickness at poles and equator, and exine surface features (Punt, 1984).For each character measured, the average measurement of 100 pollen grains was determined.Palynological data for Johreniopsis seseloides and Holandrea caucasica were obtained from literature (Salimian, 2002).
DNA sequencing and species selection.-In total, 177 accessions (157 species) of Apiaceae subfamily Apioideae were examined for nrDNA ITS sequence variation (Appendix).Ninety-two of these accessions were newly sequenced; data for the remaining 85 accessions were obtained from GenBank.The latter were chosen because preliminary phylogenetic analysis of all available ITS sequences of subfamily Apioideae in GenBank and numerous unpublished sequences (representing about 1,000 species from 250 genera) revealed that they were closely related to the five groups of Iranian genera considered herein (Downie & al., unpubl.).To date, ITS sequences comprise the most comprehensive database for Apioideae.These 1,000 accessions represent species across a broad geographic distribution and nearly all tribes and major clades of subfamily Apioideae defined on the basis of previous molecular systematic studies except those most basal within the subfamily (i.e., tribe Heteromorpheae, Annesorhiza clade, Lichtensteinia clade, and several putatively basal species of uncertain position; Downie & al., 2001;Calviño & al., 2006) and those lineages within tribes Oenantheae and Scandiceae which demonstrate rapid ITS sequence evolution (Downie & al., 1998;Petersen & al., 2002;Hardway & al., 2004).Approximately two-thirds of the 157 species included in this investigation are native to Iran (Rechinger, 1987); many other species are distributed in the greater Flora Iranica area.
Protocols for total genomic DNA extraction, PCRamplification using standard ITS primers, and DNA purification and sequencing strategies are presented elsewhere (Downie & Katz-Downie, 1996;Downie & al., 2000a).Cycle sequencing reactions were carried out using the purified PCR product, AmpliTaq DNA polymerase (Roche Molecular Systems, Alameda, California, U.S.A.) and florescent Big Dye terminators (Applied Biosystems, Foster City, California, U.S.A.).The sequencing products were resolved using an ABI 3730XL high-throughput DNA capillary sequencer.Simultaneous consideration of both DNA strands across the entire ITS region permitted unambiguous base determination.
Alignment and phylogenetic analyses.-The DNA sequences were aligned initially using the computer program CLUSTAL X (Jeanmougin & al., 1998), with default parameters for gap penalty and extension, and then adjusted manually as necessary.Gaps were positioned to minimize nucleotide mismatches.The trees were rooted with members of tribe Pleurospermeae, based on results of previous higher-level analyses of the subfamily (Downie & al., 2001).For 50 accessions obtained from GenBank, sequence data for the 5.8S rDNA region were not available.These missing data represented 6.7% of the entire matrix.Uncorrected pairwise nucleotide distances of unambiguously aligned positions were determined using the distance matrix option of PAUP* vers.4.0b10 (Swofford, 2002).Maximum parsimony (MP) analysis of the complete data matrix was implemented using PAUP* and the search strategies described elsewhere (Downie & al., 2000a).The maximum tree limit of 20,000 minimal length trees was obtained.Bootstrap (BS) values were calculated from one million replicate analyses using "fast" stepwiseaddition of taxa and only those values compatible with a majority-rule consensus tree were recorded.The number of additional steps required to force particular taxa into a monophyletic group was examined using the constraint option of PAUP*.The data matrix was also analyzed using maximum likelihood (ML), after using the program Modeltest vers.3.7 to choose an appropriate model of nucleotide substitution that best fits these data (Posada & Crandall, 1998), as selected by the Akaike Information Criterion (AIC) estimator (Posada & Buckley, 2004).The parameters appropriate for the chosen model were input into PAUP* and a heuristic search performed using random sequence addition and tree-bisection-reconnection branch swapping under ML optimization.One hundred BS replicate analyses were conducted using neighborjoining searches with ML distance estimates, using the parameters inferred by Modeltest.Lastly, Bayesian analysis was conducted using the program MrBayes vers.3.1.2(Ronquist & Huelsenbeck, 2003).The settings appropriate for the best-fit model of nucleotide substitution, as selected by MrModeltest vers.2.2 (Nylander, 2004) under the AIC estimator, were put into a MrBayes block in PAUP* (nst = 6; rates = invgamma).The priors on state frequencies and rates and variations across sites were estimated automatically by the program.Four Markov chains starting with a random tree were run for one million generations, sampling trees at every 100th generation.Trees from the first 100,000 generations were discarded as "burn-in" before stationarity was reached, prior to calculating the majority-rule consensus tree from the remaining trees.Posterior probability (PP) values for all internal tree branches were recorded.

