A revised infrageneric classification and synopsis of the Afro-Eurasian genus Moraea ( Iridaceae : Irideae )

Molecular phylogenetic studies of Moraea Mill. and the inclusion of Barnardiella Goldblatt, Galaxia Thunb., Gynandriris Parl., Hexaglottis Vent., Homeria Vent. and Roggeveldia Goldblatt in the genus have rendered the existing infrageneric classifi cation, dating from 1976, in need of substantial revision. In particular, subg. Moraea and subg. Vieusseuxia have been shown to be paraphyletic. We propose a new infrageneric classifi cation, based, as far as current data permit, on phylogenetic principles. Monophyletic subgenera and sections are circumscribed based on molecular phylogenies alone or in combination with morphological considerations. We recognize 11 subgenera, 15 sections and three series, arranged as follows in phylogenetic sequence: Plumarieae; Visciramosae (with sect. Multifoliae and sect. Visciramosae); Moraea (with sect. Moraea and sect. Polyphyllae); Galaxia (with ser. Unguiculatae, ser. Eurystigma and ser. Galaxia); Monocephalae; Acaules; Polyanthes (with sect. Serpentinae, sect. Deserticola, sect. Hexaglottis, sect. Gynandriris, sect. Polyanthes and sect. Pseudospicatae); Grandifl orae; Vieusseuxia (with sect. Integres, sect. Vieusseuxia and sect. Villosae); and Homeria (with sect. Stipanthera, sect. Flexuosae, sect. Homeria and sect. Conantherae). Most are moderately to well circumscribed at the morphological level either by fl oral or vegetative characters, except subg. Moraea, which includes a small number of unspecialized species apparently not linked by any apomorphic features. With over 27 new species described in the past 25 years and another 60 transferred to the genus, Moraea now includes 214 species. We provide a full taxonomic synopsis of the genus.


INTRODUCTION
Moraea Mill., the largest African genus of Iridaceae tribe Irideae, comprises ± 220 species (214 recognized species plus several more yet to be described).Following revisions of its southern and tropical African members (Goldblatt 1973(Goldblatt , 1976a(Goldblatt , 1977b)), Moraea was believed to be well understood and was considered to be monophyletic, largely on the basis of phenetic considerations and outgroup comparison.A close correlation between morphology and chromosome cytology in the genus, then including some 110 species, led Goldblatt (1976b) to propose an infrageneric classifi cation that used cytology as a major determinant in circumscribing subgenera and sections.Later research in Moraea and related genera led to the realization that Moraea as so circumscribed was paraphyletic when Barnardiella Goldblatt, Galaxia Thunb., Gynandriris Parl., Hexaglottis Vent., Homeria Vent.and Roggeveldia Goldblatt were recognized.A phylogenetic study of Moraea based on DNA sequence data subsequently confi rmed this conclusion, thus fully supporting its expanded circumscription (Goldblatt et al. 2002).This study and a second, more extensive one (Schnitzler et al. 2011) also showed that Goldblatt's 1976 infrageneric classifi cation of Moraea required considerable revision if a classifi cation follow-ing the principle of monophyly was to be implemented.In these molecular systematic studies some large species clusters comprised clades receiving moderate to strong support, but subg.Moraea included several disparate elements and required major restructuring to achieve a classifi cation consistent with the phylogenetic principle of monophyly.Some of the major changes include removal from subg.Vieusseuxia (D.Delaroche) Baker of sect.Polyanthes Goldblatt, which comprises a clade outside the subgenus and is sister to two lineages until now included in subg.Moraea.In addition, several individual species were shown by molecular, and sometimes cytological, data to be misplaced to subgenus or section.Re-evaluation of their morphology makes it clear that their position in the classifi cation must be revised.This paper provides a new classifi cation of Moraea based as far as possible on the principle of monophyly.Moraea now includes some 214 species, which we assign to 11 subgenera, 15 sections and three series.We also provide a full synopsis of the genus.
