Molecular phylogenetics of subfamily Urgineoideae (Hyacinthaceae): Toward a coherent generic circumscription informed by molecular, morphological, and distributional data

The taxonomy and systematics of Urgineoideae (Hyacinthaceae) have been controversial in recent decades, with contrasting taxonomic treatments proposed based on preliminary and partial studies that have focused on morphology and/or solely plastid DNA sequence data. Some authors have recognized only two genera, with a very broadly conceived Drimia , while others have accepted several genera that, although better de ﬁ ned morphologically, were doubtfully monophyletic. Here, we present phylogenetic analyses involving four plastid DNA regions ( trnL intron, trnL ‐ F spacer, matK , and the trnCGCA ‐ ycf6 intergenic region), a nuclear region ( Agt1 ), and a selection of 40 morphological characters. Our study covers 293 samples and ca. 160 species of Urgineoideae (ca. 80% of its global diversity). Bayesian inference, maximum likelihood, and maximum parsimony analyses were performed to derive the phylogenetic patterns. The combination of data yielded phylogenetic trees with 31 well ‐ de ﬁ ned clades or lineages, most corresponding to previously described genera, although some have required description or revised circumscription. As with other monocot families, a considerable degree of homoplasy was observed in morphological characters, especially in those groups with unspecialized ﬂ owers; nonetheless, consistent syndromes of traditional and novel characters are shown to support clade recognition at genus rank. The forthcoming revised classi ﬁ cation of Urgineoideae is outlined here.

The taxonomy and systematics of Hyacinthaceae have been the focus of numerous studies in recent decades (Speta, 1998a(Speta, , 1998b(Speta, , 2001Pfosser & Speta, 1999, 2001, 2004Manning et al., 2004Manning et al., , 2009Martínez-Azorín et al., 2011;Pfosser et al., 2012;Manning & Goldblatt, 2018; among others) that have generated considerable controversy regarding generic limits. The accounts of Speta (1998aSpeta ( , 1998bSpeta ( , 2001 and Pfosser & Speta (1999) combined morphological, anatomical, and molecular studies to substantiate the description of the monophyletic urgineoid genera Boosia Speta, Charybdis Speta, Duthiea Speta, Ebertia Speta, Geschollia Speta, Indurgia Speta, Ledurgia Speta, Rhadamanthopsis (Oberm.) Speta, Sekanama Speta, and Urginavia Speta, each showing a unique combination of morphological characters (Fig. 1), distribution, and evolutionary history. Shortly thereafter, Manning et al. (2004) extended the study to a second plastidial region (rbcL and trnL-F), but with limited sampling. These authors opted for a very broad Drimia, which was congruent with the whole subfamily Urgineoideae (except for Bowiea with one to two species) and accordingly synonymized all other urgineoid genera. In the process, extreme variation in morphology was included in Drimia, as reflected in the recent revision of the genus in southern Africa (Manning & Goldblatt, 2018), which delimited 20 sections. These sections generally align with earlier described urgineoid genera, with some sections circumscribed as para-or polyphyletic, as shown by prior phylogenetic accounts (Pfosser & Speta, 1999, 2001, 2004Pfosser et al., 2012). This challenge in terms of delineation can be resolved through phylogenetic analyses, such as the present study, as when the clades resolved include taxa with a homogeneous morphology, they can be recognized as distinct genera. When both quantitative and qualitative characters of flower morphology are combined with those of fruits, seeds, and vegetative forms, clades or lineages can be accepted at genus rank, as shown earlier by Martínez-Azorín et al. (2011) for the sister subfamily Ornithogaloideae.
Here, we explore an alternative classification that leads to an improved and communicable understanding of generic boundaries in the Urgineoideae, based on the combination of molecular and morphological data sets. With substantially more comprehensive sampling than has previously been achieved, our intention here is to identify clades or lineages showing clear morphologies and distribution. Our taxonomic studies have informed the recent description of eight new genera in the subfamily based on their unique morphological syndromes (Martínez-Azorín et al., 2013d, 2018a, 2019bPinter et al., 2013Pinter et al., , 2019Speta, 2016;Crouch et al., 2018). A new classification that focuses on generic circumscription in the Urgineoideae is forthcoming.

Taxon sampling
The phylogenetic analyses cover a total of 293 ingroup samples belonging to ca. 160 species from all previously recognized taxonomic groups in Urgineoideae (Table S1). Two samples of Whiteheadia bifolia (Jacq.) Baker (subfam. Hyacinthoideae) are used as an outgroup. Fresh material and herbarium specimens conserved at ABH, B, BLFU, BM, BOL, E, G, GRA, GZU, HAL, K, L, LINN, NU, NY, M, MO, NBG, P, PRE, S, SAM, TCD, UPS, W, WIND, WU, Z, ZSS, and ZT (acronyms follow Thiers, 2022) were used for morphological studies, including numerous samples currently kept in the living collection at the University of Alicante. These materials cover the full range of morphological and taxonomic variations of the Urgineoideae across its entire distribution range in Europe, Africa, and southwestern Asia. For a more detailed description of the materials and methods used for previous morphological studies undertaken on Hyacinthaceae, see Martínez-Azorín et al. (2007, 2011. Relative to data informing previous phylogenetic studies (e.g., Pfosser & Speta, 1999, 2001, 2004Manning et al., 2004;Pfosser et al., 2012), the sampling has been extended in the present work by 140-270 samples and 120-140 taxa, now covering ca. 80% of the total accepted species for the subfamily. A total of 994 new DNA sequences were obtained for this study, which were combined with some other previously published sources and accessions available from the GenBank. Plant source information and GenBank accession numbers are shown in Table S1. Author names of taxa cited in the text and tables align with IPNI (2022). Generic circumscription follows Crouch et al. (2018), Knirsch et al. (2015), Martínez-Azorín et al. (2013d, 2018a, 2019a, 2019b, 2019c, 2019d, and Pinter et al. (2013Pinter et al. ( , 2019Pinter et al. ( , 2020. Some species names used for BTU samples provided by U. Müller-Doblies & D. Müller-Doblies are inedited (ined.) and represent undescribed taxa. Other names used in this work (including the ones in the phylogenetic trees) correspond to inedited combinations, to accommodate particular taxa in the accepted taxonomic  Table S1. These unpublished combinations will be presented in a forthcoming monograph (Martínez-Azorín et al., in prep.). The nomenclature conventions and procedures follow the Shenzhen Code (ICN; Turland et al., 2018).