RESULTS
DNA sequencing and phylogenetic resolutions.
-The ITS region ranged in size from 577 bp (one accession of Ainsworthia trachycarpa Boiss.) to 610 bp (Holandrea achaica (Halácsy) Spalik, Reduron & S.R. Downie).Species with identical DNA sequences were each treated as one terminal in the phylogenetic analysis.Alignment of all ITS sequences from 177 accessions (170 terminals) resulted in a matrix of 654 positions, of which 44 were excluded from subsequent analysis because of alignment ambiguities.The remaining 610 alignment positions yielded 200 constant, 342 parsimony informative, and 68 autapomorphic characters.Alignment gaps ranged in size between 1 and 20 bp; the vast majority of these were only a single bp in size.Maximum pairwise sequence divergence values ranged from identity to 33% (the latter between Lagoecia L. and the outgroup Physospermum Cusson).
MP analysis of all unambiguously aligned positions resulted in the preset limit of 20,000 trees, each of 2,614 steps (consistency index, CI = 0.3083 and 0.2868, with and without uninformative characters, respectively; retention index, RI = 0.7642).The strict consensus of these trees, with accompanying BS values, is presented in Fig. 1.Superimposed on Fig. 1 are ten tribes recognized on the basis of previous molecular phylogenetic studies (Downie & al., 2001;Spalik & al., 2004), plus the Apioid superclade, a large and variable group of umbellifers containing tribes and other major clades of uncertain relationship (Plunkett & Downie, 1999).Also indicated are the five groups of Iranian genera circumscribed herein, one of which is now redefined on the basis of these molecular results.Based on the AIC estimator, Modeltest selected the GTR + I + G model of nucleotide substitution as best fitting these ITS data.Using the parameters inferred by Modeltest, a single ML tree was recovered having a -ln likelihood score of 14,263.187.This tree is presented in Fig. 2 along with ML distance-based BS values and the ten tribes and clades indicated previously.Relationships inferred by the Bayesian analysis were highly consistent with those presented in the MP strict consensus tree.PP values are presented on the MP strict consensus tree (Fig. 1) for those branches common to both analyses.
The group of Diplotaenia (2 spp.), Ferulago (3 spp.), Prangos (6 accessions represented by 5 spp.), and the monotypic Azilia (2 accessions) is redefined to include Bilacunaria microcarpa (M.Bieb.)Pimenov & V.N.Tikhom., Cachrys (2 spp.), Alococarpum erianthum (DC.)Riedl & Kuber, and Eriocycla olivieri (Boiss.)Wolff, and is referred to as the Cachrys group.(Numbers in parentheses here and afterwards in this section refer to numbers of species in each genus included in the phylogenetic study.)This group is variously supported, with MP and ML BS values less than 50% and a PP value of 1.00.Similarly, intergeneric relationships within this clade are generally supported poorly.The genera Prangos (upon the possible inclusion of Alococarpum erianthum, see Discussion), Cachrys, Diplotaenia, and Ferulago are each supported as monophyletic.The clade of Prangos and Alococarpum is variously supported as monophyletic (MP BS 89%; ML BS 56%; PP 1.00).Prangos acaulis (DC.)Bornm.and P. uloptera DC. accession no.3147 have identical ITS sequences and the two included accessions of P. uloptera do not unite as monophyletic (with 3.1% nucleotide sequence variation between them).Azilia eryngioides is sister group to all aforementioned taxa of the Cachrys group.
