| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
3University of Florida, Fort Lauderdale Research and Education Center, 3205 College Avenue,Fort Lauderdale, Florida 33314; 4Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3DS, UK; 5University of Florida, Department of Environmental Horticulture, 1545 Fifield Hall, Gainesville, Florida 32611
Received for publication August 10, 1998. Accepted for publication January 14, 1999.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: Amaryllidaceae • cladistic analysis • molecular systematics • moncotyledons • phylogeny • plastid DNA
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
), the 51 genera of which have been variously classified since as liliaceous or amaryllidaceous. This basic dichotomy represents the generally uncertain phylogenetic placement of many petaloid monocots until the past two decades. Seven of the 51 genera that Linneaus placed in Hexandria monogynia have since been included within a common taxonomic unit, as section Narcissi (Adanson, 1763
; de Jussieu, 1789
), family Amaryllideae (Jaume-St.-Hilaire, 1805
; Brown, 1810
), order Amaryllidaceae (Lindley, 1836
; Herbert, 1837
), tribe Amarylleae (Bentham and Hooker, 1883
), suborder Amarylleae (Baker, 1888
), subfamily Amaryllidoideae (Pax, 1888
); family Amaryllidaceae (Hutchinson, 1934
, 1959
), and subfamily Amarylloideae (Traub, 1963
). Brown (1810)
was the first to propose that the genera with superior ovaries be excluded from Amaryllidaceae, a restriction followed faithfully until Hutchinson (1934)
. Herbert (1837)
recognized that the Taccaceae were not allied to Amaryllidaceae, and Pax (1888)
formally removed Velloziaceae as part of the family (Herbert's suborder Xerophyteae). Hutchinson's (1934,
1959
) classification was the first radical recircumscription of Amaryllidaceae since Brown (1810)
. In defining the unifying character of the family to be "an umbellate inflorescence subtended by an involucre of one or more spathaceous bracts," he segregated Agavaceae, Hypoxidaceae, and Alstroemeriaceae and added tribes Agapantheae, Allieae, and Gilliesieae (Alliaceae). Takhtajan (1969)
recognized Amaryllidaceae in the narrowest sense, and maintained a distinct Alliaceae. Cronquist (1988)
and Thorne (1976)
included Amaryllidaceae within broad concepts of Liliaceae.
Concepts of familial and ordinal limits of the monocotyledons were radically challenged by Huber (1969)
, who emphasized less conspicuous characters, particularly embryological characters, over gross floral or vegetative morphology. Huber's work highlighted the heterogeneity present in many traditional monocot families, especially Liliaceae. Much of this work was refined and placed into phylogenetic context by Dahlgren and coworkers (Dahlgren and Clifford, 1982
; Dahlgren and Rasmussen, 1983
; Dahlgren, Clifford, and Yeo, 1985
). In Dahlgren, Clifford, and Yeo's (1985)
synthesis, Amaryllidaceae and Alliaceae are both recognized as members of the order Asparagales, an order of 31 families that have evolved many traits in parallel with Liliales. One of the most important and consistent characters separating these two orders is the presence of phytomelan in the seed coat of Asparagales (Huber, 1969
). To date, phylogenetic analyses of the monocotyledons, based on both morphological and gene sequence matrices, have supported this classification with some amendment (Duvall et al., 1993
; Stevenson and Loconte, 1995
; Chase et al., 1995a,
b
), but the precise relationship of Amaryllidaceae to other Asparagales remained elusive until Fay and Chase (1996)
used molecular data to argue that Amaryllidaceae, Agapanthaceae, and Alliaceae form a monophyletic group and that together they are related most closely to Hyacinthaceae s.s. and the resurrected family Themidaceae (the former tribe Brodiaeeae of Alliaceae).
Despite a lack of consensus on generic limits and tribal delimitations within the Amaryllidaceae, cladistic analysis has only rarely been applied to problems in the family, such as by Nordal and Duncan (1984)
for Haemanthus and Scadoxus, two closely related, baccate-fruited African genera, Meerow (1987a
, 1989
) for Eucrosia and Eucharis and Caliphruria, respectively, and Snijman (1994)
and Snijman and Linder (1996)
for various taxa of tribe Amaryllideae. Applying phylogenetic studies for the entire family is difficult due to homoplasy for many conspicuous characters within this highly canalized group (Meerow, 1987a
, 1989
, 1995
). This led Meerow (1995)
to conclude that "future reconstruction attempts will greatly benefit from the inclusion of molecular data."
The four most recent infrafamilial classifications (Table 1) of Amaryllidaceae are those of Traub (1963)
, Dahlgren, Clifford, and Yeo (1985)
, Müller-Doblies and Müller-Doblies (1996)
, and Meerow and Snijman (1998)
. Traub's scheme included Alliaceae, Hemerocallidaceae, and Ixioliriaceae as subfamilies, following Hutchinson (1934
, 1959)
in part. Within his subfamily Amarylloideae, he erected two informal taxa, "infrafamilies" Amarylloidinae and Pancratioidinae, both of which were polyphyletic (Meerow, 1995
). Dahlgren, Clifford, and Yeo (1985)
dispensed with any subfamilial classification above the level of tribe, recognizing eight, and treated as Amaryllidaceae only those genera in Traub's Amarylloideae. Stenomesseae and Eustephieae were combined. Meerow (1995)
resurrected Eustephieae from Stenomesseae and suggested that two new tribes may need to be recognized, Calostemmateae and Hymenocallideae. Müller-Doblies and Müller-Doblies (1996)
recognized ten tribes (among them Calostemmateae) and 19 subtribes, many of them monogeneric; Meerow and Snijman (1998)
recognize 14 tribes, with two subtribes only in one of them (Table 1).
|
recircumscribed Amaryllidaceae to include Agapanthus, previously included in Alliaceae, as subfamily Agapanthoideae. This recircumscription was based on phylogenetic analysis of rbcL sequence data, with only four genera of Amaryllidaceae s.s. included in the analysis. All the epigynous genera were treated as Amaryllidoideae. Bootstrap support for this treatment was weak (63%). The sampling within Amaryllidaceae s.s. in Fay and Chase (1996)
did not allow sufficient resolution of the generic relationships within the family, and we present here phylogenetic analyses of three plastid DNA sequence data sets for a much wider range of taxa.
