Phylogenetics of Asphodelaceae (Asparagales): an analysis of plastid rbcL and trnL-F DNA sequences

Annals of Botany 86: 935±951, 2000 doi:10.1006/anbo.2000.1262, available online at http://www.idealibrary.com on

Phylogenetics of Asphodelaceae (Asparagales): An Analysis of Plastid rbcL and trnL-F DNA Sequences M A R K W. C H A S E * {, A N E T T E Y. D E B R U I J N {, A N T HO N Y V. CO X {, GA I L R E E V E S {, PA U L A J. R U D A L L{, M A R G A R E T A . T. J O H N S O N { and L U IS E . E G UI A R T E{ {Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3DS, UK and {Centro de EcologõÂa, Universidad Nacional AutonoÂma de MeÂxico, Apartado Postal 70-275, C.U., C.P. 04510 MeÂxico, D.F., Mexico Received: 14 May 2000

Accepted: 11 July 2000

Phylogenetic relationships of Asphodelaceae were investigated by parsimony analysis of 57 monocot rbcL nucleotide sequences, including 17 genera that have at some time been assigned to the family. All genera of Asphodelaceae except for three (Hemiphylacus, Paradisea and Simethis) form a strongly supported monophyletic group with Hemerocallidaceae and Xanthorrhoeaceae as their immediate sister taxa. In a second analysis, we added 34 plastid trnL-F sequences (an intron and a spacer between two transfer RNA genes) for the Asphodelaceae clade and nearest outgroup families (Doryanthaceae, Hemerocallidaceae, Iridaceae, Ixioliriaceae, Tecophilaeaceae and Xanthorrhoeaceae) in an attempt to improve resolution and levels of internal support. The results from the separate analyses produced highly similar although not identical results. No strongly supported incongruent groups occurred, and we combined both sequence regions in one analysis, which demonstrated improved results. Strong support exists for a monophyletic subfamily Alooideae, but this leaves a paraphyletic subfamily Asphodeloideae because Bulbine/ Jodrellia alone are strongly supported as the sister group of Alooideae. Characters that have been used to separate Alooideae as a distinct group (either as here a subfamily or as a separate family by other authors), such as secondary growth and bimodal karyotypes, are found in at least some members of Asphodeloideae, particularly in Bulbine and # 2000 Annals of Botany Company Jodrellia for the karyotypes, making Alooideae less easily recognized. Key words: Alooideae, Asphodeloideae, Asphodelaceae, Asparagales, phylogenetic analysis, rbcL, trnL-F, molecular systematics.

I N T RO D U C T I O N The broad de®nition of Lilianae by Dahlgren et al. (1985) divided this largest and most diverse superorder of monocotyledons into ®ve orders. The largest of these orders is Asparagales (currently comprising 29 families), which these authors considered to be a monophyletic group but which contained families that may be para- or even polyphyletic. Dahlgren et al. (1985) chose to use narrow family circumscriptions that were more likely to be monophyletic. One of the families within Asparagales recognized by Dahlgren et al. (1985) was Asphodelaceae, which they divided into two subfamilies: Asphodeloideae and Alooideae. Many previous authors (see Smith and Van Wyk, 1998, for example) used Asphodelaceae, but there has been a general lack of con®dence in their monophyly (Dahlgren et al., 1985). Asphodelaceae may be distinguished from other lilioid monocot groups by a combination of characters: general presence of anthraquinones, lack of steroidal saponins, simultaneous microsporogenesis, atypical ovular morphology and presence of an aril. Each of these occurs in other asparagoid groups, so it may be only the combination that is distinctive. In this paper, we address the taxonomic diculties that surround Asphodelaceae, as described below. * For correspondence. Fax ‡44 (0) 208 332 5310, e-mail m.chase@ rbgkew.org.uk

0305-7364/00/110935+17 $35.00/00

Subfamily Alooideae are a distinctive group of plants noted for their often spectacular secondary growth, which in many areas of tropical and southern Africa form the dominant plants. This character was used by Dahlgren et al. (1985) to de®ne Alooideae as monophyletic, but not all of the genera and species in this subfamily exhibit this trait. Secondary growth via a secondary thickening meristem (STM) is unique to Asparagales (Rudall, 1995) and occurs commonly within the order in at least some genera of Agavaceae, Aphyllanthaceae, Convallariaceae (e.g. Dracaena Vand. ex L., Nolina Michx.), Hemerocallidaceae (e.g. Phormium J. R. Forst. & G. Forst.), Iridaceae (e.g. Nivenia Vent., Klattia Baker), Laxmanniaceae (e.g. Lomandra Labill.) and Xanthorrhoeaceae [all familial circumscriptions following the classi®cation of the Angiosperm Phylogeny Group (APG), 1998]. Except for Asphodelaceae, Hemerocallidaceae and Xanthorrhoeaceae, most of these families are not particularly closely related to each other (Chase et al., 1995, 2000; Rudall et al., 1997; Fay et al., 2000). According to Dahlgren et al. (1985, citing circumscription of the family by Schulze, 1975), Alooideae consist of seven genera: Aloe L., Astroloba Uitew., Chamaealoe A.Berger (ˆ Aloe of many authors), Gasteria Duval, Haworthia Duval, Lomatophyllum Willd. and Poellnitzia Uitew. Other authors have added Chortolirion A.Berger (Smith and van Wyk, 1998, which is considered by some as a synonym of Haworthia). Their distribution is centred in # 2000 Annals of Botany Company

936

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae

Africa and western Asia, with a remarkable southern African concentration. All Alooideae have caps of `aloin' cells at their phloem poles, which are thought to be involved in secondary metabolite production and are absent in most, but not all, Asphodeloideae (Beaumont et al., 1985). The remaining genera of Asphodelaceae, which are usually grouped in subfamily Asphodeloideae, are more varied, and there is a great deal of morphological similarity between members of this group and some members of Anthericaceae s. str. (see Chase et al., 1996 for a recircumscription of Anthericaceae), although all Anthericaceae have successive microsporogenesis whereas Asphodelaceae uniformly exhibit simultaneous microsporogenesis (Rudall et al., 1997). For example, Trachyandra Kunth (Asphodelaceae) has often been treated as a section within Chlorophytum Ker Gawl. (Anthericaceae; Nordal et al., 1990). The morphological similarity, together with the possibility that Asphodeloideae may be more closely related to other families with simultaneous microsporogenesis, led Dahlgren et al. (1985) to conclude that Asphodeloideae may be para- or even polyphyletic. Several molecular studies (Chase et al., 1993; Duvall et al., 1993a,b; Rudall et al., 1997; Fay et al., 2000) placed Knipho®a Moench, Aloe and Haworthia together and generally supported the ordinal circumscriptions of Asparagales sensu Dahlgren et al. (1985), with the exception that in all these studies Orchidaceae and Iridaceae appear in Asparagales, rather than in Liliales as in Dahlgren et al. (1985). However, higher level relationships are not the focus here. Generally, Asphodeloideae have been thought to be comprised of nine genera: Asphodeline Rchb., Asphodelus L., Bulbine Wolf, Bulbinella Kunth, Eremurus M.Bieb., Jodrellia Baijnath, Knipho®a, Simethis Kunth and Trachyandra. These have a wide Old World distribution: western Asia, Europe, and southern and tropical Africa. The exceptions to this pattern are Bulbine and Bulbinella, the former from Australia and the latter from New Zealand as well as South Africa. Some authors have disagreed with Dahlgren and coworkers' circumscription of Asphodelaceae. Brummitt (1992) divided the monocotyledons into approx. 95 families, among which were both Asphodelaceae and Aloaceae. This approach was based on a presumed di€erence in karyotype between Asphodelaceae and Aloaceae (Brummitt, pers. comm.). Aloaceae were based on x ˆ 7 with a distinctive and consistent bimodal karyotype of four long and three short pairs of chromosomes, six acrocentric with one long submetacentric (Brandham, 1973), whereas other Asphodelaceae were found not to have this particular karyotype even though some species also have bimodal karyotypes based on x ˆ 7. Brummitt (1992) also placed Hemiphylacus S.Watson and Paradisea Mazzuc. in Asphodelaceae. These genera were mistakenly included and should have been placed in Anthericaceae (Brummitt, pers. comm.). Cronquist (1981) provided a more traditional approach to circumscription of monocot families. He recognized Aloaceae as one of 15 families of Liliales but retained nearly all Asphodeloideae in Liliaceae s.l. Although he stated that `Bulbine sometimes approaches Aloaceae in

aspect', Cronquist (1981) still placed the genus within Liliaceae. He was also unsure how to treat Knipho®a but added this genus to Aloaceae, presumably because of the similar ¯ower morphology: Knipho®a and all Aloaceae have a fused, tubular perianth. This disagrees with the system of Dahlgren et al. (1985) in which Knipho®a is placed with Asphodeloideae because it lacks `aloin' cells (Beaumont et al., 1985). Aloaceae sensu Cronquist (1981) thus comprise Alooideae plus Knipho®a of Asphodeloideae of Dahlgren et al. (1985). In an attempt to clarify these issues, we conducted a DNA sequence study to evaluate the place of Asphodelaceae in the asparagoid lilies, particularly their separation from Anthericaceae, and relationships within the family. We sequenced the plastid rbcL gene for 21 taxa and analysed the data using di€ering numbers of outgroups. This region was chosen here because recent studies using this gene (Chase et al., 1993, 1995) enabled us to take advantage of the large existing database of outgroups now available. Knowing that the patterns found with rbcL, although reproducible with other data (Soltis et al., 1997; Chase and Cox, 1998; Chase et al., 2000; Fay et al., 2000), are generally weakly supported or unresolved, we have also added 34 new plastid trnL-F sequences (Taberlet et al., 1991); this region consists of an intron in the transfer-RNA gene, trnL (UAA), and the adjacent spacer between trnL and trnF (GAA). We will thus also examine the degree of congruence between these three plastid regions, one coding and two non-coding, and compare their patterns of molecular evolution. M AT E R I A L S A ND M E T H O D S Plant material Fresh plant material was used for DNA extraction. Flowers of Aloe and related genera were preferred to leaf material because mucilage present in some of the leaves interfered with DNA extractions. Voucher specimens were made for all taxa sequenced for this study; voucher information, literature citations and database accession numbers are listed in Table 1. For ®ve taxa, di€erent DNA samples (often for di€erent species) were used to produce the two sequence regions (Table 1). In the case of the rbcL sequence of Cyanastrum cordifolium, we used a di€erent species, C. hostifolium, which was subsequently described as a new genus, Kabuyea (Brummitt et al., 1998); this species is the sister species of C. cordifolium in the phylogenetic results of Brummitt et al. (1998), which should not a€ect the topology, given the sparse sampling from Tecophilaeaceae in our study. In all, 21 new rbcL and 34 new trnL-F sequences are reported for this study. DNA extraction, gene ampli®cation sequencing and alignment Approximately 0.5±1.5 g of fresh material was ground at 658C with a mortar and pestle in 2 % w/v CTAB (hexadecyltrimethylammonium bromide in 100 mM pH 8.0 Tris-HCl, 1.4 M NaCl, and 20 mM EDTA; Doyle and

