Advances in Legume Systematics 14. Classification of Caesalpinioideae . Part 2: Higher-level classification

PMCID: PMC11188994

PMID:

---

Abstract

Abstract Caesalpinioideae is the second largest subfamily of legumes ( Leguminosae ) with ca. 4680 species and 163 genera. It is an ecologically and economically important group formed of mostly woody perennials that range from large canopy emergent trees to functionally herbaceous geoxyles, lianas and shrubs, and which has a global distribution, occurring on every continent except Antarctica. Following the recent re-circumscription of 15 Caesalpinioideae genera as presented in Advances in Legume Systematics 14, Part 1, and using as a basis a phylogenomic analysis of 997 nuclear gene sequences for 420 species and all but five of the genera currently recognised in the subfamily, we present a new higher-level classification for the subfamily. The new classification of Caesalpinioideae comprises eleven tribes, all of which are either new, reinstated or re-circumscribed at this rank: Caesalpinieae Rchb. (27 genera / ca. 223 species), Campsiandreae LPWG (2 / 5–22), Cassieae Bronn (7 / 695), Ceratonieae Rchb. (4 / 6), Dimorphandreae Benth. (4 / 35), Erythrophleeae LPWG (2 /13), Gleditsieae Nakai (3 / 20), Mimoseae Bronn (100 / ca. 3510), Pterogyneae LPWG (1 / 1), Schizolobieae Nakai (8 / 42–43), Sclerolobieae Benth. & Hook. f. (5 / ca. 113). Although many of these lineages have been recognised and named in the past, either as tribes or informal generic groups, their circumscriptions have varied widely and changed over the past decades, such that all the tribes described here differ in generic membership from those previously recognised. Importantly, the approximately 3500 species and 100 genera of the former subfamily Mimosoideae are now placed in the reinstated, but newly circumscribed, tribe Mimoseae . Because of the large size and ecological importance of the tribe, we also provide a clade-based classification system for Mimoseae that includes 17 named lower-level clades. Fourteen of the 100 Mimoseae genera remain unplaced in these lower-level clades: eight are resolved in two grades and six are phylogenetically isolated monogeneric lineages. In addition to the new classification, we provide a key to genera, morphological descriptions and notes for all 163 genera, all tribes, and all named clades. The diversity of growth forms, foliage, flowers and fruits are illustrated for all genera, and for each genus we also provide a distribution map, based on quality-controlled herbarium specimen localities. A glossary for specialised terms used in legume morphology is provided. This new phylogenetically based classification of Caesalpinioideae provides a solid system for communication and a framework for downstream analyses of biogeography, trait evolution and diversification, as well as for taxonomic revision of still understudied genera.