The molecular studies of Goldblatt et al. (2002) and Schnitzler et al. (2011) are summarized here and provide the framework for the revised classifi cation.Some 161 species were included in the latter study, and three additional taxa have been sequenced for the present Bothalia 43, 1 (2013) study, over 75% of the genus (Table 1).PCR amplification and sequencing were performed as described in Goldblatt et al. (2002).Insertions/deletions (indels) were coded as present/absent following the 'simple indel coding' method of Simmons & Ochoterena (2000), implemented in SeqState (Müller 2005).The fi nal matrix comprised 3 096 characters from three plastid markers (rbcL, rps16, and trnL-F) and 218 indels.Maximum parsimony analyses were performed using the heuristic search implemented in PAUP* (v.4.0b10; Swofford 2002).Initial searches were carried out using 1 000 replicates of random taxon addition and the tree bisection and reconnection (TBR) algorithm with equal character weights, retaining a maximum of 10 trees per replicate.The resulting trees were used as starting trees in a second search using the same parameters with a limit of 10 000 trees, which were then used to reweight the characters according to the rescaled consistency index (RC).Successive searches were performed using the reweighted matrix until tree lengths reached stationarity.We performed 1 000 bootstrap replicates using equal character weights and the TBR swapping algorithm, again keeping only 10 trees at each step.Maximum likelihood analyses were performed in RAxML (v.7.2.1;Stamatakis 2006) using the BINGAMMA function, with the alignment divided into partitions according to gene regions.This process implements the GTR+Γ model for each gene with individual estimation and optimization of model parameters and a discrete morphological model as proposed by Lewis (2001) for the indels, which is comparable to the Jukes-Cantor model of nucleotide substitution.We performed 500 rapid bootstrap searches (Stamatakis et al. 2008), followed by a thorough ML search on the original alignment.Bayesian phylogenetic inference was performed using MrBayes (v.3.2.1; Huelsenbeck & Ronquist 2001).The best-fi t models of nucleotide evolution were implemented according to the Akaike Information Criterion (AIC) scores for substitution models evaluated using MrModeltest (v.2.3; Nylander 2004).For binary traits, MrBayes implements an F81-like model.Three independent runs with four chains each were run for 30 000 000 generations, sampling the Markov chain every 1 000 generations.After removal of the fi rst 3 000 000 generations as burn-in, all runs were combined to build the consensus tree.The alignment and consensus trees are available from Tree-Base (accession number S13606).

DISCUSSION
All approaches resulted in highly congruent tree topologies.The phylogenetic tree presented here refl ects the topology and branch length of the Maximum Likelihood analysis.Unless indicated otherwise, support values reported in the text (e.g. 1, 95, 92) represent Bayesian posterior probabilities (if higher than 0.5), and bootstrap support values from the Maximum Likelihood and Maximum Parsimony analyses (if higher than 50%), respectively.Most of the major species clusters correspond to existing, named infrageneric taxa but their relationships to one another are often quite different from past interpretations of their affi nities.Not entirely surprisingly, several species of uncertain affi nity based on morphological considerations remain unresolved in the molecular analyses, notably M. cooperi, M. nubigena and M. papilionacea.The relationships of other species of which the affi nities were uncertain, e.g.M. fergusoniae and M. radians, are established in the trees generated.Conversely, a few species appear in the molecular trees in positions unacceptable on morphological grounds, especially M. inclinata and M. rivulicola.These exceptions are discussed in more detail below in our proposed classifi cation.1. Subg.Plumariae: Moraea lugubris is linked with low signifi cant statistical support (pp = 0.55, 67 BP(MP)) to the M. bubalina clade (Figure 1) in an association without morphological support.This taxonomically isolated species was retrieved as sister to all other species of Moraea in an earlier analysis (Goldblatt et al. 2002).M. lugubris is unique in Moraea in having coarsely netted corm tunics not matched elsewhere in Moraea, plumose style crests and stigmas located at the distal tips of the bifurcate style branches rather than as discrete lobes abaxial to the style branches.We recognize the monotypic subg.Plumariae for M. lugubris.Both the position of the stigmatic surfaces and the plumose style branches recall the genus Ferraria, which is well established as sister genus to Moraea (Goldblatt et al. 2002(Goldblatt et al. , 2008)).2. Subg.Visciramosae: the eight species of subg.Visciramosae (six included in the current phylogenetic analysis) are united by at least three morphological synapomorphies, notably sticky stems, unique brown, ± woody, longitudually grooved corm tunics with a sticky secretion on the inner surfaces, and stigmatic lobes with a central forked tooth.The ± free but closely contiguous fi laments of some species of the alliance are also unique in Moraea, but might constitute an ancestral state (other Moraea species have fi laments partially or completely united except in some species of the very derived M. tripetala complex and in M. thomasiae, both subg.Vieusseuxia).Ferraria, sister genus to Moraea, has partially united fi laments in all species.Floral mor- phology is unremarkable in subg.Visciramosae except in the species with free fi laments and in those with the style branches reduced and lacking crests and the inner tepals as well as the outer with nectar guides (a derived condition that recurs in several other clades).M. simplex stands out here in having fi liform style branches extending between the stamens, a condition shared by the unrelated M. fi stulosa and M. monticola (assigned here to sect.Pseudospicatae of subg.Polyanthes, although neither was included in the molecular analysis).