The phylogenetic application of sequences of rbcL is well documented (e.g., Chase et al., 1993
; Olmstead and Palmer, 1994) and has been used to clarify relationships between and within a number of asparagoid families, including Themidaceae (Fay and Chase, 1996
), Asphodelaceae (de Bruijn et al., unpublished data), Alliaceae (Fay et al., unpublished data), and Orchidaceae (Cameron et al., 1999
). Within Amaryllidaceae, however, levels of resolution obtained within some major clades, particularly those from the Neotropics, were not sufficient to elucidate tribal relationships fully (Fay et al., 1995
). For this reason, we chose to combine our rbcL matrix with two for the trnL intron/trnL-F spacer region of noncoding plastid DNA, for which Taberlet et al. (1991)
had developed "universal" primers for amplification. Sequences of this region have been used in phylogenetic studies of Crassulaceae (Kim, t'Hart, and Mes, 1996
; Mes, Van Brederode and t'Hart, 1996
; Mes, Wijers, and t'Hart, 1997
), Gentianaceae (Gielly and Taberlet, 1996
; Gielly et al., 1996
), Paeoniaceae (Sang, Crawford and Stuessy, 1997
), Proteaceae (Maguire et al., 1997
), Ranunculaceae (Kita, Ueda, and Kadota, 1995
), among others, either alone or in combination with other loci. This region of the plastid genome evolves more than three times faster, on average, than rbcL (Gielly and Taberlet, 1994
) and can therefore potentially add increased resolution to a phylogeny generated by rbcL sequences.
Combining independent character matrices, whether both molecular or molecular and morphological, very often increases the resolution of the ingroup and the bootstrap support of the internal nodes of the phylogenetic trees (Chase et al., 1995b
; Olmstead and Sweere, 1994
; Rudall et al., 1998; Soltis et al., 1998
). In this paper we present the first family-wide phylogenetic analysis of Amaryllidaceae using three plastid DNA sequences, alone and in combination, and comment on the evolutionary and bigeographic implications of the results.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
RBG Kew
DNA was extracted from 1.0 g fresh, 0.2–0.25 g silica gel-dried leaves, or 
0.1 g material from herbarium sheets using the 2X CTAB method of Doyle and Doyle (1987)
. All samples were then purified on cesium chloride/ethidium bromide gradients (1.55 g/mL density). Gene amplification of the rbcL gene was carried out using forward primers that match the first 20 or 26 base pairs (bp) of the coding region and reverse primers that correspond to 20-bp sequences that begin at position 1352 or 1367 in the coding region (Table 3; Chase and al., 1995a
). The trnL-trnF region was amplified using the c and f primers of Taberlet et al. (1991)
. Amplified products were purified using Magic mini columns (Promega, Madison, Wisconsin) or QIAquick (Qiagen, Valencia, California) columns, following manufacturers protocols. Standard dideoxy methods or modified dideoxy cycle sequencing with dye terminators run on an ABI 373A or 377 automated sequencer (according to the manufacturer's protocols; Applied Biosystems, Inc., Foster City, California) were used to sequence the amplification products directly. For rbcL, both strands were sequenced for 70–90% of the exon. We have 1320 bp of rbcL sequence data for most taxa. The trnL-F region is length variable; we sequenced both strands for 70–90% of the region, obtaining between 750 and 900 bp of sequence data for most taxa.
|
). Polymerase chain reaction (PCR) protocols were those of Li and Guy (1996)
. The PCR reaction product was fractionated by electrophoresis on a 0.8% low melting point agarose gel, then the DNA was purified from gel slices with phenol and chloroform and dissolved in 10 µL of 10 mmol/L Tris (pH 8.0) buffer. A 10-µL ligation reaction was prepared containing 25 ng of pGEM-T® vector (Promega, Madison, Wisconsin), 50–100 ng PCR products, 1 µL of 10x ligase buffer [300 mmol/L Tris-HCl, pH 7.8, 100 mmol/L MgCl2, 100 mmol/L dithiothreitol, 5 mmol/L ATP, and 1.5 U T4 DNA ligase (Promega, Madison, Wisconsin)]. The ligations were incubated at 16°C overnight. One hundred microlitres of competent E. coli XL-1 Blue cells were transformed with 2.5 µL of each ligation mixture, and spread on a Luria-Bertani (LB) agar plate (100 x 15 mm) containing 50 µg/mL of ampicillin and 12.5 µL tetracycline. The plate was spread with 50 µL of 2% X-gal and 50 µL of 100 mmol/L isopropyl-beta-D-thiogalactopyranoside before using. The plate was incubated at 37°C overnight. Individual colonies were counted, and white colonies were selected to grow overnight at 37°C in LB media containing 50 µg/mL ampicillin. Plasmid DNA was digested with Pst I and Sst II, and the restricted DNAs were fractionated on a 0.8% agarose gel to verify the presence of the cloned insert. Plasmid DNA containing the cloned, amplified insert was purified, and DNA sequencing was accomplished using the Taq DyeDeoxyTM Terminator Cycle Sequencing Kit (Applied Biosystems Inc., Foster City, California) on an automated sequencer (Applied Biosystems Inc., Foster City, California) by the DNA sequencing Core of the Interdisplinary Center for Biotechnology Research at the University of Florida. Sequencing was accomplished using vector T7 and Sp6 primers along with specific primers for the rbcL gene received from G. Zurawski (Table 3). The complete sequence of both strands was determined using sets of synthetic primers (Table 3).