The length of PCR fragments was verified on a 1% agarose gel by electrophoresis. Successful PCR products were analyzed at both the technical services of University of Alicante and the company Macrogen SPAIN.
2.3 Sequence alignment and data sets Sequencher 4.1 (Gene Codes Corp., Ann Arbor, MI, USA) was used to assemble complementary strands and verify software base-calling. Sequence alignment was performed using MUSCLE (Edgar, 2004), conducted in MEGA X (v. 10.0.5) (Kumar et al., 2018) with minor manual adjustments to obtain the final aligned matrix (Material S10). The individual marker trees were studied for incongruence before combining data sets. The aligned matrix for the ycf region included 290 samples and 1475 bp, the matK matrix 294 samples and 908 bp, the trnL-F matrix 272 samples and 1200 bp, and the Agt1 matrix 166 samples and 1090 bp, although some samples of the latter region showed a hypervariable central region including large polymorphic insertions that did not allow proper alignment. Therefore, this central region was removed for those samples in the phylogenetic analyses, leaving a common 486 bp (the first 266 bp and the last 220 bp).
Molecular data sets were produced for individual makers and their combination, corresponding, respectively, to (1) the concatenated plastid (four regions) data matrix and (2) the concatenated molecular (plastid plus nuclear) sequence data matrix.

Morphological data
A selection of 40 discrete morphological characters was used to explore relationships with the phylogenetic findings as follows: 1. Bulb scales (0. Compact; 1. Loose); 2. Cataphylls (0. Leaves lacking sheathing cataphylls with transversal dark or prominent ribs; 1. Leaves surrounded at the base by sheathing cataphylls with transversal dark or prominent ribs); 3. Leaf number (0. More than one per bulb; 1. One per bulb; 2. Absent in old plants); 4. Leaf curving (0. Straight or slightly curved; 1. Distinctly coiled distally); 5. Leaf section morphology around the middle portion (0. Flattened or canaliculate; 1. Round to hemicircular); 6. Leaf proportions (0. Clearly elongated, from three to many times longer than wide; 1. Suborbicular to ovate, up to two times longer than   (Table S2, Materials S12). These data were incorporated into Mesquite v. 3.61 (Maddison & Maddison, 2022) to trace the morphological characters in relation to the obtained phylogenetic trees. The 40 characters were traced individually to evaluate apomorphies and ancestral characters for the obtained clades (Materials S13). Additionally, all characters were traced simultaneously on the phylogenetic trees (not shown).

Phylogenetic analyses
First, bayesian inference (BI) analyses were conducted using MrBayes v.3.2.7 (Ronquist et al., 2012) for individual markers and the combined data sets. To determine the best model of DNA substitutions for each independent region, jModelTest 2. 1.10 (Darriba et al., 2012) was performed, using the Akaike Information Criterion (AIC; Akaike, 1974); models with the lowest BIC (Bayesian Information Criterion) score were considered to best describe the substitution pattern. The best models were the most-parameterized models, all with a Gamma distribution (G parameter). The best model for the ycf and trnLF markers was (T92) Tamura-nei (coded as nst = 6, rates = gamma), for matK, the (GTR) General Time Reversible model (coded as nst = 6, rates = gamma), and for the Agt1 marker, the (HKY) Hasegawa-Kishino-Yano model (coded as nst = 2, rates = gamma). A partition was set to run each marker with the determined rates. For BI analysis, the Markov Chain Monte Carlo (MCMC) algorithm was run for 10 × 10 6 generations and sampled every 1000 generations for all individual analyses. Two runs were executed. The first 25% generations (burninfrac = 0.25) were excluded, and the remaining trees were used to compile a posterior probability (PP) distribution using a 50% majority-rule consensus. Additionally, two Bayesian analyses were performed combining the molecular (plastid and full molecular) data set with the 40 coded discrete morphological characters, indicating a mixed data type (DNA for the molecular and Standard for the morphological characters with a gamma rate), following the same criteria specified above for the molecular data. The results of the Bayesian analyses are shown for the concatenated plastid regions (Fig. S1) and the full molecular data set (plastid plus nuclear regions) (Fig. 2), indicating posterior probability (PP).
Second, phylogenetic analyses of the two molecular databases were obtained with maximum likelihood (ML) (Felsenstein, 1981) and maximum parsimony (MP) (Nei & Kumar, 2000), using the model indicated previously and applying, in all cases, partial deletion, as implemented in MEGA. ML analysis was conducted using the tree searching strategy based on the nearest neighbor interchange (NNI). MP analysis was performed using the Heuristic search options using the tree searching strategy based on the Subtree-Pruning-Regrafting (SPR) algorithm with search level 1 (Nei & Kumar, 2000), in which the initial trees were obtained by the random addition of sequences (10 replicates). For ML and MP methods, support was assessed by the bootstrap (Felsenstein, 1981), with 10 000 replicates holding ten trees per replicate. Clades showing bootstrap (BS) values of 50%-74% were considered as weakly supported, those with values of 75%-89% were considered as moderately supported, and those with values of 90%-100% were considered as strongly supported. www.jse.ac.cn Fig. 2. Bayesian majority-rule consensus tree of the full combined chloroplast (trnL intron, trnL-F spacer, matK, and trnCGCA-ycf6 intergenic region) and nuclear (Agt1) data sets for Urgineoideae; posterior probabilities (PP) are displayed at the nodes, and the proposed generic classification is indicated by shaded rectangles.

