Subfamily Caesalpinioideae with ca. 4,600 species in 152 genera is the second-largest subfamily of legumes (Leguminosae) and forms an ecologically and economically important group of trees, shrubs and lianas with a pantropical distribution. Despite major advances in the last few decades towards aligning genera with clades across Caesalpinioideae, generic delimitation remains in a state of considerable flux, especially across the mimosoid clade. We test the monophyly of genera across Caesalpinioideae via phylogenomic analysis of 997 nuclear genes sequenced via targeted enrichment (Hybseq) for 420 species and 147 of the 152 genera currently recognised in the subfamily. We show that 22 genera are non-monophyletic or nested in other genera and that non-monophyly is concentrated in the mimosoid clade where ca. 25% of the 90 genera are found to be non-monophyletic. We suggest two main reasons for this pervasive generic non-monophyly: (i) extensive morphological homoplasy that we document here for a handful of important traits and, particularly, the repeated evolution of distinctive fruit types that were historically emphasised in delimiting genera and (ii) this is an artefact of the lack of pantropical taxonomic syntheses and sampling in previous phylogenies and the consequent failure to identify clades that span the Old World and New World or conversely amphi-Atlantic genera that are non-monophyletic, both of which are critical for delimiting genera across this large pantropical clade. Finally, we discuss taxon delimitation in the phylogenomic era and especially how assessing patterns of gene tree conflict can provide additional insights into generic delimitation. This new phylogenomic framework provides the foundations for a series of papers reclassifying genera that are presented here in Advances in Legume Systematics (ALS) 14 Part 1, for establishing a new higher-level phylogenetic tribal and clade-based classification of Caesalpinioideae that is the focus of ALS14 Part 2 and for downstream analyses of evolutionary diversification and biogeography of this important group of legumes which are presented elsewhere.

Fabaceae, generic delimitation, mimosoid clade, monophyly, morphological homoplasy, phylogenomics

In 2017, the Legume Phylogeny Working Group established a new subfamily classification of the Leguminosae (LPWG 2017), which dealt with the longstanding problem of the paraphyly of old sense subfamily Caesalpinioideae DC. by formally dividing the family into six subfamilies: Cercidoideae LPWG, Detarioideae Burmeist., Duparquetioideae LPWG, Dialioideae LPWG, Caesalpinioideae and Papilionoideae DC. Subfamily Caesalpinioideae was especially impacted by this new classification because several large clades previously included within it were afforded subfamily rank, while at the same time the former subfamily Mimosoideae DC., which is nested within Caesalpinioideae, was subsumed within the re-circumscribed Caesalpinioideae and is now simply referred to as the mimosoid clade (LPWG 2017). The idea that Leguminosae comprises six main lineages has since been amply confirmed by phylogenomic analyses of large nuclear gene and plastome DNA sequence datasets (Koenen et al. 2020a; Zhang et al. 2020; Zhao et al. 2021) providing robust support for the six subfamilies. Establishment of this new classification has shifted the focus of current legume systematics research to development of phylogenetically-based tribal (e.g. de la Estrella et al. 2018 for Detarioideae) and clade-based (e.g. Sinou et al. 2020 for Cercidoideae) higher-level classifications and, especially, towards establishment of robust generic systems for each subfamily. Here, we present a phylogenomic backbone for the re-circumscribed subfamily Caesalpinioideae as the basis for a new higher-level and generic classification of that subfamily.

Caesalpinioideae sensu LPWG (2017) is the second largest subfamily of legumes with ca. 4,600 species currently placed in 152 genera (LPWG 2017 plus additions, see below). Within this subfamily, ca. 3,400 species and 90 genera are placed in the mimosoid clade corresponding to the former subfamily Mimosoideae, which is nested within new sense Caesalpinioideae (LPWG 2017). Caesalpinioideae has a pantropical distribution and many of its lineages form ecologically abundant or dominant elements across each of the major lowland tropical biomes – seasonally dry tropical forests (“the succulent biome” sensu Schrire et al. 2005 and Ringelberg et al. 2020), savannas and tropical rain forests – thus spanning the full lowland tropical rainfall spectrum from arid to hyper-wet, with just a small fraction of species extending into the warm temperate zone, a subset of which are frost tolerant. Caesalpinioideae species are infrequent above 2500 m elevation in the tropics and are notably absent from mid- and high-elevation tropical montane forests, with only a few exceptions (e.g. some Inga Mill. spp., Paraserianthes lophantha (Vent.) I.C. Nielsen subsp. montana (Jungh.) I.C. Nielsen). The ecological versatility of the subfamily across the lowland tropical moisture availability spectrum is matched by its great diversity of life-history strategies, from massive canopy-emergent rainforest trees to small desert shrubs, and functionally-herbaceous savanna geoxyles to woody lianas and aquatic plants (Lewis et al. 2005; LPWG 2013, 2017; Koenen et al. 2020b; Ringelberg et al. 2022). Many species are economically important because of their highly-nutritious fruits, valuable wood, nitrogen-rich leaves and other products (Lewis et al. 2005) and are especially prominent as multipurpose trees in tropical silvo-pastoral and other agroforestry systems. Several other species constitute some of the world’s most serious invasive weeds (e.g. Leucaena leucocephala (Lam.) de Wit, several Mimosa L. spp. and Acacia Mill. spp., Prosopis juliflora (Sw.) DC.). Generic diversity is highest in the Neotropics and Africa and there are important centres of species diversity in Mexico and Central America, lowland South America, Africa, Madagascar, parts of S.E. Asia and Australia. Caesalpinioideae includes some of the largest genera in the legume family, such as Acacia with > 1,000 species concentrated in dry parts of Australia and Mimosa with > 500 species mostly in the Neotropics, as well as Chamaecrista Moench and Senna Mill., each with 300+ species distributed pantropically, Inga Mill. with ca. 300 species restricted to the Neotropics, almost entirely in rainforests and Vachellia Wight & Arn. (ca. 160 species) and Senegalia Raf. (ca. 220 species), two pantropical genera concentrated in drier environments, within which the iconic umbrella-crown trees of African savannas are found.

