A Molecular Phylogenetic Study of Generic and Subgeneric Relationships in the Southwest Australian Endemics Conostylis and Blancoa ( Haemodoraceae )

We sequenced the plastid gene matK and the nuclear ribosomal spacer ITS for 39 of the 47+ species of Conostylis as well as its monotypic sister genus Blancoa, which some authors have included within Conostylis. Conostylis received 99% bootstrap support as monophyletic, with 100% support that Blancoa is its sister. Within Conostylis, the study provides strong support for two large sister clades, which we refer to as clades A (100%) and B (99%). Clade A consists of C. subgen. Conostylis plus the recently discovered C. glabra of C. subgen. Pendula sect. Divaricata (100%), and C. subgen. Pendula sect. Appendicula (100%). Clade B consists of species mostly placed within the remainder of C. subgen. Pendula but also contains members of the other small subgenera. Subgenus Pendula can be recircumscribed as monophyletic by excluding sect. Appendicula, Conostylis phathyrantha, and C. glabra and including subgen. Androstemma and subgen. Greenia. The status of the other two minor subgenera-C. subgen. Brachycaulon and C. subgen. Bicolorata-requires further investigation. Conostylis sect. Divaricata is polyphyletic. Ancient vicariance events are postulated for Conostylis involving separation of major clades in the northern and southern kwongan regions of southwestern Australia. The phylogenetic pattern in Conostylis is consistent across several lineages with the prolonged persistence of relictual taxa combined with explosive more recent speciation, the latter pronounced in the northern kwongan. There is evidence of significant divergence in major speciation mechanisms and chromosome number change among the three most species-rich subgenera/sections (dysploidy in Pendula and Appendicula vs. diploid speciation in Conostylis). Further investigation is needed to evaluate these ideas and elucidate the patterns of speciation in this most diverse genus of


