Phylogenetic analyses of DNA sequence data sampling all species of Leucochloron alongside representatives of genera of the Inga and Albizia clades of the larger ingoid clade of mimosoid legumes (sensu Koenen et al. 2020) confirm the non-monophyly of the genus Leucochloron. We show that Leucochloron bolivianum is placed in the Albizia clade, while the remaining four species of Leucochloron are placed in the Inga clade, in line with previous results. To rectify this non-monophyly, L. bolivianum is segregated as the new genus, Boliviadendron, with a single species, Boliviadendron bolivianum, narrowly endemic to the interior Andean valleys of Bolivia. We illustrate this new segregate genus, present a map of its distribution and discuss the striking lack of morphological distinctions between Boliviadendron and Leucochloron, as well as the phylogenetic and morphological affinities of Boliviadendron to the genera Enterolobium and Albizia.

Fabaceae, generic delimitation, Leucochloron, monophyly, taxonomy

The genus Leucochloron Barneby & J.W. Grimes was established by Barneby and Grimes (1996) as part of their generic system for the mimosoid legume tribe Ingeae in the New World, with four species, all of them native to Brazil: L. incuriale (Vell.) Barneby & J.W. Grimes, L. limae Barneby & J.W. Grimes, L. minarum (Glaz. ex Harms) Barneby & J.W. Grimes and L. foederale (Barneby & J.W. Grimes) Barneby & J.W. Grimes. Barneby and Grimes (1996) based their new genus on the combination of lack of armature, perulate resting buds, homomorphic flowers arranged in globose axillary capitula, broad plano-compressed stiffly papery or weakly coriaceous fruits and shiny discoid exareolate seeds, encircled by a marginal nerve or minute wing.

In May 2002, C.E. Hughes and collaborators collected material of an undescribed mimosoid legume tree from a single locality on the mid-elevation eastern flanks of the Bolivian Andes. A small number of additional collections from nearby localities have been made in subsequent years. This material shares the same combination of characters used by Barneby and Grimes (1996) to delimit Leucochloron and, hence, provided the basis for description of a fifth species, Leucochloron bolivianum C.E. Hughes & Atahuachi (Hughes and Atahuachi (2006, publ. 2007).

This apparently morphologically commodious generic home for Leucochloron bolivianum has been brought into question by phylogenetic evidence from Almeida (2014), LPWG (2017) and more recently generated DNA sequence data for 964 nuclear genes for two species of Leucochloron, L. bolivianum and L. limae, which showed that the genus Leucochloron is non-monophyletic (Koenen et al. 2020). Here, we further test this non-monophyly to ascertain where the remaining three species of Leucochloron are placed phylogenetically in relation to the two separate Leucochloron lineages identified by Koenen et al. (2020), with the aim of providing a solid basis for generic re-delimitation.

Morpho-taxonomic analyses and sample selection for DNA extraction were based on field collections and herbarium specimens from BOLV, FHO, HUEFS, K, LPB, MBM, MEXU, NY, SP and VIC (acronyms follow Thiers 2022, [continuously updated]) collections and previous taxonomic literature (Hughes and Atahuachi 2006, publ. 2007; Almeida 2014).

We sampled 47 accessions (40 of them newly sequenced here and seven using sequences from GenBank) in 19 genera, including all genera of the Albizia and Inga clades sensu Koenen et al. (2020), except Abarema s.s., plus a representative set of other lineages spanning the ingoid clade (see Suppl. material 1: Table S1, for GenBank accession numbers and voucher information). All five recognised species of Leucochloron were sampled, including seven accessions of Leucochloron bolivianum. Acacia Mill. (i.e. Acacia sensu stricto), which is also a member of the wider ingoid clade, was selected as outgroup.