Numbers of genera across Caesalpinioideae have increased progressively through the last 270 years, but are difficult to track, because of the altered delimitation of the subfamily. However, the history of generic delimitation in mimosoids illustrates the overall trajectory of numbers of genera. Von Linnaeus (1753) placed all known mimosoids in a single genus Mimosa, which was later subdivided by Willdenow (1805) into five genera: Inga, Mimosa, Schrankia Willd., Desmanthus Willd. and Acacia. In 1825, de Candolle added five more genera, but the real foundations for all subsequent work were established by Bentham (1842, 1875) notably in his ‘Revision of suborder Mimoseae’ in 1875, which recognised six tribes and 46 genera, based on examination of 1,200 species known at that time.

The legacy of Bentham’s generic system has been long-lasting. At the heart of Bentham’s system were a set of large, geographically widespread genera, including Acacia, Calliandra Benth., Pithecellobium Mart. and Prosopis L., all of which, with the advent of molecular phylogenetics, have been shown to be non-monophyletic. The disintegration of Acacia into (currently) seven segregate genera (Acacia, Acaciella Britton & Rose, Mariosousa Seigler & Ebinger, Parasenegalia Seigler & Ebinger, Pseudosenegalia Seigler & Ebinger, Senegalia and Vachellia), based on 20 years of molecular phylogenetic studies (Clarke et al. 2000; Miller and Bayer 2000, 2001, 2003; Robinson and Harris 2000; Luckow et al. 2003; Miller et al. 2003, 2013, 2017; Murphy et al. 2003; Seigler et al. 2006a, b; Brown et al. 2008; Bouchenak-Khelladi et al. 2010; Gómez-Acevedo et al. 2010; Miller and Seigler 2012; Kyalangalilwa et al. 2013; Mishler et al. 2014; Boatwright et al. 2015; Terra et al. 2017; Koenen et al. 2020b) (Figs 1 and 6–8) has been the most prominent example in legumes of the dissolution of one of Bentham’s broadly circumscribed pantropical genera. Pithecellobium and Calliandra have suffered similar fates (Barneby and Grimes 1996, 1997; Barneby 1998; de Souza et al. 2013, 2016). In contrast, although Bentham (1875) had restricted his concept of the genus Albizia Durazz. to just Old World species, Nielsen (1981) expanded the genus pantropically, creating the last big ‘dustbin genus’ of mimosoids (Koenen et al. 2020b). By far the most persistent generic delimitation problems surround those of former tribe Ingeae, where starkly contrasting generic systems and numerous generic transfers have caused much on-going confusion (reviewed by Brown 2008).

Figure 1. 

Phylogeny of Caesalpinioideae with clade names as inferred by Koenen et al. (2020b), the starting point for this study.

By 1981, the number of mimosoid genera had risen to 62 in Advances in Legume Systematics Part 1 (Elias 1981), 78 in Legumes of the World (Lewis et al. 2005) and in the most recent census (LPWG 2017) to 84, with 148 genera recognised in Caesalpinioideae as a whole.

Across the non-mimosoid Caesalpinioideae generic delimitation has also seen many changes. The most complex problems have been, without doubt, in the Caesalpinia Group and, especially, the genus Caesalpinia L. s.l. (Polhill and Vidal 1981; Lewis 1998; Gagnon et al. 2016), but these have now largely been resolved with the phylogenetically-based generic system of Gagnon et al. (2016), which recognised 26 genera, leaving just one residual generic problem in that group (see Clark et al. 2022).

Since LPWG (2017), two genera of Caesalpinioideae have been synonymised (i.e. Cathormion Hassk. within Albizia (Koenen et al. 2020b) and Lemuropisum H. Perrier within Delonix Raf. (Babineau and Bruneau 2017)) and six new genera have been segregated or resurrected (i.e. Lachesiodendron P.G. Ribeiro, L.P. Queiroz & Luckow (Ribeiro et al. 2018), Parasenegalia and Pseudosenegalia (Seigler et al. 2017), Jupunba Britton & Rose and Punjuba Britton & Rose (Soares et al. 2021) and Robrichia (Barneby & J.W. Grimes) A.R.M. Luz & E.R. Souza (de Souza et al. 2022a)), bringing the current tally of Caesalpinioideae genera to 152, of which 90 are mimosoids.

Despite this rapid on-going progress to align genera with clades in recent years, generic delimitation across Caesalpinioideae and, especially, the mimosoid clade, remains in a state of considerable flux and there is evidence to suggest that several more genera are non-monophyletic: Prosopis (Catalano et al. 2008), Dichrostachys (DC.) Wight & Arn. (Hughes et al. 2003; Luckow et al. 2005), Balizia Barneby & J.W. Grimes (Iganci et al. 2016; Koenen et al. 2020b), Zygia P. Browne (Ferm et al. 2019), Entada Adans. (Luckow et al. 2003), Caesalpinia (Gagnon et al. 2016), Albizia, Senegalia and Leucochloron Barneby & J.W. Grimes (Koenen et al. 2020b; Fig. 1). One factor that has undoubtedly contributed significantly to this widespread generic non-monophyly is the potentially pervasive homoplasy of multiple morphological characters previously used for generic delimitation, as well as reliance on only a few characters for delimiting taxa. This has led to tribes defined solely on stamen number and fusion into a staminal tube (Bentham 1875) and ‘fruit genera’, such as Calliandra, which was defined by Bentham (1875), based on its characteristic elastically dehiscent fruit. All mimosoid tribes and the genus Calliandra have since been shown to be non-monophyletic and their defining characters shown to have evolved multiple times across the subfamily (e.g. LPWG 2013; Barneby 1998). Such over-reliance on a small number of potentially homoplasious morphological characters, such as fruit type, connation and number of stamens and floral heteromorphy have likely repeatedly misled classification and resulted in widespread generic non-monophyly.