The members of subg.Visciramosae comprise two separate, well supported (both pp = 1 and > 90 BP) clades (Figure 1).One, including the multi-leaved Moraea bubalina and M. vespertina, is weakly linked with M. lugubris and the other, including the remaining species, which are characterized by presence of two foliage leaves, is placed as sister to the rest of the genus with only moderate support.The clear morphological synapomorphies that the two clades share (sticky stems, sigmoid branching, unique corm type, stigmatic appendages) provide adequate justifi cation for ignoring the weak molecular data in treating both lineages within a single subg.Visciramosa.We recognize the two lineages as separate sections, Multifoliae and Visciramosae.
3. Subg.Umbellatae: M. longifl ora (sect.Tubifl ora Goldblatt), plus several species from sect.Moraea (M.intermedia, M. margaretae, M. nana and M. umbellata) were retrieved as a well-supported clade (1, 99, 96; Figure 1) sister to M. maximiliani.The second member of sect.Tubifl ora, M. cooperi, was retrieved in an isolated but only weakly supported / unsupported (4-14 BP) position (Figure 1).We fi nd it most unlikely that the molecular topology refl ects the true relationships of M. cooperi as the morphological evidence linking it to the M. intermedia clade (i.e.subg.Umbellatae) is strong.Unique brown corm tunics, woody with plane surface (without the sticky secretion on the brown, longitudinally grooved corm tunics of subg.Visciramosa), darkly veined, pale yellow tepals, blunt outer infl orescence spathes usually not sheathing distally, and fl owers with a perianth tube are all evidently derived character states that are shared with M. longifl ora of the subgenus.Those features in M. cooperi may be homoplasious, but it seems to us unlikely that the entire set could have evolved independently.Thus we prefer to place M. cooperi in subg.Umbellatae rather than recognize it as a separate genus with the same morphological circumscription as subg.Umbellatae.We include eight species in what we designate subg.Umbellatae (one species, M. linderi, was not included in the molecular analysis).We note that M. maximiliani was retrieved as sister to the remaining members of the subg.Umbellatae clade and its position is weakly supported; nevertheless, it has the morphological synapomorphies of the subgenus hence its inclusion therein.We see no need for its recognition as a separate, monospecifi c subgenus or section in light of the morphology.
Floral diversifi cation in the subgenus parallels that in subg.Visciramosae, ranging from an unspecialized Moraea-type fl ower with well developed style branches and crests and larger outer tepals bearing nectar guides, to reduced style branches lacking crests (M.maximiliani, M. umbellata) and, in M. nana, to style branches represented by paired, fi liform arms extending on either side of the opposed stamen.The fl ower type in M. nana is convergent with that in the Hexaglottis group (subg.Polyanthes sect.Hexaglottis), in M. hexaglottis (not sequenced) and in M. pearsonii (now subg.Polyanthes sect.Pseudospicatae).

Subg.
Moraea: here much reduced in size, includes M. vegeta, type of the genus, plus M. gawleri and its close relative, M. vlokii, which together form a well-supported clade (1,100,99).The phylogenetic position of this clade (sister to the fi rst polytomy in the tree) was, however, not supported in the bootstrap analyses and thus remains uncertain.To this small alliance we provisionally add M. namaquamontana and M. indecora, a rare Namaqualand endemic, and not yet sequenced.