Congruence between data sets
Topological incongruence between cpDNA (trnL intron, trnL-F spacer, matK, and the trnCGCA-ycf6 intergenic region) and nDNA (Agt1 region) data sets were checked using two methods. First, an incongruence length difference (ILD) test (Farris et al., 1994) was performed in PAUP v.4.0.b10 (Swofford, 2002) using heuristic search options, which included 100 random addition replicates and tree-bisectionreconnection (TBR) branch swapping with MulTrees in effect and keeping 10 trees per replicate. Second, comparison of the ML phylogenetic trees of individual cpDNA and nDNA data sets was performed using MEGA, with the substitution model T92 + G (Tamura, 1992) as selected in the jModelTest, and with 1000 fast bootstrap replicates. A tanglegram comparing the ML tree of each data set was computed in Dendroscope 3.7.5 (see Huson & Scornavacca, 2012) and checked for topological conflicts on the basis of BP support ≥85% (Norup et al., 2006), but also BP support ≥75% to detect further relationships. Comparison of the combined cpDNA tree and the nDNA tree was undertaken from a reduced matrix including only the 166 taxa for which both kinds of data were fully available.

Distribution patterns
We undertook a comparative study of the phytogeographic distribution patterns of the accepted genera, some of which are newly circumscribed, to further inform the delineation of genera in respect of sister or related clades (Fig. 3). To achieve this, we subdivided the original Sudano-Zambezian Region of Takhtajan (1986), applying his principle of subdivision into subregions, but with one notable exception, which is, further subdivision of the Zambezian Subregion into three informal "sections": northern (precipitation-rich, transitional toward the tropical regions of Central Africa), eastern (precipitation-rich, including a series of highelevation mountain ranges), and southern (drier, home to arid and semi-arid savanna). The following resulting phytochoria were used to aid comparative analyses of the distribution maps of the genera: 1. the Cape Region; 2. the Karoo-Namib Region; 3. the Uzambara-Zululand Region; 4. the Madagascan Region; 5a: the southern section of the Zambezian Subregion; 5b. the northern section of the Zambezian Subregion; 5c. the eastern section of the Zambezian Subregion; 6a. the Erithraeo-Arabian Subregion; 6b. the Omano-Sindian Subregion; 7. the Guineo-Congolian Region; 8. the Sahelo-Sudanian Subregion; 9. the Saharo-Arabian Region; 10. the Mediterranean Region; 11. the Macaronesian Region; 12. the Indian Region; and 13. the Indochinese Region.

Results
Analyses of each individual matrix using BI, MP, and ML methods yielded trees with similar major topologies and support in most branches, resolving similar clades that are assimilated to genera in this work (see the Supplementary Material for BI Figs. S2, S3, S4, S6). However, early diverging relationships are sometimes collapsed or weakly supported in the individual DNA region trees. When the full cpDNA (Figs. S1, S5) and full molecular data sets (Figs. 2, S7) were analyzed, resolution improved considerably.
Use of the ILD test indicated the existence of slight incongruencies between plastid and nuclear data sets (P = 0.01), whereas comparison of individual ML trees of cpDNA and nDNA data sets yielded no remarkable conflicts (taking into account that most of the primary branches obtained in the nDNA analyses were weakly supported in unresolved positions). Consequently, as no major differences were found in the topologies of all obtained trees, and in view of the argument that combining heterogeneous data can also increase accuracy even if ILD analyses do not explicitly incorporate that heterogeneity (see Barker & Lutzoni, 2002), we accept that phylogenetic trees obtained from the combined molecular matrices (both the concatenated plastid regions and mostly the concatenated plastid plus nuclear regions) are good reconstructions of the evolutionary history of Urgineoideae. Further, our trees accord with previous partial phylogenetic analyses of the subfamily (Pfosser & Speta, 2001, 2004Pfosser et al., 2012). Notably, some authors have long disregarded ILD as an appropriate tool for testing the suitability of data set concatenation (Yoder et al., 2001;Pirie, 2015).

Molecular phylogenetic trees
We obtained comprehensive data for the plastid regions trnCGCA-ycf6 intergenic region, matK, and the trnL intron plus trnL-F spacer of the studied samples, with 290, 294, and 272 DNA sequences, respectively. The obtained phylogenetic trees, based on BI from each independent plastid region, provided several well-supported clades, although generally with inadequate support to explain their relationships (Figs. S2-S4). Concatenation of all plastid DNA regions generated an aligned matrix of 295 samples and 3583 characters. The Bayesian majority-rule consensus tree of the concatenated plastid regions is shown in Fig. S1.
MP and ML analyses from concatenation of all plastid DNA regions recovered a very similar general topology of the trees and generic relationships. However, the topology of the parsimony strict consensus tree (Fig. S5) resolved the polytomy of Thuranthos, Ledurgia, and Zingela, where the latter two genera form sister clades.
Amplification of the nuclear Agt1 region produced 166 Urgineoideae sequences covering nearly all recognized genera in the subfamily (except for Mucinaea and Triandra) and yielded useful data for phylogenetic studies. The obtained phylogenetic tree based on BI for the final Agt1 matrix (486 bp after removing the hypervariable central region) provided several well-supported clades, although generally with inadequate support to explain their relationships (Fig. S6). Concatenation of all plastid and nuclear DNA regions generated an aligned matrix of 295 samples and 4069 bp. The Bayesian majority-rule consensus tree of the concatenated plastid and nuclear regions is shown in Fig. 2 MP and ML analyses from concatenation of all molecular data sets yielded a similar topology to the Bayesian tree. However, the Parsimony analysis recovers all studied samples of Rhadamanthopsis plus Aulostemon as monophyletic (86% BS), Drimia as sister to Litanthus plus Schizobasis (77% BS), and again resolves the polytomy of Ledurgia, Thuranthos, and Zingela (Fig. S7). Moreover, Zulusia and Sekanama are not related within the general polytomy, a sample of Iosanthus sp. from central Namibia is placed out from the other samples of the genus, although within very weakly supported relationships, and Striatula is sister to Tenicroa, albeit with very low support (60% BS) (Fig. S7).