Another issue has been delimitation of the mimosoid clade with on-going uncertainties surrounding the inclusion or not of certain genera (Luckow et al. 2000, 2003; Manzanilla and Bruneau 2012). Although lacking valvate petals in bud (the putative synapomorphy of mimosoids), morphologically some members of the informal Dimorphandra group of Polhill and Vidal (1981) and Polhill (1994) show many similarities to mimosoids, with small, often numerous, regular flowers arranged in spikes or spiciform racemes, the hypanthium contracted, the anthers sagittate and introrse, the stamens becoming the most conspicuous and attractive part of the flower and pollen in tetrads in a few genera (Diptychandra Tul. and Dinizia Ducke) with possible affinities to the polyads that characterise many mimosoid lineages (Banks et al. 2010). These mimosoid-like features have prompted inclusion of some genera such as Dinizia in the mimosoid clade in the past (e.g. Burkart 1943; Luckow et al. 2000). Although none of these mimosoid-like genera has flowers with petals valvate in bud, previous molecular phylogenetic analyses have unexpectedly placed two Dimorphandra group genera in the mimosoid clade: Chidlowia Hoyle and Sympetalandra Stapf. The monospecific west African genus Chidlowia was placed with high support within the mimosoid clade in analyses based on few genetic markers (Manzanilla and Bruneau 2012; LPWG 2017), a result which was confirmed by the phylogenomic analyses of Koenen et al. (2020b; Fig. 1). The small Asian genus Sympetalandra was also recovered in the mimosoid clade in the matK tree of LPWG (2017), but was not sampled by Koenen et al. (2020b). Although support for the mimosoid clade is robust and the branch subtending that clade is long (Koenen et al. 2020b; Fig. 1), such that the monophyly of mimosoids is not in doubt, not all Caesalpinioideae genera have been included in phylogenomic analyses. By sampling widely and densely across Caesalpinioideae as a whole, we aim to further resolve which genera are placed in the mimosoid clade.

Several other issues have hindered a more complete understanding of the phylogeny and tribal / generic classification of subfamily Caesalpinioideae. First, the legacy of the traditional subfamily classification meant that taxon sampling in previous phylogenetic studies focused primarily on either old sense Caesalpinioideae (i.e. the grade subtending mimosoids (the ‘Caesalpinieae grade’ of Manzanilla and Bruneau 2012) of new sense Caesalpinioideae (Bruneau et al. 2008; Manzanilla and Bruneau 2012)), or on the mimosoid clade (e.g. Luckow et al. 2003, 2005; Koenen et al. 2020b). Few studies, apart from the family-wide analysis of plastid matK sequences (LPWG 2017), have sampled densely and widely across Caesalpinioideae as a whole. Second, several parts of the Caesalpinioideae phylogeny have been recalcitrant to phylogenetic resolution using traditional DNA sequence loci, most notably along the backbone of the grade subtending the mimosoid clade (Bruneau et al. 2008; Manzanilla and Bruneau 2012; LPWG 2017) and across the large ingoid clade sensu Koenen et al. (2020b). Third, lack of dense pantropical sampling of taxa in previous phylogenies means that the monophyly of several key genera with wide pantropical distributions, such as the ‘dustbin genus’ Albizia, has not been adequately tested and that possible sister-group relationships between New and Old World groups that are relevant to delimitation of genera may have been missed.

More robust foundations to overcome these difficulties were established by Koenen et al. (2020b) in a phylogenomic study of the mimosoid clade. By developing a clade-specific bait set (Mimobaits) for targeted enrichment of 964 nuclear genes, Koenen et al. (2020b) opened the way for generating DNA sequence datasets orders of magnitude larger than those used previously, thereby providing much enhanced phylogenetic resolution. Using these new data, Koenen et al. (2020b) established a new phylogenomic framework and recognised three large informally named higher-level clades each successively nested within Caesalpinioideae (Fig. 1). The mimosoid clade, core mimosoid clade and ingoid clade were all strongly supported by high proportions of gene trees and subtended by long branches. In addition, a set of 15 smaller informally named subclades across mimosoids were proposed by Koenen et al. (2020b) (Fig. 1) to replace the previously defined tribes and informal groups and alliances, almost all of which have been shown by numerous studies to be non-monophyletic (Luckow et al. 2003; LPWG 2013, 2017; Koenen et al. 2020b). Furthermore, although the Mimobaits bait set was designed based on RNA-seq data from species of four mimosoid genera and used initially for the mimosoid clade, the results of Koenen et al. (2020b) suggested that they work well across the non-mimosoid Caesalpinioideae, opening the way to potentially sequence these genes across the subfamily as a whole. The Koenen et al. (2020b) study also further revealed or confirmed the non-monophyly of several genera, but it lacked sufficient taxon sampling to fully test generic monophyly and sampling was largely restricted to the mimosoid clade. Here, we capitalise on these foundations using a slightly modified version of the Mimobaits gene set covering 997 nuclear genes to extend taxon sampling to 420 species from 147 of the 152 genera and establish a robust phylogenomic hypothesis for subfamily Caesalpinioideae as a whole.