Moraea garipensis and M. ramosissima constitute another small clade (1, 100, 100) included by Goldblatt (1976bGoldblatt ( , 1986a) ) in subg.Moraea.A separate subgenus for the two species seems unwarranted as the molecular topology here is not well supported and we therefore provisionally retain them as a section of subg.Moraea pending additional molecular studies.At the morphological level, the species of subg.Moraea share palecoloured corm tunics, mostly of fi ne fi bres, except for M. garipensis and M. ramosissima.Further molecular analysis may show an alternative placement for these two species, the brown corm tunics of which recall those of subg.Umbellatae.
The rare Western Cape mountain endemic Moraea nubigena appears as isolated in the molecular phylogeny (Figure 1) as it does morphologically.The acaulescent habit, solitary leaf and unbranched stem are all distinctive but may be adaptations to its montane habit.Goldblatt (1986a) linked the species to M. lugubris, but that now seems unlikely.M. nubigena has a chromosome number of n = 10 and a fugaceous fl ower, both plesiomorphic in Moraea, and fi nely fi brous corm tunics, evidently an apomorphy for subg.Moraea but the blue perianth is otherwise unknown in the subgenus.We provisionally include it in subg.Moraea, with which the corm tunics are consistent, but it is not referred to any section.We note that it is the only blue-fl owered and single-leaved member of the group.An argument can be made for the alternative, a separate subgenus for M. nubigena, but we see no value in a monospecifi c subgenus for the species.5. Subg.Monocephalae: the lineage including Moraea angusta and two immediately related species is moderately well-supported (1, 89, 65) as one of a polytomy (Figure 1).The alliance (including M. vallisavium, not sequenced) is united by unbranched stems, a single foliage leaf inserted well above ground level, a terete leaf blade, obtuse to truncate infl orescence spathes, sticky nodes, a prismatic ovary, and fl attened, discoid seeds.The alliance was treated as the rank-less group Monocephalae by Baker (1896, then including M. spathulata of subg.Grandifl ora), and as subg.Monocephalae by Goldblatt (1976b).Although M. namaquamontana is retrieved as sister to the M. angusta clade in all molecular analyses, this association is without statistical or morphological support and there is little reason to doubt that M. namaquamontana is allied to M. gawleri (subg.Moraea), with which it shares a virtually identical fl ower, a distinctive, asymmetric corm, and globose capsules.The M. angusta alliance is phylogenetically isolated and well defi ned by several morphological synapomorphies, and we continue to accord it subgeneric rank.
6. Subg.Galaxia: all species previously treated as the genus Galaxia (Goldblatt 1984a) comprise a well-supported (1,93,75) clade.The alliance has several synapomorphies, not least an acaulescent habit, fl owers always with a perianth tube, style branches short and lacking crests, and an asymmetric corm, and we recognize it as subg.Galaxia.Relationships within subg.Galaxia are more complex than refl ected in Goldblatt's (1979b) recognition of two sections.Although sect.Eurystigma (species with entire style branches) is retrieved as monophyletic, sect.Galaxia (species with fringed style branches) is shown to comprise two lineages, the M. galaxia clade and M. kamiesensis plus M. fenestralis.A strictly phylogenetic classifi cation would thus require the recognition of three or four sections, but we do not feel that this is justifi ed in such a small group and suggest that the level of series is adequate.The position of the isolated M. kamiesmontana in the tree (Figure 1) as sister to the remaining members of the clade has morphological support; it is the only member of the subgenus that has clawed tepals, the plesiomorphic condition in Moraea.We thus admit three series, Eurystigma, Galaxia and Unguiculatae, this last monospecifi c. 7. Subg.Acaules: the species of the taxonomically isolated Moraea ciliata group were retrieved with strong support in the Bayesian analysis (pp = 1) as sister to the remaining species of Moraea and we accord the group subgeneric status.Within the group M. ciliata, M. macronyx and M. tricolor are obviously closely related, with the morphologically somewhat different M. falcifolia as sister.One more species belongs here, M. longipes, until recently included in M. ciliata (Goldblatt & Manning 2009).Synapomorphies for subg.Acaules include leaves clustered at the terminal node, stem usually below ground level at fl owering time (not M. longipes), sessile fl owers raised above the leaves and infl orescence spathes on an elongating, tubular stalk; possibly the sterile base of the ovary (called a contractile pedicel in older