Sequence alignment
Sequences of rbcL were easily aligned manually because no length variation was detected. For trnL-F, two methods were employed. Sequences of several taxa representing the range of probable variation in the matrix were aligned using the Clustal option in Sequence Navigator (Applied Biosystems, Inc.), followed by manual optimisation and alignment of subsequent sequences. Alternatively, the program Sequencher (Gene Codes, Inc., Ann Arbor, Michigan) was used to align sequences of closely related taxa with subsequent builds of these smaller alignments performed manually. Copies of the aligned matrices are available from the senior author.
Analysis
Aligned matrices were analyzed using the parsimony algorithm of the software package PAUP* for Macintosh (v4.0 d59-64, Swofford, 1998
) with a successive weighting (SW; Farris, 1969
) strategy. SW was employed to globally reduce the effect of highly homoplasious base positions on the resulting topologies (Lledó et al., 1998
; Wenzel, 1997
). Whole category weights (codon or tranversion) exhibit broad and overlapping ranges of consistency (Olmstead, 1997
), whereas SW independently assesses each base position of the multiple alignment based on their consistency in the initial analysis. The initial tree search was conducted under the Fitch (equal weights; Fitch, 1971) criterion with 1000 random sequence additions and SPR (subtree pruning-regrafting) branch swapping but permitting only ten trees to be held at each step to reduce the time spent searching trees at suboptimal levels. All trees collected in the 1000 replicates were swapped on to either completion or an upper limit of 5000 trees. The characters were then reweighted by the rescaled consistency index, and a further 50 replications of random sequence additions were conducted with the weighted matrix saving 15 trees per replication. These trees were then swapped on to completion or an upper limit of 5000 trees. The resulting trees were then used to reweight the matix a second time by the rescaled consistency index, and another 50 replications of random sequence addition conducted, saving 15 trees per replication, with subsequent swapping on those trees. This cycle was repeated until two successive rounds found trees of the same length. All analyses were run with the MULPARS option and ACCTRAN optimization. Branches with zero length were collapsed if the maximum value = 0 ("amb +"). Internal support was determined by bootstrapping (5000 replicates) with the final reweighted character matrix and with the jackknife program (5000 replicates) of Farris et al. (1996)
without SW weights applied. The cut-off bootstrap percentage is 50; minimum jackknife support percentage is 63 (Farris et al., 1996
). The rbcL matrix consisted of 81 taxa, 51 Amaryllidaceae s.s. representing 48 genera, and 30 additional taxa representing 28 genera of Agapanthaceae, Alliaceae, Anthericaceae, Behniaceae, Convallariaceae, Hyacinthaceae, Laxmanniaceae, and Themidaceae, with Geitonoplesium sp. (Hemerocallidaceae) used as outgroup. The trnL-F matrix includes these same with the addition of Stemmatium narcissoides (Alliaceae).
The trnL-F region consists of an intron, a short exon, and an intergene spacer (Taberlet et al., 1991
). We combined the components of trnL-F because they are nearly all noncoding, but each of the two larger regions was analyzed separately to determine whether they were congruent. Because they were congruent (results not shown), we lumped them together as the "noncoding matrix" to compare directly with rbcL before we combined all of them.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). The Amaryllidaceae are not resolved as monophyletic in the strict consensus of all 5000 SW trees; in these Agapanthus and Amaryllidaceae tribe Amaryllideae form a polytomy with the rest of Amaryllidaceae sensu stricto (s.s.). Moreover, these clades have no bootstrap or jackknife support.
|
In many of the trees (Fig. 1), the African tribe Amaryllideae is sister to the rest of Amaryllidaceae s.s. This monophyletic group has high bootstrap and jackknife support (Fig. 1). The rest of the family forms a polytomy (Fig. 2) that includes a baccate-fruited clade (Haemantheae, including Gethyllideae), the Cyrtantheae (confined to Africa), Calostemmateae (Australasia), and a monophyletic Eurasian/American group. Of these latter, only the Eurasian/American clade has any bootstrap (62) and jackknife support (67). Calostemmateae have a bootstrap percentage of 63 but no jackknife support. Within the Haemantheae, Apodolirion and Gethyllis (Gethyllideae) are resolved as sister taxa in the Fitch topologies, but not in the SW trees (Fig. 2).
|
The trnL-F matrix alone
Of the 1389 base positions (including gaps) included in the analysis, 378 were parsimony informative. More than 5000 equally most parsimonious trees were found of length = 1540 with CI = 0.66 and RI = 0.73. SW found more than 5000 equally parsimonious trees of length = 747723 (Fitch = 1541) with CI = 0.89 (Fitch = 0.66) and RI = 0.91 (Fitch 0.73), the strict consensus of which is more resolved than the initial Fitch consensus. The trnL-F matrix (Fig. 3) resolves a monophyletic Amaryllidaceae s.s. (bootstrap and jackknife support > 80%) as sister to Alliaceae with low bootstrap (56%) and somewhat higher jackknife support (71%). Agapanthus is sister to the Amaryllidaceae/Alliaceae clade with supporting bootstrap and jackknife percentages of 81 and 87%, respectively.
|
|
Compared to the rbcL topology, the American genera are less resolved by trnL-F; Griffinia and Worsleya appear outside the clade comprising all other American taxa (Fig. 4). The petiolate Andean clade, which appears in the rbcL consensus, loses two members, Eucharis and Rauhia. Hymenocallideae are not resolved, and Leptochiton and Pamianthe are resolved as sister genera with moderate bootstrap and strong jackknife support. Hippeastreae (less Griffineae) appear with low bootstrap support (61) but with different internal resolution than with rbcL. Again, short branch lengths are characteristic of most of the internal nodes of the American clade (Fig. 4).