---

Full Text

With close to 800 genera and more than 22,000 species (LPWG 2023), the () is the third largest angiosperm family in number of species after and . Legumes include a large set of economically important food crops that provide highly nutritious sources of plant protein and micronutrients, which can greatly benefit health and livelihoods. They have been domesticated alongside grasses in different areas of the world since the beginnings of agriculture and have played a key role in its early development. Legumes are also important sources of fodder and green manure in both temperate and tropical regions, and are used for their wood, tannins, oils and resins, in the manufacture of varnishes, paints, dyes and medicines, and in the horticultural trade. Legume diversification probably started close to the Cretaceous-Paleogene boundary (ca. 66 Ma) (Koenen et al. 2020b, 2021), giving rise to one of the most spectacular examples of evolutionary diversification in plants. Modern legumes are exceptionally diverse, morphologically, physiologically and ecologically (Lewis et al. 2005; LPWG 2017).
In 2017, the Legume Phylogeny Working Group (LPWG 2017) revised the higher-level classification of the family and recognised six monophyletic subfamilies within the monophyletic . Under the LPWG classification, subfamily DC. was re-circumscribed, and LPWG, Burmeist., LPWG and LPWG (all of which were previously part of sensu lato at different ranks) were recognised as distinct subfamilies along with an unchanged DC. The former subfamily DC., which is phylogenetically nested within , was subsumed within the re-circumscribed and has since been referred to as the mimosoid clade (LPWG 2017). The idea that comprises six main lineages, corresponding to these six subfamilies, is now widely accepted and has been confirmed by recent phylogenomic analyses of large nuclear gene and plastome DNA sequence datasets (Koenen et al. 2020b; Zhang et al. 2020; Zhao et al. 2021), which show robust support for all six subfamilies. Phylogenomic evidence suggests that the six subfamilies likely diverged very rapidly such that gene tree conflict obscures relationships among some of the subfamilies, but is supported as sister to (Koenen et al. 2020b, 2021).
Within subfamilies, new phylogenies of many legume groups have unequivocally demonstrated the non-monophyly of the tribes recognised in the classifications of Polhill and Raven (1981), Polhill (1994) and Lewis et al. (2005), and the need for new classifications (LPWG 2013, 2017). Following publication of the LPWG (2017) subfamily classification, phylogenetically-based tribal and clade-based higher-level classifications were developed for subfamily (Estrella et al. 2018) and informally for (Sinou et al. 2020), but complete higher-level phylogenetically-based classifications are still lacking for and . Subfamily , comprising a single species, requires no classification, and , with just 18 genera, may not be easily amenable to, or in need of, additional higher-level subdivisions, although phylogenetic studies are ongoing (e.g., Falcão et al. 2023). For , the largest of the subfamilies, despite ongoing progress in the understanding of phylogenetic relationships (Wojciechowski et al. 2004; Cardoso et al. 2012, 2013; Wojciechowski 2013; Zhao et al. 2021; Choi et al. 2022), more data and more complete taxon sampling are needed before a robust and stable phylogenetically-based classification system can be fully developed. For , although some questions persist about the monophyly and placement of a small subset of genera and some on-going uncertainty surrounding generic delimitation and relationships (LPWG 2013, 2017; Koenen et al. 2020a), recent work has clarified most of these problems (Ringelberg et al. 2022), many of which were resolved in a series of papers in Advances in Legume Systematics 14, Part 1 (Hughes et al. 2022a).
A new higher-level classification of subfamily is therefore now both feasible and timely. Here we use the phylogenomic backbones for subfamily from Koenen et al. (2020a) and Ringelberg et al. (2022) as the basis for developing a new higher-level classification of the subfamily. This new phylogenetic classification provides a solid system for communication and a framework for downstream analyses of biogeography, trait evolution and diversification (e.g., Faria et al. 2022; Ringelberg et al. 2022, 2023), as well as for guiding efforts towards fully revising the taxonomy of still understudied genera.
sensu LPWG (2017) is the second largest subfamily of legumes with ca. 4680 species placed in 163 genera (Hughes et al. 2022a; LPWG 2022; Ringelberg et al. 2022). Within this subfamily, ca. 3500 species and 100 genera are placed in the former subfamily (the mimosoid clade of LPWG 2017), which we here recognise as the reinstated, but newly circumscribed, tribe , a tribal name first published by Bronn in 1822 (see below). date to the late Paleocene when the subfamily is known from fossil bipinnate leaves from Colombia (Wing et al. 2009; Herrera et al. 2019). These fossils indicate that were an abundant element in the earliest Neotropical rain forests in the Paleocene and time-calibrated legume phylogenies suggest that started to diversify around 58 million years ago (Lavin et al. 2005; Bruneau et al. 2008; Koenen et al. 2021). have thus diversified throughout the Cenozoic and now comprise diverse, abundant, and sometimes dominant elements across all major lowland tropical biomes, including rain forests, savannas and seasonally dry forests (Figs 1, 2, 3).
A genus richness across floristic realms (according to Liu et al. 2023). The numbers within the circles represent the total number of genera in each realm. The numbers on the lines represent the number of genera shared between two realms (> 10 genera) B Number of genera in the floristic subrealms (sensu Liu et al. 2023). The numbers associated with the two polygons indicate the number of genera restricted to the two major blocks of tropical and subtropical areas in the New World and the Old World (maps modified from Liu et al. 2023, CC BY 4.0).
Map showing the global distribution of species richness. Numbers of species per one degree latitude / longitude grid cell. Infraspecific taxa are not counted individually but are included at the species level. All maps in this special issue are based on quality-controlled occurrence data from digitised herbarium specimens and floristic surveys (see Suppl. material 1 for details on occurrence data and methods used to generate maps).
are almost entirely woody perennials, but they are extremely diverse in stature and habit – including lianas, trees of all sizes, up to rain forest canopy emergents (e.g., Ducke, Ducke), shrubs, functionally herbaceous geoxyles, and two herbaceous aquatic species ( Lour.). Similarly, the subfamily is highly diverse in floral and fruit morphology. One of the hallmarks of , as in many other plant groups, is repeated morphological and ecological convergences whereby similar leaf, flower and fruit morphologies, and ecological adaptations, have apparently been reinvented multiple times across lineages and through time (Ringelberg et al. 2022, 2023).
is the only legume subfamily that has bipinnate leaves, which are prevalent but not universal across the subfamily (see Glossary, Schemes 1–7). A minority of genera have species with pinnate leaves, and leaves modified into phyllodes occur in most species of the large, mainly Australian, genus Mill. and in a few species in other unrelated genera including Mill. and L. The leaves themselves, especially the bipinnate leaf, can be extremely large (e.g., Vogel leaves are > 1 m long) to highly reduced; aphyllous, or nearly aphyllous, species occur in some genera [e.g., , , (L.) Moench, Raf., Burkart]. Across all legumes, seismonasty, i.e., leaf movements prompted by touch, is known only within subfamily , in the genera and (tribe ). Extrafloral nectaries (EFNs) are present in the majority of (Scheme 2), are morphologically extremely diverse, and often conspicuous and abundant on the petiole or leaf rachides between pinnae or leaflet pairs, and in a few genera (e.g., F. Muell., Britton & Rose) on floral bracts (Marazzi et al. 2019). A subset of genera are armed with prickles, spines or thorns (Scheme 1), but armature is highly variable, has clearly evolved multiple times across the subfamily, and can vary within clades and even within genera (Hughes et al. 2022a; Ringelberg et al. 2022). Most genera of the mimosoid clade (tribe here) are confirmed nodulators, whereas just nine of the 63 non-mimosoid genera in the subfamily are currently known to nodulate. These nine genera are phylogenetically intermingled with confirmed non-nodulating genera, suggesting multiple evolutionary transitions between non-nodulating and nodulating lineages (Faria et al. 2022). Analyses of gene duplications have shown that several whole genome duplications (WGDs) occurred during the early evolution of the family (Cannon et al. 2015; Stai et al. 2019; Koenen et al. 2021; Zhao et al. 2021), although the exact number and placement of these WGDs remain uncertain. Within , polyploidisation has also occurred numerous times during the Neogene in several genera across the subfamily [e.g., Benth., (A. DC.) Wight & Arn., , Wight & Arn., ; Dahmer et al. 2011; Govindarajulu et al. 2011a; Simon et al. 2011].
Across the subfamily, inflorescences and flowers are morphologically highly variable. The inflorescences can be racemose, paniculate or in fascicles and the have characteristic capitate or spicate and frequently heteromorphic inflorescences, often with some sterile flowers, some of which develop showy staminodia (Schemes 3, 4). Although usually bisexual, flowers can also be unisexual, and in the inflorescences can include a mixture of both bisexual and unisexual flowers with or without sterile flowers. The flowers are generally pentamerous, but there are many variations [3–6 (8) sepals or petals], and in some species, sepals and/or petals are absent ( L.). Flowers are generally radially symmetrical in several tribes, including the , but in other clades the flowers are bilaterally symmetrical or asymmetrical. Although a majority of flowers are bee pollinated, specialised bat, bird, butterfly and moth pollinated flowers are also common (Arroyo 1981). In addition to having species with pollen in the more typical tricolporate monads, is the only subfamily of legumes with taxa where pollen is arranged in polyads (Scheme 6). In the pollen arrangement is extremely variable across and sometimes within genera, with pollen in monads, tetrads, bi-tetrads and polyads. Fruit morphology is particularly homoplasious, and in the has proved misleading for generic delimitation (Borges et al. 2022; Ringelberg et al. 2022; Souza et al. 2022b). This diversity of fruit morphology (Schemes 6, 7) reflects adaptations to different seed dispersal syndromes, including passive, elastic and explosive dehiscence, as well as seed dispersal by water, wind, large herbivores, ants, and birds.
The subfamily is most diverse in lowland tropical and subtropical regions, only rarely occurring above 2500 m elevation, but a minority of genera have species in warm temperate zones that are not prone to severe frosts across the Americas, Europe, Asia, and Australia. More than half of genera naturally occur in the Americas (104 of 163 genera), of which 84 are endemic. Africa (including Madagascar) has the second highest number of genera, with 59 genera, 29 of which are endemic, followed by Asia (40 genera, 7 endemic), and Australia and the Pacific (27 genera, 6 endemic; See details in Tables 1, 2).
genera richness across global floristic realms and subrealms (according to Liu et al. 2023).
The Neotropical floristic realm (sensu Liu et al. 2023) has 101 genera with native species, of which 71 are endemic to this realm (Table 1). Liu et al. (2023) divided the Neotropical realm into the tropical Brazilian subrealm (81 genera, 25 endemic) and the Subtropical American subrealm (76 genera, 17 endemic). The Subtropical American subrealm includes the northern (69 genera, 13 endemic) and southern (35 genera, 3 endemic) extremes of the Neotropical region (Fig. 1). The genera (Benth.) Engelm. & A. Gray and Burkart are restricted to this subrealm and have an amphitropical distribution, whereas the genera Klotzsch and Willd. have a similar distribution, but with a few species reaching the Brazilian subrealm. The second largest floristic realm for genera is the African one, with 59 genera, of which 30 are endemic, primarily in the Guineo-Congolian subrealm (41 genera, 14 endemic). Within the Indo-Malesian realm, there are 40 genera (11 endemic), almost evenly distributed between the Indian (29 genera, 4 endemic) and the Malaysian subrealms (32 genera, 5 endemic).
Nine genera have a pantropical distribution, occurring in all tropical floristic realms [ L., , Adans., L., , (Vogel) Benth., Raf., , ; Table 2]. The genera and R. Br. occur in all tropical floristic realms except for the Australian realm. Seven genera ( C. Presl, R. Vig., L., Barneby & J.W. Grimes, L., Benth., Cav.) represent transatlantic disjunctions as they are exclusively distributed in the Neotropical and African realms.
Species richness, determined from occurrence records, indicate south-central Mexico and Central America, central-eastern Brazil, and southwestern Australia to be the most diverse regions (Fig. 2), with multiple one-degree grid cells in these regions containing more than a hundred species. However, patterns of generic richness (Fig. 3) show that the high species diversity in Australia is largely attributable to the hyper-diverse genus , with over a thousand species. Australia is therefore characterised by high species but low generic richness. Important hotspots of diversity also occur in continental Africa and Madagascar. Asia is the least diverse tropical continent, although it contains multiple species-rich lineages such as , especially in South East Asia. The spatially biased availability of digitised occurrence data (Meyer et al. 2016) leads to an underestimation of richness across large parts of tropical Africa, India, continental South East Asia, and the Amazon, as is apparent in Figs 2, 3. Nevertheless, our analyses (Figs 2, 3) represent an accurate depiction of relative differences in continental richness patterns (Table 1).
The wide geographical distribution of is matched by an equally wide ecological amplitude across the full precipitation spectrum of the tropics, spanning a 100-fold gradient in mean annual precipitation from arid deserts to seasonally dry tropical forests and savannas, and tropical rainforests (Schrire et al. 2005a; Gagnon et al. 2019; Ringelberg et al. 2023). Although species are important components of many wet regions of the world, it is notable that some of the major hotspots of species and especially generic diversity coincide with areas dominated by seasonally dry vegetation in southern Mexico, north-eastern Brazil, and northern and south-western Madagascar, these being key areas of the succulent biome sensu Schrire et al. (2005a) and Ringelberg et al. (2020), plus seasonally dry subtropical south-western Australia. The subfamily thus has no obvious overriding wet or dry affinity, but rather has switched between wet and dry tropical biomes multiple times and has diversified substantially within each (Ringelberg et al. 2023). This ecological adaptability is undoubtedly, at least in part, a function of the evolutionary lability of life history strategies, including adaptations to fire, which has allowed species to become important, diverse, and abundant elements of all lowland tropical biomes. It is also clear that have been able to disperse across oceans numerous times to reach all tropical continents and the majority of lower latitude islands and island archipelagos (Figs 1, 2, 3). In contrast to this wide adaptability across tropical precipitation regimes and vegetation types, show high tropical niche conservatism and very limited adaptability to cold temperatures and frost with just a small subset of lineages and species extending into temperate vegetation (Ringelberg et al. 2023).
Stability is one of the most important qualities of any taxonomic classification. It is therefore crucial that the phylogenetic framework used for assigning names to clades is robust and unlikely to change with sampling of additional taxa or genomic regions in the future. Based on the number and identity of the taxa included, the size of the genomic dataset, and the phylogenomic methods used to infer the phylogeny and assess its robustness, the phylogenetic framework employed here is currently the best available for taxonomic classification of . This is confirmed by its overall agreement with previous smaller-scale phylogenies (see details below) and other recent independent phylogenomic studies (Zhang et al. 2020; Zhao et al. 2021). Furthermore, throughout this compendium the phylogeny is presented in such a way to allow easy assessment of underlying genomic support for all nodes subtending named clades, and in general clades named here are subtended by well-supported nodes on long branches. Absolute stability can never be guaranteed, and sampling of additional taxa might well result in different topologies and generic re-delimitation in some parts of the tree, such as in the grade (Terra et al. 2022) or the clade (Brown et al. 2022; Demeulenaere et al. 2022). Nevertheless, we consider the phylogenomic framework robust, and an adequate basis for the new classification presented here.
The classification proposed here uses as its framework the most comprehensively sampled phylogenetic analysis of to date (Figs 4, 5, Suppl. materials 2, 3). This new phylogeny is based on Koenen et al. (2020a) and Ringelberg et al. (2022, 2023). By developing a clade-specific bait set for targeted enrichment of 964 nuclear genes, Koenen et al. (2020a) generated a DNA sequence dataset an order of magnitude larger than those used previously, thereby providing the greatly enhanced phylogenetic resolution required for classifying tribe . Capitalising on these foundations using a slightly modified version of the gene set covering 997 nuclear genes, and importantly extending the taxon sampling to include 300 additional species covering not only but also most genera of non-mimosoid , as well as conducting transcontinental sampling of genera that occur across different continents, Ringelberg et al. (2022, 2023) established a robust phylogenomic hypothesis for subfamily as a whole. These studies revealed or confirmed the non-monophyly of 22 genera, and this was the basis for the re-circumscription of 15 of these genera presented in Advances in Legume Systematics 14, Part 1 (Hughes et al. 2022a).
Phylogeny of showing the tribal classification presented here. The names and phylogenetic placements of all 63 non-Mimoseae genera are shown and known generic non-monophyly is indicated with terminal names of non-monophyletic genus in bold. The most likely placements for four unsampled genera are indicated with dashed lines; see respective treatments for details. Tribe has been collapsed (see Fig. 5). Branch lengths are expressed in coalescent units, and terminal branch lengths have been assigned an arbitrary uniform length for visual clarity. Monophyletic genera are represented by single branches; see Suppl. material 2 for a phylogeny with all accessions. See Suppl. material 3 for gene tree support across the phylogeny. The phylogeny is a pruned version of the backbone phylogeny of Ringelberg et al. (2023), where full details of the data and phylogenomic analysis methods are presented.
Phylogeny of tribe showing the clade-based classification of the tribe with two named higher-level and 17 named lower-level clades. The names and phylogenetic placements of all 100 genera are shown, and known generic non-monophyly is indicated with terminal names of non-monophyletic genera in bold. The most likely placement of the unsampled genus is indicated with a dashed line; see clade treatment (page 319) for details. Branch lengths are expressed in coalescent units, and terminal branch lengths have been assigned an arbitrary uniform length for visual clarity. Monophyletic genera are represented by single branches; see Suppl. material 2 for a phylogeny with all accessions. See Suppl. material 3 for gene tree support across the phylogeny. The phylogeny is a pruned version of the backbone phylogeny of Ringelberg et al. (2023) where full details of the data and phylogenomic analysis methods are presented.
The phylogenomic analysis presented here includes 420 species representing all but five of the 163 genera. The five missing genera are: Aubl., which has three species and is likely a member of tribe (e.g., Bruneau et al. 2008; LPWG 2017; Kates et al. 2024); Tul., placed here in a phylogenetically isolated, monospecific tribe (e.g., Bruneau et al. 2001, 2008; Manzanilla and Bruneau 2012; Zhang et al. 2020; Zhao et al. 2021); Harms and Gagnon & G.P. Lewis, both monospecific genera of tribe (Gagnon et al. 2016); and C. Presl, a monospecific genus in the clade of tribe (Lima et al. 2022). Although only about 10% of species were sampled in the analyses underlying the phylogenies presented here, several lower-level phylogenetic analyses of specific clades have been published and provide additional support for the groupings presented (Ringelberg et al. 2022). Furthermore, taxon sampling was specifically designed to cover taxonomic diversity spanning the root nodes of subclades and genera (Koenen et al. 2020a; Ringelberg et al. 2022, 2023).
A common feature of phylogenomic analyses employing large numbers of genes is the presence of conflict among gene trees, i.e., phylogenies based on individual genes (Salichos and Rokas 2013; Koenen et al. 2020a, 2020b, 2021; Zhang et al. 2020). Such gene tree conflict is widespread across many nodes in the phylogeny presented here (Koenen et al. 2020a; Ringelberg et al. 2022, 2023). The main cause of this conflict appears to be lack of signal for many nodes in individual gene trees. In such cases, the relevant node in the species tree is only supported by a relatively small number of gene trees, but there is no strong support among the gene trees for any of the alternative, conflicting topologies. The presence of this type of gene tree conflict, indicative of lack of signal rather than true gene tree disagreement, does not preclude naming a clade subtended by such a node, as there is no strong reason to assume that including additional accessions or genomic regions would result in different relationships (Koenen et al. 2020a; Ringelberg et al. 2022). However, in a few places across the tree there is stronger support for alternative conflicting topologies (Koenen et al. 2020a; Ringelberg et al. 2022, 2023). In general, such instances of strong conflict are not found in nodes subtending clades named here, but rather in the relationships within clades [e.g., in parts of the (Fig. 34) and clade (Fig. 225)] and between clades [e.g., the relationships between tribes , , and (Suppl. material 3) or between the and clades and Stapf and Hoyle (Fig. 114)]. Similarly, strong gene tree conflict, including between the nuclear and chloroplast genomes, may affect generic delimitation in some parts of the tree, such as in (Terra et al. 2022) and Schott (Ringelberg et al. 2022). Where relevant for the new classification presented in this compendium, these instances of strong gene tree conflict are described below.
The reference phylogeny used here as the basis for the new classification was inferred using ASTRAL (Zhang et al. 2018), deploying the multi-species coalescent approach based on individual gene trees, which performs well on datasets with inter-genic conflict (Jiang et al. 2020). We always report the non-significant (i.e., > 0.05) outcomes of the ASTRAL polytomy test (Sayyari and Mirarab 2018), which tests for each node whether the polytomy null model can be rejected. Because conventional phylogenetic support metrics, such as bootstrap support, tend to be inflated in large phylogenomic datasets (Rokas and Carroll 2006), we report support for nodes in the species tree using measures of individual gene tree conflict and concordance, calculated using PhyParts (Smith et al. 2015) (Suppl. material 3). The impacts of phylogenomic methods and the presence of conflict among different gene and species trees on taxonomic decisions were further discussed in Ringelberg et al. (2022). Full details of the phylogenomic data and analyses were presented in Ringelberg et al. (2023).
Under the LPWG (2017) subfamilial classification, subfamily was the most difficult and controversial to delimit because of the inclusion of the formerly recognised, widely accepted and morphologically distinctive subfamily . Abandoning the well-known , an important disadvantage of adopting the six-subfamily classification, was mitigated by continuing to recognise this lineage as a named clade, informally referred to simply as the mimosoid clade until now (LPWG 2017), but here formally reinstated as the re-circumscribed and expanded tribe within the new Linnean tribal classification proposed here.
Although have traditionally been diagnosed by a series of diagnostic features, notably valvate petal aestivation and flowers with a reduced perianth and showy androecium, mostly clustered in compact inflorescences, the morphological distinctions between the mimosoid clade and some genera of the subtending grade of caesalpinioid lineages are not always clear-cut. For example, , once considered to be in , is placed outside the mimosoid clade in molecular phylogenetic and phylogenomic analyses (Luckow et al. 2003; Bruneau et al. 2008; Ringelberg et al. 2022). Conversely, , which has always been considered a non-mimosoid caesalpinioid legume (Polhill and Vidal 1981; Lewis 2005b), is placed within the mimosoid clade in all molecular phylogenetic analyses (Manzanilla and Bruneau 2012; LPWG 2017; Koenen et al. 2020a; Ringelberg et al. 2022). However, the long branch subtending what could be considered equivalent to the old subfamily (plus or minus these few genera) (Fig. 4), together with the strong phylogenetic support for the clade, provide ample justification for recognising it at the tribal level.
The new classification proposed here thus follows a traditional Linnean approach but is complemented by a clade-based classification of the large tribe . Rank-free naming of clades within subfamilies and tribes has been prevalent in the legume literature, with many examples of clade names that have become widely used and accepted, such as the dalbergioid clade (Lavin et al. 2001) and the inverted repeat [IR]-lacking clade (Wojciechowski et al. 2000) of ; the clade (Herendeen et al. 2003b) and the ingoid clade (Koenen et al. 2020a) of ; and the Bauhinia and Phanera clades of (Sinou et al. 2020). Naming clades provides useful additional information even after a fully developed and stable subfamily and tribal classification is established. As noted by Wojciechowski (2013), use of Linnean names does not preclude a system that also defines and names clades and their overall relationships outside of the Linnean framework. Instead, the two can be considered complementary for developing a stable, flexible and useful classification of legumes.
The new classification of subfamily comprises eleven tribes, which are either new, reinstated or re-circumscribed at this rank: Rchb., Bronn, LPWG, Rchb., Benth., LPWG, Nakai, Bronn, LPWG, Nakai, and Benth. & Hook. f. (Fig. 4). Although many of these lineages have been recognised and named in the past, either as tribes or informal generic groups, their circumscriptions have varied widely and changed over the past decades, such that all the tribes described here differ in generic membership from those previously recognised (Table 3).
Comparison of the new phylogeny-based classification for with classifications for these genera published in Advances in Legume Systematics, Part 1 (Polhill and Raven 1981) and Legumes of the World (Lewis et al. 2005).
as defined here includes elements from three previously recognised major groups: part of old sense tribe , part of old sense tribe , and the nested subfamily . This broad clade has been referred to as the — or MCC clade (Doyle 2011, 2012), or the — or GCM clade (Marazzi et al. 2012). In 2017, was chosen over as the preferred name for this large clade, even though the two names were published at the same date (LPWG 2017). By choosing the name , this left open the option of recognising the morphologically distinct mimosoid clade at the tribal level, as proposed here.
