Laboratory procedures

Total genomic DNA was extracted from fresh or silica-gel dried material using the 2× CTAB method of Doyle and Doyle (1987) with minor modifications. Total DNA was purified by MOBIO minicolums and mostly kept in 0·1× TE buffer (10 mm Tris–HCl, 1 mm EDTA, pH 8·0). Some samples were purified on CsCl₂/ethidium bromide (1·55 g mL⁻¹) density gradients (Creeth and Denborough, 1970).

Some samples, especially those from southern Africa, were obtained from the DNA banks of the Royal Botanic Gardens, Kew (www.rbgkew.org.uk/data/dnaBank/homepage.html) and National Botanic Gardens, Kirstenbosch (NBG) (www.nbi.ac.za/frames/researchfram.htm) (Table 1). The DNA for the remainder of samples was isolated and deposited at the DNA banks of Alicante University and the Royal Botanic Gardens, Kew (Table 1).

The trnL intron and trnL-F spacer (the trnL-F region; hereafter trnL-F) were amplified, mostly in one piece, using primers c and f (Taberlet et al., 1991), though some taxa required the use of the internal primers d and e (Taberlet et al., 1991). Amplification of the rbcL gene was performed using the primers 1F and 1360R, though some taxa required the use of internal primers, 427F and 724R (Lledó et al., 1998). ITS was amplified using the 18Z and AB102 primers (Douzery et al., 1999; Pires, 2000). When the sequences were not obtained in one reaction, the internal primers ITS2 and ITS3 (White et al., 1990) were also used.

A total volume of 25 µL was used in the PCR amplification process as follows: 21 µL of Abgene PCR mastermix (1·25 U Thermoprime Plus DNA polymerase, 75 mm Tris–HCl, 20 mm (NH₄)₂SO₄, 2·5 mm MgCl₂, 0·01 % Tween^® 20, 0·2 mm each dNTP); 0·5 µL of bovine serum albumin (BSA; 0·4 %); 10 mm of each primer; 0·5 µL of water; and approx. 40 ng of DNA template. The ITS amplifications were done using 1·5 mm MgCl₂ Abgene PCR mastermix and 1 µL of dimethyl sulphoxide (DMSO). Amplifications were performed on a 9700 GeneAmp thermocycler (Applied Biosystems), with the following programme: 4 min at 94 °C, followed by 28 cycles of 94 °C for 1 min, 48 °C for 30 s for rbcL, or 50 °C for 1 min for trnL-F, 72 °C for 1 min, and a final extension 72 °C for 7 min. For ITS, a touchdown programme was used with a initial 94 °C for 5 min, following by 12 cycles of 94 °C for 30 s, 59 °C for 1 min decreasing 1 °C in each cycle plus 18 cycles at 48 °C for 1 min, 72 °C 1 min, and a final extension 72 °C for 7 min. PCR products were purified with MOBIO minicolumns following the manufacturer's protocol. Sequencing data were obtained using ABI Prism Big Dye Terminator Cycle Sequencing Kits (Applied Biosystems) and run on an ABI 3700 Genetic Analyser according to manufacturer's protocols.

All ITS sequences are original (Table 1), whereas some of those for trnL-F and rbcL (mostly from southern African taxa) and the matK matrix were obtained from John Manning (South African National Biodiversity Institute, Kirstenbosch) and Félix Forest (Royal Botanic Gardens, Kew). The plastid (except matK) sequences of the Eurasian species are all original. All new sequences were deposited in GenBank (see Table 1).

ITS trees

Amplification of the ITS region was successful for most species and yielded useful data for phylogenetic studies in Ornithogaloideae. However, it was not possible to sequence Pseudogaltonia, the position of which is thus unknown. The ITS matrix contained 959 characters, of which 500 were variable and 337 (67·4 %) were potentially phylogenetically informative.

A total of 5590 MP trees was obtained (1539 steps, CI = 0·52, RI = 0·76). The topology of the parsimony strict consensus tree (not shown) is similar to the Bayesian tree. Therefore, only one of the post-burn-in trees obtained from a Bayesian analysis under the GTR + I + G model is shown (Fig. 1). Bayesian PP and parsimony BS values are well correlated, the Bayesian being consistently higher (Fig. 1).

Open in a separate window

Fig. 1.

Bayesian inference analysis obtained from ITS sequences in the phylogenetic analysis of Ornithogaloideae. Phylogenetic relationships shown by one of the post-burn-in topologies. Taxa are identified following the traditional treatment of Ornithogalum and other related genera (cf. Obermeyer, 1978; Zahariadi, 1980; Müller-Doblies and Müller-Doblies, 1996). The new proposed generic circumscription of the present study is shown on the right side. Branch lengths are shown above branches, and posterior probabilities and bootstrap percentages are shown below. Arrows indicate branches that collapse in the 50 % majority rule consensus tree obtained by Bayesian analysis.

Several main clades can be distinguished (Fig. 1). One formed by Neopatersonia plus O. rotatum U. Müll.-Doblies & D. Müll.-Doblies (0·97 PP; 99 % BS) is sister to a large clade including the remainder of the subfamily (0·97 PP; 56 % BS).

Successively sister groups appear in that large group. First, there is a well-supported clade (1·00 PP; 93 % BS) that can be divided into two subclades. One (1·00 PP; 80 % BS) comprises the Dipcadi clade (0·99 PP; 76 % BS) sister to the Battandiera clade (0·89 PP; 77 % BS). The other (0·87 PP; 68 % BS) includes Stellarioides, Trimelopter, Albuca and Coilonox species, and is divided into two groups. One includes Albuca plus Coilonox (1·0 PP; 66 % BS) and shows little internal resolution, though Coilonox (O. polyphyllum Jacq.-O. suaveolens Jacq.) appears as a clade (1·00 PP; 60 % BS). The other comprises Stellarioides plus Trimelopter (0·80 PP; 53 % BS), in which the monophyly of both subclades is strongly supported (1·00 PP, 98 % BS and 1·00 PP, 100 % BS, respectively).

Secondly, a large clade not fully resolved includes two subclades. The first one includes the Madagascan endemic O. convallarioides H.Perrier (=Avonsera Speta) and a clade comprising Eliokarmos, Ethesia and Galtonia (1·00 PP; 58 % BS); however, their phylogenetic relationships are not resolved. In this latter clade, two groups are identified. One (0·95 PP; 96 % BS) includes Galtonia, O. saundersiae (= Zahariadia Speta) and two species of O. subgenus Urophyllon (sensu Obermeyer, 1978): O. haalenbergense U.Müll.-Doblies & D.Müll.-Doblies and O. xanthochlorum Baker (0·95 PP; 96 % BS). The other is highly supported (1·00 PP; 100 % BS) and corresponds to Eliokarmos (= O. subgenus Aspasia groups Aspasiae and Hispidae, sensu Obermeyer, 1978), with several internally well-supported groups. Finally, the second subclade includes all Eurasian and north African taxa of Ornithogalum (including Honorius, Loncomelos, Melomphis and Cathissa) plus the southern and eastern African O. subgenus Aspasia group Angustifoliati (sensu Obermeyer, 1978) that corresponds to Nicipe Raf. and Elsiea F.M.Leight. This clade is strongly supported (1·00 PP; 95 % BS) though the internal relationships are not fully resolved. Only some clades are well supported, namely those corresponding to Ornithogalum plus Honorius (1·00 PP; 99 % BS), Loncomelos (1·00 PP; 97 % BS) and Nicipe (1·00 PP; 81 % BS).

Combined trees

Combination of trnL-F, rbcL, matK and ITS regions generated a total matrix with 4242 characters, of which 1198 were variable and 653 (54·5 %) were potentially phylogenetically informative. A total of 1761 MP trees was obtained (2713 steps, CI = 0·58, RI = 0·78). The topology of the parsimony strict consensus tree (not shown) is quite similar to the Bayesian consensus tree obtained under the GTR + I + G model. Phylogenetic relationships are shown by one of the post-burn-in topologies (Fig. 2). Bayesian PP and parsimony BS values are well correlated, the Bayesian being consistently higher (Fig. 2).

Open in a separate window

Fig. 2.