literature) that retracts after fl owering.The alliance was treated as sect.Acaules by Goldblatt (1976b), following Baker (1896).8. Subg.Polyanthes: the Moraea serpentina-M.tortilis species pair (1, 100, 100) is weakly placed as sister to the remaining species in the genus, but without statistical support.On morphological grounds, we include the six main clades (Figure 1) of this large group, including M. serpentine-M.tortilis, as sections of subg.Polyanthes.The alliance is united morphologically by distinctive blackish, wiry corm tunic fi bres, but the multi-leaved habit (reduced in some species to a single leaf) and fugaceous fl ower are plesiomorphic.We note that, in addition, many species of the subgenus have an apomorphic, included ovary, but the condition is reversed in two sections.The subgenus is mixed for chromosome number: a base number of x = 20 is ancestral and basic for two sections, Serpentinae and Deserticola (but reduction to x = 6 (or 5) has occurred within sect.Deserticola) and x = 6 is almost exclusive in sect.Hexaglottis, sect.Gynandriris, sect.Polyanthes and sect.Pseudospicatae (one species of sect.Polyanthes, M. inclinata has n = 6 and 11).
In the molecular analyses, Moraea rivulicola is retrieved in an unresolved polytomy in subg.Polyanthes, but without any statistical support.This species has all the morphological hallmarks of subg.Vieusseuxia as we noted in an earlier molecular analysis (Goldblatt et al. 2002) and sequences of a second and third sample of the species alters this topology hardly at all, ruling out experimental error.Another species that appears misplaced here on morphological grounds is M. exilifl ora, which almost certainly belongs in sect.Polyanthes according to morphology.We have no explanation for the apparently anomalous position of these two species in our trees, but to suggest that their morphology is somehow convergent with those clades to which morphology suggest they belong, is unacceptable.The remaining species of the clade, which comprise sect.Gynandriris (Goldblatt 1998), and originally the genus Gynandriris (Goldblatt 1980b), have apomorphic, translucent infl orescence spathes and a sessile ovary with a tubular, sterile beak, which leaves no doubt that it is a monophyletic alliance.The clade receives moderate support in the molecular analyses (69 and 51 BP, respectively; pp = 0.6, but excluding M. exilifl ora).Two species of sect.Gynandriris, M. sisyrinchium and M. mediterranea (= G. monophylla Boiss.& Heldr.ex Klatt and not sequenced), occur in the Mediterranean Basin and Middle East and constitute the only species of Moraea that occur north of the Sahara.Lewis (1959) and Goldblatt (1987)] has the derived, Hexaglottis-type fl ower with subequal, spreading tepals with short claws and style branches reduced to fi liform arms extended on either side of the opposed stamen.The inclusion of M. papilionacea in the clade (Figure 1) receives only weak support in the bootstrap analyses (55 and 42 BP, respectively) and has no morphological basis, but the affi nities of this species are uncertain and we prefer to retain it as unplaced in subg.Moraea, to which it was assigned by Goldblatt (1976bGoldblatt ( , 1986a)).

Sect. Hexaglottis [the genus Hexaglottis sensu
Sect.Pseudospicatae (1, 71, 51), sect.Deserticola (1, 80, 60) and sect.Polyanthes (1,91,78) are retrieved as monophyletic in the molecular analysis and are also morphologically coherent.The members of sect.Polyanthes sensu Goldblatt (1976bGoldblatt ( , 1986a) fall in two lineages; the fi rst (here sect.Polyanthes) mostly with multiple leaves and the second (sect.Pseudospicatae) with a single leaf.Both clades include species with typical Moraea-type fl owers as well as others with reduced style branches and style crests.We include M. inclinata in sect.Pseudospicatae without hesitation, despite its anomalous position in the tree, sister to M. cooperi plus the rest of the genus.It has a single leaf, blue fl ower, exserted ovary and capsule, and basic chromosome number of x = 6, all of which accord with this decision.It is allied to, and sometimes has been included in, M. natalensis of sect.Pseudospicatae.Although sect.Deserticola (unranked Subracemosae Baker) likewise has species with the typical Moraea fl ower and others with reduced style branches and crests (the M. rigidi-folia-M.herrei clade, BP 54), all members consistently have the ovary included in the spathes.The ovary is sessile in M. herrei and has a sterile beak, also present in a less developed state in M. rigidifolia and the M. fugax group (1, 96, 90).Sect.Deserticola closely resembles sect.Serpentinae in several morphological features and we would not be surprised if additional molecular data show them to be a clade.The available molecular data, however, preclude their union.