Total genomic DNA was extracted from silica-dried or herbarium leaf tissue using a modified 2 × CTAB protocol (Doyle and Doyle 1987). DNA samples with low amplification were purified using Sepharose CL-6B (Sigma, St. Louis, Missouri, USA), following the manufacturer’s protocol. We sequenced eight loci which have been used in previous phylogenetic studies of generic relationships in the ingoid clade (e.g., Souza et al. 2013): the nuclear ribosomal 5.8S subunit and flanking ITS1 and ITS2 spacers, the external transcribed spacer ETS and the plastid loci rps16, psbA-trnH, rpL32-trnL, trnL intron, trnL-F and trnD-T (see Suppl. material 1: Table S2 for primer sequences and amplification protocols). The ITS amplifications were performed with primers 17SE and 26SE or primers ITS92 and ITS4. The trnD-T region was amplified in two independent reactions combining primers: trnD^guc + trnE^uuc and trnY^gua + trnT^ggu (Shaw et al. 2007).

Polymerase Chain Reactions (PCR) were performed using TopTaq Master Mix Kit (QIAGEN GmbH, Hilden, Germany) according to the manufacturer’s protocol, in 10–15 μl of final reaction volume. For the nrITS and ETS amplifications, 1.0 M betaine and 2% DMSO (dimethylsulphoxide; 2% of the preparation volume) were added. PCR products were purified by enzymatic treatments with Exonuclease I and Shrimp alkaline phosphatase (kit ExoSapIT, GE Healthcare Buckinghamshire, U.K.). Sequencing reactions were performed using the same primers used for PCR and Big Dye Terminator kit v.3.1 (Applied Biosystems, Foster City, California, U.S.A.), on the ABI3130XL Analyzer (Applied BioSystems), at the Laboratório de Sistemática Molecular de Plantas (LAMOL), of the Universidade Estadual de Feira de Santana (UEFS), Bahia, Brazil.

Electropherograms were assembled in Geneious 5.3.6 (Drummond et al. 2010) using default assembly settings. Consensus sequences were aligned in MUSCLE (Edgar 2004) using Geneious, with manual adjustments. We analysed three datasets: 1) the combined plastid loci, 2) combined nuclear loci and 3) combined plastid + nuclear loci. Incongruence between datasets was assessed via 1,000 replicates of the Homogeneity of Participation Test (Felsenstein 1985) using PAUP v. 4.0b10 for Windows (Swofford 2002).

Maximum Parsimony (MP) analyses were performed using PAUP v.4.0 (Swofford 2002), using Fitch parsimony (all characters unordered and equally weighted; Fitch 1971), 1,000 replications of taxon addition and resampling by tree bisection-reconnection (TBR) branch swapping, saving up to 15 trees per replication to prevent extensive searches in sub-optimal islands. Trees from the first search were used as a starting point for a second search using the same parameters retaining a maximum of 10,000 trees. Non-parametric bootstrapping (BS; Felsenstein 1985) was used to estimate clade support, assessed through 1,000 replications, simple taxon-addition and the TBR algorithm, saving up to 15 trees per replicate.

Maximum Likelihood (ML) analyses were performed using RAxML v.8 (Stamatakis 2014), on the CIPRES Science Gateway v.3.3 (Miller et al. 2010), using the GTR + CAT evolutionary model and estimating GAMMA distribution during the run. Branch support was evaluated using 1,000 rapid bootstrap replicates.

Best-fitting substitution models for each data partition were selected using the Akaike Information Criterion (AIC) using MrModeltest v.2.3 (Nylander 2004) (Suppl. material 1: Table S2). Bayesian Inference (BI) analysis was performed on MrBayes v.3.2.6 (Ronquist and Huelsenbeck 2003; Ronquist et al. 2012), on the CIPRES Science Gateway v.3.3 (Miller et al. 2010). Two separate runs of a Metropolis-coupled Markov Chain Monte Carlo (MCMC) were initiated with random starting trees and four simultaneous chains set at default temperatures (Ronquist and Huelsenbeck 2003) for 30 million generations, sampling once every 10³ generations. Convergence was assessed using Tracer v.1.6 (Rambaut et al. 2014) to ensure effective sample sizes (ESS) ≥ 200. MrBayes was also used to summarise 75% post burn-in trees sampled from a 50% majority rule consensus tree that included posterior probabilities (PP) as branch support estimates. All trees from MP, BI and ML were visualised and edited in FigTree 1.4.4 (Rambaut 2018).