The Ferula group is maintained to include Dorema, Ferula, and Leutea.The genus Dorema (7 spp.) occurs within a paraphyletic Ferula (9 spp.) in tribe Scandiceae in both MP and ML trees.In the Bayesian tree, these taxa form a 4-branched polytomy with the clade of Cuminum L., Daucus L., Laser Borkh.ex P. Gaertn., B. Mey.& Schreb., and Polylophium Boiss.(all members of Scandiceae subtribe Daucinae).In all trees, Dorema comprises a subclade with two species of Ferula : F. feruloides (Steud.)Korovin and F. tenuisecta Korovin ex Pavlov.Also arising from within Ferula but near the base of the Ferula group is Leutea petiolaris (DC.)Pimenov, the nomenclatural type of Leutea.While the Ferula group is resolved in both MP and ML trees, it is only supported weakly.However, the Ferula group is distinct from Scandiceae subtribe Daucinae when differences in branch lengths are considered, with those leading to subtribe Daucinae being the longest and best supported in Scandiceae and those of the Ferula group being relatively short (Fig. 2).Dorema is monophyletic in a subset of the 20,000 maximally parsimonious trees.Interspecific sequence divergence values in Dorema ranged from identity (D. glabrum, D. aucheri, D. ammoniacum, and Ferula feruloides all had identical sequences, as did D. aitchisonii and D. kopetdaghense) to 1.2% (between D. aureum and D. hyrcanum).Nucleotide sequence divergence between D. kopetdaghense and D. hyrcanum was 0.9%.The monotypic genus Haussknechtia occurs elsewhere, sister group to Demavendia pastinacifolia (Boiss.& Hausskn.ex Boiss.)Pimenov in tribe Pimpinelleae, and is confirmed not to be a member of the Ferula group.
The species of Johrenia (3 spp.) comprise a clade (MP BS 98%; 58% ML BS; PP 1.00) with Johreniopsis (2 spp.) and Holandrea (5 spp.) in the previously delimited Johrenia group.None of these genera is monophyletic.Constraining Johrenia to monophyly results in trees four steps longer than those without the constraint invoked.The narrow endemic species Johrenia golestanica is sister group to all other members of the Johrenia group.Upon the exclusion of J. golestanica, Holandrea is paraphyletic and all other examined species of Johrenia and Johreniopsis arise from within it.The clade of Johrenia, Johreniopsis, and Holandrea is isolated from Peucedanum officinale L., the nomenclatural type of Peucedanum, and several generic segregates of Peucedanum s.l.(i.e., Cervaria Wolf, Demavendia, Imperatoria L., Leutea, and Thysselinum Raf.).
Several additional relationships and phylogenetic placements deserve comment, even though many of these taxa are not components of the Iran flora.These genera are distributed widely in Europe and Asia, while some of them also occur in North America and Africa.The genera Ferulopsis Kitag., Trinia Hoffm., and Vicatia DC. are new members of Apiaceae tribe Selineae (sensu Spalik & al., 2004).The large genus Seseli L. is not monophyletic, as reported previously (Downie & al., 1998;Spalik & al., 2004).Expanded sampling within tribe Tordylieae (i.e., the Heracleum clade of Downie & al., 2001) 1.These species are very similar in leaf morphology, but differ in their fruits.The transfer of C. marginatum to the monotypic genus Kalakia (Alava, 1987) was based on its characteristic thickened fruit margins, yet there are also pronounced differences between the fruits of C. anethoides and C. erythraeum.In the former, the mericarps are compressed laterally, whereas in the latter they are compressed dorsally (as they are in K. marginata).Cymbocarpum and Kalakia show many similarities in other vegetative features (i.e., habit, stem color, stem furrows, outer petal length, and bract division).Ducrosia anethifolia is similar to Cymbocarpum and Kalakia in its petals and shape of leaf segments.However, D. anethifolia differs from the other two genera in that its mericarps are subglobular and its leaves do not wither early.In general, distinguishing among Cymbocarpum, Kalakia and Ducrosia using leaf characters alone is very difficult.Similarly, fruit characters, such as the orientation of fruit compression and type of fruit margin, are also not useful in separating these genera.