Combined molecular and morphological data sets
The Bayesian analyses combining the molecular (plastid and full molecular) data sets with the 40 coded discrete morphological characters are shown in Figs. S8 and S9, respectively, and yielded similar general topologies to the plastid and full molecular data sets alone, although with improved resolution in some clades. When the plastid data set was combined with morphological characters (Fig. S8), Rhadamanthopsis in the sense of this work obtained full support (1.00 PP) relative to the studied sample of Aulostemon. However, this solution is not recovered when the full molecular data sets were combined with morphology (Fig. S9). In this latter analysis, Drimia (1.00 PP) and Squilla (1.00 PP) appear as sister clades with a combined support of 0.85 PP, and the polytomy of Ledurgia, Thuranthos, and Zingela (Fig. 2) is dissolved, with the latter two genera in combination forming a clade with moderate support (0.83 PP).
Bowiea is consistently retrieved as sister to the remaining Urgineoideae and is identified by its distinctly branched and fleshy inflorescence, long-lasting flowers with tepals that remain at the base of the mature capsule, the conical ovary with a semi-inferior appearance ( Fig. 1.4), and green pedicels supporting dry dehisced capsules. Sister to the remaining Urgineoideae, the Rhadamanthus clade includes the species considered by Salisbury (1866), Dyer (1934), and Nordenstam (1970, with the exclusion of Rhadamanthus platyphyllus B. Nord. that belongs to Striatula and Rhadamanthus cyanelloides Baker that represents the monotypic Sagittanthera). These Rhadamanthus species share stamens connivent to the gynoecium, and anthers dehiscing by apical porelike slits. Moreover, Urginea ciliata, Urginea rigidifolia, Urginea muirii, and Drimia cochlearis form a well-supported subclade that was recognized by Manning & Goldblatt (2018) as Drimia sect. Sclerophyllae J.C. Manning & Goldblatt, based on their nodding globular buds, patent to suberect stellate flowers, spreading filaments, and complete dehiscence of anthers. We suggest accepting them as Rhadamanthus based on the   (Fig. 3). However, further studies are required, including a complete sampling in the genus, to evaluate possible alternatives.
The monotypic Mucinaea consistently represents an independent lineage, supported by the unique combination of bright purplish-pink tepals with a basal green marking encircled by a white ring (Fig. 1.12); anthers opening as an apical pore or slit; a nonbarred, purple amplexicaul cataphyll; and the bulb structure (Pinter et al., 2013).
Although Sagittanthera represents a fully supported clade to include samples of R. cyanelloides being easily identified by the large, connate anthers (unique in Hyacinthaceae) that dehisce by apical pores (Figs. 1.16, 3), presence of distinct bracteoles, and the leaves keeled abaxially, among other characters (Martínez-Azorín et al., 2013d).
Urginavia in the sense of Speta (1998b) and the present study includes species with bulbs composed of leathery scales (usually yellowish when dry), which are imbricate and produce white silky threads when broken, usually long racemose inflorescences, distinct bracteoles, withered tepals persisting below the developing capsule, and flattened black seeds. These species occur south of the Sahara Desert, and mostly fit sect. Urginavia (Speta) J.C. Manning & Goldblatt (Manning & Goldblatt, 2018). However, our phylogenetic results reveal that Urginea multisetosa Baker and Urginavia echinostachya Baker (placed in the polyphyletic sect. Ledebouriopsis by Manning & Goldblatt, 2018) also belong to Urginavia and share the diagnostic characters of that genus.
Sekanama sensu Speta (2001) includes Urginea sanguinea Schinz, Urginea burkei Baker, and Urginea delagoensis Baker. However, our results place samples of U. sanguinea in a wellsupported clade and samples of U. delagoensis, U. lydenburgensis, and Drimia edwardsii in a clade that, in some analyses, is related to the former, but with very low support (Fig. 2). Important differences in distribution and morphology exist between the two groups. Sekanama sanguinea has a more northern and generally western distribution and shows hysteranthous leaves, elongated raceme with a much shorter peduncle, white stellate flowers ( Fig. 1.18), withered tepals persisting at the base of the capsule, and flat and wide, elliptic seeds, whilst U. delagoensis, U. lydenburgensis, and D. edwardsii have a more southern distribution, and produce synanthous leaves; an elongated peduncle; subcampanulate, pale brown, or carneous to greenish flowers ( Fig. 1.30); withered tepals persisting atop the capsule; narrowly ellipsoid capsules; and narrowly lanceolate seeds. Therefore, we suggest restricting Sekanama to include S. sanguinea and S. burkei and propose the new genus Zulusia Mart.-Azorín et al. (ined.) to accommodate U. delagoensis, U. lydenburgensis, and D. edwardsii (cf. Crouch & Martínez-Azorín, 2015), a solution also supported by their different chromosome counts .
Our phylogenetic results consistently indicate a wellsupported clade that includes Iosanthus, Spirophyllos, Urginea, Indurgia, Vera-duthiea, and Ebertia. Among those related clades, Iosanthus sensu Martínez-Azorín et al. (2019b) is monotypic and includes the small toxic plant Ornithogalum toxicarium C. Archer & R.H. Archer (Fig. 1.9). Our phylogenetic results consistently place a sample of this species as sister to two samples of Drimia khubusensis P.C. van Wyk & J.C. Manning. Moreover, our trees usually show all samples of those two taxa as sister to an undescribed species from central Namibia (Fig. 2). These three species characteristically share a relatively small plant size; hypogeal, compact bulbs; filiform leaves; lack of bracteoles; short and few-flowered inflorescence; free tepals; and capsule valves reflexed to completely expose the flattened, discoid, and winged seeds. Despite some morphological differences in flower structure (Manning & Goldblatt, 2018), we tentatively propose to expand Iosanthus to comprise those three latter species, to provide the most conservative solution.