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae Doyle, 1987). Lipids were removed with SEVAG (24 : 1 chloroform : isoamyl alcohol), and DNA was precipitated with ethanol at ÿ208C, followed by equilibrium centrifugation in CsCl-ethidium bromide gradients (1.55 g ml ÿ1) to clean the DNA and remove RNA. The rbcL gene was ampli®ed (Saiki et al., 1987) using primers described by Olmstead et al. (1992) and Muasya et al. (1998). The trnL-F region was ampli®ed with primers c and f of Taberlet et al. (1991). After ampli®cation, the double-stranded product was puri®ed using `Wizard' minicolumns (Promega, Ltd.), according to the manufacturer's protocols. Gene sequencing was carried out using standard dideoxy-chain termination reactions (Sanger et al., 1977). For manual sequencing, a set of eight internal primers was used to determine the sequence of each rbcL except for the ®rst 26 base pairs (bp), which matches the forward ampli®cation primer, and the last 72 bp; the downstream reverse primer used in other rbcL studies does not work routinely in the asparagoid lilies so we used an internal one that anneals to a 20 bp region beginning at position 1352 in the Nicotiana tabacum rbcL sequence. For some of some of the rbcL and all of the trnL-F sequences, we used modi®ed dideoxy cycle-sequencing methods according to the manufacturer's protocols (PE Applied Biosystems, Inc.), scaling down to a quarter of the recommended reaction volume and run on a ABI 377 automated sequencer (PE Applied Biosystems, Inc.). The resulting ®les were edited and assembled using Sequence Navigator and Autoassembler software (PE Applied Biosystems, Inc.). Using these methods, a set of four primers was sucient to determine an rbcL sequence (as in Muasya et al., 1998); for trnL-F, the two ampli®cation primers were then used as sequencing primers which, due to the shorter length of this region, were generally sucient to produce complete sequences for each strand. Alignment of rbcL sequences is easily accomplished because there was no length variation observed for these taxa. Non-coding regions of DNA, like the trnL-F region, are subject to insertions and deletions (indels), and so must be aligned before analysis. There are software programmes available for this purpose, but none of them take into account the di€erent patterns of indel evolution. For example, indels are of three major types: unique, 5±30 (exceptionally more) bp repeats, and simple-sequence repeats of a single base (in plastid DNA these are usually A/T) or two-three bases (AT, ATT or AAT). Each of these indel types must be ®rst identi®ed before an accurate interpretation of alignment is possible, and therefore we manually aligned these regions. Some regions were excluded from the analyses (A and B; see below) because we could not obtain an unambiguous alignment of all taxa (the aligned matrix is available from the EBI database, accession number DS42597, and MWC, [email protected]). All rbcL and trnL-F sequences are also individually available from the same source (accession numbers AJ290254AJ290321; Table 1). Cladistic analysis Three separate sets of analyses were carried out. The ®rst (analysis A) comprised rbcL sequences of 54 taxa

937

representing 21 families of asparagoid lilies. Orchidaceae, Hypoxidaceae, Asteliaceae and Blandfordiaceae were speci®ed as the outgroup, based on the ®ndings of Chase et al. (1995, 2000). We repeated our analysis (analysis B) using only a subset of the nearest families and adding three taxa that had subsequently become available. For the B analysis, we also included the trnL-F region. In both analyses, rbcL base positions from 1±30 and 1352 onwards were excluded because many of the sequences used were not determined for these regions. Analyses were carried out using PAUP* version 4.0d64. (Phylogenetic Analysis Using Parsimony; Swo€ord, 1998). In both sets of analyses, a tree search was performed under the Fitch criterion (unordered, equal weights; Fitch, 1971; subsequently termed `Fitch search'), using 1000 replicates of random taxon entry-order and SPR (subtree pruning regrafting) branch-swapping with the MULPARS option on (multiple parsimony, i.e. keeping more than one tree at each step), but using the HOLD option set to permit no more than ten trees retained per replicate to minimize the time spent swapping on large sets of trees. After completing these 1000 replicates, we used all trees found as starting trees for a search to completion (using SPR and with MULPARS on), unless too many trees were found, as in analysis A, in which the search was stopped due to memory limits after 39 900 trees were found. For all searches, internal support was assessed using 1000 bootstrap replicates (Felsenstein, 1985). Each replicate was run with simple taxon-entry order, and a tree limit of ten, SPR branch swapping with MULPARS but permitting only ten trees to be held at each step (again to minimize the time spent swapping on each replicate). Bootstrap support is indicated as percentage of trees found that contain that group of taxa. Percentages of less than 50 % are not reported because there is no signi®cance in a group being found in less than 50 % of the replicates. Branches with less than 50 % are also not likely to be present in the strict consensus tree, which means that such groups have no internal support. We use the following descriptions for categories of bootstrap support: weak, 50±74 %; moderate, 75±84 %; strong, 85±100 %. In the B set of analyses, we included 34 taxa that covered Asphodelaceae as well as the nearest families indicated by analysis A. We analysed each matrix separately, and then because they were in general agreement about relationships, we combined them in a single analysis. Search strategies were as described above, but in addition successive approximations weighting (SW; Farris, 1969) was then used to down-weight base positions that changed excessively. Reweighting was based on the rescaled consistency index (RC) with a base weight of one on the best trees. Rounds of re-weighting and analysis (ten replicates of random taxon-entry order per round) were continued until the same tree length (the assigned weights) was obtained in two successive rounds. Because analysis A was used only to indicate which taxa should be used in the B analyses, we did not implement SW on that search. Due to so many regions in which gaps were ambiguous, we did not code gaps in the matrix from the B analysis of trnL-F. However by limiting the search to just Alooideae plus Bulbine/Jodrellia, in which we could, in our opinion,

938

T A B L E 1. Material studied, voucher data, DNA sequence documentation and karyotype data and references Cytology reference

Z69238 Z73700 L10253

60 Y 30 N grad 30 Y and N 60, 58 N grad (60) Y

Brandham, 1969 Fernandez and DavinÄa, 1971 Cave, 1970 Kaneko, 1970 Tamura, 1995

Fay and Chase, 1996 Fay and Chase, 1996

Z69205 Z69203

14 N 12 N

Karavokyrou and Tzanoudakis, 1991 Vijayavalli and Mathew, 1990

A

Duvall et al., 1993b

L05032

22 ±

Sveshenikovka and Zemskova, 1988

Chase 515, K

A

Chase et al., 1995

Z69225

(18) N grad (16) N

Bradley 7331, GMUF Chase 826, K

A A

Duvall et al., 1993b Chase et al., 1995

L05031 Z69229

28 N 30 N 48 N

Stedje and Nordal, 1994 Nordal et al., 1990 Kativa, 1994 Wetschnig, 1988 SatoÃ, 1942

Duvall 19920604, UCR Chase 668, K

A A

Duvall et al., 1993b This paper

L05028 Z73688

20 N grad 112 N

Rudall et al., 1998 Rudall et al., 1998

Collector/number, herbarium

Agavaceae Agave celsii Hook. Camassia leichtlinii S.Watson Chlorogalum pomeridianum Kunth Hosta rectifolia Nakai Yucca recurvifolia Salisb.

Eguiarte 6, MEXU Chase 483, K Chase 838, K Duvall 19920601, UCR Eguiarte 1, MEXU

A A A A A

Duvall et al., 1993b This paper This paper Duvall et al., 1993a Duvall et al., 1993b

Alliaceae Allium subhirsutum L. Tulbaghia violacea Harv.

Chase 439, K Chase 248, NCU

A A

Amaryllidaceae Clivia miniata Regel

Bradley 24976, GMUF

Anthericaceae Anthericum liliago L. Chlorophytum comosum (Thunb.) Jacques Paradisea liliastrum Bertol. Asparagaceae Asparagus ocinalis L. Hemiphylacus latifolia S.Watson Asphodelaceae Aloe bakeri Scott-Elliot Aloe vera (L.) Burm.

EBI accession rbcL/trnL-F{

A, B A, B

This paper/this paper Duvall et al., 1993b/this paper

Z73680/AJ290254 & AJ290288 L05029/AJ290255 & AJ290289

14 Y 14 Y

Brandham, 1971 Vijayavalli and Mathew, 1990

Asphodeline lutea Rchb.

Chase 695, K Bradley 24977, GMUF (rbcL) Chase 797, K (trnL-F) Chase 300, K

A, B

This paper/this paper

Z73681/AJ290256 & AJ290290

Asphodelus aestivus Brot. Astroloba foliolosa (Haw.) Uitewaal Bulbine semibarbata (R.Br.) Haw. Bulbine succulenta Compton Bulbine wiesei L.I.Hall Bulbinella cauda-felis L. Eremurus himalaicus Baker

Chase 482, K Chase 684, K Dixon s.n., WA Chase 291, K Chase 3504, K Chase 297, K Chase 490, K

A, B A, B B A, B B A, B A, B

This paper/this paper This paper/this paper No rbcL/this paper This paper/this paper This paper This paper/this paper This paper/this paper

Z73682/AJ290257 & AJ290291 Z73683/AJ290258 & AJ290292 no rbcL/AJ290259 & AJ290293 Z73684/AJ290260 & AJ290294 Y17333/AJ290261 & AJ290295 Z73685/AJ290262 & AJ290296 Z73686/AJ290263 & AJ290297

Gasteria liliputiana Poelln. Haworthia subfasciata Baker Hawarthia coartacta Haw. Jodrellia macrocarpa Baijnath Knipho®a uvaria (L.) Hook. Lomatophyllum purpureum Th.Dur. & Schinz Poellnitzia rubri¯ora (L.Bolus) Uitewaal Trachyandra sp.