In their treatment of tribe in Advances in Legume Systematics Part 1, Polhill and Vidal (1981) recognised eight informal generic groups, based primarily on differences in floral morphology. Six of these generic groups, namely the group, group, group, group, group, and group, are here recognised at the tribal level, albeit with modified generic compositions because most of Polhill and Vidal’s named groups have been shown to be non-monophyletic in subsequent phylogenetic analyses using molecular data (e.g., Bruneau et al. 2001, 2008) (Table 3). The monospecific group of Polhill and Vidal (1981) groups with members of tribe , rather than being considered a distinct tribe. In addition, the sensu Polhill and Vidal (1981) included the genus C. Presl. (as a distinct monogeneric group), which has since been shown to be placed in subfamily (Bruneau et al. 2001; LPWG 2017). Tribe of Polhill and Vidal (1981) had been considered to be paraphyletic, at least implicitly, for some time (Polhill and Vidal 1981; Lewis 1998) and this has since been confirmed by phylogenetic analyses which found the tribe to be polyphyletic (Bruneau et al. 2001). The other major group that forms part of is what was considered tribe by Irwin and Barneby (1981). Within the tribe, they recognised five disparate subtribes, H.S. Irwin & Barneby, H.S. Irwin & Barneby, H.S. Irwin & Barneby, Wight & Arn., and H.S. Irwin & Barneby, of which only the latter two have been placed in the (sensu LPWG 2017) in phylogenetic analyses (Table 3). and together (along with ) comprise subfamily , and the monospecific was raised to subfamily rank (LPWG 2017). Thus the grade of non-mimosoid caesalpinioid lineages that subtend tribe in are here recognised as distinct tribes.
Tribe comprises six species in four genera. The four genera had previously been placed in distinct generic groups and tribes: Wight ex Arn. in its own generic group of by Polhill and Vidal (1981); the morphologically distinct, unisexual, and apetalous genus (Tucker 1992) in subtribe of (Irwin and Barneby 1981); Humbert and Urb. in the group of , although none of these placements were considered definitive. Phylogenetic analyses of morphological and plastid sequence data showed the four genera to form a clade and to be closely related to the trigeneric clade here treated as tribe , and together the two clades were placed in the informally named clade (Herendeen et al. 2003a, 2003b; Haston et al. 2005; Bruneau et al. 2008), but subsequent combined plastid and nuclear sequence analyses did not support the monophyly of these two groups together (Manzanilla and Bruneau 2012; Zhang et al. 2020; Zhao et al. 2021). The phylogenomic analyses of Ringelberg et al. (2022) (Fig. 4) clearly indicate that each of these two clades is strongly supported as monophyletic but that they are not grouped together, supporting their recognition as distinct tribes.
These recent molecular analyses have highlighted several previously unsuspected morphological synapomorphies for , the most striking of which is a bipinnate leaf (although mostly once pinnate in ) terminating in a triad of pinnae arising from the same point at the apex of the rachis (Herendeen et al. 2003b; Herendeen and Herrera 2019). Tribe has a highly disjunct and unusual geographic distribution occurring in Hispaniola (), Madagascar (), tropical (South-)East Asia (), and north-eastern Africa and the Mediterranean () (Herendeen et al. 2003b; Tribe , page 62).
The simplest of the informal generic groups recognised by Polhill and Vidal (1981) was the group comprising two primarily north temperate genera, J. Clayton and Lam. The group is supported as monophyletic, and together with the South African Sim, is here formally re-circumscribed as tribe . was previously placed in tribe by Cowan and Polhill (1981) but was later resolved as sister to and in phylogenetic analyses using plastid (Bruneau et al. 2001, 2008) and nuclear sequences (one locus, Manzanilla and Bruneau 2012), as well as in all recent phylogenomic analyses (Zhang et al. 2020; Zhao et al. 2021; Ringelberg et al. 2022; Fig. 4).
The 20 species of the three genera of tribe occur in warm temperate regions, with several disjunctions between North and South America (), South Africa (), and North America and Asia (). Tribe is characterised by several vegetative and floral synapomorphies, such as a tubular hypanthium and sepals with trichomes on the inner surface (Herendeen et al. 2003a; Tribe , page 70).
is here recognised as a new tribe comprising just the single species Tul. Although not included in the phylogenomic analyses of Ringelberg et al. (2022), previous molecular phylogenetic analyses based on plastid and/or nuclear DNA sequence data always resolve this monospecific genus as a phylogenetically isolated lineage, on a long branch, poorly supported relative to and , but generally in the clade that comprises all except and (Bruneau et al. 2001, 2008; Haston et al. 2003, 2005; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; Zhang et al. 2020; Zhao et al. 2021). thus appears to be a classic depauperon (Donoghue and Sanderson 2015), i.e., an old, species-poor lineage. The species is highly distinct morphologically (e.g., imparipinnate leaves with alternate leaflets and a well-formed rachis extension, small flowers in dense catkin-like racemes, the style laterally displaced at the apex of the ovary, fruits a one-seeded winged samara) and cytogenetically (2n = 20), sharing little in common with , or other (Tribe , page 78). is an important tree of South American tropical and subtropical dry forests.
In Advances in Legume Systematics Part 1, Irwin and Barneby (1981) recognised five subtribes in their tribe , including subtribe comprising the genera , , and . These three genera alongside Spruce ex Benth., Schott, and Ducke, previously placed in the group of by Polhill and Vidal (1981) (Table 3), form a robustly supported clade in phylogenetic analyses (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; LPWG 2017; Zhang et al. 2020; Ringelberg et al. 2022) (Fig. 4), here recognised as tribe . The genus [also placed in the group by Polhill and Vidal (1981)], although not sampled by Ringelberg et al. (2022), is generally resolved as a member of this clade albeit with weak support (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; LPWG 2017) and is here placed in the tribe .
Tribe is the largest non-mimosoid clade (in terms of species richness) in subfamily , with 695 species, the vast majority of which are found in the genera (361 species) and (287 species) (Tribe , page 83). Although broadly distributed across the tropics, most of the genera and species are found in the New World. The clade is characterised by singly pinnate or bifoliolate leaves and, in several genera, stomata on both sides of the leaflets (Herendeen et al. 2003a; Bruneau et al. 2008). Several taxa in this clade, including most species of and , as well as and , are well-known for having notably prominent, conspicuous, abundant and unusual extrafloral nectaries (Marazzi and Sanderson 2010; Marazzi et al. 2019; Cota 2020a). Two genera are known to nodulate, and (Faria et al. 2022). None of the genera in , , and are known to nodulate.
The group defined by Polhill and Vidal (1981) is similar to recognised here, except that , Rose and H. Perrier are now resolved in a separate clade (Bruneau et al. 2001, 2008; Haston et al. 2005), here treated as tribe , although has since been synonymised under Raf. (Bruneau and Babineau 2017). Long-standing uncertainty surrounding delimitation of the genus L. and other genera in the group (Lewis 1998; Lewis 2005b) has been resolved with the new generic system of Gagnon et al. (2015, 2016) and subsequent reinstatement of the genus Adans. (Clark et al. 2022).
Tribe comprises ca. 223 species in 27 genera. Although two of these genera, and , were not sampled by Ringelberg et al. (2022), the analyses of Gagnon et al. (2016) clearly resolved as sister to , and in a clade unresolved with and the lineage that combines Adans., Tod., , R.Br. ex Wight & Arn. and .
Species of are highly diverse in growth forms, defence mechanisms, fruit morphologies, and pollination and seed dispersal syndromes (Gagnon et al. 2016). Although there are no clear morphological synapomorphies for the tribe, a diagnostic combination of characteristics is often found, including the presence of glandular trichomes, prickles or spines, bilaterally symmetrical flowers with a modified boat-shaped lower sepal, and free stamens crowded around the pistil (Tribe , page 103). The clade is pantropically distributed, with a marked affinity for the succulent biome (Gagnon et al. 2019).
Tribe as here circumscribed matches the core group (i.e., group s.s.) first recovered phylogenetically by Haston et al. (2003), and subsequently found in several other studies (Haston et al. 2005; Bruneau et al. 2008; Manzanilla and Bruneau 2012; Babineau and Bruneau 2017; Zhang et al. 2020; Zhao et al. 2021; Ringelberg et al. 2022). This clade differs from the informal group recognised by Polhill and Vidal (1981; 13 genera) and Polhill (1994; 16 genera), by excluding four genera now placed in tribe (, , , and ), three now in tribe ( Schrad., Ducke, and Rizzini & A. Mattos) and Benth. (now in tribe ). as circumscribed here also includes and , two genera previously placed in the group by Polhill and Vidal (1981), but which Lewis and Schrire (1995) had suggested might not be part of a more strictly defined group, and which Polhill (1994) had included in the group. In addition, the tribe includes M. Sousa, described by Sousa (2005). Although was resolved as part of this clade (Haston et al. 2005; Bruneau et al. 2008), Babineau and Bruneau (2017) found it to be nested within and synonymised this monospecific genus under .
Tribe contains ca. 42 species in eight genera. It has a pantropical distribution, and a considerable portion of the clade (i.e., the – subclade; Fig. 4) is strictly conserved within the succulent biome (Ringelberg et al. 2020). is not defined by any morphological synapomorphies, but most species in the clade have bipinnate leaves, yellow petals, narrow seeds, characteristic spreading umbrella-like, flat-topped tree crowns, and smooth, thin and either pale silvery metallic grey or green bark (Haston et al. 2005; Tribe , page 146).
The generic composition of tribe as treated here has not been recovered previously, although its constituent genera have often been associated with each other based on morphological and molecular data. The five genera of the were placed in two generic groups of tribe by Polhill and Vidal (1981) and Polhill (1994) based on morphology: Tul. and Aubl. (now including Vogel, but earlier considered distinct from ) were placed in the group, whereas , , and were placed in the group. Subsequent molecular phylogenetic studies generally also resolved the five genera in two separate clades, but with different generic composition from those of the informal groups of Polhill and Vidal (1981) and Polhill (1994). One strongly supported clade grouped , and , as found here (Fig. 4), and a separate less well supported clade (or grade) included and (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; LPWG 2017; Zhang et al. 2020), which has sometimes been resolved as part of a grade subtending the mimosoid clade (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012). In the recent phylogenomic analyses of Ringelberg et al. (2022; Fig. 4), the tribe is subtended by a short branch with notable gene tree conflict, whereas the , and subclade is supported by a long branch. This short branch and gene tree conflict likely explain why the five genera have not been resolved as a clade, but rather as two separate clades in previous phylogenies. In addition, there is evidence to suggest that there may be cytonuclear discordance. In recent phylogenomic analyses, although not all genera have been sampled, plastid data strongly support the , and subclade, with and forming a lineage subtending the mimosoid clade (Zhang et al. 2020), whereas nuclear sequence data group the two lineages as a clade (Zhao et al. 2021; Ringelberg et al. 2022).
As defined here, tribe is restricted to the Neotropics, and comprises ca. 113 species, most of which are in the genus (80–90 species; Tribe , page 165). Several species of are known to form close co-evolutionary associations with ants (Chomicki et al. 2015). Morphologically, each of the five genera is highly distinct in leaf morphology (pinnate or bipinnate leaves), floral symmetry (radial or bilateral), pollination syndrome (bees or birds), pollen presentation (monads or tetrads), fruit morphology, seed morphology (winged or non-winged), and dispersal syndrome (autochory, hydrochory, or anemochory). Thus, the tribe is not defined by obvious morphological synapomorphies, although there is a tendency for the occurrence of distinctively divided or foliaceous stipules (except for ; Tribe , page 165) and nodulation with a fixation thread type of nodule anatomy in , and , three of the nine non-mimosoid genera known to nodulate (Sprent 2000; Faria et al. 2022).
The four genera of the , , Benth., Harms, and Benth., form a clade in most molecular phylogenetic studies (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; Ringelberg et al. 2022), corresponding to the group A of Bruneau et al. (2008). This is a narrower definition than the morphologically-based informal group sensu Polhill and Vidal (1981), Polhill (1994) and Lewis (2005b), which also included and Harms (now tribe ), and (now placed in tribe ), as well as and (now in tribe ), a group subsequently shown to be non-monophyletic (Bruneau et al. 2001, 2008; Manzanilla and Bruneau 2012; Zhang et al. 2020). The group sensu Polhill and Vidal (1981) comprised a diverse assemblage of genera, many of which share certain characteristics with tribe (e.g., bipinnate leaves, numerous, small, regular flowers in spiciform racemes, and introrse sagittate anthers; Elias 1981a; Polhill and Vidal 1981; Luckow et al. 2000) and was considered a ‘‘transitional link’’ between the then caesalpinioids and mimosoids (Polhill and Vidal 1981; Luckow et al. 2000, 2003). As newly circumscribed, tribe is morphologically more coherent, including four genera, all with spicate inflorescences and pentamerous, diplostemonous flowers.
The four genera of the contain 35 species. However, 26 are in s.l., which is paraphyletic (e.g., LPWG 2017; Ringelberg et al. 2022; see Tribe , page 177), suggesting that generic re-delimitation will be necessary. The clade has an amphi-Atlantic distribution (Neotropics and tropical Africa) spanning a variety of biomes. Nodulation is reported in two species of , whereas and are confirmed as non-nodulators (Faria et al. 2022).
It is also notable that while tribes , and are each supported as monophyletic in recent phylogenomic analyses, the relationships among these three lineages are weakly supported and characterised by high gene tree conflict (Suppl. material 3; Ringelberg et al. 2022: Fig. 3).
and , previously placed in and the group of respectively (Table 3), have only been recovered as sister genera (Fig. 4) in one previous study (Zhang et al. 2020), although the two genera have generally been resolved in the same large clade that included and subtending lineages (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012). Morphologically, was previously considered to be a member of subfamily (Lewis and Elias 1981; Pohill 1994; Luckow 2005), but molecular phylogenetic studies have consistently placed the genus among the grade of non-mimosoid genera subtending the mimosoid clade, albeit with varying sister group relationships (Luckow et al. 2000, 2003; Wojciechowski et al. 2004; Manzanilla and Bruneau 2012; Zhang et al. 2020; Ringelberg et al. 2022; Fig. 4). These two genera exhibit disparate morphology, although they share perigynous flowers with a tubular hypanthium and showy stamens exserted from the corolla. This morphological distinctiveness is mirrored in the molecular analyses, where relatively long branches subtend the two genera. Although we here place these two morphologically disparate genera together in tribe , Ringelberg et al. (2022) noted that long-branch attraction could play a role in grouping and together in a clade. A similar phylogenetic pattern is observed in the plastid phylogenomic analyses of Zhang et al. (2020), in which the two genera also form a clade subtended by a short branch.
Tribe comprises 5 to 22 species (the genus needs to be revised because several species are of dubious taxonomic status), only two of which are in . The tribe is restricted to tropical rainforests in South America. Nodulation is reported in one species each of and (Faria et al. 2022), and is known to be absent in the other species of .
and have rarely been recovered as sister genera before (Herendeen et al. 2003a). Nevertheless, most previous studies based primarily on plastid sequence data placed these two genera as successive sisters to the mimosoid clade (Bruneau et al. 2001, 2008; Luckow et al. 2003; Bouchenak-Khelladi et al. 2010; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; Kyalangalilwa et al. 2013; Zhang et al. 2020), as found in the plastid phylogeny of Ringelberg et al. (2022), indicating another case of possible cytonuclear discordance and potentially explaining why the two genera have not previously been grouped. No clear morphological synapomorphy has been identified for the clade, but and share a combination of morphological traits only rarely found in non-Mimoseae (e.g., bipinnate leaves, small pedicellate perigynous flowers in dense spicate racemes), and both genera have highly toxic alkaloids and saponins (Tribe , page 193).
Recognising the mimosoid clade as the newly circumscribed tribe results in by far the largest tribe in subfamily in terms of numbers of species (ca. 3500) and genera (100). Previous tribal classifications of the mimosoid clade (i.e., former subfamily ) recognised five tribes: , Benth., Ingeae Benth., (Wight & Arn.) Benth., and Mimozygantheae Burkart (Elias 1981a). In formulating the new tribal classification for subfamily , the recognition of the mimosoid clade as the reinstated and re-circumscribed tribe is proposed for two main reasons. First, four of the five former tribes of are now known to be non-monophyletic and the fifth (the monospecific Mimozygantheae) to be nested within the former (Luckow et al. 2003, 2005; LPWG 2013, 2017; Koenen et al. 2020a; Ringelberg et al. 2022). Second, the ladder-like phylogenetic structure within the mimosoid clade means that any finer-scale tribal divisions would inevitably result in an undesirable proliferation of many small Linnean tribes, including a large number of monogeneric tribes (Koenen et al. 2020a).
There has been ongoing debate about which genera are included in the mimosoid clade (Luckow et al. 2000, 2003; Lewis et al. 2005; Manzanilla and Bruneau 2012). As circumscribed here, tribe matches the former subfamily (Bentham 1865; Hutchinson 1964; Polhill and Raven 1981; Lewis et al. 2005), with three exceptions (Koenen et al. 2020a; Ringelberg et al. 2022; Tribe , page 201). First, , previously considered to be a non-mimosoid caesalpinioid based on morphology (Polhill and Vidal 1981; Polhill 1994; Lewis 2005b), is firmly nested within the as an isolated early-diverging lineage (Fig. 5; Ringelberg et al. 2022). Similarly, , once considered part of (Lewis 2005b), is also nested within the , as initially found by Manzanilla and Bruneau (2012) and supported by LPWG (2017), Koenen et al. (2020a), and Ringelberg et al. (2022) (Fig. 5). Finally, was previously considered a genus of (Polhill and Vidal 1981; Polhill 1994; Luckow 2005), but is resolved in the non-mimosoid (now in tribe , page 187).
Tribe is diagnosed by valvate petal aestivation (with exceptions in and ), bipinnate leaves (except and a few scattered species in other genera), flowers that are relatively small with reduced perianth and showy androecium, mostly clustered in compact inflorescences that are commonly capitate or spicate (Tribe , page 201), and the presence of symbiosome-type (as opposed to fixation-thread-type) root nodules (Faria et al. 2022).
Tribe , as circumscribed here, is robustly supported as monophyletic and is subtended by a relatively long branch (Fig. 4). Within the phylogeny takes the form of an extensive unbalanced ladder-like topology (Fig. 5), which is not readily amenable to division into a manageable number of rank-based Linnean taxa. However, given the large size of the tribe, some form of classificatory structure is needed, and here we present a clade-based classification system for the tribe with two nested higher-level named clades – a core mimosoid clade and the ingoid clade – alongside a set of 17 named lower-level clades following Koenen et al. (2020a) and Ringelberg et al. (2022) (Fig. 5). Fourteen of the 100 genera of remain unplaced in any of these lower-level clades, eight of which are resolved in two grades, and six of which are phylogenetically isolated monogeneric lineages (Fig. 5).
The core mimosoid clade as delimited by Koenen et al. (2020a) is well supported, subtended by a notably long branch, and includes all except the and clades, the two monogeneric lineages and , and the grade (Fig. 5). The core mimosoid clade includes all with armature, with the exceptions of a single spinescent species of and of two species of Harms, which are armed with modified woody tendrils, and which occur outside the clade (Koenen et al. 2020a).
The ingoid clade, delimited by Koenen et al. (2020a), is also strongly supported, includes ca. 2000 species, i.e., almost two-thirds of species, and comprises genera of the grade and nine named clades (, , , , , , , , and clades) (Fig. 5). The ingoid clade groups all genera of with polystemonous flowers except . A synandrous androecium is exclusively found in this clade and characterises most, but not all of its genera (Koenen et al. 2020a). Based on the limited sample of mimosoid chloroplast genome sequences currently available, an expanded Inverted Repeat region of the chloroplast genome is also restricted to this clade (Dugas et al. 2015; Wang et al. 2017).
Based on the phylogeny of subfamily presented here, we recognise eleven tribes (Fig. 4) and 17 formally named clades within tribe (Fig. 5). The new classification proposed here recognises clades that are strongly supported in the phylogenomic analyses of Koenen et al. (2020a) and Ringelberg et al. (2022), many of which were already known from earlier phylogenetic studies. This classification is proposed and endorsed by the legume systematics community as reflected in the use of the Legume Phylogeny Working Group (LPWG) as the authority of the new tribes. Although an uncommon practice in botanical nomenclature, ascribing the new tribe names to the collective known as the “Legume Phylogeny Working Group” is accepted under the botanical code as stipulated in Chapter VI, Section 1 (Author Citations) and follows the approach previously used for the LPWG (2017) subfamily classification. The Legume Phylogeny Working Group authorship gives due credit to the legume systematics community and reflects the important collaborative contributions from multiple research groups over the decades that have laid the foundations for this classification.
We provide a key to genera, as well as taxonomic descriptions and notes for tribes, named clades, and all 163 genera, and illustrate the diversity of growth forms, foliage, flowers and fruits for nearly all genera. We also provide a distribution map of the native range for each genus, based on quality-controlled herbarium specimen localities and floristic surveys. The occurrence data for are from Ringelberg et al. (2023), whereas the remainder were newly assembled here. See Suppl. material 1 for sources of occurrence data and detailed data cleaning protocols. All occurrence data have been made available on Zenodo (https://zenodo.org/doi/10.5281/zenodo.8407862), unless stated otherwise. A tree file of the phylogeny presented here is available from Ringelberg et al. (2022, 2023).
Trees, shrubs, lianas, suffruticose or functionally herbaceous, occasionally aquatic, either unarmed or commonly armed with prickles, spines, or thorns; specialised extrafloral nectaries often present on the petiole and/or on the primary and secondary leaf rachides, usually between pinnae or leaflet pairs, more rarely stipular or bracteal. Stipules in lateral position and free or absent, usually entire, less frequently divided or spinescent. Leaves usually pulvinate, bipinnate, otherwise pinnate (sometimes both types on the same plant) and then mostly paripinnate, rarely imparipinnate, less often bifoliolate, modified into phyllodes or lacking, arrangement of the pinnae and leaflets mostly opposite, rarely alternate; stipels rare and not to be confused with the more commonly present paraphyllidia. Inflorescences globose or ellipsoid capitula, spicate, paniculate, racemose or in fascicles; bracteoles commonly small or absent. Flowers usually bisexual, rarely unisexual (species dioecious or monoecious), or bisexual flowers combined with unisexual and/or sterile flowers in heteromorphic inflorescences (), radially, less frequently bilaterally symmetrical, or asymmetrical; hypanthium lacking or cupular, rarely tubular; sepals (3) 5 (6–8), free or fused; petals (3) 5 (6–8), free or fused, the sepals or petals or both sometimes lacking, aestivation valvate () or imbricate and then the adaxial petal innermost; stamens commonly diplostemonous or haplostemonous, sometimes reduced to 3 or 4 (in some species), frequently polystemonous (to 100+ in some ), free or fused, sometimes heteromorphic, some or all sometimes modified or staminodial, anthers basifixed or dorsifixed, often with a stipitate or sessile apical gland, dehiscing via longitudinal slits or apical or basal poricidal slits or pores; pollen in tricolporate monads, or commonly in tetrads, bitetrads or polyads (most ); gynoecium uni- or rarely polycarpellate, 1–many-ovulate. Fruit typically dry and dehiscent, either a legume (dehiscent along both sutures) or a follicle (dehiscent along the adaxial suture only), or dry and segmented into one-seeded articles, either without a persistent margin (a lomentum) or with a persistent margins forming a replum like a frame (a craspedium), sometimes indehiscent and somewhat fleshy (an indehiscent legume), or dry and winged (a samara), when dehiscent with papery, leathery, or woody valves, dehiscence passive (inert valves), elastic, or explosive (the valves becoming curved, spirally coiled, or arched backwards), less frequently with entire valves, but with the endocarp segmented into one-seeded articles (a cryptolomentum). Seeds usually with an open (U-shaped) or closed (O-shaped) pleurogram on both faces or lacking a pleurogram, sometimes with a fleshy aril or sarcotesta, sometimes winged, in cross section terete (with a 1:1 ratio), compressed (with more or less 2:1 ratio) or flattened (with > 4:1 ratio; Gunn 1991); hilum usually apical, lens usually inconspicuous; embryo straight.
in its emended circumscription contains eleven tribes, 163 genera and ca. 4680 species. Tribes: Rchb. (27 genera / ca. 223 species), LPWG (2 / 5–22), Bronn (7 / 695), Rchb. (4 / 6), Benth. (4 / 35), LPWG (2 / 13), Nakai (3 / 20), Bronn (100 / ca. 3510), LPWG (1 / 1), Nakai (8 / 42–43), Benth. & Hook. f. (5 / ca. 113) (Fig. 4).
Pantropical, common in both wet and dry regions, with a handful of species extending to the temperate zone, less frequently frost-tolerant ( J. Clayton, Lam. and some species of Mill., Willd. and Mill.). species are infrequent above 2500 m in the tropics and are largely absent from mid- and high-elevation tropical montane forests. Generic diversity is highest in the Neotropics, and there are important centres of high species diversity in Mexico and Central America, central-east South America, Africa, Madagascar, parts of South East Asia and Australia (Fig. 3, Table 1).
This clade was referred to as the — (MCC clade) (Doyle 2011, 2012) or the — (GCM clade) (Marazzi et al. 2012). In LPWG (2017), was chosen over as the preferred name for the MCC clade, leaving open the option for naming the morphologically distinct mimosoid clade at the tribal level, as is done here (Fig. 5; see , page 201).
In the following treatments, the type species of different taxa are presented. Their homotypic (nomenclatural) synonyms are indicated by the identity symbol (≡) and heterotypic (taxonomic) synonyms by the equality symbol (=), as specified in the International Code of Nomenclature for algae, fungi, and plants (Turland et al. 2018).
Some sections of the key, for particular groups, were developed and modified from previous publications [e.g., Brenan (1955), Lewis and Elias (1981), Polhill and Vidal (1981), Nielsen (1992), Barneby and Grimes (1996), Queiroz (2009), Gagnon et al. (2016), Brown et al. (2022), Hughes et al. (2022b), Lima et al. (2022), Soares et al. (2022), Souza et al. (2022b), Tamayo-Cen et al. (2022)]. The distribution data used in some couplets refers to the native geographic range of the genus in question. Genera that key out in more than one place in the key are indicated by an asterisk (*). See Glossary, Schemes 1–7, for definitions and illustrations of selected morphological terms pertinent to .
Figs 6
, 7
, 8
, 9
, 10
, 11