This new phylogeny provides the basis for testing the monophyly of genera (the main focus of this paper and of this Special Issue Advances in Legume Systematics (ALS) 14, Part 1), establishing a new higher-level classification of the subfamily (the focus of ALS 14, Part 2) and for downstream analyses of biogeography, trait evolution and diversification (de Faria et al. 2022; Ringelberg et al. 2022). Caesalpinioideae provides an excellent clade for investigating evolutionary diversification and phylogenetic turnover across the lowland tropics (Lavin et al. 2004; Gagnon et al. 2019; Ringelberg et al. 2020, 2022), as well as the evolution of several prominent plant functional traits including compound leaves, armature, extrafloral nectaries and ant associations (Marazzi et al. 2019), agglomeration of pollen into polyads, plant growth forms (Gagnon et al. 2019), floral morphology and pollination syndromes, fruit morphology and seed dispersal syndromes and the ability to form nitrogen-fixing root nodule symbiosis (Sprent et al. 2017; de Faria et al. 2022). However, all of these opportunities require a robust and well-sampled subfamily-wide phylogeny of Caesalpinioideae. In turn, some of these traits have been used for generic delimitation in the past and, in this paper, we also evaluate a handful of such traits in a preliminary way by mapping them on to the phylogeny.

The authors thank B. Adhikari, D. Lorence, B. Marazzi, É. de Souza, the G, K, JRAU, L, MEL, NY, FHO, P, RB, WAG and Z Herbaria, the Millennium Seed Bank, Kew, the South African National Botanical Institute, the Direction de Environment, New Caledonia and the National Tropical Botanical Garden, Hawaii, U.S.A. for provision of leaf or DNA samples; P. Ribeiro, É. de Souza, M. Morim, M. Simon and F. Bonadeu in Brazil and staff of the Kew Madagascar Conservation Centre and Botanical and Zoological Garden of Tsimbazaza, Madagascar for fieldwork support; the national regulatory authorities of Brazil for access to DNA of Brazilian plants (authorised through SISGEN n° A464716), and Madagascar (research permit 006/14/MEF/SG/DGF/DCB.SAP/SCB); U. Grossniklaus, V. Gagliardini, Dept Plant & Microbial Biology, Univ. Zurich for use of their TapeStation; the S3IT, Univ. Zurich for use of the ScienceCloud computational infrastructure; and the Club Botanique de Toliara, Madagascar, the Flora Mendocina project, Argentina, B. Maslin, G. Sankowsky, M. Simon, and X. Cornejo for permission to use photos included in Figs 13–15. This work was supported by the Swiss National Science Foundation through grants 310003A_156140 and 31003A_182453/1 to CEH and an Early.Postdoc.Mobility fellowship P2ZHP3_199693 to EJMK, the Claraz Schenkung Foundation, Switzerland to CEH, the Natural Sciences and Engineering Research Council of Canada, NSERC, to AB, FAPESB, Brazil (PTX0004 & APP0096 to LPdQ) and CNPq, Brazil (480530/2012-2 & 311847/2021-8 to JRI and PROTAX 440487 to LPdQ).

Generic non-monophyly in Caesalpinioideae – towards a new generic system for the subfamily

Caesalpinia

Divergent circumscriptions of the genus Caesalpinia L. were largely resolved by Gagnon et al. (2016) who reduced Caesalpinia to ca. nine species and established a new generic system for the Caesalpinia Group as a whole, with 26 genera plus their ‘Ticanto clade’ (Caesalpinia crista L. and allies) as a putative 27^th genus. This 27^th genus accounts for the non-monophyly of Caesalpinia in our analysis (Fig. 2) with Caesalpinia crista representing the Ticanto clade that is re-instated as a genus in this Special Issue by Clark et al. (2022).

Dimorphandra

In line with previous studies (Luckow et al. 2005; LPWG 2017), Dimorphandra Schott is non-monophyletic in the nuclear phylogeny (Fig. 3), but robustly supported (99% bootstrap support (BS)) as monophyletic in the plastid tree (Suppl. material 3), indicating cytonuclear discordance. This implies either splitting Dimorphandra into two genera or sinking Mora Schomb. ex Benth., Stachyothyrsus Harms and Burkea Benth. into Dimorphandra (which predates these other three genera). Evidence suggests splitting Dimorphandra as the preferred option. First, the three Dimorphandra species sampled here represent the three morphologically delimited subgenera (da Silva 1986) with representatives of these subgenera intermingled with other genera rendering Dimorphandra polyphyletic in the legume-wide matK phylogeny (LPWG 2017) and Burkea and Mora are not closely related to Dimorphandra in the plastid phylogeny (Suppl. material 3; Stachyothyrsus is not included in the plastid analysis). Second, while Mora has been included in Dimorphandra based on morphological similarities (Sandwith 1932; van Steenis 1975), the two genera differ in floral, seed and pod morphology and have generally been treated as distinct (Sandwith 1932; van Steenis 1975; da Silva 1986). African Stachyothyrsus and Burkea are morphologically (van Steenis 1975) and geographically distinct from South American Dimorphandra and Mora. All of this suggests that Dimorphandra will need to be split into two genera or potentially three, although the robustly supported sister group relationship between D. davisii and D. macrostachya (internode certainty 0.77, subtended by a long branch) would perhaps favour two genera, rather than three. Additional taxon sampling, to test the monophyly of the three subgenera, is required before taxonomic re-arrangements can be made. If the genus is to be split, the name Dimorphandra would remain attached to subgenus Dimorphandra, here represented by D. gardneriana Tul. Dimorphandra exaltata Schott is the type species of the genus. The names of the other two subgenera, Phaneropsia Tulasne and Pocillum Tulasne, would be available for the remaining species. Both names originate from the same publication (Tulasne 1844), but since Pocillum also refers to a genus of fungi (Kirk et al. 2008), Phaneropsia would be the more suitable generic name for the species not in Dimorphandra s.s. However, as taxon names have no priority at different rank (Turland et al. 2018), a new generic name may also be proposed.