Chase 276, NCU Bradley 24978, GMUF (rbcL) Chase 3859, K (trnL-F) Chase 3941, K Chase 120, NCU Chase 694, K Chase 669, K Chase 1027, K

A, B A, B

This paper/this paper Duvall et al., 1993b/this paper

Z73687/AJ290264 & AJ290298 L05035/AJ290265 & AJ290299

28 (14) Y 28 N 28,84 N 14 Y 26, 28, 52, 54, 78 Y (14) Y (14) Y 12 N (14) Y 14 N 14 Y 14 Y

This paper (MAT Johnson, pers. obs.) Tzanoudakis and Kypriotakis, 1987 Diaz Lifante, 1996 Brandham, 1973 Watson, 1986 Stedje and Nordal, 1994 Stedje and Nordal, 1994 This paper (MAT Johnson, pers. obs.) Tamura, 1995 Zakirova and Nafanailova, 1990 Vosa and Bennett, 1990 Chinappa and Semple, 1976

A, A, A, A, A,

This paper/this paper Chase et al., 1993/this paper This paper/this paper This paper/this paper This paper/this paper

Y17335/AJ290266 & AJ290300 Z73689/AJ290267 & AJ290301 Z73690/AJ290269 & AJ290303 Z73691/AJ290268 & AJ290302 Z73692/AJ290270 & AJ290304

14 12 14 14 14 14

Stedje and Nordal, 1994 Vijayavalli and Mathew, 1990 Brandham, 1971 Smith, 1991 Tamura, 1995 Stedje and Nordal, 1994

B B B B B

N N grad Y Y Y Y

Asteliaceae Milligania stylosa F. Muell. ex Benth.

Chase 511, K

A

This paper

Z73693

None

None

Blandfordiaceae Blandfordia punicea Sweet

Chase 519, K

A

This paper

Z73694

(34, 54) N

Di Fulvo and Cave, 1964

Convallariaceae Comospermum yedoeÈnsis (Maxim. ex Franch. & Sav.) Rauschert Danae racemosa Moench Dasylirion longissimum Lem.

Chase 833, K

A

This paper

Z73679

Chase 121, NCU Eguiarte 7, MEXU

A A

Chase et al., 1993 Duvall et al., 1993b

Z73708

Nolina (Beaucarnea) recurvata (Lem.) Hemsl.

Peterson s.n., US

A

Duvall et al., 1993a

L05030

Polygonatum hookeri Baker Sansevieria cylindrica Bojer

Chase 492, K unknown

A A

This paper Chase et al., 1993

Z73695 Z73698

40 N 40 N 40 Y (38) N 38 N/Y (38) ? (38) N (38) ? 38 ± 30 N 112 N grad

This paper (MAT Johnson, pers. obs.) SatoÃ, 1942 [as Alectorus yedoensis] Tamura, 1995 Johnson and Gale, 1983 SatoÃ, 1942 Tamura, 1995 Johnson and Gale, 1983 Cave, 1964 Whittaker, 1934 Therman, 1953 Vijayavalli and Mathew, 1990

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae

2n ˆ karyotype Y, N}

Matrix*

DNA sequence reference{ rbcL/trnL-F

Taxon

Clements 4812, CANB Albert s.n., K Chase O-199, K 243187, TRT (rbcL) Chase 1378, K (trnL-F) Chase 447, K Unvouchered Chase 1575,K Chase 192, NCU Chase 3064, K

Orchidaceae Apostasia stylidioides Rchb.f. Cypripedium irapeanum Lex. in La Llave & Lex. Epipactis helleborine Crantz

Tecophilaeaceae Cyanastrum cordifolium Oliv./ Kabuyea hostifolia (Engl.) Brummitt Tecophilaea cyanocrocus Leyb. Walleria mackenzii J.Kirk Zephyra elegans D.Don

Xanthorrhoeaceae Xanthorrhoea resinosa Pers. Xanthorrhoea minor R.Br. A, B A, B

Chase et al., 1993/this paper This paper/this paper

Z73710/AJ290271 & AJ290305 Y17339/AJ290272 & AJ290306

Chase et al., 1995/this paper Z73709/AJ290276 & AJ290310 Brummitt et al., 1998/this paper Y17338/AJ290279 & AJ290313 Brummitt et al., 1998/this paper Y17340/AJ290277 & AJ290311

A, B A, B A, B

Z73696/AJ290278 & AJ290312

Z73705 Z73706 Z73707

Z73704/AJ290280 & AJ290314

Z77287/AJ290283 & AJ290317 Z77289/AJ290287 & AJ290321 AJ277879/AJ290284 & AJ290318 AJ277880/AJ290286 & AJ290320

Chase et al., 1993/this paper

Chase et al., 1995 Chase et al., 1993 Chase et al., 1995

Chase et al., 1995

Chase et al., 1995/this paper Chase et al., 1995/this paper This paper/this paper This paper/this paper

L05037/AJ290282 & AJ290316

Z77282/AJ290285 & AJ290316

Z73702

Z73701

Z69237 L05038

Z69230

Z69231/AJ290275 & AJ290309

L05036/AJ290274 & AJ290308

Not submitted/AJ290273 & AJ290307

Z73697/AJ290281 & AJ290315

A, B

A A A

A, B

B B B B

Duvall et al., 1993/this paper

A, A, A, A,

This paper/this paper

A, B

Chase et al., 1993

Chase et al., 1993

Chase et al., 1993 Duvall et al., 1993b

A, B

A

A

A A

Chase et al., 1995

Fay and Chase 1996/this paper

A, B

A

Duvall et al., 1993b/this paper

Duvall et al., 1993b/this paper

Chase et al., 1995/this paper

A

A, B

A, B

N grad Y Y ±

± ± ± ± ± ± ±

(22)Y 22 Y

24 N

22 N

(48) N 20 N 20, 38, 40 Y

24 ± 72 N

24 28 34 36 44 48 60

(36) ± (18) N (42) Y 14,28,53,54,55,56, 67,70,78,80,81,85,87,89, 92,94, 98, 102,105 N

20 N 26, 30

54 ±

48 N 24 N

22, 33 N

32 N 32 N

48 48 36 34

Briggs, 1966 Briggs, 1966

Brummitt et al., 1998

Brummitt et al., 1998

Okada, 1988 Malakhova, 1990 Yokota, 1987

Zakirova and Nafanailova, 1990 Karihaloo and Koul, 1984

Goldblatt, 1982

Mitra, 1966 SatoÃ, 1942 Naranjo, 1975 Zimudzi, 1994

Bruyns and Vosa, 1987 Hara,1969

SatoÃ, 1942

This paper (MAT Johnson, pers. obs.) Locatelli-Lanzara et al. 1971

Jin, 1986

Vijayavalli and Mathew, 1990 Tanaka, 1981

Briggs, 1966 Cave, 1964 Whittaker, 1934 Newman, 1929

Family taxonomy is that of Angiosperm Phylogeny Group (1998). *Analysis: A, analysis of a large rbcL dataset containing 53 lilioid monocots; B, analysis of the smaller rbcL and trnL-F dataset, containing Asphodelaceae and 17 outgroup taxa. {Publication citing this rbcL sequence for the ®rst time. {trnL-F sequences were submitted as two pieces, the trnL intron and the trnL-F intergenic spacer. }Bimodal tendency: Y, yes; N, no; Y/N, some species bimodal, others not; N grad, graduated karyotype; ? ˆ not analysable; Blank, not seen.

Chase 489, K

Bruhl s.n., TAS Goldblatt s., MO UNSW 21494 Orchard 35, MO

Goldblatt 9500, MO (rbcL) Goldblatt 9362, MO (trnL-F) Bradley 25976, GMUF (rbcL) Chase I-100, K (trnL-F)

Ixioliriaceae Ixiolirion tataricum Schult.

Isophysis tasmanica T.Moore Nivenia corymbosa Baker Patersonia glabrata R.Br. Witsenia maura Thunb.

Iridaceae Aristea glauca Klatt Aristea coerulea Vahl Iris Xgermanica L./ Iris unguicularis Poir.

Chase 108, NCU

Chase 205, NCU

Hypoxidaceae Curculigo capitata Kuntze

Hypoxis leptocarpa Engelm.

Chase 176, NCU None

Chase 801, K

Hyacinthaceae Bowiea volubilis Harvey ex Hook.f. Scilla socialis (Baker) Jessop

Herreriaceae Herreria montevidensis Klotzsch ex Griseb.

Chase 3869, K (trnL-F) Duvall 19920601, UCR (rbcL) Chase 3833, K (trnL-F) Ortiz s.n. (SANT)

Eguiarte 8, MEXU (rbcL)

Hemerocallideae Dianella cf. ensifolia (L.) D.Don

Dianella ensifolia (L.) D.Don Hemerocallis fulva L. Hemerocallis littorea Makino Simethis mattiazzii (Vand.) Sacc.

Chase 188, NCU

Doryanthaceae Doryanthes excelsea CorreÃa

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae 939

940

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae

unambiguously align the complete sequences, we wished to determine if we could improve the estimate of relationships in Alooideae by including the indels in the analysis. We coded indels using the PaupGap program of A. V. Cox (1997; available on request from MWC) to produce a gap matrix of 88 presence/absence characters coded as A (absent) and T ( present) so that they could be run with the same character coding as used for the DNA sequences, and this gap matrix was included with the combined matrix (C analyses). Because of the reduction in taxon number, this portion of the analysis used the Branch and Bound (B&B) algorithm of PAUP, with furthest addition of sequences, MULPARS on (no HOLD limit set), and the upper bound calculated stepwise (the standard B&B settings). We used MacClade 3.06 (Maddison and Maddison, 1992) to calculate the number of steps, consistency index (CI) and retention index (RI) for each codon position in rbcL and the numbers of steps at each base position in rbcL and trnL-F. To evaluate frequencies and performance of transitions and transversions, we used the step matrix given below to down-weight transitions to zero, thereby giving us the contribution of transversions to tree length, CI and RI, and from this we calculated the numbers of transitions as well as their CI and RI. For all comparisons of matrices, we used only the taxa for which we had both rbcL and trnL-F sequences (i.e. we thus left out Bulbine semibarbata (R.Br.) Haw). [A] [C] [G] [T]

A ± 1 0 1

C 1 ± 1 0

G 0 1 ± 1

T 1 0 1 ±

R E S U LT S Analysis A The length of rbcL included in analysis A was 1321 bp, of which 236 were variable and 207 were potentially informative. Heuristic search (the larger rbcL only matrix), comprising 54 asparagoid monocots (see Table 1), resulted in a shortest tree length of 1011 steps, with more than 39 900 equally parsimonious trees at this length. CI (consistency index; with autapomorphies included) ˆ 0.51 and RI (retention index) ˆ 0.67. Because we reach no conclusions based on CI, reporting it without autapomorphies serves no useful purpose. One of the equally most parsimonious trees is shown in Fig. 1 to illustrate general relationships and branch length distributions. Bootstrap percentages are indicated in parentheses above each branch; arrowheads indicate groups not present in the strict consensus tree. The tree shown does not contradict the topology of the asparagoid clade reported by Duvall et al. (1993b) and Chase et al. (1993, 1995). Asphodelaceae are grouped together in one strongly supported monophyletic clade (100 % bootstrap support; BS). Only the two genera mistakenly added to Asphodelaceae by Brummitt (1992), Hemiphylacus and Paradisea, are not included in this clade. Hemiphylacus comes out in a strongly supported clade with