INTRODUCTION
Haemodoraceae, as with many families of monocots, have had a tortuous taxonomic history, which prompted Ornduff (1979) to comment that "the substantial and continuing disagreements concerning the limits of the Haemodoraceae and the relationships of the genera assigned to it indicate that the family merits additional detailed study."Such studies were carried out subsequently.Today, in light of molecular and other data, Haemodoraceae are placed in Commelinales with Commelinaceae, Hanguanaceae, Philydraceae, and Pontederiaceae (Simpson 1990(Simpson , 1998;;Hopper et al. 1999;Chase et al. 2000;Angiosperm Phylogeny Group II [APG II] 2003).
No morphological synapomorphy is known for the family, and the common occurrence of arylphenalenones, once thought to be unique to Haemodoraceae (Cooke and Segal 1955;Cooke and Edwards 1981;Simpson 1990), is now also known from Musaceae, Pontederiaceae, and Strelitziaceae (Greca et al. 1992;Opitz and Schneider 2002;Opitz et a!. 2002;Otalvaro et a!. 2002).Nevertheless, there is 100% bootstrap support for the monophyly of Haemodoraceae based on trnL-F plastid DNA sequences (Hopper et a!. 1999).They are a small family with many highly divergent, relictual taxa, as well as a few genera demonstrating a more recent evolutionary radiation.
Two groups have been documented within Haemodoraceae from the earliest descriptions of the family.Whereas Takhtajan ( 1997) advocated recognition of these as separate families (i.e., Conostylidaceae and Haemodoraceae), most authors have advocated tribal status for them (reviewed by Simpson 1990).We found these to be clades (Hopper et al. 1999) and recognize them as subfamilies, following Dahlgren et al. (1985) and Macfarlane et al. (1987).
Haemodoraceae subfamily Haemodoroideae are widespread, with eight small genera and ca.30 species in South Africa, the Americas, and Australia-New Guinea.Haemodoroideae have three (rarely one) fertile stamens and mostly glabrous flowers, and their rootstocks usually have red pigments (hence the common name bloodroots).Conostylidoideae are endemic to the southwestern Australian global biodiversity hotspot (Myers et al. 2000;Hopper and Gioia 2004) and have six genera with ca.77 species.There are six stamens in Conostylidoideae, flowers are usually pubescent, and there is no red pigment in their rootstocks.
Conostylis is the largest genus in Haemodoraceae, with 45 species recognized by Hopper et al. (1987), and a few more are awaiting description based on recent research (Hopper, unpubl. data), all endemic to southwestern Australia (Fig. 1).The genus contains a few relictual, divergent species as well as several species complexes in which patterns of geo- graphlcal variati o n, natura l hybridi zation, and dys plo idy and/ or po lypl oidy are evident.
E uropean bota ni sts in the nineteenth centur y faced considerable d iffi c ulties in understanding variati o n in th ese complexes fro m specimens collected mainl y by Ja mes Drummo nd and Ludw ig Pre iss.So me 24 spec ies named, for exa mple, by Lindley (1840), Endlic her ( 1846), and Bentham ( l 873), are no w regarded as synonymous with c urre ntly recogni zed species .
Thi s unu suall y hi g h level of synonymy was largely resolved by Green ( 1960), who provided a generi c revisio n based o n herbarium and fi e ld studies together with anato mical (Green 1959) and cyto logical observati o ns.Green ( 1960) recogni zed o nly 23 spec ies, inc luding two new endemics from the the n recentl y developed northe rn kwongan agri cultura l region cente red o n Eneabba and Mt.Lesue ur.
Subsequentl y, the genus has been studi ed in the fi e ld and co llected extensively, resulting in the descriptio n of 15 new species, the re instatement to spec ies statu s of six names regard ed by Green ( 1960) as syno nyms, and the e le vatio n o f one subspecies to full species status (Ho pper 1978(Ho pper , 1980(Ho pper , 1982, L987a ;, L987a ;Green 1982).A total of 22 species additio na l to the 23 recogni zed by Green ( 1960) were described in th e Flora of A ustralia treatme nt of Hopper et al. (1 987).Mo re-over, 3 1 subspecies were a lso recogni zed in the Flora of A ustralia, most described for the first time the re.
Thus, as with ma ny gene ra ende mic to southweste rn Austra li a, more than half the species of Conostylis have been described o r re instated in just the past two decades during an exciting phase of modern botanica l di scovery (Ho pper 2003(Ho pper , 2004;;Ho pper and Gi o ia 2004).Co nseque ntl y, much remains to be in vesti gated to de velop a sound evo lutio nary unde rstanding of the genus.Earl y wo rk on natura l hybridizati on (Ho pper 1977;Krauss and Hopper 2001 ) ;Simpson 1990, l 998).In our earlie r stud y of trnL-F seque nces in Haemodoraceae, Blancoa was siste r to three spec ies fro m di vergent subgenera of Conostylis, bu t with o nl y mode rate bootstrap support.Further work on this generi c re lati o nship was recomme nded .M oreover, the di -vision of Conostylis into six subgenera, although building on earlier ideas of Lindley (1840), Bentham (1873), Green (1960), andGeerinck (1969), has yet to be evaluated with molecular data.
For these reasons, we embarked upon a study of generic and subgeneric relationships in Conostylis and the allied genus Blancoa.Given that sequence divergence in the plastid region trnL-F found in an earlier study of Haemodoraceae (Hopper et al. 1999) was not high in the three species of Conostylis investigated, we used the plastid matK and nuclear ribosomal spacer ITS in this study because they are known in other groups to have more variable sites than trnL-F (e.g., Kores et al. 2001 ).It was anticipated that such a molecular phylogenetic study of Conostylis and Blancoa would also shed light on the phytogeographic history of southwestern Australia, enabling hypotheses on patterns of speciation to be evaluated ( cf.Hopper 1979Hopper , 1992;;James 1981;James and Hopper 1981;Hopper and Gioia 2004).

Plant Material
Sources of plant material and locations of vouchers are listed in Table I.Some 39 of the 47 + species of Conostylis were sampled, together with Blancoa canescens.These taxa represented all subgenera and sections within Conostylis and most morphologically divergent species known in the genus, as well as several closely related taxa within species complexes.Outgroups were chosen from within subfamily Conostylidoideae based on the earlier trnL-F study.They included Anigozanthos (3 spp.), Macropidia (l sp.), Phlebocarya (1 sp.), and Tribonanthes (1 sp.).