A comparison of selected morphological features for seven Iranian representatives from four genera provision- A comparison of selected morphological characters for three species of Johrenia and one species each of Johreniopsis and Holandrea is presented in Table 3, with the primary purpose of assessing the relationship of the narrow endemic Johrenia golestanica to its congeners.Johrenia golestanica is unique in its white petals, grayish-green and finely striate stems, sub-elliptic (to sub-orbicular) fruits with non-prominent dorsal ribs, and a preference for growing in disturbed areas.Johreniopsis is morphologically similar to Johrenia.Furthermore, with its yellow petals, once-to twice-pinnate leaves, and prominent mericarp ribs, Johreniopsis is morphologically more similar to Johrenia paucijuga and J. aromatica than it is to J. golestanica.Comparisons of petiole anatomy   4; Fig. 3) and pollen grains (Table 5; Fig. 3) also support the separation of J. golestanica from its congeners.Unlike the other two species of Johrenia where the vascular bundles are partially or wholly surrounded by sclerenchyma cells, the vascular bundles of J. golestanica are without sclerenchyma tissue.Sclerenchyma at the base of the xylem, however, is present in Johreniopsis seseloides and Holandrea caucasica.Pollen of Johrenia, Johreniopsis, and Holandrea is of the oval-type (sensu Cerceau-Larrival, 1971) and in J. golestanica the grains are slightly smaller (in length and width) than they are in its congeners.The morphology of eight species of Dorema from Iran was examined (D. kopetdaghense was considered  a distinct species).All species are morphologically very similar, although each can be distinguished by several characters (results not shown).As examples, Dorema ammoniacum and D. aitchisonii are similar in their leaf segments (shape, pubescence, margin, width), peduncle and pedicel length and surface features, and having woolly fruits prior to ripening.They both differ from other species in the genus, however, in several of these attributes.Dorema ammoniacum and D. aitchisonii are separated by mericarp length, wing width, and degree of stem internode swelling (the stem is swollen in D. aitchisonii but not in D. ammoniacum).The size of the leaves in these two species shows great variation.Dorema aucheri has similar leaf characters to the two aforementioned species, but is separated from them by fruit size and surface texture.Dorema kopetdaghense and D. hyrcanum are morphologically distinct; the former differs in its smaller leaf segments that are lanceolate with entire margins, its pubescent immature fruits, and shorter peduncles.Dorema glabrum is also distinct morphologically, with its glabrous leaf segments and long pedicels.Similarities between Dorema and Ferula species are apparent and have been presented elsewhere (e.g., Schischkin, 1951;Pimenov, 1988).
A comparison of morphological features of Opopanax hispidus, O. persicus, and Smyrniopsis aucheri of the Opopanax group is presented in Table 6.These species share several morphological attributes, such as those of the leaves and inflorescences.These features include once-or twice-pinnate basal leaves, leaf segments oblong or elliptic, acute, crenate and with cartilage at their margins, hispidus leaves, type of cork on the stem, and inflorescences of male lateral umbels and central hermaphrodite umbels.There are some obvious differences between Opopanax and Smyrniopsis, such as dorsally versus laterally compressed mericarps and solid versus hollow petioles in cross-section.The stems of O. persicus are smooth, whereas those of O. hispidus are densely hispid (especially on the basal portions), and those of S. aucheri are sparsely hispidus.Inflorescence shape in O. persicus is dense thyrsoid, in O. hispidus it is lax thyrsoid, and in S. aucheri it is conical.

DISCUSSION
The limitations of nrDNA ITS sequence data in resolving tribal-level relationships within Apiaceae subfamily Apioideae have been discussed elsewhere (Downie & al., 1998;Katz-Downie & al., 1999).Their homoplastic nature, high levels of nucleotide sequence divergence in some lineages, and small size of the region all conspire   Downie & al., 2001).To date, phylogenies of the Apioideae inferred from ITS sequence data are generally consistent with those inferred from chloroplast markers, but until the latter are available for a broad sampling of taxa from the Apioid superclade the relationships among its tribes and the tribal placements of the Cachrys and Opopanax groups cannot be resolved.The Cachrys group may constitute a new, major lineage of Apiaceae, but its recognition now as a tribe is premature in the absence of supporting plastid data.The major objective of this study was to ascertain taxonomic and phylogenetic relationships within these five genus groups.Sister group relationships of each of these five groups are not discussed in detail because of the poor resolution of relationships, limited sampling outside of these analyzed clades, and the large size and taxonomic complexity of the subfamily Apioideae.Below, we discuss the taxonomic implications of our study for each of these five genus groups.Cachrys group.-The Cachrys group, as presently circumscribed, comprises a heterogeneous assemblage of genera and we are not yet aware of any obvious morphological character that would support its monophyly.Moreover, the clade is not very well supported in the MP and ML trees, nor are the relationships resolved among its constituent genera.The genera Prangos and Ferulago are large and, together with Cachrys and its segregates, extremely complex taxonomically.The species of Bilacunaria ( = Hippomarathrum Link) was treated previously in Cachrys (as C. microcarpa M. Bieb.) and this relationship is not necessarily refuted by the ITS results because B. microcarpa is a weakly supported sister group to Cachrys in the ML tree.Azilia eryngioides is a sister group to all other members of this clade, coincident with its highly unusual morphology.We maintain the monotypic Azilia as a distinct genus given its isolated position in all ITS-derived cladograms and its unique morphology.The monotypic genus Alococarpum falls within the genus Prangos and in the MP and Bayesian trees this clade is very well supported.Alococarpum erianthum was treated previously in Cachrys (as C. eriantha DC.), but its placement in Prangos seems better justified on the basis of their similar morphologies (Leute, 1987).However, before A. erianthum is transferred into Prangos, supporting evidence from cpDNA is required.