Bayesian inference analysis of the combination of rbcL, trnL-F, matK and ITS sequences in the phylogenetic analysis of Ornithogaloideae. Phylogenetic relationships shown by one of the post-burn-in topologies. Taxa are identified following the traditional treatment of Ornithogalum and other related genera (cf. Obermeyer, 1978; Zahariadi, 1980; Müller-Doblies and Müller-Doblies, 1996). The new proposed generic circumscription of the present study is shown on the right side. Branch lengths are shown above branches, and posterior probabilities and bootstrap percentages are shown below. Arrows indicate branches that collapse in the 50 % majority rule consensus tree obtained by Bayesian analysis.

The trees in Figs 1 and ​and22 generally have a similar topology, but clade support is substantially higher in the latter. However, the combined tree (Fig. 2) shows two clades placed in quite different positions compared with the ITS tree (Fig. 1). Firstly, the Neopatersonia plus O. rotatum clade (1·00 PP; 100 % BS) is in the Bayesian combined tree (Fig. 2) nested within Ornithogaleae (Clade C), sister to the group including the Eurasian and North African species plus Nicipe and Elsiea (0·98 PP). That clade is, however, placed as sister to the rest of Ornithogaleae in the MP combined tree (not shown), with moderate support (88 % BS). Secondly, the Trimelopter clade (O. unifolium Retz. plus O. etesiogaripense U.Müll.-Doblies & D.Müll.-Doblies) is in the combined tree (by both MP and Bayesian analyses) sister to the clade comprising Stellarioides, Coilonox and Albuca.

Two main clades can be identified in Fig. 2. The first main clade (Clade A + B), which corresponds to Albuceae plus Dipcadieae (sensu Manning et al., 2009a), receives moderate support (0·71 PP; 75 % BS) and includes the genera Pseudogaltonia, Dipcadi, Battandiera, Trimelopter, Albuca, Coilonox and Stellarioides as recognized here. It can be divided into two subclades. Pseudogaltonia and Dipcadi species form a strongly supported group (1·0 PP; 94 % BS). In the Dipcadi clade, segregation of the Mediterranean D. serotinum Medik. from the South African representatives is strongly supported (1·0 PP; 100 % BS). A second clade (1·00 PP; 65 % BS) comprises Battandiera, Trimelopter, Stellarioides, Albuca and Coilonox. Within that clade, as sister to the rest, is a well-supported group (1·0 PP; 99 % BS) including O. seineri (Engl. & K.Krause) Oberm., O. pulchrum Schinz, O. stapffii Schinz and O. amoenum Batt., which are here included in Battandiera. The next branch groups O. etesiogaripense and O. unifolium (Trimelopter clade) (1·0 PP; 100 % BS) together, sister to Albuca, Coilonox and Stellarioides (1·0 PP; 88 % BS). The Stellarioides species (O. longebracteatum Jacq., O. sordidum Baker and O. tenuifolium Redouté) constitute a fully supported subclade (1·0 PP; 100 % BS), sister to Albuca plus Coilonox (1·0 PP; 99 % BS). Although Albuca was not supported as monophyletic in the heuristic MP analysis, the Bayesian one yielded two fully supported monophyletic groups: Coilonox (1·00 PP) and Albuca (1·00 PP). In Albuca two main clades appear, one of which matches the former Pallastema Salisb (0·99 PP; 55 % BS).

The second main clade in the combined tree (Clade C), which corresponds to Ornithogaleae (sensu Manning et al., 2009a), is well supported (1·0 PP; 88 % BS). However, relationships of the principal lineages are not well resolved. First, a large clade (1·0 PP; 69 % BS) comprises the species of the here-accepted genera Galtonia, Ethesia and Eliokarmos. Each of these three genera form a well-supported clade. Species of Galtonia sensu stricto plus O. saundersiae Baker form a strongly supported subclade (1·0 PP; 97 % BS), and therefore we consider all of them as members of Galtonia. This group is sister to the one formed by O. haalenbergense U.Müll.-Doblies & D.Müll.-Doblies and O. xanthochlorum Baker (1·0 PP; 100 % BS), here segregated as Ethesia. Sister to these two genera is a strongly supported subclade (1·0 PP; 100 % BS), which we accept as Eliokarmos, that includes species of O. subgenus Aspasia groups Aspasiae and Hispidae (sensu Obermeyer, 1978). Moreover, in Eliokarmos two strongly supported subgroups (each one 1·0 PP; 100 % BS) are identified which show morphological peculiarities as discussed below: one comprising O. maculatum Jacq. and O. rupestre L.f., and the other the rest of the species of Eliokarmos.

The position of the Madagascan O. convallarioides (=Avonsera) is not resolved, as it constitutes a separate clade in an unresolved position in Ornithogaleae. As said before, the Neopatersonia clade falls sister to a well-supported third main clade (1·0 PP; 99 % BS) that comprises the rest of representatives from Europe, Middle East and North Africa. This latter clade includes all samples of Ornithogalum, Honorius, Loncomelos, Melomphis and Cathissa, plus some species from southern and eastern Africa belonging to Elsiea and Nicipe. Within this wide group, O. esterhuyseniae Oberm. (=Elsiea) is strongly supported (1·0 PP; 99 % BS) as sister to the rest of taxa. The species of Nicipe form a strongly supported subclade (1·0 PP; 97 % BS), sister to the rest of the Eurasian and North African taxa (0·91 PP; 61 % BS). The species of Cathissa form a clade (1·0 PP; 62 % BS). Relationships of Melomphis, which is represented here only by O. arabicum L., are not resolved. Finally, the species of Loncomelos form a strongly supported subclade (1·0 PP; 99 % BS), which is sister to Honorius plus Ornithogalum (1·0 PP; 94 % BS), with Ornithogalum being rather well supported (1·0 PP; 68 %BS).

Why the need for another new arrangement?

Following studies by Speta (1998a) and Pfosser and Speta (1999), 13–15 genera were accepted within Ornithogaloideae (Table 2). This arrangement was based exclusively on morphology and/or sequencing of a single plastid region (trnL-F). Some genera as circumscribed by these authors (e.g. Stellarioides, Eliokarmos and Neopatersonia) have, however, been demostrated to be para- or polyphyletic in recent studies (including the present one, as discussed below).

The phylogenetic studies of Hyacinthaceae by Manning et al. (2004) added the rbcL region to the previous studies with trnL-F, and improved knowledge of the phylogenetic relationships of some groups. They reported a strict consensus tree with many polytomies, notably in Ornithogaloideae, perhaps due to the low nucleotide variation in these regions (Lihová et al., 2004; Shaw et al., 2005; Pirie et al., 2007). Nonetheless, Manning et al. (2004) adopted a ‘maximally stable arrangement’, with a single genus Ornithogalum for the whole subfamily. According to Lebatha et al. (2006), the polytomies in the strict consensus trees for Hyacinthaceae show a conflict or lack of hierarchy among groups, though it would not justify their fusion. These latter authors also pointed out that morphology can be useful in elucidating phylogenetic relationships in the family.

The addition of another plastid region (matK) resulted in a better-resolved topology and in a new taxonomic arrangement (Manning et al., 2009a). Their new proposal was produced in a scenario in which three more or less contradictory taxonomies were discussed. The first (Option 1) retains as far as possible genera accepted by Speta (1998a), with the recircumscription or addition of as many genera as necessary to preserve monophyly. The second (Option 2) assumes only one genus, Ornithogalum, following Manning et al. (2004). The third (Option 3) is an intermediate, medium-conservative solution in which only four genera are accepted, corresponding to the four main lineages found in the plastid tree. All three options were extensively discussed, and finally the solution with four genera was favoured (Table 2). However, different criteria appear to be used to assign generic rank to particular clades. For instance, the sister groups Pseudogaltonia and Dipcadi are segregated as proper genera on the basis of morphological characters that are of a similar degree (or even weaker) than those separating other related clades, such as Neopatersonia–Ornithogalum, Galtonia–-Ornithogalum, Battandiera–Albuca or Coilonox–Albuca, which are treated as subgenera or sections.