DNA Extraction, Gene Amplification and Sequencing
DNA was extracted from 1.0 g fresh or 0.2-0.25 g silica gel-dried leaves (Chase and Hills 1991) using the 2X CTAB method of Doyle and Doyle (1987), except that all samples were purified on cesium chloride/ethidium bromide gradients (1.55 g/mL density).
Amplification of matK and ITS from total DNA was carried out using the primers of H. Sun et al. (2001) andY.Sun et al. (1994 ), respectively.Amplified products were cleaned using Magic Miniprep columns (Promega, Annandale, New South Wales, Australia), following protocols provided by the manufacturer.
Modified dideoxy cycle sequencing with dye terminators run on an ABI 377 automated DNA Sequencer (according to the manufacturer's protocols; Applied Biosystems, Inc., Scoresby, Victoria, Australia) was used to sequence the amplification products directly.

Sequence Alignment
Sequences were added to the matrix and aligned manually following the guidelines of Kelchner (2000).The aligned matrix is available from m.fay@kew.org.

Data Analysis
We analyzed the combined matrix using heuristic searches with PAUP* vers.4.0bl0 (Swofford 2001) using the following strategy: 1000 replicates of randomized taxon entry with subtree-pruning-regrafting (SPR) swapping and a tree limit of 20 trees per replicate to reduce the time spent swapping on suboptimal islands of trees.We then used bootstrapping (Felsenstein l985a,b) to estimate internal support with 500 replicates of simple taxon addition, again with a limit of 20 trees per replicate and SPR swapping.
Two separate analyses were done; both using combined matK and ITS data.Analysis l (Fig. 2) includes all 64 taxa indicated in Table I. Analysis 2 (Fig. 3) includes a subset of 50 of the taxa of Table l, omitting those for which sequences are missing.For both analyses we show one randomly selected tree to illustrate branch lengths (DELTRAN optimization due to problems with ACCTRAN optimization in PAUP* vers.4.0b10) and indicate branches not found in the strict consensus with an asterisk.We report all bootstrap percentages (% BS) >50.

RESULTS
We present here the results of analyses l and 2 (both using the combined matK and ITS data) (Fig. 2, 3).Both analyses corroborate the hypothesis that Conostylis is monophyletic, supported by 99% bootstrap.Moreover, Blancoa is corroborated as sister to Conostylis with 100% bootstrap support in the two analyses.Within Conostylis, two major clades, labeled "/1:' and "B," were resolved, with strong bootstrap support in both analyses (Fig. 2, 3).In clade A, Conostylis sect.Appendicula fell as sister to C. subgen.Conostylis (100% bootstrap in analysis 2).The enigmatic, recently discovered C. glabra, which presently keys out as a member of C. sect.Divaricata within C. subgen.Pendula, was embedded within C. subgen.Conostylis as sister to C. serrulata subsp.serrulata with 95% bootstrap support in analysis 2 (Fig. 3).Conostylis misera was also part of a trichotomy (in the strict consensus trees) at the base of the C. subgen.Conostylis clade.A large clade with 95% bootstrap support (in analysis 2) constituted the majority of this subgenus, but relationships among species were poorly resolved except for a clade comprising C. bracteata, C. candicans subsp.candicans (Fig. lA) and C. paucifiora subsp.euryrhipis, which received 98% support in analysis 2 (Fig. 3).

Generic Relationships
Our study affirms the earlier work on plastid sequences of a much smaller sample of species (Hopper et al. 1999) Table l.Sources of plant material.Classification follows Hopper et al. (1987).Voucher numbers are for S. D. Hopper specimens at Kings Park and Botanic Garden (KPBG) unless otherwise stated.that Conostylis is monophyletic but with higher bootstrap support (99% ).Moreover, we obtained 100% support that Blancoa is sister to Conostylis.Given our more comprehensive sampling of species, and the far greater number of parsimony informative characters in the matK-ITS sequence data than in Simpson's (1990) morphological study, we can discount his hypothesis that Blancoa is embedded within Conostylis.Blancoa differs from Conostylis in its large, tubular, pendulous red flowers borne on a unilateral raceme, the ridged adnation of the upper ovary to the perianth at the septa forming three deep nectar pockets, and the ovules several per locule in two rows on a short, vertically elongate, non-dilated placenta (Hopper 1987b).
Blancoa canescens is pollinated by birds and honey possums (Keighery 1980;Brown et al. 1997).Its phylogenetic position as sister to Conostylis, and its highly modified flowers, indicate that the evolution of adaptations for pollination by vertebrates must have been an old event, predating those that gave rise independently to species of Conostylis with  (Backlund and Bremer 1998).However, in this case we prefer to advocate nomenclatural stability.Blancoa was first named as a distinct genus by Lindley ( 1840), with most authors subsequently upholding this view.Consequently, least disruption to nomenclatural use and maximal retrieval of information from the literature will occur if Blancoa continues to be recognized at generic rank (Backlund and Bremer 1998).