Urginea, as typified by Adamson et al. (1944) in Urginea fugax Steinh., is narrowed to include the latter species and Urginea ollivieri Maire, being restricted to the western Mediterranean basin, where it forms a morphologically consistent group characterized by filiform leaves; diurnal, stellate, patent to suberect flowers with free tepals (Fig. 1.26); spreading filaments; and flattened, ellipsoid seeds. Our samples of U. fugax and U. ollivieri form a strongly supported clade that is sister to a clade comprising various samples of U. noctiflora from Morocco, a relationship reported earlier by Pfosser et al. (2006). The latter species differs from Urginea by the distinctly coiled leaves (a unique character in Urgineoideae); nocturnal nodding flowers with straight filaments that are connivent to the style and cross at their middle ( Fig. 1.19); and patent capsules. We, therefore, propose the description of a new genus named Spirophyllos Mart.- Azorín et al. (ined.) to accommodate U. noctiflora, based on the distinct differences noted above as well as its genetic divergence and different habitats and distributions.
The latter two genera are consistently shown in our trees as sister to samples of Indurgia in the sense of Yadav et al. (2019), who accommodated only southeast Asian members of Urgineoideae in their Drimia sect. Indurgia (Speta) J.C. Manning & Lekhak. Indurgia can be identified by the combination of caducous bracts; lack of bracteoles; nocturnal and nodding flowers or sometimes diurnal and spreading; suberect to spreading filaments; erect, usually thickened, subclavate style with truncate stigma (Fig. 1.8); apiculate capsule valves; and ellipsoid, flattened, and winged seeds.
Speta (2001) 2021) include taxa from southern and central Africa and the southern Arabian Peninsula, characterized by maculate leaves (at least at their base); lack of bracteoles (rarely present); nodding, nocturnal flowers; tepals strongly reflexed; filaments incurved along the lower half, connivent to the style in the middle section and spreading distally ( Fig. 1.28); style distinctly deflexed; and flattened subelliptic seeds. Our phylogenetic results agree with those of Pfosser and Speta (1999, 2001, 2004 in retrieving samples of this genus in a well-supported clade, being sister to Ebertia. Moreover, as reported earlier by Pfosser et al. (2006), U. noctiflora requires segregation from Vera-duthiea sensu Speta (2016) (as already noted under Spirophyllos), as this species does not present the typical leaf maculation of Vera-duthiea.
Ebertia Speta includes the tropical African taxa Urginea pauciflora Baker and Urginea nana Oyewole. This genus is characterized by filiform, proteranthous leaves; short peduncle and condensed few-flowered raceme; shortly spurred bracts; nocturnal, campanulate flowers; tepals shortly connate at the base; filaments shorter than tepals; pedicels of ripe capsules laterally recurved; and flattened black seeds. Our results recover three samples of Ebertia in a well-supported clade.
Triandra pellabergensis Mart.-Azorín et al. constitutes an isolated lineage supported by the presence of only three stamens per flower (unique in Hyacinthaceae) ( Fig. 1.24), among other characters (cf. Martínez-Azorín et al., 2021). This genus approaches Urginea revoluta in flower morphology, although the latter produces the usual six stamens per flower and is only very distantly related in our phylogenetic studies. Further species and expanded genetic studies are required to elucidate the taxonomic placement of U. revoluta. Sister to Triandra appears a perfectly supported clade fitting with the Madagascan endemic Rhodocodon in the sense of Baker (1881) and Knirsch et al. (2015Knirsch et al. ( , 2016Knirsch et al. ( , 2019, characterized by urceolate to campanulate flowers ( Fig. 1.15) (lasting for 3-7 days); tepals connate for most of their length and persisting at the base of capsules; adnate filaments; and seeds subellipsoidal and usually with a distinct raphe, or rarely compressed (Brudermann et al., 2018).
Tenicroa is a distinct genus accepted historically by most researchers, including by Pinter et al. (2020), in the latest revision of that genus, one easily characterized by mostly synanthous leaves with transversally striate-raised sheathing cataphylls; stellate flowers with free tepals; suberect stamens with subbasifixed anthers, and an elongate, deflexed, and curved-sigmoid style (Fig. 1.22). Our Tenicroa samples form a fully supported clade that is usually related to Urgineopsis.
Within the remaining clades, Squilla in the sense of Steinheil (1836) was treated in recent times as Charybdis Speta (1998b) nom. nov. to replace Squilla, a name considered by Speta to be an orthographic variant of both Scilla L. and Skilla Raf. However, typification by Rafinesque of Scilla maritima L. renders Charybdis illegitimate and unavailable for use (cf. Martínez-Azorín & Crespo, 2016a;Crespo et al., 2020;Martínez-Azorín et al., 2022). Martínez-Azorín & Crespo (2016b) have recently requested a binding decision on whether Scilla L. and Squilla Steinh. are sufficiently alike to be considered orthographic variants. It seems that most members of the Committee will accept Squilla as not being confusable with Scilla (W. Appelquist, pers. comm.), in which case the name Squilla would be available for the current concept of Charybdis, as already accepted by Martínez-Azorín et al. (2022). Previous phylogenetic analyses (Pfosser & Speta, 2001, 2004Pfosser et al., 2012) place numerous samples of Squilla (as Charybdis) in a strongly supported clade, supporting acceptance of this group as an independent genus (Speta, 1998b;Pfosser & Speta, 2001, 2004Conti et al., 2005;Jeanmonod & Gamisans, 2007;Bacchetta et al., 2012;Ali et al., 2013;Véla et al., 2016). We found our 22 samples of Squilla to form a strongly supported clade in an isolated position within Urgineoideae, and therefore, we recognize this genus based on the hysteranthous leaves; presence of distinct bracteoles; and flattened and winged seeds, together with their Mediterranean distribution (Martínez-Azorín et al., 2022).