Cymbocarpum group.-Molecular and morphological studies support the close relationship among Cymbocarpum, Kalakia, and Ducrosia.Cymbocarpum is monophyletic and a strongly supported sister group to Kalakia in the ITS trees.These two genera are morphologically very similar and the differences noted in fruit morphology between them are just as great as those between C. anethoides and C. erythraeum.ITS sequence divergence between Cymbocarpum and Kalakia marginata is low relative to comparisons between Cymbocarpum/ Kalakia and Ducrosia.The erection of the monotypic genus Kalakia solely on the basis of its thickened fruit margin is dubious, as similar tumid margins also occur in Ducrosia and elsewhere in Apioideae.Given the overall similarity between Cymbocarpum and Kalakia and their relatively low genetic divergence, we do not support Kalakia as a distinct, monotypic genus.Instead, we maintain three species in the Iranian genus Cymbocarpum.Three species also comprise Ducrosia, but only D. anethifolia was included in this study.Its other two species must be examined to confirm their taxonomic placements alongside D. anethifolia.
Ferula group.-The recognition of Dorema kopetdaghense as a synonym of D. hyrcanum, as proposed by Rechinger (1987) in Flora Iranica, is not supported by our results.Sequence divergence between these two species is among the highest in infrageneric pairwise comparisons and we consider them morphologically distinct at the species level.Therefore, Dorema kopetdaghense and D. hyrcanum should be retained as separate species.Molecular data support the monophyly of Iranian Dorema (upon the inclusion of F. feruloides and F. tenuisecta) and its close relationship to Ferula in tribe Scandiceae.Indeed, three species of Dorema and Ferula feruloides have identical ITS sequences.Ferula feruloides is treated in Ferula subgenus Dorematoides (Rgl.& Schmalh.)Korovin, an apparently polyphyletic group, but whose plants bear simple umbels like those of Dorema (Schischkin, 1951).The close relationship between Dorema and Ferula was suggested previously based on fruit anatomical, morphological, and immunological data (Schischkin, 1951;Pimenov, 1988;Shneyer & al., 1995).The placement of D. aucheri as a sister group to Seseli mucronatum in tribe Selineae, as indicated by Valiejo-Roman & al. (2006) on the basis of ITS sequence comparisons, must be regarded as spurious.In both MP and ML trees, Dorema arises from within a paraphyletic Ferula.This same relationship is supported by Kurzyna-Mlynik & al. (in press) upon further study of ITS sequences and a comprehensive sampling of Ferula.Therefore, we suggest that the eight species of Dorema from Iran be transferred into Ferula, but before such nomenclatural changes are implemented supporting evidence from cpDNA is required.Similarly, the genus Leutea may also be submerged within Ferula.Six species are recognized in the genus and all are almost entirely restricted to the Flora Iranica area.Leutea is a segregate of Peucedanum and its affinity to several other peucedanoid genera has been inferred, such as Johrenia.
Our results indicate that Leutea is allied with Ferula, as suggested previously (Pimenov, 1987b;Shneyer & al., 1995).However, we refrain from making this transfer now because only one species of Leutea was included in our study and its placement within the Ferula subclade is unclear.
The molecular results support the placement of the monotypic genus Haussknechtia adjacent to Demavendia in tribe Pimpinelleae, well away from Dorema and Opopanax for which affinities were suggested previously (Rechinger, 1987;Pimenov, 1988;Valiejo-Roman & al., 2006).Haussknechtia and Demavendia are morphologically similar and very distinct from all examined species of Dorema.Recently, Valiejo-Roman & al. (2006), on the basis of ITS sequence comparisons, showed a close relationship among Haussknechtia, Demavendia, and Zeravschania.Such a relationship has also been supported by fruit anatomy (Pimenov & al., 2006).