Subgeneric Relationships within Conostylis
In light of the molecular data, significant modification to the subgeneric treatment of Conostylis by Hopper et al. (1987)  Clearly, however, all subgenera with two or more species recognized by Hopper eta!. (1987)  Placement of C. stylidioides in clade A (Fig. 2) is interesting, as it has been hypothesized to be an allopolyploid hybrid of C. aculeata and C. candicans (Hopper 1978(Hopper , 1995;;Krauss and Hopper 2001).The highly divergent C. stylidioides fell outside a weakly supported (59% in Analysis I) clade containing its putative parents, its precise relationships unclear in an unresolved polytomy comprising the C. aculeata-C.candicans clade, a novel strongly supported clade (84% in analysis I) of C. glabra, C. juncea, C. laxiflora, and C. serrulata subsp.serrulata, and two other divergent taxa-C.misera and C. serrulata subsp.magna.Conostylis stylidioides is one of the three most divergent taxa of Conostylis revealed in this study, its number of unique base pair changes (22) approximating those for the monotypic C. subgen.Bicolorata (C.vaginata) and C. subgen.Brachycaulon (C.breviscapa).If it is an allopolyploid, it must be of considerable antiquity, or its genome underwent considerable and rapid reorganization following its hybrid genesis (Krauss and Hopper 2001).
On the other hand, C. subgen.Pendula, with 25 species, has two of its three sections polyphyletic.Only C. sect.Appendicula was retrieved in our study as a clade.Even this section, however, proved to be sister to C. subgen.Conostylis in clade A rather than in clade B with other taxa in C. subgen.Pendula in which C. sect.Appendicula was originally placed.
The diagnostic characters for C. subgen.Pendula are the ovules few and pendulous from a peltate placenta or rarely several on the lateral and lower sides of a globose placenta (Hopper eta!. 1987).On the basis of our results, ovule num-ber and placentation are subject to convergent evolution, with similar ovule number reductions having arisen independently in several lineages of Conostylis.Conostylis subgen.Pendula needs to be much more narrowly circumscribed or its diagnostic characters redefined to parallel its relationships in our study.
The arrangement of taxa in clade B calls for a major reassessment of the subgeneric treatment of Conostylis.Conostylis phathyrantha, the type species of C. sect.Divaricata, is sister to the monotypic C. subgen.Bicolorata (C.vaginata) and C. subgen.Brachycaulon (C.breviscapa) in a moderately supported clade (71% in analysis 1, 63% in analysis 2) sister to all other taxa in clade B. Elsewhere in clade B, there is a large unresolved polytomy comprising four single species (C.albescens, C. androstemma, C. micrantha, and C. rogeri) and three strongly supported clades.
Four vertebrate-pollinated species with large tubular flowers are widely scattered in clade B rather than forming the two monophyletic subgenera: C. subgen.Greenia (C.bealiana, Fig. 1B, C. albescens) and C. subgen.Androstemma (C.androstemma, C. argentea).Multiple origins of vertebrate pollination are indicated here, and these subgenera should be abandoned as polyphyletic.However, it is possible that a tubular perianth could represent an apomorphy for C. albescens, C. androstemma, and C. bealiana, given that their interrelationships are unresolved in strict consensus trees (Fig. 2, 3).A tubular perianth definitely evolved independently in C. argentea, according to our analyses.
Further data are needed from additional gene sequences and taxa before the many polytomies in the consensus tree are resolved.Consequently, we refrain here from making formal changes to the infrageneric classification of Conostylis.However, we do note that C. subgen.Pendula could be recircumscribed as monophyletic by excluding C. sect.Appendicula, C. phathyrantha, and C. glabra and including C. subgen.Androstemma and subgen.Greenia.
With a few exceptions, precise sister species relationships are not identifiable in our study.We can say, however, that some subspecies merit elevation to specific rank based on their molecular phylogenetic placement in our consensus tree (e.g., C. setigera subsp.dasys, C. serrulata subsp.magna).