As pointed out by Manning et al. (2009a), recognition of many monophyletic genera (as favoured here), or acceptance of only four genera with many subgenera and sections (as they proposed), are both equally valid on theoretical grounds. However, Manning et al. (2009a) rejected the solution with many genera (Option 1) since it would require radical taxonomic changes including ‘the creation of six or seven new genera, and the unprecedented dismemberment of Ornithogalum’ and the need for a large number of combinations in ‘genera that are either new or have not been used in modern times’. In their opinion, all these ‘obligate’ new nomenclatural proposals would be a handicap to acceptance of such an arrangement, and hence it would not be supported by botanists (mainly South African), since it is more disruptive and appears to lack nomenclatural stability.

In their arrangement, the genera Albuca and Ornithogalum are hard to recognize, since they are broadly defined from a morphological point of view. However, they are divided in nine subgenera, 13 sections and 12 series, mostly on the basis of morphological affinities, although one (A. subgenus Albuca) is not supported as monophyletic in their trees (cf. Manning et al., 2009a).

Contrary to Stedje (2001b), the arrangement with many genera favoured here is much more consistent with morphology, and matches the molecular phylogenetic data. The new arrangement is not handicapped by any previous treatment and represents a logical circumscription of clades only when sufficient morphological evidence is combined with molecular data and monophyletic groups, as claimed by Manning et al. (2009a). Any a priori constraint (e.g. criticism by local taxonomists or apparent nomenclatural unstability) is unwarranted and counterproductive in taxonomy and systematics. Nomenclatural stability is warranted only by solid establishment of consistent taxonomic groups and a sensible application of rules in the International Code of Botanical Nomenclature (ICBN), and does not depend necessarily on taxonomic arrangements themselves. A proposed new arrangement will become stable only when it is based on a successful combination of phylogenetic, morphological and biogeographical information, and when it is comprehensive and overlooks taxonomic prejudices and/or regional botanical practices.

On the basis of these considerations, we accept 19 phylogenetically monophyletic groups, which are here treated as separate genera and which are supported by syndromes of morphological characteristics as shown below. The most informative phylogenetically characters correspond namely to the colour of, and presence of, a band on tepals, capsule shape, seed morphology and seed disposition in the capsule locules. Other characters such as perigone features, tepal fusion, disposition of leaves and bulb shape and size, are homoplasious in the subfamily. However, contrary to the arguments of Stedje (2001b) and Manning et al. (2009a), they are useful for taxonomy and do not generate inappropriate or unstable associations when they are properly combined with the other relevant features.

Homoplasy is common in most large families of monocots (e.g. Orchidaceae and Amaryllidaceae, among others), and Hyacinthaceae are no exception. Contrary to arguments of Stedje (2001b) and Manning et al. (2009a), broader generic circumscriptions make it more difficult to understand relationships among species groups, from morphological and phylogenetic points of view, since they underestimate the value of particular apomorphies. Recognition of groups which are too wide (such as Albuca and/or Ornithogalum, in Options 2 and 3 sensu Manning et al., 2009a), although monophyletic, only increases heterogeneity in each and causes difficulties in morphological definition. The existence of small genera (monotypic or with two or three species) is not necessarily an inconvenience; in contrast, when small clades are well defined morphologically, molecularly and phylogenetically, they can merit generic rank, depending on the general criteria used. Other subfamilies of Hyacinthaceae (e.g. Hyacinthoideae) are treated by Manning et al. (2004) in a narrower way, with ≤11 genera being recognized. Morphological syndromes characterizing genera in Hyacinthoideae are similar to those allowing recognition of 19 genera in Ornithogaloideae. The use of two different criteria in these close subfamilies appears to be inappropriate, since homoplasy is equally common in both groups. Therefore, the arrangement favoured here (with many genera, based on clades obtained in the combined trees) is more natural and easier to understand from a global view of the family. Furthermore, as asserted by Müller-Doblies (2006) ‘a splitting of Ornithogalum seems to be by far the better taxonomic solution of the molecular data than the sinking of Albuca and other genera into Ornithogalum, which results in a huge taxonomic dustbin’.

A discussion of clades accepted at the genus rank in the taxonomic arrangement follows below. Disagreements and/or incongruence with previous, recent comprehensive treatments (Speta, 1998a; Manning et al., 2004, 2009a) are also discussed.

The Pseudogaltonia clade

For a long time, Pseudogaltonia has been accepted as including a single species, P. clavata (Mast.) Phillips, distributed throughout arid regions of north-west South Africa (Richtersveld), Namibia and Botswana. It was first described within the genus Galtonia, which grows exclusively in precipitation-rich, summer-rainfall, high-elevation regions of the Drakensberg, Low Drakensberg, Southern Berg and Natal Midlands in the eastern provinces of South Africa (KwaZulu-Natal, Free State, Mpumalanga and Eastern Cape). However, most later experts on Hyacinthaceae recognized Pseudogaltonia as a separate genus (Kuntze, 1886; Durand, 1890; Phillips, 1935; Hilliard and Burt, 1988; Speta, 1998a, b; Manning et al., 2009a, b), based on biogeography and morphological features, including the large bulb with cartilaginous and fibrous outer tunics, the racemose pyramidal inflorescence, tepals fused to form a long tube (for about three-quarters of their length) and somewhat curved and swollen at the base (autapomorphy), stamens exserted and inserted at the mouth of the perianth tube (autapomorphy), style thin and >2 cm long (autapomorphy), capsule quadrate, truncate and deeply trilobate and uniseriate seeds with five to seven angled polygonal testa cells (see Phillips, 1935; Jessop, 1975; Speta, 1998a; Manning et al., 2009a, b). Recently, an additional species, P. liliiflora J.C.Manning & Goldblatt has been described (cf. Manning et al., 2009b) from Richtersveld, which is morphologically close to the former. Our phylogenetic studies provide further evidence for P. clavata being placed far from Galtonia as sister to the Dipcadi clade, as previously pointed out (Pfosser and Speta, 1999; Manning et al., 2004, 2009a; Wetschnig et al., 2007). Accordingly, Pseudogaltonia is here accepted at the generic rank.

The Dipcadi clade

Dipcadi has been widely accepted by most researchers on Ornithogaloideae (Heywood, 1980; Valdés, 1987; Speta, 1998a, b; Manning et al., 2002, 2009a). It is characterized by two autapomorphies (the usually unilateral raceme and tepals curved outwards in the apical part) and by other features such as the tepals fused into a tube approx. one-quarter to one-third (to two-thirds) of their length, stamens adnate to the tube and included, capsule quadrate in outline, as broad as or broader than long, truncate at the apex, and seeds discoid, flattened, horizontally stacked and uniseriate in each locule with five to seven angled polygonal testa cells (cf. Obermeyer, 1964; Jessop, 1975; Heywood, 1980; Valdés, 1987; Speta, 1998a; Manning et al., 2002, 2009a). It is important to note that some species of Albuca produce almost unilateral racemes due to the decumbent growth of scapes, a pattern which is quite different from that found in Dipcadi.

All the species of Dipcadi studied in these phylogenetic analyses form a fully supported clade, sister to Pseudogaltonia. These results are identical to those obtained by Pfosser and Speta (1999) and Manning et al. (2004). The former authors recognized both genera, whereas the latter authors combined them under Ornithogalum. Recently Manning et al. (2009a) reversed their previous opinion (cf. Manning et al., 2004) and recognized the generic identity of both Pseudogaltonia and Dipcadi. Based on the data presented here and on unequivocal morphological and phylogenetic evidence, the traditional concept of Dipcadi, following Pfosser and Speta (1999), is maintained. Manning et al. (2009a) suggested two series within Dipcadi, one including species with caudate apical appendages (D. series Uropetalon), and the other including those lacking these appendages (D. series Dipcadi). However, their phylogenetic tree does not provide support for that classification, since neither series was monophyletic. Our phylogenetic results also recovered the polyphyly of these series.

The Battandiera clade

Manning et al. (2004, 2009a) showed that three species of Ornithogalum occurring in the desert regions of Namibia and South Africa (O. seineri, O. pulchrum and O. stapffii) form a well supported clade, sister to the rest of Albuceae. This result is also found in our combined analysis. However, our ITS data (Fig. 1) placed this clade as sister to Dipcadi, with moderate support. This can be explained because all these latter taxa share the same type of discoid, flat seeds, horizontally stacked and uniseriate in each locule, with angled polygonal or slightly undulate testa cells (Fig. 4I and J), a rare character in the subfamily which supports their close phylogenetic relationship. The uncertain position of this group raises questions about the circumscription of Dipcadieae (sensu Manning et al., 2009a), and even the necessity for its segregation from Albuceae. In Fig. 2, these tribes are thus grouped as ‘Clade A + B’.