Johrenia group.-Molecular, morphological, petiole anatomical, and palynological data support a distant relationship between Johrenia golestanica and its congeners.Thus, J. golestanica may constitute a new, monotypic genus pending confirmation from additional sampling of Johrenia and Johreniopsis and supporting evidence from cpDNA.These taxa are distantly related to Peucedanum s.str.(P.officinale L.), but are very closely related to Holandrea, whose species were previously attributable to Peucedanum s.l.(Reduron & al., 1997;Spalik & al., 2004).In this study, Johrenia, Johreniopsis, and Holandrea are each not monophyletic, yet the assemblage comprises a strongly supported monophyletic group in the MP and Bayesian trees.These genera are placed in tribe Selineae along with Peucedanum and its segregates (Cervaria, Imperatoria, Thysselinum) and are distantly related to Leutea (tribe Scandiceae) and Demavendia (tribe Pimpinelleae).Johrenia and Johreniopsis constitute new members of tribe Selineae.The association of Holandrea with Johrenia and Johreniopsis is not surprising given their similar morphologies, thus future studies of this group must include Holandrea, as well.Pending further investigation, Holandrea may be transferred into Johrenia, as this genus has priority over Johreniopsis.However, the results of multivariate analyses of 34 morphological characters for 33 species of Johrenia, Johreniopsis, Holandrea, and other genera revealed that Holandrea should be treated as a section of Johreniopsis (Pimenov, Kljuykov & Ostroumova, unpubl.).Continued investigations of this group are necessary.
Opopanax group.-Historically, Opopanax and Smyrniopsis were placed in different tribes (Drude, 1898), yet molecular data revealed them to be a monophyletic group (e.g., Katz-Downie & al., 1999).Their monophyly is supported upon additional sampling.Similarly, these taxa share similar morphological attributes, especially with regard to their leaf segments and inflorescences.These genera are allied with Petroedmondia syriaca which was previously referable to Smyrniopsis (S. cachroides).Boissier (1872) recognized Smyrniopsis aucheri and S. cachroides in two sections; however, their recognition as distinct genera is justified based on the molecular phylogenies presented herein and the high sequence divergence values between them.In addition, there is a possible alliance of these taxa with the western European species Magydaris panacifolia and the taxonomically isolated, monotypic Turkish genus Crenosciadium Boiss.& Heldr.ex Boiss.The latter was considered by Bentham (1867) as closely allied to and possibly congeneric with Opopanax and these genera are treated adjacently in the Flora of Turkey (Davis, 1972).Further studies of the Opopanax group would benefit by the inclusion of Petroedmondia, Magydaris, and Crenosciadium.
Other relationships.-Tribe Tordylieae is one of only a few traditionally recognized higher-level taxa of Apiaceae subfamily Apioideae that has been confirmed as monophyletic on the basis of phylogenetic analysis of molecular data.Its largest genus, Heracleum, with 65 species (Pimenov & Leonov, 1993), is not monophyletic.The type of the genus, H. sphondylium, and its allies form the core Heracleum clade, albeit this group is not very well supported.Ten species of Heracleum occur Iran, of which we have included only one.The Chinese species Heracleum candicans may represent a new genus.Heracleum pedatum has been treated previously as Vanasushava pedata (Wight) P.K. Mukh.& Constance and is sister group to Tetrataenium.The next largest genus in the tribe, Pastinaca, appears paraphyletic, with Malabaila ( = Leio tulus) nested within; this well supported clade is referred to as Pastinaca s.l., or the Pastinaca group (Valiejo-Roman & al., 2006), and requires further study.
In conclusion, the phylogenetic relationships within five groups of genera native to the Flora Iranica region have been assessed using molecular data, with supplementary comparative information coming from morphology and, for the Johrenia group, petiole anatomy and palynology.Given the rich diversity of Apiaceae in Iran, as well as its numerous endemic species and genera (many of which remain to be investigated using molecular data), further studies of these plants are necessary.Nomenclatural changes are suggested, but only if the relationships proposed using ITS data are corroborated by cpDNA evidence.To date, the treatment of Apiaceae in Flora Iranica is the best available for the region, yet many of its taxonomic accounts are still provisional (Hedge & Lamond, 1987).Our results should stimulate further detailed studies on these plants.While this manuscript was in preparation, a study appeared on the taxonomic relationships of Iranian Apiaceae using molecular data (Valiejo-Roman & al., 2006).This study included only a few exemplars of the Cymbocarpum and Johrenia groups and nothing from the Opopanax group; sampling of the Cachrys group was similar to our study.However, the one species of Dorema included in that investigation fell within tribe Selineae and not Scandiceae.Clearly, given the size and complexity of subfamily Apioideae in Iran, more work needs to be done to produce a modern classification of these plants.