Chromosome Number Change
Knowledge of chromosome numbers in Conostylidoideae has accumulated since the pioneering count of n = 6 for Anigozanthos .flavidusby Stenar ( 1927).Green (1960) made a major contribution, reporting counts for 24 species of southwestern Australian Conostylidoideae, including confirmation of n = 6 for Anigozanthos, new records of n = 4, 5, 7, 8 and 14 in Conostylis, and n = 8 for Blancoa.James (unpubl. data, 1971)

Macropidia fuliginosa Tribonanthes uniflora
number n = 6 and noted some meiotic irregularities associated with pollen sterility in the hybrids.Hopper (1978) reported a case of allopolyploidy (n = 16) in C. stylidioides.
More comprehensive data for Conostylis and related genera were then reported briefly by Hopper et al. (1987), who listed previously unpublished counts of n = 5, 6, 7, 8, 14, 16, 21 and 28 for Conostylis, n = 8 for Blancoa, and n = 6 for Anigozanthos and Macropidia.Simpson (1990) reported other unpublished counts including n = 7 for Phlebocarya (vouchers: Hopper 840 PERTH, Keighery 691 PERTH), and n = 7 for Tribonanthes (attributed to G. Keighery pers.comm.).Although n = 7 for Phlebocarya is consistent with our own data (Hopper unpubl.data), we have obtained n = 11 for two species of Tribonanthes (Hopper and Stone unpubl. data).Moreover, G. J. Keighery (pers. comm.) has no record of a count of n = 7 for Tribonanthes, and the report of n = 7 for the genus was an error in Simpson 1990 (Simpson pers. comm.).Now that a well-corroborated molecular analysis of Conostylis is in place, the opportunity to explore the evolution of chromosome number change presents itself.Based on outgroup comparison, it is most likely that the ancestor of Conostylis had a chromosome number of n = 8 (Green 1960;Simpson 1990).Blancoa is n = 8, Phlebocarya n = 7, Macropidia n = 6, and Anigozanthos n = 6.Only Tribonanthes, sister to all the above genera in subfamily Conostylidoideae, has a higher number (n = 11).However, members of subfamily Haemodoroideae have higher numbers still, leading to two contrasting models of cytoevolution in Haemodoraceae.Either the common ancestor had n = 8, and the higher numbers in subfamily Haemodoroideae and Tribonanthes are due to ancient palaeopolyploidy (Simpson 1990), or the common ancestor had n = 16, and numbers less than this are due to dysploid decreases.
In Conostylis, one species has n = 16 (C.stylidoides), interpreted as tetraploid on n = 8 (Hopper 1978(Hopper , 1995;;Krauss and Hopper 2001).However, an alternative hypothesis is that n = 16 is a plesiomorphic number retained in C.  (Hopper et al. 1987) and the frequent occurrence of natural hybridization observed in the C. aculeata-C.candicans complex, with only minor lev-els of pollen sterility in hybrids recorded (e.g., Hopper 1977;Krauss and Hopper 2001).Indeed, the only other species of the genus outside C. subgen.Conostylis with numerous ovules per locule is C. androstemma, deeply embedded in clade B among species with few per locule, indicating that the high ovule number in C. androstemma is a parallelism.Moreover, the only case of natural hybridization outside of C. subgen.Conostylis found among the many sympatric combinations of species is a hybrid of C. resinosa X C. tomentosa (voucher Hopper 2473, PERTH) in C. sect.Appendicula.This hybrid was highly sterile, with pollen fertility estimated at 3% (Hopper unpubl. data).
There appears to be significant genetic system divergence among the higher clades in Conostylis.Indeed, in C. sect.Appendicula and C. subgen.Pendula, the two other taxa with species complexes like those in C. subgen.Conostylis, the low ovule numbers are correlated with dysploid chromosome number reduction series and a virtual absence of natural hybridization when congeners are sympatric.The exceptional speciation in southwestern Australian genera of Haemodoraceae exhibited by Conostylis appears to have been based on divergent processes among the three most species-rich subgeneric taxa (sect.Appendicula, subgen.Conostylis, and subgen.Pendula).