Asparagus L. (Rudall et al., 1998; BS 100 %), and Paradisea is in Anthericaceae (BS 100 %), an anity already noted by Dahlgren et al. (1985). Simethis, included in Asphodelaceae by Dahlgren et al. (1985), is strongly supported (BS 96 %) as sister to Hemerocallis in Hemerocallidaceae. Apart from Asphodelaceae, other moderate to strongly supported groups within Asparagales are: Orchidaceae, Hypoxidaceae, Iridaceae, Tecophilaeaceae and a large group in which Asphodelaceae, Hemerocallidaceae and Xanthorrhoeaceae are included (99 %) with the `higher asparagoids' (Chase et al., 1995; Rudall et al., 1997). Within these `higher asparagoids', several other groupings are well supported: a clade comprising Agavaceae, Anthericaceae and Herreriaceae (BS 100 %), within which all Anthericaceae s. str. were clustered, Convallariaceae (BS 94 %), Hyacinthaceae (BS 80 %), and Alliaceae (BS 86 %). Indicated in the tree (Fig. 1) is a `higher asparagoid' clade that is present in all shortest trees, comprising families with successive microsporogenesis. Only Hypoxidaceae and Xanthorrhoeaceae, which also have successive microsporogenesis, fall in the lower asparagoids, otherwise all characterized by simultaneous microsporogenesis (Rudall et al., 1997). It would also be possible for the switch to successive microsporogenesis to have occurred earlier in the tree (open bar), such that a reversion to simultaneous microsporogenesis occurred instead at the branch just above Xanthorrhoea (open bar) and to successive (open bar) in Hemerocallis. Each hypothesis requires four steps. We favour the former explanation (see Discussion). All but one of these changes occur in clades consistent in all shortest trees, so that there is no distortion of patterns caused by optimization of this character on just one tree of the 39 900 retained. Within Asphodelaceae (BS 100 %), there is a wellsupported subclade (BS 96 %) comprising all genera of Alooideae. The members of this group appear closely related and little diverged. The only other group within Asphodelaceae that is well supported is Asphodelus and Asphodeline (BS 93 %). Analysis BÐrbcL of Asphodelaceae plus closest relatives In this rbcL matrix, 169 positions are variable, of which 136 are potentially informative. In the Fitch analysis (equal weights, unordered; Fitch, 1971), 651 optimal trees were found. The tree length was 615 steps with CI ˆ 0.63 and RI ˆ 0.75. SW identi®ed one shortest tree as optimal (Fig. 2); its SW length was 294.10 steps, CI ˆ 0.88 and RI ˆ 0.92 (Fitch length of this tree is 617 steps, two steps longer than the shortest Fitch trees, with a CI ˆ 0.63 and RI ˆ 0.75). Bootstrap percentages (calculated under the Fitch criterion) are indicated in parentheses below each branch; solid arrowheads indicate branches not present in the Fitch strict consensus tree. Although it may seem illogical to show Fitch bootstrap percentages on a SW tree, this is done to show which groups lack internal support under minimal weighting assumptions (equal weights). No groups with bootstrap percentages greater than 50 % are di€erently arranged in either the shortest Fitch or SW trees, so there is no inconsistency in showing Fitch bootstrap

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae

941

F I G . 1. One of more than 39 900 equally most parsimonious trees from analysis A (rbcL only), with 54 taxa including 14 genera of Asphodelaceae. Numbers of substitutions (ACCTRAN optimization) are indicated above each branch, and bootstrap percentages over 50 % are given in parentheses. Solid arrowheads indicate branches that are not present in the strict consensus tree, and bars show points at which changes in microsporogenesis type might have taken place (two scenarios: one with solid bars and the other with open bars; i.e. ACCTRAN and DELTRAN optimizations). Tree length is 1011 steps with CI ˆ 0.51 and RI ˆ 0.67. Family names are in accordance with the Angiosperm Phylogeny Group (1998).

942

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae

F I G . 2. Comparison of trnL-F and rbcL results for the B series of analyses. On the right is the single rbcL tree found with successive weighting (SW); its Fitch tree length is 617 steps with CI ˆ 0.63 and RI ˆ 0.75 (SW length is 294.10135 steps with CI ˆ 0.88 and RI ˆ 0.92). Numbers above the branches are lengths optimized under the Fitch criterion (ACCTRAN optimization), and those below the branches are Fitch bootstrap percentages over 50 %. Solid arrowheads indicate branches not present in the Fitch strict consensus tree of 651 equally most parsimonious trees (615 steps with CI ˆ 0.63 and RI ˆ 0.75). On the left is one of the three most parsimonious SW trees for the trnL-F matrix. Their Fitch tree length is 790 steps with CI ˆ 0.72 and RI ˆ 0.79 (SW length is 447.38863 steps with CI ˆ 0.91 and RI ˆ 0.94). Numbers on the branches are the same as for the rbcL portion. Open arrowheads indicate branches not found in all three shortest SW trees; solid arrowheads indicate branches not found in all six shortest Fitch trees (789 steps with CI ˆ 0.72 and RI ˆ 0.79). The two matrices produced highly similar results; the only di€erences being the positions of Ixiolirion and the sister group relationships of Bulbinella, Eremurus, Knipho®a and Trachyandra, all with bootstrap percentages of less than 50 %.

percentages on one SW tree. The two sets of trees (Fitch and SW) are highly similar in the groups they contain, as evidenced by the arrowheads marking inconsistent groups among the optimal trees. This reduced rbcL analysis contained another Xanthorrhoea species (X. minor R. Br.) and fewer, generally less distantly related outgroups. Nevertheless, it is still unclear whether Hemerocallidaceae or Xanthorrhoeaceae or both are sister to Asphodelaceae. The addition of an another species of Bulbine (B. wiesei L.I.Hall, also from Africa) and Jodrellia achieved slightly better resolution within Asphodelaceae. There is weak support (BS 68 %) for the clade containing Bulbine and Jodrellia being sister to Alooideae. There is stronger

support (BS 97 %) for the grouping of Asphodelus and Asphodeline. Alooideae now have BS 100 %, but within this group relationships are still unresolved in the Fitch analysis. Analysis BÐtrnL-F The length of the aligned trnL-F matrix for 34 taxa was 1145 bp after ambiguously aligned regions were excluded (the complete matrix was 1428 base pairs). Of these, 446 positions were variable, but only 213 were potentially informative. There were six shortest Fitch trees of 789 steps with a CI ˆ 0.72 and an RI ˆ 0.79; SW found three trees of 447.39 steps with CI ˆ 0.91 and RI ˆ 0.94 (Fitch length of

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae these trees was 790 steps, one step longer than the shortest Fitch trees, with a CI ˆ 0.72 and RI ˆ 0.79). One of these trees is illustrated in Fig. 2; branches not present in the SW strict consensus tree are marked with an open arrowhead, whereas those not present in the Fitch strict consensus tree are indicated by solid arrowheads. We also performed separate analyses of the trnL intron and the trnL-F intergene spacer (results not shown) to be certain that these two functionally distinct regions did not contain strongly supported (BS 85 % or greater), incongruent patterns, which they did not; in general, they produced much less resolution, as would be expected due to the smaller number of variable sites within each separate region. The overall patterns observed in the trnL-F trees are highly similar to those found with rbcL (Fig. 2), and the only patterns that con¯ict have less than 50 % bootstrap support and thus constitute `soft' incongruence (Seelanan et al., 1997), which is probably due to sampling error (too few variable sites). All strongly supported (as evaluated with the bootstrap) patterns found in one analysis agree with those in the other. The major di€erences are thus not in topology but in levels of support; for example, the sister group relationship of Jodrellia to Bulbine is well supported with trnL-F (BS 99 %) compared to only moderate support with rbcL (BS 83 %) and this pair's relationship to Alooideae (BS 93 % vs. 68 %). The exclusion of the wellsupported pair, Asphodeline and Asphodelus, is also well supported with trnL-F (BS 92 %) whereas with rbcL this relationship receives less than 50 % bootstrap. The same lack of a clear pattern concerning the relationships of Bulbinella, Eremurus, Knipho®a and Trachyandra occurs for trnL-F as well as for rbcL. To further examine the monophyly of Bulbine, which has species in both Africa and Australia, we sequenced one of the Australian species, B. semibarbata, which falls with moderate support (BS 84 %) as sister to B. succulenta Compton within a strongly supported Bulbine (BS 89 %). Analysis BÐcombined analysis of rbcL and trnL-F Because of the highly congruent separate analyses, we directly combined both matrices. Analysis of this matrix produced four Fitch trees of 1416 steps with CI ˆ 0.68 and CI ˆ 0.77. SW identi®ed one of these four trees as optimal (Fig. 3; i.e. it has the same Fitch length); this tree had 733.31 steps with a CI ˆ 0.90 and RI ˆ 0.93. Branches not present in the Fitch strict consensus tree are indicated by solid arrowheads. The general relationships found are again much like those of the highly similar separate analyses except that there is greater bootstrap support for groups previously supported, and some new groups receive minimal support for the ®rst time (e.g. the pairs Eremurus/Trachyandra, 65 %, and Bulbinella/Knipho®a, 76 %). In spite of the increased number of variable sites in the combined analysis, the sister-group relationship of Hemerocallidaceae to Asphodelaceae is found in only three of the four shortest Fitch trees as well as the single SW tree and should thus still be considered tenuous. Likewise, the

943

position of highly sequence divergent Ixiolirion Herb. does not receive a bootstrap percentage over 50 % (it falls in a di€erent position in the rbcL and trnL-F/combined trees; Figs 2, 3). Within Alooideae, the extremely low levels of sequence divergence continue to complicate estimates of relationships, although the pair Aloe/Lomatophyllum excluding the other four genera is strongly supported (100 %). Gasteria/Haworthia putatively share only a single change in the 2466 base pairs of plastid DNA analysed here (Fig. 3). Analysis CÐcombined data for Alooideae including gaps Excluding all the taxa more distantly related to Alooideae made it possible to align the whole trnL-F region, but this added only three additional potentially informative sites (104 vs. 101), so little was gained from this exercise, but it does demonstrate that being able to align the more length-variable regions for all taxa does not add a substantial amount of additional information. Fitch B&B analysis produced only one tree of 338 steps (only 166 steps without autapomorphies; there were a large number of gaps found in single taxa) with CI ˆ 0.83 (0.66 without autapomorphies) and RI ˆ 0.68 (Fig. 4). Although this matrix resulted in only one tree, it changed little except estimates of divergence. Most of the bootstrap percentages were also unchanged. Haworthia and Gasteria no longer form a pair, but there is only weak support (BS 60 %) for this. Indels indicating that Lomatophyllum is close to Aloe are abundant, as are those that separate the genera of Alooideae from Asphodeloideae, but the sequence data alone clearly supported these relationships. Most of the indels are autapomorphies with this level of sampling. Adding more species would probably make many of these synapomorphies, so using indels as characters has the possibility of adding a great deal of information to more extensively sampled studies within these genera, but it added little to this analysis. Patterns of molecular evolution We compared some aspects of molecular evolution on matrices containing only taxa for which both sequences were present (Table 2). Although the aligned matrices of both regions were identical in length (1428 bp), a large portion of this had to be excluded for trnL-F because of ambiguity in the alignment. Of the portion that could be aligned, 1145 bp, 165 (14 %) were potentially informative, whereas with rbcL, only 10 % of positions were potentially informative (136 of 1321 bp). Each variable site in rbcL changed on average 4.5 times (tree length, 615 steps, divided by the number of variable positions), whereas for trnL-F each variable site changed 4.7 times (773 steps, the tree length without B. semibarbata, divided by 165, the number of variable characters). Thus, the per site rate of evolution for the two regions is basically the same. The proteincoding gene, rbcL, has fewer variable sites overall (136 positions vs. 165), but they change at nearly the same rate as do the variable sites in trnL-F.