The combined matrix
More than 5000 equally most parsimonious trees were found of length = 2546 with CI = 0.64 and RI = 0.71. SW found more than 5000 equally parsimonious trees of length = 1194297 (Fitch = 2546) with CI = 0.89 (Fitch = 0.64) and RI = 0.89 (Fitch = 0.71). The strict consensus of the weighted trees is more resolved than the initial Fitch consensus. Agapanthus is sister to Amaryllidaceae in the combined topologies (Fig. 5), albeit with low bootstrap support (60%). A monophyletic Alliaceae is sister to the former clade, with a bootstrap of 79% and jackknife of 77%. Both Amaryllideae and Haemantheae are well-supported tribal clades (Figs. 5–6), with higher bootstrap and jackknife percentages than in either of the separate analyses, and the former resolves as sister to the rest of Amaryllidaceae s.s.
|
|
Within the American clade (Fig. 6), a distinct Andean subclade has weak bootstrap support (68). Eustephieae have no consensus, bootstrap, or jackknife support. A well-supported Hippeastreae are in a polytomy with Griffinia, Worsleya, and the Andean clade. In Hippeastreae, a distinct Zephyranthinae and Rhodophiala/Traubia are well supported. Within the Andean clade, the resolution of Hymenocallideae, observed in the rbcL trees (Fig. 2), is lost, with Ismene, however, remaining monophyletic. Leptochiton and Pamianthe are weakly supported sister taxa. Eucharis and Rauhia fail to join the rest of a weakly supported petiolate-leafed subclade that is sister group to Hymenocallis (marked by a single synapomorphy).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
presented a smaller rbcL analysis and argued for the inclusion of Agapanthus as a monotypic subfamily within Amaryllidaceae. The sister-group status of Agapanthus to Amaryllidaceae s.s. is only weakly supported by our combined matrix (bootstrap = 60%). Based on these data, it would be possible to argue for recognizing Amaryllidaceae in a modified Hutchinsonian (1934)
sense, i.e., with three subfamilies, Allioideae, Agapanthoideae, and Amarylloideae.
Backlund and Bremer (1998)
discussed the issue of monogeneric families and how best to treat them. They generated a set of guiding principles for classification: (1) primary principle of monophyly and (2) a set of secondary principles: (a) maximizing stability, (b) maximizing phylogenetic information (= minimizing redundancy), (c) maximizing support for monophyly, and (d) maximizing ease of identification. Principle 1 is considered most important by Backlund and Bremer (1998)
, as it is by most modern systematists. However, the secondary principles will vary in importance among different taxa.
Monophyly is maximized by either treating Agapanthus as a monogeneric family or accepting Amaryllidaceae in the Hutchinsonian sense. However, the support for a broad concept of Amaryllidaceae (including Alliaceae and Agapanthaceae) is only moderate (bootstrap = 79%, jackknife = 77%). The only morphological character that unites all three families is the pseudo-umbellate inflorescence (homoplasious with Themidaceae), whereas the Alliaceae are readily marked by their solid styles and sulfonated compounds, and the Amaryllidaceae have inferior ovaries and unique alkaloid chemistry. Maximizing stability in this case seems rather moot, given that Agapanthus has been maintained for years as part of Alliaceae, a classification that violates the primary principle of monophyly, and Amaryllidaceae and Alliaceae have been united on and off again over the last two centuries. However, maximizing phylogenetic information and ease of identification are best served by treating Agapanthus as the sole genus of a separate family (Agapanthaceae Voight), while maintaining the independent status of Alliaceae.
The combined analysis supports most of the other relationships hypothesized by Fay and Chase (1996)
. The sister-group status of Themidaceae and Hyacinthaceae is confirmed with good support, and there is bootstrap support in the combined analysis for this clade as sister group to Amaryllidaceae/Alliaceae/Agapanthaceae, although Antheridaceae/Behniaceae resolves outside of this clade. It should be noted that, with this level of sampling of the broader Asparagales, these relationships are largely a matter of outgroup selection.
Relationships within Amaryllidaceae
Within Amaryllidaceae s.s., several groups are well supported within all of the analyses, some of which correspond to traditionally accepted tribes of the family. The most unexpected resolution concerns the sister status of the Eurasian/American clades. This is only supported in the combined analysis, and resolution of this group in relation to the remaining African and Australasian clades is still elusive because of short branch lengths in this portion of the trees (Figs. 2, 4, 6).
In a survey of internal morphology of American and African Amaryllidaceae, Arroyo and Cutler (1984)
noted several characters that separated American genera from African. All American species surveyed have scapes with collenchyma, a one-layered rhizodermis, and obvolute bracts. All Amaryllideae (entirely African with the exception of pantropical Crinum) have schlerenchyma in the scape, a multilayered rhizodermis, and equitant bracts. Haemanthus and Cyrtanthus exhibit scape and root anatomy of the American species but the equitant bracts of Amaryllideae (Arroyo and Cutler, 1984
). Calostemmateae (Calostemma and Proiphys), which were not discussed by Arroyo and Cutler (1984)
, have equitant bracts. Many of the Eurasian genera have fused spathe bracts, which obscures the pattern of their coherence, but both Lycoris and Pancratium species with free bracts show the equitant condition. Obvolute bracts may thus be a synapomorphy of the American clade.
Two American subclades are found in the consensus of the combined analysis (Fig. 5), with both Griffinia and Worsleya forming a polytomy with them. The more weakly supported Andean subclade (tribes Eucharideae, Stenomesseae and Eustephieae) is characterized by 2n = 46 chromosomes, which has been interpreted as a tetraploid derivation from an ancestral 2n = 22 (Meerow, 1985,
1987a
, c
, 1989
). The strongly supported Hippeastreae is characterized for the most part by x = 6 or 11, with diploid chromosome numbers of 22, 24 or less. The short branch lengths and numerous polytomies in the Andean group (Fig. 6) may indicate that they are a relatively young clade with an evolutionary history tied closely to the geologically recent Andean uplift (Meerow, 1987c
).