Open in a separate window

Fig. 4.

Morphological variation in several genera of Ornithogaloideae: (A) Honorius boucheanum (from Sturm, Deutschlands Flora in Abbildungen: tab. 33. 1796, as Ornithogalum boucheanum); (B) Nicipe juncifolia (from Jacquin, Plantarum Rariorum Horti Caesarei Schoenbrunnensis Descriptiones et Icones, 1: tab. 90. 1797, as O. juncifolium); (C) Elsiea corymbosa (from Leighton in J. S. African Bot. 10: 56. 1944); (D) Ethesia prasina (from Lindl., Bot. Reg. 2: tab. 158. 1816, as O. prasinum); (E) E. xanthochlora (from Pretoria National Herbarium PREART 0002737, as O. xanthochlorum); (F) Trimelopter unifolium (from Jacquin, Icones Plantarum Rariorum, 2(16): tab. 429. 1795, as O. fuscatum); (G and H) T. dyeri (pictures by A. P. Dold); (I) Battandiera amoena (picture by Stephen L. Jury); (J) capsule of B. amoena; (K) capsules and seeds of B. pulchra; (L) Stellarioides longebracteata; (M) S. sessiliflora; (N) Capsules and seeds of S. longebracteata; (O) capsules and seeds of S. sessiliflora.

All three species in this clade are characterized by tepals with an evident longitudinal band visible on both sides, the wide capsule, quadrate and truncate at the apex, with darker transverse nerves, and the peculiar seed testa, shape and disposition (cf. Leighton, 1945; Obermeyer, 1978; Manning et al., 2009a), similar to those of Dipcadi. In our combined phylogenetic trees, and after the expansion of sampling to include more Eurasian and north African species, O. amoenum (Fig. 4I and J) was found to be deeply embedded in this clade. It is a rare and rarely collected plant which comes from the desert regions of North Africa (Morocco and Algeria), and shares the habitat, capsule and seed morphology with O. seineri, O. pulchrum and O. stapffii. These peculiar features, different from all Eurasian species of Ornithogalum, led Maire (1926) to describe the monospecific genus Battandiera to include O. amoenum. Similarly, the Namibian and South African species of this clade were segregated in the past in different taxa (Table 2). In our phylogenetic studies, the clade O. seineri–O. amoenum receives strong support. This fact, together with a clear morphological and phylogenetic difference (cf. Obermeyer, 1978; Müller-Doblies and Müller-Doblies, 1996; Manning et al., 2009a) with regard to related taxa, leads us to recognize Battandiera as a separate genus, though with some necessary recircumscription. Some species of this group, including O. amoenum, O. stapffii, O. candidum Oberm., O. tubiforme (Oberm.) Oberm. and O. rautanenii Schinz, occurring mainly in the western arid regions of Africa (Namibia and Morocco–Algeria), have tepals fused in a short tube ≤5 mm long, whereas others such as O. seineri, O. donaldsonii (Rendle) Greenway and O. pulchrum have free tepals and grow mostly in the eastern regions of Africa (from western Namibia to Kenya). The existence of fused or free tepals in this genus can be explained by the former state being plesiomorphic in the subfamily and having evolved independently in several clades (Fig. 3C). The cited morphological and biogeographical differences are also supported by our phylogenetic analysis, since O. amoenum plus O. stapffii form a clade sister to the O. seineri plus O. pulchrum clade, both of which are strongly supported. Further studies, including wider sampling, are being carried out to clarify internal relationships among members of this genus.

Inclusion of Battandiera in Albuca (as A. subgenus Namibiogalum) as Manning et al. (2009a) suggested, results in a highly heterogeneous Albuca, which does not reflect the clear morphological divergence with highly informative phylogenetic characters (e.g. capsule shape, seed morphology and arrangement in the capsule). Moreover, that inclusion would not be possible on the basis of the ITS tree obtained, as stated above. Segregation of Battandiera reduces heterogeneity in genera of the subfamily (cf. Manning et al., 2009a), making Albuca more coherent. Moreover, circumscription of O. subgenus Namibiogalum sensu Manning et al. (2009a), with the inclusion of O. prasinum and O. toxicarium C.Archer & R.H.Archer, is not consistent with morphology (cf. Lindley, 1816; Archer and Archer, 1999) or our combined trees. Ornithogalum prasinum and O. toxicarium are not related to one another, and are allied to other groups as shown below.

The Trimelopter clade

The Trimelopter clade includes O. unifolium and O. etesiogaripense, two species occurring in arid regions of western South Africa and southern Namibia (Sperrgebiet). These taxa are differentiated from the rest of the subfamily by the presence of a single, elliptic to narrowly oblong leaf (exceptionally two or three leaves), usually flattened against the ground, an apomorphy of the group (cf. Manning et al., 2009a); the ovaries and capsules have three prominent carpels, each with two longitudinal keels down the back and a groove between (Fig. 4G and H); and the seeds are unequally compressed or semi-discoid (cf. Dyer, 1931; Leighton, 1944; Obermeyer, 1978; Manning et al., 2009a). Due to their morphological peculiarities, these plants have been treated at various taxonomic ranks (Table 2). Rafinesque (1837) described the genus Trimelopter, including only O. fuscatum Jacq. (=O. unifolium Retz.) (Fig. 4F), which Obermeyer (1978) included (as O. unifolium) in O. subgenus Urophyllon. Müller-Doblies and Müller-Doblies (1996) described O. subgenus Urophyllon section Monarchos to accommodate 11 related species. Finally, Manning et al. (2009a) recognized section Monarchos as Albuca subgenus Monarchos. The phylogenetic study by Manning et al. (2004, 2009a) found strong support for the clade of O. unifolium and O. etesiogaripense (97–100 % BS) and revealed its sister relationship with clades including Albuca, Coilonox and Stellarioides sensu stricto. Our phylogenetic study reveals the same relationship, also with a strong support (Fig. 2). Due to the clear morphological and phylogenetic identity of this clade, the genus Trimelopter is accepted here, following Rafinesque (1837), though it is expanded to include the species of O. section Monarchos (sensu Müller-Doblies and Müller-Doblies, 1996), with the exclusion of O. rotatum because of its disruptive morphology and phylogenetic position. This latter species is evidently connected to taxa in the Neopatersonia clade, as discussed later.

The Albuca clade

The well-established genus Albuca comprises at least 60 species, as traditionally circumscribed (Speta, 1998a), though this number will probably reach approx. 100. They are readily characterized by their floral structures, which clearly differentiate them from the rest of the subfamily. Flowers are nodding, patent or erect, and the tepals are usually dimorphic: the inner ones are erect and connivent, overtopping the stamen and gynoecium, whereas the outers are erect–patent. The filaments have basal constrictions (apomorphy), the ovary has paraseptal crests, the styles are thick and prismatic (autapomorphy) or narrower and tapering, the capsule is ovoid or oblong, acute or obtuse, and the seeds are flattened, semi-lunate or discoid with smooth or rarely granulate puzzle-like testa (cf. Baker, 1897a, 1898, 1904; Jessop, 1975; Müller-Doblies, 1994, 1995; Speta, 1998a; Manning et al., 1999, 2002, 2009a). Based on a combined trnL-F plus rbcL phylogeny, Manning et al. (2004) combined all these species into Ornithogalum. When they extended the sampling of taxa and sequenced other plastid regions, they considered the traditionally accepted species of Albuca as A. subgenus Albuca, though in their tree it was not clearly monophyletic, without the inclusion of the taxa of Coilonox.