A distinct group with nodding campanulate flowers was recognized as Rhadamanthopsis at the subgenus (Obermeyer, 1980a) or genus (Speta, 2001) level to include two species: Rhadamanthopsis namibensis and Rhadamanthopsis karooicus (Oberm.) Speta. These species are characterized by diurnal, nodding, and campanulate flowers; tepals connate for about 1/3 to 2/5 of their length and free, suberect, apical lobes ( Fig. 1.13); stamens included and connivent to the style, with adnate filaments; loculicidal dehiscence of their anthers (instead of by apical pores or slits) and distinct bracteoles, differing substantially from Rhadamanthus (as interpreted in this paper). Other species agreeing morphologically with Rhadamanthopsis were described, including Drimia hyacinthoides Baker (1874), Ornithogalum haworthioides Baker (1878) (≡Drimia bolusii Baker 1897; not to be confused with Drimia haworthioides Baker 1875), and Drimia monophylla Oberm. ex J.C. Manning & Goldblatt. The phylogenetic analyses by Pfosser & Speta (1999, 2001, 2004 and Pfosser et al. (2012) found samples of Rhadamanthopsis to form a clade with moderate support, including samples of "Karoophila bolusii" (a genus name not formally published). Our phylogenetic results consistently place samples morphologically fitting Rhadamanthopsis into three fully supported clades, which usually form a polytomy, where a sample of Aulostemon is also related. One clade includes samples of the namibian R. namibensis, another comprises the Namaqualand samples of R. karooicus and relatives, and the last clade accommodates the southeastern South African D. hyacinthoides, D. monophylla, and O. haworthioides. Although some morphological differences in vegetative characters exist among the taxa included in these three biogeographic subclades and their polyphyletic relationships are revealed in some analyses (Fig. 2), we propose to accept Rhadamanthopsis to include all species characterized by the distinct flower morphology detailed above. The placement of the R. namibensis clade is diverse in our analyses and sometimes it is recovered as an independent clade, although with its relationships very weakly supported or collapsed. However, when morphological data are included in the plastid phylogenetic analyses, Rhadamanthopsis recovers monophyly in the sense of this work (Fig. S8). Furthermore, the published chromosome numbers (2n = 16, 18) for this genus differ from common chromosome counts in the subfamily (2n = 20; x = 10) . The required new combinations in the genus will be effected in a forthcoming monograph.
Aulostemon, although related to Rhadamanthopsis in the phylogenetic analyses, is readily differentiated by its stellate flowers; green pedicels supporting dry dehisced capsules, free tepals with a green basal macula; filaments connate to form a long tube above the perigone ( Fig. 1.1) (a unique and diagnostic character in Hyacinthaceae); and free anthers, among other characters (Martínez-Azorín et al., 2017).
Another clade that consistently resolved in our phylogenetic analyses includes both Litanthus and Schizobasis as sister, fully supported lineages, corroborating the findings of Pfosser & Speta (2001, 2004 and Pfosser et al. (2012). This sister relationship, at first sight surprising based on their different flower and inflorescence morphologies, is supported by both the elongation of the anther connective into a small, translucent, membranous flap and the angled seeds (Manning & Goldblatt, 2018). However, Litanthus in the sense of Harvey (1844), Manning et al. (2013), andMartínez-Azorín et al. (2015b) is easily characterized by 1(2)-flowered inflorescence; two subopposite spurred bracts; nodding, tubular flowers with tepals connate into a long tube ( Fig. 1.11); stamens with adnate very short filaments; trigonous, minute seeds; and most notably, stigma with six tiny, erect teeth (a unique character in Urgineoideae). Schizobasis is also highly distinctive on account of its slender, wiry, flexuose, branched inflorescences (Baker, 1873;Manning et al., 2014); it shares the latter character with Bowiea, although clearly differing in both sexual and vegetative morphology as noted above.
Another clade with full support, although with weakly supported relationships, includes the species of Drimia s.str. that constitute a morphologically compact group. The inclusion of numerous taxa in it after its original description by Willdenow (1799) blurred the morphological characterization of the genus and created considerable instability in generic circumscriptions in the Urgineoideae. This primarily stems from differing perceptions of the significance to be accorded to the extent of tepal connation, from nearly free to connate in a distinct tube (Huber, 1969;Jessop, 1977;Stedje, 1987Stedje, , 2001aStedje, , 2001bDeb & Dasgupta, 1982;Manning et al., 2004). However, when recovering its original sense, Drimia is easily recognized by the tepals connate in a cylindrical tube with linear, elongate, narrowly subspathulate, strongly reflexed lobes; adnate filaments that arise at the mouth of the tube; and stamens commonly curved and closely appressed to the style (Fig. 1.5), among other characters.
The originally monotypic Urgineopsis accommodated Urgineopsis salteri R.H. Compton (Compton, 1930), and although Speta (1980) intended to effect later the combination Urgineopsis modesta (Baker) Speta, our analyses place the latter species in Boosia. Our phylogenetic trees recover some species of Urgineopsis in the sense of Martínez-Azorín et al. (2019a) as monophyletic and well supported, based on their connate tepals forming a campanulate and usually widely open tube, and the spreading and slightly incurved filaments that arise at the apex of the tepal tube ( Fig. 1.27). This genus is related to Tenicroa in some of our analyses (Fig. S1), with which it shares a general distribution. Urgineopsis was reduced to synonymy in Drimia by Jessop (1977), who argued that the degree of fusion of tepals is not a consistent character useful in defining genera in the Urgineoideae, as a continuum of connation degrees is observable. As noted above, we concur that this character alone should not be used for generic circumscription in the Urgineoideae due to a certain degree of homoplasy, but the correct combination of morphological characters and phylogenetic evidence supports the acceptance of several genera at that rank, including Urgineopsis.