944

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae

F I G . 3. The single SW tree found in the C analysis of the combined matrices. This tree is one of the four produced by the Fitch analyses; their tree length is 1416 steps with CI ˆ 0.68 and RI ˆ 0.77. Numbers above the branches are Fitch lengths (ACCTRAN optimization), and those below the branches are bootstrap percentages over 50 %. Solid arrowheads indicate branches not present in the Fitch strict consensus tree of all four trees. The only taxa for which some uncertainty exists are the relative positions of Asphodelaceae, Hemerocallidaceae and Xanthorrhoeaceae and the same four genera that were problems in the separate analyses (Bulbinella, Eremurus, Knipho®a and Trachyandra).

Patterns of support are not much di€erent, although there are a few more supported groups with trnL-F, and these have slightly higher bootstrap percentages than those for rbcL. If we can assume that the combined tree is more accurate than either of the separate trees because of the

higher levels of support, then comparing each of these to the combined tree indicates that the trnL-F tree is the more accurate: the combined tree length is 1388 steps, of which 628 are contributed by rbcL and 777 by trnL-F; this means that rbcL analyses alone had 13 undetected steps (615 steps

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae

945

T A B L E 2. Comparison of molecular evolution for the three plastid regions analysed here rbcL

trnL-F (both regions, plus 30 trnL exon)

trnL intron

trnL-F spacer

No. variable/informative positions Tree length (steps)* CI RI Number of missed steps{

169/136 615 0.62 0.73 13

446/233 773{ 0.72 0.79 4

200/110 351 0.71 0.78 3

235/121 410 0.74 0.80 1

Steps/codon position{ CI/codon position{ RI/codon position{

1st 109 0.65 0.66

Number of steps{ CI{ RI{ ts : tv

ts 372 0.63 0.83 1.45

2nd 106 0.63 0.62

3rd 413 0.61 0.76

tv 256 0.61 0.64

ts 190 0.75 0.79 1.18

tv 161 0.68 0.78

ts 226 0.73 0.81 1.23

tv 184 0.73 0.77

*All statistics calculated from analyses in which Bulbine semibarbata was excluded so that comparable sets of taxa were treated for all three regions. {Twelve steps occurred in the trnL 30 exon, so the number for the trnL intron and trnL-F intergenic spacer do not add up to the total for the complete trnL-F region. {As assessed on the single tree found with SW on the combined matrix (Fig. 3). ts, Transition; tv, transversion.

F I G . 4. The single shortest tree found with rbcL and trnL-F by including complete length of trnL-F sequences due to the ability to align all regions within this restricted matrix and coding all gaps for use in the analysis. This tree has 338 steps (166 steps if autapomorphies are excluded) with a CI ˆ 0.83 (0.66 excluding autapomorphies) and RI ˆ 0.68. It di€ers from the combined tree shown in Fig. 3 only in the relative positions of Haworthia and Gasteria. Numbers above the branches are Fitch lengths (ACCTRAN optimization), and those below the branches are bootstrap percentages over 50 %.

when analysed alone vs. 628 on the combined analysis; i.e. rbcL had 2 % undetected substitutions) whereas trnL-F had only four such steps (only 0.5 %). It should be expected that a region with a higher percentage undetected homoplasy would provide a weaker overall result and would deviate more from the combined analysis. This was also the pattern observed by Chase and Cox (1998) in comparisons between rbcL, atpB, and 18S rDNA. This phenomenon is simply a result of weak patterns (too few variable sites), not of true incongruence between the patterns observed for the two matrices, and this sort of underestimate should always be anticipated for partitions of larger matrices. We would like

to distinguish between `di€erences' in tree topologies, the `soft' incongruence of Seelanan et al. (1997) and `incongruence', reserving the latter term to mean `strongly supported and incongruent'. If all of these plastid DNA sequence data can be assumed to be the result of the same evolutionary process, then the pattern underlying them is the same for all partitions (exon, intron and intergenic spacer) and the degree to which they deviate must therefore be due to simple sampling error, which should never produce high bootstrap percentages (90 % or more). Thus, the di€erences of topology observed here in the separate analyses are not considered evidence of matrix incongruence. In the context of plastid DNA (which is uniparentally inherited), strongly supported and incongruent patterns from partitions must be due to either di€erent functional constraints or horizontal transmission of some segments of the plastid genome (which seems the more unlikely explanation at this taxonomic level). The overall matrices for the three regions performed rather similarly as measured by RI: rbcL ˆ 0.73, trnL intron ˆ 0.78, trnL-F intergene spacer ˆ 0.77 (all based on ACCTRAN optimizations on the single SW tree illustrated from the combined analysis). This indicates that changes in all three had reasonable and similar levels of signal associated with them, so no relative matrix weighting appears appropriate. For protein coding genes such as rbcL, change is not expected to be evenly distributed at all three codon positions, and this was observed here: ®rst positions changed 109 times (17 %), second positions changed at least 106 times (17 %), whereas third positions changed the most, 413 times (66 %). However, the CI for each of these was rather similar: 0.65, 0.63 and 0.61, respectively, and the RI for third positions was highest: 0.66, 0.62 and 0.76. Thus third positions, at which the majority of substitutions occur

946

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae

and which weighting based on frequency would downweight relative to change at ®rst and second codon positions, performed the best. The other frequently weighted substitutions are transversions. By using the step matrix indicated in the Materials and Methods, we could separate the two types of substitutions occurring at each variable site. For rbcL, there was a 1.45 excess of transitions (Table 2), but CI and RI of transitions were higher than those for transversions: CI ˆ 0.63 and 0.61, RI ˆ 0.83 and 0.64 (transitions and transversions, respectively). A similar pattern was observed for the trnL-F intron/intergene spacer as well, although the relative frequencies of transitions and transversions were more similar as would be expected in non-coding plastid DNA (Morton, 1997), 1.18 and 1.23 for the intron and spacer, respectively. Again, the more frequent transitions performed better, although performance was not as dramatically di€erent as for rbcL, which might be expected since their frequencies are more similar: for the intron, CI ˆ 0.75 and 0.68, RI ˆ 0.80 and 0.78; for the spacer, CI ˆ 0.72 and 0.73, and RI ˆ 0.81 and 0.78.

sampling and an additional one or two sequence regions will be required). Because of the large data base of rbcL sequences available (nearly every family and nearly complete sets of genera for many families outside of the orchids in Asparagales, Dioscoreales, Liliales, and Pandanales; taxonomy of APG, 1998), an immense amount of focus can be gained by sequencing just a few taxa for rbcL (we added only 24 new rbcL sequences whereas for trnL-F we needed to add 34 taxa). Although rbcL is now an `old fashioned' locus, there is no reason to stop sequencing it (Chase and Albert, 1998), and subsequent studies using genes from all three genomes (Qiu et al., 1999) have corroborated patterns obtained ®rst with rbcL alone. Since most studies are now sequencing more than one region so that improved estimates are produced, it makes sense to include rbcL as one of those regions (Chase and Albert, 1998) because patterns of relationships and support derived from it are not signi®cantly worse than commonly sequenced non-coding regions, such as trnL-F. Weighting

DISCUSSION Molecular evolution Several authors (e.g. Taberlet et al., 1991) have suggested that for analyses at lower taxonomic levels non-coding regions of plastid DNA should be preferred to proteincoding genes such as rbcL because they evolve faster, presumably due to the lower levels of constraint imposed by function on non-coding DNA. The evidence for this has been based solely on pair-wise comparisons of the number of variable sites observed for rbcL and regions like trnL-F (e.g. Taberlet et al., 1991). The results of our study indicate that rates of change at variable sites di€er insigni®cantly (4.5 vs. 4.7 changes; see above). Although trnL-F does have slightly more variable sites than rbcL, the di€erence is not as great as would have been expected from the conclusions of Taberlet et al. (1991), and it is not rate of change that is of interest but rather number of characters (more characters providing more evidence). Even when we could align the total trnL-F region, the number of informative sites increased only slightly (101 to 104). On the whole, if we had to select one region over the other, we would favour trnL-F, but there is no overwhelming reason based on rates of change for choosing a non-coding region over rbcL. Although a somewhat greater number of characters can be obtained from non-coding regions, the amount of time spent in aligning these length-variable sequences is immense compared to length-conserved rbcL; in many respects, the bene®t of the slight increase in number of variable sites is more than compensated for by the amount of time spent in aligning trnL-F. An argument can be made that phylogenetic patterns are clearer for trnL-F than rbcL (based on the number and level of supported clades), but neither region alone provided a totally sucient result, so the best situation is to have both (in fact, to resolve robustly the phylogenetic patterns of several of these genera, more taxon