Four recognized tribes of Amaryllidaceae are consistently resolved by the plastid DNA sequences, and all receive strong bootstrap and jackknife support in at least the combined analysis. These are the Amaryllideae, Haemantheae, Calostemmateae, and Hippeastreae.
Amaryllideae
This tribe, with much of its generic diversity confined to South Africa is sister to the rest of the Amaryllidaceae and has high bootstrap and jackknife support. Compared to other tribes in Amaryllidaceae, Amaryllideae are marked by a large number of synapomorphies (Snijman and Linder, 1996
): extensible fibers in the leaf tissue, bisulculate pollen with spinulose exines, scapes with a sclerenchymatous sheath, unitegmic or ategmic ovules, and nondormant, water-rich, nonphytomelanous seeds with chlorophyllous embryos. A few of the genera extend outside of South Africa proper, but only Crinum, with seeds well adapted for oceanic dispersal (Koshimizu, 1930
), ranges through Asia, Australia, and America. Snijman and Linder's (1996)
phylogenetic analysis of the tribe based on morphological, seed anatomical, and cytological data resulted in recognition of two monophyletic subtribes: Crininae (Boophone, Crinum, Ammocharis, and Cybistetes) and Amaryllidinae (Amaryllis, Nerine, Brunsvigia, Crossyne, Hessea, Strumaria, and Carpolyza). Müller-Doblies and Müller-Doblies (1996)
recognized four subtribes with little discussion and no phylogenetic analysis: Crininae, Boophoninae, Amaryllidinae, and Strumariinae, the latter two containing several segregate genera from Hessea and Strumaria (Table 1). Our sampling of this tribe is incomplete, and therefore we feel it is premature to attach a great deal of confidence to the generic sister relationships seen here. Four genera of Snijman and Linder's (1996)
Amaryllidinae do form a weakly supported clade (Fig. 5) with Amaryllis as sister to the rest of the tribe in the rbcL and combined analyses. Identical positioning of Amaryllis occurred in Snijman's (1992)
cladistic analyses if tribe Hippeastreae was used as the outgroup, with Haemantheae as outgroup (Snijman and Linder, 1996
), and also both outgroups used (Snijman, 1992
). Amaryllis resolves as sister to a clade containing the other genera they ultimately placed, with Amaryllis, in subtribe Amaryllidinae. Müller-Doblies and Müller-Doblies' (1996)
concept of Amaryllidinae [Amaryllis, Nerine, and Namaquanula (= Hessea)] would make their subtribe Strumariinae paraphyletic and Amaryllidinae polyphyletic.
Haemantheae
This baccate-fruited tribe is another morphologically well-marked group with strong molecular support. The limits of the tribe, however, have been controversial. Müller-Doblies and Müller-Doblies (1996)
insisted on retaining Cyrtanthus in the tribe, albeit as a monotypic subtribe, Cyrtanthinae. The basis for uniting Cyrtanthus with the Haemantheae has always been weak, chiefly the shared chromosome number with Haemanthus (2n = 16; Ising, 1970
; Vosa and Snijman, 1984
) and its strictly African range. This diploid number also occurs in some Hippeastreae (Flory, 1977
; Grau and Bayer, 1991
). Uniting Cyrtanthus with Haemantheae has no molecular support in our analyses, and we believe that Cyrtanthus, the only solely African genus with the flattened, winged, phytomelanous seed so common in the American clade, should be recognized as a monotypic tribe (Traub, 1963
; Dahlgren, Clifford, and Yeo, 1985
; Meerow and Snijman, 1998
). Recognition of Gethyllideae as a distinct tribe (Müller-Doblies and Müller-Doblies, 1996
; Meerow and Snijman, 1998
), however, is not supported by the molecular data. Although the large, elongate, baccate fruits and small hard seeds of Apodolirion and Gethyllis are a departure from the berries and large succulent seeds of the rest of Haemantheae, the two genera, though resolved as sister taxa (Figs. 4, 6), are firmly embedded within Haemantheae. Recognizing them as a distinct taxon would render the rest of Haemantheae paraphyletic. Haemantheae are the only tribe of Amaryllidaceae that contain rhizomatous genera (Cryptostephanus and Scadoxus in part), a condition that occurs in the sister family Agapanthaceae. This has generally been conceived as a plesiomorphy within the family (Nordal and Duncan, 1981; Meerow, 1995
, 1997
; Müller-Doblies and Müller-Doblies, 1996
). In the rbcL and combined consensus trees, the three "bulbless" genera form a grade at the base of the Haemantheae, which would support this hypothesis, although Cryptostephanus, the only member of the tribe with the ancestral state of a phytomelanous testa, is not the first branch in the grade. Haemanthus and Scadoxus, which have been treated as one genus in the past (e.g., Hutchinson, 1934
, 1959
; Traub, 1963
), are sister genera only in the rbcL topologies (Fig. 2). The position of Scadoxus, the only genus of the tribe polymorphic for the rhizomatous state, as the final terminal taxon in the "bulbless" grade seems reasonable. In any event, all three matrices render recognition of a subtribe Cliviinae for Clivia and Cryptostephanus by Müller-Doblies and Müller-Doblies (1996)
as paraphyletic. Any further insight on the internal relationships within Haemantheae requires additional sampling.
Calostemmateae
Calostemmateae, treated as part of a polyphyletic Eucharideae by Hutchinson (1934
, 1959)
, Traub (1963)
, and Dahlgren, Clifford, and Yeo (1985)
, were first suggested as a distinct lineage by Meerow (1989)
and formally recognized by Müller-Doblies and Müller-Doblies (1996)
. The tribe consists of two Australasian genera (Proiphys, forest understory herbs of Malaysia, Indonesia, the Philippines and tropical Australia, and Calostemma, endemic to Australia). A few species of Crinum, with the broadest distribution of any genus in the family, are the only other members of Amaryllidaceae present in Australia. The indehiscent capsules of both genera are similar in appearance to the unripe berry-fruits of Scadoxus and Haemanthus (Haemantheae), but early in the development of the seed, the embryo germinates precociously, and a bulbil forms within the capsule and functions as the mature propagule (Rendle, 1901
). The two genera exhibit the equitant bract condition of the African and Eurasian genera.