While separate analyses of the nuclear ITS + ETS and plastid loci yielded gene trees that were, in many cases, poorly resolved (Suppl. material 1: Table S3, Suppl. material 2: Figs S1–S4), the combined nuclear + plastid analysis (Fig. 1) is sufficiently resolved to demonstrate that the genus Leucochloron is not monophyletic, with all seven accessions of L. bolivianum forming a robustly supported clade placed in a clade with Albizia Durazz., Enterolobium Mart. and Samanea (Benth.) Merr. and the other four species of Leucochloron forming a robustly supported clade that is placed in a clade with Inga Mill., Zygia P. Browne and Macrosamanea Britton & Rose ex Britton & Killip. (Fig. 1).

Figure 1. 

Majority-rule (50%) consensus tree from Bayesian analysis of the combined nuclear (ITS, ETS) and plastid (psbA-trnH, rpL32, rps16, trnD-T, trnL-F) data for Boliviadendron and related genera. Values above branches are Bayesian Posterior Probabilities (PP), below are Bootstrap Support (BS) percentages from the maximum parsimony (left) and Maximum Likelihood analyses (right). Branches supported by PP ≥ 95% are in bold. Only BS values ≥ 50% are shown; - indicates BS ≤ 50%; * indicates PP = 1.0 or BS = 100%.

The non-monophyly of the genus Leucochloron is supported by two independent phylogenetic studies, one presented here sampling all five species currently assigned to the genus, based on a small set of widely-used nrDNA and plastid DNA sequence loci (Fig. 1) and the other sampling just two species, but based on phylogenomic analyses of 964 nuclear gene sequences and a broad sample of genera across the whole mimosoid clade (Koenen et al. 2020). This non-monophyly is further confirmed in a more recent phylogeny constructed using a slightly expanded set of 997 nuclear genes, derived from a modified version of the Mimobaits gene set of Koenen et al. (2020), which sampled all but five of the 151 genera of subfamily Caesalpinioideae (Ringelberg et al. 2022). This thorough sampling of both taxa and genes demonstrates beyond doubt that Leucochloron is indeed non-monophyletic, with L. bolivianum placed in the Albizia clade (sensu Koenen et al. 2020) and the remaining four species of Leucochloron forming a robustly-supported clade in the Inga clade (sensu Koenen et al. 2020; Fig. 1). These results provide strong evidence supporting segregation of L. bolivianum as a new genus, here named Boliviadendron.

This robustly-supported non-monophyly is unexpected, given the close morphological similarities between L. bolivianum and the other four species of Leucochloron. Indeed, there appear to be very few qualitative morphological differences separating Boliviadendron (i.e., L. bolivianum) from Leucochloron, suggesting that the original combination of character states used by Barneby and Grimes (1996) to delimit Leucochloron has essentially been recapitulated in Boliviadendron. These striking morphological similarities between Leucochloron and Boliviadendron must, thus, be viewed as further evidence of the extensive morphological homoplasy evident across mimosoids which has frequently misled generic delimitation (Ringelberg et al. 2022). For example, the presence of minute wing-like rims of seeds, which was one of the characters supporting placement of L. bolivianum in the genus Leucochloron (Hughes and Atahuachi 2006, publ. 2007), have been documented across phylogenetically scattered mimosoid legume lineages, including the genera Mimozyganthus Burkart (Luckow et al. 2005), Archidendropsis I.C. Nielsen and Leucochloron and now here in the phylogenetically distinct lineage Boliviadendron. This suggests that this character, just like seeds with true wings which also occur across a disparate set of unrelated mimosoid genera, is homoplasious and evolutionarily labile.