Xylia and Calpocalyx

The non-monophyly of Xylia with Calpocalyx nested within it was documented using matK sequences (LPWG 2017) and is confirmed here (Fig. 4). This does not come as a great surprise, as these genera have always been considered closely related (Villiers 1984; Lewis et al. 2005). They have overlapping geographical and ecological distributions mainly in the tropical rainforests of central and western Africa (although Xylia has a wider distribution in Africa, Madagascar and Asia). The two genera also share a suite of morphological characteristics (Villiers 1984; Luckow et al. 2003), including robust woody sickle-shaped explosively dehiscent fruits (Fig. 13), a chromosome count of 2n = 12 (Goldblatt and Davidse 1977) and pollen grains in small-sized polyads (Jumah 1991). Since the name Xylia (Bentham 1841) predates Calpocalyx (Engler and Prantl 1897) and given the morphological and ecological similarities of the two genera, the most straightforward solution to the non-monophyly presented here would be the transfer of the species of Calpocalyx to Xylia. However, this apparently straightforward incorporation of Calpocalyx into Xylia is complicated by the name Esclerona Raf., an apparently valid name predating Xylia, raising the possibility of proposing conservation of the name Xylia prior to merging these two genera.

Entada and Elephantorrhiza

A close relationship between Entada Adans. and Elephantorrhiza Benth. has long been suggested in all molecular phylogenies that sampled these genera (e.g. Luckow et al. 2003; Koenen et al. 2020b). With denser sampling of species, it has become clear that Elephantorrhiza is nested within Entada (LPWG 2017), a result that is confirmed here (Fig. 4) and which provides the basis for re-circumscription of Entada to include Elephantorrhiza by O’Donnell et al. (2022) in this Special Issue.

Prosopis

One of the most striking and robustly supported examples of generic non-monophyly in our analyses is Prosopis s.l. whose species are placed in four separate lineages (Figs 4 and 5). The nodes supporting this non-monophyly are some of the most robustly supported across the Caesalpinioideae phylogeny as a whole (Fig. 5). This shows that P. africana is not closely related to the rest of Prosopis s.l., but is placed in a grade with other monospecific or species-poor genera subtending the core mimosoid clade (Fig. 4), confirming results from earlier studies (Catalano et al. 2008; LPWG 2017; Koenen et al. 2020b). The rest of Old World Prosopis (three species) is sister to the Indo-Nepalese genus Indopiptadenia and New World Prosopis has the Namibian – S. African Xerocladia nested within it (Fig. 5). A new generic classification of Prosopis s.l., accounting for this non-monophyly, is presented in this Special Issue by Hughes et al. (2022a).

Desmanthus

The non-monophyly of Desmanthus with the monospecific Hawaiian endemic Kanaloa Lorence & K.R. Wood nested within it (Fig. 5) mirrors earlier phylogenies (Hughes et al. 2003; Luckow et al. 2003, 2005) and is in line with the morphological distinctiveness of Desmanthus balsensis J.L. Contreras from the remaining species of Desmanthus (Contreras Jiménez 1986; Luckow 1993). A new monospecific segregate genus to account for this non-monophyly is proposed in this Special Issue by Hughes et al. (2022b).

Dichrostachys , Gagnebina and Alantsilodendron

Dichrostachys (DC.) Wight & Arn. and Alantsilodendron Villiers are both recovered as non-monophyletic in our sparsely sampled analysis (Fig. 5), raising questions about the monophyly of Gagnebina Neck. ex DC., here represented by just a single species. The Malagasy members of these three genera (all species in our phylogeny, except D. cinerea R. Vig.) cluster together in a clade characterised by very short branches and extensive gene tree conflict (Fig. 5) suggestive of an early burst model of diversification typical of a rapid radiation on Madagascar (Aebli 2015). Previous molecular phylogenetic studies have also found at least some of these genera to be non-monophyletic (Hughes et al. 2003; Luckow et al. 2003, 2005; Aebli 2015) and some species have been transferred between genera based on morphology (Lewis and Guinet 1986). Each of these genera contains several other species from Madagascar not sampled here. While a parsimonious solution could be to merge the three genera into Gagnebina (de Candolle 1825) (a name predating Dichrostachys (Wight and Walker-Arnott 1834) and Alantsilodendron (Villiers 1994)), such a move would result in a highly variable genus, with no consistent morphological character to distinguish it. A forthcoming monograph (Luckow, unpublished data) will resolve the non-monophyly of these genera by transferring two species of Dichrostachys to Alantsilodendron and seven to a new genus (Phillipson et al. 2022). Additional sampling of non-Malagasy species of Dichrostachys would also be important, especially Australian D. spicata, as it has been placed as sister to the combined Dichrostachys / Gagnebina / Alantsilodendron + Calliandropsis nervosa (Britton & Rose) H.M. Hern. & Guinet clade in several studies (Hughes et al. 2003; Luckow et al. 2003, 2005; Aebli 2015). The African species D. dehiscens Balf. f. and D. kirkii Benth. also need to be sampled as they share a dehiscent fruit type with members of the new Madagascan genus.