Subg.Polyanthes is sister to a polytomy formed by four large clades that include the remaining species in the genus.Two of these coincide exactly with subg.Grandifl ora sensu Goldblatt (1976b) and subg.Homeria sensu Goldblatt (1981); and we continue to treat them as subgenera rather than sections because of their large size.9. Subg.Grandifl orae: this subgenus (1, 87, 59) includes some 28 species of eastern, southern and tropical Africa.Most are tall, have unbranched stems and large, evidently unspecialized fl owers that are long-lived in almost all species.All have a solitary leaf, distinctive, fl attened, discoid seeds and a karyotype of large, subacrocentric chromosomes with a base number of x = 6.With only half its species sequenced for the molecular study, subg.Grandifl ora is still inadequately sampled and no clear morphological patterns are evident in the tree topology.10.Subg.Vieusseuxia: we treat the remaining two clades as constituting subg.Vieusseuxia, which has until now included sect.Vieusseuxia of Goldblatt (1976b) plus the monospecifi c sect.Integres.All species share a basic chromosome number of x = 6, but often a somewhat variable karyotype, although always of relatively large chromosomes.Members of the subgenus have derived, long-lived fl owers (a fugaceous fl ower is plesiomorphic in Moraea), a single foliage leaf, also derived (multiple leaves only in M. fergusoniae must be interpreted as a striking reversal), and inner tepals that are typically much reduced, often 3-lobed or even absent.
There is no morphological support for the two separate clades (1, 87, 60 and 0.79, 53, respectively; Figure 1), and there are few obvious subclades in subg.Vieusseuxia that correlate with morphological patterns.A separate analysis of the alliance including additional molecular markers is probably needed to better resolve the phylogeny.
Moraea fergusoniae (included in subg.Moraea by Goldblatt 1976b) is inconsistent in subg.Vieusseuxia in its several leaves, but is shown to be nested in the subgenus, where its basic chromosome number, x = 6, and trilobed inner tepals accord perfectly.We include it here without hesitation.The Cape species, M. thomasiae, separated by Goldblatt (1986a) as sect.Integres on account of its entire inner tepals, yellow fl owers (rare in the subgenus) and ± free stamens, is morphologically isolated in subg.Vieusseuxia and we continue to refer it to a separate section.In the tree (Figure 1) it is allied with a second winter-rainfall species, M. regalis (possibly misplaced here and without morphological support), plus all the eastern southern African (and summer rainfall area) species of subg.Vieusseuxia, which constitute a well supported (1,87,60) lineage.It is biogeographically signifi cant that these eastern species constitute a clade, thus representing a minor radiation of the genus in the region.