Another clade with strong support includes two sister and fully supported genera, Austronea sensu Martínez-Azorín et al. (2018a, 2019c and Fusifilum sensu Müller-Doblies et al. (2001), although with Fusifilum magicum being related to Urginea revoluta in our trees (Fig. 2). Both Austronea and Fusifilum share some general morphological characters, although their flower and inflorescence morphologies allow them to be readily distinguished. Austronea is characterized by its capitate to subcorymbose inflorescences that commonly nod at early development stages (one of the best diagnostic characters of the genus), the green to yelloworange ovary ( Fig. 1.2), and seeds trigonous in outline and tetrahedrally folded. On the other hand, Fusifilum differs in the white, fusiform filaments that are distinctly papillate basally; white flowers; white ovary, sometimes tinged with purple ( Fig. 1.6) (one of the best diagnostic characters of the group); and ellipsoid flattened seeds.
Finally, the remaining samples constitute a fully supported clade where samples of Geschollia in the sense of Martínez-Azorín et al. (2019d) form a fully supported clade and share the main diagnostic characters of the genus, such as the single terete leaf (rarely 2) and comparatively small capsules with small polygonal or irregularly compressed, angled seeds, among other characters. The other samples are grouped in a clade with strong support with two fully supported subclades. One subclade includes Urginea macrocentra Baker (Fig. 1.3), corresponding to the monotypic Boosia sensu Speta (2001), plus other species distributed along southeastern South Africa and Lesotho, such as Drimia flagellaris T.J. Edwards et al., Urginea modesta Baker, U. rubella Baker, U. saniensis Hilliard & B.L. Burtt, and U. tenella Baker. Despite Boosia being described originally as monotypic to accommodate a peculiar species with very long and colored bract spurs, and a single, terete, corky leaf, consideration of the morphological characters of these subclades reveals that they differ from the related Geschollia by a syndrome of morphological characters: multiple leaves per bulb (rarely one); often very long spurs on the basal bracts; pedicels that remain photosynthetic when capsules are completely ripe; and elongated, flattened seeds. We accordingly propose expanding Boosia to include several related species, the new combinations for which will be effected in a forthcoming monograph. Finally, the other subclade includes samples from western South Africa and comprises two samples of F. magicum (the only species of the genus sensu Müller-Doblies et al. (2001) that dissolves its monophyly), the sample "H847 Boosia sp." presented by Speta (2001, 2004) from Swellendam (which we were unable to study morphologically), and a sample of Urginea revoluta. The latter species is not a member of Urginea in the sense of the current work based on morphology, phylogenetic evidence, and distribution ranges and shows morphological affinities to Triandra. Further studies are needed to provide more insight into the relationships and statuses of taxa from western South Africa that resolve in the latter subclade.

Distribution patterns
Subfamily Urgineoideae is mostly restricted to Africa, Madagascar, the Mediterranean, and southwestern Asia, with two main centers of diversity-one in southern Africa, where Urgineoideae originated ca. 48 Ma ago, and the other in the Mediterranean Basin, representing a secondary center of diversity that was formed ca. 17-20 Mya by colonization from Africa (Buerki et al., 2012;Ali et al., 2013). This dispersal was facilitated by both low-elevation arid and high-elevation montane corridors linking the ancestral region of southern Africa and the Mediterranean, via East Africa (Martínez-Azorín et al., 2010;Buerki et al., 2012;Ali et al., 2013). The Ex-Africa scenario represented by Indurgia is indicative of the emergence of yet another secondary center of diversity in India and SE Asia.
Among the 31 clades or lineages accepted as genera in the present study, of which some are newly circumscribed here, several are restricted to the southern regions of Africa, thus reflecting their taxonomic independence (Austronea, Aulostemon, Fusifilum, Geschollia, Iosanthus, Litanthus, Mucinaea, Rhadamanthopsis, Rhadamanthus, Sagittanthera, Sekanama, Striatula, Tenicroa, Thuranthos, Triandra, and Urgineopsis) (regions 1, 2, 3, and 5a in Fig. 3). Among them, some are endemic to certain regions, such as Iosanthus, Mucinaea, and Triandra to the Karoo-Namib Region (region 2 in Fig. 3); Aulostemon, Sagittanthera, and Zulusia to the Uzambara-Zululand Region (region 3 in Fig. 3); Rhodocodon to the Madagascan Region (region 4 in Fig. 3); and Sekanama to the southern section of the Zambezian Subregion (region 5a in Fig. 3). Tenicroa and Urgineopsis are restricted to the Cape plus Karoo-Namib Regions (regions 1 and 2 in Fig. 3); Austronea, Geschollia, and Rhadamanthopsis share distribution with Tenicroa and Urgineopsis but also extend to the Uzambara-Zululand Region (region 3 in Fig. 3); and Fusifilum, Litanthus, Rhadamanthus, and Thuranthos also share the distribution of Austronea, Geschollia, and Rhadamanthopsis but spread further to the southern section of the Zambezian Subregion (region 5a in Fig. 3). Striatula grows in the Cape and Karoo-Namib Regions and the southern section of the Zambezian Subregion (regions 1, 2, and 5a in Fig. 3). Zingela is restricted to the Uzambara-Zululand Region and the southern section of the Zambezian Subregion (regions 3 and 5a in Fig. 3), and Boosia shares distribution with Zingela but also extends to the eastern section of the Zambezian Subregion (region 5c in Fig. 3). Drimia, Bowiea, and Schizobasis are also present in the southern regions of Africa but extend northwards to the Zambezian Subregion (regions 5a, 5b, and 5c in Fig. 3), although Bowiea is not present in the Cape Region. Urginavia widely occurs in the southern regions of Africa and extends to northern and eastern sections of the Zambezian Subregion, the Erithraeo-Arabian Subregion, the Guineo-Congolian Region, and the Sahelo-Sudanian Subregion (regions 5b, 5c, 6a, 7, and 8 in Fig. 3). Vera-duthiea is distributed along the Uzambara-Zululand Region, the southern and eastern sections of the Zambezian Subregion, the Guineo-Congolian Region, and the Sahelo-Sudanian Subregion, extending beyond Africa to the Erithraeo-Arabian and Omano-Sindian Subregions in southern Yemen and the Dhofar mountains in Oman (regions 3, 5a, 5c, 6a, 6b, 7, and 8 in Fig. 3), to share the latter Subregion with Indurgia.