Stryphnodendron and Pseudopiptadenia

Our analyses support the monophyly of the Stryphnodendron clade sensu Koenen et al. (2020b) comprising the genera Parapiptadenia Brenan, Pityrocarpa (Benth. & Hook.f.) Britton & Rose, Pseudopiptadenia Rauschert and Stryphnodendron Mart. (Fig. 6) and presumably Microlobius C. Presl., which, although not sampled here, has been shown to be nested within or sister to Stryphnodendron (Ribeiro et al. 2018; Simon et al. 2016; see also Lima et al. 2022). Of these genera, only Parapiptadenia is monophyletic in our analyses, although Pityrocarpa is here only represented by a single taxon (Fig. 6). Stryphnodendron is non-monophyletic as S. duckeanum Occhioni does not group with the rest of the genus (Fig. 6), in line with flower, fruit and branching characteristics that suggested transfer of S. duckeanum to another genus (Scalon 2007) and with previous molecular phylogenies showing S. duckeanum separated from the rest of Stryphnodendron (Jobson and Luckow 2007; Simon et al. 2016; Ribeiro et al. 2018; Sauter 2019). Similarly, Pseudopiptadenia is also non-monophyletic with P. schumanniana placed as sister to the single sampled species of Pityrocarpa, rather than forming a clade with Pseudopiptadenia contorta (DC.) G.P. Lewis & M.P. Lima and P. psilostachya (DC.) G.P. Lewis & M.P. Lima (Fig. 6). Several previous molecular phylogenies also found Pseudopiptadenia to be non-monophyletic – however, those studies did not include P. schumanniana and found P. brenanii G.P. Lewis & M.P. Lima (not sampled here) to be the outlier instead (Simon et al. 2016; Ribeiro et al. 2018). The sparsely sampled backbone phylogeny of the Stryphnodendron clade presented here provides the foundations for more densely sampled analyses and re-delimitation of both Stryphnodendron (Lima et al. 2022) and Pseudopiptadenia / Pityrocarpa (Borges et al. 2022) in this Special Issue.

The remaining genera in the Stryphnodendron and Mimosa clades are all monophyletic (Fig. 6), confirming previous phylogenetic studies and taxonomic rearrangements, including segregation of Lachesiodendron P.G. Ribeiro, L.P. Queiroz & Luckow from Piptadenia (Ribeiro et al. 2018), as well as placement of amphi-Atlantic Adenopodia C. Presl as sister to Mimosa and the sister group relationships amongst the main clades of Mimosa (Simon et al. 2011).

Senegalia and allied genera

The striking cytonuclear discordance whereby Senegalia Raf. appears as non-monophyletic in the analyses of nuclear gene sequences, but as monophyletic in the analyses of plastomes, was first revealed by Koenen et al. (2020b), a result confirmed here by sampling more species of Senegalia, plus the closely related Mariosousa, Parasenegalia and Pseudosenegalia (Fig. 7). In the nuclear gene analyses, the two clades of Senegalia plus these three other genera and the incompletely known Albizia leonardii Britton & Rose ex Barneby & J.W. Grimes form a paraphyletic grade with very short and poorly supported or unsupported internal branches (Fig. 7). The complex and intriguing issues these features raise for delimitation of Senegalia are explored by Terra et al. (2022), who conclude that sequencing of more species is required.

Calliandra

Following reduction of Bentham’s (1875) broad trans-continental circumscription of Calliandra Benth. to just the New World species by Barneby (1998), five genera have been segregated to account for the majority of the Old World species. Now just a handful of Old World species remain to be resolved, including the Asian Calliandra sp. nov. (Poilane 9150), that, as expected, does not group together with the New World Calliandra s.s., but is instead sister to the Indian monospecific genus Sanjappa É.R. Souza & M.V. Krishnaraj in the Zapoteca clade (Fig. 7). Bentham (1875) included four Asian species in Calliandra (de Souza et al. 2013), which share the apically dehiscent pods of Calliandra (Fig. 13a–f), but in other respects present anomalies, especially in the configuration of their polyads. The identities of these Asian Calliandra species have long been considered ambiguous (Barneby 1998). Two of these Asian species have been assigned to different genera (C. cynometroides Bedd. to Sanjappa (de Souza et al. 2016) and C. geminata (Wight & Arn.) Benth. to Thailentadopsis Kosterm. (Lewis and Schrire 2003)), while the generic placement of the remaining species, C. umbrosa (Wall.) Benth., remains unknown. The fourth species, C. griffithii Baker ex Benth., is now considered a subspecies of C. umbrosa (Paul 1979). Calliandra umbrosa has never been included in a molecular phylogenetic analysis (de Souza et al. 2013, 2016) and, unfortunately, sequencing of C. umbrosa was unsuccessful in this study. However, polyad, leaf, corolla and pod morphology, plus the presence of facultatively spinescent stipules, distinguish C. umbrosa from other genera, suggesting that it should potentially be assigned to a new genus (de Souza et al. 2013, 2016). Until DNA sequences of C. umbrosa can be obtained to ascertain its relationship to Calliandra sp. nov., this residual non-monophyly of the genus Calliandra cannot be resolved.