The most inclusive crown clade containing Urb. and L., but not Sim, (Splitg.) Sandwith or L. (Fig. 6).
The tribal name was first published by Reichenbach (1832) to accommodate the genus . Irwin and Barneby (1981) placed in its own new subtribe H.S. Irwin & Barneby of tribe , and they would have accorded the genus tribal rank, i.e., a reinstatement of Reichenbachʼs (1832) taxon, “if not dissuaded by others” (Irwin and Barneby 1981: 98). Herendeen et al. (2003b), based on molecular and morphological data, grouped seven disparate caesalpinioid genera in their “ clade”, which included two major subclades: the — clade and the — clade. More recent studies do not support the monophyly of these two clades together (Manzanilla and Bruneau 2012; LPWG 2013, 2017; Ringelberg et al. 2022) and they are here recognised under the two reinstated tribes and , respectively. Tribe is resolved as sister to the rest of (Fig. 6) and comprises six species in four genera (two of the genera are monospecific).
Generic relationships in tribe . Left part of figure shows complete genus-level phylogeny with the indicated with a red rectangle. Branch lengths are expressed in coalescent units and terminal branches were assigned an arbitrary uniform length for visual clarity. Support for relationships is based on fractions of supporting and conflicting gene trees: pie charts show gene tree support and conflict per node (blue representing supporting gene trees, green gene trees supporting the most common alternative topology, red gene trees supporting further alternative topologies, grey gene trees uninformative for this node), and numbers above pie charts are Internode Certainty All support values [both calculated with PhyParts (Smith et al. 2015)]. If present, red numbers below pie charts are non-significant (i.e. > 0.05) outcomes of ASTRAL’s polytomy test (Sayyari and Mirarab 2018), which tests for each node whether the polytomy null model can be rejected. Monophyletic genera are represented by single branches; see Suppl. material 2 for a phylogeny with all accessions. See Suppl. material 3 for gene tree support across the phylogeny. The phylogeny is a pruned version of the backbone phylogeny of Ringelberg et al. (2023), where full details of the data and phylogenomic analysis methods are presented.
Only two characters have been found to be shared by members of tribe : bipinnate leaves with a single terminal pinna (Fig. 7E; although leaves are usually singly pinnate) and three of the four genera (all except , which is sister to the three other genera) have a large deletion in the plastid trnL intron (Bruneau et al. 2001). Three of the genera ( is the exception) are dioecious (although occasional bisexual flowers occur). and are sister genera, both have simple inflorescences, a haplostemonous androecium, glabrous stamen filaments, alternate first seedling leaves, septate wood fibres, and a chromosome base number of x = 12. They differ in unisexual () versus bisexual () flowers, absence of petals () versus petals present (), and fruit type: indehiscent () versus dehiscent (). Three of the four genera are confirmed as non-nodulating and the status of is not known (Sprent 2001; Faria et al. 2022).
Figs 7
, 8