Morphological characters which weakly separate Boliviadendron from Leucochloron s.s. are: (i) the leaflet base is more evidently asymmetrical in Boliviadendron than in the species of Leucochloron and the under-surface of Boliviadendron leaflets has 1–2 (–3) prominent primary veins, but otherwise the venation is not evident vs. the evident reticulate secondary and tertiary venation on the lower leaflet surface in Leucochloron species; (ii) in Boliviadendron, the upper leaflet surfaces are consistently blotched purple-black, while leaflets of Leucochloron s.s. species are strongly discolorous, with the upper surface drying dark brown and often glossy; (iii) the indumentum in Boliviadendron tends to be shorter and white, especially on the corollas and calyces which have fine white, appressed, silky trichomes vs. the generally more ferruginous and longer indumentum, occasionally with golden and/or white hairs intermixed, of Leucochloron species; (iv) the number of pollen grains per polyad is constant in Boliviadendron at 16 (counted on the isotype Hughes 2423 and Wood 21618 at K), while in Leucochloron s.s., it is variable (even within species) with 16, 18, 24 or 32 grains.

In the Koenen et al. (2020) phylogeny, Boliviadendron appeared as sister to a clade comprising Enterolobium and Albizia s.s. (i.e., Old World Albizia). In the phylogeny presented here, Boliviadendron is also placed in a clade with Enterolobium and Albizia s.s. (Fig. 1) in line with the findings of Koenen et al. (2020). This close relationship of Boliviadendron to Enterolobium is reflected in the closely similar leaf and flower morphologies of these two genera, including the asymmetrically displaced primary vein of the leaflets (which also characterises many species of Albizia s.s.), the blotched purple-black upper leaflet surfaces and superficially very similar globose axillary capitula of homomorphic flowers immersed amongst coevally developing leaves. However, in fruit, these two genera are immediately distinguishable by the characteristically thickened indehiscent curved ‘ear-pod’ fruits of Enterolobium vs. the plano-compressed, inertly dehiscent fruits with chartaceous valves of Boliviadendron. This means that material of Boliviadendron lacking fruits can be difficult to distinguish from Enterolobium. Indeed, one of the collections, cited as L. bolivianum by Hughes and Atahuachi (2006, publ. 2007) (J.R.I. Wood 22134 at K, which has leaves and flowers only), has since been correctly identified as Enterolobium contortisiliquum (Vell.) Morong. Closer examination of the inflorescences of these two taxa, which can occur in close geographical proximity in Bolivia, shows that the flowers of Boliviadendron are essentially sessile, while those of Enterolobium contortisiliquum are pedicellate, on more robust inflorescences. The plano-compressed fruits with stiffly papery chartaceous or weakly coriaceous valves dehiscing inertly along both sutures in Boliviadendron are strongly reminiscent of the fruits of the other closely related genus, Albizia s.s., but the marked lack of a pleurogram on the seeds in Boliviadendron clearly distinguishes the genus from Albizia. It is notable that polyads, like those of Boliviadendron, are also 16-celled in Albizia s.s. (although only few species of the genus have been studied for their pollen characteristics), but can be 16 or 32-celled in Enterolobium.

We thank the Laboratório de Sistemática Molecular de Plantas (LAMOL-UEFS) and the Jodrell Laboratory of the Royal Botanic Gardens, Kew staff for support with laboratory procedures and all the colleagues who kindly provided samples for DNA extraction. This work was supported by the FAPESB (grants: PTX0004/2016 and APP0096/2016), SISBIOTA (563084/2010-3) and the Swiss National Science Foundation through grants 310003A_156140 and 31003A_182453/1 to Colin Hughes and an Early.Postdoc.Mobility fellowship P2ZHP3_199693 to Erik Koenen. We are grateful to curators of all herbaria cited for providing access to their collections and to João Iganci and Luciano de Queiroz for comments and careful revision of the text.