This pattern exemplifies endemic speciation in the southwestern region as a whole, which has become known as one of the earth's 25 global biodiversity hot spots (Myers et al. 2000), and has an estimated 8000 species of vascular plants, with 50% endemism (Hopper and Gioia 2004).Hypotheses to account for this exceptional species richness occupy an increasingly diversified literature (references in Hopper and Gioia 2004).Hopper (1979), Lamont et al. (1984), and Cowling et al. (1996) emphasized ecological and geohistorical aspects of speciation, but this approach needs to be combined with a population genetics perspective for an adequate understanding of how populations diverge and ultimately become species.It has become clear that cytoevolution has precipitated or accompanied speciation in a diversity of southwestern Australian groups (e.g., Stylidium Sw. ex Willd.: Coates and James 1996;Coates et al. 2003).
Increasing evidence indicates that the majority of southwestern plants have had unparalleled opportunities to persist in small populations of limited dispersal ability for long periods of time, albeit in the face of significant climatic stresses and recurrent fire (Hopper et al. 1996).James (1981James ( , 1992) ) proposed that the inbreeding and self-fertilization resulting from small population size have placed selection pressures on the evolution of mechanisms that conserve heterozygous genotypes.These mechanisms range from endogenous genetic system responses such as genomic coalescence (e.g., dysploid chromosome number reduction that minimizes the number of randomly assorting elements and increases retention of heterozygous supergenes at meiosis), to exogenous responses like adaptation to mobile vertebrate pollinators that increase the likelihood of outcrossing (an estimated 15% of the southwestern flora is vertebrate pollinated: Keigh-ery1982; Hopper and Burbidge 1986).
An incidental byproduct of those taxa that move down the endogenous response route would be the establishment of barriers to gene exchange among populations-changes in chromosome number or structure selected to maintain heterozygosity within isolated daughter populations would result in meiotic aberrations and sterility in hybrids on secondary contact of daughter and parental populations (Grant 1981 ).Consequently, fecundity would be low and hybridization would be rare.Rapid and prolific microspeciation would occur.Conversely, taxa that evolve towards vertebrate pollination or other exogenous means of maintaining heterozygosity might be expected to display high levels of hybridization and fecundity associated with chromosomal constancy and much less meiotic aberration.Divergence would require isolation by distance (allopatry), either through ecological divergence, long-distance dispersal, or vicariance imposed by landscape or climatic change.
These predictions hold well in the major clades of Conostylis.Thus, C. subgen.Conostylis has adopted exogenous means of maintaining heterozygosity, primarily through ecological divergence in allopatry of sister taxa (Hopper 1977;Hopper et a!. 1987).Conostylis sect.Appendicula and C. subgen.Pendula, in contrast, exhibit the features predicted for clades that have followed the endogenous route of conserving heterozygosity through chromosomal and other genetic system change.
The fundamental question, of course, remains as to the actual genetic basis for such divergent evolutionary patterns.What predisposes clades to follow either the endogenous or exogenous route to conserving heterozygosity or adaptive genotypes under intense selection?Excluding artifacts of classification, why are there so many examples of sister taxa, one monotypic and the other highly species-rich (e.g., Blancoa and Conostylis), in so many independently evolved plant lineages?With the emergence of contemporary techniques enabling DNA sequences to be linked and visualized to positions on chromosomes, we may have the tools to test hypotheses in this most fundamental arena of speciation theory.Most of these species, as well as Blancoa canescens, occur in deep, well-drained, nutrient-poor sands in the northern or southern kwongan of the southwest's Transitional Rainfall and Southeast Coastal Floristic Provinces (Hopper and Gioia 2004), rather than in the laterite, rich loams, or peaty swamps of the forested High Rainfall Province of the deeper southwest.An ancient vicariance event is postulated, with the ancestor of clade A (Fig. 2, 3) confined to the northern kwongan, and that of clade B confined to the southern kwongan.