Shrub or small tree, with well-developed brachylasts. Stipules spinescent, caducous. Leaves pinnate when juvenile, bipinnate with a terminal pinna when mature, leaflets opposite to subopposite, sessile. Inflorescence a sparsely branching panicle of spikes arising from a thickened woody brachyblast. Flowers unisexual, staminate flowers with a prominent pistillode, pistillate flowers with staminoidia; sepals free to a very short hypanthium; petals 5 (6); a short cupuliform disk present centrally; stamens (or staminodes) number more than twice sepal number (12 or more per flower); pollen markedly irregular and coarsely reticulate, porate with prominent pores; ovary with an appressed rust-coloured indumentum, stigma terminal and capitate. Fruits oblong-ellipsoid, subterete, thick-walled, indehiscent, 1–few-seeded. Seeds ovate, compressed, surrounded by copious pulp, pleurogram lacking (Fig. 7A).
Monospecific (), endemic to Hispaniola (Dominican Republic, Haiti, and the Island of Gonâve) (Fig. 8).
Distribution of based on quality-controlled digitised herbarium records. See Suppl. material 1 for the source of occurrence data.
Originally placed by Urban (1923) in the , was subsequently transferred to the by Urban (1928). The genus was placed in the eclectic group of by Polhill and Vidal (1981), along with , both with unisexual flowers and petals not covered by the sepals in bud, but phylogenetic analyses clearly resolved as part of the “ clade”, either sister to the genera (Herendeen et al. 2003b) or sister to those of (Bruneau et al. 2008), the latter as in Ringelberg et al. (2022).
Lewis (2005b); Urban (1928).
Figs 7
, 9