Ledurgia is endemic to the Guineo-Congolian Region (region 7 in Fig. 3) and Ebertia occurs along the eastern section of the Zambezian Subregion, and the Guineo-Congolian and Sahelo-Sudanian Regions (regions 5c, 7, and 8 in Fig. 3) (Oyewole, 1989;Friis & Vollesen, 1999;Speta, 2001). Urginea is mostly restricted to the Mediterranean Region (region 10 in Fig. 3) and Squilla is widely distributed along the Mediterranean, extending to the Macaronesian Region (regions 10 and 11 in Fig. 3) (Pfosser & Speta, 2004;Martínez-Azorín et al., 2022). Spirophyllos is endemic to a narrow strip in desert habitats in northern Morocco and Algeria, south of the Atlas mountain range, included in the Saharo-Arabian Region (region 9 in Fig. 3). Finally, the only genus occurring in Asia is Indurgia, being mostly restricted to India and Thailand in the Indian and Indo-Chinese Regions and just entering the eastern part of the Omano-Sindian Subregion (regions 6b, 12, and 13 in Fig. 3).
Based on our phylogenetic findings, the genera Ebertia, Indurgia, Iosanthus, Vera-duthiea, Spirophyllos, and Urginea consistently place in a strongly supported clade, with the southern African lineages of Vera-duthiea and Iosanthus postulated to have given rise to the remaining northern hemisphere representatives of both the western Mediterranean and the southwestern Asian lineages. Our studies consistently place the Madagascan Rhodocodon as related to the northwestern South African Triandra. Urginavia and Ledurgia, both present in central Africa, are related to the southern African Thuranthos and Zingela.

Supplementary Material
The following supplementary material is available online for this article at http://onlinelibrary.wiley.com/doi/10.1111/jse. 12905/suppinfo: Fig. S1. Phylogram of the Bayesian majority-rule consensus tree of the concatenated plastid regions (trnL intron, trnL-F spacer, matK, trnCGCA-ycf6) data set for Urgineoideae displaying branch lengths; posterior probabilities (PP) are shown at the nodes; clade labels follow Fig. 2. Fig. S2. Phylogram of the Bayesian majority-rule consensus tree of the plastid trnCGCA-ycf6 region for Urgineoideae displaying branch lengths; posterior probabilities (PP) are shown at the nodes; clade labels follow Fig. 2.  Fig. S3. Phylogram of the Bayesian majority-rule consensus tree of the plastid matK region for Urgineoideae displaying branch lengths; posterior probabilities (PP) are shown at the nodes; clade labels follow Fig. 2.  Fig. S4. Phylogram of the Bayesian majority-rule consensus tree of the plastid trnL intron, trnL-F spacer region for Urgineoideae displaying branch lengths; posterior probabilities (PP) are shown at the nodes; clade labels follow Fig. 2. Fig. S5. Maximum Parsimony majority-rule consensus tree of the concatenated plastid regions (trnL intron, trnL-F spacer, matK and trnCGCA-ycf6) data set for Urgineoideae; bootstrap support (BS) values are shown at the nodes; clade labels follow Fig. 2. Fig. S6. Phylogram of the Bayesian majority-rule consensus tree of the nuclear Agt1 region for Urgineoideae displaying branch lengths; posterior probabilities (PP) are shown at the nodes; clade labels follow Fig. 2. Fig. S7. Maximum Parsimony majority-rule consensus tree of the concatenated plastid (trnL intron, trnL-F spacer, matK and trnCGCA-ycf6) and nuclear (Agt1) data set for Urgineoideae; bootstrap support (BS) values are shown at the nodes; clade labels follow Fig. 2. Fig. S8. Phylogram of the Bayesian majority-rule consensus tree of the concatenated plastid (trnL intron, trnL-F spacer, matK, and trnCGCA-ycf6) plus morpho (40 coded characters indicated in Table S2) data set for Urgineoideae displaying branch lengths; posterior probabilities (PP) are shown at the nodes; clade labels follow Fig. 2. Fig. S9. Phylogram of the Bayesian majority-rule consensus tree of the concatenated plastid (trnL intron, trnL-F spacer, matK, and trnCGCA-ycf6), nuclear (Agt1) and morpho (40 coded characters indicated in Table S2) data set for Urgineoideae displaying branch lengths; posterior probabilities (PP) are shown at the nodes; clade labels follow Fig. 2. Supplementary materials S10. Nexus file of the Bayesian analyses for the complete molecular data set for Urgineoideae. Supplementary materials S11. Nexus file of the Bayesian analyses for the complete molecular and morphological data set for Urgineoideae. Supplementary materials S12. Mesquite file where the studied 40 morphological characters are plotted onto the phylogenetic tree as in Figure 2. Supplementary materials S13. File including the 40 plotted phylogenetic trees with the studied morphological characters using Mesquite. Table S1. Data on the studied samples of Urgineoideae, including sample number, taxonomy, locality details, voucher codes, and Genbank numbers for each DNA sequence. Table S2. Morphological matrix for the studied samples in Urgineoideae with 40 coded morphological characters, as follows: 1. Bulb scales (0. Compact; 1. Loose); 2. Cataphylls (0. Leaves lacking sheathing cataphylls with transversal dark or prominent ribs; 1. Leaves surrounded at base by sheathing cataphylls with transversal dark or prominent ribs); 3. Leaf number (0. More than one per bulb; 1. One per bulb; 2. Absent in old plants); 4. Leaf curving (0. Straight or slightly curved; 1. Distinctly coiled distally); 5. Leaf section morphology around the middle portion (0. Flattened or canaliculate; 1. Round to hemicircular); 6. Leaf proportions