In the Bayesian analysis, the present data provide a much better resolved tree, in which the species of Albuca form a clade sister to Coilonox. This fact, together with the remarkable morphological differences between the groups, supports recognition of Albuca at the generic rank, as suggested by Müller-Doblies (2006) and adopted here. Manning et al. (2009a) suggested four sections within Albuca subgenus Albuca, which we choose not to accept here because they have not been shown to be monophyletic. In the combined trees, species of Albuca form two strongly supported groups. One of these fits the concept of the former genus Pallastema Salisb., which is well characterized by the slender, almost cylindrical style, and the slightly zygomorphic flowers with all stamen similar and fertile. The other group includes a mixture of taxa formerly referred either to subgenera Albuca, Falconera and Leptostyla (cf. Baker, 1897a; Müller-Doblies, 2006), or to sections Albuca, Falconera or Branciona (cf. Manning et al., 2009a). However, none of these receives support in our phylogenetic trees, perhaps due to the still poor sampling of all these groups. Further research is currently being carried out to clarify this point.

The Coilonox clade

Some species of Ornithogalum (e.g. O. albucoides (Ait.) Thunb., O. secundum Jacq., O. suaveolens, etc.) were segregated in the past at different ranks (Table 2). At the generic rank, Coilonox (Rafinesque, 1837) has priority. Species of Coilonox possess some similarities with Albuca that led several authors to describe species of the former in the latter (cf. Manning et al., 2009a). Although in general terms Coilonox and Albuca share the longer than wide capsules, the style longer than the ovary and stamens, and the phenology of the flowers, closing at night (cf. Manning et al., 2009a), Coilonox does not show the floral structures typical of Albuca and hence is quite simple to differentiate (cf. Manning et al., 2009a). In fact, as Müller-Doblies (2006) pointed out ‘there are no real transitions between both genera’. Our phylogenetic results demonstrate the monophyly of both groups and, together with the clear morphological differences, support recognition of both genera, which can be easily differentiated by their floral structure.

The Stellarioides clade

Ornithogalum longebracteatum, O. sordidum and O. tenuifolium are characterized by large bulbs, sometimes epigeous, with papery and greyish or fleshy and green outer tunics. They usually produce aerial or hypogeous bulbils and have long, narrow and dense cylindrical inflorescences (Fig. 4L and M), flowers small and numerous, capsule subglobose or widely ovoid to obovoid, trilobate, and seeds flattened or irregularly compressed, with sharpened edges, obliquely stalked in each locule (Fig. 4N) and a puzzle-like testa (cf. Leighton, 1945; Obermeyer, 1978; Speta, 1998a; Manning et al., 2009a). In Africa, the species of this group are mostly distributed in areas characterized by abundant, mainly summer rainfall. They also occur in regions with a bimodal autumn–summer rainfall regime (O. tenuifolium) or in semi-arid Mediterranean-climate areas, ranging from high-altitude warm-temperate regions of highveld in South Africa, through the mountains of southern and East Africa as far north as Ethiopia and Morocco–Algeria (O. sessiliflorum Desf., Fig. 4M and O). In South Africa, O. longebracteatum is usually found in dense thickets and edges of both subtropical and afromontane forests (Manning et al., 2009a). All these species have been recognized at various ranks (Table 2), and Stellarioides (Medikus, 1790) has priority at the generic level. They share a unique syndrome of morphological characters, including the dense and many-flowered inflorescences, with small flowers (tepals ≤12 mm long), almost sessile to shortly/pedicellate and small capsules ≤1·4 cm long. The present results show that O. longebracteatum, O. sordidum and O. tenuifolium form a fully supported clade, sister to the Albuca–Coilonox alliance. With the cited morphological syndrome, this justifies acceptance of Stellarioides at the generic rank. In contrast, recircumscription of Albuca to include Stellarioides as suggested by Manning et al. (2009a) greatly increases the heterogeneity of Albuca.

Speta's (1998a, 2001) concept of Stellarioides is almost coincident with that accepted here, and not with that in the tree of Manning et al. (2009a), in which several taxa (e.g. O. etesiogaripense, O. haalenbergense, O. pulchrum, O. rotatum, O. seineri, O. stapffii and O. unifolium) are treated as members of Stellarioides, supposedly following Speta (1998a). Nonetheless, the treatment of Speta (2001) is in need of a few changes: first, the exclusion of S. donaldsonii (≡Albuca donaldsonii), which is included here in Battandiera, and, second, the exclusion of Trimelopter and Ardernia Salisb. from synonymy, which are synonyms and accepted here as a different genus. It is worth mentioning here that Stedje (2001b) used some taxa of the present clade to exemplify her criticism of Speta's treatment. In her opinion O. tenuifolium Redouté, O. sessiliflorum and O. narbonense where supposedly undistinguishable from a morphological point of view (only by minor quantitative characters), but apparently they should be included in three different genera (Stellarioides, Cathissa and Loncomelos, respectively) following Speta. However, evidence is here shown that the first two species indeed belong to Stellarioides, whereas O. narbonense is a member of Loncomelos (see below), all showing clear morphological differences and being placed in clades far apart from each other.

The Galtonia clade

Four species are traditionally included in the genus Galtonia (cf. Hilliard and Burt, 1988): G. candicans Decne., G. princeps Decne., G. regalis Hilliard & B.L.Burtt and G. viridiflora Verdoorn. They are found in summer-rainfall, high-altitude regions of the Drakensberg, Low Drakensberg, Southern Berg and Natal Midlands (South Africa and Lesotho; Hilliard and Burt, 1988; PRECIS database). Galtonia is characterized by large leaves sheathing the stem, racemose inflorescence, with flowers nodding, tepals fused into a campanulate tube for approx. half of their length, filaments cylindrical and adnate, capsule lanceolate cylindric and acute, and seeds irregularly compressed or flattened with smooth puzzle-like testa (cf. Verdoorn, 1955; Jessop, 1975, Hilliard and Burt, 1988; Riddles and Condy, 2001; Manning et al., 2009a). Because of the unique combination of morphological features, many authors (e.g. Decaisne, 1880; Verdoorn, 1955; Hilliard and Burt, 1986, 1988; Speta, 1998a, b; Riddles and Condy, 2001) recognized Galtonia as a genus distinct from Ornithogalum. Taxa belonging to Galtonia had been also classified as Hyacinthus (Baker, 1870) or Ornithogalum subgenus Galtonia section Galtonia (Manning et al., 2009a). In the early molecular phylogenetic work of Pfosser and Speta (1999), Galtonia appeared to be polyphyletic, since G. candicans was related to Stellarioides and Albuca and G. princeps and G. viridiflora fell as sister to Ornithogalum saundersiae, a species which has been included in O. subgenus Urophyllon (Obermeyer, 1978; Müller-Doblies and Müller-Doblies, 1996) or in the monospecific genus Zahariadia Speta (nom. nud.). Manning et al. (2004, 2009a) presented evidence for the monophyly of Galtonia, with the necessary inclusion of O. saundersiae. In their studies G. candicans, G. viridiflora and O. saundersiae formed a strongly supported clade, sister to O. xanthochlorum and O. haalenbergense. First, they (Manning et al., 2004) synonymized both genera, though later they (Manning et al., 2009a) segregated Galtonia as a subgenus of Ornithogalum. Our molecular data corroborate their later results and show Galtonia plus O. saundersiae to constitute a strongly supported clade, which is sister to that comprising O. xanthochlorum and O. haalenbergense. Connections between O. saundersiae and Galtonia are surprising, since O. saundersiae has a pseudocorymbose inflorescence, erect flowers and free tepals and filaments (cf. Leighton, 1945; Obermeyer, 1978), traits all different from typical Galtonia. Moreover, chromosome numbers support separation of these groups, being 2n = 16 in Galtonia (Speta, 1998a; Forrest and Jong, 2004; Wetschnig et al., 2007) and 2n = 12, 14 in O. saundersiae (Obermeyer, 1978; Speta, 1998a). However, some morphological features, including the large and sheathing leaves, fleshy tepals, acute capsules and irregularly compressed seeds and biogeography are shared. Although some plastid molecular trees (Pfosser and Speta, 1999; Wetschnig et al., 2007) showed both genera to be monophyletic, our combined results place O. saundersiae inside Galtonia. Therefore, until new molecular evidence is available, we suggest treating it as a member of Galtonia. Related to this matter, Hammet and Murray (1993) found that some tetraploid (2n = 32) cultivars of Galtonia candicans produced inflorescences up to 2 m, with erect flowers, somewhat similar to those of O. saundersiae, which support the discussion above.