Fig. 1 (
Fig. 1 (overleaf).Strict consensus of 20,000 minimal-length 2,614-step trees derived from maximum parsimony analysis of nrDNA ITS sequences from 177 accessions (170 terminals) of Apiaceae subfamily Apioideae (CIs = 0.3083 and 0.2868, with and without uninformative characters, respectively; RI = 0.7642).The tree inferred by Bayesian analysis of these data was highly consistent with the MP strict consensus tree.Numbers represent bootstrap values (%) and posterior probabilities for those branches common to both analyses.Bootstrap estimates were calculated from one million replicate analyses using "fast" stepwise-addition of taxa; only those values compatible with the majority-rule consensus tree are recorded.The ten tribes recognized on the basis of previous molecular phylogenetic studies are indicated, as are the Apioid superclade and the five groups of Iranian genera circumscribed herein.

Fig. 2 (
Fig. 2 (previous page).Single tree inferred from maximum likelihood analysis of nrDNA ITS sequences from 177 accessions (170 terminals) of Apiaceae subfamily Apioideae (-ln likelihood = 14,263.187).Branch lengths are proportional to the number of expected nucleotide substitutions per site (note scale bar) under a GTR + I + G model of nucleotide substitution.Numbers represent bootstrap estimates calculated from 100 replicate analyses using neighbor-joining searches and maximum likelihood distance.The ten tribes recognized on the basis of previous molecular phylogenetic studies are indicated, as are the Apioid superclade and the five groups of Iranian genera circumscribed herein.

Table 1 . Morphological comparisons within the Cymbocarpum group: Cymbocarpum anethoides, C. erythraeum, Kalakia marginata, and Ducrosia anethifolia.
Cachrys group(Diplotaenia, Prangos, Ferulago, Azilia)is presented in Table2.Sampling of this genus group was not comprehensive but is adequate to compare critical morphological differences between Azilia and its generic allies, as suggested by published studies.Diplotaenia cachrydifolia and D. damavandica are morphologically very similar, with noted differences between them limited to leaf segment shape and width.Inflorescences of both species are whorled, with lateral umbels bearing male and hermaphrodite flowers and central umbels bearing only hermaphrodite flowers.Diplotaenia shares with Prangos similar leaf and inflorescence characters.Indeed, D. cachrydifolia is very similar to P. ferulacea (L.) Lindl. in leaf morphology.Their fruits, however, are quite different.In Diplotaenia, the fruits are strongly dorsally compressed and have prominent dorsal ribs.In Prangos, the fruits are moderately compressed laterally and their prominent dorsal ribs are extended into wings.Ferulago stellata Boiss. is very similar to the other two genera, with its similar inflorescence structure and leaf form.In contrast, Azilia eryngioides stands apart from all other members of this group.Its leaf segments are orbicular to reniform and large with spiny margins and its inflorescences are paniculate with hermaphrodite umbels.Azilia is clearly distinguished morphologically from Diplotaenia, Prangos, and Ferulago (and indeed any other genus in the family), and any close ally based on morphological similarity is not apparent in our study.

Table 6 . Morphological comparisons within the Opopanax group: Opopanax hispidus, Opopanax persicus, and Smyrni- opsis aucheri.
reduce the utility of these data in resolving deep-level relationships within the subfamily.A glance at the trees in Figs. 1 and 2 will show poor resolution and branch support for the most basal lineages within the Apioid superclade.Minor differences in topology are also apparent among the trees with regard to the relationships inferred among the tribes and major clades.In this study, one of our goals was to elucidate the phylogenetic placements of five groups of genera native to the Flora Iranica area.Based on the phylogenetic results, the Cymbocarpum group falls within tribe Tordylieae, the Johrenia group is placed within tribe Selineae, and the Ferula group allies with tribe Scandiceae.Unclear, however, are the tribal placements of the Cachrys and Opopanax groups.Both of these fall within the Apioid superclade, but not in any previously recognized tribe supported strongly as monophyletic on the basis of molecular phylogenetic studies (reviewed in to