We segregate the remaining species of subg.Vieusseuxia (BP 98) in two sections based primarily on morphology.These are the small sect.Villosae, which includes the species clustered around M. villosa and which share outer tepal limbs as wide or wider than long and usually both leaves and stems pilose (not M. caeca).The remaining species fall by default in sect.Vieusseuxia: molecular analysis provides modest support for this alliance (1,73,62; excluding M. calcicola).The isolated position of M. algoensis (sister to sect.Vieusseuxia plus sect.Villosae ) and the position of M. barkerae in the M. villosa clade receive no morphological support and both are provisionally referred to sect.Vieusseuxia.11.Subg.Homeria: this exclusively southern African alliance is well supported (1,94,88) in the molecular analysis and is united morphologically by its hard, black corm tunic fi bres and by specialized fl owers in which the style branches and crests are reduced, the crests sometimes lacking, and the inner tepals are scarcely smaller than the outer (the Homeria-type fl ower sensu Goldblatt (1986c)).Both fugaceous and long-lived fl owers are present in the subgenus.Basic chromosome number is x = 6, but a few species have n = 5 or 4 (and 2n =9).Leaf number is reduced in some species from several to consistently one leaf, evidently the derived condition.The inclusion of M. radians in the subg.Homeria clade (originally a separate genus Sessilistigma) is consistent with Goldblatt's (1991) reduction of the genus in Homeria.M. fl exicaulis, originally assigned to a monotypic section of Moraea (Goldblatt 1982) is likewise retrieved within the subg.Homeria clade, which is consistent with its chromosome number and karyotype.We have no hesitation in including it in subg.Homeria.Its rather unusual morphology (the stem is strongly fl exuose and the fl owers have well developed style branches and crests) is somewhat anomalous for the subgenus, but recall the species pair M. knersvlaktensis and M. schlechteri (originally treated as Homeria sect.Namaquanae by Goldblatt (1981)) and we provisionally unite the three species in an enlarged sect.Flexuosae (the earliest name in Moraea for the section).There is no molecular support for the inclusion of M. louisabolusiae with these species as proposed by Goldblatt (1981, as H. bolusiae) based on morphology.We place M. louisabolusiae in sect.Homeria close to M. patens; a treatment consistent with both morphology and the molecular tree topology.Within the subgenus, the group of species around M. cookii also merits taxonomic recognition and we treat the alliance as sect.Stipantherae Goldblatt (1981).It then comprised fi ve species with the single leaf clasping the stem for some distance, fugaceous fl owers and, in four species, the fi laments free distally.Of the fi ve species M. britteniae and M. marlothii fall outside the clade and their continued inclusion in sect.Stipantherae must be viewed as provisional.
There is no molecular support for species groups based on fl ower shape, namely those with very short tepal claws clasping the base of the fi lament column (e.g.sect.Conantherae including Moraea bifi da, M. miniata, M. pendula and several more) versus those with longer claws forming a wide, deep bowl (e.g.sect.Homeria, including M. collina, M. minor, M. ochroleuca, M. pyrophila and others).In fact, apart from three members of sect.Stipantherae and two of sect.Flexuosae, there are no signifi cant species alliances retrieved in the molecular trees that merit taxonomic recognition and we fall back on the existing classifi cation based on morphology in which two more sections were recognized (Goldblatt 1981 Flowers pink with white cup; tepals prominently clawed; fi laments partly united; style branches with stigma lobes entire. Corm tunics mostly of hard, black fi bres.Foliage leaves several to solitary.Flowers fugaceous, mostly blue to mauve, also yellow or white; stamens united in lower half; ovary often included, style branches well developed with prominent crest or much reduced, sometimes as paired or single fi liform arms.Capsules oblongellipsoid, sometimes beaked, exserted or included.Basic chromosome number x = 10 but four sections have x = 6.Plants unbranched or branched, branches sometimes sessile.Foliage leaves solitary (sometimes 2 in M. fugax group).Flowers usually blue to mauve or violet (sometimes yellow or white); tepal claws short or ± as long as limbs; stamens united in lower half; style branches well developed with erect crests or reduced, without crests (as paired or fi liform arms extended horizontally in M. hexaglottis).Ovary and capsules included, sometimes beaked.
Plants unbranched or branched, secondary branches sometimes sessile.Foliage leaves solitary (2 in M. callista, which may be misplaced here).Flowers usually blue to mauve or violet; tepal claws short or ± as long as limbs; stamens united in lower half; style branches well developed with erect crests or reduced, without crests, sometimes as paired or single fi liform arms extended horizontally.Ovary and capsules exserted or included.

FIGURE 1 .
FIGURE 1.-Maximum Likelihood phylogeny of Moraea.Support values at nodes are Bayesian posterior probabilities, and bootstrap values from the Maximum Likelihood and Maximum Parsimony analyses respectively if pp > 0.5, or BS > 50%.The subgeneric classifi cation is indicated by black bars on the right.

TABLE 1 .
Schnitzler et al. (2011)es sequenced in this study with voucher information and GenBank/EMBL accession number for each gene region.Accession numbers for taxa not listed here can be found inGoldblatt et al. (2002)andSchnitzler et al. (2011).