The Ethesia clade

The genus Ethesia was described by Rafinesque (1837) to include O. prasinum as the sole species (Fig. 4D). Later, Obermeyer (1978) and Müller-Doblies and Müller-Doblies (1996) classified O. prasinum, O. xanthochlorum (Fig. 4E) and other related species within O. subgenus Urophyllon. Manning et al. (2009a) presented molecular phylogenetic evidence for a close relationship between O. haalenbergense, similar to O. prasinum (cf. Müller-Doblies and Müller-Doblies, 1996), and O. xanthochlorum, both forming a clade sister to Galtonia, and taxonomically interpreted as O. subgenus Galtonia section Xanthochlora. The present molecular phylogenetic study corroborates the results of Manning et al. (2009a). Both species are well differentiated from Galtonia by their smaller size; leaves lacking sheaths, sometimes not coinciding with the flowers; tepals greenish; filaments petaloid, white, showy and spreading; capsule obcordate or widely globose and truncate to retuse at the apex; and seeds compresso-angulose, ovate or discoid with one or two ribs and biseriate per locule (cf. Lindley, 1816; Leighton, 1945; Obermeyer, 1978; Manning et al., 2009a). Moreover, Galtonia and Ethesia have contrasting distribution areas, being found in summer-rainfall and winter-rainfall regions, respectively. Because of the differences in crucial morphological characters we recognize the genus Ethesia (Rafinesque, 1837), although with a minimal recircumscription. Inclusion of O. prasinum in Battandiera (=A. subgenus Namibiogalum) as proposed by Manning et al. (2009a) is not justified on the basis of the morphology discussed above (cf. Lindley, 1816). On the contrary, this species appears to be closely connected to O. haalenbergense (cf. Müller-Doblies and Müller-Doblies, 1996).

The Eliokarmos clade

Ornithogalum conicum Jacq., O. thyrsoides Jacq., O. dubium Houtt., O. maculatum and other related taxa (cf. Manning et al., 2007) form a well-characterized group, defined by their ciliate or fimbriate (rarely glabrous) leaves; membranous, ovoid, wide and petaloid bracts; large colourful flowers; concolourous, white, yellow, orange or reddish tepals, sometimes with basal or apical maculae; filaments usually expanded or winged; capsules lanceolate or fusiform, usually hidden among the dry perianth segments; and small abundant and irregularly compressed seeds (0·5–2·0 mm), with papillate to echinulate or puzzle-like testa (cf. Baker, 1897b; Leighton, 1944, 1945; Obermeyer, 1978; Manning et al., 2007). Some have great ornamental value and are widely cultivated (Obermeyer, 1978). These species have been recognized at different ranks (Table 2), with Eliokarmos, Tomoxis Raf. and Lomaresis Raf. being published first and simultaneously at the generic level (Rafinesque, 1837). Eliokarmos has been recently used by Speta (1998a) at that rank.

Species such as O. constrictum F.M.Leight., O. bicornutum F.M.Leight., O. hispidum Hornem. and other relatives produce capsules and seeds similar to those of section Aspasia, but they differ in the long sheathing and ascending leaves, with patent blades arranged at different levels and a tubular cataphyll along the stem, usually not coinciding with the flowers, bracts narrower and apiculate or aristate, flowers smaller, and tepals usually having a darker longitudinal band only apparent on the abaxial side (cf. Obermeyer, 1978). These species were included in O. subgenus Aspasia group Hispidae (cf. Obemeyer, 1978) or in O. section Hispidaspasia (cf. Müller-Doblies and Müller-Doblies, 1996).

The phylogenetic studies of Manning et al. (2004, 2009a) revealed that all these species form a strongly supported clade. In this clade, two well-supported subclades can be identified: one comprising O. maculatum and O. rupestre, and the other grouping the rest of species of sections Aspasia and Hispidaspasia. Our phylogenetic data confirm the findings of Manning et al. (2009a), which would support the independent status of Phaeocles Salisb. (Salisbury, 1866), including O. maculatum and O. rupestre, apart from the rest of section. These taxa are peculiar in their glabrous leaves, a rare character in the group. However, no other relevant differences to the rest of the related taxa seem to exist, and therefore we prefer to accept a single genus Eliokarmos (including Phaeocles) to group all taxa discussed in these paragraphs. This concept is similar to that proposed by Speta (1998a), with the exclusion of taxa belonging to O. group Angustifoliati (sensu Obermeyer, 1978) that is phylogenetically distant from this group and which we recognize as the genus Nicipe. Further studies and a more extensive sampling in this group are needed to shed more light on phylogenetic relationships within this clade and on eventual recognition of subgeneric ranks.

The Avonsera clade

Ornithogalum convallarioides is endemic to Madagascar and is characterized by its wide, pseudopetiolate leaves, loosely racemose inflorescence (usually with two flowers per bract in the lower part), small bracts; tepals completely white and connate at the basal part, filaments tapering, connate and adnate to the perianth, capsules retuse and trilobate and seeds large and ovoid (cf. Speta, 1998a; Wetschnig et al., 2007). The occurrence of two flowers per bract is an autapomorphy within Ornithogaloideae, though it is shared with the genus Oziroë Raf. (Guaglianone and Arroyo-Leuenberger, 2002), the only representative of Oziroëoideae (Speta, 1998a, b). The trilobate capsules have a peculiar structure with each lobe having a longitudinal central furrow (cf. Wetschnig et al., 2007; fig. 3·5), much resembling the six-lobed capsules of Ornithogalum sensu stricto. The genus Avonsera was described by Speta (1998b) to group O. convallarioides (the type species) and Resnova lachenalioides (Baker) Van der Merwe, arguing that it was the only genus having a multinerved perigone in Hyacinthoideae (the subfamily into which it was erroneously misplaced). Molecular phylogenetic studies by Wetschnig et al. (2007) and Manning et al. (2009a) demonstrated Avonsera (sensu Speta, l.c.) to be polyphyletic. Resnova species were true Hyacinthoideae, sister to species of Ledebouria, whereas A. convallarioides nested unequivocally within Ornithogaloideae. Therefore, Manning et al. (2009a) proposed creation of subgenus Avonsera within Ornithogalum. Our phylogenetic analyses show O. convallarioides to be an isolated clade, probably sister to the Mediterranean species of Ornithogalum plus those of Elsiea and O. subgenus Aspasia group Angustifoliati (Obermeyer, 1978), though its relationships are not fully resolved or supported. These morphological, phylogenetic and biogeographical (Malagasy endemism) characters warrant recognition of the so-far monotypic Avonsera as an independent genus.

The Neopatersonia clade

Schönland (1912) described the genus Neopatersonia to include N. uitenhagensis Schönland. Later, Lewis (1952) added N. namaquensis G.J.Lewis and N. falcata G.J.Lewis, to complete the three traditionally accepted species in the genus. As usually defined, Neopatersonia is characterized by the tepals fused in a short basal tube, filaments adnate and somewhat connate at the base (autapomorphy), stigma long trilobate and somewhat reflexed, capsule lanceolate, acute, and seeds subglobose or urceolate with pitted puzzle-like or granulate testa (cf. Schönland, 1912; Lewis, 1952; Jessop, 1975). Different classifications have treated them (Table 2) as Neopatersonia (Schönland, 1912; Lewis, 1952; Speta, 1998a, b), Ornithogalum subgenus Neopatersonia (Hutchinson, 1934), O. subgenus Ornithogalum section Neopatersonia (Manning et al., 2009a), or as members of a widely defined Ornithogalum (Manning et al., 2004). Speta (1998a, b) accepted Neopatersonia as a member of Hyacinthaceae, though with doubts about its possible inclusion in Ornithogaloideae. Molecular studies (Manning et al., 2004, 2009a) revealed a close relationship between Neopatersonia and O. rotatum, a species first included in O. section Monarchos (Müller-Doblies and Müller-Doblies, 1996), and recently placed in O. section Neopatersonia, according to molecular data (Manning et al., 2009a). In this analysis Neopatersonia plus O. rotatum appear as the sister group of the clade formed by the European species of Ornithogalum plus the African Nicipe. After combining the nuclear and plastid data, our phylogenetic analyses also show clear support for Neopatersonia in the traditional sense, to which O. rotatum is sister, being both sister to other European and African clades of ornithogaloids. Relationships of O. rotatum to Neopatersonia are supported by shared floral characters (e.g. tepals connate at the base, filaments connate and adnate), and inclusion of that species into Neopatersonia is therefore justified.