Unarmed trees or shrubs, dioecious; brachyblasts absent. Stipules inconspicuous, minute and caducous. Leaves bipinnate, ending in a terminal pinna; leaflets alternate. Inflorescences axillary, spike-like racemes of small, subsessile flowers, usually aggregated into panicles. Flowers unisexual, actinomorphic, 4-merous (sepals and petals 4 per flower), greenish; sepals equal, petals equal, both whorls imbricate in young bud; androecium diplostemonous in staminate flowers, with one whorl of 4 fertile stamens, their filaments with an apical tuft of hairs behind the anthers, and one whorl of 4 hairy staminodes, lacking anthers; pollen with a scabrate-punctate sculpture pattern; pistillate flowers with a stipitate, compressed-fusiform ovary, stigma capitate and bilobed. Fruits membranous, indehiscent, compressed, 4-winged (in two unequal pairs), 1-seeded (Fig. 7B). Seeds trapezoid, subterete, club-shaped, pleurogram lacking.
Two species, both endemic to Madagascar (Fig. 9).
Distribution of based on quality-controlled digitised herbarium records. See Suppl. material 1 for the source of occurrence data.
Although placed in the group of the by Polhill and Vidal (1981), the genus was resolved as sister to an – clade in a study of the “ clade” by Herendeen et al. (2003b), sharing with these two genera a bipinnate leaf with an unequal leaflet base, and this relationship is supported in the phylogenomic analysis of Ringelberg et al. (2022) (Fig. 6).
Du Puy and Rabevohitra (2002); Lewis (2005b).
Figs 7
, 10