Pithecellobium and allies

While the Pithecellobium alliance is the only one of the informal alliances of Barneby and Grimes (1996) whose monophyly has withstood the test of phylogenomic analysis (Koenen et al. 2020b), other than Pithecellobium Mart. itself, our sparsely sampled phylogeny of this clade suggests that the monophyly of the four other genera placed in the Pithecellobium clade (Painteria Britton & Rose, Havardia Small, Ebenopsis Britton & Rose and Sphinga Barneby & J.W. Grimes) is doubtful and needs to be further tested with more complete taxon sampling (Fig. 7). Even with our limited taxon sampling, Painteria and Havardia are clearly non-monophyletic (Fig. 7), raising significant doubts about the taxonomic status of Ebenopsis and Sphinga, which are both represented by only one species in our trees. Painteria is especially poorly distinguished from Havardia; Sphinga was originally described in Havardia and previous studies (Nielsen 1981; Polhill 1994) placed all four genera in a more broadly defined Havardia (Brown 2008). Such a solution might, therefore, seem sensible, but together they form a paraphyletic grade in our phylogenies (Fig. 7), suggesting that unless all four genera were to be sunk back into Pithecellobium (from which they were segregated (Barneby and Grimes 1996)), these four genera require at least three names, as they are divided over three (poorly-supported) lineages: one comprising Spinga acatlensis (Benth.) Barneby & J.W. Grimes and Havardia campylacantha (L. Rico & M. Sousa) Barneby & J.W. Grimes, one Painteria leptophylla (DC.) Britton & Rose, Pa. elachistophylla (A. Gray ex S. Watson) Britton & Rose and Ebenopsis confinis (Standl.) Britton & Rose and one H. pallens (Benth.) Britton & Rose, which is the type species of Havardia and sister to Pithecellobium. Clearly, taxon sampling in our phylogeny is too limited to draw firm taxonomic conclusions. A new phylogeny of the Pithecellobium clade, presented here in this Special Issue, is used as the basis for erecting two new genera to account for these generic non-monophyly issues (Tamayo-Cen et al. 2022). This new phylogeny, based on a small set of DNA sequence loci, but with denser taxon sampling than that encompassed here, is not fully congruent with the phylogenomic backbone presented in Fig. 7.

The Archidendron clade

The genera and lineages of the large Archidendron clade comprising Acacia Mill., Archidendron F. Muell. and six smaller genera (Fig. 8; Koenen et al. 2020b), together make up over one third of all mimosoid species and are restricted to Australasia. Relationships across the backbone of this clade are complex and generally poorly resolved with very short branches and high levels of gene tree conflict and lack of phylogenetic signal across a significant fraction of genes (Fig. 8), such that the topologies across different analytical approaches can differ. This suggests that some nodes across this backbone should better be viewed as putative polytomies. Three genera in this clade, Wallaceodendron Koord., Pararchidendron I.C. Nielsen and Paraserianthes I.C. Nielsen, are monospecific. Falcataria (I.C. Nielsen) Barneby & J.W. Grimes comprises three species but is represented by only one taxon in our phylogeny, so no conclusion can, therefore, be made about its monophyly, although our results support the segregation of this genus from Paraserianthes (Barneby and Grimes 1996; Brown et al. 2011). Three of the four remaining genera are monophyletic: Acacia, Archidendron and Serianthes Benth. (confirming the results of Demeulenaere et al. (2022) in this Special Issue). However, the monophyly of Archidendron remains doubtful as it is supported by few gene trees and opposed by many (Fig. 8) and the genus is not monophyletic in the plastid tree (Suppl. material 3). This is very much in line with previous findings of a non-monophyletic Archidendron (Brown et al. 2008, 2011; Iganci et al. 2016; LPWG 2017). The likely non-monophyly of Archidendron is explored in more detail in this Special Issue by Brown et al. (2022). It is notable that the two well-supported Archidendron subclades found here are replicated by Brown et al. (2022), where their morphological and geographical identities are discussed in detail. Finally, the non-monophyly of Archidendropsis I.C. Nielsen, documented and addressed in this Special Issue by Brown et al. (2022), is confirmed by the much larger phylogenomic dataset analysed here (Fig. 8).

Our results weakly support Paraserianthes lophantha as sister to Acacia (Fig. 8), in line with earlier findings (Brown et al. 2008, 2011; Koenen et al. 2020b) and shared morphological similarities including hard seeds that are stimulated to germinate by fire (Brown et al. 2011), minute anthers and numerous stamens (Barneby and Grimes 1996). As P. lophantha contains two geographically disjunct subspecies, P. lophantha subsp. montana (Jungh.) I.C. Nielsen in Indonesia and P. lophantha subsp. lophantha (the subspecies sequenced here) in southern Australia (Brown et al. 2011), sequencing the missing subspecies would be worthwhile to check that the two cluster together as sister to Acacia. However, it is important to note that this relationship is sensitive to the type of dataset and phylogenetic method: the ASTRAL trees (Fig. 8) recover P. lophantha as the sister of Acacia, whereas the nuclear RAxML phylogenies (Ringelberg et al. 2022) find a sister relationship between Acacia and Archidendron plus Archidendropsis xanthoxylon (C.T. White & W.D. Francis) I.C. Nielsen, the PhyloBayes gene jack-knifing phylogeny (Ringelberg et al. 2022) resolves the whole Archidendron clade as one large polytomy lacking a clear sister lineage to Acacia and the plastid tree (Suppl. material 3) recovers Archidendropsis xanthoxylon as sole sister of Acacia. Furthermore, P. lophantha and several species of Archidendron are also identified as species often changing positions across trees by RogueNarok (Aberer et al. 2013). The high levels of intergenic conflict, very short branches, extremely low bootstrap support values especially in the nucleotide RAxML phylogenies, lack of concordance and signal amongst the gene trees and failure to reject a polytomy by ASTRAL (Fig. 8), all suggest that the backbone of the Archidendron clade should perhaps best be viewed as one large polytomy, as depicted in the PhyloBayes consensus tree (Ringelberg et al. 2022). However, the number (eight in the PhyloBayes phylogeny) and precise identity of lineages arising from this tangle remain unclear and relationships amongst the genera of this clade remain highly uncertain pending additional taxon sampling and detailed investigation of the causes of gene tree conflict and possible evidence for introgression.