Hillis (1996, 1998) and others have observed that weighted parsimony performs better than Fitch parsimony, but the question has been on what basis should weights be assessed. Weighting has generally been based upon frequencies of change (Albert et al., 1993, among many), and this is why we wished to observe performance of each of the categories of change so that we could compare it with rates of change. The summary consistency index is not a useful measure of matrix performance (it is simply a measure of homoplasy and does not indicate distribution of homoplasy on the resulting trees; i.e. widely spaced homoplasy is not confounding as is closely spaced homoplasy). The summary retention index is thus a better measure of matrix performance because it informs us of the extent to which patterns of homoplasy are consistent with the tree. On the issue of how to weight to improve results with parsimony analyses, it is clear from the performance of the di€erent types of substitutions that frequency and performance are negatively correlated in the data presented here [and in those examined by Savolainen et al. (2000) for plastid rbcL and atpB across all seed plants and Richardson et al. (2000) for rbcL and trnL-F in Rhamnaceae and related families]. Hillis' simulations (1998) indicated the same thing: faster evolving sites perform better. Thus weighting based on frequency will have exactly the opposite of the desired e€ect because it gives more weight to the less reliable changes in every case. Weighting transitions via step matrices is a timeconsuming method of analysis, so the form of weighting performed here, successive approximations weighting (Farris, 1969), is a simpler though probably slightly less satisfactory alternative. A priori determination of appropriate weights is probably to be preferred (less chance of circularity), but if performance is to be the basis of weighting, then a posteriori weighting is a necessity. Given that the e€ect on trees selected with the coarser form of weighting each position is minimal (as demonstrated here;

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae

Systematic implications of the combined rbcL and trnL-F results Despite the misgivings stated by Dahlgren et al. (1985), these analyses demonstrate clearly that, apart from three genera, Asphodelaceae are a monophyletic group and quite distinct from other groups of asparagoid lilies, particularly Anthericaceae s. str. (see also Smith and van Wyk, 1998). Two families now appear to be their closest relatives: Hemerocallidaceae (sensu Rudall et al., 1997) and Xanthorrhoeaceae. Dahlgren et al. (1985) had already suggested that Phormiaceae, Hemerocallidaceae (which we now consider synonyms; APG, 1998) and perhaps Tecophilaeaceae (which they cited as Cyanastraceae; these are embedded in Tecophilaeaceae here and in Brummitt et al., 1998) were more closely related to Asphodelaceae than Anthericaceae. This view was based on both Hemerocallidaceae and Tecophilaeaceae having simultaneous microsporogenesis. Anthericaceae sensu Dahlgren et al. (1985) was grossly polyphyletic, and Chase et al. (1996) split these into several natural groups; some also have simultaneous microsporogenesis (e.g. Boryaceae) whereas others have successive (e.g. Anthericaceae s.s., Laxmanniaceae). Again we demonstrate here that microsporogenesis pattern divides the asparagoid lilies into two sets of families: a paraphyletic group that is mostly simultaneous and a monophyletic clade that is entirely successive (Fig. 1; the `higher asparagoids'). This was also found by Chase et al. (1995), Rudall et al. (1997) and Fay et al. (2000). The latter group includes Asparagaceae, Hyacinthaceae, Anthericaceae s.s., Alliaceae, Amaryllidaceae, Herreriaceae, Convallariaceae and Agavaceae. In the group that has mostly simultaneous

60 rbcL

50 40 30 20 Number of substitutions

SW selected one of the Fitch trees in the combined analysis), the e€ects of more ®nely adjusted weights would be negligible and therefore not necessary (employing a step matrix would also greatly slow down the analysis). Another relative measure of performance for a DNA region is the percentage of characters that change less than four times (each position can change up to three times without potentially any loss of pattern). In this regard, rbcL performs somewhat worse, as measured by the reduced tree lengths seen with SW: the SW tree for rbcL is a 52 % reduction in length (due to down-weighting of homoplasious characters) whereas trnL-F experienced only a 42 % reduction in tree length. This same thing can been seen in Fig. 5, in which 82 % of the potentially phylogenetically informative characters in trnL-F changed two±three times whereas for rbcL this was only 68 %. Successive approximations weighting [or the implied weighting criterion of Golobo€ (1997) which is another form of weighting based on an initial assessment of congruence] is thus much more sensitive to performance than any form of whole-category weights. The only caveat that we would attach to performance-based weighting is that reasonable levels of support should be present; if phylogenetic patterns are obscure, particularly if levels of sequence divergence are low (two±three substitutions per internal branch across many groups), then the estimates for performance will be jeopardized and probably unreliable.

947

10 0

1

2

3

4

5

6

7

8

9

80 70 trnL-F

60 50 40 30 20 10 0

1

2

3

4

5

6

7

Number of steps F I G . 5. The number of base positions relative to the number of substitutions. Note that trnL-F had more positions changing fewer times than rbcL.

microsporogenesis, the successive type is present several times (i.e. in Hypoxidaceae, Xanthorrhoea and some Iridaceae and Orchidaceae). Some of these are certainly reversals (those within Iridaceae and Orchidaceae), and we favour this interpretation in general rather than the reverse because the change from simultaneous to successive (as plotted on Fig. 1) is more frequent across the angiosperms. It is possible that the alternative scenario with open bars (Fig. 1) occurred, but it seems the less likely. Although molecular data clearly mark Asphodelaceae as a monophyletic group and the available morphological and cytological data generally agree with these ®ndings, it is dicult to ®nd any diagnostic characters to delimit Asphodelaceae (Smith and van Wyk, 1998). In many instances (i.e. Agavaceae, Asparagaceae and Hyacinthaceae; Table 1; Greilhuber, 1995), clear bimodal karyotypes exist, but yet other taxa form a graduated series, which demonstrates that clear designations are impossible, as demonstrated in Bulbine semibarbata (Table 1). This species was described

948

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae

by Watson (1986) as having ten larger pairs of chromosomes graded in size and three much smaller pairs. Greilhuber (1995) also discussed the problem of de®ning what constitutes a `bimodal' karyotype. It is clear that bimodal conditions have evolved independently many times in Asparagales and are obviously the end point of continua rather than a single distinguishable character state. Nonetheless all Alooideae have distinctive bimodal karyotypes based on x ˆ 7 with the same proportion of large to small chromosomes, which are nearly all acrocentric (one submetacentric pair), whereas Asphodeloideae have varying proportions of large to small chromosomes (see Table 1). Although in Bulbine, Eremurus and Trachyandra there are species with bimodal karyotypes of the same number (2n ˆ 14) as Alooideae, these karyotypes are di€erent (Table 1). Base chromosome numbers show a similar variety: x ˆ 7 uniformly for all Alooideae, whereas several basic numbers are found within Asphodeloideae, with the numbers varying from 2n ˆ 12 to 84. In both cytological aspects, several genera of Asphodeloideae are much more diverse than all genera of Alooideae. `Aloin' cells, as described earlier, are present in all Alooideae but absent in most Asphodeloideae. Similarly, anthraquinones have been reported as generally present in Asphodelaceae (Dahlgren et al., 1985), but are in fact absent in many of its genera (Aloe, Astroloba, Lomatophyllum, Asphodelus, Asphodeline and Bulbine; Beaumont et al., 1985) and present in Hemerocallidaceae and Xanthorrhoeaceae (Kite et al., 2000). The absence of steroidal saponins is shared with all other lower asparagoid families and is probably plesiomorphic. Microsporogenesis type is the same for the whole family but also for most of its close relatives (i.e. Simethis and Dianella), so none of these characters alone or perhaps even in combination is sucient to distinguish Asphodelaceae from other families of Asparagales. Secondary growth by means of a secondary thickening meristem mentioned by Dahlgren et al. (1985) as a characteristic of some but not all Asphodelaceae, is also present in Agavaceae, Convallariaceae, Iridaceae, Laxmanniaceae and Xanthorrhoeaceae. Although con®ned to Asparagales amongst monocotyledons, it has presumably evolved independently in most of these families (Rudall, 1995). Dahlgren et al. (1985) also described how sympodial growth pushes terminal in¯orescences into a pseudo-lateral position for all members of Alooideae. In Asphodeloideae, however, Bulbine is the only genus that clearly has this type of in¯orescence (Chase and de Bruijn, pers. obs.). All other genera are sympodial so this form of growth simply represents a shift of the origin of the next sympodium from the base to the apex of the previous sympodium. This feature has not been reported outside Asphodelaceae (Dahlgren et al., 1985) but must occur in other arborescent genera such as Yucca (Agavaceae) and Cordyline (Laxmanniaceae). Nearly all asparagoid lilies (like most monocots) are sympodial, and to become arborescent this pseudo-lateral position of in¯orescences is likely to be a requirement. The last morphological feature that we shall discuss here is a combination of seed characteristics that might be useful

to distinguish Asphodelaceae. These are: presence of an aril (also in Johnsonieae of Hemerocallidaceae); an endosperm with lipids and aleurone rather than starch; and an embryo that is three quarters of the length of the endosperm. No other family seems to have the same combination of characteristics (Dahlgren et al., 1985), and this would make them useful to delimit Asphodelaceae. However, these characteristics require more thorough investigation, and, except for the aril, they are not useful for ®eld identi®cation. The genus Trachyandra is representative of the problems in using macromorphological characters to delimit these two families; it is so like Chlorophytum that it has sometimes been included in this genus (Willis, 1973; Nordal et al., 1990). In our analyses, Trachyandra comes with Asphodelaceae, which was suspected by Dahlgren et al. (1985) due to its chemistry and microsporogenesis type. Nordal et al. (1990) studied the African species of Trachyandra and Chlorophytum (Anthericaceae s.s.) to ®nd ®eld characters but could not identify any. The gross similarity of many of these asparagoid families presents a paradox: if they are to be lumped into a single family to maintain monophyly then this family will also have to include readily distinguishable groups such as Orchidaceae, Xanthorrhoeaceae, Iridaceae and Tecophilaeaceae. However, if they are retained as separate families, then there are no readily observed traits by which they can be distinguished. We favour the narrower circumscriptions, but there will be many taxonomists who will not be pleased by the prospect of continuing to use families they can only know in the ®eld by means of ®rst knowing the genus or species and working back to the family. Resolution within Asphodelaceae was improved slightly in analysis B, which is probably due to better sampling. The di€erences in sequence divergence and position for the Bulbine species and Jodrellia con®rm the segregation of the latter from Bulbine. This segregation was originally based on di€erences in ¯oral morphology (Baijnath, 1976). Resolution among members of Alooideae was not substantially improved in analysis B or C, and the circumscription of these genera needs a thorough investigation. Astroloba has been considered by some authors as a synonym of Haworthia (Brummitt, 1992), but this does not seem likely based on these results in which it forms a strongly supported sister genus to Poellnitzia (Fig. 4). The generic groupings that we obtained are in agreement with the six informal sets recognized by Smith and van Wyk (1998): (1) Trachyandra only; (2) Asphodelus and Asphodeline; (3) Eremurus only; (4) Bulbine, Bulbinella, and Jodrellia; (5) Knipho®a only; and (6) Aloe, Astroloba, Gasteria, Haworthia, Lomatophyllum, Poellnitzia, and presumably Chortolirion (not sampled here). These authors were aware of the results presented here (they cited a published abstract; de Bruijn et al., 1995), so it cannot be stated that their conclusions and ours are independent. Phylogenetic reconstruction, tree shape and Asphodeloideae Huelsenbeck and Kirkpatrick (1996) described a bias of phylogenetic reconstruction methods towards tree estimates