Hippeastreae
All but two of the genera treated by Meerow and Snijman (1998)
as part of Hippeastreae are resolved as a well-supported monophyletic clade in all the analyses (Figs. 2, 4, 6). The two genera that lie outside of this clade are Worsleya and Griffinia, both Brazilian endemics, exhibiting the rare character of blue-range pigmentation in the flowers. The variable positioning of these two in the various analyses is interesting in itself. In the trnL-F topologies (Fig. 4), Worsleya is part of the basal polytomy within the Eurasian/American clade, whereas Griffinia weakly resolves as sister to the Mediterranean Hannonia (the latter on a long terminal branch). In the rbcL consensus (Fig. 2), Worsleya resolves as sister to Chlidanthus (Eustephieae), whereas Griffinia remains unresolved along with the rest of Eustephieae. In the combined analysis (Fig. 6), both are positioned within the American clade, but unresolved with either Hippeastreae s.s. or the weakly supported tetraploid Andean clade. The failure of Worsleya or Griffinia to resolve as part of Hippeastreae in any of the analyses casts doubt on Müller-Doblies and Müller-Doblies' (1996)
submergence of Worsleya in Hippeastrum and weakens Meerow and Snijman's (1998)
retention of both genera in tribe Hippeastreae.
Another unexpected indication of relationship occurs within the tetraploid Andean clade, where a distinct petiolate-leafed subclade is resolved in the rbcL topologies (Fig. 2). This resolution is not retained by the trnL-F and combined analyses in which Eucharis and Rauhia are pulled from this group. Nonetheless, a core of petiolate genera remain monophyletic, with weak support in the combined analysis. Despite the fact that petiolate leaves have evolved independently several times elsewhere in the Amaryllidaceae (Amaryllideae, Calostemmataceae, Haemantheae, Hymenocallideae, and Hippeastreae), the molecular data begin to indicate that it may be a synapomorphy for this group. Hymenocallideae as a distinct tribe receive weak support (55%, one synapomorphy) in the rbcL matrix only, and the rest of Stenomesseae is poorly resolved by all matrices.
Within the Eurasian clade of the combined analysis (Fig. 6), Lycorideae appears as sister to the rest, although without support. This tribe represents the more or less temperate Asian component of the family, with Lycoris ranging from Korea, through China, Myanamar, and Japan, and Ungernia restricted to the mountains of central Asia. Müller-Doblies and Müller-Doblies (1978)
described similarities in the bulb anatomy of Ungernia and Sternbergia, a possible synapomorphy between Lycorideae and the rest of this clade. One genus of the Eurasian clade, Pancratium, is represented throughout Africa, tropical Asia, as well as Mediterranean Europe and the Middle East, a distribution that could signify a more ancestral position within the Eurasian clade. The current data, however, do not support this resolution for Pancratium, with only a single species from the Canary Islands represented in the analyses. The presence of Pancratium in Africa may thus be secondary, though it is the only genus outside of the African tribes Amaryllideae and Haemantheae with external trichomes (Björnstad, 1973
).
Sister relationships of Hannonia and Vagaria receive good bootstrap and jackknife support, as does the traditional alliance of Galanthus and Leucojum. Crespo et al. (1995)
, using ITS sequences, refuted Müller-Doblies and Müller-Doblies' placement of Lapiedra in Pancratieae, and our data support a closer relationship with Galantheae or Narcisseae for this genus. However, concepts of Galantheae, Narcisseae, and Pancratieae presented in Müller-Doblies and Müller-Doblies (1996)
or Meerow and Snijman (1998)
are not resolved in any of the three analyses, and we believe that caution should be used before categorical statements are made about tribal lineages within this group.
The low internal branch lengths throughout the Amaryllidaceae, except in some of the deepest branches, are a striking contrast to the other asparagalean families included in the analysis (Figs. 1, 3, 5). The significance of this is not clear. It could mean that a great deal of the modern diversity in the family is of relatively recent occurrence (as is likely, for example, within the Andean clade), or else base substitution rates in the chloroplast genome are lower within the family than for other Asparagales.
Character state evolution in the Amaryllidaceae
By optimizing morphological or other "traditional" characters onto a gene tree, one is able to gain insight about putative transformation series or state polarities that have characterized the evolution of the group under study. This can be useful for constructing a character state matrix for an ingroup in which rampant homoplasy in such characters confounds the endeavor. Certain characters that have been used to justify older intrafamilial classifications of Amaryllidaceae do show stability within some of the clades resolved by the combined analysis, while others appear extremely homoplasious (Fig. 7).
|
Presence or absence of bulbs
The bulbless condition occurs in the sister group to Amaryllidaceae, the monogeneric Agapanthaceae. It is also the character state for the only South African subfamily of Alliaceae, Tulbaghoideae (Fay and Chase, 1996
). In Amaryllidaceae, the absence of bulbs characterizes only three genera, Clivia, Cryptostephanus, and Scadoxus, but the latter also includes species that form a true bulb. If this is a symplesiomorphy as most have interpreted it (Nordal and Duncan, 1984
; Müller-Doblies and Müller-Doblies, 1996
; Meerow and Snijman, 1998
), the bulbous state has evolved at least three times in the family, in Amaryllideae, Haemantheae, and within the ancestral stock for the rest of the family.