Albizia

At the start of this study, the genus Albizia was dubbed the last pantropical so-called ‘dustbin’ genus pending resolution (Koenen et al. 2020b). Here, we show that Albizia s.l. is rampantly non-monophyletic, most notably because the bulk of the Old and New World species are placed in separate clades (Figs 9 and 10). This Old World – New World split is remedied in this Special Issue by Aviles et al. (2022) who resurrect the genus Pseudalbizzia Britton & Rose for the majority of the New World species placed in Barneby’s Albizia section Arthrosamanea, with Albizia s.s. now restricted to just the Old World species, which still includes ca. 90 spp. (Koenen et al., unpubl. data). Furthermore, the disparate placements of several other species of Albizia across the phylogeny, viz: Albizia carbonaria Britton (Fig. 8), the long-neglected African Albizia species previously often placed in Cathormion or Samanea (Benth.) Merr. (Figs 9 and 11) and Albizia leonardii (Fig. 7), are all accounted for with new generic placements and nomenclatural combinations (Koenen 2022b; Soares et al. 2022), one synonymisation (Terra et al. 2022) and a new segregate genus (Koenen 2022a), all of them being published in this Special Issue.

Abarema , Hydrochorea and Balizia

The recent re-circumscription of Abarema Pittier to include just two species and transfer of the remaining species to the re-instated Punjuba Britton & Rose and Jupunba Britton & Rose (Guerra et al. 2016, 2019; Iganci et al. 2016; Soares et al. 2021), is broadly supported here (Figs 9 and 11), except for the anomalous placement of Jupunba macradenia (Pittier) M.V.B. Soares, M.P. Morim & Iganci which is sister to the Hydrochorea + Balizia clade (Fig. 9). This placement is unexpected and somewhat suspect considering J. macradenia is firmly placed in Jupunba in Soares et al. (2021). As found by Iganci et al. (2016), Koenen et al. (2020b) and Soares et al. (2021), Balizia is non-monophyletic with the genus Hydrochorea plus two African species of Albizia nested within it (Fig. 9). Hydrochorea is re-circumscribed to accommodate all these elements by Soares et al. (2022) in this Special Issue.

Leucochloron

Koenen et al. (2020b) showed that Leucochloron is polyphyletic and that result is confirmed here, split between the Albizia and Inga clades (Figs 10 and 11). A new segregate genus to account for this non-monophyly is proposed in this Special Issue by de Souza et al. (2022b).

Zygia , Macrosamanea and Inga

Alongside Archidendron, the large Neotropical, mainly rainforest genus Zygia remains one of the least well-documented genera of mimosoids, with many species known from incomplete material (Barneby and Grimes 1997). Previous work by Ferm et al. (2019) showed that, while the bulk of genus Zygia is monophyletic, a handful of outlier species have affinities to other genera: Zygia ocumarensis (Pittier) Barneby & J.W. Grimes is sister to Macrosamanea Britton & Rose ex Britton & Killip, Marmaroxylon magdalenae Killip ex. L. Rico (treated as a synonym of Z. ocumarensis by Barneby and Grimes (1997)) is nested in Jupunba and Z. inundata and Z. sabatieri are together sister to Inga. With the exception of M. magdalenae, which is not included in this study, these placements are confirmed here with phylogenomic data (Figs 11 and 12) and reflect the morphological distinctiveness of these species from the rest of the genus (Barneby and Grimes 1997; Ferm et al. 2019) which prompted placements in their own separate monospecific sections of Zygia (Barneby and Grimes 1997). New nomenclatural combinations to deal with these outlier Zygia species are still pending. We suggest that Zygia ocumarensis should best be transferred to Macrosamanea, as it shares bipinnate leaves with multiple pairs of pinnae and an absence of cauli-/ramiflory (which is almost universal in Zygia) with several species of Macrosamanea (Barneby and Grimes 1996; Ferm et al. 2019). The identity of Marmaroxylon magdalenae needs to be re-evaluated, but the evidence of Ferm et al. (2019), who sampled the type material, suggests it should be transferred to Jupunba. The generic placements of Z. inundata and Z. sabatieri are more contentious. Arguments can be made to transfer Z. inundata to Inga (Ferm et al. 2019): it was originally described in Inga and it shares once-pinnate leaves and absence of cauli-/ramiflory with Inga (Barneby and Grimes 1997; Ferm et al. 2019). However, Z. inundata was placed as the sole sister of Inga in the plastid tree (Suppl. material 3) and by Ferm et al. (2019), whereas the nuclear gene data suggest that Z. inundata is sister to Z. sabatieri and together these two species form the sister clade of Inga (Fig. 12). Zygia sabatieri has bipinnate leaves and both Z. sabatieri and Z. inundata have dehiscent pods, characteristics that distinguish these species from Inga with its uniformly once-pinnate leaves and indehiscent pods. In order to maintain a morphologically coherent and homogeneous Inga with respect to these diagnostic characters, segregating Z. inundata and Z. sabatieri as a new genus would appear to be advantageous. Ingopsis Barneby & J.W. Grimes and Pseudocojoba Barneby & J.W. Grimes, the names for the monospecific sections containing Z. inundata and Z. sabatieri, respectively (Barneby and Grimes 1997), are two available names, of which Ingopsis would be preferable given the morphological and phylogenetic proximity of this clade to Inga and the lack of a close relationship to Cojoba Britton & Rose. However, since these sectional names have no priority at generic rank (Turland et al. 2018), alternatively, a new name could equally be proposed. Finally, while Zygia s.s. was reasonably well sampled by Ferm et al. (2019) and also in the current study (Fig. 12), alongside further herbarium taxonomic work and field studies to clarify species, denser phylogenetic taxon sampling is desirable, in particular to include Z. eperuetorum (Sandwith) Barneby & J.W. Grimes. This species is known only from the Essequibo Valley in Guyana, was placed in its own section by Barneby and Grimes (1997), has an unusual combination of morphological characters not found elsewhere in Zygia and the fruit remains unknown. Zygia eperuetorum may well, therefore, represent an additional separate lineage that could potentially merit recognition as a distinct genus.