The Elsiea clade

Leighton and Salter (1944) described the genus Elsiea for peculiar plants growing in permanently wet habitats, with poorly developed and subrhizomatous bulbs, connate tepals and adnate filaments (Fig. 4C). Two species were first included in the new genus: E. corymbosa F.M.Leight. (currently O. esterhuyseniae) and E. flanaganii (Baker) F.M.Leight. (currently O. paludosum Baker). However, Obermeyer (1978) found this segregation unjustified, since she was not able to corroborate the fusion of perianth segments. Müller-Doblies and Müller-Doblies (1996) placed both taxa, together with O. flexuosum (Thunb.) U.Mull.-Doblies & D.Müll.-Doblies and O. dregeanum Kunth, in series Flexuosa of O. subgenus Urophyllon section Helogalum U.Müll.-Doblies & D.Müll.-Doblies, on the basis of the poorly developed bulbs. However, O. synanthifolium F.M.Leight., which belongs to O. subgenus Aspasia group Aspasiae, also shows this type of bulb and also occurs in permanently wet habitats (cf. Manning et al., 2007). Consequently, this bulb character appears to be a morphological convergence to that peculiar environment (cf. Manning et al., 2009a), and thus it is not suitable for systematics. Our phylogenetic studies confirm this idea, since O. esterhuyseniae falls in an isolated position in its own clade, far away from O. paludosum, which is placed together with taxa of O. section Linaspasia (the Nicipe clade). These results agree with those presented by Manning et al. (2009a), who proposed O. subgenus Ornithogalum section Elsiea to include only O. esterhuyseniae, and also demonstrate that proposals by Leighton and Salter (1944) and Müller-Doblies and Müller-Doblies (1996) are not supported from a phylogenetic point of view. However, phylogenetic divergence of these groups is reasonable on the basis of morphological and biogeographical evidence. First, O. esterhuyseniae (=Elsiea corymbosa) has a corymbose inflorescence, tepals white and connate, filaments adnate, ovary obovoid and stipitate, and capsule black (cf. Obermeyer, 1978; Manning et al., 2009a), and it is restricted to a small mountainous area in south-west South Africa. Secondly, O. paludosum, O. flexuosum and O. dregeanum have long and narrow racemose to spicate inflorescences, tepals and filaments free, ovary sessile, and capsule not black coloured (cf. Obermeyer, 1978), and occur in the eastern part of south Africa, as far north as Zambia and Malawi (cf. Obermeyer, 1978; Müller-Doblies and Müller-Doblies, 1996). The isolated phylogenetic position of O. esterhuyseniae, together with its characteristic syndrome of morphological features, warrants recognition of Elsiea as a monotypic genus.

The Nicipe clade

This clade includes O. juncifolium Jacq. (Fig. 4B), O. graminifolium Thunb., O. gracillimum R.E.Fr., O. sephtonii Hilliard & B.L.Burtt. and O. paludosum, a group of taxa growing in eastern and southern Africa and forming a strongly supported clade. They have been recognized in the past at different ranks (Table 2), including Nicipe (Rafinesque, 1837), O. group Angustifoliati (Obermeyer, 1978) and O. section Linaspasia, Vaginaspasia, Nanaspasia and Helogalum (Müller-Doblies and Müller-Doblies, 1996). Some morphological features characterize this clade, such as bulbs mostly with membranous tunics, usually ending abruptly in an epigeous neck, occasionally surrounded by tubular, transversally banded and epigeous cataphylls; leaves usually occurring with flowers, numerous, all emerging almost at the same level from the bulb neck, often filiform with enrolled margins, rarely flat and broader or even coriaceous and boat-shaped; inflorescence racemose, cylindrical to subcorymbose; bracts small and aristate, auriculate and sometimes denticulate at the base; tepals white to yellowish on the adaxial side, and usually with a dark longitudinal band on the abaxial one, almost free to fused ≤ one-third; capsule trigonous-elliptical to trilobate-turbinate, usually hidden among the dry perianth segments; and seeds variable in size and irregularly compressed (cf. Leighton, 1944, 1945; Obermeyer, 1978; Müller-Doblies and Müller-Doblies, 1996; Manning et al., 2009a).

According to this evidence, we accept this group at the generic rank, with the name Nicipe. As here circumscribed, this genus includes several groups that are quite well characterized morphologically and have peculiar distribution patterns. Species from eastern South Africa to Kenya (O. sephtonii and O. gracillimum) produce turbinate capsules, trilobate in section, and form a clade in our combined tree. Ornithogalum baurii Baker and O. diphyllum Baker share a similar fruit structure, as pointed out by Manning et al. (2007). Species from the western to eastern Cape (O. graminifolium, O. juncifolium and O. paludosum) also form a monophyletic group, and they have elliptical capsules, trigonous in section with blunt angles. Related to this group, species of O. series Zebrinella, such as O. monophyllum Baker, O. anguinum F.M.Leight. ex Oberm. and O. zebrinellum U.Müll.-Doblies & D.Müll.-Doblies show characteristic tubular, transversally banded cataphylls.

Nicipe as recognized here is mostly congruent with O. group Angustifoliati (Obermeyer, 1978), though it also includes O. pilosum L.f., a species Obermeyer placed in O. group Hispidae. That concept is supported by both molecular data (M. Martínez-Azorín et al., unpubl. res.) and morphology, as this species lacks the typical vaginate, superposed and clearly distant leaves characterizing the latter group, as pointed out by Müller-Doblies and Müller-Doblies (1996). Although the traditional concept of O. group Angustifoliati included plants with narrow to filiform leaves, Nicipe as treated here includes a wide range of leaf morphology, usually associated with particular habitats. Taxa such as O. baurii, O. pilosum, O. natalense Baker and O. aloiforme Oberm. have wide, oblong-lanceolate and flat leaves, whereas others including O. paludosum and O. flexuosum have flattened to terete, somewhat succulent leaves. These species grow mostly on moist soils or high mountain environments. Ornithogalum britteniae F.M.Leight. ex Oberm. is an extremely rare plant with evergreen, distichous, coriaceous, boat-shapped leaves with fimbriate margins resembling those of some Gasteria Duval, which grows in grassy karroid dwarf shrubland (Dold, 2003). Another group of species (O. annae-ameliae U.Müll.-Doblies & D.Müll.-Doblies, O. campanulatum U.Müll.-Doblies & D.Müll.-Doblies, O. synadelphicum U.Müll.-Doblies & D.Müll.-Doblies and O. verae U.Müll.-Doblies & D.Müll.-Doblies) have peculiar floral features, such as tepals fused, and filaments adnate to perigone and/or connate into a short tube. Connation of perigone segments is a plesiomorphy which has evolved several times in different genera of the subfamily (Fig. 3C), including Pseudogaltonia, Dipcadi, Battandiera, Galtonia, Avonsera, Neopatersonia and Elsiea. Furthermore, Battandiera and Galtonia include species with either connate or almost free tepals, as in Nicipe.

From a phylogenetic point of view, Nicipe is sister to the rest of Eurasian genera of Ornithogaloideae, and it could constitute the link with the lineages from southern Africa (cf. Manning et al., 2009a), where the origin of this genus would be placed (Feinbrun, 1941). Furthermore, the Kenyan species O. gracillimum provides evidence for this (Fig. 2). Migrations across East Africa would have connected floristic regions of both hemispheres in two different ways. On the one hand, the ‘arid corridor’ or ‘arid track’ migration hypothesis (cf. Verdcourt, 1969; Thulin, 1994; Jürgens, 1997; Linder et al., 2005; Hoyo and Pedrola-Monfort, 2006) and, on the other, the high-altitude corridor (cf. Linder et al., 2005; Devos et al., 2009), which jointly would have favoured disjunctions in genera such as Colchicum L. (including Androcymbium Willd.), Euryops (Cass.) Cass., Delosperma N.E.Br., Matthiola W.T.Aiton, Albuca, Battandiera, Dipcadi, Stellarioides and Nicipe itself.