Unarmed evergreen tree, bark pale grey, smooth,

---

Sections

"[{\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B507\", \"B442\", \"B443\", \"B483\", \"B505\"], \"section\": \"\\ufeff\\ufeffIntroduction - Classification of subfamily\", \"text\": \"With close to 800 genera and more than 22,000 species (LPWG 2023), the  () is the third largest angiosperm family in number of species after  and . Legumes include a large set of economically important food crops that provide highly nutritious sources of plant protein and micronutrients, which can greatly benefit health and livelihoods. They have been domesticated alongside grasses in different areas of the world since the beginnings of agriculture and have played a key role in its early development. Legumes are also important sources of fodder and green manure in both temperate and tropical regions, and are used for their wood, tannins, oils and resins, in the manufacture of varnishes, paints, dyes and medicines, and in the horticultural trade. Legume diversification probably started close to the Cretaceous-Paleogene boundary (ca. 66 Ma) (Koenen et al. 2020b, 2021), giving rise to one of the most spectacular examples of evolutionary diversification in plants. Modern legumes are exceptionally diverse, morphologically, physiologically and ecologically (Lewis et al. 2005; LPWG 2017).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B505\", \"B505\", \"B442\", \"B931\", \"B932\", \"B442\", \"B443\"], \"section\": \"\\ufeff\\ufeffIntroduction - Classification of subfamily\", \"text\": \"In 2017, the Legume Phylogeny Working Group (LPWG 2017) revised the higher-level classification of the family and recognised six monophyletic subfamilies within the monophyletic . Under the LPWG classification, subfamily  DC. was re-circumscribed, and  LPWG,  Burmeist.,  LPWG and  LPWG (all of which were previously part of  sensu lato at different ranks) were recognised as distinct subfamilies along with an unchanged  DC. The former subfamily  DC., which is phylogenetically nested within , was subsumed within the re-circumscribed  and has since been referred to as the mimosoid clade (LPWG 2017). The idea that  comprises six main lineages, corresponding to these six subfamilies, is now widely accepted and has been confirmed by recent phylogenomic analyses of large nuclear gene and plastome DNA sequence datasets (Koenen et al. 2020b; Zhang et al. 2020; Zhao et al. 2021), which show robust support for all six subfamilies. Phylogenomic evidence suggests that the six subfamilies likely diverged very rapidly such that gene tree conflict obscures relationships among some of the subfamilies, but  is supported as sister to  (Koenen et al. 2020b, 2021).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B668\", \"B667\", \"B483\", \"B504\", \"B505\", \"B505\", \"B266\", \"B796\", \"B267\", \"B919\", \"B157\", \"B158\", \"B917\", \"B932\", \"B183\", \"B504\", \"B505\", \"B441\", \"B712\", \"B388\"], \"section\": \"\\ufeff\\ufeffIntroduction - Classification of subfamily\", \"text\": \"Within subfamilies, new phylogenies of many legume groups have unequivocally demonstrated the non-monophyly of the tribes recognised in the classifications of Polhill and Raven (1981), Polhill (1994) and Lewis et al. (2005), and the need for new classifications (LPWG 2013, 2017). Following publication of the LPWG (2017) subfamily classification, phylogenetically-based tribal and clade-based higher-level classifications were developed for subfamily  (Estrella et al. 2018) and informally for  (Sinou et al. 2020), but complete higher-level phylogenetically-based classifications are still lacking for  and . Subfamily , comprising a single species, requires no classification, and , with just 18 genera, may not be easily amenable to, or in need of, additional higher-level subdivisions, although phylogenetic studies are ongoing (e.g., Falc\\u00e3o et al. 2023). For , the largest of the subfamilies, despite ongoing progress in the understanding of phylogenetic relationships (Wojciechowski et al. 2004; Cardoso et al. 2012, 2013; Wojciechowski 2013; Zhao et al. 2021; Choi et al. 2022), more data and more complete taxon sampling are needed before a robust and stable phylogenetically-based classification system can be fully developed. For , although some questions persist about the monophyly and placement of a small subset of genera and some on-going uncertainty surrounding generic delimitation and relationships (LPWG 2013, 2017; Koenen et al. 2020a), recent work has clarified most of these problems (Ringelberg et al. 2022), many of which were resolved in a series of papers in Advances in Legume Systematics 14, Part 1 (Hughes et al. 2022a).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B441\", \"B712\", \"B268\", \"B712\", \"B713\"], \"section\": \"\\ufeff\\ufeffIntroduction - Classification of subfamily\", \"text\": \"A new higher-level classification of subfamily  is therefore now both feasible and timely. Here we use the phylogenomic backbones for subfamily  from Koenen et al. (2020a) and Ringelberg et al. (2022) as the basis for developing a new higher-level classification of the subfamily. This new phylogenetic classification provides a solid system for communication and a framework for downstream analyses of biogeography, trait evolution and diversification (e.g., Faria et al. 2022; Ringelberg et al. 2022, 2023), as well as for guiding efforts towards fully revising the taxonomy of still understudied genera.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B505\", \"B388\", \"B506\", \"B712\", \"B505\", \"B915\", \"B361\", \"B131\", \"B443\", \"F1\", \"F2\", \"F3\"], \"section\": \"\\ufeff\\ufeffSubfamily : Diversity and distribution\", \"text\": \" sensu LPWG (2017) is the second largest subfamily of legumes with ca. 4680 species placed in 163 genera (Hughes et al. 2022a; LPWG 2022; Ringelberg et al. 2022). Within this subfamily, ca. 3500 species and 100 genera are placed in the former subfamily  (the mimosoid clade of LPWG 2017), which we here recognise as the reinstated, but newly circumscribed, tribe , a tribal name first published by Bronn in 1822 (see below).  date to the late Paleocene when the subfamily is known from fossil bipinnate leaves from Colombia (Wing et al. 2009; Herrera et al. 2019). These fossils indicate that  were an abundant element in the earliest Neotropical rain forests in the Paleocene and time-calibrated legume phylogenies suggest that  started to diversify around 58 million years ago (Lavin et al. 2005; Bruneau et al. 2008; Koenen et al. 2021).  have thus diversified throughout the Cenozoic and now comprise diverse, abundant, and sometimes dominant elements across all major lowland tropical biomes, including rain forests, savannas and seasonally dry forests (Figs 1, 2, 3).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B498\", \"B498\", \"B498\"], \"section\": \"\", \"text\": \"A genus richness across floristic realms (according to Liu et al. 2023). The numbers within the circles represent the total number of genera in each realm. The numbers on the lines represent the number of genera shared between two realms (> 10 genera) B Number of  genera in the floristic subrealms (sensu Liu et al. 2023). The numbers associated with the two polygons indicate the number of genera restricted to the two major blocks of tropical and subtropical areas in the New World and the Old World (maps modified from Liu et al. 2023, CC BY 4.0).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"S1\"], \"section\": \"\", \"text\": \"Map showing the global distribution of  species richness. Numbers of  species per one degree latitude / longitude grid cell. Infraspecific taxa are not counted individually but are included at the species level. All maps in this special issue are based on quality-controlled occurrence data from digitised herbarium specimens and floristic surveys (see Suppl. material 1 for details on occurrence data and methods used to generate maps).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B712\", \"B713\"], \"section\": \"\\ufeff\\ufeffSubfamily : Diversity and distribution\", \"text\": \" are almost entirely woody perennials, but they are extremely diverse in stature and habit \\u2013 including lianas, trees of all sizes, up to rain forest canopy emergents (e.g.,  Ducke,  Ducke), shrubs, functionally herbaceous geoxyles, and two herbaceous aquatic species ( Lour.). Similarly, the subfamily is highly diverse in floral and fruit morphology. One of the hallmarks of , as in many other plant groups, is repeated morphological and ecological convergences whereby similar leaf, flower and fruit morphologies, and ecological adaptations, have apparently been reinvented multiple times across lineages and through time (Ringelberg et al. 2022, 2023).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B533\", \"B388\", \"B712\", \"B268\", \"B153\", \"B443\", \"B932\", \"B213\", \"B318\", \"B782\"], \"section\": \"\\ufeff\\ufeffSubfamily : Diversity and distribution\", \"text\": \" is the only legume subfamily that has bipinnate leaves, which are prevalent but not universal across the subfamily (see Glossary, Schemes 1\\u20137). A minority of genera have species with pinnate leaves, and leaves modified into phyllodes occur in most species of the large, mainly Australian, genus  Mill. and in a few species in other unrelated genera including  Mill. and  L. The leaves themselves, especially the bipinnate leaf, can be extremely large (e.g.,  Vogel leaves are > 1 m long) to highly reduced; aphyllous, or nearly aphyllous, species occur in some genera [e.g., , ,  (L.) Moench,  Raf.,  Burkart]. Across all legumes, seismonasty, i.e., leaf movements prompted by touch, is known only within subfamily , in the genera  and  (tribe ). Extrafloral nectaries (EFNs) are present in the majority of  (Scheme 2), are morphologically extremely diverse, and often conspicuous and abundant on the petiole or leaf rachides between pinnae or leaflet pairs, and in a few genera (e.g.,  F. Muell.,  Britton & Rose) on floral bracts (Marazzi et al. 2019). A subset of  genera are armed with prickles, spines or thorns (Scheme 1), but armature is highly variable, has clearly evolved multiple times across the subfamily, and can vary within clades and even within genera (Hughes et al. 2022a; Ringelberg et al. 2022). Most genera of the mimosoid clade (tribe  here) are confirmed nodulators, whereas just nine of the 63 non-mimosoid genera in the subfamily are currently known to nodulate. These nine genera are phylogenetically intermingled with confirmed non-nodulating genera, suggesting multiple evolutionary transitions between non-nodulating and nodulating lineages (Faria et al. 2022). Analyses of gene duplications have shown that several whole genome duplications (WGDs) occurred during the early evolution of the family  (Cannon et al. 2015; Stai et al. 2019; Koenen et al. 2021; Zhao et al. 2021), although the exact number and placement of these WGDs remain uncertain. Within , polyploidisation has also occurred numerous times during the Neogene in several genera across the subfamily [e.g.,  Benth.,  (A. DC.) Wight & Arn., ,  Wight & Arn., ; Dahmer et al. 2011; Govindarajulu et al. 2011a; Simon et al. 2011].\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B20\", \"B93\", \"B712\", \"B821\"], \"section\": \"\\ufeff\\ufeffSubfamily : Diversity and distribution\", \"text\": \"Across the subfamily, inflorescences and flowers are morphologically highly variable. The inflorescences can be racemose, paniculate or in fascicles and the  have characteristic capitate or spicate and frequently heteromorphic inflorescences, often with some sterile flowers, some of which develop showy staminodia (Schemes 3, 4). Although usually bisexual, flowers can also be unisexual, and in the  inflorescences can include a mixture of both bisexual and unisexual flowers with or without sterile flowers. The flowers are generally pentamerous, but there are many variations [3\\u20136 (8) sepals or petals], and in some species, sepals and/or petals are absent ( L.). Flowers are generally radially symmetrical in several  tribes, including the , but in other clades the flowers are bilaterally symmetrical or asymmetrical. Although a majority of  flowers are bee pollinated, specialised bat, bird, butterfly and moth pollinated flowers are also common (Arroyo 1981). In addition to having species with pollen in the more typical tricolporate monads,  is the only subfamily of legumes with taxa where pollen is arranged in polyads (Scheme 6). In  the pollen arrangement is extremely variable across and sometimes within genera, with pollen in monads, tetrads, bi-tetrads and polyads. Fruit morphology is particularly homoplasious, and in the  has proved misleading for generic delimitation (Borges et al. 2022; Ringelberg et al. 2022; Souza et al. 2022b). This diversity of fruit morphology (Schemes 6, 7) reflects adaptations to different seed dispersal syndromes, including passive, elastic and explosive dehiscence, as well as seed dispersal by water, wind, large herbivores, ants, and birds.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"T1\", \"T2\"], \"section\": \"\\ufeff\\ufeffSubfamily : Diversity and distribution\", \"text\": \"The subfamily is most diverse in lowland tropical and subtropical regions, only rarely occurring above 2500 m elevation, but a minority of genera have species in warm temperate zones that are not prone to severe frosts across the Americas, Europe, Asia, and Australia. More than half of  genera naturally occur in the Americas (104 of 163 genera), of which 84 are endemic. Africa (including Madagascar) has the second highest number of  genera, with 59 genera, 29 of which are endemic, followed by Asia (40 genera, 7 endemic), and Australia and the Pacific (27 genera, 6 endemic; See details in Tables 1, 2).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B498\"], \"section\": \"\", \"text\": \" genera richness across global floristic realms and subrealms (according to Liu et al. 2023).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B498\", \"T1\", \"B498\", \"F1\"], \"section\": \"\\ufeff\\ufeffSubfamily : Diversity and distribution\", \"text\": \"The Neotropical floristic realm (sensu Liu et al. 2023) has 101  genera with native species, of which 71 are endemic to this realm (Table 1). Liu et al. (2023) divided the Neotropical realm into the tropical Brazilian subrealm (81 genera, 25 endemic) and the Subtropical American subrealm (76 genera, 17 endemic). The Subtropical American subrealm includes the northern (69 genera, 13 endemic) and southern (35 genera, 3 endemic) extremes of the Neotropical region (Fig. 1). The genera  (Benth.) Engelm. & A. Gray and  Burkart are restricted to this subrealm and have an amphitropical distribution, whereas the genera  Klotzsch and  Willd. have a similar distribution, but with a few species reaching the Brazilian subrealm. The second largest floristic realm for  genera is the African one, with 59 genera, of which 30 are endemic, primarily in the Guineo-Congolian subrealm (41 genera, 14 endemic). Within the Indo-Malesian realm, there are 40 genera (11 endemic), almost evenly distributed between the Indian (29 genera, 4 endemic) and the Malaysian subrealms (32 genera, 5 endemic).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"T2\"], \"section\": \"\\ufeff\\ufeffSubfamily : Diversity and distribution\", \"text\": \"Nine genera have a pantropical distribution, occurring in all tropical floristic realms [ L., ,  Adans.,  L., ,  (Vogel) Benth.,  Raf., , ; Table 2]. The genera  and  R. Br. occur in all tropical floristic realms except for the Australian realm. Seven genera ( C. Presl,  R. Vig.,  L.,  Barneby & J.W. Grimes,  L.,  Benth.,  Cav.) represent transatlantic disjunctions as they are exclusively distributed in the Neotropical and African realms.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"F2\", \"F3\", \"B561\", \"F2\", \"F3\", \"F2\", \"F3\", \"T1\"], \"section\": \"\\ufeff\\ufeffSubfamily : Diversity and distribution\", \"text\": \"Species richness, determined from occurrence records, indicate south-central Mexico and Central America, central-eastern Brazil, and southwestern Australia to be the most diverse regions (Fig. 2), with multiple one-degree grid cells in these regions containing more than a hundred  species. However, patterns of generic richness (Fig. 3) show that the high species diversity in Australia is largely attributable to the hyper-diverse genus , with over a thousand species. Australia is therefore characterised by high species but low generic richness. Important hotspots of  diversity also occur in continental Africa and Madagascar. Asia is the least diverse tropical continent, although it contains multiple species-rich lineages such as , especially in South East Asia. The spatially biased availability of digitised occurrence data (Meyer et al. 2016) leads to an underestimation of  richness across large parts of tropical Africa, India, continental South East Asia, and the Amazon, as is apparent in Figs 2, 3. Nevertheless, our analyses (Figs 2, 3) represent an accurate depiction of relative differences in  continental richness patterns (Table 1).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B752\", \"B300\", \"B713\", \"B752\", \"B711\", \"B713\", \"F1\", \"F2\", \"F3\", \"B713\"], \"section\": \"\\ufeff\\ufeffSubfamily : Diversity and distribution\", \"text\": \"The wide geographical distribution of  is matched by an equally wide ecological amplitude across the full precipitation spectrum of the tropics, spanning a 100-fold gradient in mean annual precipitation from arid deserts to seasonally dry tropical forests and savannas, and tropical rainforests (Schrire et al. 2005a; Gagnon et al. 2019; Ringelberg et al. 2023). Although  species are important components of many wet regions of the world, it is notable that some of the major hotspots of species and especially generic diversity coincide with areas dominated by seasonally dry vegetation in southern Mexico, north-eastern Brazil, and northern and south-western Madagascar, these being key areas of the succulent biome sensu Schrire et al. (2005a) and Ringelberg et al. (2020), plus seasonally dry subtropical south-western Australia. The subfamily thus has no obvious overriding wet or dry affinity, but rather has switched between wet and dry tropical biomes multiple times and has diversified substantially within each (Ringelberg et al. 2023). This ecological adaptability is undoubtedly, at least in part, a function of the evolutionary lability of life history strategies, including adaptations to fire, which has allowed  species to become important, diverse, and abundant elements of all lowland tropical biomes. It is also clear that  have been able to disperse across oceans numerous times to reach all tropical continents and the majority of lower latitude islands and island archipelagos (Figs 1, 2, 3). In contrast to this wide adaptability across tropical precipitation regimes and vegetation types,  show high tropical niche conservatism and very limited adaptability to cold temperatures and frost with just a small subset of lineages and species extending into temperate vegetation (Ringelberg et al. 2023).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B931\", \"B932\", \"B848\", \"B123\", \"B215\"], \"section\": \"\\ufeffReference phylogeny\", \"text\": \"Stability is one of the most important qualities of any taxonomic classification. It is therefore crucial that the phylogenetic framework used for assigning names to clades is robust and unlikely to change with sampling of additional taxa or genomic regions in the future. Based on the number and identity of the taxa included, the size of the genomic dataset, and the phylogenomic methods used to infer the phylogeny and assess its robustness, the phylogenetic framework employed here is currently the best available for taxonomic classification of . This is confirmed by its overall agreement with previous smaller-scale phylogenies (see details below) and other recent independent phylogenomic studies (Zhang et al. 2020; Zhao et al. 2021). Furthermore, throughout this compendium the phylogeny is presented in such a way to allow easy assessment of underlying genomic support for all nodes subtending named clades, and in general clades named here are subtended by well-supported nodes on long branches. Absolute stability can never be guaranteed, and sampling of additional taxa might well result in different topologies and generic re-delimitation in some parts of the tree, such as in the  grade (Terra et al. 2022) or the  clade (Brown et al. 2022; Demeulenaere et al. 2022). Nevertheless, we consider the phylogenomic framework robust, and an adequate basis for the new classification presented here.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"F4\", \"F5\", \"S2\", \"S3\", \"B441\", \"B712\", \"B713\", \"B441\", \"B712\", \"B713\", \"B388\"], \"section\": \"\\ufeffReference phylogeny\", \"text\": \"The classification proposed here uses as its framework the most comprehensively sampled phylogenetic analysis of  to date (Figs 4, 5, Suppl. materials 2, 3). This new phylogeny is based on Koenen et al. (2020a) and Ringelberg et al. (2022, 2023). By developing a clade-specific bait set for targeted enrichment of 964 nuclear genes, Koenen et al. (2020a) generated a DNA sequence dataset an order of magnitude larger than those used previously, thereby providing the greatly enhanced phylogenetic resolution required for classifying tribe . Capitalising on these foundations using a slightly modified version of the gene set covering 997 nuclear genes, and importantly extending the taxon sampling to include 300 additional species covering not only  but also most genera of non-mimosoid , as well as conducting transcontinental sampling of genera that occur across different continents, Ringelberg et al. (2022, 2023) established a robust phylogenomic hypothesis for subfamily  as a whole. These studies revealed or confirmed the non-monophyly of 22 genera, and this was the basis for the re-circumscription of 15 of these genera presented in Advances in Legume Systematics 14, Part 1 (Hughes et al. 2022a).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"F5\", \"S2\", \"S3\", \"B713\"], \"section\": \"\", \"text\": \"Phylogeny of  showing the tribal classification presented here. The names and phylogenetic placements of all 63 non-Mimoseae  genera are shown and known generic non-monophyly is indicated with terminal names of non-monophyletic genus in bold. The most likely placements for four unsampled genera are indicated with dashed lines; see respective treatments for details. Tribe  has been collapsed (see Fig. 5). Branch lengths are expressed in coalescent units, and terminal branch lengths have been assigned an arbitrary uniform length for visual clarity. Monophyletic genera are represented by single branches; see Suppl. material 2 for a phylogeny with all accessions. See Suppl. material 3 for gene tree support across the phylogeny. The phylogeny is a pruned version of the backbone phylogeny of Ringelberg et al. (2023), where full details of the data and phylogenomic analysis methods are presented.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"S2\", \"S3\", \"B713\"], \"section\": \"\", \"text\": \"Phylogeny of tribe  showing the clade-based classification of the tribe with two named higher-level and 17 named lower-level clades. The names and phylogenetic placements of all 100  genera are shown, and known generic non-monophyly is indicated with terminal names of non-monophyletic genera in bold. The most likely placement of the unsampled genus  is indicated with a dashed line; see  clade treatment (page 319) for details. Branch lengths are expressed in coalescent units, and terminal branch lengths have been assigned an arbitrary uniform length for visual clarity. Monophyletic genera are represented by single branches; see Suppl. material 2 for a phylogeny with all accessions. See Suppl. material 3 for gene tree support across the phylogeny. The phylogeny is a pruned version of the backbone phylogeny of Ringelberg et al. (2023) where full details of the data and phylogenomic analysis methods are presented.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B131\", \"B505\", \"B430\", \"B130\", \"B131\", \"B526\", \"B931\", \"B932\", \"B299\", \"B489\", \"B712\", \"B441\", \"B712\", \"B713\"], \"section\": \"\\ufeffReference phylogeny\", \"text\": \"The phylogenomic analysis presented here includes 420  species representing all but five of the 163 genera. The five missing genera are:  Aubl., which has three species and is likely a member of tribe  (e.g., Bruneau et al. 2008; LPWG 2017; Kates et al. 2024);  Tul., placed here in a phylogenetically isolated, monospecific tribe (e.g., Bruneau et al. 2001, 2008; Manzanilla and Bruneau 2012; Zhang et al. 2020; Zhao et al. 2021);  Harms and  Gagnon & G.P. Lewis, both monospecific genera of tribe  (Gagnon et al. 2016); and  C. Presl, a monospecific genus in the  clade of tribe  (Lima et al. 2022). Although only about 10% of species were sampled in the analyses underlying the phylogenies presented here, several lower-level phylogenetic analyses of specific clades have been published and provide additional support for the groupings presented (Ringelberg et al. 2022). Furthermore, taxon sampling was specifically designed to cover taxonomic diversity spanning the root nodes of subclades and genera (Koenen et al. 2020a; Ringelberg et al. 2022, 2023).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B736\", \"B441\", \"B442\", \"B443\", \"B931\", \"B441\", \"B712\", \"B713\", \"B441\", \"B712\", \"B441\", \"B712\", \"B713\", \"F41\", \"F232\", \"S3\", \"F121\", \"B848\", \"B712\"], \"section\": \"\\ufeffReference phylogeny\", \"text\": \"A common feature of phylogenomic analyses employing large numbers of genes is the presence of conflict among gene trees, i.e., phylogenies based on individual genes (Salichos and Rokas 2013; Koenen et al. 2020a, 2020b, 2021; Zhang et al. 2020). Such gene tree conflict is widespread across many nodes in the  phylogeny presented here (Koenen et al. 2020a; Ringelberg et al. 2022, 2023). The main cause of this conflict appears to be lack of signal for many nodes in individual gene trees. In such cases, the relevant node in the species tree is only supported by a relatively small number of gene trees, but there is no strong support among the gene trees for any of the alternative, conflicting topologies. The presence of this type of gene tree conflict, indicative of lack of signal rather than true gene tree disagreement, does not preclude naming a clade subtended by such a node, as there is no strong reason to assume that including additional accessions or genomic regions would result in different relationships (Koenen et al. 2020a; Ringelberg et al. 2022). However, in a few places across the tree there is stronger support for alternative conflicting topologies (Koenen et al. 2020a; Ringelberg et al. 2022, 2023). In general, such instances of strong conflict are not found in nodes subtending clades named here, but rather in the relationships within clades [e.g., in parts of the  (Fig. 34) and  clade (Fig. 225)] and between clades [e.g., the relationships between tribes , , and  (Suppl. material 3) or between the  and  clades and  Stapf and  Hoyle (Fig. 114)]. Similarly, strong gene tree conflict, including between the nuclear and chloroplast genomes, may affect generic delimitation in some parts of the tree, such as in  (Terra et al. 2022) and  Schott (Ringelberg et al. 2022). Where relevant for the new classification presented in this compendium, these instances of strong gene tree conflict are described below.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B930\", \"B419\", \"B746\", \"B722\", \"B798\", \"S3\", \"B712\", \"B713\"], \"section\": \"\\ufeffReference phylogeny\", \"text\": \"The reference phylogeny used here as the basis for the new classification was inferred using ASTRAL (Zhang et al. 2018), deploying the multi-species coalescent approach based on individual gene trees, which performs well on datasets with inter-genic conflict (Jiang et al. 2020). We always report the non-significant (i.e., > 0.05) outcomes of the ASTRAL polytomy test (Sayyari and Mirarab 2018), which tests for each node whether the polytomy null model can be rejected. Because conventional phylogenetic support metrics, such as bootstrap support, tend to be inflated in large phylogenomic datasets (Rokas and Carroll 2006), we report support for nodes in the species tree using measures of individual gene tree conflict and concordance, calculated using PhyParts (Smith et al. 2015) (Suppl. material 3). The impacts of phylogenomic methods and the presence of conflict among different gene and species trees on taxonomic decisions were further discussed in Ringelberg et al. (2022). Full details of the phylogenomic data and analyses were presented in Ringelberg et al. (2023).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B505\", \"B505\"], \"section\": \"\\ufeffIntegrating tribal and clade-based classifications\", \"text\": \"Under the LPWG (2017) subfamilial classification, subfamily  was the most difficult and controversial to delimit because of the inclusion of the formerly recognised, widely accepted and morphologically distinctive subfamily . Abandoning the well-known , an important disadvantage of adopting the six-subfamily classification, was mitigated by continuing to recognise this lineage as a named clade, informally referred to simply as the mimosoid clade until now (LPWG 2017), but here formally reinstated as the re-circumscribed and expanded tribe  within the new Linnean tribal classification proposed here.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B516\", \"B131\", \"B712\", \"B669\", \"B471\", \"B526\", \"B505\", \"B441\", \"B712\", \"F4\"], \"section\": \"\\ufeffIntegrating tribal and clade-based classifications\", \"text\": \"Although  have traditionally been diagnosed by a series of diagnostic features, notably valvate petal aestivation and flowers with a reduced perianth and showy androecium, mostly clustered in compact inflorescences, the morphological distinctions between the mimosoid clade and some genera of the subtending grade of caesalpinioid lineages are not always clear-cut. For example, , once considered to be in , is placed outside the mimosoid clade in molecular phylogenetic and phylogenomic analyses (Luckow et al. 2003; Bruneau et al. 2008; Ringelberg et al. 2022). Conversely, , which has always been considered a non-mimosoid caesalpinioid legume (Polhill and Vidal 1981; Lewis 2005b), is placed within the mimosoid clade in all molecular phylogenetic analyses (Manzanilla and Bruneau 2012; LPWG 2017; Koenen et al. 2020a; Ringelberg et al. 2022). However, the long branch subtending what could be considered equivalent to the old subfamily  (plus or minus these few genera) (Fig. 4), together with the strong phylogenetic support for the clade, provide ample justification for recognising it at the tribal level.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B456\", \"B918\", \"B357\", \"B441\", \"B796\", \"B917\"], \"section\": \"\\ufeffIntegrating tribal and clade-based classifications\", \"text\": \"The new classification proposed here thus follows a traditional Linnean approach but is complemented by a clade-based classification of the large tribe . Rank-free naming of clades within subfamilies and tribes has been prevalent in the legume literature, with many examples of clade names that have become widely used and accepted, such as the dalbergioid clade (Lavin et al. 2001) and the inverted repeat [IR]-lacking clade (Wojciechowski et al. 2000) of ; the  clade (Herendeen et al. 2003b) and the ingoid clade (Koenen et al. 2020a) of ; and the Bauhinia and Phanera clades of  (Sinou et al. 2020). Naming clades provides useful additional information even after a fully developed and stable subfamily and tribal classification is established. As noted by Wojciechowski (2013), use of Linnean names does not preclude a system that also defines and names clades and their overall relationships outside of the Linnean framework. Instead, the two can be considered complementary for developing a stable, flexible and useful classification of legumes.