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae that are more unbalanced than the `true' tree. This occurs especially when there are di€erences in divergence rates, as we found for Alooideae and Asphodeloideae, in which the low levels of nucleotide divergence within Alooideae ( for rbcL 0.30±1.67 %; Fig. 2) are in sharp contrast with that of Asphodeloideae (1.89±4.38 %; Fig. 2). The paraphyly of Asphodeloideae in our study could result from this tendency toward unbalanced trees. To evaluate this, a constraint analysis was carried out under the assumption of a monophyletic Asphodeloideae. The resulting single tree was ten steps less parsimonious, and there was a slight (0.1) decrease in both CI and RI. This is not much less parsimonious than the trees that were found without constraint, but it is opposed in the unconstrained analysis by a 92 % bootstrap excluding Asphodeline/Asphodelus from the clade composed of the rest of the family as well as a 96 % bootstrap that places Bulbine/Jodrellia as sister to Alooideae. In addition (as discussed above), there are no characters that mark a monophyletic Asphodeloideae. In fact, our review of morphological and cytological data shows a similar pattern to the sequence results in which Asphodeloideae are much more divergent and heterogeneous than Alooideae; this could be seen as an indication that, rather than being recognized as a subfamily, Alooideae might better be considered a single genus (and monophyly of these genera has not yet been con®rmed by phylogenetic analysis of any type of data). To summarize, Asphodelaceae have been found to be a monophyletic group with two families, Hemerocallidaceae and Xanthorrhoeaceae, as their closest relatives and a distant relationship to the morphologically similar Anthericaceae. Support for two subfamilies is lacking, but there is strong support for Bulbine/Jodrellia alone as sister to the genera of Alooideae. Asparagales show a clear division based on microsporogenesis type: a monophyletic group with successive microsporogenesis and a paraphyletic grade with simultaneous microsporogenesis, within which reversal to the successive type has occurred at least three times. Further study of all these families will be required to identify synapomorphies (see Fay et al., 2000, for a discussion of this topic), but the level of sequence divergence and the pattern of relationship support the continued recognition of these groups at the family level (Dahlgren et al., 1985) rather than lumping them all, including Orchidaceae and Iridaceae, into a single unit. The basic framework of families laid out by Dahlgren et al. (1985) and supported by the results of Chase et al. (1995), Rudall et al. (1997) and Fay et al. (2000) as well as here, warrants continued recognition of Asphodelaceae and Anthericaceae despite the fact that these families will not be easy to distinguish in the ®eld. AC K N OW L E D GE M E N T S The authors would like to thank Tony Hall of the Alpine Section, Clive Foster of the Jodrell Glass and their co-workers at Royal Botanic Gardens, Kew, for growing much of the material used and Himansu (Snowy) Baijnath for providing material of Jodrellia. Material of outgroup taxa was provided by the Royal Botanic Gardens, Sydney,

949

Christopher Quinn and Paul Gadek (University of New South Wales, Australia) and Peter Goldblatt (Missouri Botanical Garden, St. Louis, USA).

L I T E R AT U R E C I T E D Albert VA, Chase MW, Mishler BA. 1993. Character-state weighting for cladistic analysis of protein-coding DNA sequences. Annals of the Missouri Botanical Garden 80: 752±766. Angiosperm Phylogeny Group. 1998. An ordinal classi®cation for the families of ¯owering plants. Annals of the Missouri Botanical Garden 85: 531±553. Baijnath H. 1976. Jodrellia, a new genus of Liliaceae from tropical Africa. Kew Bulletin 32: 571±578. Beaumont J, Cutler DF, Reynolds T, Vaughan JG. 1985. The secretory tissue of aloes and their allies. Israel Journal of Botany 34: 265±282. Brandham PE. 1969. Inversion heterozygosity and sub-chromatid exchange in Agave sensu stricta. Chromosoma (Berlin) 26: 270±286. Brandham PE. 1971. The chromosomes of the Liliaceae II. Kew Bulletin 25: 381±399. Brandham PE. 1973. The chromosomes of the Liliaceae III. Kew Bulletin 28: 341±348. Briggs BG. 1966. Chromosome numbers of some Australian monocotyledons. Contributions from the New South Wales National Herbarium 4: 24±34. Bruijn A de, Cox AV, Chase MW. 1995. Molecular systematics of Asphodelaceae (Asparagales: Lilianae) [abstract]. American Journal of Botany 82: 124. Brummitt RK. 1992. Vascular plant families and genera. Richmond: Royal Botanic Gardens, Kew. Brummitt RK, Banks H, Johnson MAT, Doherty KA, Jones K, Chase MW, Rudall PJ. 1998. Taxonomy of Cyanastroideae (Tecophilaeaceae): a multidisciplinary approach. Kew Bulletin 53: 769±803. Bruyns P, Vosa CG. 1987. Taxonomic and cytological notes on Bowiea Hook. F. and allied genera (Liliaceae). Caryologia 40: 287±297. Cave MS. 1955. Sporogenesis and the female gametophyte of Phormium tenax. Phytomorphology 5: 247±253. Cave MS. 1964. Cytological observations on some genera of the Agavaceae. MadronÄo 17: 163±169. Cave MS. 1970. Chromosomes of California Liliaceae. University of California Publications in Botany 57: 1±58. Chase MW, Albert VA. 1998. A perspective on the contribution of plastid rbcL DNA sequences to angiosperm phylogenetics. In: Soltis DE, Soltis PS, Doyle JJ, eds. Molecular systematics of plants II: DNA sequencing. Boston: Kluwer Academic Publishers, 488±507. Chase MW, Cox AV. 1998. Gene sequences, collaboration, and analysis of large data sets. Australian Systematic Botany 11: 215±229. Chase MW, Rudall PJ, Conran JG. 1996. New circumscriptions and a new family of asparagoid lilies: genera formerly included in Anthericaceae. Kew Bulletin 51: 667±680. Chase MW, Duvall MR, Hills HG, Conran JG, Cox AV, Eguiarte LE, Hartwell J, Fay MF, Caddick LR, Cameron KM, Hoot S. 1995. Molecular phylogenetics of Lilianae. In: Rudall PJ, Cribb PJ, Cutler DF, Humphries CJ, eds. Monocotyledons: systematics and evolution. Richmond: Royal Botanic Gardens, Kew, 109±137. Chase MW, Soltis DE, Soltis PS, Rudall PJ, Fay MF, Hahn WH, Sullivan S, Joseph J, Molvray M, Kores PJ, Givnish TJ, Sytsma KJ, Pires JC. 2000. Higher-level systematics of the monocotyledons: an assessment of current knowledge and a new classi®cation. In: Wilson KL, Morrison DA, eds. Monocots: systematics and evolution. Collingwood, Victoria, Australia: CSIRO Publishing, 3±16. Chase MW, Soltis DE, Olmstead RG, and 39 others. 1993. Phylogenetics of seed plants: an analysis of nucleotide sequences from the plastid gene rbcL. Annals of the Missouri Botanical Garden 80: 528±580.

950

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae

Chinappa CC, Semple JC. 1976. Chromosomal aberrations in a diploid Haworthia subfasciata. Canadian Journal of Genetics and Cytology 18: 281±286. Cox AV. 1997. PaupGap version 1.0: program and documentation. Richmond: Royal Botanic Gardens, Kew. Cronquist A. 1981. An integrated system of classi®cation of ¯owering plants. New York: Columbia University Press. Dahlgren RMT, Cli€ord HT, Yeo PF. 1985. The families of the monocotyledons, structure, evolution and taxonomy. Berlin: Springer-Verlag. Dõ az Lifante Z. 1996. A karyological study of Asphodelus L. (Asphodelaceae) from the western Mediterranean. Botanical Journal of the Linnean Society 121: 285±344. Di Fulvio TS, Cave MS. 1964. Embryology of Blandfordia nobilis Smith (Liliaceae), with special reference to its taxonomic position. Phytomorphology 14: 487±489. Doyle JJ, Doyle JL. 1987. A rapid DNA isolation procedure for small quantities of leaf tissue. Phytochemical Bulletin 19: 11±15. Duvall MR, Learn GH Jr, Eguiarte LE, Clegg MT. 1993a. Phylogenetic analysis of rbcL sequences identi®es Acorus calamus as the primal extant monocotyledon. Proceedings of the National Academy of Sciences, USA 90: 4641±4644. Duvall MR, Clegg MT, Chase MW, Clark WD, Kress WJ, Hills HG, Eguiarte LE, Smith JF, Gaut BS, Zimmer EA, Learn GH Jr. 1993b. Phylogenetic hypotheses for the monocotyledons constructed from rbcL sequence data. Annals of the Missouri Botanical Garden 80: 607±619. Farris JS. 1969. A successive approximations weighting approach to character weighting. Systematic Zoology 18: 374±385. Fay MF, Chase MW. 1996. Resurrection of Themidaceae for the Brodiaea alliance and recircumscription of Alliaceae, Amaryllidaceae and Agapanthoideae. Taxon 45: 441±451. Fay MF, Rudall PJ, Sullivan S, Stobart KL, de Bruijn A, QamaruzZaman F, Hong W-P, Joseph J, Hahn WJ, Conran JG, Chase MW. 2000. Phylogenetic studies of Asparagales based on four plastid DNA loci. In: Wilson KL, Morrison DA, eds. Monocots: systematics and evolution. Collingwood, Victoria, Australia: CSIRO Publishing, 360±371. Felsenstein J. 1985. Con®dence limits on phylogenetics: an approach using the bootstrap. Evolution 39: 783±791. Fernandez A, DavinÄa JR. 1971. Heterochromatin and genome size in Fortunatia and Camassia (Hyacinthaceae). Kew Bulletin 46: 307±316. Fitch WM. 1971. Toward de®ning the course of evolution: minimal change for a speci®c tree topology. Systematic Zoology 20: 406±416. Goldblatt P. 1982. Chromosome cytology in relation to suprageneric systematics of neotropical Iridaceae. Systematic Botany 7: 186±198. Golobo€ PA. 1997. Self-weighted optimization: tree searches and character state reconstructions under implied transformation costs. Cladistics 12: 225±245. Greilhuber J. 1995. Chromosomes of the monocotyledons (general aspects). In: Rudall PJ, Cribb PJ, Cutler DF, Humphries CJ, eds. Monocotyledons: systematics and evolution. Richmond: Royal Botanic Gardens, Kew, 379±414. Hara H. 1969. Variation in Paris polyphylla Smith. with reference to other Asiatic species. Journal of the Faculty of Science, University of Tokyo, Section III, Botany 10: 141±180. Hillis DM. 1996. Inferring complex phylogenies. Nature 383: 130±131. Hillis DM. 1998. Taxonomic sampling, phylogenetic accuracy, and investigator bias. Systematic Biology 47: 3±8. Huelsenbeck JP, Kirkpatrick M. 1996. Do phylogenetic methods produce trees with biased shapes?. Evolution 50: 1418±1424. Jin X. 1986. The chromosomes of Hemerocallis (Liliaceae). Kew Bulletin 41: 379±391. Johnson MAT, Gale RMO. 1983. Observations on the leaf-anatomy, pollen, cytology and propagation of Calibanus hookeri (Lem.) Trelease. Bradleya 1: 25±32. Kaneko K. 1970. Cytological studies on some species of Hosta VI. Karyotypes of H. undulata. H. undulata var. erromena, H. undulata var. albomarginata, H. tardiva and H. sacra. Botanical Magazine (Tokyo Botanical Society) 83: 27±35.