Petiolate leaves
Petiolate or, more accurately, pseudopetiolate leaves are widepread throughout the Asparagales, and this character exhibits a great deal of homoplasy within Amaryllidaceae (Meerow and Snijman, 1998
). At the extreme, one-to-few petiolate species occur in otherwise lorate-leafed genera (e.g., Crinum, Hymenocallis). The state may occur throughout a genus, but renders a tribe polymorphic (Calostemmateae, Haemantheae, Griffineae). In the tetraploid Andean clade, a subclade is defined by the synapomorphy of a petiolate leaf in the rbcL trees, but Eucharis and Rauhia pull away with trnL-F and in the combined analyses.
Mesophyll palisade
It has been suggested that the presence or absence of a distinct palisade layer in the leaf mesophyll may have systematic significance (Arroyo and Cutler, 1984
; Artyushenko, 1989
). Petiolate-leafed taxa never have palisade chlorenchyma (Meerow and Snijman, 1998
). It is characteristic of Amaryllideae (Crinum is polymorphic), but absent in Haemantheae (the state in unknown for Gethyllis and Apodolirion). In Calostemmateae, it is present in Calostemma but absent in the petiolate Proiphys (Meerow, unpublished data). Palisade almost universally occurs in the Eurasian clade. It is absent in Leucojum (Artyushenko, 1989
), and Galanthus is polymorphic (Davis and Barnett, 1997
). Within the American clade, it is wholly characteristic of the Eustephieae (Arroyo and Cutler, 1984
; Meerow and Snijman, 1998
) but occurs only sporadically within the Hippeastreae (Arroyo and Cutler, 1994
). The state of Worsleya is not known. Outside of Eustephieae, a distinct palisade is absent from the Andean tetraploid clade (Meerow, 1987a
, 1989
). The inference based on the distribution of this character state on our topology (Fig. 7) is that a distinct palisade is plesiomorphic within the family, though the state within Agapanthaceae, sister to Amaryllidaceae, has not to our knowledge been reported.
Pubescence
The presence of trichomes on the external parts of Amaryllidaceae is common only in some Amaryllideae and Haemantheae and one African species of Pancratium (Arroyo and Cutler, 1984
; Meerow and Snijman, 1998
). It is completely unknown in the American clade, Cyrtantheae, and Calostemmateae. It may have evolved independently in the three clades within which it occurs.
Scape characters
Solid scapes are the predominant condition in Amaryllidaceae as occurs in Agapanthaceae as well. Hollow scapes are almost universally characteristic of Hippeastreae, and thus appears to be a synapomorphy for that tribe. The only other genera within which hollow scapes occur are Leucojum and Cyrtanthus both of which are polymorphic for the character (Traub, 1963
; Reid and Dyer, 1984
). As discussed previously, obvolute spathe bracts seem to be apomorphic for the American clade, and the presence of schlerenchyma in the scape is an autapomorphy for Amaryllideae.
Floral symmetry
Zygomorphic and actinomorphic flowers occur in the Amaryllidaceae, and several genera (Crinum, Cyrtanthus, Phycella) are polymorphic. Snijman and Linder (1996)
consider actinomorphy the apomorphic condition in Amaryllideae. The flowers of Agapanthaceae are zygomorphic. Within Haemantheae, only Clivia is zygomorphic. In the American clade, zygomorphy is the rule in the "hippeastroid" subclade. Pyrolirion and Zephyranthes (including Haylockia) are the only genera characterized exclusively by actinomorphic flowers, while Phycella (not included in the sequence analyses) is polymorphic. In the Andean subclade, only Eucrosia, Plagiolirion, and Ismene subgenus Elisena are exclusively zygomorphic; Rauhia is polymorphic. The Eurasian clade is on the whole actinomorphic; only Lycoris is characterized by zygomorphic flowers. The mosaic occurrence of actinomorphy throughout the family (Fig. 7) and the occurrence of polymorphic genera suggest that transformations between the two states of floral symmetry may be easily modified by pollinator-mediated selection, and perhaps controlled by one or few genes.
Paraperigone
The "paraperigone" is an anomalous secondary outgrowth of the perianthal meristem with ramifying vasculature (Arber, 1939
; Singh, 1972
), not to be confused with a similar-looking structure formed by staminal connation (see below). It is most well developed (and typified) by the corona of Narcissus. Such a well-developed paraperigone occurs in only one other genus, the Chilean endemic Placea (Hippeastreae). However, a homologous series of fimbrae, scales, or a continuous callose ring occurs in Cryptostephanus (Haemantheae), one or two species of Cyrtanthus, and variably thoughout Lycorideae and Hippeastreae. It has thus probably evolved at least three times (Haemantheae, Cyrtantheae, and the Eurasian/American clade), but from a meristematic potential that is deep rooted in the family. Polymorphism for this character within genera may suggest that it is easily lost.
Staminal connation
The fusion of the staminal filaments was the single most important character with which Traub (1957
, 1963
) justified recognizing his "infrafamily" Pancratioidinae, a subfamilial taxon that in fact was glaringly polyphyletic. Though staminal connation is a widespread character state within the Andean tetraploid clade (Fig. 7), it is paralleled elsewhere in the family, particularly in Amaryllideae subtribe Amaryllidinae (Snijman and Linder, 1996
), the Calostemmateae, in Gethyllis, and some species of Cyrtanthus (Reid and Dyer, 1984
). In the Eurasian clade it occurs in Pancratium, the flower morphology in general of which bears striking resemblance to several Andean genera (Hymenocallideae pro parte, Paramongaia, and Pamianthe). Meerow and Dehgan (1985)
attemepted to link these so-called "pancratioid" genera by pollen morphology, but a more parsimonious explanation may be convergence for pollinator specificity (Morton, 1965
; Bauml, 1979
; Grant, 1983
). However, Pancratium and these Andean genera are monophyletic in a larger sense (as part of the Eurasian/American clade), and the exact position of Pancratium within the Eurasian subclade is still not strongly resolved (Fig. 6).