The Cathissa clade

The Mediterranean species traditionally placed in O. subgenus Cathissa (O. concinnum Salisb., O. broteroi M.Laínz and O. reverchonii Lange ex Willk.) are characterized by their leaves glabrous, lanceolate to tapering; tepals white and ovoid or obovoid; filaments lanceolate to tapering; capsule ovoid to elliptical; and seeds irregularly flattened with papillate or puzzle-like testa (Figs 5A–C and ​and6A–C)6A–C) (cf. Zahariadi, 1980; Pastor, 1987; Moret et al., 1990; Martínez-Azorín et al., 2006, 2007). These species were placed in Cathissa (Salisbury, 1866; Speta, 1998a, b, 2001), a group treated as a genus, subgenus or section (Table 2). Our phylogenetic results confirm the monophyly of this group of three currently accepted taxa (Martínez-Azorín et al., 2007), which are well characterized morphologically and are restricted to the western part of the Iberian Peninsula and western Morocco. Acceptance of Cathissa at the generic rank is thus justified.

Open in a separate window

Fig. 5.

Inflorescence and flower features of most of the Eurasian genera of Ornithogaloideae: (A) Cathissa reverchonii; (B) C. concinna; (C) C. unifolia; (D) Melomphis arabica; (E) Loncomelos narbonense; (F) L. pyrenaicus; (G) Ornithogalum baeticum; (H) O. bourgaeanum; (I) O. divergens.

Open in a separate window

Fig. 6.

Capsule (lateral and section views) and seed features of most of the Eurasian genera of Ornithogaloideae: (A) Cathissa reverchonii; (B) C. concinna; (C) C. unifolia; (D) Melomphis arabica; (E) Loncomelos narbonense; (F) L. pyrenaicus; (G) Ornithogalum baeticum; (H) O. bourgaeanum; (I) O. divergens.

The Melomphis clade

In our phylogenetic analyses, O. arabicum, traditionally included in Melomphis (Rafinesque, 1837; Garbari et al., 2008), Caruelia Parl. (Parlatore, 1854), O. subgenus Caruelia (Parl.) Baker (Baker, 1872) or Myanthe Salisb. (Salisbury, 1866), appears as an independent clade, the position of which is not fully resolved as previously demonstrated by Manning et al. (2009a). Melomphis species are distributed in the Mediterranean basin eastwards, and are characterized by the corymbose to pseudocorymbose inflorescence, globose and shining black ovary, plain white tepals, globose and truncate to retuse capsule and irregularly compressed seeds with granulate testa (Figs 5D and 6D). Some authors have recently accepted the independence of this genus (Garbari et al., 2008). Although a wider sampling is needed to clarify relationships to other ornithogaloid clades, notable morphological differences exist to justify segregation of Melomphis at the generic rank. The globose and shining ovary of these species is similar to that occurring in the South African O. saundersiae (cf. Baker, 1891), here treated as Galtonia. Therefore, it can be interpreted as a morphological convergence between these groups, which are distant from each other in the phylogenetic analyses. Morphological features exist which allow easy differentiation of these groups, including tepals fleshy, reflexed and persistent for a long time in fruit, and capsule acute and ovate in O. saundersiae.

The Loncomelos clade

The species traditionally included in O. subgenus Beryllis are characterized by their concolourous leaves long and tapering; inflorescence racemose, much longer than wide; tepals white or rarely yellowish with a green longitudinal band on the abaxial side; filament expanded in the lower half; capsule trigonous with obtuse angles, ovate-lanceolate; and seeds biseriate, horizontally disposed and irregularly compressed, with papillate, rugose testa (Figs 5E and F and 6E and F) (cf. Baker, 1872; Zahariadi, 1980; Cullen, 1984; Wittmann, 1985; Feinbrun, 1986; Pastor, 1987; Moret et al., 1990; Rechinger, 1990; Speta, 1998a; Martínez-Azorín et al., 2010a). These plants have been placed in different taxa (Table 2), including the genus Loncomelos (Rafinesque, 1837). Our phylogenetic results support the monophyly of this group of plants. Ornithogalum narbonense L. and O. pyrenaicum L. form a strongly supported clade sister to Ornithogalum sensu stricto plus Honorius, which is clearly differentiated by the morphology of inflorescences, capsules and seeds. The inclusion of Loncomelos in the synonymy of Ornithogalum section Ornithogalum, as proposed by Manning et al. (2009a), generates a highly heterogeneous group mixing different capsule and seed morphology, characters that have been treated as highly informative phylogenetically. Therefore, we adopt the generic rank for Loncomelos, following the concept of Wittmann (1985) and Speta (1998a), on the basis of clear morphological and phylogenetic evidence. Similarly, segregation of Loncomelos has been accepted increasingly in recent times (cf. Speta, 1998a, b, 2000, 2006; Spies, 2004; Peruzzi et al., 2007; Ravenna, 2007; Garbari et al., 2008).

Maire (1958) included the north African O. sessiliflorum in O. subgenus Beryllis, mainly on the basis of its long and narrow inflorescence, whereas Stedje (2001b) connected it to Cathissa. However, this species shows other evident differences with regard to the rest of the Eurasian and north African taxa of both cited groups, such as the flowers with a green band visible on both sides of tepals, the subglobose capsule, and the compressed seeds obliquely stalked in each locule and with puzzle-like testa (Fig. 4M and O). All those features strongly connect it to Stellarioides, which led Speta (2001) to establish the combination S. sessiliflora (Desf.) Speta. The exclusion of O. sessiliflorum from O. subgenus Beryllis (=Loncomelos) gives homogeneity to the latter as regards to testa morphology (only granulate or rugose, never puzzle-like as reported by Moret et al., 1990).

The Honorius clade

Ornithogalum nutans L. is characterized by the racemose inflorescence; filaments tapering, with two apical lobes overtopping the anther; the capsule nodding at maturity (autapomorphy) and with six weak ribs in section and globose seeds with prominently reticulate testa (Fig. 4A) (cf. Baker, 1872; Zahariadi, 1980; Uysal et al., 2005). Due to these features, it was segregated at different ranks (Table 2), of which Honorius (Gray, 1821) has priority at the generic rank. Our phylogenetic results place O. nutans as sister to the Ornithogalum sensu stricto clade. Phylogenetic and morphological evidence justifies its segregation at the generic rank, as accepted by some recent authors (cf. Holub, 1976; Speta, 2000; Garbari et al., 2008). The recently described O. chetikianum Uysal, Ertuğrul & Dural, a Turkish taxon close to O. nutans, is said to have leaves with a longitudinal greyish band, which connects it to many representatives of Ornithogalum sensu stricto. A wider sampling of related taxa from Anatolia and the Middle East is still needed to improve our knowledge of phylogenetic relationships within Honorius.

This work was supported by the Ministerio de Educación y Ciencia (Spanish Government) (grant number AP2002-0214) and the Fundación Ramón Areces (Spain). We thank Laco Mucina (Curtin University of Technology, Perth) for his interesting biogeographical comments and corrections and additions to the early manuscript. We are grateful to John Manning (South African National Biodiversity Institute, Kirstenbosch) and Félix Forest (Jodrell Laboratory, Royal Botanic Gardens, Kew) for sharing their unpublished plastid sequences. M^a Dolores Lledó (Jodrell Laboratory, Royal Botanic Gardens, Kew) helped with laboratory procedures. Katherine Challis (Royal Botanic Gardens, Kew) kindly supplied information of protologues of several taxa. Lorenzo Peruzzi (Università degli Studi di Pisa) and M^a Montserrat Martínez-Ortega (Universidad de Salamanca) sent important samples for the phylogenetic analyses. Stephen L. Jury (University of Reading) sent pictures and vouchers of B. amoena. Nigel P. Barker (Department of Botany, Rhodes University, Grahamstown) and Tony P. Dold (Albany Museum, Schönland Herbarium, Grahamstown) provided valuable assistance with some South African genera. Two anonymous referees and the handling editor, Anne Brysting, made constructive criticisms that improved the text.