Kativa S. 1994. Anthericaceae in Zimbabwe: a study on cytology and reproduction. In: Seyani JH, Chikuni AC, eds. Proceedings of the XIIIth Plenary Meeting of AETFAT, vol 1. Zomba, Malawi: National Herbarium and Botanic Gardens of Malawi, 525±534. Karavokyrou E, Tzanoudakis D. 1991. The genus Allium in Greece: II. A cytotaxonomical study of the E. Aegean species. Botanika Chronika (Patras) 10: 777±784. Karihaloo JL, Koul AK. 1984. Bearing of cytology on the systematic position of Ixiolirion (Fischer) Herb. (Amaryllidaceae). Cytologia 49: 39±47. Kite GC, Grayer RJ, Rudall PJ, Simmonds SJ. 2000. The potential for chemical characters in monocotyledon systematics. In: Wilson KL, Morrison DA, eds. Monocots: systematics and evolution. Collingwood, Victoria, Australia: CSIRO Publishing, 101±113. Locatelli-Lanzara P, Marchi P, Moskova R. 1971. Simethis mattiazzi (Vand.) Saccardo s. l. Informatore Botanico Italiano 3: 149±151. Maddison WP, Maddison DR. 1992. MacClade: analysis of phylogeny and character evolution, Version 3.06. Sunderland, MA: Sinauer Associates. Malakhova LA. 1990. Kariologocheskij analiz prirodnykh populjacij redkich i ischezajushchikh rastenij na juge Tomskoj Oblasti. Bjulleten' Glavnogo Botaniceskogo Sada 155: 60±66. Mitra K. 1966. Karyological studies in Curculigo orchioides. Science and Culture 32: 201±202. Morton BR. 1997. The in¯uence of neighbouring base composition on substitutions in plant chloroplast coding sequences. Molecular Biology and Evolution 14: 189±194. Muasya AM, Simpson DA, Culham A, Chase MW. 1998. An assessment of suprageneric phylogeny in Cyperaceae using rbcL DNA sequences. Plant Systematics and Evolution 211: 257±271. Naranjo CA. 1975. Chromosome studies in Hypoxis L. (Hypoxidaceae) I. Karyotype of H. decumbens L (Horn). Phyton 33: 45±49. Newman IV. 1929. Life history of Doryanthes excelsa. II. The gametophytes, seed production, chromosome number and general conclusion. Proceedings of the Linnaean Society of New South Wales 54: 411±435. Nordal I, Eriksen TE, Fosby M. 1990. Studies on the generic delimitation of Anthericaceae. Proceedings of the XIIth Plenary Meeting of AETFAT. Hamburg: Mitteilungen aus dem Institut fuÈr allgemeine Botanik in Hamburg: 23B, 535±559. Okada H. 1988. Karyomorphological observations of Apostasia nuda and Neuwidia veratrifolia (Apostasioideae, Orchidaceae). Japanese Journal of Botany 63: 344±350. Olmstead RG, Michaels HJ, Scott KM, Palmer JD. 1992. Monophyly of the Asteridae and identi®cation of their major lineages inferred from DNA sequences of rbcL. Annals of the Missouri Botanical Garden 79: 249±265. Qiu Y-L, Lee J, Bernasconi-Quadroni F, Soltis DE, Soltis PS, Zanis M, Chen Z, Savolainen V, Chase MW. 1999. The earliest angiosperms: evidence from mitochondrial, plastid and nuclear genomes. Nature 402: 404±407. Richardson JE, Fay MF, Cronk QCB, Bowman D, Chase MW. 2000. A molecular analysis of Rhamnaceae using rbcL and trnL-F plastid DNA sequences. American Journal of Botany 87: 1309±1324. Rudall PJ. 1995. New records of secondary thickening in monocotyledons. IAWA Journal 16: 261±268. Rudall PJ, Engleman EM, Hanson L, Chase MW. 1998. Systematics of Hemiphylacus, Anemarrhena and Asparagaceae. Plant Systematics and Evolution 211: 181±199. Rudall PJ, Furness CA, Chase MW, Fay MF. 1997. Microsporogenesis and pollen sulcus type in Asparagales (Lilianae). Canadian Journal of Botany 75: 408±430. Saiki RK, Gelfand DH, Sto€el S, Scharf SJ, Higuichi R, Horn GT, Mullis KM, Erlich HA. 1987. Primer-directed enzymatic ampli®cation of DNA with a thermostable DNA polymerase. Science 239: 487±491. Sanger F, Nicklen S, Coulson AR. 1977. DNA sequencing with chainterminating inhibitors. Proceedings of the National Academy of Sciences, USA 74: 5463±5467. Satoà D. 1942. Karyotype alteration and phylogeny in Liliaceae and allied families. Japanese Journal of Botany 7: 57±161. Savolainen V, Chase MW, Hoot SB, Morton CM, Soltis DE, Bayer C, Fay MF, de Bruijn AY, Sullivan S, Qiu Y-L. 2000. Phylogenetics of

Chase et al.ÐMolecular Phylogenetics of Asphodelaceae ¯owering plants based upon a combined analysis of plastid atpB and rbcL gene sequences. Systematic Biology 49: 306±362. Schulze W. 1975. BeitraÈge zur Taxonomie der Lilii¯oren. I. Asphodelaceae. Wissenschaftliche Zeitschrift der Friedrich-SchillerUniversitaÈt Jena/ThuÈringen 24: 403±415. Seelanan T, Schnabel A, Wendel J. 1997. Congruence and consensus in the cotton tribe (Malvaceae). Systematic Botany 22: 259±290. Smith GF. 1991. The chromosomes of Chortilirion and Poelnitzia (Asphodelaceae: Alooideae). Bothalia 21: 171±175. Smith GF, van Wyk B-E. 1998. Asphodelaceae. In: Kubitzki K, ed. The families and genera of vascular plants. Berlin: Springer-Verlag, 30±140. Soltis DE, Hibsch-Jetter C, Soltis PS, Chase MW, Farris JS. 1997. Molecular phylogenetic relationships among angiosperms: an overview based on rbcL and 18S rDNA sequences. In: Iwatsuki K, Raven PH, eds. Evolution and diversi®cation of land plants. Tokyo: Springer-Verlag, 157±178. Stedje B, Nordal I. 1994. A contribution to the discussion of the family delimitation of Anthericaceae versus Asphodelaceae. In: Seyani JH, Chikuni AC, eds. Proceedings of the XIIIth Plenary Meeting of AETFAT, vol 1. Zomba, Malawi: National Herbarium and Botanic Gardens of Malawi, 513±524. Sveshnikova LI, Zemskova EA. 1988. Chromosome numbers in some members of the Amaryllidaceae. Botaniceskij Zurnal (Moscow and Leningrad) 73: 1207±1208. Swo€ord DL. 1998. PAUP*: Phylogenetic analysis using parsimony, version 4.0d65. Sunderland, MA: Sinauer Associates. Taberlet P, Gielly L, Pautou G, Bouvet J. 1991. Universal primers for ampli®cation of three non-coding regions of chloroplast DNA. Plant Molecular Biology 17: 1105±1109. Tamura MN. 1995. A karyological review of the orders Asparagales and Liliales (Monocotyledonae). Feddes Repertorium 106: 83±111.

951

Tanaka NY. 1981. Studies on chromosome arrangement in some higher plants. I. Interphase chromosomes in three liliaceous plants. Cytologia 46: 343±357. Therman E. 1953. Chromosomal evolution in the genus Polygonatum. Hereditas 39: 277±288. Tzanoudakis D, Kypriotakis Z. 1987. Chromosome studies in the Greek ¯ora III. Karyotypes of eight Aegean species. Botanica Helvetica 97: 229±237. Vijayavalli B, Mathew PM. 1990. Cytotaxonomy of the Liliaceae and allied families. Kerala, India: Continental Publishers. Vosa CG, Bennett ST. 1990. Chromosome studies in the southern African ¯ora. 58-94. Chromosome evolution in the genus Gasteria Duval. Caryologia 43: 235±247. Watson EM. 1986. Cytoevolutionary studies in the genus Bulbine Wolf (Liliaceae). II. The Australian annual taxa (B. semibarbata s.l.). Australian Journal of Botany 34: 505±522. Wetschnig W. 1988. Chromosomenzahlen. KaÈrntner Gefassplanzen (Teil 1). Carinthia II 178: 391±401. Whittaker TW. 1934. Chromosome constitution in certain monocotyledons. Journal of the Arnold Arboretum 15: 135±153. Willis JC. 1973. A dictionary of the ¯owering plants and ferns. Cambridge: Cambridge University Press. Yokota M. 1987. Karyotypes and phylogeny in Orchidinae and allied subtribes. In: Saito K, Tanaka R. Proceedings of the 12th world orchid conference. Tokyo: 12th World Orchid Conference Organizing Committee, 70±79. Zakirova RO, Nafanailova II. 1990. Chromosome numbers in the members of some families of vascular plants in the ¯ora of Kazakhstan. Botaniceskij Zurnal (Moscow and Leningrad) 75: 438±439. Zimudzi C. 1994. The cytology and reproduction of the genus Hypoxis L. In: Seyani JH, Chikuni AC, eds. Proceedings of the XIIIth Plenary Meeting of AETFAT, vol 1. Zomba, Malawi: National Herbarium and Botanic Gardens of Malawi, 535±543.