Phytogeography of Clades in
These two kwongan areas are separated today by the high rainfall forested regions, for which the boundaries extended even further inland during more mesic periods of the Tertiary and Quaternary.There are also complex edaphic barriers in mosaic pattern interspersed through the current wheat belt that separate the large expanses of sands and laterite found in the northern and southern kwongan.The moderating maritime influence on the western and southern coasts since the Jurassic undoubtedly has facilitated the persistence of taxa through arid periods in the northern and southern kwongan, whereas the intervening wheat belt has fewer mesic refuges and may have endured greater extinction rates when rainfall was lower than at present.
There is mounting evidence that many independent clades in southwestern Australian vascular plants exhibit vicariance between the northern and southern kwongan regions, with divergence at varying taxonomic levels from vicarious subspecies to vicarious genera (Hopper 1979;Hopper and Gioia 2004).Consequently, the phytogeographic patterns in major clades of Conostylis and between Conostylis and Blancoa are not unexpected.
Within subgenera of Conostylis, as noted above, it is difficult to elaborate on the phytogeography of speciation without better resolution of the molecular pattern.However, the phylogenetic pattern is consistent across several lineages with a hypothesis of the prolonged persistence of relictual taxa combined with explosive more recent speciation, the latter pronounced in the northern kwongan of the Transitional Rainfall Province (Hopper and Gioia 2004).Further investigation is needed to evaluate these ideas and hopefully elucidate speciation in this most diverse genus of the Haemodoraceae.
and seed bi o logy (Ti eu et a l.2001 a, b) already indi cated the pote nti al for future studies .The generi c re lati onships of Conostylis a nd subgeneri c taxo no my of the genus re ma in controversia l or in need of further e luc idati on.Whereas most authors recogni zed the monotypic Blancoa as sister to Conostylis (e.g., Lindley 1840 ; Be ntham 1873; Ke ighe ry 1980 ; Hopper 1987b ; Ho pper et a l.1999; Fig. 1), oth ers have considered Blancoa a syno nym of Conostylis (Mue ll er 1873; Geerinck 1969; Macfarl a ne 1987 is needed.Two major clades emerged within the genus-clade A corresponding to C. subgen.Conostylis + C. sect.Appendicula (of C. subgen.Pendula) and the enigmatic C. glabra and clade B including all other species in a large C. sect.Catospora of C. subgen.Pendula and the remaining four minor subgenera.Each of these two major clades received very strong bootstrap support, but neither has any recognizable morphological synapomorphies.
are paraphyletic.Of the two largest subgenera, C. subgen.Conostylis (14 spp.) was retrieved intact in clade A in our study, requiring only the addition of C. glabra to become monophyletic.Although C. glabra is not yet formally described, this robust local endemic keys out as a third member of C. sect.Divaricata within C. subgen.Pendula.The other two species of this section, C. phathyrantha and C. neocymosa (Fig. 1C, D), occur at divergent points in clade B. It is clear that the loosely cymose inflorescences with flowers on long pedicels seen in C. phathyrantha, C. neocymosa, and C. glabra have evolved independently in three divergent lineages of Conostylis.Conostylis sect.Divaricata as circumscribed by Hopper et a!.(1987) should be abandoned as polyphyletic.With the inclusion of C. glabra, the monophyletic C. subgen.Conostylis is diagnosed by numerous ovules per locule borne on a globose placenta and a chromosome number of n = 8 (except for C. stylidioides which has n = 16).
Fig. 2.-Tree from the combined matK and ITS analysis.Number of changes above the line, bootstrap values below.Nodes marked by an asterisk(*) collapsed in the consensus tree.Subgeneric taxa from Hopper eta!. (1987) indicated as: P-C = Conostylis subgen.Pendula sect.Catospora; P-D = C. subgen.Pendula sect.Divaricata; P-A = C. subgen.Pendula sect.Appendicula; G = C. subgen.Greenia; A = C. subgen.Androstemma; Br = C. subgen.Brachycaulon; Bi = C. subgen.Bicolorata; C = C. subgen.Conostylis.Large asterisk beside Conostylis taxa indicate possession of a tubular perianth.Note monophyletic groups A (containing clades I and 2), and B (containing clades 3 and 4). 100 stylidioides and in no other species.This seems unlikely because all other taxa in C. subgen.Conostylis are uniformly diploid as n = 8.Circumstantial but not yet definitive evidence for an allopolyploid origin of C. stylidioides exists (Krauss and Hopper 2001), whereas there is none for the massive extinction of taxa with n = 15 to n = 9 in C. subgen.Conostylis required to support the hypothesis of C. stylidoides as a plesiomorphic diploid of n = 16, and other species in the subgenus derived from dysploid reduction to and stabilization on n = 8.Significant correlations with the chromosomal uniformity on n = 8 (excluding C. stylidioides) in C. subgen.Conostylis are the numerous ovules per locule

Table I .
Continued.