Fruit and seed characters
Fruit and seed morphology have been an important focus of experimentation within the family. Baccate fruits have apparently evolved only once, despite the difference in gross morphology between the long, aromatic fruit of Gethyllis and Apodolirion and the berries of the rest of Haemantheae. Phytomelan [the ancestral state for all Asparagales (Huber, 1969
)] has been lost from the testa as many as five times in the Amaryllidaceae: in Amaryllideae, Griffineae, Hymenocallideae, Haemantheae, and Calostemmateae [in Calostemmateae a true seed never forms, but an integumentary rudiment is present (Rendle, 1901
)]. In both Haemantheae and Hymenocallideae, phytomelan is found around the seeds of one genus each (Cryptostephanus and Leptochiton, respectively). The loss of one integument [or both, as been controversially reported for some Crinum (Prillieux, 1858
; von Schlimbach, 1924
; Tomita, 1931
; Markötter, 1936
, but see Snijman and Linder, 1996
)] is synapomorphic for Amaryllideae.
A flattened, winged seed, which occurs in Agapanthaceae, is very common in the American clade, but otherwise occurs only in Ungernia (Lycorideae) and Cyrtantheae. The most similar type of seed to this is the D-shaped seed of Worsleya and some Pancratium. A dry, hard, wedge-shaped or irregularly round seed is characteristic of most of the Eurasian clade (except Lycorideae), frequently with an elaiosome at the chalazal end. Among all genera of the family, Pancratium is the most polymorphic for seed type (Werker and Fahn, 1975
).
Characterization of certain seeds of Amaryllidaceae as fleshy (regardless of whether phytomelan is present) has led, in the past, to false homologies (see discussion in Meerow, 1989
). Truly fleshy seeds occur in Amaryllideae (in which case the bulk of the seed volume is endosperm; Rendle, 1901
), Hymenocallideae (the fleshy portion is integumentary; Whitehead and Brown, 1940
), and some water-rich Haemantheae. But an "intermediate" state occurs in a number of genera in which the seed is round, turgid, but not really fleshy (the seed will burst under pressure rather than give way), and contains copious, oily endosperm. This type of morphology is found in Cryptostephanus (Haemantheae), Lycoris (Lycorideae), Eucharideae sensu Meerow (1989)
, Griffinia, and a single species of Hippeastrum.
Chromosome number
A chromosome number of 2n = 22 is considered plesiomorphic in Amaryllidaceae due to the broad occurrence in many of the tribes of the family (Goldblatt, 1976
; Flory, 1977
; Meerow, 1984
, 1987b
). Andean-centered genera in the tribes Eucharideae, Eustephieae, Hymenocallideae, and Stenomesseae are characterized by a somatic chromosome number of 2n = 46 or presumptive derivations thereof (Di Fulvio, 1973
, Flory, 1977
; Williams, 1981
; Meerow, 1987a
, b
). Resolution of these genera as a clade in our plastid DNA trees supports the interpretation of a monophyletic polyploid origin for these tribes from an ancestor with 2n = 22 via chromosome fragmentation or duplication and subsequent doubling or vice versa (Satô, 1938
; Lakshmi, 1978
; Meerow, 1987b
).
Biogeographic implications
Raven and Axelrod (1974)
postulated a western Gondwanaland origin for Amaryllidaceae sensu Huber (1969)
, and this is supported by the plastid DNA phylogeny. The deepest branches of the topology originate in Africa, including the sister group of the family Agapanthus. Africa has also been the site of considerable innovation in the family's history as well, as typified by the Afrocentric tribes Amaryllideae, Haemantheae, and Cyrtantheae. Most of the diversity within those three tribes is, however, centered in South Africa, and thus may reflect radiation engendered by the more recent paleoclimatic and geological history of Africa encompassing Neogene and later times (Axelrod, 1972
; Raven and Axelrod, 1974
). The increased aridity of the African climate and the uplift of the continental mass beginning near the end of the Oligocene, further abetted by Quaternary climatic fluctuations, were catastrophic to many elements of the African flora, but it may have been a selective pressure for diversity among groups of geophytes capable of adapting to increasing drought. The geophyte richness of South Africa is well documented (Goldblatt, 1978
), and the Cape region has been suggested as a possible refuge for certain African plant and animal groups as the tropical flora of the continent was impoverished (Raven and Axelrod, 1974
). However, the three basal genera of the baccate-fruited Haemantheae according to our combined analysis (Fig. 6), Clivia, Cryptostephanus, and Scadoxus, are all forest understory taxa, do not form bulbs, and are at least in part (Scadoxus, Cryptostephanus) elements of tropical vegetation farther north. Cryptostephanus does not occur in South Africa at all, and this is the only genus of Haemantheae in which the plesiomorphic state of a phytomelanous testa occurs.
The Calostemmateae, the only exclusively Australasian element of the family, may have been isolated from the African lineages as Australia separated from western Gondwanaland (Raven and Axelrod, 1974
). Direct migration between Africa and Australia may have persisted up through the close of the early Cretaceous, although India and Madagascar may have provided a less direct corridor up until the late Cretaceous (Raven and Axelrod, 1974
). That the Calosternmateae remains within the unresolved grade of otherwise African tribes would suggest relative antiquity for the lineage. Crinum is the only amaryllid that is known to occur on Madagascar, despite the island's probable role as a refuge for taxa decimated by the Neogene African extinctions, whereas indigenous Indian amaryllids are restricted to Crinum and two to three species of Pancratium. The adaptations of Crinum for long-distance dispersal have been demonstrated (Koshimizu, 1930
), and Pancratium may have been able to directly enter India from either Africa or Eurasia during the late Cretaceous or early Eocene (Raven and Axelrod, 1974
).
The sister relationship of the Eurasian/Mediterranean clade to the American genera raises the interesting question of when and wher