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"F4\", \"T3\"], \"section\": \"\\ufeffThe new classification\", \"text\": \"The new classification of subfamily  comprises eleven tribes, which are either new, reinstated or re-circumscribed at this rank:  Rchb.,  Bronn,  LPWG,  Rchb.,  Benth.,  LPWG,  Nakai,  Bronn,  LPWG,  Nakai, and  Benth. & Hook. f. (Fig. 4). Although many of these lineages have been recognised and named in the past, either as tribes or informal generic groups, their circumscriptions have varied widely and changed over the past decades, such that all the tribes described here differ in generic membership from those previously recognised (Table 3).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B668\", \"B483\"], \"section\": \"\", \"text\": \"Comparison of the new phylogeny-based classification for  with classifications for these genera published in Advances in Legume Systematics, Part 1 (Polhill and Raven 1981) and Legumes of the World (Lewis et al. 2005).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B224\", \"B225\", \"B531\", \"B505\"], \"section\": \"\\ufeffThe new classification\", \"text\": \" as defined here includes elements from three previously recognised major groups: part of old sense tribe , part of old sense tribe , and the nested subfamily . This broad clade has been referred to as the -- or MCC clade (Doyle 2011, 2012), or the -- or GCM clade (Marazzi et al. 2012). In 2017,  was chosen over  as the preferred name for this large clade, even though the two names were published at the same date (LPWG 2017). By choosing the name , this left open the option of recognising the morphologically distinct mimosoid clade at the tribal level, as proposed here.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B669\", \"B130\", \"B131\", \"T3\", \"B669\", \"B669\", \"B130\", \"B505\", \"B669\", \"B669\", \"B469\", \"B130\", \"B406\", \"B505\", \"T3\", \"B505\"], \"section\": \"\\ufeffThe new classification\", \"text\": \"In their treatment of tribe  in Advances in Legume Systematics Part 1, Polhill and Vidal (1981) recognised eight informal generic groups, based primarily on differences in floral morphology. Six of these generic groups, namely the  group,  group,  group,  group,  group, and  group, are here recognised at the tribal level, albeit with modified generic compositions because most of Polhill and Vidal\\u2019s named groups have been shown to be non-monophyletic in subsequent phylogenetic analyses using molecular data (e.g., Bruneau et al. 2001, 2008) (Table 3). The monospecific  group of Polhill and Vidal (1981) groups with members of tribe , rather than being considered a distinct tribe. In addition, the  sensu Polhill and Vidal (1981) included the genus  C. Presl. (as a distinct monogeneric group), which has since been shown to be placed in subfamily  (Bruneau et al. 2001; LPWG 2017). Tribe  of Polhill and Vidal (1981) had been considered to be paraphyletic, at least implicitly, for some time (Polhill and Vidal 1981; Lewis 1998) and this has since been confirmed by phylogenetic analyses which found the tribe to be polyphyletic (Bruneau et al. 2001). The other major group that forms part of  is what was considered tribe  by Irwin and Barneby (1981). Within the tribe, they recognised five disparate subtribes,  H.S. Irwin & Barneby,  H.S. Irwin & Barneby,  H.S. Irwin & Barneby,  Wight & Arn., and  H.S. Irwin & Barneby, of which only the latter two have been placed in the  (sensu LPWG 2017) in phylogenetic analyses (Table 3).  and  together (along with ) comprise subfamily , and the monospecific  was raised to subfamily rank (LPWG 2017). Thus the grade of non-mimosoid caesalpinioid lineages that subtend tribe  in  are here recognised as distinct tribes.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B669\", \"B859\", \"B406\", \"B356\", \"B357\", \"B350\", \"B131\", \"B526\", \"B931\", \"B932\", \"B712\", \"F4\"], \"section\": \"\\ufeffTribe\", \"text\": \"Tribe  comprises six species in four genera. The four genera had previously been placed in distinct generic groups and tribes:  Wight ex Arn. in its own generic group of  by Polhill and Vidal (1981); the morphologically distinct, unisexual, and apetalous genus  (Tucker 1992) in subtribe  of  (Irwin and Barneby 1981);  Humbert and  Urb. in the  group of , although none of these placements were considered definitive. Phylogenetic analyses of morphological and plastid sequence data showed the four genera to form a clade and to be closely related to the trigeneric clade here treated as tribe , and together the two clades were placed in the informally named  clade (Herendeen et al. 2003a, 2003b; Haston et al. 2005; Bruneau et al. 2008), but subsequent combined plastid and nuclear sequence analyses did not support the monophyly of these two groups together (Manzanilla and Bruneau 2012; Zhang et al. 2020; Zhao et al. 2021). The phylogenomic analyses of Ringelberg et al. (2022) (Fig. 4) clearly indicate that each of these two clades is strongly supported as monophyletic but that they are not grouped together, supporting their recognition as distinct tribes.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B357\", \"B354\", \"B357\"], \"section\": \"\\ufeffTribe\", \"text\": \"These recent molecular analyses have highlighted several previously unsuspected morphological synapomorphies for , the most striking of which is a bipinnate leaf (although mostly once pinnate in ) terminating in a triad of pinnae arising from the same point at the apex of the rachis (Herendeen et al. 2003b; Herendeen and Herrera 2019). Tribe  has a highly disjunct and unusual geographic distribution occurring in Hispaniola (), Madagascar (), tropical (South-)East Asia (), and north-eastern Africa and the Mediterranean () (Herendeen et al. 2003b; Tribe , page 62).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B669\", \"B210\", \"B130\", \"B131\", \"B526\", \"B931\", \"B932\", \"B712\", \"F4\"], \"section\": \"\\ufeffTribe\", \"text\": \"The simplest of the informal generic groups recognised by Polhill and Vidal (1981) was the  group comprising two primarily north temperate genera,  J. Clayton and  Lam. The group is supported as monophyletic, and together with the South African  Sim, is here formally re-circumscribed as tribe .  was previously placed in tribe  by Cowan and Polhill (1981) but was later resolved as sister to  and  in phylogenetic analyses using plastid (Bruneau et al. 2001, 2008) and nuclear sequences (one locus, Manzanilla and Bruneau 2012), as well as in all recent phylogenomic analyses (Zhang et al. 2020; Zhao et al. 2021; Ringelberg et al. 2022; Fig. 4).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B356\"], \"section\": \"\\ufeffTribe\", \"text\": \"The 20 species of the three genera of tribe  occur in warm temperate regions, with several disjunctions between North and South America (), South Africa (), and North America and Asia (). Tribe  is characterised by several vegetative and floral synapomorphies, such as a tubular hypanthium and sepals with trichomes on the inner surface (Herendeen et al. 2003a; Tribe , page 70).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B712\", \"B130\", \"B131\", \"B349\", \"B350\", \"B528\", \"B526\", \"B931\", \"B932\", \"B222\"], \"section\": \"\\ufeffTribe\", \"text\": \" is here recognised as a new tribe comprising just the single species  Tul. Although not included in the phylogenomic analyses of Ringelberg et al. (2022), previous molecular phylogenetic analyses based on plastid and/or nuclear DNA sequence data always resolve this monospecific genus as a phylogenetically isolated lineage, on a long branch, poorly supported relative to  and , but generally in the clade that comprises all  except  and  (Bruneau et al. 2001, 2008; Haston et al. 2003, 2005; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; Zhang et al. 2020; Zhao et al. 2021).  thus appears to be a classic depauperon (Donoghue and Sanderson 2015), i.e., an old, species-poor lineage. The species is highly distinct morphologically (e.g., imparipinnate leaves with alternate leaflets and a well-formed rachis extension, small flowers in dense catkin-like racemes, the style laterally displaced at the apex of the ovary, fruits a one-seeded winged samara) and cytogenetically (2n = 20), sharing little in common with ,  or other  (Tribe , page 78).  is an important tree of South American tropical and subtropical dry forests.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B406\", \"B669\", \"T3\", \"B131\", \"B528\", \"B526\", \"B505\", \"B931\", \"B712\", \"F4\", \"B669\", \"B712\", \"B131\", \"B528\", \"B526\", \"B505\"], \"section\": \"\\ufeffTribe\", \"text\": \"In Advances in Legume Systematics Part 1, Irwin and Barneby (1981) recognised five subtribes in their tribe , including subtribe  comprising the genera , , and . These three genera alongside  Spruce ex Benth.,  Schott, and  Ducke, previously placed in the  group of  by Polhill and Vidal (1981) (Table 3), form a robustly supported clade in phylogenetic analyses (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; LPWG 2017; Zhang et al. 2020; Ringelberg et al. 2022) (Fig. 4), here recognised as tribe . The genus  [also placed in the  group by Polhill and Vidal (1981)], although not sampled by Ringelberg et al. (2022), is generally resolved as a member of this clade albeit with weak support (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; LPWG 2017) and is here placed in the tribe .\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B356\", \"B131\", \"B528\", \"B533\", \"B200\", \"B268\"], \"section\": \"\\ufeffTribe\", \"text\": \"Tribe  is the largest non-mimosoid clade (in terms of species richness) in subfamily , with 695 species, the vast majority of which are found in the genera  (361 species) and  (287 species) (Tribe , page 83). Although broadly distributed across the tropics, most of the genera and species are found in the New World. The clade is characterised by singly pinnate or bifoliolate leaves and, in several genera, stomata on both sides of the leaflets (Herendeen et al. 2003a; Bruneau et al. 2008). Several taxa in this clade, including most species of  and , as well as  and , are well-known for having notably prominent, conspicuous, abundant and unusual extrafloral nectaries (Marazzi and Sanderson 2010; Marazzi et al. 2019; Cota 2020a). Two  genera are known to nodulate,  and  (Faria et al. 2022). None of the genera in , , and  are known to nodulate.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B669\", \"B130\", \"B131\", \"B350\", \"B469\", \"B471\", \"B298\", \"B299\", \"B191\"], \"section\": \"\\ufeffTribe\", \"text\": \"The  group defined by Polhill and Vidal (1981) is similar to  recognised here, except that ,  Rose and  H. Perrier are now resolved in a separate clade (Bruneau et al. 2001, 2008; Haston et al. 2005), here treated as tribe , although  has since been synonymised under  Raf. (Bruneau and Babineau 2017). Long-standing uncertainty surrounding delimitation of the genus  L. and other genera in the  group (Lewis 1998; Lewis 2005b) has been resolved with the new generic system of Gagnon et al. (2015, 2016) and subsequent reinstatement of the genus  Adans. (Clark et al. 2022).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B712\", \"B299\"], \"section\": \"\\ufeffTribe\", \"text\": \"Tribe  comprises ca. 223 species in 27 genera. Although two of these genera,  and , were not sampled by Ringelberg et al. (2022), the analyses of Gagnon et al. (2016) clearly resolved  as sister to , and  in a clade unresolved with  and the lineage that combines  Adans.,  Tod., ,  R.Br. ex Wight & Arn. and .\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B299\", \"B300\"], \"section\": \"\\ufeffTribe\", \"text\": \"Species of  are highly diverse in growth forms, defence mechanisms, fruit morphologies, and pollination and seed dispersal syndromes (Gagnon et al. 2016). Although there are no clear morphological synapomorphies for the tribe, a diagnostic combination of characteristics is often found, including the presence of glandular trichomes, prickles or spines, bilaterally symmetrical flowers with a modified boat-shaped lower sepal, and free stamens crowded around the pistil (Tribe , page 103). The clade is pantropically distributed, with a marked affinity for the succulent biome (Gagnon et al. 2019).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B349\", \"B350\", \"B131\", \"B526\", \"B28\", \"B931\", \"B932\", \"B712\", \"B669\", \"B667\", \"B669\", \"B479\", \"B667\", \"B814\", \"B350\", \"B131\", \"B28\"], \"section\": \"\\ufeffTribe\", \"text\": \"Tribe  as here circumscribed matches the core  group (i.e.,  group s.s.) first recovered phylogenetically by Haston et al. (2003), and subsequently found in several other studies (Haston et al. 2005; Bruneau et al. 2008; Manzanilla and Bruneau 2012; Babineau and Bruneau 2017; Zhang et al. 2020; Zhao et al. 2021; Ringelberg et al. 2022). This clade differs from the informal  group recognised by Polhill and Vidal (1981; 13 genera) and Polhill (1994; 16 genera), by excluding four genera now placed in tribe  (, , , and ), three now in tribe  ( Schrad.,  Ducke, and  Rizzini & A. Mattos) and  Benth. (now in tribe ).  as circumscribed here also includes  and , two genera previously placed in the  group by Polhill and Vidal (1981), but which Lewis and Schrire (1995) had suggested might not be part of a more strictly defined  group, and which Polhill (1994) had included in the  group. In addition, the tribe includes  M. Sousa, described by Sousa (2005). Although  was resolved as part of this clade (Haston et al. 2005; Bruneau et al. 2008), Babineau and Bruneau (2017) found it to be nested within  and synonymised this monospecific genus under .\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"F4\", \"B711\", \"B350\"], \"section\": \"\\ufeffTribe\", \"text\": \"Tribe  contains ca. 42 species in eight genera. It has a pantropical distribution, and a considerable portion of the clade (i.e., the  \\u2013  subclade; Fig. 4) is strictly conserved within the succulent biome (Ringelberg et al. 2020).  is not defined by any morphological synapomorphies, but most species in the clade have bipinnate leaves, yellow petals, narrow seeds, characteristic spreading umbrella-like, flat-topped tree crowns, and smooth, thin and either pale silvery metallic grey or green bark (Haston et al. 2005; Tribe , page 146).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B669\", \"B667\", \"B669\", \"B667\", \"F4\", \"B131\", \"B528\", \"B526\", \"B505\", \"B931\", \"B131\", \"B528\", \"B526\", \"B712\", \"F4\", \"B931\", \"B932\", \"B712\"], \"section\": \"\\ufeffTribe\", \"text\": \"The generic composition of tribe  as treated here has not been recovered previously, although its constituent genera have often been associated with each other based on morphological and molecular data. The five genera of the  were placed in two generic groups of tribe  by Polhill and Vidal (1981) and Polhill (1994) based on morphology:  Tul. and  Aubl. (now including  Vogel, but earlier considered distinct from ) were placed in the  group, whereas , , and  were placed in the  group. Subsequent molecular phylogenetic studies generally also resolved the five genera in two separate clades, but with different generic composition from those of the informal groups of Polhill and Vidal (1981) and Polhill (1994). One strongly supported clade grouped ,  and , as found here (Fig. 4), and a separate less well supported clade (or grade) included  and  (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; LPWG 2017; Zhang et al. 2020), which has sometimes been resolved as part of a grade subtending the mimosoid clade (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012). In the recent phylogenomic analyses of Ringelberg et al. (2022; Fig. 4), the tribe is subtended by a short branch with notable gene tree conflict, whereas the ,  and  subclade is supported by a long branch. This short branch and gene tree conflict likely explain why the five genera have not been resolved as a clade, but rather as two separate clades in previous phylogenies. In addition, there is evidence to suggest that there may be cytonuclear discordance. In recent phylogenomic analyses, although not all genera have been sampled, plastid data strongly support the ,  and  subclade, with  and  forming a lineage subtending the mimosoid clade (Zhang et al. 2020), whereas nuclear sequence data group the two lineages as a clade (Zhao et al. 2021; Ringelberg et al. 2022).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B184\", \"B268\"], \"section\": \"\\ufeffTribe\", \"text\": \"As defined here, tribe  is restricted to the Neotropics, and comprises ca. 113 species, most of which are in the genus  (80\\u201390 species; Tribe , page 165). Several species of  are known to form close co-evolutionary associations with ants (Chomicki et al. 2015). Morphologically, each of the five genera is highly distinct in leaf morphology (pinnate or bipinnate leaves), floral symmetry (radial or bilateral), pollination syndrome (bees or birds), pollen presentation (monads or tetrads), fruit morphology, seed morphology (winged or non-winged), and dispersal syndrome (autochory, hydrochory, or anemochory). Thus, the tribe is not defined by obvious morphological synapomorphies, although there is a tendency for the occurrence of distinctively divided or foliaceous stipules (except for ; Tribe , page 165) and nodulation with a fixation thread type of nodule anatomy in ,  and , three of the nine non-mimosoid  genera known to nodulate (Sprent 2000; Faria et al. 2022).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B131\", \"B528\", \"B526\", \"B712\", \"B131\", \"B669\", \"B667\", \"B471\", \"B130\", \"B131\", \"B526\", \"B931\", \"B669\", \"B261\", \"B669\", \"B515\", \"B669\", \"B515\", \"B516\"], \"section\": \"\\ufeffTribe\", \"text\": \"The four genera of the , ,  Benth.,  Harms, and  Benth., form a clade in most molecular phylogenetic studies (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; Ringelberg et al. 2022), corresponding to the  group A of Bruneau et al. (2008). This is a narrower definition than the morphologically-based informal  group sensu Polhill and Vidal (1981), Polhill (1994) and Lewis (2005b), which also included  and  Harms (now tribe ),  and  (now placed in tribe ), as well as  and  (now in tribe ), a group subsequently shown to be non-monophyletic (Bruneau et al. 2001, 2008; Manzanilla and Bruneau 2012; Zhang et al. 2020). The  group sensu Polhill and Vidal (1981) comprised a diverse assemblage of genera, many of which share certain characteristics with tribe  (e.g., bipinnate leaves, numerous, small, regular flowers in spiciform racemes, and introrse sagittate anthers; Elias 1981a; Polhill and Vidal 1981; Luckow et al. 2000) and was considered a \\u2018\\u2018transitional link\\u2019\\u2019 between the then caesalpinioids and mimosoids (Polhill and Vidal 1981; Luckow et al. 2000, 2003). As newly circumscribed, tribe  is morphologically more coherent, including four genera, all with spicate inflorescences and pentamerous, diplostemonous flowers.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B505\", \"B712\", \"B268\"], \"section\": \"\\ufeffTribe\", \"text\": \"The four genera of the  contain 35 species. However, 26 are in  s.l., which is paraphyletic (e.g., LPWG 2017; Ringelberg et al. 2022; see Tribe , page 177), suggesting that generic re-delimitation will be necessary. The clade has an amphi-Atlantic distribution (Neotropics and tropical Africa) spanning a variety of biomes. Nodulation is reported in two species of , whereas  and  are confirmed as non-nodulators (Faria et al. 2022).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"S3\", \"B712\", \"F3\"], \"section\": \"\\ufeffTribe\", \"text\": \"It is also notable that while tribes ,  and  are each supported as monophyletic in recent phylogenomic analyses, the relationships among these three lineages are weakly supported and characterised by high gene tree conflict (Suppl. material 3; Ringelberg et al. 2022: Fig. 3).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"T3\", \"F4\", \"B931\", \"B131\", \"B528\", \"B526\", \"B475\", \"B511\", \"B515\", \"B516\", \"B919\", \"B526\", \"B931\", \"B712\", \"F4\", \"B712\", \"B931\"], \"section\": \"\\ufeffTribe\", \"text\": \" and , previously placed in  and the  group of  respectively (Table 3), have only been recovered as sister genera (Fig. 4) in one previous study (Zhang et al. 2020), although the two genera have generally been resolved in the same large clade that included  and subtending lineages (Bruneau et al. 2008; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012). Morphologically,  was previously considered to be a member of subfamily  (Lewis and Elias 1981; Pohill 1994; Luckow 2005), but molecular phylogenetic studies have consistently placed the genus among the grade of non-mimosoid  genera subtending the mimosoid clade, albeit with varying sister group relationships (Luckow et al. 2000, 2003; Wojciechowski et al. 2004; Manzanilla and Bruneau 2012; Zhang et al. 2020; Ringelberg et al. 2022; Fig. 4). These two genera exhibit disparate morphology, although they share perigynous flowers with a tubular hypanthium and showy stamens exserted from the corolla. This morphological distinctiveness is mirrored in the molecular analyses, where relatively long branches subtend the two genera. Although we here place these two morphologically disparate genera together in tribe , Ringelberg et al. (2022) noted that long-branch attraction could play a role in grouping  and  together in a clade. A similar phylogenetic pattern is observed in the plastid phylogenomic analyses of Zhang et al. (2020), in which the two genera also form a clade subtended by a short branch.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B268\"], \"section\": \"\\ufeffTribe\", \"text\": \"Tribe  comprises 5 to 22 species (the genus  needs to be revised because several species are of dubious taxonomic status), only two of which are in . The tribe is restricted to tropical rainforests in South America. Nodulation is reported in one species each of  and  (Faria et al. 2022), and is known to be absent in the other species of .\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B356\", \"B130\", \"B131\", \"B516\", \"B97\", \"B528\", \"B526\", \"B449\", \"B931\", \"B712\"], \"section\": \"\\ufeffTribe\", \"text\": \" and  have rarely been recovered as sister genera before (Herendeen et al. 2003a). Nevertheless, most previous studies based primarily on plastid sequence data placed these two genera as successive sisters to the mimosoid clade (Bruneau et al. 2001, 2008; Luckow et al. 2003; Bouchenak-Khelladi et al. 2010; Marazzi and Sanderson 2010; Manzanilla and Bruneau 2012; Kyalangalilwa et al. 2013; Zhang et al. 2020), as found in the plastid phylogeny of Ringelberg et al. (2022), indicating another case of possible cytonuclear discordance and potentially explaining why the two genera have not previously been grouped. No clear morphological synapomorphy has been identified for the clade, but  and  share a combination of morphological traits only rarely found in non-Mimoseae  (e.g., bipinnate leaves, small pedicellate perigynous flowers in dense spicate racemes), and both genera have highly toxic alkaloids and saponins (Tribe , page 193).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B261\", \"B516\", \"B517\", \"B504\", \"B505\", \"B441\", \"B712\", \"B441\"], \"section\": \"\\ufeffTribe\", \"text\": \"Recognising the mimosoid clade as the newly circumscribed tribe  results in by far the largest tribe in subfamily  in terms of numbers of species (ca. 3500) and genera (100). Previous tribal classifications of the mimosoid clade (i.e., former subfamily ) recognised five tribes: ,  Benth., Ingeae Benth.,  (Wight & Arn.) Benth., and Mimozygantheae Burkart (Elias 1981a). In formulating the new tribal classification for subfamily , the recognition of the mimosoid clade as the reinstated and re-circumscribed tribe  is proposed for two main reasons. First, four of the five former tribes of  are now known to be non-monophyletic and the fifth (the monospecific Mimozygantheae) to be nested within the former  (Luckow et al. 2003, 2005; LPWG 2013, 2017; Koenen et al. 2020a; Ringelberg et al. 2022). Second, the ladder-like phylogenetic structure within the mimosoid clade means that any finer-scale tribal divisions would inevitably result in an undesirable proliferation of many small Linnean tribes, including a large number of monogeneric tribes (Koenen et al. 2020a).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B515\", \"B516\", \"B483\", \"B526\", \"B70\", \"B396\", \"B668\", \"B483\", \"B441\", \"B712\", \"B669\", \"B667\", \"B471\", \"F5\", \"B712\", \"B471\", \"B526\", \"B505\", \"B441\", \"B712\", \"F5\", \"B669\", \"B667\", \"B511\"], \"section\": \"\\ufeffTribe\", \"text\": \"There has been ongoing debate about which genera are included in the mimosoid clade (Luckow et al. 2000, 2003; Lewis et al. 2005; Manzanilla and Bruneau 2012). As circumscribed here, tribe  matches the former subfamily  (Bentham 1865; Hutchinson 1964; Polhill and Raven 1981; Lewis et al. 2005), with three exceptions (Koenen et al. 2020a; Ringelberg et al. 2022; Tribe , page 201). First, , previously considered to be a non-mimosoid caesalpinioid based on morphology (Polhill and Vidal 1981; Polhill 1994; Lewis 2005b), is firmly nested within the  as an isolated early-diverging lineage (Fig. 5; Ringelberg et al. 2022). Similarly, , once considered part of  (Lewis 2005b), is also nested within the , as initially found by Manzanilla and Bruneau (2012) and supported by LPWG (2017), Koenen et al. (2020a), and Ringelberg et al. (2022) (Fig. 5). Finally,  was previously considered a genus of  (Polhill and Vidal 1981; Polhill 1994; Luckow 2005), but is resolved in the non-mimosoid  (now in tribe , page 187).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B268\"], \"section\": \"\\ufeffTribe\", \"text\": \"Tribe  is diagnosed by valvate petal aestivation (with exceptions in  and ), bipinnate leaves (except  and a few scattered species in other genera), flowers that are relatively small with reduced perianth and showy androecium, mostly clustered in compact inflorescences that are commonly capitate or spicate (Tribe , page 201), and the presence of symbiosome-type (as opposed to fixation-thread-type) root nodules (Faria et al. 2022).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"F4\", \"F5\", \"B441\", \"B712\", \"F5\", \"F5\"], \"section\": \"\\ufeffTribe\", \"text\": \"Tribe , as circumscribed here, is robustly supported as monophyletic and is subtended by a relatively long branch (Fig. 4). Within  the phylogeny takes the form of an extensive unbalanced ladder-like topology (Fig. 5), which is not readily amenable to division into a manageable number of rank-based Linnean taxa. However, given the large size of the tribe, some form of classificatory structure is needed, and here we present a clade-based classification system for the tribe with two nested higher-level named clades \\u2013 a core mimosoid clade and the ingoid clade \\u2013 alongside a set of 17 named lower-level clades following Koenen et al. (2020a) and Ringelberg et al. (2022) (Fig. 5). Fourteen of the 100 genera of  remain unplaced in any of these lower-level clades, eight of which are resolved in two grades, and six of which are phylogenetically isolated monogeneric lineages (Fig. 5).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B441\", \"F5\", \"B441\"], \"section\": \"\\ufeffTribe\", \"text\": \"The core mimosoid clade as delimited by Koenen et al. (2020a) is well supported, subtended by a notably long branch, and includes all  except the  and  clades, the two monogeneric lineages  and , and the  grade (Fig. 5). The core mimosoid clade includes all  with armature, with the exceptions of a single spinescent species of  and of two species of  Harms, which are armed with modified woody tendrils, and which occur outside the clade (Koenen et al. 2020a).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B441\", \"F5\", \"B441\", \"B241\", \"B906\"], \"section\": \"\\ufeffTribe\", \"text\": \"The ingoid clade, delimited by Koenen et al. (2020a), is also strongly supported, includes ca. 2000 species, i.e., almost two-thirds of  species, and comprises genera of the  grade and nine named clades (, , , , , , , , and  clades) (Fig. 5). The ingoid clade groups all genera of  with polystemonous flowers except . A synandrous androecium is exclusively found in this clade and characterises most, but not all of its genera (Koenen et al. 2020a). Based on the limited sample of mimosoid chloroplast genome sequences currently available, an expanded Inverted Repeat region of the chloroplast genome is also restricted to this clade (Dugas et al. 2015; Wang et al. 2017).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"F4\", \"F5\", \"B441\", \"B712\", \"B505\"], \"section\": \"\\ufeffTaxonomy\", \"text\": \"Based on the phylogeny of subfamily  presented here, we recognise eleven tribes (Fig. 4) and 17 formally named clades within tribe  (Fig. 5). The new classification proposed here recognises clades that are strongly supported in the phylogenomic analyses of Koenen et al. (2020a) and Ringelberg et al. (2022), many of which were already known from earlier phylogenetic studies. This classification is proposed and endorsed by the legume systematics community as reflected in the use of the Legume Phylogeny Working Group (LPWG) as the authority of the new tribes. Although an uncommon practice in botanical nomenclature, ascribing the new tribe names to the collective known as the \\u201cLegume Phylogeny Working Group\\u201d is accepted under the botanical code as stipulated in Chapter VI, Section 1 (Author Citations) and follows the approach previously used for the LPWG (2017) subfamily classification. The Legume Phylogeny Working Group authorship gives due credit to the legume systematics community and reflects the important collaborative contributions from multiple research groups over the decades that have laid the foundations for this classification.\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B713\", \"S1\", \"B712\", \"B713\"], \"section\": \"\\ufeffTaxonomy\", \"text\": \"We provide a key to genera, as well as taxonomic descriptions and notes for tribes, named clades, and all 163 genera, and illustrate the diversity of growth forms, foliage, flowers and fruits for nearly all genera. We also provide a distribution map of the native range for each genus, based on quality-controlled herbarium specimen localities and floristic surveys. The occurrence data for  are from Ringelberg et al. (2023), whereas the remainder were newly assembled here. See Suppl. material 1 for sources of occurrence data and detailed data cleaning protocols. All occurrence data have been made available on Zenodo (https://zenodo.org/doi/10.5281/zenodo.8407862), unless stated otherwise. A tree file of the  phylogeny presented here is available from Ringelberg et al. (2022, 2023).\"}, {\"pmc\": \"PMC11188994\", \"pmid\": \"\", \"reference_ids\": [\"B339\"], \"section\": \"Description.\", \"text\": \"Trees, shrubs, lianas, suffruticose or functionally herbaceous, occasionally aquatic, either unarmed or commonly armed with prickles, spines, or thorns; specialised extrafloral nectaries often present on the petiole and/or on the primary and secondary leaf rachides, usually between pinnae or leaflet pairs, more rarely stipular or bracteal. Stipules in lateral position and free or absent, usually entire, less frequently div

Metadata

"{}"